GUIA DE ESTUDIO: Gestión Comparada Del Desarrollo Sustentable

Universidad Autónoma de Nuevo León Facultad de Ciencias Políticas y Relaciones Internacionales

GUIA DE ESTUDIO: Gestión Comparada del Desarrollo Sustentable

Docente Responsable MSc. Paulina Jiménez Quintana

Academia: Desarrollo Sustentable Coordinador: Dra. Ana María Romo

06/02/2020

TEMAS

1. Panorama general de los servicios ambientales en las decisiones de los líderes 1.1. De la teoría a la implementación 1.2. Teoría del cambio

2. Interpretación y estmación del valor de los servicios ambientales 2.1 La importancia de la valuación de la naturaleza. 2.2 Cuestiones filosóficas: valores, derechos y toma de decisiones. 2.3 Tipos de valores ambientales.

3. Valuación de múltiples servicios ambientales: una herramienta integrativa para la vida real 3.1 El problema con las hojas de balance incompletas actuales. 3.2 La revolución de la toma de decisiones. 3.3 El enfoque ecológico de función-productividad y la herramienta investigación.

4. Estudios de casos de modelos multi-nivel para la valuación de los servicios ambientales. 4.1 Servicios para el suministro de agua del poder hídrico y la irrigación. 4.2 Servicios para el secuestro y almacenamiento de carbono. 4.3 Servicios para el abastecimiento y regulación en la agricultura. 4.4 Servicios para la polinización de cultivos. 4.5 Servicios para el ecoturismo y la recreación. 4.6 Servicios culturales y valores non-use 4.7 Servicios de biodiversidad terrestre. 4.8 Servicios para la conservación marina

5. Extensión, aplicación y la siguiente generación de valuación de servicios ambientales. 5.1. La información que necesitan los managers 5.2. Pobreza y la distribución del capital natural. 5.3. Impactos del cambio climático en la valuación de los servicios ambientales.

6. Incorporando el valor de los servicios ambientales en las decisiones de los líderes.

Bibliografía: Kareiva, P., Tallis, H., Ricketts, T., Daily, G., Polasky, S. (2011). Natural Capital: Theory and Practice of Mapping Ecosystem Services (Chapters 1 and 2). Oxford University Press Inc. New York. U.S.A.

Natural Capital This page intentionally left blank Natural Capital Theory & Practice of Mapping Ecosystem Services

EDITED BY

Peter Kareiva The Nature Conservancy and Santa Clara University, USA Heather Tallis Natural Capital Project, Stanford University, USA Taylor H. Ricketts World Wildlife Fund, USA Gretchen C. Daily Stanford University, USA Stephen Polasky University of Minnesota, USA

1 1 Great Clarendon Street, Oxford ox2 6dp Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide in Oxford New York Auckland Cape Town Dar es Salaam Hong Kong Karachi Kuala Lumpur Madrid Melbourne Mexico City Nairobi New Delhi Shanghai Taipei Toronto With off ces in Argentina Austria Brazil Chile Czech Republic France Greece Guatemala Hungary Italy Japan Poland Portugal Singapore South Korea Switzerland Thailand Turkey Ukraine Vietnam Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries Published in the United States by Oxford University Press Inc., New York © Oxford University Press 2011 The moral rights of the authors have been asserted Database right Oxford University Press (maker) First published 2011 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this book in any other binding or cover and you must impose the same condition on any acquirer British Library Cataloguing in Publication Data Data available Library of Congress Cataloging in Publication Data Library of Congress Control Number: 2010942945 Typeset by SPI Publisher Services, Pondicherry, India Printed in Great Britain on acid-free paper by CPI Antony Rowe, Chippenham, Wiltshire

ISBN 978-0-19-958899-2 (Hbk.) 978-0-19-958900-5 (Pbk.)

1 3 5 7 9 10 8 6 4 2 Contents

List of contributors xi Foreword (Hal Mooney) xv How to read this book xvii Acknowledgments xviii

Section I: A vision for ecosystem services in decisions 1: Mainstreaming natural capital into decisions 3 Gretchen C. Daily, Peter M. Kareiva, Stephen Polasky, Taylor H. Ricketts, and Heather Tallis 1.1 Mainstreaming ecosystem services into decisions 3 1.2 What is new today that makes us think we can succeed? 4 1.3 Moving from theory to implementation 5 1.4 Using ecosystem production functions to map and assess natural capital 6 1.5 Roadmap to the book 6 Box 1.1: The everyday meaning of natural capital to the world’s rural poor (M. Sanjayan ) 7 1.6 Open questions and future directions 9 Box 1.2: Sorting among options for a more sustainable world ( Stephen R. Carpenter ) 10 1.7 A general theory of change 12 References 12 2: Interpreting and estimating the value of ecosystem services 15 Lawrence H. Goulder and Donald Kennedy 2.1 Introduction: why is valuing nature important? 15 2.2 Philosophical issues: values, rights, and decision-making 16 2.3 Measuring ecosystem values 20 2.4 Some case studies 27 Box 2.1: Designing coastal protection based on the valuation of natural coastal ecosystems (R. K. Turner ) 29 2.5 Conclusions 31 References 33 3: Assessing multiple ecosystem services: an integrated tool for the real world 34 Heather Tallis and Stephen Polasky 3.1 Today’s decision-making: the problem with incomplete balance sheets 34 3.2 The decision-making revolution 34 3.3 The ecological production function approach 35 vi C O N T E N T S

3.4 InVEST: mapping and valuing ecosystem services with ecological production functions and economic valuation 37 Box 3.1: Unsung ecosystem service heroes: seed dispersal and pest control ( Liba Pejchar ) 39 3.5 Future directions and open questions 47 References 48

Section II: Multi-tiered models for ecosystem services 4: Water supply as an ecosystem service for hydropower and irrigation 53 Guillermo Mendoza, Driss Ennaanay, Marc Conte, Michael Todd Walter, David Freyberg, Stacie Wolny, Lauren Hay, Sue White, Erik Nelson, and Luis Solorzano 4.1 Introduction 53 4.2 Tier 1 water supply model 54 Box 4.1 Can we apply our simple model where groundwater really matters? (Heather Tallis, Yukuan Wang, and Driss Ennaanay ) 54 4.3 Tier 1 valuation 59 4.4 Limitations of the tier 1 water yield models 62 4.5 Tier 2 water supply model 62 4.6 Tier 2 valuation model 65 4.7 Sensitivity analyses and testing of tier 1 water supply models 65 4.8 Next steps 70 References 71 5: Valuing land cover impact on storm peak mitigation 73 Driss Ennaanay, Marc Conte, Kenneth Brooks, John Nieber, Manu Sharma, Stacie Wolny, and Guillermo Mendoza 5.1 Introduction 73 5.2 Tier 1 biophysical model 74 5.3 Tier 1 valuation 78 Box 5.1: Integrated f ood risk management: gaining ecosystem services and increasing revenue (David Harrison ) 80 5.4 Tier 2 supply and use model 84 5.5 Tier 2 valuation 85 5.6 Limitations and next steps 86 References 87 6: Retention of nutrients and sediment by vegetation 89 Marc Conte, Driss Ennaanay, Guillermo Mendoza, Michael Todd Walter, Stacie Wolny, David Freyberg, Erik Nelson, and Luis Solorzano 6.1 Introduction 89 6.2 Tier 1 biophysical models 90 6.3 Tier 1 economic valuation 96 6.4 Tier 2 biophysical models 99 6.5 Tier 2 economic valuation models 102 6.6 Constraints and limitations 104 6.7 Testing tier 1 models 105 Box 6.1: China forestry programs take aim at more than f oods (Christine Tam ) 107 6.8 Next steps 108 References 109 CONTENTS vii

7: Terrestrial carbon sequestration and storage 111 Marc Conte, Erik Nelson, Karen Carney, Cinzia Fissore, Nasser Olwero, Andrew J. Plantinga, Bill Stanley, and Taylor Ricketts 7.1 Introduction 111 7.2 Tier 1 supply model 112 Box 7.1: Noel Kempff case study: capturing carbon f nance ( Bill Stanley and Nicole Virgilio ) 115 7.3 Tier 1 valuation model: an avoided economic damage approach 118 Box 7.2: Valuing the Arc: measuring and monitoring forest carbon for offsetting (Andrew R. Marshall and P. K. T. Munishi ) 119 7.4 Tier 2 supply model 121 7.5 Tier 2 valuation: an application of the avoided economic damage approach 122 7.6 Limitations and next steps 124 References 126 8: The provisioning value of timber and non-timber forest products 129 Erik Nelson, Claire Montgomery, Marc Conte, and Stephen Polasky 8.1 Introduction 129 Box 8.1: Wildlife conservation, corridor restoration, and community incentives: a paradigm from the Terai Arc landscape ( Eric Wikramanayake, Rajendra Gurung, and Eric Dinerstein ) 130 8.2 The supply, use, and value of forests’ provisioning service in tier 1 132 8.3 The supply, use, and value of forests’ provisioning service in tier 2 141 8.4 Limitations and next steps 146 References 147 9: Provisioning and regulatory ecosystem service values in agriculture 150 Erik Nelson, Stanley Wood, Jawoo Koo , and Stephen Polasky 9.1 Introduction 150 9.2 Def ning agricultural scenarios 151 9.3 Tier 1 152 9.4 Tier 2 158 9.5 Mapping the impacts of agriculture on important ecological processes 161 9.6 Uncertainty 162 9.7 Limitations and next steps 163 References 164 10: Crop pollination services 168 Eric Lonsdorf, Taylor Ricketts, Claire Kremen, Rachel Winfree, Sarah Greenleaf, and Neal Williams 10.1 Introduction 168 Box 10.1: Assessing the monetary value of global crop pollination services (Nicola Gallai, Bernard E. Vaissière, Simon G. Potts, and Jean-Michel Salles ) 169 10.2 Tier 1 supply model 173 10.3 Tier 1 farm abundance map 175 10.4 Tier 1 valuation model 175 10.5 Tier 2 supply model 176 10.6 Tier 2 farm abundance map 177 10.7 Tier 2 valuation model 177 10.8 Sensitivity analysis and model validation 178 viii C O N T E N T S

10.9 Limitations and next steps 182 Box 10.2: Pollination services: beyond agriculture ( Berry Brosi ) 183 References 185 11: Nature-based tourism and recreation 188 W. L. (Vic) Adamowicz, Robin Naidoo, Erik Nelson, Stephen Polasky, and Jing Zhang 11.1 Nature-based tourism and recreation values in context 188 11.2 Tier 1 tourism supply and use model 190 11.3 Tier 2 tourism supply and use model 193 11.4 Tier 1 and 2 use value 197 Box 11.1: How the economics of tourism justif es forest protection in Amazonian Peru (Christopher Kirkby, Renzo Giudice, Brett Day, Kerry Turner, Bridaldo Silveira Soares-Filho, Hermann Oliveira-Rodrigues, and Douglas W. Yu ) 198 11.5 State-of-the-art tourism value 201 11.6 Limitations and next steps 202 References 204 12: Cultural services and non-use values 206 Kai M. A. Chan, Joshua Goldstein, Terre Satterf eld, Neil Hannahs, Kekuewa Kikiloi, Robin Naidoo, Nathan Vadeboncoeur, and Ulalia Woodside 12.1 Introduction 206 Box 12.1: The sacred geography of Kawagebo (Jianzhong Ma and Christine Tam ) 211 12.2 Methods: integrating cultural services and non-use values into decisions 213 Box 12.2: People of color and love of nature ( Hazel Wong ) 221 12.3 Limitations and next steps 225 References 226 13: Terrestrial biodiversity 229 Erik Nelson, D. Richard Cameron, James Regetz, Stephen Polasky, and Gretchen C. Daily 13.1 Introduction 229 13.2 Tier 1: habitat-quality and rarity model 229 Box 13.1: Integrating biodiversity and agriculture: a success story in South Asia (Jai Ranganathan and Gretchen C. Daily ) 230 13.3 Tier 2 models of terrestrial biodiversity 233 13.4 Tier 1 and 2 examples with sensitivity analysis 236 13.5 Limitations and next steps 242 References 244

Section III: Extensions, applications, and the next generation of ecosystem service assessments 14: Putting ecosystem service models to work: conservation, management, and trade-offs 249 Stephen Polasky, Giorgio Caldarone, T. Ka’eo Duarte, Joshua Goldstein, Neil Hannahs, Taylor Ricketts, and Heather Tallis 14.1 Introduction 249 14.2 Applying ecosystem service and biodiversity models in management and conservation contexts 250 Box 14.1: Plight of a people ( Neil Hannahs ) 256 14.3 Extending the frontier: challenges facing ecosystem management 260 References 262 CONTENTS ix

15: How much information do managers need? The sensitivity of ecosystem service decisions to model complexity 264 Heather Tallis and Stephen Polasky 15.1 Introduction 264 Box 15.1: How much data do we need to support our models: a case study using biodiversity mapping and conservation planning (Craig Groves and Edward Game ) 265 15.2 Testing agreement between simple and complex ecosystem service models 268 15.3 Future directions and open questions 276 References 277 16: Poverty and the distribution of ecosystem services 278 Heather Tallis, Stefano Pagiola, Wei Zhang, Sabina Shaikh, Erik Nelson, Charlotte Stanton, and Priya Shyamsundar 16.1 Introduction 278 16.2 Ecosystem services and the poor 279 Box 16.1: Can the natural capital of agroecosystems provide a pathway out of poverty? (C. Peter Timmer ) 280 Box 16.2: Poverty and ecosystem service mapping at work in Kenya ( Norbert Henninger and Florence Landsberg ) 282 16.3 Mapping poverty and ecosystem services 283 16.4 Case studies 285 16.5 Including institutions: the way forward 293 References 293 17: Ecosystem service assessments for marine conservation 296 Anne D. Guerry, Mark L. Plummer, Mary H. Ruckelshaus, and Chris J. Harvey 17.1 Introduction 296 17.2 Ecosystem services provided by marine environments 297 Box 17.1: Nonlinear wave attenuation and the economic value of mangrove land-use choices (Edward B. Barbier ) 299 Box 17.2: Valuation of coral reefs in the Caribbean ( Emily Cooper and Lauretta Burke ) 301 17.3 Mapping and modeling the f ow of marine ecosystem services: a case study of Puget Sound 303 17.4 Future directions 317 References 318 18: Modeling the impacts of climate change on ecosystem services 323 Joshua J. Lawler, Erik Nelson, Marc Conte, Sarah L. Shafer, Driss Ennaanay, and Guillermo Mendoza 18.1 Introduction 323 18.2 Previous analyses of climate-driven changes in ecosystem services 323 18.3 Using ecosystem-service models to evaluate the impact of climate change on natural and human systems 324 Box 18.1: An estimate of the effects of climate change on global agricultural ecosystem services (David Lobell ) 325 18.4 Climate impacts on ecosystem-services in the Willamette Basin of Oregon 326 18.5 Discussion and conclusions 335 References 337 x C O N T E N T S

19: Incorporating ecosystem services in decisions 339 Emily McKenzie, Frances Irwin, Janet Ranganathan, Craig Hanson, Carolyn Kousky, Karen Bennett, Susan Ruffo, Marc Conte, James Salzman, and Jouni Paavola 19.1 Introduction 339 19.2 Putting ecosystem services on the agenda 340 Box 19.1: An assessment of ecosystem services helps a paper and packaging business respond to emerging risks ( Craig Hanson ) 343 19.3 Instruments for sustaining and enhancing ecosystem services 344 19.4 Choosing the right instrument 348 19.5 Building stronger organization 349 Box 19.2: Cultural evolution as an enabling condition for the use of ecosystem services in decisions ( Paul R. Ehrlich, Lee D. Ross, and Gretchen C. Daily ) 350 19.6 Future directions 352 References 352

Index 357 Contributors

W. L. (Vic) Adamowicz — Department of Rural Gretchen C. Daily — Department of Biology and Economy, University of Alberta, Edmonton, Natural Capital Project, Woods Institute for the Alberta, Canada T6G 2H1 Environment, Stanford University, 371 Serra Mall, Edward B. Barbier —Department of Economics and Stanford, CA 94305-5020, USA Finance, University of Wyoming, Laramie, WY Brett Day — Centre for Social and Economic Research 82071-3985, USA on the Global Environment (CSERGE), School of Karen Bennett — World Resources Institute, 10 G Environmental Sciences, University of East Anglia, Street NE, Suite 800, Washington, DC 20002, USA Norwich, NR4 7TJ, UK Kenneth Brooks —Department of Forest Resources, Eric Dinerstein —Conservation Science Program, University of Minnesota, 1530 Cleveland Avenue World Wildlife Fund-US, 1240, 24th Street NW, North, Saint Paul, MN 55108-6112, USA Washington, DC 20037, USA Berry Brosi— Department of Environmental T. Ka’eo Duarte — Land Assets Division, Kamehameha Studies, Emory University, 400 Dowman Dr., Ste. Schools, 567 South King Street, Suite 200, Honolulu, E510, Atlanta, GA 30322, USA HI 96813, USA Lauretta Burke — World Resources Institute, 10 G Paul R. Ehrlich —Department of Biology, 371 Serra Street NE, Suite 800, Washington, DC 20002, USA Mall, Stanford University, Stanford, CA 94305- Giorgio Caldarone — Land Assets Division, 5020, USA Kamehameha Schools, 567 South King Street, Driss Ennaanay —Department of Biology and Suite 200, Honolulu, HI 96813, USA Natural Capital Project, Woods Institute for the D. Richard Cameron —The Nature Conservancy— Environment, 371 Serra Mall, Stanford University, California, 201 Mission Street, 4th Floor, San Stanford, CA 94305-5020, USA Francisco, CA 94105, USA Cinzia Fissore —Department of Soil, Water, and Karen Carney — Stratus Consulting, Inc., 1920 L Street Climate, University of Minnesota, 439 Borlaug NW, Suite 420, Washington, DC, 20036, USA Hall, Saint Paul, MN 55108, USA Stephen R. Carpenter — Center for Limnology, David Freyberg — Department of Civil and University of Wisconsin, 680 North Park Street, Environmental Engineering, 473 Via Ortega, Madison WI 53706, USA Stanford University, Stanford, CA 94305, USA Kai M. A. Chan — Institute for Resources, Environment, Nicola Gallai — INRA, UMR406 Abeilles & and Sustainability, University of British Columbia, Environnement, 84914 Avignon Cedex 9, France Vancouver, BC, Canada and INRA, UMR LAMETA, 2 place Viala, 34060 Marc Conte — Department of Biology and Natural Montpellier Cedex 1, France Capital Project, Woods Institute for the Environment, Edward Game — The Nature Conservancy, PO Box 371 Serra Mall, Stanford University, Stanford, CA 5681, West End, QLD 4101, Australia 94305-5020, USA Renzo Giudice — Centre for Ecology, Evolution Emily Cooper — World Resources Institute, 10G and Conservation (CEEC), School of Biological Street NE, Suite 800, Washington, DC 20002, Sciences, University of East Anglia, Norwich, USA NR4 7TJ, UK

xi xii C O N T R I B U T O R S

Joshua Goldstein — Department of Human Dimen- Centre for Ecology, Evolution and Conservation sions of Natural Resources, Colorado State (CEEC), School of Biological Sciences, University of University, Fort Collins, CO 80523, USA East Anglia, Norwich, NR4 7TJ, UK, AND Centre Lawrence H. Goulder—Environmental and Resou- for Social and Economic Research on the Global rce Economics, Stanford University, Stanford, CA Environment (CSERGE), School of Environmental 94305, USA Sciences, University of East Anglia, Norwich, NR4 Sarah Greenleaf — Department of Biological 7TJ, UK Sciences, California State University, 6000 J Street, Jawoo Koo —International Food Policy Research Sacramento, CA 95819, USA Institute, 2033 K Street NW, Washington, DC Craig Groves — The Nature Conservancy, 520 E. 20006, USA Babcock Street, Bozeman, MT 59715, USA Carolyn Kousky —Resources for the Future, 1616 P Anne D. Guerry — Department of Biology and Street NW, Washington, DC 20036, USA Natural Capital Project, Woods Institute for the Claire Kremen —Department of Environmental Environment, 371 Serra Mall, Stanford University, Science, Policy and Management, University of Stanford, CA 94305-5020, USA AND Conservation California, 217 Wellman Hall, Berkeley, CA 94720- Biology Division, NOAA Northwest Fisheries 3114, USA Science Center, 2725 Montlake Boulevard E, Florence Landsberg— World Resources Institute, 10 Seattle, WA 98112, USA G Street NE Suite 800, Washington, DC 20002, Rajendra Gurung — World Wildlife Fund Nepal USA Program, Baluwatar, Nepal Joshua J. Lawler — School of Forest Resources, Neil Hannahs — Kamehameha Schools, 567 South University of Washington, Box 352100, Seattle, King Street, Suite 200, Honolulu, HI 96813, WA 98195, USA USA David Lobell — Environmental Earth System Science, Craig Hanson — World Resources Institute, 10 G Stanford University, 473 Via Ortega, Stanford CA Street NE, Suite 800, Washington, DC 20002, 94305, USA USA Eric Lonsdorf — Urban Wildlife Institute, Lincoln David Harrison — The Nature Conservancy, USA Park Zoo, 2001 North Clark Street, Chicago, IL Chris J. Harvey —Conservation Biology Division, 60614, USA NOAA Northwest Fisheries Science Center, 2725 Jianzhong Ma — The Nature Conservancy, China Montlake Blvd E, Seattle, WA 98112, USA Program, Yunnan Provincial Meteorological Lauren Hay —USGS Denver Federal Center, Building, 8th Floor, 77 Xi Chang Road, Kunming, Lakewood, CO, USA Yunnan Province, People’s Republic of China Norbert Henninger — World Resources Institute, 10 650034 G Street NE Suite 800, Washington DC 20002, Andrew R. Marshall — Research Fellow, Environ- USA ment Department, University of York, Heslington, Frances Irwin — World Resources Institute, 10 G York YO10 5DD, UK Street NE, Suite 800, Washington, DC 20002, Emily McKenzie — World Wildlife Fund and The USA Natural Capital Project, 1250 24th Street NW, Peter Kareiva — The Nature Conservancy, USA Washington, DC 20009, USA Donald Kennedy — Environmental Sciences, Guillermo Mendoza — National Research Council— Stanford University, Stanford, CA 94305, USA Research Associateships Program Fellow, 500 5th Kekuewa Kikiloi — Land Assets Division, Street NW, Washington, DC 20001 USA Kamehameha Schools, Honolulu, HI, USA Harold Mooney —Department of Biology, Stanford Christopher Kirkby — State Key Laboratory of University, 371 Serra Mall, Stanford, CA 94305- Genetic Resources and Evolution; Ecology, 5020, USA Conservation and Environment Center (ECEC), Claire Montgomery — Department of Forest Kunming Institute of Zoology, Chinese Academy Resources, Oregon State University, 205 Peavy of Science, Kunming, Yunnan, 650223, China, AND Hall, Corvallis, OR 97331, USA CONTRIBUTORS xiii

P.K.T. Munishi —Department of Forest Biology, James Regetz —National Center for Ecological Faculty of Forestry and Nature Conservation, Analysis and Synthesis (NCEAS), University of Sokoine University of Agriculture (SUA), PO Box California, 735 State Street, Suite 300, Santa 3010, Morogoro, Tanzania Barbara, CA 93101, USA Robin Naidoo — Conservation Science Program, Taylor H. Ricketts —World Wildlife Fund-US, 1250 World Wildlife Fund-US, 1250 24th Street NW, 24th Street NW, Washington, DC 20090, USA Washington, DC 20009, USA Lee D. Ross — Department of Psychology, Bldg Erik Nelson — Department of Economics, Bowdoin 420-Rm 380, Stanford University, Stanford, CA College, 9700 College Station, Brunswick, Maine 94305-2130, USA 04011–8497, USA Mary H. Ruckelshaus — Natural Capital Project, John Nieber — Department of Bioproducts and Woods Institute for the Environment, Stanford Biosystems Engineering, University of Minnesota, University, 371 Serra Mall, Stanford, CA 94305– 1390 Eckles Avenue, St Paul, MN 55108-3005, 5020, USA USA Susan Ruffo — The Nature Conservancy, 4245 North Hermann Oliveira-Rodrigues — CSR, Universidade Fairfax Drive, Suite 100, Arlington, VA 22203, Federal de Minas Gerais, Belo Horizonte, 31270- USA 901, MG, Brazil Jean-Michel Salles — CNRS, UMR LAMETA, 2 Nasser Olwero — Conservation Science Program, place Viala, 34060 Montpellier Cedex 1, France World Wildlife Fund-US, 1250 24th Street NW, James Salzman — Law School and Nicholas School Washington, DC 20037-1193, USA of the Environment and Earth Sciences, Duke Jouni Paavola — School of Earth and Environment, University, P.O. Box 90360, Durham, NC 27708, University of Leeds, Leeds, LS2 9JT, UK USA Stefano Pagiola — Economics Unit, Socially M. Sanjayan —The Nature Conservancy, 1011 Sustainable Development Department, Latin Poplar Street, Missoula, MT 59802, USA America and Caribbean, World Bank Terre Satterf eld —Institute for Resources, Environ- Liba Pejchar — Warner College of Natural Resources, ment, and Sustainability, University of British Colorado State University, Fort Collins, CO 80534- Columbia, Vancouver, BC, Canada 1401, USA Sarah L. Shafer — U.S. Geological Survey, 3200 SW Andrew J. Plantinga — Department of Agricultural Jefferson Way, Corvallis, OR 97331, USA and Resource Economics, Oregon State University, Sabina Shaikh —Public Policy Studies and Program 213 Ballard Extension Hall, Corvallis, OR 97331- on Global Environment, University of Chicago 4501, USA and RCF Economic Consulting, USA Mark L. Plummer — Conservation Biology Division, Manu Sharma —Natural Capital Project, Woods NOAA Northwest Fisheries Science Center, 2725 Institute for the Environment, 371 Serra Mall, Montlake Blvd E, Seattle, WA 98112, USA Stanford University, Stanford, CA 94305-5020, USA Stephen Polasky — Department of Applied Priya Shyamsundar — South Asian Network for Economics, Department of Ecology, Evolution Development and Environmental Economics and Behavior, University of Minnesota, 1994 Britaldo Silveira Soares-Filho — State Key Labora- Buford Avenue, St. Paul, MN 55108, USA tory of Genetic Resources and Evolution; Ecology, Simon G. Potts — Centre for Agri-Environmental Conservation and Environment Center (ECEC), Research, University of Reading, RG6 6AR, Kunming Institute of Zoology, Chinese Academy UK of Science, Kunming, Yunnan, 650223, China Jai Ranganathan —National Center for Ecological Luis Solorzano —Gordon and Betty Moore Analysis and Synthesis, 735 State Street, Suite Foundation, 1661 Page Mill Road, Palo Alto, CA 300, Santa Barbara, CA 93101, USA 94304, USA Janet Ranganathan — World Resources Institute, 10 Bill Stanley — Climate Change Team, The Nature G Street NE, Suite 800, Washington, DC 20002, Conservancy, 4245 North Fairfax Drive, USA Arlington, VA 22203, USA xiv CONTRIBUTORS

Charlotte Stanton — Emmett Interdisciplinary Eric Wikramanayake — Conservation Science Program in Environment and Resources, 397 Program, World Wildlife Fund-US, 1240 24th Panama Mall, Stanford University, Stanford, CA Street NW, Washington, DC 20037, USA 94305, USA Neal Williams — Department of Biology, Bryn Heather Tallis —Natural Capital Project, Woods Mawr University, Park Science Building, 101 N. Institute for the Environment, 371 Serra Mall, Merion Avenue, Bryn Mawr, PA 19010, USA Stanford University, Stanford, CA 94305, USA Rachel Winfree —Department of Entomology, State Christine Tam —Natural Capital Project, Woods University New Brunswick, 119 Blake Hall 93 Institute for the Environment, 371 Serra Mall, Lipman Drive, Rutgers, NJ 08901, USA Stanford University, Stanford, CA 94305, USA Stacie Wolny — Department of Biology and Natural C. Peter Timmer —Harvard Professor of Capital Project, Woods Institute for the Development Studies, emeritus, P.O. Box 1402, Environment, 371 Serra Mall, Stanford University, Kenwood, CA 95452, USA Stanford, CA 94305, USA R. Kerry Turner —Centre for Social and Economic Stanley Wood— International Food Policy Research Research on the Global Environment (CSERGE), Institute, 2033 K Street NW, Washington, DC School of Environmental Sciences, University of 20006, USA East Anglia, Norwich, NR4 7TJ, UK Ulalia Woodside — Land Assets Division, Bernard E. Vaissière — INRA, UMR406 Abeilles Kamehameha Schools, Honolulu, HI, USA & Environnement, 84914 Avignon Cedex 9, France Hazel Wong — The Nature Conservancy, USA Nathan Vadeboncoeur — Institute for Resources, Douglas W. Yu — State Key Laboratory of Genetic Environment, and Sustainability, University of Resources and Evolution; Ecology, Conservation British Columbia, Vancouver, BC, Canada and Environment Center (ECEC), Kunming Nicole Virgilio — Climate Change Team, The Nature Institute of Zoology, Chinese Academy of Science, Conservancy, 4245 North Fairfax Drive, Arlington, Kunming, Yunnan, 650223, China, AND Centre VA 22203, USA for Ecology, Evolution and Conservation (CEEC), Michael Todd Walter — Department of Biological School of Biological Sciences, University of East and Environmental Engineering, Riley Robb Hall, Anglia, Norwich, NR4 7TJ, UK Cornell University, Ithaca, NY 14853-5701, USA Jing Zhang — Department of Bioresource Policy, Yukuan Wang —Institute of Mountain Hazards, Business and Economics, University of Chinese Academy of Sciences, No.9, Block 4, Saskatchewan, Saskatoon, Saskatchewan, Renminnan Road, Chengdu, China Canada, S7N 5A8 Sue White — School of Applied Sciences, Building 53, Wei Zhang —International Food Policy Research Cranf eld University, Cranf eld, Bedfordshire, UK Institute Foreword

Getting there the policy-making community, one of the targeted audiences. Throughout the world ecosystem serv- In 1997 two books were published that focused on ice concepts are now being incorporated into devel- the signif cance of biological diversity for the welfare opment and strategic planning. The concept that of humankind. One was Yvonne Baskin’s delightful ecosystem services benef t society has resonated book on The Work of Nature: How the Diversity of Life with an extraordinary breadth of constituencies, Sustains Us ( Baskin 1997 ) and the second, the highly including the development community that has tra- inf uential volume edited by Gretchen Daily on ditionally viewed environmental priorities as an Nature’s Services: Societal Dependence on Natural impediment to development. Ecosystems (Daily 1997 ). These volumes provided a So, the case has been made, but the means to compelling rationale for conserving biological diver- practical utilization of the concepts and f ndings of sity as not only a social responsibility of society but the MA are not suff ciently developed for easy also as a necessity for human prosperity and sur- implementation. It was noted in the MA summary vival. The crucial interface between biological diver- that, “the scientif c and assessment tools and mod- sity and ecosystem services to human well-being had els available to undertake a cross-scale integrated f nally been made explicit in these seminal publica- assessments and to project future changes in eco- tions. These books marked a turning point in ecologi- system changes in ecosystem services are only now cal science and conservation. being developed” (MA 2005 ). The tools needed to The idea that ecosystem services provide an carry out assessments at local levels, in the frame- imperative for conservation became the launching work of the MA, have been aided greatly by the pad for the monumental Millennium Ecosystem recent publication of A Manual for Assessment Assessment (MA). Using the ecosystem service par- Practitioners ( Ash et al . 2010 ). adigm, the MA took a global view of the status and Now, this volume represents a major leap forward trends of ecosystems and the services they provide, in providing tools to utilize ecosystem service con- plausible scenarios of the capacity of ecosystems to cepts in decision-making. It has been produced by a deliver services in the future, and the response team with an unusual history. In 2006, three veter- options available to society that would lead to the ans of the MA formed a unique partnership to take continuance of the delivery of the vital services that the next step toward bringing ecosystem services underpin human endeavors. science into practice. Gretchen Daily (Stanford The MA was developed under the auspices of the University), Peter Kareiva (The Nature Conser- UN, and guided by wide representation including vancy), and Taylor Ricketts (World Wildlife Fund) those from intergovernmental conventions, NGO’s, founded the Natural Capital Project, dedicated to and industry. Over 1300 scientists were involved in producing quantitative tools for spatially explicit the production of a number of products that received valuation of ecosystem services, and applying them wide distribution. The social process that led to its in major resource decisions worldwide. The project development, as well as the resultant publications, has blended the muscle of a research university with sparked the interest of a wide audience including the practical perspectives and global networks of the

xv xvi F O R E W O R D two largest conservation organizations. From the in the search for practical approaches that will outset, Stephen Polasky (University of Minnesota) become mainstream and will further the goal of and Heather Tallis (Stanford University) have co-led society of conserving the biological diversity that the modeling efforts, and Minnesota has recently produces the ecosystem services vital for our future joined as a fourth formal partner. well-being. This team has produced a series of models for an Hal Mooney array of key ecosystem services that can be used in concert to provide scenarios of the land-use deci- A s h , N . , B l a n c o , H . et al., Eds. (2010). Ecosystems and sions on the subsequent delivery of a bundle of Human Well-Being. A Manual for Assessment Practioners. services. These models were designed for use by a Island Press, Washington, DC. wide range of practitioners and have the capacity to Baskin, Y. (1997). The Work of Nature. How the Diversity utilize input data of differing levels of resolution. of Life Sustains Us. Island Press, Washington, DC. The Natural Capital Project has not only developed Daily, G.D., Ed. (1997). Nature’s Services. Societal Dependence these models, but has been applying them in many on Natural Ecosystems. Island Press, Washington, DC. regions in the world. MA (2005). Millennium Ecosystem Assessment. Eco- There is no doubt that the application of the mate- systems and Human Well-being. Synthesis. Island Press, rial in this volume will provide a major step forward Washington, DC. How to read this book

This book is an outgrowth of the Natural Capital Capital website: http://naturalcapitalproject.org/ Project, which seeks to ensure that nature’s value is publications.html. accounted for in all of our business, policy, and We recognize that thousands of scientists and pol- development decisions. In this sense, the book icy makers around the world are striving to incorpo- depicts the “state of science” relevant to the Natural rate nature’s value into their work. There are Capital Project. Together, the chapters lay the scien- hundreds of stories that reveal personal and institu- tif c foundation for our project, and are linked by tional discoveries surrounding nature’s benef ts. In that common goal. More importantly, however, order to capture some of this exciting diversity of each chapter represents a general contribution to experience regarding the value of nature, we include modeling ecosystem services and connecting them essays throughout the book—set off in boxes—that to resource management. Each is written to “stand tell some story apart from the technical details of the alone” as much as possible. The book therefore need main text. These boxes can be read on their own. not be read front to back—read the chapters of inter- They are not intended to be part of the main course— est to you. they are appetizers. In order to cover a wide array of topics concern- Finally, we encourage all readers to join the coming nature’s benef ts and how those benef ts might munity of scientists and policy makers working to shape public and private decisions, we have had to create a world molded by better-informed decisions ask authors to leave out a lot of important detail. To that factor in the services nature provides for human f nd more details on the models one can go to well-being. We invite you all to go to the Natural http://naturalcapitalproject.org/InVEST.html, and Capital Project website and join a community of to f nd publications in technical journals about vari- users for the models described in this book. ous applications of these models go to the Natural The Editors

xvii Acknowledgments

This book was possible only because of the Walt Reid, Hal Mooney, Jane Lubchenco, Mary generous support of several foundations and Ruckelshaus, Buzz Thompson, Jeff Koseff, Jim donors. We are grateful for the support of the Salzman, Chuck Katz, Kerry Turner, Joshua National Science Foundation, the John D. and Galdstein, Neil Hannahs, Dennis White, Joshua Catherine T. MacArthur Foundation, the David Lawler, Jimmy Kagan, Stacie Wolny, Sue White, and Lucille Packard Foundation, the Gordon and Andrew Balmford, Neil Burgess, Mattieu Rouget, Betty Moore Foundation, the Google Inc. Charitable Kai Chan, Rebecca Shaw, Nasser Olwero, Jim Giving Fund of Tides Foundation, the Leverhulme Regetz, Yukuan Wang, Hua Zheng, Zhiyun Ouyang, Trust, the Resources Legacy Fund Foundation, the Li Shuzhuo, Mark Plummer, Robin Naidoo, George Winslow Foundation, and the National Center for Jambiya, Silvia Benitez, Eric Lonsdorf, Dick Ecological Analysis and Synthesis. Several indi- Cameron, and many others. We are also deeply viduals have also donated their time, advice, and grateful to the practitioners at WWF and TNC who f nancial support. In particular, we could not even continue to f eld test our ideas and give critical, have begun this project without start-up funding real-world feedback to our efforts. Finally, our from Vicki and Roger Sant, and Peter and Helen employers have given us the freedom to occasion- Bing. ally ignore our day jobs and get this book done Advice is cheap, but good advice is hard to come and for their tolerance and patience we are grate- by. We are lucky to have received advice from ful to Stanford University, WWF, The Nature some of the leading thinkers in conservation and Conservancy, and the University of Minnesota. science, including Paul Ehrlich, Steve McCormick, The Editors

xviii SECTION I A vision for ecosystem services in decisions This page intentionally left blank CHAPTER 1 Mainstreaming natural capital into decisions

Gretchen C. Daily, Peter M. Kareiva, Stephen Polasky, Taylor H. Ricketts, and Heather Tallis

1.1 Mainstreaming ecosystem services human well-being (MA 2005). The Millennium into decisions Ecosystem Assessment took a giant step forward in developing a widely shared vision, a conceptual The past several decades have produced tremen- framework, and a synthesis of existing knowledge. dous change in how people think about the envi- It spawned a suite of further efforts, including an ronment and human development. The focus of Intergovernmental Science-Policy Platform for environmental issues in the 1960s and 1970s was on Biodiversity and Ecosystem Services ( Mooney and air and water pollution with an immediate impact Mace 2009 ; Larigauderie and Mooney 2010 ). By on the local surroundings. Actions to reduce pollu- almost any measure—scientif c papers published, tion occurred primarily in relatively wealthy coun- media mentions, Google search trends—awareness tries able to afford it. of natural capital and efforts to sustain it have sky- Now the focus has expanded to encompass the rocketed since the Millennium Assessment. benef ts from (and losses to) living natural capital: The Millennium Ecosystem Assessment’s vision Earth’s lands and waters and their biodiversity. is starting to take hold. China, for instance, has Food and f ber production, provision of clean water, invested over 700 billion yuan (approximately $100 maintenance of a livable climate, security from billion) in ecosystem service payments over 1998– f oods, the basis for many pharmaceuticals, and 2010 ( Zhang et al. 2000 ; Liu et al. 2008 ). In addition, appreciation of the wonders and beauty of the natu- China has established a new system of “ecosystem ral world are a few of the many dimensions of function conservation areas,” spanning 25% of the human well-being that hinge on living natural capi- nation’s land area where the most vital elements of tal ( Daily 1997 ). living natural capital will be protected for securing The importance of maintaining natural capital for and harmonizing human and natural well-being. the ecosystem service benef ts that f ow from it is It is not just giant modernizing nations that are increasingly seen as vital in both poor and rich bringing a new view to nature. The value of natural countries alike. Indeed, declining natural capital capital is being included in decisions taken by com- poses a direct threat to rural poor since they depend munity leaders, traditional cultures, and global cor- closely on the environment for their livelihood porations. For example, payments for watershed ( Dasgupta 2010 ). After spending decades struggling service projects make up a signif cant portion of to fence off nature from people, conservation is existing ecosystem services schemes (many others emerging on the global stage with a new vision that relate to carbon) (Goldman et al. 2010 ). These schemes emphasizes the importance of connecting nature typically involve water users paying upstream land and people (Kareiva and Marvier 2007). managers for the delivery of clean, consistent water One of the largest efforts to date, the Millennium supplies (Brauman et al. 2 0 0 7 ; P o r r a s et al. 2007 ; Ecosystem Assessment, illustrated the many ways W u n d e r et al. 2008 ), and have in some places become in which natural systems are vital assets critical for more extensive and sophisticated in design (e.g., Nel

3 4 MAINSTREAMING NATURAL CAPITAL INTO DECISIONS

et al. 2009 ). In Hawai`i, policies and payments for a But who and what will catalyze the next giant wide array of services is being promoted through step forward? Part of the answer lies with improv- local watershed agreements, the state’s House ing science. The natural and social science commu- Concurrent Resolution on Ecosystem Services nities need to attack a set of diff cult and compelling (passed in 2006), the state’s Climate Bill (passed in issues: How can such complex processes as the role 2007), and leadership of the state’s largest private of forests in f ood control or crop pollination be landowner, Kamehameha Schools (Chapter 14 ). quantif ed accurately? How can such diverse values Companies including Coca-Cola, LaFarge, and as are embodied in cultural services be character- Mondi are evaluating the role of ecosystem services ized meaningfully? How can we make credible pro- within their supply chains and working to invest jections of natural capital under scenarios of change, in them ( Varga 2009 , WRI 2010, McKenzie et al. , such as in population, climate, or resource manage- Chapter 19 this volume). ment? And how can we build the capacity in civil Including the value of ecosystem services in the society and institutions—and in deep aspects of decisions of governments, corporations, tradi- human beliefs and behavior—to take account of tional cultures, and individuals does not replace or ecosystem services and natural capital? undermine the intrinsic value of nature, nor the This book tackles these science issues, while moral imperative to conserve it (e.g., Leopold 1949 ; acknowledging the many other social and political Norton 1987 ; Ehrenfeld 1988; Rolston 2000 ). elements to the problem. It is intended to supply Instead, valuing ecosystem services and natural one of the catalysts required for a new approach capital complements these moral concerns, broad- that harmonizes conservation and development. ening our understanding of the roles nature plays in our lives and the reasons for conserving it. If we 1.2 What is new today that makes can add how nature contributes to human well- us think we can succeed? being to the arguments for conservation, why wouldn’t we? An appreciation of ecosystems as valuable capital While the recent transformation in the way peo- assets traces back to Plato and doubtless much ear- ple think about nature and human development lier ( Mooney and Ehrlich 1997 ), and the current has been productive, the urgent challenge now is in research agenda on ecosystem services continues moving from ideas to action on a broad scale long-standing lines of work. For example, renewa- ( Carpenter et al. 2006 , 2009 ). Mainstreaming ecosys- ble resources have been an active area of study in tem services into everyday decisions requires a sys- economics since at least the 1950s, when Gordon tematic method for characterizing their value—and ( 1954 ), Scott ( 1955 ), and Schaefer ( 1957 ) character- the change in value resulting from alternative ized harvesting a biological stock and the problems polices or human activities. Unlike the well-estab- of open-access f sheries. In the 1960s and 1970s, lished accounting tools we apply to measure the economists set out to measure “the value of services value of traditional economic goods and services, that natural areas provide” ( Krutilla and Fisher we have no ready set of accounting tools to measure 1975 , p. 12) that included the value of renewable the value of ecosystem services (MA 2005; NRC resources ( Krutilla 1967 ; Clark 1990 ), non-renewa- 2005; Mäler et al. 2008 ). Absent these, ecosystem ble resources ( Dasgupta and Heal 1979 ), and envi- services are invariably undervalued or not valued ronmental amenities (Freeman 1993 ). More recent at all—by governments, businesses, and the public advances in a broad range of areas, such as in ecol- ( Daily et al. 2000 ; Balmford et al. 2002; NRC 2005; ogy and global change, economics, policy and insti- Dasgupta 2010 ). The result is continued losses in tutions, and especially their integration, have natural capital and biodiversity. Often, it is only broadened this work to include a wider set of eco- after their loss that we recognize the importance of system services and an examination of the set of ecosystem services, such as in the wake of Hurricane human actions needed to maintain the f ow the Katrina or cyclones in India (Stokstad 2005; Das and services (e.g., Dasgupta 2001 ; MA 2005; NRC 2005; Vincent 2009 ). Ruhl et al. 2007 ). MOVING FROM THEORY TO IMPLEMENTATION 5

Four big advances of the past decade promise to ing the multitude of benef ts derived from services make an old good idea a new beacon for real change. to various segments of society; understanding the First, the Millennium Ecosystem Assessment repre- decision-making process of individuals, corpora- sented a visionary and seminal step in global tions, and governments; integrating research with science—it was the f rst comprehensive global institutional design and policy implementation; assessment of the status and trends of all of the and crafting policy interventions that are designed world’s major ecosystem services. The key f nding for learning and improvement through time. Each of this assessment was that two-thirds of the world’s of these alone is a complex task; together they form ecosystem services were declining, a f nding that a daunting but critically important agenda requir- captured the attention of world leaders (MA 2005). ing a global collaboration. Second, the science of ecosystem functions and processes has made huge advances so that we can 1.3 Moving from theory now model (albeit with uncertainty) the impacts of to implementation land use and resource management decisions on a wide variety of ecosystem processes. Ecological sci- In moving from theory to practical implementation, ence has also become adept at spatially explicit Figure 1.1 presents a framework of the role that eco- modeling, which is essential for mapping ecosys- system services can play in decision-making (Daily tem services and their f ows to people (e.g., Chan et al. 2009 ). This framework connects the science of et al. 2006 ; Rokityanskiy et al. 2007 ; Bennett et al. quantifying services with valuation and policy to 2009 ; Nelson et al. 2009 ; Harrison et al. 2010 ). devise payment schemes and management actions Third, economic valuation methods have been that take account of ecosystem services. applied to the spatial provision of ecosystem serv- Though the framework is a continuous loop, we ices to estimate the monetary value of benef ts and start with the decisions oval to emphasize our the distribution of those benef ts to various seg- focus. After all, the main point of understanding ments of society (NRC 2005; Naidoo and Ricketts and valuing natural capital and ecosystem services 2006 ). In addition, qualitative and quantitative is improving natural resource decisions. So we methods from other social sciences have been start—and end—there. These decisions encourage applied to gain better understanding of the social and constrain actions relating to the use of land, and cultural importance of ecosystem services (e.g., water, and other elements of natural capital. MA 2005; US EPA 2009). Continuing clockwise around Figure 1.1 , “bio- Lastly, experiments in payments for ecosystem physical sciences” are central to understanding the services ( Pagiola et al . 2002 ; Pagiola and Platais 2007 ; Wunder et al. 2008 ), in ecosystem-based management (Barbier et al. 2008 ), and in regional plan- Decisions Actions & ning give us the empirical data for evaluating Incentives Scenarios approaches to valuing ecosystem services and incorporating values into decision-making. There is a growing recognition that bundling together of Institutions Ecosystems ecosystem services and explicit attention to trade- offs will both better inform decisions, and help Information Biophysical diverse stakeholders to appreciate the perspectives Models of others (e.g., Boody et al. 2005 ; Naidoo and Ricketts Value Services 2006 ; Egoh et al. 2008 ; Bennett et al. 2009 ; Nelson et al. 2009 ). Economic & Our challenge today is to build on this founda- Cultural Models tion and integrate ecosystem services into real Figure 1.1 A framework showing how ecosystem services can be decisions. Doing so requires understanding the integrated into decision-making. One could link any two ovals, in any interlinked; joint production of services; quantify- direction; we present the simplest version here. 6 MAINSTREAMING NATURAL CAPITAL INTO DECISIONS link between decisions and ecosystems , and along 1.4 Using ecosystem production with economics and social science, the links between functions to map and assess natural ecosystems and services. We study the former link capital with classic ecology and conservation biology to, for example, estimate impacts of land-use change There are several methods for mapping and assign- on biodiversity (e.g., Daily et al. 2001 ; Steffan- ing value to ecosystem services, each with its own Dewenter et al. 2007 ). And we pursue the latter link advantages and limitations. The initial valuation with “ecological production functions” that relate, work in the f eld of ecosystem services primarily say, forest condition and management practices to used what is called the benef t transfer approach the supply of carbon storage, pollination, and other (e.g., Costanza et al. 1997 ). This approach typically ecosystem services (e.g., Ricketts et al. 2004 ). uses empirical estimates of the value of goods pro- Social sciences are also central to understanding duced from some habitat type and transfers those the value of services to people (“economic and cul- benef ts to similar habitats elsewhere, including tural models”). Economic valuation techniques are anywhere in the world (Costanza et al. 1997 ). Local commonly used for this link, to place monetary knowledge can be used to adjust the benef ts value on natural capital. Value is often not fully cap- because one knows, for instance, that the west coast tured in monetary terms, though, so it is important marshes of North America are less productive than to characterize value in multiple dimensions, the east coast marshes and so forth. The general including, for example, health, livelihood support, idea, however, is to use lookup tables of benef ts per cultural signif cance (e.g., Dasgupta 2001 ). This will unit area of habitat type, and thereby quantify over- help ensure that valuation and broader decision- all natural capital. making approaches are inclusive of the range of An alternative method favored in this book is benef ts and people concerned ( Heal 2000a , 2000b ). called a “production function approach.” Instead of Finally, valuing ecosystem services provides use- relying on lookup tables, we build models that pre- ful information that can help shape institutions dict local ecosystem service supply based on land (e.g., agricultural markets, subsidies, land-use poli- cover, land use, ecosystem attributes, human cies, conservation NGOs) to guide resource man- demand, etc. These functions are analogous to those agement and policy. Having the right institutions long used in agriculture, which relate amounts of can create incentives so that the decisions of indi- water, fertilizer, and labor to resulting crop yield. In viduals, communities, corporations, and govern- our view, production functions have key advan- ments promote widely shared values. The links tages over benef ts transfer, and we delve into these between the value , institutions , and decisions further in Chapter 3 . ovals are much more the art and politics of social change than science, though scientists can inform 1.5 Roadmap to the book these debates if they target specif c decisions and are attuned to the social and political contexts. Our book begins with three chapters that introduce This idealized framework is helpful in clarifying the core approach and hypotheses of our work on the many frontiers of research and implementation natural capital. Chapter 2 examines the philosophi- en route to operationalizing ecosystem services into cal bases for ecosystem service value and explores decisions (see also Carpenter et al. 2009 ). This ways of measuring such value, distinguishing alter- includes continued biophysical research on the native approaches and highlighting some ethical impacts of human actions on ecosystems, all the issues underlying the choices among them. It also way to studies on the way landowners respond to explores the strengths and weaknesses of these conservation incentives. Chapters in this book touch measurement approaches, and indicates which on all ovals and all arrows within the framework, approaches are best suited to the different types of but the core chapters focus on moving from ecosys- value conferred by ecosystem services. Chapter 3 tems to services, and from services to value, using then introduces the modeling approach we have production functions and valuation techniques. developed, which strives to integrate many differ- ROADMAP TO THE BOOK 7 ent ecosystem services, to do so over scales appro- include enough math in each chapter to make the priate to important resource decisions, and to assess modeling approach clear. And we have implemented trade-offs among services on real landscapes. All tier 1 equations into a modeling tool available for resource decisions involve these trade-offs (e.g., free download at http://invest.ecoinformatics.org. between biological carbon sequestration and stream The f nal section of the book is based on potential f ow; Jackson et al. 2005 ). Yet, all too often, the applications of our approach to modeling and map- importance of trade-offs among services is lost in ping natural capital. Applications are messy and decision-making, with the result that unintended demanding, and require links to other f elds of sci- consequences arise while pursuing what at f rst ence as well as policy. In this “getting real” section seems like a good idea. of the book, we discuss trade-offs (Chapter 14 ), dif- The middle section of the book delves into details f cult choices about how complicated or detailed for each of the core models of ecosystem services. models need to be (Chapter 15 ), the implicit but The specif c services we model are water supply for rarely quantif ed link between ecosystem services hydropower and irrigation ( Chapter 4 ), f ood dam- and poverty (Chapter 16 ), the challenge of extend- age avoidance ( Chapter 5 ), water pollution regula- ing our approach to marine ecosystems (Chapter 17 ), tion (Chapter 6 ), carbon storage and sequestration assessing the impacts of climate change on ecosys- ( Chapter 7 ), production of timber and non-timber tem services ( Chapter 18 ), and ideas for how all of forest products ( Chapter 8 ), agricultural production this science might actually enter into decision-mak- ( Chapter 9 ), crop pollination ( Chapter 10 ), enhance- ing and policy ( Chapter 19 ). ment of recreation and tourism ( Chapter 11 ), and In all chapters we include short essays by con- provision of cultural services ( Chapter 12 ). We also tributors who are using concepts of natural capital model biodiversity, as an ecosystem attribute in their conservation and policy work. We include ( Chapter 13 ). these essays to emphasize that our models play Like all early efforts in modeling, we try to strike only one small part in a world of innovation sur- a balance between scientif c rigor, data availability, rounding natural capital. There has never been a and practical usability. Some will object that the more exciting time for conservation and ecosystem models are overly simplistic; others will f nd them science than now—but some of that excitement is hopelessly complicated (indeed, reviewers have shrouded in equations and modeling details. It is in made both arguments for almost every chapter). We our essays that one can f nd evidence of the tipping offer two tiers of models for each ecosystem service. point that is before us, in contemplating the African Tier 1 is the simplest credible model we could devise, boy cooking a monkey ( Box 1.1 , this chapter); the with data needs that can be met even in data-poor f rst national exchange for carbon storage credits regions that are often so fundamental to both con- (Chapter 7 ); the hopes we pin on agroecosystems as servation and human livelihoods. Tier 2 models highways out of poverty ( Chapter 16 ); vastly differ- offer more complexity, specif city, and realism for ent options for a sustainable future (Box 1.2 , this users and places with the data to support them. We chapter); and many others.

Box 1.1 The everyday meaning of natural capital to the world’s rural poor

M. Sanjayan stumbled onto this unfolding scene in a village in Sierra Leone, West Africa, whose inhabitants are amongst the The boy is no more than 10 years old, bare legs scarred by poorest in the world despite being surrounded by a wealth tropical parasites, clad only in dingy shorts despite the of biodiversity. The monkey is a Cercopithecus of some sort, threatening rain ( Figure 1.A.1 ). He is engrossed in his task perhaps a white-nosed Guenon, a relatively common of carefully burning the fur off a dead monkey. I have crop-raiding monkey in these parts. Holding it carefully in

continues 8 MAINSTREAMING NATURAL CAPITAL INTO DECISIONS

Box 1.1 continued

Figure 1.A.1 Young boy preparing to eat a monkey, rural Sierra Leone. Photo by M. Sanjayan.

both hands, the boy slowly turns it as the pelt singes and As I see it, six basic services provide most of the daily curls into soft gray ash. It is clearly a delicate task, with the needs of extremely poor rural people. Fresh water is the f ames struggling to catch the rain-soaked pieces of stick most obvious and its procurement is taxing, particularly to fed into the weak f re. Occasionally the boy pauses and, women, who bear most of the load. While there are taps with a piece of tin sheet metal, furiously fans the smoky sprouting in many rural villages in Africa, few connect to mess. An acrid odor hangs in the heavy air. sustainable water sources. Fuel wood, collected from The monkey will soon be food. It will be dismembered, forests, plantations, or local groves, is indispensible for the every bit from nose to tail, thrown into a pot with some heating and cooking needs of 40% of rural homes. okra, peppers, or other meager vegetables, and a few drops Gathering it is the second most taxing chore (after water) of palm oil—a stew ultimately yielding, based on the small that impacts daily life. Fisheries provide protein to 20% of crush of spectators, what I estimate to be about two the world’s population. On the Ganges River in India, for tablespoons of meat protein for each person. A small example, ten million people in 2000 villages depend on monkey in a big pot. Fascinated by the boy’s handiwork, f shing to both meet their daily needs and provide jobs. I pull out my camera and snap a photo—and immediately Fertility of soils, and its natural renewal through processes feel a little shabby about it. The boy just giggles. of nutrient cycling, is essential to places untouched by the Spend any amount of time quietly observing the daily Green Revolution and does not involve the consequences rituals of rural village life in any tropical country and you of industrial fertilization in terms of energy use and cannot but be impressed by the magnitude and diversity of nutrient overload. Forest products, like meat for protein services people derive from their immediate surroundings; from forest animals, fruit, honey, medicinal plants, and nature if you will. However, these services provided by f ber, have a myriad of vital uses. Fodder, in terms of grass nature are nearly always taken for granted. Local and browse for livestock, is important in rural communities communities are usually myopic in their understanding of because it is one of the few ways through which the poor similar communities, encircling what are to them virtually can access the global economy. Livestock is the common endless resources, and governments who can see the big bank for rural populations. These “6 Fs” of nature (six free picture are reluctant to acknowledge it lest they expose services) are part of the staple packet of goods and their shortcomings. Plus, changes are usually slow and services that virtually every rural community depends upon accumulate over a long period of time, thus hiding and that governments conveniently ignore, and non-prof ts cumulative impacts in the imperfect memories of the underestimate the importance of. Lose them and people elders. will suffer. OPEN QUESTIONS AND FUTURE DIRECTIONS 9

Later, when I look through my day’s pictures on my to more diff cult-to-assess services like fodder, we need to camera, it is not the dead monkey that draws my attention. ensure that we are adequately capturing the roughly It’s the sheet of tin metal, the one the boy used to fan the one-sixth of the world’s population who live on the very f ames that I focus on. It’s a piece of a signboard—one of margins of national or global economies but whose needs millions that line roads all over Africa, from charities, for such services are not just dire, but virtually irreplaceable. governments, and religious organizations. On this one, The value of water to a woman in West Africa trying to framed against a white background, are the blue prepare a meal is far higher (though not in strictly monetary letters—WFP—suspended over the image of a bundle of terms) than that to a woman f lling up her swimming pool grain. The World Food Programme. Here is this kid, fanning in California. Valuation of ecosystem services must properly the f ames with a sign board for a humanitarian organiza- recognize, and incorporate, this vast social net that nature tion, to burn the fur off a dead monkey from the forest, provides to the poor. which is what he is actually going to eat. For the poor, I think about returning to the village and bartering for nature often provides when governments and institutions the sign; hanging in my off ce, it would be a powerful don’t suff ce, and that is a powerful lesson. reminder of this insight. But I have nothing useful to trade. As we begin to better map, quantify, and assign Nothing sustainable. For now, the boy needs it more than monetary value on nature’s services, from carbon to water, I do.

1.6 Open questions and future have major inf uences on ecosystems and affect the directions services produced. Changes in economic conditions or fads in human behavior can similarly cause major This book and the modeling approaches we intro- changes in systems (e.g., f nancial crises). The occur- duce are only a beginning. We anticipate the research rence of each of these and other potential distur- community adding other ecosystem services over bances is diff cult to predict but virtually certain to time, as well as continually improving the models come about. Understanding their likely impacts on and data for those presented in this book. These ecological and social systems will help us prepare additions and improvements will come from con- for them. fronting these and other models with a variety of A second major area for further development is in real-world data and challenges. Here we mention relating ecosystem condition to human health. The two of the many key arenas in which further under- relationships between biophysical attributes of eco- standing is crucial. systems and human communities are complex First, major advances in methods and tools are ( Myers and Patz 2009 ). Destruction of ecosystems needed to incorporate dynamic effects, as well as can at times improve aspects of community health. shocks and uncertainty. Dynamic changes (e.g., in For example, draining swamps can reduce habitat climate and in the nitrogen cycle) and changes aris- for the mosquito vector that transmits the parasite ing through economic development and evolving that causes malaria. On the other hand, ecosystems human preferences over time are important to provide many services that sustain human health, include. The possibility of feedbacks within ecosys- for which substitutes are not available at the tems, and between ecosystem services and human required scale, such as purif cation and regulation behavior, are important areas for further develop- of drinking water f ow; regulation of air quality; ment. Feedback effects can give rise to thresholds nutrition (especially of protein and micronutrients); and rapid changes in systems that can fundamen- psychological benef ts; and, in complex ways, regu- tally alter system states ( Scheffer et al. 2001 ). The lation of vector-borne disease (Levy et al. in press). ability to incorporate shocks and the possibility of To date, there is little rigorous research establishing surprises is another area where further develop- the links between ecosystem conditions and human ment is needed. Fires, droughts, and disease all can health. 10 MAINSTREAMING NATURAL CAPITAL INTO DECISIONS

Box 1.2 Sorting among options for a more sustainable world

Stephen R. Carpenter 2050 under these and other scenarios. Technogarden and Adapting Mosaic were the most successful scenarios for “What gives you the most hope for the environment over maintaining ecosystem services. However, these two the next 50 years?” When we asked that question of 59 scenarios represent very different policies that lead to global leaders in 2003, we expected great variability in the different bundles of global ecosystem services by 2050. answers. To our surprise, the answers fell rather cleanly into In Technogarden , society addresses global environmental three clusters. Some respondents were not worried and problems such as climate heating, materials cycles, and simply had faith in economic growth, being convinced that nutrient mobilization through innovations in energy a sustainable environment would follow automatically from production, buildings, transportation, and agriculture economic development. Two clusters stood out, however, ( Figure 1.B.1 ). Improvements in agriculture and urban because they envisioned futures that were less automatic, design make it possible to feed and house humanity and that would need some guidance if we were to achieve without extensive new conversion of wild lands. Market a hopeful outcome. The f rst of these “we need change” mechanisms and sophisticated economic instruments are clusters thought that innovation and investments in deployed to manage ecosystem services. International environmentally friendly technology was the key. The cooperation on incentives for better technology lays the second “we need change” group felt that governance foundation for improved cooperation on other problems of should be restructured to motivate local innovation and the global commons, such as pelagic marine f sheries, learning, and thereby create sustainable landscapes from disease containment, and conservation of antibiotics. the bottom up. Expanded access to education leads to smaller family sizes These two clusters of ideas became the Technogarden and thereby slows the pace of population growth. Even and Adapting Mosaic scenarios of the Millennium though many ecosystem services are in sustainable Ecosystem Assessment (MA 2005). We evaluated the condition by 2050, there are some downsides. Some condition of 24 global ecosystem services from 2000 to aesthetic, cultural, and spiritual aspects of ecosystems are

Figure 1.B.1 Depiction of the Technogarden scenario of the Millennium Ecosystem Assessment. From Ecosystems and Human Well-Being: Scenarios , by the Millennium Ecosystem Assessment. Copyright © 2005 Millennium Ecosystem Assessment. Reproduced by permission of Island Press, Washington, D.C. OPEN QUESTIONS AND FUTURE DIRECTIONS 11

lost or irreversibly changed. Local ecological knowledge is are strong. Innovations and news of failed experiments can sometimes lost as management becomes more centralized. spread rapidly. The global network makes rather fast Unexpected consequences of technology lead to some big progress on improving practices for ecosystem services. accidents. Nonetheless, Technogarden offers many Ironically, however, the withering of global institutions successes in management of ecosystem services . hinders progress on problems of the global commons such Adapting Mosaic begins with reorganization of as climate heating and pelagic marine f sheries. governance around institutions tailored to naturally The scenarios of the Millennium Ecosystem Assessment occurring clusters of ecosystem services (Figure 1.B.2). are not a prescription for solving the world’s problems. For example, the Headwaters of the Missisippi River in They are more like hypotheses to be tested. So why not North America (Minnesota and Wisconsin, plus parts of combine the best of Technogarden a n d Adapting Mosaic ? eastern Iowa and northern Illinois) organizes around The global commons problems are critical, and improved sustainable agriculture mostly for local consumption, technology will be needed to create better energy sources, ecosystem management for abundant clean freshwater, agriculture, transportation, and infrastructure. Half of and urban areas known for environmentally friendly humanity lives in cities. The USA alone will replace most of high-technology and biotechnology industries. Within the its infrastructure by 2030, and in the 21st century the overarching Headwaters region, responsibility for world will erect more buildings than in the entire history of ecosystem management organizes around sub-water- our species before 2000. This reconstruction is an sheds at the smallest spatial extent, and major ecore- opportunity for lowering the impact of cities on the global gions at an intermediate spatial extent. Governance of commons and on the rural regions that feed the cities and most other regions of the world undergoes similar absorb their waste. At the same time we are addressing adjustments to accommodate natural patterns of the global commons, there are many benef ts available ecosystem services. from multiscale adaptive management of landscapes. Adapting Mosaic stresses local innovation and learning These benef ts can be f nanced through appropriate pricing by doing to improve ecosystem services. Even though of the ecosystem services that rural regions and wild global economic linkages are sparse, information linkages regions provide.

Figure 1.B.2 Depiction of the Adapting Mosaic scenario of the Millennium Ecosystem Assessment. From Ecosystems and Human Well-Being: Scenarios , by the Millennium Ecosystem Assessment. Copyright © 2005 Millennium Ecosystem Assessment. Reproduced by permission of Island Press, Washington, D.C. continues 12 MAINSTREAMING NATURAL CAPITAL INTO DECISIONS

Box 1.2 continued Even though the Millennium Ecosystem Assessment did not combinations of ecosystem services are biophysically possible, compute an optimal path to 2050, it is likely that the best mix the trade-offs among different bundles of ecosystem services of options is some combination of Technogarden and possible from a given region, and the institutional frameworks Adapting Mosaic . The challenge is to understand which that enable ongoing f ows of ecosystem services.

beginning to provide tools and methods that will 1.7 A general theory of change reduce the transaction costs. And there are enough Mainstreaming natural capital into decisions is a policy experiments underway that compelling long-term proposition, requiring co-evolving examples of natural capital stewardship enhancing advances in knowledge, social institutions, and cul- human well-being should be forthcoming. ture. Certainly a single book is not suff cient for Our book targets the f rst element of our theory: achieving this. We propose instead that our book to make quantifying and valuing natural capital contributes to an overall theory of change ( Bradach straightforward and routine. Science is not every- et al. 2008 ) involving three key, broad elements. thing, but both modeling and empirical science pro- First, businesses, governments, and individuals vide the foundation for action. The models we rely must f nd it easy to inculcate ecosystem services on are not a fait accompli —they are the f rst step in and natural capital into their decisions, and the an iterative process between basic science and methods for doing so must be transparent, credible, application to real-world problems. That is why we and predictable. In many cases, sectors of society highlight case studies in which valuation of natural are open to the idea of ecosystem services and natu- capital is being used to inf uence land and water ral capital, but simply do not know how to take the management. Science by itself cannot change the idea and use it in a concrete way. world, but science plus the vision and action of Second, there need to be examples of projects or leaders can—and that is what we seek. enterprises that—as a result of properly valuing ecosystem services and natural capital—end up References with improved decisions, institutions, and human well-being. These examples both test our science Balmford, A., Bruner, A., Cooper, P., et al. (2002). Economic reasons for conserving wild nature. Science , 297 , 950–3. against real-world problems and produce compel- Barbier, E., Koch, E., Silliman, B., et al. (2008). Coastal eco- ling stories of how an ecosystem services approach system-based management with non-linear ecological made a difference. functions and values. Science , 319 , 321–3. Lastly, political and thought leaders must appre- Bennett, E., Peterson, G., and Gordon, L. (2009). ciate these examples of success and spread the Understanding relationships among multiple ecosys- word. This is where the lessons of a set of examples tem services. Ecology Letters , 12 , 1394–1404. can be mainstreamed into the myriad decisions—by Boody, G., Vondracek, B., Andow, D., et al. (2005). businesses, governments, farmers, and banks—that Multifunctional agriculture in the United States. are made every year and that impact our natural BioScience , 55 , 27–38. world. This is where the impact of scattered projects Bradach, J., Tierney, T., and Stone, N. (2008). Delivering on can be magnif ed into worldwide change. the promise of nonprof ts. Harvard Business Review , Dec. , 88–97. None of these steps are complicated, and our the- Brauman, K., Daily, G. C., Duarte, T. K., and Mooney, H. A. ory of change does not require a brilliant and novel (2007). The nature and value of ecosystem services: An strategy. In fact, we are convinced that all three overview highlighting hydrologic services. Annual ingredients are within striking distance. The envi- Review in Environment and Resources , 32 , 67–98. ronmental movement has a much bigger and more Carpenter, S., DeFries, R., Dietz T., et al. (2006). Millennium diverse and powerful community behind it now Ecosystem Assessment: research needs. Science , 314 , than ever before ( Daily and Matson 2008 ). Science is 257–8. A GENERAL THEORY OF CHANGE 13

Carpenter, S., Mooney, H., Agard, J., et al. (2009). Science Gordon, H. (1954). The economic theory of a common- for managing ecosystem services: Beyond the property resource: the f shery. Journal of Political Economy , Millennium Ecosystem Assessment. Proceedings of the 62 , 124–42. National Academy of Sciences, USA , 106 , 1305–12. Harrison, J., Bouwman, A., Mayorga, E., and Seitzinger, S. Chan, K. M. A., Shaw, M. R., Cameron, D. R., et al. (2006). (2010). Magnitudes and sources of dissolved inorganic Conservation planning for ecosystem services. PLoS phosphorus inputs to surface fresh waters and the Biol , 4 , e379. coastal zone: A new global model. Global Biogeochemical Clark, C. (1990). Mathematical Bioeconomics: The Optimal Cycles , 24 , 271–9. Management of Renewable Resources, 2nd edn. Wiley, Heal, G. (2000a). Nature and the Marketplace: Capturing the New York. Value of Ecosystem Services. Island Press, Washington, Costanza, R., d’Arge, R., de Groot, R., et al. (1997). The DC. value of the world’s ecosystem services and natural Heal, G. (2000b). Valuing ecosystem services. Ecosystems , capital. Nature , 387 , 253–60. 3 , 24–30. Daily, G. C., Ed. (1997). Nature’s Services: Societal Dependence Jackson, R. B., Jobággy, E., Avissar, R., et al. (2005) Trading on Natural Ecosystems , Island Press, Washington, DC. water for carbon with biological carbon sequestration. Daily, G. C., Söderqvist, T., Aniyar, S., et al. (2000). The value Science , 310 , 1944–7. of nature and the nature of value. Science , 289 , 395–6. Kareiva, P., and Marvier, M. (2007). Conservation for the Daily, G. C., Ehrlich, P. R., and Sánchez-Azofeifa, G. (2001). people. Scienti f c American , 297 , 50–7. Countryside biogeography: Use of human-dominated Krutilla, J. (1967). Conservation reconsidered. American habitats by the avifauna of southern Costa Rica. Economic Review , 47 , 777–86. Ecological Applications , 11 , 1–13. Krutilla, J., and Fisher, A. (1975). The Economics of Natural Daily, G. C., and Matson, P. (2008). Ecosystem services: Environments: Studies in the Valuation of Commodity and From theory to implementation. Special Feature. Amenity Resources. Johns Hopkins University Press, Proceedings of the National Academy of Sciences, USA , 105 , Baltimore. 9455–6. Larigauderie, A., and Mooney, H. (2010). The Inter- Daily, G. C., Polasky, S., Goldstein, J., et al. (2009). governmental Science-Policy Platform on Biodiversity Ecosystem services in decision-making: time to deliver. and Ecosystem Services: moving a step closer to an Frontiers in Ecology and the Environment , 7 , 21–8. IPCC-like mechanism for biodiversity. Current Opinion Das, S., and Vincent, J. (2009). Mangroves protected vil- in Environmental Sustainability , 2 , 9–14. lages and reduced death toll during Indian super Leopold, A. (1949). A Sand County Almanac . Oxford cyclone. Proceedings of the National Academy of Sciences, University Press, Oxford. USA , 106 , 7357–60. Levy, K., Daily, G. C., and Myers, S. (in press). Ecosystem Dasgupta, P., and Heal, G. (1979). Economic Theory and services and human health: A conceptual framework. Exhaustible Resources. Cambridge University Press, Liu, J., Li, S., Ouyang, Z., Tam, C. and Chen, X. (2008). Cambridge. Ecological and socioeconomic effects of China’s policies Dasgupta, P. (2001). Human Well-Being and the Natural for ecosystem services. Proceedings of the National Environment , Oxford University Press, Oxford. Academy of Sciences, USA , 105 , 9489–94. Dasgupta, P. (2010). Nature’s role in sustaining economic MA (Millennium Ecosystem Assessment). (2005). Eco- development. Philosophical Transactions of the Royal systems and Human Well-being: The Assessment Series (Four Society B , 365 , 5–11. Volumes and Summary). Island Press, Washington, DC. Egoh, B., Reyers, B., Rouget, M., et al. (2008). Mapping eco- Mäler K.-G., Aniyar, S., and Jansson, A. (2008). Accounting system services for planning and management. for ecosystem services as a way to understand the Agriculture Ecosystems and Environment , 127 , 135–40. requirements for sustainable development. Proceedings Ehrenfeld, D. (1988). Why put a value on biodiversity? In of the National Academy of Sciences USA , 105 , 9501–06. Biodiversity, E. O. Wilson, Ed., National Academy Press, Mooney, H., and Ehrlich, P. (1997). Ecosystem services: A Washington, DC. fragmentary history. In G. Daily, Ed., Nature’s Services . Freeman, A. M., III (1993). The Measurement of Environmental Island Press, Washington, DC. and Resource Values: Theory and Methods . Resources for Mooney, H. and Mace, G. (2009). Biodiversity policy chal- the Future, Washington, DC. lenges. Science, 325 , 1474. Goldman, R., Benítez, S., Calvache, A., and Rámos, A. Myers, S. and Patz, J. (2009). Emerging threats to human (2010). Water funds: protecting watersheds for nature and health from global environmental change. Annual people. The Nature Conservancy, Arlington. Review of Environment and Resources , 34 , 223–52. 14 MAINSTREAMING NATURAL CAPITAL INTO DECISIONS

Naidoo, R. and Ricketts, T. H. (2006). Mapping the eco- Ruhl, J., Kraft, S., and Lant, C. (2007). The Law and Policy of nomic costs and benef ts of conservation. PLoS Biology , Ecosystem Services . Island Press, Washington, DC. 4 , e360. Schaefer, M. B. (1957). Some considerations of population Nel, D., Marais, C., and Blignaut, J. (2009). Water neutral- dynamics and economics in relation to the manage- ity: A f rst quantitative framework for investing in water ment of marine f shes. Journal of the Fisheries Research in South Africa. Conservation Letters , 2 , 11–18. Board of Canada , 14 , 6 6 9 – 8 1 . Nelson, E., Mendoza, G., Regetz, J., et al. (2009). Modeling Scheffer, M., Carpenter, S. R., Foley, J. et al. (2001). multiple ecosystem services, biodiversity conservation, Catastrophic shifts in ecosystems. Nature , 413 , 591–6. commodity production, and tradeoffs at landscape Scott, A. (1955). The f shery: the objectives of sole owner- scales. Frontiers in Ecology and the Environment , 7 , 4–11. ship. Journal of Political Economy , 63 , 116–24. Norton, B. (1987). Why Preserve Natural Variety? Princeton Steffan-Dewenter, I., Kessler, M., Barkmann, J., et al. (2007). University Press, Princeton. Tradeoffs between income, biodiversity, and ecosystem NRC (National Research Council) (2005). Valuing Ecosystem functioning during tropical rainforest conversion and Services: Toward Better Environmental Decision-Making . agroforestry intensif cation. Proceedings of the National National Academies Press, Washington, DC. Academy of Sciences USA , 104 , 4973–8. Pagiola, S., Bishop, J., and Landell-Mills, N. (2002). Selling Stokstad, E. (2005). Louisiana’s wetlands still struggle for Forest Environmental Services . Earthscan, London. survival. Science , 310 , 1264–6. Pagiola, S. and Platais, G. (2007). Payments for Environmental US EPA (United States Environmental Protection Agency) Services: From Theory to Practice . World Bank, Washington, Science Advisory Board. (2009). Valuing the Protection of DC. Ecological Systems and Services . EPA-SAB-09-012. US Porras, I., Grieg-Gran, M., and Neves, N. (2007). All That EPA, Washington, DC. Glitters: A Review of Payments for Watershed Services in Varga, A. (2009). Payments for Ecosystem Service: An Analysis Developing Countries. International Institute of of Cross Cutting Issues in Ten Case Studies. Columbia Environment and Development, London. University, New York. Ricketts, T. H., Daily, G. C., Ehrlich, P. R., and Michener, C. WRI (World Resources Institute). (2010). Lafarge Presque (2004). Economic value of tropical forest to coffee pro- Island Quarry: Corporate Ecosystem Services Review. World duction. Proceedings of the National Academy of Sciences, Resources Institute, Washington, DC. USA , 34 , 12579–82. Wunder, S., Engel, S., and Pagiola, S. (2008). Taking Rokityanskiy, D., Benítez, P., Kraxner, F., et al. (2007). stock: a comparative analysis of payments for Geographically explicit global modeling of land-use environmental service programs in developed and change, carbon sequestration, and biomass supply. developing countries. Ecological Economics , 65 , Technological Forecasting and Social Change , 74 , 1057–82. 834–52. Rolston, H., III (2000). The land ethic at the turn of the mil- Zhang, P., Shao, G., Zhao, G., et al. (2000). China’s forest lennium. Biodiversity and Conservation , 9 , 1045–58. policy for the 21st century. Science , 288 , 2135–6. CHAPTER 2 Interpreting and estimating the value of ecosystem services

Lawrence H. Goulder and Donald Kennedy

basis, we contrast competing approaches to value 2.1 Introduction: why is valuing nature and bring out some ethical issues underlying the important? choice among different approaches. Many of the critical ecosystem services generated The other component is to lay out various meth- by natural capital (such as pollination services, ods for measuring the values of ecosystem services, f ood control, water f ltration, and provision of hab- and to consider the strengths and limitations of itat for biodiversity) are externalities—they are not these approaches. Quantitative assessments of eco- given a price in markets. As a result, unfettered system service value have become much more markets often lead to the compromising or collapse widespread in recent years. The expanding litera- of ecosystems, much to the detriment of human ture now includes estimates of the value of such welfare. Oftentimes society would benef t from ecosystem services as pollination, pest control, and greater protection of ecosystems and their services water purif cation. These assessments are begin- than results from unregulated markets. ning to play a signif cant role in the formulation of Public policy has a crucial role to play in regulat- land-use policies. ing or inf uencing markets so as to prevent them Setting out the values of ecosystem services to from producing unfortunate societal outcomes. Yet society provides a basis for making public policy decisions about such public policies are often con- decisions. However, it is not the only basis. As we tentious. Agricultural interests will vie for greater discuss brief y below, it may make sense to con- ability to purchase wetlands and convert them sider as well whether a policy decision is consist- through drainage to agricultural land. Urban devel- ent with preserving the intrinsic rights of the opers will push to convert open space to new hous- various organisms or ecosystems that might be ing tracts. affected by the decision. If intrinsic rights are Perhaps the most important basis for supporting involved, it is reasonable to restrict the set of seri- a policy that would protect otherwise threatened ous alternatives to those that are consistent with ecosystem services is evidence that society gains these rights. more value from such protections than it gives up. The chapter is organized as follows. Section 2.2 Providing such evidence requires an understanding examines alternative philosophical foundations for of the biophysical processes involved, that is, the valuing living things and ecosystems. It also con- various services offered by the ecosystem in ques- siders how attention to intrinsic rights might sup- tion. It also requires an assessment of the benef ts to plement or even offer an alternative to a well-being—or values to society—of these ecosys- consideration of values. This philosophical discus- tem services. sion lays the groundwork for Section 2.3 ’s examina- This chapter clarif es how such an assessment of tion of empirical valuation methods. In Section 2.4 ecosystem services can be made. It has two main we indicate some valuation problems that arise in a components. One is to examine the philosophical few specif c real-world cases. Our f nal section basis of ecosystem service value. In considering this draws conclusions.

15 16 INTERPRETING AND ESTIMATING THE VALUE OF ECOSYSTEM SERVICES

2.2 Philosophical issues: values, rights, nutrients for other living things that in turn feed and decision-making humans). The anthropocentric approach does not restrict value to forms of nature that are consumed: 2.2.1 Competing philosophical approaches to there are both consumptive and non-consumptive use value values. Examples of the former are the values that 2.2.1.1 The anthropocentric approach might be attached to ducks insofar as they provide From what do nature’s values derive? When we food. Examples of the latter are the values ducks claim that a given living thing or species or habitat provide in the form of pleasure to bird-watchers. is worth such-and-such, what is the basis of that Satisfactions also include non-use values: values claim? that involve no actual direct or indirect physical Among US policy analysts, the prevailing involvement with the natural thing in question. The approach to value is anthropocentric. This approach most important value of this type may be existence claims that natural things (indeed, all things) have value (or passive use value)—the satisfaction one value to the extent that they confer satisfactions to enjoys from the pure contemplation of the existence humans. It stipulates that value is based on the of some entity. For example, a New Jersey resident ability to give utility (or well-being) to humans. who has never seen the Grand Canyon and who Economists tend to support this viewpoint which, never intends to visit it can derive satisfaction sim- as we discuss below, is inherent in benef t–cost ply from knowing it exists. As another example, analysis. many people experienced a loss of satisfaction or At f rst blush, this anthropocentric approach well-being simply from learning of the ecological might seem inconsistent with safeguarding the damage resulting from the Bluewater Horizon oil planet or protecting non-human forms of life. But spill in the Gulf of Mexico in 2010. This was a loss of the approach does not necessarily imply a ruthless existence value. exploitation of nature. On the contrary, it can be The array of services provided by ecosystems consistent with the fervent protection of non-human spans all of these categories of values. The pest con- things, both individually and as collectivities. After trol and f ood control services they offer have direct all, we may feel that the protection of nature or par- use value to nearby agricultural producers. Their ticular non-human forms of life is important to our provision of habitat for migratory birds confers an satisfaction or well-being, and thus we may place a indirect use value for people who enjoy watching high value on these forms. The anthropocentric them (non-consumptive) or hunting them (con- approach does not rule out our making substantial sumptive). Ecosystems also yield an existence value: sacrif ces to protect and maintain other living wetlands, for example, provide such value to peo- things. However, it asserts that we should assign ple who simply appreciate the fact that wetlands or value (and therefore help other forms of life) only their services exist. insofar as we humans gain satisfaction or well-being The fundamental assertion of the anthropocentric from doing so. The notion of satisfaction here should approach is that the value of a given species or form be interpreted broadly, to encompass not only mun- of nature to an individual is entirely based on its dane enjoyments (as with consuming plants or ani- ability to yield satisfaction to that person (directly mals for food) but also more lofty pursuits (such as or indirectly). Benef t–cost analysis invokes the marveling at the beauty of an eagle). anthropocentric approach, while introducing a fur- Anthropocentric value can be categorized accord- ther assumption—that the value to society of the ing to the way the satisfaction is generated. Use natural thing is the sum of the values it confers to value refers to satisfaction that involves (directly or persons. indirectly) a physical encountering with the object Benef t–cost analysis offers a rather convenient in question. There are direct use values (for exam- way of measuring the overall social values of alter- ple, the satisfaction from catching or eating trout) as native policies. Thus it provides a basis for making well as indirect use values (for example, the value diff cult policy decisions. It seeks to ascertain in that can be attached to plankton because it provides monetary terms the gain or loss of satisfaction to PHILOSOPHICAL ISSUES: VALUES, RIGHTS, AND DECISION-MAKING 17 different groups of humans under each of various 2.2.1.2 A biocentric approach policy alternatives. Under each alternative, it adds The biocentric approach offers another basis for up the gains and subtracts the losses, and then value. It asserts that value consists in the ability to compares the net gains across policy options. provide well-being or utility to humans and to other Importantly, benef t–cost analysis often does not species. Under the anthropocentric approach, the differentiate between one person’s valuation of a well-being of other species counts only indirectly: given species and another’s—that is, each person’s such well-being is important only to the extent that valuation receives the same weight as another’s. it contributes to human well-being. In contrast, Many times, no attempt is made to correct for dif- the biocentric approach gives weight directly to the ferences in awareness, education, or “enlighten- well-being of other species. Thus, it allows for the ment” among individuals. The preferences of possibility that another species will have value even people who have no concern for future generations, if it does not confer satisfaction directly or indirectly or who have no sense of the ecological implications to humans. This independent value is sometimes of their actions, are often counted equally with those referred to as intrinsic value. of people who are more altruistic or who recognize Defenders of the anthropocentric approach more fully the fragility of ecosystems. point out that since human beings are the domi- S u c h b e n e f t–cost analyses are non-discriminat- nant species on the planet; they are obliged to ing, perhaps to a fault. Consider the fact that pref- def ne ethical principles in terms of human wants erences change. They may change for a given and needs. However, biocentrists can counter by person over his or her lifetime, or from generation pointing out the following implication of anthro- to generation. To impute values for future genera- pocentric logic. Suppose that representatives of tions (such as the value that future generations another species should arrive from outer space, might place on certain ecosystem functions), bene- a species clearly superior to human beings in f t–cost analysis must impute preferences to these intelligence, perceptiveness, and technological generations. Clearly, this can only involve guess- know-how. To the extent that defenders of anthro- work. Usually benef t–cost analyses assume that pocentrism have invoked the “dominant species” future generations’ preferences are similar to those argument, consistency would require humans to of the current generation. Costanza et al. ( 1 9 9 5 ) allocate some decision-making authority to this indicate that preferences seem to evolve toward an other species, no matter whether humans like increasing concern for sustainability. They consider their decisions or not. Human well-being would the notion that this natural evolution of preferences count only insofar as it served the well-being of ought to be accounted for in social decisions—that the superior species. This may seem troubling to more evolved, developed preferences deserve many of us! What if the dominant species felt it greater weight in analyses of policy options. was best to exterminate humans? This reductio ad However, some benef t–cost analyses do in fact absurdum argument has been invoked to support a give special attention to the assessments offered by biocentric approach that gives weight directly to a experts. range of species. Many ecologists are uneasy with the tendency of While the biocentric approach may have some benef t–cost assessments to give considerable appeal, it is diff cult to implement. As discussed weight to valuations made by relatively unin- below, “willingness to pay” offers a measure of the formed individuals. There is a basic appeal to the change in well-being to humans generated by a idea that the preferences of some individuals— given policy change to protect nature or environ- particularly those who are better informed or have mental quality. No comparable measure is currently more relevant expertise—should count more. But available for assessing changes in satisfaction to it is very diff cult to arrive at an objective standard other species or communities of them. Also, it is dif- for “relevant expertise.” Philosophers offer vary- f cult to draw a clear line between biocentric value ing viewpoints as to what’s appropriate here (NRC and certain anthropocentric values. When individu- 2004). als call for a biocentric approach, they may actually 18 INTERPRETING AND ESTIMATING THE VALUE OF ECOSYSTEM SERVICES be expressing the anthropocentric satisfaction they dential housing) had any value at all, then accord- would gain if that approach were followed. For ing to the anthropocentric approach this is the best example, when someone calls for the preservation option for society. of a given habitat on the grounds that the species Based on examples of this sort, some philoso- residing there has intrinsic value, that individual phers argue that the fate of other species becomes may really be revealing the (anthropocentric) exist- too precarious when it must depend on a link to ence value that the species provides. It thus becomes human values or satisfactions. (See, for example, diff cult to distinguish biocentric intrinsic value Skidmore 2001.) An intrinsic rights approach pro- from existence value, which suggests that the bio- vides an entirely different basis for decision- making. centric approach may be superf uous. When intrinsic rights are involved, then the appropriate social decision must respect those rights. Attention to intrinsic rights can in some cases com- 2.2.2 Intrinsic rights: a further consideration plement the weighing of values. In such cases, pol- Under the value-based approaches just discussed, icy makers would f rst restrict the set of options to social decisions are to be made based on a com- be considered to those that respect intrinsic rights. parison of values. If Policy A generates greater Within this restricted set, policy makers would then value than Policy B, then Policy A should be given choose the option that yielded the highest value. preference over Policy B. Consider in particular In other cases, an attention to intrinsic rights is the anthropocentric approach to value. If a given a fully dispositive. This applies, for example, when species or other element of nature does not convey any change to a given habitat would violate the satisfaction to human beings directly or indirectly, claimed intrinsic rights of the species that currently then according to this approach it should be given reside there. In such circumstances, a defender of no value. It must produce no use value, either intrinsic rights could argue that the value-based directly or indirectly. Thus, it must be something approach is inappropriate: any comparison of ben- we do not enjoy eating (there is no consumptive ef ts (values gained) and costs (values sacrif ced) is use value) and something we do not enjoy observ- not justif ed. Many analysts argue that species and ing (there is no non-consumptive use value). In natural communities have intrinsic rights to exist addition, the organism must not serve any positive and prosper. They claim that, consequently, society ecosystem function (there must be no indirect use should uphold these rights irrespective of the val- value). Also it must be the case that we are certain ues gained (benef ts) or sacrif ced (costs) in the that humans’ tastes and ecosystem function will process. not change to give rise to a future use value. To Arguments for intrinsic rights are not entirely complete the picture, the organism must also have independent of references to well-being or satisfac- zero existence value—humans must not enjoy con- tions of other species. For example, in Animal templating this thing. Is there any real-world Liberation, ethicist Peter Singer argues that non- organism that f ts this picture? Perhaps some lowly human animals have the basic right to be spared of species of cockroach comes close. Whether it suffering deliberately caused by humans (Singer exactly f ts the picture is not important. The key 1975). This argument is grounded in the notion that, point is that such a creature would be given virtu- like humans, other animals are sentient creatures, ally no value in a benef t–cost analysis. This means capable of experiencing pleasure and pain, and that that if we are considering a development project there is something fundamentally wrong about that threatens its existence, this threat does not causing pain to any creature. However, even though cause us to refrain from undertaking the project. the call for intrinsic rights may be based on a con- As long as there are some benef ts from the project cern for well-being, it proposes a very different and no other, “signif cant” form of life is put at basis for decision-making: the appropriate social risk, we would not prevent the loss of this particu- decision must respect intrinsic rights. A policy that lar species. If destroying the habitat and putting satisf es a benef t–cost test should be rejected if it the area involved to an alternative use (e.g., resi- violates intrinsic rights. PHILOSOPHICAL ISSUES: VALUES, RIGHTS, AND DECISION-MAKING 19

the broader public: to protect every species would 2.2.3 Public policy’s inconsistent approach require far more funds than the public generally to decision-making wishes to devote to this purpose. People want spe- When should a value-based approach be employed, cies protection, but they also want funds for other and when should attention to intrinsic rights sup- things (e.g., education, defense, and their own ply the primary basis for decision-making? And consumption). which of the two approaches do societies in fact The case of the Endangered Species Act is not adopt? United States environmental policy adopts unusual. The US Clean Air offers another example. both the anthropocentric value approach (via bene- The mandate is broad: setting air quality standards f t–cost analysis) and an intrinsic rights approach, tight enough to assure an “adequate margin of and often acts inconsistently. Oftentimes the man- safety” to all individuals. There is no reference to a date for a particular environmental law will embrace comparison of benef ts and costs. Yet in the actual the intrinsic rights approach, but actual implemen- establishment of the standards, the EPA paid close tation yields to a value-based approach. attention to costs, and the ultimate standards A key example is the US Endangered Species Act, imposed were not tight enough to prevent serious passed in 1973 after a previous Act was brought up health problems or premature mortality to the most to date and linked to the Convention on International pollution-sensitive individuals. Trade in Endangered Species. In addition to def n- In many instances there is a fundamental incon- ing the status of “endangered” and “threatened,” it sistency between the stated objectives of environ- made eligible for protection all plants and inverte- mental laws and the way the laws are implemented. brates, and prohibited the “taking” of all endan- Lawmakers and the public may experience rewards gered animals. “Taking” included destruction of from establishing broad mandates that declare essential habitat. Federal agencies were required to intrinsic rights. At the same time, they are free to use their authority to conserve listed species and implement the laws much more restrictively; so that were prohibited from undertaking actions that people need not sacrif ce as many other things as would jeopardize listed species or modify their crit- they would had they enforced intrinsic rights fully. ical habitats. We do not suggest that society must choose There is an assumption here that certain species between the universal application of an intrinsic under threat have an intrinsic right to exist. rights approach and the across-the-board adoption However, when it comes to actual implementation of a value-based approach. In certain circumstances, of the Endangered Species Act, the Congress only it may be best to invoke intrinsic rights. In the Clean allocated funds suff cient for protecting a small Water Act, Congress essentially found that the pop- fraction of species that may qualify for the designa- ulation of the United States had an intrinsic right to tion of threatened or endangered. Based on threat clean water. In other circumstances the appeals to criteria and the availability of appropriated funds, intrinsic rates are fairly rare; for example, there are the Interior Department decided that some species few claims that farmers enjoying the pollination are more worthy of protection than others. In effect, services provided to agriculture by bees inhabiting it adopted an anthropogenic, value-based approach, nearby natural habitats have an intrinsic right to with the priorities ref ecting the range of values that such services. Are these pollination services some- people place on different species. Charismatic meg- what less “fundamental” than the various services afauna like the wolf or the peregrine falcon get more offered by clean water? Does this explain the rela- protection than the white-footed mouse. The anthro- tive infrequency of claims that woodland-based pocentric, value-based approach involved in imple- pollination services are an intrinsic right? mentation contradicts the intrinsic rights basis of Even if one adopts a value-based approach rather the mandate declared by Congress. than an intrinsic rights approach in making policy One might be tempted to fault Congress for fail- decisions, this does not necessarily preclude ing to allocate enough funds to protect all species. invoking additional evaluation criteria in the However, the allocation ref ects the preferences of decision process. Benef t–cost analysis considers 20 INTERPRETING AND ESTIMATING THE VALUE OF ECOSYSTEM SERVICES the aggregate values gained (benef ts) and values natural amenity is revealed by the amount that peo- sacrif ced (costs) of a given policy option. It usually ple would be willing to pay or sacrif ce in order to does not focus on how the benef ts and costs are dis- enjoy it. Willingness to pay is thus the measure of tributed across members of society (rich vs. poor, satisfaction. current generations vs. future generations, etc.; see It is important to be clear as to what is meant by Chapter 16 for further discussion). The distribution “willingness to pay.” It is not always an actual, of policy impacts is important and deserves atten- expressed willingness; it is not restricted to what tion as well. Other evaluative criteria (minimization we observe from people’s actual payments in mar- of risk and political feasibility) can also be impor- ket transactions. Rather, it represents a kind of tant. Thus, the results of a benef t–cost study psychological equivalence. Suppose a project may not be suff cient to settle the question of which would improve water quality an individual enjoys. policy is best. That individual’s willingness to pay is the income sacrif ce that just brings the individual back to his 2.3 Measuring ecosystem values or her original utility level after the improvement in water quality. More formally, the willingness to Although attention to values need not be the sole pay W for a given improvement to environmental criterion for decision-making, we believe it is suff - quality Q i s t h e v a l u e W that leads to the following ciently important to justify a focus on how to meas- equality: ure various values. Here we describe central methods for measuring anthropocentric value. U(Q + DQ, Y – W) = U(Q, Y), (2.1) Considerable progress has been made over the years in developing such methods. However, the where U stands for the individual’s utility and Y science is far from perfect. Controversies persist. is the individual’s income. Willingness to pay Ecosystem services are especially diff cult to expresses the maximum payment an individual measure for the same reason that ecosystems are would make that just compensates for (or undoes) threatened. Many of the services provided by eco- the utility gain from the environmental improve- systems are positive externalities. The f ood control ment. It is the size of the payment that, if made, benef ts, water f ltration services, and species-sus- would keep the person’s utility from changing. It is taining services offered by ecosystems are usually not necessarily what people say they are willing to external to the parties involved in the market deci- pay. In some cases, markets indicate individuals’ sion as to whether and at what price a given habitat true willingness to pay as def ned above: for exam- will be sold. As a result, the habitats that support ple, the market price of a tomato might indicate complex ecosystems tend to be sold too cheaply in what consumers are willing to pay at the margin the absence of public intervention, since important for this product. But in other circumstances research- social benef ts are not captured in the price. Public ers need to rely on other, more indirect methods to attention to the values of these (largely external) fathom it. benef ts is important for providing support for reasonable public policies to protect important habi- 2.3.2 Methods for measuring the values tats. This makes it all the more important to of ecosystem services determine the values of these services. Above, we have distinguished various types of ecosystem service values. The ecosystem services 2.3.1 Willingness to pay themselves can be placed in various categories as As indicated, under the anthropocentric approach well. As in the Millennium Ecosystem Assessment the value of a given living thing is the amount of (2003), we will distinguish provisioning, regulating, human satisfaction that thing provides. How could and other services offered by ecosystems. Below, we such satisfaction be measured? Nearly every empir- consider the types of values associated with each of ical approach assumes that the value of a given these major categories of service. Different types of MEASURING ECOSYSTEM VALUES 21

Table 2.1 Ecosystem services and valuation methods

Services Types of values offered Valuation method

Provisioning services Sustenance of plants and animals Direct use values —Consumptive Direct valuations based on market prices —Non-consumptive Indirect valuations (revealed expenditure methods, contingent valuation method) Indirect use values (No valuations necessary if plants/animals with direct values are counted) Regulating services Water f ltration, f ood control, pest control, Direct and indirect use values Estimation of service’s contribution to prof t (holding all else pollination, climate stabilization constant)

Other services Generation of spiritual, esthetic, and cultural Existence value Indirect valuations (contingent valuation method) satisfaction Direct, non-consumptive use value Indirect valuations (revealed expenditure methods, contingent valuation method) Recreational services Non-consumptive direct use value Indirect valuations (revealed expenditure methods, contingent (e.g., from bird-watching) valuation method) Generation of option value * Option value Empirical assessments of individual risk-aversion

* Option value represents a component of the overall value offered by a potential future ecosystem service, supplementing other values attributed to this potential service. See discussion text. valuation techniques are called for, depending on Direct, consumptive use values. When direct use the category of service involved. Table 2.1 shows values are involved, two main valuation methods the relationships between service types and valua- may apply. In the case of direct consumptive use tion methods. values, one can employ direct valuation methods based on market prices. When natural ecosystems 2.3.2.1 Valuing the provisioning services provide a habitat for animals that are harvested and of ecosystems sold commercially, the commercial market value A s s u g g e s t e d b y T a b l e 2 . 1, a g e n e r a l t y p e o f s e r v i c e provides a gauge of the value of the habitat services. provided by ecosystems is the sustenance of plants For example, part of the value of marine ecosystems and animals. In choosing a method for valuing this is conveyed by the value of the commercial f sh that type of service, it helps to distinguish plants and ani- they help sustain. Of course, this only represents a mals with direct use values f r o m t h o s e w i t h indirect portion of the value of the ecosystem—namely, the use values. Examples of the former are plants or ani- value of the ecosystem’s potential to sustain those mals that are consumed as food or that directly offer f sh that have a market value. recreational values (sightseeing, nature-watching, There is an important difference between the mar- etc.). Examples of the latter are plants and animals ginal and total value associated with market prices (such as organisms that are lower on the food chain) or the willingness to pay of consumers in markets. that help sustain other plants and animals that we Economists regard the prices that people are willing enjoy directly. To give specif c examples: ecosystems to pay as indicators of the marginal value—the generate direct use values by supporting the various value they place on the last unit purchased. Consider types of birds that we either enjoy non-consump- what a homeowner would be willing to pay for resi- tively as bird-watchers or consumptively as bird- dential water in a given month. He might be willing hunters. They generate indirect use values by to pay a huge sum for the privilege of consuming supporting the life of various plants or insects that the f rst ten cubic feet, because doing without them in turn enable birds to thrive. would deprive him of even the most fundamental 22 INTERPRETING AND ESTIMATING THE VALUE OF ECOSYSTEM SERVICES

(and valuable) uses of water for that month: drink- is an expression of the marginal willingness to pay, ing water, the occasional shower, etc. The next ten or marginal value. (In the example of Figure 2.1 , the cubic feet would probably not be worth quite as user would demand 400 cubic feet of water per much. They would allow him additional opportu- month at this price.) nities to f ll a glass from the faucet, and an extra The total value of the water consumed is much shower or two, but these would not be as critical to more than the price, however. The total value is the him (or to the people with whom he associates!) as area under the marginal willingness-to-pay sched- the f rst ten cubic feet. Thus the marginal value of ule (the sum of areas I and II in the diagram). Note water—the amount one is willing to pay for each that to ascertain total value (as opposed to marginal successive increment—falls steadily. value), researchers need to have information on the Figure 2.1 displays a typical willingness-to-pay entire marginal willingness-to-pay schedule (or schedule. The f rst cubic foot is shown to be worth a demand curve), not just the price paid. A main chal- great deal more than the f ftieth, which in turn is lenge of empirical valuation techniques is to trace worth much more than the hundredth. In reality, of out marginal willingness-to-pay schedules. course, households do not have to purchase each In the context of commercial products of ecosys- unit of water at its marginal value. If they did, they tems, this means that market prices represent only would be charged larger amounts for the f rst incre- the marginal value of these products. The value of ments than for later ones. Instead, utilities charge the total sales of these products corresponds to area households a given price per unit of water, regard- II in Figure 2.1 . Note that this is less than the total less of how much they consume. value to consumers, which is the sum of areas I and In Figure 2.1 , the horizontal line at $0.02 repre- II. Thus market sales understate the overall value of sents the price charged for the water. (We use this the commercially viable forms of life supported by number arbitrarily.) The standard economic ecosystems. assumption is that users will continue to purchase water until the marginal value of the water (or mar- Direct, non-consumptive use values. Within the ginal willingness to pay) is equal to the marginal category of direct use values from living things sacrif ce (or price). In these circumstances, the price maintained by ecosystems, we have another case to consider: the case where the life forms are used non- consumptively. For such uses, the relevant markets do not usually arise, and thus it is not possible to gauge values directly by observing market prices. Markets tend to arise for goods or services that are 0.15 excludable: the failure to pay for the good or service implies an inability to enjoy or consume the good. For non-consumptive use values (like bird- Marginal 0.10 Willingness to Pay watching), it is diff cult to establish a market because

Dollars people cannot easily be excluded from enjoying the good or service. In these cases, it is necessary to apply more inferential methods to ascertain the 0.05 I relevant values. 0.02 Revealed expenditure methods represent a broad II category of inferential approaches (NRC 2004). 0 Revealed expenditure methods have been applied 100 200 300 400 to ascertain some of the values provided by parks, Water used per month (cubic feet) lakes, and rivers—or, equivalently, the costs that

Figure 2.1 Relationship between water use and marginal willingness to results from the loss of these elements of nature. pay. From Daily ( 1997 ) reproduced by permission of Island Press, Here we describe one of the f rst and simplest Washington, DC. revealed expenditure methods: the travel cost MEASURING ECOSYSTEM VALUES 23 method (Freeman 1993). In recent years several services provided by these life forms. In fact, there more general and sophisticated approaches have is no need to include the values of these services in tended to supplant the travel cost method, but the an accounting of the overall value of an ecosystem! basic logic of the newer approaches is the same as These values are already captured in the values that of the travel cost method. attached to the life forms that humans enjoy. Non-consumptive uses are not directly bought or Consider the value of certain plants whose fruits are sold in markets; prices are not usually charged for eaten by birds and other “higher” life forms; assume their use. In those instances when use prices are humans obtain no direct use value from these charged (through entry fees, etc.), the prices are plants. If we abide by the anthropocentric approach unlikely to be good indicators of (marginal) value. to value, then there is no value to these plants over That is because the users of these resources actually and above the value that we attach to the higher life “pay” more than the entry fees to use them. For forms to which they contribute. To add their indirect example, the cost of a family visit to Yosemite use values to the direct use values would be double National Park is much greater than the daily use counting. The accounting here is perfectly analogous fee. The travel cost method recognizes that by add- to the economic valuation of net economic output, ing to the entry fee (if any) the transportation cost which disregards the value of intermediate inputs, and time cost expended to visit a particular site, one that is, inputs used up in the process of producing can ascertain the overall travel cost. This method f nal goods such as consumer goods and capital regards the overall travel cost as a measure of the goods. marginal willingness to pay by a visitor to the park; this is considered to be the same as the marginal 2.3.2.2 Valuing the regulating services of ecosystems value of the park to the visitor. The underlying Table 2.1 lists four examples of production inputs assumption is that people will continue to visit the from ecosystems: water f ltration, f ood control, pol- park until the value of the last unit (that is, the mar- lination, and climate stabilization. These services ginal value) is just equal to the travel cost. are inputs to the sustained production of agricul- It is also possible to employ survey methods, tural products in the sense that it would be diff cult such as the contingent valuation method, to deter- to maintain agricultural production without rela- mine how much value people place on the non- tively pure water, f ood control, pest control, or a consumptive uses. In contingent valuation stable climate. assessments of value, interviewees are asked what An appropriate measure of the value of produc- they would be willing to pay in order to prevent tive inputs is the additional economic income or some real or hypothetical amenity. For an exposi- prof t that they provide, holding everything else tion of this approach, see Mitchell and Carson constant. Thus, for example, the value of pest con- (1989). Many economists distrust results from sur- trol services provided by ecosystems is their contri- vey approaches, claiming that individuals’ asserted bution to prof ts. To assess these values, agronomists preferences in the hypothetical circumstances and other researchers develop models in which the posed by surveys bear no systematic relationship prof tability of various agricultural products is to their true preferences. Defenders of survey meth- assessed in the presence and absence of ecosystem- ods counter that, in many cases, surveys are the provided pest control, a key production input. The only method available. This “only game in town” difference is the value of the pest control services. argument may have force when existence values Similarly, one can gauge the value of f ood control are involved, as discussed below. services by comparing prof tability in the presence and absence of such services. A favorable climate Indirect use values. Ecosystems contain many can be considered a productive input. Numerous living organisms that support other, often “higher” studies have aimed to assess the damage from cli- forms of life that provide direct or indirect value to mate change to agriculture by comparing yields humans. It could be assumed that the value of and agricultural prof ts under current climate with ecosystem services should include the values of the those that would apply after predicted future 24 INTERPRETING AND ESTIMATING THE VALUE OF ECOSYSTEM SERVICES

climate change (e.g. Mendelsohn et al. 1994 ; ered inf nite. In fact, although the loss of f ood con- Schlenker et al. 2005). The damage from a changed trol services would cause a signif cant loss of prof t climate is equivalent to the monetized benef t or to farmers, the loss would not be inf nite. value from avoiding this change. Although avoided cost is not a measure of value, Pollination services are another example of an it is still important information. It indicates the net important production input. In the Central Valley of advantage of having access to the productive input northern California, various specialty crops, includ- provided by ecosystems, as opposed to having to ing melons, nuts, and tree fruits, depend upon the achieve the same input through an alternative. It pollination services supplied by wild bees whose provides a rationale for preserving the ecosystem population is maintained by breeding sites in service. For example, when the New York City nearby “natural” areas such as undeveloped forest- Water District struggled with how to preserve water land. The value of these pollination services is the quality in the Catskills, it determined that it was far additional prof t generated by the populations of less costly to do so by restoring the ecosystems sur- wild bees. rounding the city’s upstate reservoirs rather than by These pollination services have declined over constructing a new water treatment plant. The very time as larger and larger proportions of the region high avoided cost motivated the decision to pursue have been developed for agricultural purposes. One ecosystem-generated water quality control (Daily can only assume that at some point in the develop- and Ellison 2002). mental history of this unusually productive agriculture, wild insect populations alone were suff cient 2.3.2.3 Valuing ecosystem services offering non-use to guarantee some base level of pollination services values and thus guarantee yields adequate to keep the Other important services include the generation of farmers in business. Since that is clearly no longer spiritual, esthetic, and cultural satisfaction, the pro- the case, farmers now have to substitute a costly vision of recreational services, and the generation of alternative—pollination services from the bees sup- option value. Recreational services provide a non- plied by commercial beekeepers. This is now an consumptive direct use value. For example, a economically signif cant activity in these regions. In National Park offers opportunities for hiking, swim- this example, the avoided cost is the difference in ming, and bird-watching. Park visitors engaging in cost between the case where farmers enjoyed free these activities physically encounter the ecosystems pollination services from wild bees and the case involved (implying a use value), but (hopefully) do where they must pay for the services of bees hus- not use up the hiking trails, lakes, or birds in the banded by commercial beekeepers. process of enjoying them. The non-consumptive It is sometimes suggested that one can place a use values from these activities can be estimated value on production inputs by examining what using the methods described for such values under costs or expenditures agricultural producers man- Section 2.1 above. age to avoid by having these inputs and thus not Ecosystems also provide services with values having to substitute other inputs for them. For other than use values—values that do not derive example, where ecosystems provide effective pest from a physical encounter with the item of nature control, farmers can avoid having to pay for alter- in question. The values provided here are non-use native pest control methods such as the use of syn- values. There are two main types of non-use thetic pesticides. In fact, avoided cost is not a value. theoretically valid indicator of value. To see this, consider the following (extreme) situation. Suppose Existence value. This is the value that derives it were inf nitely costly for farmers to f nd an alter- from the sheer contemplation of the existence of native to wetlands in providing f ood control. If ecosystems. While much of our enjoyment of avoided cost indicated value, then the value of the biodiversity involves use value—that is, it derives wetlands’ f ood control services would be consid- from a physical encountering with various plants MEASURING ECOSYSTEM VALUES 25 and animals—we also derive satisfaction from It is much easier to def ne option value than to simply recognizing that these forms of nature exist. measure it. Its measurement requires a gauging of Thus existence value is an important element of the individuals’ risk-aversion, and this may depend on value people attach to nature or the functioning of the specif c context: persons are not equally averse ecosystems. It may ref ect the spiritual, esthetic, or to different types of risk. For an empirical assess- cultural satisfaction we obtain when we contemplate ment of option value, see Cameron (1992). the diversity, beauty, complexity, or power of nature. 2.3.3 Marginal vs. total value Survey approaches such as contingent valuation assessments may be the only way of ascertaining In discussions of ecosystems, one often might existence value, since actual market and non-market have in mind their total value. However, in many behavior gives little hint of its magnitude. As men- real-world circumstances, the policy debate con- tioned, survey approaches are controversial. Yet, cerns the change in value o r marginal loss of value when it comes to existence values, surveys may be that results from alteration or conversion of a part the only way of ascertaining values because peo- of the region that occupies an ecosystem. In bene- ple’s actions do not leave a “behavioral trail” from f t–cost analyses, when a portion of the ecosystem which their valuations can be inferred. In this lim- is threatened with conversion, it may be more ited space we cannot offer an appraisal of survey important to know the change or loss of ecosys- approaches. However, we can point out what seems tem value associated with such conversion than to to be the key underlying question: Is the informa- know the total value of the entire original ecosys- tion obtained from surveys, however imperfect, tem. Does a “minor” encroachment on the land better than no information at all? area of an ecosystem generate small losses in ecosystem value, or do small encroachments precipi- Option value. As developed in the economics tate large damages? literature (e.g., Bishop 1982), the term “option To examine this issue, we can begin with a very value” refers to a premium that people are willing large area of a (relatively) undisturbed ecosystem. to pay to preserve an environmental amenity, over The value we place on a given amount of area lost and above the mean value (or expected value) of the to other uses depends on the area of this system. Let use values anticipated from the amenity. Suppose, A represent the land area of our ecosystem, and for example, a habitat is threatened with destruction. suppose that the initial area is A 0. This ecosystem, And suppose that, if the habitat is preserved, there valued for its natural beauty and its biological is a 50% chance you would visit it, and a 50% chance diversity, is being decreased marginally in area you would not. If you were to visit it, you would through conversion to farmland. Suppose f rst derive a use value of 10; if you didn’t you would (counter to fact) that this decrease takes place with- enjoy no use value. In this case the expected value out changing the ecosystem’s character through species of the use value is 5. But you might be willing to loss. Since a larger area is worth more than a small pay, say, 7 to ensure the preservation of the habitat. one, the marginal value of each withdrawn unit If so, your option value is 2 (7–5). This premium rises gradually as the area ( A ) decreases. But in the ref ects individual risk-aversion: in the absence of limit, an area of size 0 is worthless, and tiny areas risk-aversion, people’s willingness to pay would are less attractive because they have a rather zoo- equal the mean use value (its expected value), and like character. Thus at small values of A , the mar- option value would be 0. ginal value begins to fall again. This relationship is We follow general practice in subsuming option shown in the path labeled “1” in Figure 2.2 . The value under the general category of non-use values. relationship between area and value expresses the However, the case can be made that option value is pure ecosystem-scale effect. so closely connected with (potential) use that it In fact we know that the biological diversity of should be placed in the use-value category. the ecosystem—one of the features contributing to 26 INTERPRETING AND ESTIMATING THE VALUE OF ECOSYSTEM SERVICES

practical matter, species values become proxies for ecosystem values: the Endangered Species Act in the United States is an embodiment of this principle in policy. Of course we regularly justify large expen- (1) ditures to save some species (e.g., the Black Rhinoceros) but not others (there is no Save the Furbish Lousewort Society).

Total Value Total (2) On what basis do we assign value to species? The (3) following are some axes along which different people make selections. Taxonomic proximity. We like animals that are like us. Primates attract human attention not only Area (A) A0 because there may be utility in the relationship Figure 2.2 Relationship between loss of habitat area and loss of (“animal models” for human disease) but because ecosystem value viewed through three different dimensions: ecosystem we respond to their quasi-human qualities. scale effect (1), diversity effect (2), and species effects (3). A o denotes the initial area of the ecosystem and the curves show how value declines as Rarity. All other things being equal, we have that initial area is reduced. In all cases there is some threshold area that is more interest in rare things than in common ones. too small to support any value. This is not simply a matter of vulnerability, although it is true that rare organisms are more vulnerable to its value to nature lovers—is not area-independent. extinction than abundant ones. Rarity itself can be The relationship, established mainly in studies on the attraction; in some sense animals and plants in islands and (to a more limited extent) on tropical nature are “collectibles,” if only in the sense of f nd- forests, is a non-linear one. The precise form varies, ing and listing them, and collections of the rare are but in a variety of studies the number of species lost more desired than collections of the commonplace. is slight until a quarter to a half of the area is lost, Indeed, “collection” in the form of listing is a motive and rises precipitously after about three-quarters of with powerful economic consequences. Many bird- the area is lost. The effect on marginal value is to watchers will undertake extreme expenditures to exaggerate the loss of ecosystem value as A is visit ecosystems harboring rare species for the pur- reduced. The impact of the loss in numbers of spe- pose of expanding their “life-lists.” cies as A is reduced may be termed the diversity Genetic uniqueness . If a species represents a unique effect. This effect is taken into account in the path evolutionary line—is, for example, the only extant labeled “2” in Figure 2.2 . As indicated by the differ- member of its genus or family—then it may be enti- ences between paths 1 and 2, this intensif es the tled to higher value than it would otherwise. marginal loss of value from a given reduction in A . Scientists especially would favor the use of this A third effect needs to be considered. The species criterion. in ecosystem A are not considered to be of equal Importance to ecosystem function . Certain species value to humans. People seem to care more about (often called “keystone” species) create conditions eagles and panthers than about mosses and bacte- that permit the maintenance of the entire ecosys- ria. We also know that species are related to one tem. The dominant trees in a forest, or birds that dig another in a complex, co-evolved web of dependen- nest-holes in trees that are used by other species, or cies: prey and predator, plant and pollinator. Trophic insects vital to the pollination of a dominant plant, relationships are also vitally important. Often, would be examples. higher order species on the food chain have the most exacting environmental requirements, and are How can these preferences be related to the mar- thus valuable indicators of the health of the entire ginal value calculation? Biological diversity is ecosystem; they or others may also be critical “key- reduced as A shrinks, but species do not fall out stone” species because they are located at the randomly; certain kinds tend to drop out relatively center of a network of interdependencies. Thus, as a early, others only when A becomes quite small. For SOME CASE STUDIES 27 conservation biologists and others, this means that vision of habitat for numerous species. The values wise policies cannot be made unless some value is of these and many other services are very diff cult attached to the different kinds. If, for example, the to quantify. Perhaps even more important to the ones we view as most valuable did well in relatively measurement challenge is the complexity of the net- small areas, we might argue for a patchwork of little work that links wetlands to groundwater and parks, whereas if the opposite were true we would thence to streams and lakes and other “navigable insist on large refuges. waters.” Obviously the number of possible criteria is large In theory, society could decide on the desired level enough to prohibit development of a precise rela- of protection of wetlands and their ecosystem serv- tionship among area, species loss, and value. ices by calculating the values of the numerous eco- However, larger organisms with broad ranges that system services, determining the extent of wetlands are especially area-sensitive would be likely to be that maximizes the net benef ts from these services rarer, on average closer taxonomically to humans, minus the opportunity cost to society, and then favored for “charm,” and important to ecosystem implementing laws that protect just this amount of function. Thus it is reasonable to assume a species- wetland. In fact, wetland protection has largely composition effect : t h a t a s A is reduced, the species ignored valuation. In the United States, the law does lost early in the reduction are more valuable than not invoke an explicit comparison of benef ts or those lost later. When this effect is taken into account, costs as a basis for the protection offered; this may the marginal loss from a reduction in species area is partly ref ect the measurement diff culties just men- even greater than indicated by path 2. Path 3 incor- tioned. Indeed, US law does not even acknowledge porates this effect (and the others). Indeed, our anal- cost as a consideration in determining the extent to ysis applies specif cally to the simple case in which which wetlands or their services are to be protected. A is reduced by shrinkage from the outside edges. In Rather, these broad ecosystem services are treated many situations, the reduction occurs by fragmenta- like a public “right,” something to be safeguarded tion—a patch here, a patch there, leading to a check- irrespective of the cost of protection. erboard of “natural” and “modif ed” areas. The new That right is protected by two public agencies habitats provided by “edge effects” can raise local under several laws. The Clean Water Act, adminis- biodiversity (at least transiently). In the longer run tered through the Environmental Protection Agency, the area/diversity rule will apply over the entire has several sections devoted to wetlands protection, region, but the value of species lost may differ. In and these refer to the general set of the “navigable recent studies of plant diversity in grassland patches, waters” referred to above. The Food Quality the f rst species lost are the most effective, narrow- Protection Act has a provision colloquially known niche competitors: fragmentation gives an advan- as the “swampbuster” clause that prohibits the tage to those species adept at dispersal and at rapid drainage or alteration of wetlands for farming pur- colonization ( Tilman 1997 ). poses. That provision is administered by the US Clearly some of these relationships are uncertain, Army Corps of Engineers. These two agencies are and the exercise could be applied to real natural required to issue permits when a landowner under- areas only after substantial research. However, it takes measures that would contribute f ll or pollut- points up the importance of thinking about value in ants to wetlands that lay within the drainage system marginal rather than aggregate terms, and suggests of the owner’s property. In an important wetlands a discipline that could be applied in the framing of case called Borden Ranch, the Supreme Court ruled general conservation policy. that a California farmer could not be issued a permit for a technique of plowing called “deep rip- 2.4 Some case studies ping,” on the grounds that it violated provisions of the Clean Water Act. In short, the Court found that 2.4.1 Wetlands in the United States the connection of the groundwater under the farm- Wetlands provide important ecosystem services, er’s plow to the navigable waters could not be dis- including f ood control, water purif cation, and pro- turbed or interrupted. 28 INTERPRETING AND ESTIMATING THE VALUE OF ECOSYSTEM SERVICES

In a much later case, the Supreme Court again This problem has been analyzed by Barbier and a split about a proposal to f ll some wetlands near team of f fteen authors (Barbier et al. 2008; Box 17.1 Lake St. Claire in Michigan. This particular wetland in this book) from various international institutions. was some distance from the lake, which clearly f t Their analysis examined the trade-offs between any ordinary def nition of a “navigable water.” Once conservation of coastal mangroves in Thailand and again, the issue rested on the question on whether a the conversion of mangrove lands to shrimp aquac- wetland that is distant from clearly navigable sur- ulture. The authors of the study started with the face waters is nevertheless entitled to protection. reasonable assumption that the buffering capacity The responsibility, again shared by the Environmental of the mangroves would depend on their area. On Protection Agency and the Army Corps of Engineers, measuring the ocean wave attenuation at various was complicated once again by the ambiguity of the distances inland from the shore, they showed that Clean Water Act’s language. A four-four split on the the storm buffering service provided by the man- Court was eventually decided by Justice Kennedy, groves was in fact nonlinear owing to the shape of who wrote in his opinion that the Clean Water Act the declining wave attenuation. intended to apply its provisions to the nation’s Using the estimated nonlinear relationship, waters generally, not restricted to surface waters. Barbier et al. (2008) found that the policy that pro- However, he also argued that the case should be vides the greatest overall benef t to the local popu- ultimately decided scientif cally by the federal agen- lation is one that prohibits shrimp aquaculture in cies responsible for applying the protection. much, but not all, of the area in question, and thus This case exemplif es the diff culty of interpreting reserves some area for shrimp farming ponds. Congressional intention. It also highlights the diff - Some economic benef ts to aquaculture investors culties of measuring ecosystem values in a network were retained. (In contrast, the assumption of a of rivers, lakes, wetlands, and groundwater that is linear relationship between area conserved and diffuse, interconnected, and complex. In the face of buffering capacity would have suggested that pro- such diff culties, policy-makers might prefer simply hibition of all shrimp farming was optimal.) This to establish a broad right to protection, rather than signif cant effort at ecosystem service valuation aim to compare values gained and values sacrif ced provided the basis for a solution favoring both under alternative levels of protection. stakeholders—both conservation groups and investors in aquaculture. 2.4.2 Vegetation and coastal protection 2.4.3 The Galapagos Islands About one-third of the world’s population lives in coastal areas or small islands, and they are at risk A third example, international in character, is pro- from buffeting and damage from storms and vided by the Galapagos Islands. This archipelago, extreme weather events, like the hurricanes that located 600 miles west of the Ecuador coast, consists regularly visit Caribbean and Gulf of Mexico coast- of thirteen large islands and a number of smaller lines and the recent tsunamis that swept across ones. All are of recent volcanic origin (100 000 to a Indonesia and coastlines in the south Pacif c and 1 000 000 years old), and they contain a unique Indian Oceans. In coastal areas, societies often must assemblage of plants and animals. They were vis- make diff cult choices between economic develop- ited by Charles Darwin during the voyage of the ment activities and risk-reducing conservation Beagle , and now are an important site for contempo- measures. Decision-makers practicing ecosystem- rary studies of evolutionary biology. based management are required to address both of Managed as a National Park by Ecuador since these competing needs in a way that balances the the 1950s, the islands have also become a favorite welfare of residents. In order to do this, they must destination for tourists, who explore the islands be able to measure the values associated with the from boats and debark on the islands to follow ecosystem services provided by conservation meas- carefully marked trails in the company of trained ures (see Box 2.1 ). naturalist-guides. With the growth in popularity of SOME CASE STUDIES 29

Box 2.1 Designing coastal protection based on the valuation of natural coastal ecosystems

R. K. Turner regulating, provisioning, and cultural services. Building on this platform, it can be argued that when the focus is on Depending on the precise de f nition used, coastal zones national accounting (Boyd and Banzhaf 2007), or landscape occupy around 20% of the earth’s surface but host more management (Wallace 2007), or in our example valuation of than 45% of the global population and 75% of the world’s service benef ts (Fisher et al. 2008), further elaboration is megacities (> 10 million inhabitants). The zone’s required for actual choice-making involving human welfare. underpinning coastal ecosystems—coral reefs, mangroves, A key step is the separation of ecosystem processes and salt marshes and other wetlands, sea grasses and seaweed functions into intermediate and f nal services, with the latter beds, beaches and sand dunes, estuaries and lagoons, forests yielding welfare benef ts (see Table 2.A.1 ). and grasslands—necessary to sustain human occupation, are The generation of services and the enjoyment of bene f ts highly diverse, productive, and biocomplex. These ecosystems is spatially conditioned and therefore a key step in any provide a range of services, such as, nutrient and sediment evaluation process must be the setting of the ecosystems in storage, water f ow regulation and quality control, and storm their appropriate contexts. The valuation process must also and erosion buffering, summarized in Table 2.A.1 be restricted to marginal gains/losses and should avoid (Crossland et al. 2 0 0 5) . C o a s t a l z o n e s a r e i m p a c t e d b y double counting and should note possible threshold effects dynamic environmental changes that occur both ways across and any nonlinearity between change in ecosystem services the land–ocean boundary. The natural and anthropogenic and habitat variables such as size of area. Failure to account drivers of change (including climate change) cause impacts for these limitations will lead to under-/overestimated ranging from erosion, siltation, eutrophication, and economic values and unnecessarily polarized cost–benef t over-f shing to expansion of the built environment, and decision choices (Turner 2007; Barbier et al. 2 0 0 8 ) . inundation due to sea level rise. All coastal zone natural Traditional sea defense and coastal protection strategies in capital assets have suffered signif cant losses over the past Europe have sought to provide rigid engineered “hold-the- three decades (e.g., 50% of marshes lost or degraded, 35% line” protection for people, property, and other assets against of mangroves, and 30% of reefs) (MA 2005). The the vagaries of dynamic coastal environments. Given the consequences for services and economic value of this loss at growing awareness of the consequences of climate change a the margin are considerable but have yet to be properly policy switch is taking place toward a “coping strategy” recognized and more precisely quantif ed and evaluated based on a mixed approach with engineered protection (Daily 1997 ; Turner et al. 2 0 0 3 ; B a r b i e r et al. 2 0 0 8 ) . focused on high commercial value areas, and the rest of the The ecosystem services (storm buffering) valuation coastline left to adapt to change more naturally. Measures example illustrated below is drawn from the European such as “managed realignment,” which involves the coastal zone context and in particular the east coast of deliberate breaching of engineered defenses to allow coastal England, which is one of UK’s most “at-risk” areas from migration and the creation of extended intertidal marshes climate change and other impacts. The European coastal and mudbanks, at the expense of agricultural land, are now zone is around 600 000 km2 (within 10 km of shorelines) being tested. Testing sites have been carefully chosen to and home to 80 million people and 280 major cities. The minimize impacts on existing people, property, and annual value of coastal tourism alone is 75 billion EUR (The environmental assets that enjoy engineered protection from Changing Faces of Europe’s Coastal Areas (2006): http:// the sea. The question is do they represent cost-effective reports.eea.europa.eu/eea_report_2006_6/en). In response strategies for society? to the multitude of pressures bearing down on coastal Managed realignment schemes yield benef ts in terms of areas, coastal protection and sea defense policy in the UK ecosystem services. They generate carbon storage benef ts and Europe is being reappraised and reoriented toward a (via saltmarsh creation) that can be valued in terms of the more f exible and adaptive strategy anchored to an damage costs avoided per tonne of CO . The sites also ecosystem services approach and decision support system 2 serve to improve f sheries’ productivity via nursery areas (Turner and Daily 2008; Turner et al. 2008). and this gain can be valued via market prices for The Millennium Ecosystem Assessment (MA 2005) states commercial species. There are also general recreation and that “ecosystem services are the benef ts people obtain amenity benef ts related to walking, bird-watching, and from ecosystems” and subdivides them into supporting, other recreational activities, as well as biodiversity continues 30 INTERPRETING AND ESTIMATING THE VALUE OF ECOSYSTEM SERVICES

Box 2.1 continued Table 2.A.1 Coastal zone services * and benef ts

Intermediate services Final services Benef ts Econ valuation methods

e.g: e.g: e.g: e.g: • Geodynamics: sediment and • Creation of beaches, • Flood/storm buffering • Market prices/damage cost avoided nutrient cycling and transport dunes, estuaries • Shoreline stabilization/erosion control • Production function market prices • Primary production • Sediment, nutrients, • Fish production • Survey-based contingent valuation/ • Water cycling contaminants • Biodiversity maintenance choice experimentation • Climate mitigation retention/storage • Carbon storage • Damage cost avoided • Biomass export • Amenity and recreation activities • Travel cost, hedonic pricing, survey • Maintenance of f sh • Cultural/heritage conservation based CV or CE nurseries and refuges • Survey based CV or CE • Regulation of water f ow and quality • Carbon sequestration • Recreation and amenity • Cultural heritage

*. European coastal areas including estuaries and saltmarshes.

maintenance and existence value benef ts. An indication of The costs and benef ts of managed realignment is given the composite value of some of these amenity and related as follows: benef ts can be estimated by transferring benef ts data T 1 ⎡⎤lCmr() mr+ C mr from the published literature if the spatial and other PV mr k, t m, t (2.A.2) t = ∑ t ⎢⎥mragr mr contextual variables are similar, or, more properly, by t0= (1+ r ) ⎣⎦⎢⎥−−()()aLt agr, t aB h e, t conducting site-specif c contingent valuation/choice mr experiment studies to estimate willingness-to-pay values. where PV t is the present value of managed realignment mr Finally, the maintenance costs of the existing engineered schemes (£million), l is the length of managed mr defenses will be saved as realignment schemes are realignment (km), Ck,t is the capital cost of realignment, mr mr implemented. On the costs side, secondary defenses may C m,t is the maintenance costs, at is the agricultural land agr mr be required further inland and there are opportunity costs lost, L agr,t is the forgone agricultural land value, ah is the associated with any agricultural land that is sacrif ced as area of intertidal habitat created, and Be,t is the ecosystem –1 the old defenses are breached. value benef ts (£ha ). Cost–benef t analysis of managed realignment schemes Finally, the overall CBA result is found as: took the following approach. NPVmr= () PV sq− PV mr (2.A.3) The “status quo” existing protection system is appraised ttt as follows: mr where NPVt is the net present value of managed T 1 realignment compared to hold-the-line for a given stretch PVsq [(I sq sq )], (2.A.1) tm= −∑ t C , t of coastline at time t (£million). t (1+r ) =0 The analysis was carried out with data from a number of sq w h e r e PV t is the present value of total costs of current different estuaries, and indicates that appropriately sited defenses (£million), r is the discount rate, l sq i s t h e l e n g t h schemes do represent gains in economic eff ciency if declining sq of defenses and Cm,t is the maintenance costs discount rates are applied over a 100-year time horizon (see (£km –1 y r–1 ) . Turner et al. 2 0 0 7 a n d L u i s e t t i et al. 2 0 0 8 f o r f u r t h e r d e t a i l s ) . CONCLUSIONS 31

“ecotourism,” the Galapagos now attract over Thus it is not surprising that a sometimes violent 150 000 visitors each year. controversy has arisen over the protection of the There is a resident population on several of the islands. When the government closed the sea larger islands, with a few service industries and a cucumber f shery in 1994 because the catch limit subsistence economy that depends on agriculture was being vastly exceeded, f shermen and some and f shing. These have been augmented by other other local residents seized the Darwin Station and direct uses that compete with the “natural” state of took scientists hostage. In a political controversy the larger islands, whereas recent reports suggest over a bill that would have given the islanders more that the less occupied islands are doing better than local autonomy (and relaxed many of the ecological they did 25 years ago. A signif cant f shery for sea protections) there was another takeover. The tense cucumbers, a delicacy prized in Asian and French historical contest between extraction and conserva- cuisine, developed in the 1990s and still exists, tion in the Galapagos is, at least with respect to this although the catch is declining. Illegal long-line particular indirect use value, the result of distribu- shark f shing continues to create a problem. Not tional effects. The economic potential of ecotourism only do these activities threaten the intertidal fauna, is almost certainly greater than that of the resource- they pose signif cant risks to the terrestrial ecosys- extraction uses. Yet the residents retain most of the tem through the introduction of “exotic” species rents from the second, and little from the f rst. and destructive camping on some islands. A second use value stems from the (uncertain) Fortunately, the Park’s protection system is much future benef ts that would emanate from the scien- more effective now than in the past, and efforts to tif c research underway on the Galapagos. The large eradicate goats, cats, and other invasive species are number of endemic species found there, and the continuing, most effectively on the four smaller, recentness of their evolutionary divergence from less-inhabited islands. mainland relatives, make the islands a living labo- Arrayed against these direct, consumptive use ratory for studies of species formation. Important values are two other values. The f rst is the direct, recent work (see Grant 1986) depends on the integ- non-consumptive use value from ecotourism, rity of the ecosystems of certain islands. Calculating which brings signif cant revenue. A sample calcula- its value, of course, would be extremely diff cult. tion of this value would be that the average visitor Finally, there are two important non-use values. (a week on a boat is a typical excursion) spends First, as in the case of the wetland example, people well more than $10 000. If the visitor is from the who have never been to the Galapagos and never USA, additional revenue will accrue to the expect to may experience a loss of existence value Ecuadorian economy through accommodations on that they would willingly pay to avoid. The unique the mainland, the f ight to the islands, and (if a quality of the islands and the considerable publicity national carrier is used) the f ight to Quito or they have received as a mecca for naturalists gives Guayaquil. A total per-visit value of $15 000 would this consideration a weight it might lack in less spe- be a reasonable f gure for the “overseas” visitor: if cial areas. In addition, in the presence of uncertainty, half were Ecuadorian nationals and half from else- people might be willing to pay a premium (over where, the value of the industry would approach and above the expected future use value) to ensure one billion annually. the preservation of the unique f ora and fauna of the Local residents, however, would make quite a islands. This is the option value. different calculation. The shops and restaurants at Puerto Ayora collect some money, and the support 2.5 Conclusions of the Darwin Station by tourists f ows into the local economy. Some boat operators are islanders, and Society must often make diff cult choices about how some services for all vessels are locally provided. and how much to protect natural capital and the However, the vast majority of the revenue f ows to ecosystem services such capital generates. Perhaps tour operators, many of them non-Ecuadorian, and the most important basis for supporting a policy to other off-island entities. that would protect otherwise threatened ecosystem 32 INTERPRETING AND ESTIMATING THE VALUE OF ECOSYSTEM SERVICES services is evidence that society gains more value rate of such conversion. According to economic from such protections than it gives up. This requires theory, the tax rate should be set equal to the mar- an assessment of the values that human beings ginal value of the ecosystem services provided by place on such services—values that often are not the natural capital or ecosystem in question. This expressed in markets. tax rate would lead to a lowered frequency of con- In this chapter we have aimed to clarify the philo- version that maximizes the net gain to society from sophical underpinnings of these values. The most intact wetlands—the benef t to agriculture minus prevalent and perhaps most workable philosophi- the lost ecosystem service value. In many cases, the cal basis is anthropocentric—value consists in the tax would prevent any conversion from taking ability to provide satisfaction or well-being to place. These are instances in which the marginal humans. Although anthropocentric, this approach value of the existing ecosystem services exceeds the is consistent with society’s making great sacrif ces marginal value generated by any conversion or in order to protect valued species and ecosystem alternative use of the natural capital. services. Although our discussion acknowledges a key We have also indicated the various types of value role for benef t–cost analysis in the valuation of eco- generated by ecosystem services, and laid out prin- system services, we would emphasize that such cipal empirical methods for measuring these val- analysis is usually not suff cient for deciding policy. ues. None of the empirical approaches is perfect; Benef t–cost analyses yield useful information on the uncertainties in measurement can be vast. aggregate net benef ts under alternative policy sce- However, even with the imperfections the methods narios, but usually ignore issues of fairness or dis- generally are good enough to provide a basis for tribution. These analyses need to be accompanied public policies to protect ecosystem services. The by an assessment of the distribution of the gains InVEST models described throughout this book and losses, both across the current generation and represent the frontier in valuing ecosystem services. between current and future generations. If a pro- These models exemplify the substantial progress of posed policy clearly would lead to serious inequi- the past decade in researchers’ abilities to depict the ties, it is reasonable to reject the policy, even if it gains and losses associated with the protection of a passes a benef t–cost test. wide range of ecosystems and their services. The topics of valuation and policy choice raise a Many of the benef ts from ecosystem services are number of imponderables. In arriving at the social not captured by unregulated markets. An individu- value of a given option, should every person’s will- al’s private gain from protecting ecosystem services ingness to pay count equally, or should some mem- falls short of the value of such protection to society. bers be given more weight than others? Are the Hence private markets tend to fail to provide suff - preferences of sophisticated ecologists worth more cient protection, and there is an important role for than those of city-dwellers who evidence neither public policy to protect these services. knowledge of nor interest in “nature”? How much Two types of public policy could stem from the weight should we give to the preferences or well- information offered by the InVEST models and being of future generations, as compared to that of other empirical studies. One is the introduction of current inhabitants of the planet? How can we quantitative restrictions that restrict the way natu- gauge the preferences of future generations in ral capital gets used or converted and thereby pro- attempts to ascertain the gains or losses they might tect the services such capital generates. Limits on experience under different policies? wetland conversion, for example, help protect the The fact that these questions have no easy answers various ecosystem services (f ood control, water need not make us pessimistic about the prospects purif cation, and habitat provision) that wetlands for sensible public policy. We can go a long way offer. Another approach is the introduction of prices toward improving policy-making by calling atten- for ecosystem services—prices that the market tion to the underlying philosophical questions, by would not generate on its own. For example, a tax developing empirical methods that generate better on wetland conversion would serve to reduce the information about the gains and losses at stake CONCLUSIONS 33 under alternative public policies, and by develop- Mendelsohn, R., Nordhaus, W., and Shaw, D. (1994) The ing channels for communicating this information to impact of global warming on agriculture: a Ricardian the general public. analysis. American Economic Review , 84 , 753–71. Millennium Ecosystem Assessment (MA) (2003). Concepts References of ecosystem value and valuation approaches. In Ecosystems and Human Well-being, A Framework for Assessment; A Report Barbier, E., Koch, E., Silliman, B., et al. (2008). Coastal eco- of the Conceptual Framework Working Group of the Millennium system-based management with nonlinear ecological Ecosystem Assessment . Island Press, Washington, DC. functions and values. Science , 319 , 321–3. Mitchell, R. C., and Carson, R. T. (1989). Using Surveys to Bishop, R. C. (1982). Option value: an exposition and Value Public Goods: The Contingent Valuation Method . extension. Land Economics , 58 , 1–15. Resources for the Future, Washington, DC. Boyd, J., and Banzhaf, S. (2007). What are ecosystem serv- National Research Council (NRC) (2004). Valuing Ecosystem ices? Ecological Economics , 63 , 616–26. Services: Toward Better Environmental Decision-Making . Cameron, T. (1992). Nonuser resource values. American National Academies Press, Washington, DC. Journal of Agricultural Economics , 74 , 1133–7. Schlenker, W., Hanemann, W. M., and Fisher, A. (2005). Costanza, R., Norton, B., and Bishop. R. C. (1995) The evo- Will US agriculture really benef t from global warming? lution of preferences: Why sovereign preferences may not lead American Economic Review , 95 , 395–406. to sustainable policies and what to do about it . SCASSS Singer, P. (1975). Animal Liberation . Random House, workshop on Economics. Ethics and the Environment, New York. Sweden. Skidmore, J. (2001). Duties to animals: The failure of Kant’s Crossland, C. J., Kremer, H. H., Lindeboom, H. J., et al., Moral Theory. Journal of Value Inquiry , 35 , 541–59. Eds. (2005). Coastal f uxes in the anthropocene , IGBP Series. Tilman, D. (1997). Biodiversity and ecosystem functioning. Springer, Berlin. In G. C. Daily, Ed., Nature’s Services: Societal Dependence Daily, G. C., Ed. (1997). Nature’s services: Societal depend- on Natural Ecosystems . Island Press, Washington, DC. ence on natural ecosystems. Island Press, Washington, Turner, R. K. (2007). Limits to CBA in UK and European DC. environmental policy: Retrospects and future prospects. Daily, G. C., and Ellison, K. (2002). The New Economy of Environmental and Resource Economics , 37 , 253–69. Nature. Island Press, Washington, DC. Turner, R. K., Burgess, D., Hadley, D., et al. (2007). A cost- Fisher, B., Turner, K., Zylstra, M., et al. (2008). Ecosystem benef t appraisal of coastal managed realignment policy. services and economic theory: integration for policy- Global Environmental Change , 17 , 397–407. relevant research, Ecological Applications , 18 , 2050–67. Turner, R. K., and Daily, G. C. (2008). The ecosystem serv- Freeman, A. M. (1993). The Measurement of Environmental ices framework and natural capital conservation. and Resources Values: Theory and Methods. Resources for Environmental and Resource Economics , 39 , 25–35. the Future, Washington, DC. Turner, R. K., Georgiou, S., and Fisher B. (2008). Valuing Grant, P. (1986). Ecology and Evolution of Darwin’s Finches . ecosystem services: The case of multifunctional wetlands . Princeton University Press, Princeton. Earthscan, London. Luisetti, T., Turner, R. K., and Bateman, I. J. (2008) An Wallace, K. J. (2007). Classif cation of ecosystem services: ecosystems’ services approach to assess managed realign- Problems and solutions. Biological Conservation , 139 , ment coastal policy in england, CSERGE Working Paper, 235–46. ECM 2008–04,CSERGE, University of East Anglia, Watson, R. A. (1983) A critique of anti-anthropocentric bio- Norwich. centrism. Environmental Ethics , 5 , 245–56.

CHAPTER 3 Assessing multiple ecosystem services: an integrated tool for the real world

Heather Tallis and Stephen Polasky

3.1 Today’s decision-making: the problem oping their land as real estate. They typically do not with incomplete balance sheets receive f nancial rewards for providing public goods from ecosystems, such as pollution f ltration or f ood Conservation and natural resource management mitigation. For example, in tropical coastal ecosys- have been dominated by approaches that focus on a tems, mangroves are routinely cleared for shrimp single sector and a single objective. These approaches aquaculture. People clearing the mangroves receive often fail to include a wider set of consequences of high market prices for the shrimp they produce but decision-making. For example, maximizing prof t they do not bear the full costs associated with the from industrial production often leads to negative loss of habitat for coastal f sheries, storm surge pro- impacts on air quality and human health. Maximizing tection (Das and Vincent 2009 ), pollution f ltration, agricultural production often leads to poor water or the loss of other ecosystem services provided by quality and in some cases losses of productivity in mangrove ecosystems. A more complete accounting downstream f sheries. Maximizing biodiversity con- shows that maintaining mangroves generally pro- servation can come at the cost of local jobs, food provides greater benef ts for society than shrimp aquac- duction, and other important benef ts. ulture provides ( Sathirathai and Barbier 2001 ; Many of these consequences are the result of Barbier 2007 ). A single-sector approach, which management decisions that overlook the broad ignores the multitude of connections among compo- suite of ecosystem conditions and processes that nents of natural and social systems, generally fails to sustain and enrich human life. Ecosystem services, provide as high a value to society as would manage- def ned as the contribution of ecosystem conditions ment that accounted for the full range of social and processes to human well-being, include the benef ts. production of goods (such as agricultural crops, seafood, timber, and natural pharmaceuticals), 3.2 The decision-making revolution processes that control variability and support life (climate regulation, f ood mitigation, pollination, In this chapter, we outline the major challenges in and the provision of soil fertility and clean water), taking an integrated approach to decision-making, enrich cultural life (recreational opportunities, and and present a new spatially explicit modeling tool satisfaction of aesthetic and spiritual needs), and that takes critical steps toward addressing these preserve options ( Ehrlich and Ehrlich 1981 ; Daily challenges. The inclusion of ecosystem services in 1997 ; MA 2005). decision-making provides a framework that ena- In most cases and for most services, there is little bles managers to broaden their perspectives by con- incentive for business managers and local landown- sidering the multiple, interlinked consequences of ers to account for the provision of ecosystem serv- their decisions. Our approach builds from the ices in their decision-making. Landowners receive Millennium Ecosystem Assessment (MA), which f nancial rewards for producing crops or for devel- contributed substantially to our understanding of

34 THE ECOLOGICAL PRODUCTION FUNCTION APPROACH 35 how to take such an approach at a global scale (MA 3.3 The ecological production 2005). Within a year of its completion, f ndings from function approach the MA were incorporated into the Convention on Biological Diversity, the RAMSAR Convention An ecological production function specif es the out- on Wetlands, and the Convention to Combat put of ecosystem services provided (“produced”) Desertif cation (Boerner 2007 ). Despite the overall by an ecosystem given its condition and processes. success of the MA at the global scale, we are still left These functions vary spatially due to site- and eco- with the grand challenge of bringing useful models system-specif c relationships. Once an ecological and information to bear at local, regional, and production function is specif ed, researchers can national scales where most decisions are made. quantify the impact of landscape change on the Although ecosystem service assessment has been level of ecosystem service outputs. In the twentieth attempted at sub-global scales ( Imperial 1999 ) there century, human alteration of ecosystems on a large are no systematic tools that can be applied in a gen- scale, such as the conversion of native ecosystems eral, consistent way across sites at the spatial scales to monoculture agriculture, led to an increase in and time frames relevant to major decisions affect- some provisioning services (e.g., food production) ing ecosystems. at the expense of many regulating, supporting, and One of the most challenging aspects of creating cultural services ( Vitousek et al. 1997 ; MA 2005). such tools is integrating robust ecological models Most applications of an ecological production and understanding in “ecological production func- function modeling approach have been done at tions” that def ne how the spatial extent, structure, small scales or for a small set of services. There are and functioning of ecosystems determine the pro- a growing number of such studies (e.g., Ellis and duction of ecosystem services. This challenge is par- Fisher 1987 ; Barbier and Strand 1998 ; Wilson and ticularly acute with ecosystem functions and Carpenter 1999 ; Barbier 2000 ; Kaiser and Roumasset services that operate across ecosystem boundaries 2002 ; Ricketts et al. 2004 ) and useful overviews and (such as nutrient transport from land to sea) and summaries have been compiled ( Pagiola et al. 2004 ; across spatial scales ( Engel et al. 2008 ). NRC 2005; Barbier 2007 ). One of the most challeng- A second major challenge of including ecosystem ing tenets of production functions is to integrate services in specif c decisions is generating estimates modeling across multiple services. The essential of the value of ecosystem services in economic and next step toward informing decision-making is a other terms. This task requires linking ecological systematic approach that combines the rigor of the models and understanding with social and eco- small-scale studies with the breadth of broad-scale nomic methods to reveal the values people hold for assessments. Recent work has taken strides in this different ecosystem services (NRC 2005). This task vein (e.g., Boody et al. 2005 ; Jackson et al. 2005 ; Antle is especially hard for the many ecosystem services and Stoorvogel 2006 ; Naidoo and Ricketts 2006 ; that generate global public benef ts, such as climate Nelson et al. 2008 , 2009 ). regulation or existence value of species, for which There are some cases where an understanding of there are no market prices or other readily available ecological production functions alone is suff cient. signals of value. Over the past 40 years or so, econ- Many government agencies make decisions about omists have developed a number of methods and what activities will be allowed based on whether or tools for “non-market valuation” that can be not they meet an environmental standard. For applied to estimate the value of ecosystem services example, an agency may assess how activities will (Freeman 2003 , NRC 2005). Whether for market or likely affect the ability of an ecosystem to meet non-market values, appropriately linking social water quality standards. Their decision is not based and economic valuation with ecological produc- on how much it would cost to treat that water for tion functions is necessary to ensure that values consumption or the value of access to clean drink- ref ect underlying ecological conditions. We ing water, but rather on the expected change in con- describe these challenges further below and offer taminant levels or the likelihood of crossing a plausible solutions. contamination threshold. In these cases, simply 36 ASSESSING MULTIPLE ECOSYSTEM SERVICES knowing how ecosystem services will change in national attention especially following Costanza biophysical terms is informative and useful. et al. (1997 ), which estimated the monetary value Other decisions are tied to f nancial costs and of ecosystem services for the entire planet. The benef ts, and many decision-makers are conditioned most common application of the benef t transfer to analyzing policy alternatives in terms of the net approach uses estimates of the value of services benef ts measured in monetary terms. One concern per unit area from a single or small number of about keeping measurement of ecosystem services locations and applies this value to other locations in biophysical units is that services not measured in with the same ecosystem type (e.g., Costanza et al. monetary terms may not be given full weight in 1 9 9 7 ; K o n a r s k a et al. 2002 ; Troy and Wilson 2006; decision-making or may be ignored altogether T u r n e r et al. 2007 ). This approach relies on existing ( Daily 1997 ). In these cases it can be very useful to estimates and does not require any additional combine ecological production functions with eco- analysis, which is a distinct advantage if decisions nomic valuation methods to estimate and report the are imminent and primary data collection is not monetary value of ecosystem services. feasible ( Wilson and Hoehn 2006 ). Some ecological production function approaches The assumption of constant per hectare values that have been combined with appropriate market prices makes the benef t transfer approach so tractable, and non-market valuation methods to estimate eco- however, has signif cant disadvantages that limit its nomic value, and illustrate the change in the mone- social, economic, and ecological realism. In some tary value of services with changing environmental cases, the benef ts transfer approach provides an esti- conditions (e.g., Swallow 1994 ; Naidoo and Ricketts mate of total economic value rather than estimates of 2006 ; Barbier 2007 ). Ecological production functions value for individual services (e.g., Konarska et al . can be used to determine how the provision of vari- 2002 ). When limited to estimates of total economic ous market goods and services change as ecosystem value, we cannot analyze how the provision and conditions or processes change. The value of the value of each individual service will change under changes in output of marketed goods and services new conditions. If a wetland is converted to agricul- can be evaluated using market prices for marginal tural land, how does this subsequently affect the pro- changes or by using changes in consumer and provision of clean drinking water, f oods downstream, ducer surplus for non-marginal changes ( Just et al. climate regulation, or soil fertility? Without service- 2004 ; Barbier 2007 ). Non-market valuation meth- specif c information, it is impossible to design effec- ods, including revealed preference and stated pref- tive policies or payment programs that ensure the erence methods, can be used for ecosystem services continued provision of ecosystem services. that are not traded in markets (Freeman 2003 ; NRC Further, assuming that every hectare of a given 2005). At present, however, we lack comprehensive habitat type is of equal value ignores well-demon- studies that tie together economic valuation meth- strated differences between sites in terms of scar- ods with ecological production functions to esti- city, spatial conf guration, size, quality of habitat, mate the monetary value of ecosystem services both number of nearby people, or their social practices for a broad range of ecosystem services and at a and preferences, all of which may be crucial in broad geographic scale (NRC 2005, but see Naidoo determining the value of ecosystem services. These and Ricketts 2006 ). simplif cations mean that area-based benef t esti- Detailed non-market valuation studies appro- mates that rely on transferring values from a study priate to a particular service in a particular place site (where original valuation took place) to a pol- tend to be time- and-resource intensive, limiting icy site cannot consider important spatial aspects the applicability to broad-scale assessments. of land use, habitat distribution and geometry, or Benef t transfer provides a less time-intensive economic insights on the importance of proximity approach than production functions for generat- and value. For these reasons, we do not believe ing broad-scale monetary estimates of ecosystem that application of benef ts transfer based on value services. Benef t transfer studies of the value of per hectare by habitat type is a good direction to ecosystem services have garnered signif cant inter- pursue. INVEST: MAPPING AND VALUING ECOSYSTEM SERVICES 37

A more promising approach to benef t transfer is • Is spatially explicit; to use a value function approach, which estimates • Provides output in both biophysical and mone- value as a function of ecological and socio-economic tary terms; conditions, rather than a unit value approach that • Is scenario driven; depends only on habitat type. Value functions can • Clearly reveals relationships among multiple be estimated from data at a single site or from a services; and meta-analysis of a number of sites (Rosenberger and • Has a modular, tiered approach to deal f exibly Phipps 2007 ). Although the f eld of benef t trans- with data availability and the state of system fer is moving toward value function approaches knowledge. that include socio-demographic characteristics, Several of these features are discussed in greater environmental attitudes, and biophysical contexts detail below. at study and policy sites ( Wilson and Hoehn 2006 ; Rosenberger and Stanley 2006 ; Eshet et al. 2007 ), examples of sound applications are still hard to f nd 3.4.1 A multiple ecosystem service approach ( Plummer 2009 ). The creation of general tools capable of integrat- Managers are often forced to make trade-offs among ing ecological production functions with valuation sectors and goals. A fundamental socio-economic for multiple services would allow a major advance truth is that a manager cannot simultaneously max- toward integrated decision-making in diverse con- imize returns for all sectors of society at once. As the texts across scales. At present, the lack of such tools old saying goes, we cannot have our cake and eat it makes ecosystem approaches to most resource deci- too. Despite the ubiquity of trade-offs, managers sions expensive, time-consuming, unwieldy, and frequently lack a set of tools to inform them about diff cult to replicate. In the next section, we describe the trade-offs they face. Often, mental assessments the development of a new general tool aimed at f ll- of the existence and magnitude of trade-offs are ing this gap. wrong and lead to decisions that result in poor outcomes. Although management actions that strike a balance among goals may be plausible, these actions 3.4 InVEST: mapping and valuing are often hard to identify in a highly charged politi- ecosystem services with ecological cal environment. Arguments commonly based on production functions and economic qualitative assumptions are diff cult to balance in a valuation systematic and clearly understood way. A formal modeling framework that can reveal the likely rela- The Natural Capital Project (http://www. tionship among services can help dispel incorrect naturalcapitalproject.org ) has developed a new tool assumptions and identify management options that designed to facilitate integrated decision-making, provide a high level of a range of ecosystem serv- bringing together credible, useful models based on ices. There is growing evidence that decision-mak- ecological production functions and economic valu- ers are ready for this kind of approach, if only they ation methods. The intention is to incorporate bio- had tools to help them move forward (see examples physical and economic information about ecosystem in Section 3.4.4 ). services into conservation and natural resource Trade-offs among sectors, or among ecosystem decisions at an appropriate scale. The tool is called services exist because services are not perfectly cor- InVEST, for Integrated Valuation of Ecosystem related. Using data from the Willamette Basin in Services and Tradeoffs. We have built in several key Oregon, Nelson et al. ( 2008 ) found that targeting features that make this a f exible yet scientif cally policies to provide carbon sequestration, by limit- grounded tool. InVEST is a set of computer-based ing enrollment to landowners who would grow for- models that: ests on their land, was effective at increasing carbon • F o c u s e s o n e c o s y s t e m s e r v i c e s t h e m s e l v e s , r a t h e r storage but not effective for species conservation. than on the underlying biophysical processes alone; Alternatively, targeting policies to meet species 38 ASSESSING MULTIPLE ECOSYSTEM SERVICES

conservation objectives, by limiting enrollment to production, provision of pollination for agricultural landowners who would restore rare habitat types crops, recreation and tourism, and provision of cul- (e.g., oak savannah and prairie), was effective at tural values and non-use values ( Table 3.1 ). We also increasing species conservation but not effective for provide models for terrestrial biodiversity, as an carbon sequestration. More generally, the MA (2005) attribute of natural systems that underpins the found pervasive trade-offs between provisioning delivery of ecosystem services ( Chapter 13 ). services (e.g., food and timber production) and Additional ecosystem services besides those other types of services (regulating, supporting, and listed above should be considered in many natural cultural services). However, some trade-offs are resource decision-making processes (see Box 3.1 for more a consequence of past land-use decisions than example). We have focused initially on the above- of the underlying potential of the socio-ecological mentioned subset because of their global impor- system. Polasky et al. ( 2008 ) demonstrated that tance, relevance to major decisions being made higher levels of both biodiversity conservation and currently, and proximity of many of these services value of marketed commodities could be achieved to markets. As the modeling effort progresses our by rearranging the spatial pattern of activities on a aim is to include other services that likely provide landscape. A modeling framework that allows value to society. assessments of biodiversity and multiple ecosystem services can identify policies or geographies that 3.4.2 Ecological processes vs. ecosystem can lead to win–win outcomes, where all objectives services: a critical distinction can be increased relative to the status quo, and to those situations where outcomes necessarily lead to Ecological processes are essential for the provision trade-offs. of ecosystem services but processes are not synony- We address the need to reveal and quantify syn- mous with services. Until there is some person ergies and trade-offs by providing models for a somewhere benef ting from an ecological process, it suite of ecosystem services in InVEST. The services is only a process and not an ecosystem service ( Luck we currently model are provision of hydropower et al. 2009 ). This distinction is critical yet often over- and irrigation water, mitigation of storm peak f ows, looked. Extensive research has been applied to the avoidance of reservoir sedimentation, regulation of modeling and measurement of ecological processes, water quality, climate regulation through carbon and it is tempting to simply apply those models to storage and sequestration, timber production, pro- ecosystem service-related decisions. However, eco- duction of non-timber forest products, agriculture logical processes tell us only about the ecological

Table 3.1 Classif cation of ecosystem services modeled in InVEST

Service MA classif cation Valuation technique Chapter

Provision of water for hydropower Supporting Market valuation 4 Provision of water for irrigation Supporting Market valuation 4 Storm peak mitigation Regulating Avoided damages 5 Water purif cation: nutrient retention Regulating Avoided damages 6 Avoided reservoir sedimentation Regulating Avoided damages 6 Carbon sequestration Regulating Social value, Market valuation 7 Timber production Provisioning Market valuation 8 Non-timber forest product production (NTFP) Provisioning Market valuation 8 Agricultural production Provisioning Market valuation 9 Pollination of agricultural crops Supporting Market valuation 10 Recreation/tourism Cultural/aesthetic Market and non-market valuation 11 Cultural/aesthetic Cultural/aesthetic None 12

Valuation technique identif es the general approach used to derive value once InVEST has generated the amount of ecosystem service. INVEST: MAPPING AND VALUING ECOSYSTEM SERVICES 39

Box 3.1 Unsung ecosystem service heroes: seed dispersal and pest control

Liba Pejchar Eurasian Jay, which bury 4500–11 000 acorns per year at depths ideal for germination and in areas where light At dusk in Mexico, a bat f ashes low over a coffee conditions are perfect for growth. The fraction of acorns plantation, swiping a moth off a leaf and disappearing into that are buried and forgotten have a far greater chance of the night. In a large city park halfway around the world, a germinating than passively dispersed acorns that typically jay screeches shrilly as it snaps up an acorn and swoops suffer 100% predation by mammals. In the absence of the away, burrowing it neatly for the winter. In Texas, 100 jays, the replacement cost of this service (using human million bats pour out from caves and from under bridges, labor to plant oak seedlings) for the park alone would be feeding in a frenzy over 10 000 acres of cotton plantations. 1.5–6.7 million SEK per pair of jays (approximately And in a Hawaiian forest, a thrush f utters down, nabs a US$200 000–950 000 per pair ( Hougner et al. 2006 )). red berry, and swoops to a perch, dropping the seed in In this and other rare cases, seed dispersal and pest alarm when the shadow of a hawk passes overhead. What control services have been quantif ed in monetary terms by do all of these actions have in common? They all involve comparing costs and benef ts to the human equivalent (i.e., the feeding habits of birds and bats, our winged cousins. hand-planting seeds or applying pesticides). For example, But they also illustrate two frequently unrecognized breeding colonies of Brazilian free-tailed bats ( Tadarida ecosystem services: seed dispersal and pest control, services brasiliensis ) feed on extraordinary numbers of insects in that are provided for free every day, all over the world. south-central Texas ( Figure 3.A.1 ). Lactating females Understanding how bird and bat species provide key consume the insect equivalent of two-thirds of their body functions in ecosystems opens up new opportunities to mass every hot summer night. Taking into account the value the provision of little-known ecosystem services. The private and social costs of pesticides that would be acorn-toting Eurasian Jay ( Garrulus glandarius ) is a neat required in the absence of these insectivores, cotton example of such an opportunity ( Figure 3.A.1 ). A city park farmers benef t from bats to the tune of approximately near Stockholm supports one of the largest populations of US$750 000 per year on a harvest worth US$4–6 million giant oaks ( Quercus spp.) in Europe—a keystone species ( Cleveland et al. 2006 ). Building on ecological that harbors high biodiversity and plays an important role understanding to demonstrate the economic value of wild in the cultural landscape of Sweden. Natural regeneration nature is crucial for incorporating conservation of these and long-distance dispersal is largely dependent on the organisms and processes into decision-making .

(a) (b)

Figure 3.A.1 (a) Eurasion Jays ( Garrulus glandarius ) and (b) Brazilian free-tailed bats ( Tadarida brasiliensis ) provide valuable seed dispersal and pest control services.

production function, which is only the supply o f of ecosystem services. We def ne the term “use” ecosystem services. It is critical to include the quite broadly. “Use” includes not only the consump- demand for services as well. Where are the people tion of physical goods (e.g., agricultural crops, f sh), who enjoy services, and how much do they use? The but also the recreational, cultural, spiritual, and aes- combination of supply and demand generates “use” thetic appreciation of nature (non-consumptive use 40 ASSESSING MULTIPLE ECOSYSTEM SERVICES value), as well as option, existence, and bequest val- habitat in an agricultural landscape may house ues that do not generate a benef t from any type of bee populations, but if there are no agricultural consumptive or non-consumptive use (“non-use” f elds within foraging distance with a crop in need values). To model use of ecosystem services, we of pollination, then that native habitat patch does need to integrate analysis of the supply of ecosystem not provide pollination benef ts for crops at that services (ecological processes) with analysis of the time. So, in this case, a model of native pollinator location, type, and intensity of demand for services. meta-population dynamics could give us a very This is true for all categories of ecosystem service clear sense of how much pollination service could be ( Table 3.1 ). supplied by patches of native habitat in an agricul- Water-related services provide good examples for tural landscape. Until that information is paired how to think about the distinction between ecologi- with information on the identity of crops grown, cal processes (supply) and ecosystem services (sup- their distribution in the area, and the crop-specif c ply and demand). Consider water purif cation for yield benef ts of pollination, we cannot estimate the drinking, a regulating service. Many useful eco- amount of pollination service actually being pro- logical production function models exist that can vided at a given time. help us predict the concentration of contaminants InVEST deals with this challenge by using a three- in waterways (e.g., Soil and Water Assessment step modeling process. First, the ecological produc- Tool (SWAT), Gassman et al. 2007 ; Annualized tion function, or the supply side of ecosystem Agricultural Non-Point Source (AnnAGNPS), Yuan services, is modeled ( Table 3.2 ). These models et al. 2006 ). However, the provision of clean drink- require biological, physical, geological, and other ing water is not a service unless there is someone kinds of inputs, and draw heavily from existing there who wishes to drink it. This does not mean knowledge. For example, our water-related service that a natural system providing clean water in a models start out with similar fundamental hydro- remote area with no people does not provide any logic processes as those included in models such as services. Clean water in remote areas can maintain SWAT ( Gassman et al. 2007 ). Our model for avoided biodiversity or provide ecological functions that reservoir sedimentation draws heavily from the underpin other ecosystem services, and may also be Universal Soil Loss Equation (USLE) (Brooks et al. a value as clean water for drinking in the future . 1982 ). Our biodiversity model is based on species- However, if no one currently makes use of the water area relationships in extensive native habitats and for drinking, then there is no clean drinking water in countrysides (Connor and McCoy 1979 ; Pereira ecosystem service in that particular place at the cur- and Daily 2006 ). The outputs from this step of mod- rent time. Ecological processes (the production eling are in biophysical units and represent the level function) must be connected to benef ciaries to gain of each ecological process supported by each part of an accurate picture of the level of use of the service the landscape ( Table 3.2 ). For example, our model of provided. irrigation f rst predicts the total amount of surface Consider another example involving the pollina- runoff from each parcel on the landscape. This rep- tion of agricultural crops, a supporting service. resents the supply of water for all potential con- Many agricultural crops require insect pollination sumptive uses (Table 3.3 ). Outputs like this can be (e.g. almonds, strawberries), but many other useful for model calibration or for understanding crops do not (e.g., rice, corn). A patch of native the maximum potential level of service.

Table 3.2 Three-step structure of InVEST ecosystem service models

Modeling step Model inputs Model outputs Units

Ecological process Geological, morphological, biological, etc. Supply Biophysical Use Socio-economic, management characteristics, etc. Level of use—intermediate service Biophysical Socio-economic, management characteristics, etc. Level of use—f nal service Final product Valuation Financial Value Monetary INVEST: MAPPING AND VALUING ECOSYSTEM SERVICES 41

Table 3.3 Examples of ecosystem service outputs and units for one f nal service (timber production) and two intermediate services (crop pollination and provision of irrigation water)

Output Service

Timber production Crop pollination Provision of irrigation water

Supply Standing stock of wood (ft3 ha–1 ) Insect abundance (# insects ha–1 ) Surface runoff (vol ha–1 ) Use—intermediate service None Insect abundance contributing to crop Runoff available and used for irrigation pollination (# of insects ha–1 ) (vol ha–1 ) Use—f nal service Harvested wood (ft 3 ha –1 ) Crop yield due to insects (kg crop ha–1 ) Crop yield due to runoff used for irrigation (kg ha–1 ) Value NPV of harvested timber ($ha–1 ) NPV of additional crop yield ($ha–1 ) NPV of additional crop yield ($ha–1 )

NPV = net present value.

The second step of modeling determines the use of the intermediate service. Irrigation water is an of ecosystem services. This step incorporates socio- input to agricultural production (f nal service), so economic, management, and other kinds of data we can also use our agricultural production model on demand for ecosystem services with informa- to estimate how much additional crop yield can be tion on supply. Use of an ecosystem service is the expected given that additional amount of water level of supply in an area actually demanded by available from irrigation. This output (additional people for the service of interest (Table 3.2 ). For yield from irrigation, in kg ha–1 ) is the level of use of example, in our irrigation model, we f rst consider the f nal service (the f nal product) ( Table 3.3 ). how much water supply is available, which is In addition to mapping and quantifying the sup- determined by surface runoff generated upstream ply and use of ecosystem services, InVEST also has of agricultural f elds with subtractions for with- the capacity to estimate their value. As discussed in drawals upstream for other consumptive uses such Section 3.3 , we use market and non-market valua- as for drinking or industry. We then consider how tion approaches to arrive at monetary values for much water demanded by crops in the region is ecosystem service provision (Table 3.2 ). Our focus not met by rainfall. There may be a large amount in estimating value is on the social value of each of water available for irrigation (after other con- service, which captures the total value of the serv- sumptive uses), but if there are no crops left with a ice to society as opposed to the value it offers to the water def cit after factoring in rainfall, there is no owner of the service-providing land. While the demand for irrigation. In such a case, although social value of services for which markets exist there is a supply of irrigation water, because there (e.g., provisioning services) can be estimated using is no demand there is no resulting use and there- market prices, estimating the social value of serv- fore no ecosystem service provided. This example ices for which there are not markets requires alter- illustrates why it is so important to combine sup- native methods. Numerous techniques have been ply and demand in the use step in modeling. It is developed and ref ned in recent decades to esti- only by combining supply and demand to deter- mate the value of non-market goods (e.g., hedonic mine use that we quantify the level of outputs of price models, travel cost models (random utility ecosystem services. models), choice experiments including both contin- If the service of interest is an intermediate service, gent valuation and conjoint analysis). To reduce the there are two possible model outputs: one for the data collection and analytical burden on InVEST use of the intermediate service and one for the use users, the default valuation methodologies we use of the f nal service ( Table 3.2 ). In our irrigation in InVEST are typically associated with market example, we can report the amount of water used prices for commodities traded in markets, or the for crop irrigation, which would have the units of damages avoided by the maintenance of service volume per hectare (vol ha–1 ). This is the level of use provision. In each case, the valuation methodologies 42 ASSESSING MULTIPLE ECOSYSTEM SERVICES described in the following chapters for services is not an appropriate method for valuing an ecosys- represent only one of the many viable valuation tem service. options. Users with greater sophistication and The details of how supply, demand, and eco- access to valuation studies may wish to utilize nomic valuation are combined in models for these other valuation approaches. and other services can be found in later chapters of Our irrigation example demonstrates the InVEST this volume (see also Table 3.3 for other examples). valuation approach for many provisioning services. The monetary value of increased water for irriga- 3.4.3 Spatially explicit ecosystem service tion requires identifying the increase in the revenue modeling from agricultural production that would result from an increase in available water input ( Table 3.3 ). This Because the value of ecosystem services is determined additional value can arise because more land can be by both the location of ecological processes that pro- irrigated or because existing cropland can receive vide services (supply) and the location of people who more water, which could increase yields of crops or demand and use the services, any ecosystem service allow more water-intensive but higher value crops modeling effort should be spatially explicit. We con- to be grown. sider two key elements of space in the application of Consider another example of avoided reservoir InVEST: the role of spatial pattern and heterogeneity sedimentation that demonstrates the methodology in the landscape in determining the provision of serv- of damages avoided for valuing non-market services, and the scale across which different services act. ices. The role that vegetation and management Often, decision-makers want to know where to invest practices play in keeping sediment out of water- or how to target programs to get the greatest return ways can provide services to society including from their investment. For instance, where should avoided infrastructure maintenance costs (as reser- protected areas be located to gain the largest biodi- voirs silt in and require dredging) and avoided versity and climate regulation co-benef ts? Should a f ood risk (as rivers or reservoirs silt in and lose new agricultural subsidy program to control water their capacity to control or buffer f oods). To esti- quality be targeted at riparian areas in headwaters or mate the value of avoided siltation in reservoirs, we further downstream? Will a tree planting program in calculate supply by modeling how much erosion a poor district help with f ood control? All of these control is provided by a landscape based on questions have a spatial element, but many existing enhanced USLE equations. We then calculate use by biophysical process models are non-spatial and do adding demand, represented by the location of res- not allow analysis of the landscape locations best for ervoirs and their characteristics (dead volume, investment. All of the models in InVEST focus on remaining lifetime of the dam, etc.). Finally, we identifying how much each parcel on the landscape derive the value of this ecosystem service through contributes to each service. avoided dredging cost calculations. Secondly, we must consider the scale across which Strictly speaking, damages avoided (or replace- services are provided. Some services, such as polli- ment costs) are estimates of costs not estimates of nation and some water-related services, are pro- benef ts and so need to be used with caution. vided at a very local scale, while other services, Avoided damages (replacement cost) can be used as such as climate regulation, are provided at a global a measure of value of ecosystem services only when scale. Trees f xing carbon in the Amazon forest are there are at least two ways of providing an equiva- providing a global benef t. Each model in InVEST lent quantity and quality of an ecosystem service looks across the appropriate scale for the service of (one supplied by ecosystems and an alternative interest. For example, the pollination model uses supply via a human-engineering approach), and the foraging distance of pollinators to delineate the where the benef ts of the service exceed the costs landscape for assessment, while the carbon seques- providing the service via the human-engineering tration model assumes that tree growth on any par- approach (Shabman and Batie 1978 ; NRC 2005). If cel provides a benef t no matter where it is located these conditions are violated then damages avoided since the global atmosphere is well-mixed. INVEST: MAPPING AND VALUING ECOSYSTEM SERVICES 43

Considerations of scale raise two important incentives to people who control the provision of issues, one related to modeling and one related to services so that they can recognize the benef ts that policy. It may be more diff cult to apply models for their actions provide to others. Such policies can be local services since input data on land use and cover explored through the development of scenarios. patterns need to be at a high enough resolution to capture important features of the service. One 3.4.4 Scenario-driven modeling: making would not learn much from the crop pollination a decision-relevant tool model if native pollinators at the site of interest for- aged 1.5 km, but land use and cover data were only To be effective in a decision-making or policy arena, available at 10 km2 resolution. When modeling mul- analyses should be relevant to the needs and ques- tiple services, the scale of data resolution should tions of managers and decision-makers. To apply correspond to the f nest scale of resolution for the InVEST in such situations, we envision embedding ecosystem services of interest. the modeling within a stakeholder engagement In terms of policy, the scale and location of the process that allows managers to identify the choices provision of ecosystem services and the scale and of interest to them (Figure 3.1 ). InVEST is designed location of benef ciaries of the services are often dis- to work with many different kinds of scenarios connected. This is the case for trees f xing carbon in derived through many types of stakeholder engage- the Amazon rainforest that are enjoyed (and “used”) ment processes. by people all around the globe. For other services, InVEST can take input from stakeholders to con- such as pollination or provision of clean water, ben- sider a wide range of land use and resource man- ef ts are fairly local. However, even where provision agement alternatives. Each ecosystem service model and benef ts of services are local, supply and uses land use and land cover (LULC) patterns as demand may be disconnected in space. For exam- inputs to predict biodiversity and the production of ple, upstream landowners may divert water or ecosystem services across a landscape. This means increase nutrient loading that harms downstream that we can consider choices that affect the type of water users. Such spatial disconnects between pro- land cover (urban, wetland, closed-canopy decidu- vision and benef ts have important implications for ous forest, etc.) and choices that keep land cover policy. Explicit policies may be needed to give the same but alter management practices on any

Staging

Scenarios (D Management, Climate, Population)

Models

Biodiversity Provisioning Regulating Cultural Supporting Food Species Climate Recreation Pollination Timber Habitats Stability Tradition Fresh water Flood Community Control

Stakeholder Engagement Outputs ~ Biophysical, Economic, Cultural

Tradeoff Balance Maps Curves Sheets

Figure 3.1 Conceptual model for applying InVEST as part of a stakeholder process. Stakeholders give input to the process at every step. They produce future options that are turned into scenarios, they identify the ecosystem services of interest and help determine the level of model complexity needed for the questions of interest, they provide input data for the models, and they request particular types of outputs and then assess those outputs. If results spur further questions or ideas for alternative scenarios, the entire process can be repeated. 44 ASSESSING MULTIPLE ECOSYSTEM SERVICES particular part of the landscape (change in release can be built to include these drivers of change in addi- pattern from an existing dam, change in crop type tion to management practices. Many efforts have now planted in existing agricultural areas, change in fer- down-scaled global climate models and used regional tilizer type or amount used, change in rotation time predictions to drive vegetation patterns, giving us in existing plantation forests, etc.). Most natural maps of likely future land cover and climate. Similarly, resource management decisions will have effects on a variety of research groups are in the process of turn- land use and cover patterns, either directly or indi- ing numeric projections of population change into rectly, so sensitivity of InVEST models to LULC pat- spatially explicit human population density estimates terns translates into broad sensitivity to management or urban/rural area extents (e.g., Salvatore et al. 2005 ). choices. When models or maps of these drivers are available, For InVEST to assess the choices that managers or they can be combined with any approaches that other stakeholders want to consider, we need to project management impacts, giving scenarios that translate those choices into likely future land use represent all three major drivers of future change. We and cover patterns. There are many different meth- demonstrate one such application of InVEST with cli- ods for turning choices into LULC patterns. When mate scenarios in Chapter 18 . In the next section we decision-makers hold full control over the area of explore the specif cs of several real-world decisions interest, their own planning processes usually being addressed through scenario analysis with include the development of scenarios that can be InVEST. assessed by InVEST for likely ecosystem service provision. When the area of interest is a more com- 3.4.5 Examples of scenario development and plex landscape with multiple ownership and multi- InVEST application ple drivers of change, more complex scenario generation options are available. One can develop In many cases, managers want to know the likely predictions of how landowners will react given the outcomes of a proposed program, policy, or man- various market forces, institutions, and incentives agement action before they decide how to proceed that they face at regional (e.g., Nelson et al. 2008 ; (Ghazoul 2007 ). Maps of ecosystem service provi- Sohl and Sayler 2008 ), national (e.g., Veldkamp and sion can be used for planning, priority setting, Fresco 1996 ), or global scales (e.g., Alcamo et al. determining compensation or offset levels, design- 1994 ). For example, we can use models of land- ing policies or monitoring programs, or identifying owner decision-making to predict how landowners which members of society are controlling the provi- would react to changes in crop prices or to govern- sion of ecosystem services and which members are ment policies to generate scenarios of land use and receiving benef ts. Here, we consider several man- land cover that can be input into InVEST models. In agement questions and explore how InVEST can more demonstrative applications, users may want help answer them. These are all real cases where to explore what is possible on a particular landscape InVEST and its precursors are being applied. given the fundamental ecosystem service relationships in place. In these cases, landscape optimiza- The Chinese national government has a pro- tion modeling can be used to f nd solutions that gram to set up conservation areas to protect maximize the provision of a combination of serv- the natural capital that supports human well- ices. We have used this approach to maximize a being. Where should these areas be located? measure of species conservation for a given value of commodity production (and vice versa) using early This is an optimization question, like those com- versions of InVEST models ( Polasky et al. 2008 ). monly asked in the conservation arena. To answer We have emphasized management decisions as this question, we need to know which areas of the drivers of landscape change, but there are obviously landscape provide the highest level of services at other factors at work. Climate change and human the least cost. We can f rst use production function population growth will strongly inf uence LULC pat- and economic valuation models to estimate ecosys- terns and climate conditions in the future. Scenarios tem service levels and values. We can then feed INVEST: MAPPING AND VALUING ECOSYSTEM SERVICES 45 these maps into optimization algorithms that deter- and people. In many mitigation cases, offsets may be mine which parts of the landscape, if protected or designed with the intention of providing services, managed in a certain way, meet goals the policy but the actual connection to people is not considered maker sets for ecosystem service provision at the explicitly. In Florida for example, a wetland mitiga- lowest cost (e.g., number of hectares, purchase tion bank was created on a nearshore island to com- value of the land, opportunity cost of foregone pensate for development in a coastal watershed. activities). For those familiar with conservation People in this area get drinking water from groundwa- planning, this could be done by running MARXAN, ter wells. The wetland bank was meant to replace bio- or a similar program, with ecosystem service tar- diversity and water f ltration services for clean drinking gets. While InVEST does not include an optimiza- water lost through development, but the island hous- tion routine, outputs of InVEST easily can be used ing the bank was located over a different aquifer with in standard optimization packages to answer ques- no inhabitants. There was technically no net loss of tions of this sort. wetland area or function, but the drinking water- related services provided to people were lost entirely. The Nature Conservancy has completed an InVEST models can be used to identify where on a analysis of priority sites for conservation action landscape services are produced and used. in the Willamette Basin, Oregon (USA). They Colombia’s Ministry has generated scenarios that want to work f rst in the places that also give represent proposed mines and other development the highest possible benef t for human well- projects and InVEST in being used to identify how being. Where should they work? much service will be gained or lost, and where. InVEST is also being used to identify other areas in This is also a priority setting question, but one that the same landscape that provide similar levels of places biodiversity conservation above ecosystem service to the same people, thus allowing targeted service provision. In this case, we do not have to offsetting that will come closer to ensuring no net apply production function models to the whole loss of ecosystem services. landscape but can instead focus analyses on the priority sites for biodiversity conservation. The options A state agency in Oregon (USA) could design for conf gurations of priority areas become scenar- a subsidy that would motivate conservation of ios and InVEST can be used to estimate the levels natural habitat on private lands. The objective and values of services that society will receive from would be to achieve biodiversity protection conservation of each site. Planners can then choose and climate regulation (through carbon where to act. sequestration) with this single program. How should they target payments to get the best Colombia’s Ministry of the Environment is returns for both goals on a f xed budget? How responsible for permitting and licensing all much more return could they get if they major production sectors in the country. They increased the budget? are considering requiring offsets and mitigation for biodiversity and ecosystem service In data-rich environments like Oregon, scenarios damages caused by development in sectors can be developed by computer-based land transi- such as mining, agriculture, oil and gas, and tion models. This was the approach taken by Nelson infrastructure. If the expanded approach et al. ( 2008 ). Scenarios were created to represent f ve becomes law, how much mitigation should be possible subsidy programs, each designed to target required for a given development plan or set a different set of landowners: (i) all landowners, of permits, and where should the mitigation (ii) owners of land in riparian zones, (iii) owners be done to adequately offset the damages? who could restore certain rare types of habitat, (iv) InVEST has great potential for application in the mit- owners who could restore forests with large poten- igation and offsetting arena because it explicitly con- tial to sequester carbon, and (v) owners with parcels siders the connection between ecosystem processes with high species conservation value. Researchers 46 ASSESSING MULTIPLE ECOSYSTEM SERVICES projected land use change trajectories based on his- terms of ecosystem services not presently val- toric land change data in the region and altered the ued? Do some alternatives offer more equita- land change patterns based on an econometric ble distribution of benef ts to residents? model that determined which landowners would Most natural resource management decisions have likely participate in each subsidy program. InVEST major impacts on the balance of society, inf uencing and its precursor models were used to project who receives higher income streams, who has access changes in biodiversity and service provision under to markets, who must follow restrictive regulations, each policy scenario, clearly revealing returns asso- or even who has the right of access or ownership. In ciated with each policy design and showed the this case, the policies implemented in the Basin will trade-off in biodiversity conservation and ecosys- determine the balance of revenues among agricul- tem service provision under alternative policies. ture, timber, and real estate sectors, and which areas will be developed or conserved. To better under- Kamehameha Schools, the largest private stand the kinds of policies people living in the Basin landowner in Hawai’i, seeks to manage their would like to see implemented, the Consortium led lands to balance economic, environmental, a multi-stakeholder process to create three plausible educational, cultural, and community values. scenarios for the future that fell within the bounds They are considering several possible manage- of society’s comfort level. The three scenarios were ment plans for providing these values on an assessed with InVEST to reveal the trade-offs and iconic 26 000-ha parcel of land, the North synergies among commodities and ecosystem serv- Shore of O`ahu. Which plan will yield the best ices that can be expected under each possible future results? ( Nelson et al. 2009 ). In some cases, like this one, the key stakeholder has Going one step further, we can combine produc- control over the whole area of analysis. In this case, tion function models such as InVEST with socio- hand-drawing scenarios is an alternative to compu- economic data to understand how each possible ter simulation of scenarios since there is little in the future will affect social equity. Such analyses can land use and cover patterns that is left to outside help identify options that may lead to inequities social drivers or chance. We worked with before social groups or sectors are marginalized. Kamehameha Schools to mark up a map of the par- Preliminary analyses of this type are discussed in cel, identifying which parts of the landscape would Chapter 16 . be under different uses in the alternate plans. We then rendered these ideas with GIS software, creat- 3.4.6 Tiered modeling: f exibility for ing landscape scenarios that were assessed with a data-limited world InVEST. Managers will use the multiple service outputs of InVEST to identify how well each possible There is always a trade-off in modeling between scenario measures up against their multiple objec- making a model more complex and detailed and tives. These results will help inform the type of keeping it simple. Simple models require fewer management plan implemented. data, are often less prone to parameter estimation errors and subsequent error propagation, and can The Pacif c Northwest Ecosystem Research be easier to explain and understand. Complex mod- Consortium, a group of representatives from els require more information, but they often inspire government agencies, non-government greater conf dence because they more faithfully organizations, and production sectors, is cur- depict the details and underlying intricacies of proc- rently trying to align the policy trajectory of esses. Different applications and different users will the Willamette Basin (Oregon, USA) with the have specif c needs for either complicated or simple desires of its inhabitants. Which future will models of ecosystem services and valuation. For provide the greatest benef ts to society, both this reason, we have developed a tiered system of in terms of current market commodities and in models in InVEST ( Figure 3.2 ). FUTURE DIRECTIONS AND OPEN QUESTIONS 47

Modeling Reality Simple Complex Data Reality

Tier 1 Tier 2 Tier 3

Figure 3.2 A tiered approach to modeling ecosystem services. Given the diff culty of matching models with the desired level of complexity with often sparse data, we created InVEST with different types of models. Tier 1 models are simple and require few data. Tier 2 models are more complex and require more data. Tier 3 models add even greater conf dence and will often be site-specif c models created by other research groups.

Tier 1 models are the simplest models. We devel- lines will help clarify when greater model complex- oped these models to require few data and be easy ity is essential in decision-making and when it is to understand and explain to others, yet to retain unnecessary. suff cient credibility to guide management deci- InVEST provides a general framework within sions. Their distinguishing feature is a reliance on which one can mix and match tier 1 and tier 2 mod- readily available data that are generally accessible els depending on differing data availability or need everywhere in the world. Tier 1 models can draw for precision among services. By mixing tiers, information from the published literature, global InVEST allows users to customize its application to data sets, site-specif c data sources, local traditional specif c problems. While we have developed only knowledge, or expert opinion. Because of their sim- tier 1 and tier 2 models, it is possible to also use plicity, these models will be most appropriately InVEST with what we call tier 3 models, research applied in scoping and planning activities where level, state-of-the-art models (e.g., the CENTURY the purpose is to understand the general lay of the model ( Parton et al. 1994 ); SWAT model ( Gassman land. They may also be used in planning stages of et al. 2007 )). payment for ecosystem service programs or policy design to estimate likely returns from alternative 3.5 Future directions and open management options or to place focus on different questions parts of a landscape. In many applications the predictions of tier 1 InVEST provides a means for addressing many of models may be too crude, or too prone to errors of the current challenges of mapping and valuing averaging or aggregating, to meet the needs of deci- multiple ecosystem services to create change in con- sion-makers. For these cases, we provide more servation and natural resource management, but it detailed tier 2 models. These models require more is not a panacea. Many of the components of InVEST data, have more parameters, and are more time- are relatively new and untested. Methods for assess- consuming and diff cult to apply. However, tier 2 ing the validity and reliability of model predictions models are likely to be seen as “better” in the sense at landscape levels are needed. The models will of addressing more ecological complexity, allowing continue to undergo modif cation and updating as for greater spatial and temporal heterogeneity and experience and applications increase. generating more ref ned estimates. Instead of repre- A large unmet challenge in ecosystem service senting a world of “average trees” and the “average assessment and integrated decision-making is to pollinator” or habitat types sans species lists, tier 2 understand the distribution of ecosystem services models disaggregate groups or time steps to include among different groups in society and how this dis- age structure of trees, a variety of pollinator guilds tribution will likely change as a consequence of or species, monthly precipitation patterns, and so management decisions. While it is important to on. We discuss when tier 1 or tier 2 models are most know the total amount of ecosystem services pro- appropriate in Chapter 15 . Further work along these vided and their overall value to society, it is also 48 ASSESSING MULTIPLE ECOSYSTEM SERVICES important to know who benef ts from the provision to protect and maintain ecosystems. Addressing of services and their social and economic status. this mismatch and providing proper incentives for Without information about the distribution of ben- provision of ecosystem services will require changes ef ts from ecosystem services, management deci- in policies of local and national governments and in sions can lead to serious unintended consequences international agreements. Some progress on these for equity and well-being. This concern is especially fronts can be seen, as with the recent emergence of strong for management actions that negatively carbon markets, international policy discussions on affect underprivileged segments of society ( Pagiola Reduced Emissions from Deforestation in et al . 2005 ). In developing countries, for example, Developing Countries (REDD), and expansion of establishing a new national park or conservation programs of Payments for Ecosystem Services area has in some cases resulted in removing and (PES). Successfully linking the science of mapping separating people and their work (Kareiva and and valuing ecosystem services with proper institu- Marvier 2007 ), leading to a decline in their liveli- tions and policies will likely remain a major chal- hood or well-being. lenge for decades to come. Whether a particular social group wins, loses, or Finally, there is the need to expand InVEST to remains unaffected by a decision is determined by include models for other key terrestrial (see Box 3.1 ), several factors. Of utmost importance is access. freshwater, and marine ecosystem services. Useful Underprivileged members of society will not benef t groundwork is being laid for comprehensive mod- or lose from changes in ecosystem services if they eling of marine ecosystem services (Chapter 17 ) and cannot access those services. Access has two critical the Natural Capital Project is now developing components in this context: physical overlap in space InVEST models for f sheries production, shoreline and legal rights (and their enforcement). InVEST can protection, marine tourism and recreation, and show clearly where services will be provided on a marine biodiversity. landscape and how their provision is likely to change A user’s guide for InVEST can be found at http:// in space. This ability can give insights into the spatial invest.ecoinformatics.org and there is a growing overlap part of access. Rapid advances are being community that seeks to develop and apply multi- made in the mapping of social indicators of poverty objective ecosystem service models to decision- (CIESIN 2006; World Resources Institute 2007 ), and making. Whether one uses InVEST or any of several we can draw from these approaches to ask where other ecosystem service modeling approaches (e.g., ecosystem services and the poor overlap on the land- Artif cial Intelligence for Ecosystem Services scape. We provide some examples of this kind of (ARIES; ARIES 2008 ), EcoMetrix ( Primozich 2008 ), exercise in Chapter 16 . However, we currently do not Multiscale Integrated Models of Ecosystem Services have a standardized way for bringing information (MIMES; unpublished)), the greatest challenge is about institutions and their level of enforcement into not in developing the models, but in linking the a mapping and valuation context. Developing ways models pragmatically to everyday land- and to represent and predict the interactions among eco- resource-use decisions. system services, people, and institutions will be critical to the assessment of the distributional effects of References management decisions. A related challenge lies in altering f nancial and ARIES. (2008). ARIES white paper. University of Vermont, institutional infrastructure to give incentives that Burlington. Alcamo, J., Kreileman, G. J. J., Krol, M., et al . (1994). maintain and enhance the provision of ecosystem Modeling the global society-biosphere-climate system, services. Often there is a spatial or temporal mis- 1. Model description and testing. Water Air and Soil match between those who control the provision of Pollution , 76 , 1–35. ecosystem services and those who benef t from the Antle, J. M., and Stoorvogel, J. J. (2006). Incorporating sys- services. Without the ability to connect the demand tems dynamics and spatial heterogeneity in integrated for services with the supply from those who control assessment of agricultural production systems. it, there will be insuff cient incentive for suppliers Environment and Development Economics , 11, 39–58. FUTURE DIRECTIONS AND OPEN QUESTIONS 49

Barbier, E. B., and Strand, I. (1998). Valuing mangrove- Freeman, A. M. I. (2003). The Measurement of Environmental f shery linkages: A case study of Campeche, Mexico. and Resource Values: Theory and Methods . Resources for Environmental and Resource Economics , 12, 151–66. the Future, Washington, DC. Barbier, E. B. (2000). Valuing the environment as input: Gassman, P. W., Reyes, M. R., Green, C. H., et al. (2007). Applications to mangrove-f shery linkages. Ecological The soil and water assessment tool: historical develop- Economics , 35, 47–61. ment, applications and future research directions. Barbier, E. B. (2007). Valuing ecosystem services as pro- Transactions of the American Society of Agricultural and ductive inputs. Economic Policy , 22, 177–229. Biological Engineers , 50, 1211–50. Boerner, B. (2007). Ecosystem services, agriculture, and Ghazoul, J. (2007). Recognising the complexities of ecosys- rural poverty in the Eastern Brazilian Amazon: tem management and the ecosystem service concept. Interrelationships and policy prescriptions. Amsterdam, Gaia , 16 , 215–21. The Netherlands. Hougner, C., Colding, J., and Soderqvist, T. (2006). Boody, G., Vondracek, B., Andow, D. A., et al. (2005). Economic valuation of a seed dispersal service in the Multifunctional agriculture in the United States. Stockholm National Urban Park, Sweden. Ecological Bioscience , 55, 27–38. Economics , 59, 364–74. Brooks, K. N., Gregersen, H. M., Berglund, E. R., et al. Imperial, M. T. (1999). Institutional analysis and ecosys- (1982). Economic evaluation of watershed projects—an tem-based management: The institutional analysis and overview methodology and application. Water Resources development framework. Environmental Management , Bulletin , 18, 245. 24, 449–65. Center for International Earth Science Information Jackson, R. B., Jobbagy, E. G., Avissar, R., et al. (2005). Network (CIESIN). (2006). Where the poor are: An atlas of Trading water for carbon with biological sequestration. poverty. The Earth Institute at Columbia University, Science , 310, 1944–7. New York. Just, R. E., Hueth, D. L., and Schmitz, A. (2004). The Welfare Cleveland, C. J., Betke, M., Federico, P. et al. (2006). Economics of Public Policy: A Practical Approach to Project Economic value of the pest control service provided by and Policy Evaluation. Edward Elgar, Cheltenham, UK the Brazilian free-tailed bats in south-central Texas. and Northampton, MA. Frontiers in Ecology and Evolution , 4, 238–43. Kaiser, B., and Roumasset, J. (2002). Valuing indirect eco- Connor, E. F., and McCoy, E. D. (1979). The statistics system services: the case of tropical watersheds. and biology of the species–area relationship. American Environment and Development Economics , 7, 701–14. Naturalist , 113, 791–833. Kareiva, P., and Marvier, M. (2007). Conversation for the Costanza, R., d’Arge, R., de Groot, R., et al. (1997). The people—Pitting nature and biodiversity against people value of the world’s ecosystem services and natural makes little sense. Many conservationists now argue capital. Nature , 387, 253–60. that human health and well-being should be central to Daily, G. C. (1997). Nature’s Services. Island Press, conservation efforts. Scientif c American, 297, 50–7. Washington, DC. Konarska, K. M., Sutton, P. C., and Castellon, M. (2002). Das, S., and Vincent, J. R. (2009). Mangroves protected vil- Evaluationg scale dependence of ecosystem service val- lages and reduced death toll during Indian super uation: a comparison of NOAA-AVHRR and Landsat cyclone. Proceedings of the National Academy of Sciences of TM datasets. Ecological Economics, 41, 491–507. the United States of America . Early Edition , 1–4. Luck, G. W., Harrington, R., Harrison, P. A. et al. (2009). Ehrlich, P. R., and Ehrlich, A. H. (1981). Extinction: The Quantifying the contribution of organisms to the provi- Causes and Consequences of the Disappearance of Species . sion of ecosystem services. BioScience , 59 , 223–35. Random House, New York. Millennium Ecosystem Assessment (MA). (2005). Ellis, G. M., and Fisher, A. C. (1987). Valuing the einvron- Ecosystems and Human Well-being: Synthesis. Island Press, ment as an input. Journal of Environmental Management , Washington, DC. 25, 149–56. Naidoo, R., and Ricketts, T. H. (2006). Mapping the eco- Engel, S., Pagiola, S., and Wunder, S. (2008). Designing nomic costs and benef ts of conservation. PLoS Biology , payments for environmental services in theory and 4, 2153–64. practice: An overview of the issues. Ecological Economics , National Research Council (NRC). (2005). Valuing ecosys- 65, 663–74. tem services: Toward better environmental decision-making. Eshet, T., Baron, M. G., and Shechter, M. (2007). Exploring National Academies Press, Washington, DC. benef t transfer: disamenities of waste transfer stations. Nelson, E., Polasky, S., Lewis, D. J., et al. (2008). Eff ciency Environmental and Resource Economics , 37 , 521–47. of incentives to jointly increase carbon sequestration 50 ASSESSING MULTIPLE ECOSYSTEM SERVICES

and species conservation on a landscape. Proceedings of Salvatore, M., Pozzi, F., Ataman, E., et al. (2005). Mapping the National Academy of Sciences of the United States of global urban and rural population distributions. Food and America , 105, 9471–6. Agriculture Organization, Rome. Nelson, E. N., Mendoza, G. M., Regetz, J., et al. (2009). Sathirathai, S. and Barbier, E. B. (2001). Valuing mangrove Modeling multiple ecosystem services, biodiversity conservation in southern Thailand. Contemporary conservation, commodity production and tradeoffs at Economic Policy , 19, 109–22. landscape scales. Frontiers in Ecology and the Environment , Shabman, L. A., and Batie, S. S. (1978). Economic value of 7 , 4–11. natural coastal wetlands: a critique. Coastal Zone Pagiola, S., von Ritter, K., and Bishop, J. (2004). How Much Management Journal , 4 , 231–47. is an Ecosystem Worth? Assessing the Economic Value of Sohl, T. and Sayler, K. (2008). Using the FORE-SCE model Conservation. The World Bank, Washington, DC. to project land-cover change in the southeastern United Pagiola, S., Arcenas, A., and Platais, G. (2005). Can pay- States. Ecological Modelling , 219 , 49–65. ments for environmental services help reduce poverty? Swallow, S. K. (1994). Renewable and nonrenewable An exploration of the issues and the evidence to date resource theory applied to coastal agriculture, forest, from Latin America. World Development , 33, 237–53. wetland and f shery linkages. Marine Resource Economics , Parton, W. J., Schimel, D. S., Ojima, D. S., et al . (1994). 9, 291–310. Quantitative modeling of soil forming processes. In R. B. Troy, A. and Wilson, M. A. (2006). Mapping ecosystem Bryant and R. W. Arnold et al . , E d s . Special Publication , services: Practical challenges and opportunities in link- pp. 147–67. Soil Science Society of America, Madison, WI. ing GIS and value transfer. Ecological Economics , 60, Pereira, H. M., and Daily, G. C. (2006). Modeling biodiver- 435–49. sity dynamics in countryside landscapes. Ecology , 87, Turner, W. R., Brandon, K., Brooks, T. M., et al . (2007). 1877–85. Global conservation of biodiveristy and ecosystem serv- Plummer, M. (2009). Assessing benef t transfer for the ices. Bioscience , 57, 868–73. valuation of ecosystem services. Frontiers in Ecology and Veldkamp, A., and Fresco, L. O. (1996). CLUE-CR: An the Environment , 7 , 38–45. integrated multi-scale model to simulate land use Polasky, S., Nelson, E., Camm, J., et al. (2008). Where to put change scenarios in Costa Rica. Ecological Modeling , 91, things? Spatial land management to sustain biodiver- 231–48. sity and economic returns. Biological Conservation , 141, Vitousek, P. M., Aber, J. D., Howarth, R. W., et al. (1997). 1505–24. Human alteration of the global nitrogen cycle: Sources Primozich, D. (2008). Developing the Willamette ecosystem and consequences. Ecological Applications , 7, 737–50. marketplace. Willamette Partnerhsip, Salem. Wilson, M. A. and Carpenter, S. R. (1999). Economic valu- Ricketts, H. T., Daily, G. C., Ehrlich, P. R., et al . (2004). ation of freshwater ecosystem services in the United Economic value of tropical forest to coffee production. States: 1971–1997. Ecological Applications , 9, 772–83. Proceedings of the National Academy of Sciences of the Wilson, M. A. and Hoehn, J. P. (2006). Valuing environ- United States of America , 101, 12579–82. mental goods and services using benef t transfer: The Rosenberger, R. S., and Stanley, T. D. (2006). Measurement, state-of-the-art and science. Ecological Economics , 60 , generalization, and publication: Sources of error in ben- 335–42. ef t transfers and their management. Ecological Economics , World Resources Institute. (2007). Nature’s Benef ts in 60 , 372–8. Kenya: An atlas of ecosystem services and human well-being. Rosenberger, R. S. and Phipps, T. T. (2007). Correspondence World Resources Institute, Washington, DC. and convergence in benef t transfer accuracy: A meta- Yuan, Y., Bingner, R. L., and Boydstun, J. (2006). analytic review of the literature. In: S. Navrud, and R. Development of TMDL watershed implementation plan Ready, Eds., Environmental Values Transfer: Issues and using Annualized AGNPS. Land Use and Water Resources Methods . Springer, Dordrecht. Research , 6 , 2.1–2.8.

SECTION II Multi-tiered models for ecosystem services This page intentionally left blank CHAPTER 4 Water supply as an ecosystem service for hydropower and irrigation

Guillermo Mendoza, Driss Ennaanay, Marc Conte, Michael Todd Walter, David Freyberg, Stacie Wolny, Lauren Hay, Sue White, Erik Nelson, and Luis Solorzano

4.1 Introduction Although there is a lot still to be learned about the Water is necessary for all life. Precipitation through connections between land management, vegetation stream runoff and groundwater that is tapped via cover, and water yield (e.g., Chomitz and Kumari wells provide the world’s supply of water for drink- 1998 ; Bruijnzeel 2001; Bosch and Hewlett 1982 ; ing, irrigation, and hydropower generation. The Oyebande 1998), there is a demand for science- spatial and temporal availability of this water is based decision-making regarding policies, pay- strongly inf uenced by watershed geomorphology, ments, or activities that can alter water use. vegetation, and land and water management prac- Decision-makers need a credible and convenient tices. As such, natural capital (in this case, vegeta- methodology that explicitly links land use to water tion and soil) can support the provision of water delivery (Tallis et al . 2008). Several complex models services by regulating the amount and timing of are available for simulating water yield—most water availability. In this chapter, we presents mod- notably Hydrological Simulation Program—Fortran els that link land use and land cover (LULC) and (HSPF; Donigian et al. 1984) and Soil and Water several other key attributes to the quantity of sur- Assessment Tool (SWAT; Arnold et al. 1998 ). SWAT face water available for irrigation and for hydro- and HSPF are quasi-process-based hydrology mod- power production. els that are data- and time-intensive; they are diff - We focus on irrigation and hydropower because cult to apply in data-poor regions of the world, or in of their global economic signif cance. The vast situations lacking technically sophisticated support majority of water used worldwide is used for irriga- staff to calibrate the models. For this reason there is tion, accounting for up to 85% of fresh water use in a need for simpler models that can be more easily developing nations (IWMI 2001). The benef t of applied, especially in a context of examining many access to irrigation is the associated increase in crop ecosystem services at once, where trade-offs and yield: irrigated agriculture provides 40% of the glo- relative comparisons may be suff cient. bal food production (FAO, 2003). Water supply for We present two models that explicitly connect hydropower is also an economically signif cant land use and land cover to the regulation of surface service, and one likely to become even more impor- water f ows. We describe methods for converting tant because of its presumed low carbon emissions. modeled f ows into the level and value of two eco- For example, hydropower supplies 9% of the US system services: the regulation of water f ow for electricity and 49% of all renewable energy used in hydropower production and the regulation of water the USA (Edwards 2003). Some countries depend f ow for agricultural crop irrigation. We provide almost exclusively on hydropower, such as Tanzania two sets of models called tier 1 and tier 2 models. where hydropower accounts for more than 60% of The tier 1 models are the simplest and require the total generated capacity (Lyimo 2005). fewest data. They provide annual average outputs

53 54 WATER SUPPLY AS AN ECOSYSTEM SERVICE FOR HYDROPOWER AND IRRIGATION and do not require daily precipitation data, yet pro- not evaporate or transpire. In developing a model duce predictions that closely match more complex designed to accommodate areas with minimal models. The tier 2 models are more complex and access to data, we utilize a water balance model that provide predictions on a daily time step. is drawn from globally available data on annual precipitation and dryness indices that partition the 4.2 Tier 1 water supply model water balance for any place in the world (Budyko and Zubenok 1961; Milly 1994; Zhang et al . 2001). 4.2.1 Modeling water yield The model we describe is for surface water and The tier 1 water yield model is designed to evaluate does not separate groundwater, which will require how land use and land cover affect annual surface another approach (see Box 4.1 ). water yield across a landscape. We def ne the water Our water balance model is based on the hypoth- yield on a landscape as all precipitation that does esis that water yield can be approximated solely by

Box 4.1 Can we apply our simple model where groundwater really matters?

Heather Tallis, Yukuan Wang, and Driss Ennaanay grasslands, and15% in shrubs. Woodlands, paddy f elds, and residential and agricultural lands account for the Groundwater makes up 100 times more of the world’s remaining area. Soils are also diverse, including yellow, freshwater than surface water does (30 and 0.3%, yellow brown, mountain brown, dark brown, sub-alpine respectively) ( Gleick 1996 ). Groundwater also constitutes meadow, alpine meadow, and limestone soils. 30% of streamf ow, on average, around the globe ( Zektser We compared tier 1 annual average yield estimates to and Loaiciga 1993 ), although this percent varies observed streamf ow data (summarized to annual average dramatically by region. In regions of the USA where runoff) for the years 1995 to 2005. Model inputs included groundwater–surface water interactions are signif cant, average annual precipitation, potential evapotranspiration groundwater contribution to streamf ow can reach up to (calculated using Hamon method, grids generated from 90% ( Winter et al . 1998 ). The tier 1 water yield model does f ve weather stations within and surrounding the not account for such interactions, instead predicting the watershed), soil data, and a land use land cover map from total water depth generated from a parcel (combined 2005. Un-calibrated model runs showed poor agreement surface water, shallow groundwater, and deep with observed streamf ow levels, with our tier 1 estimates groundwater) based on characteristics of the watershed of falling 32% below observed levels. This was expected, interest. As part of the model development process, we given the high contribution of basef ow apparent from wanted to test whether our simplif ed approach could be monthly hydrograph and precipitation analyses. It is likely useful in regions where groundwater–surface water that the groundwater aquifer contributing to streamf ow interactions are signif cant. We selected a watershed in in this basin extends well beyond the borders of the Boaxing County, China (part of the Upper Yangtze River modeled watershed, so our estimates of yield based solely Basin) where an 11-year time series of river discharge on precipitation and watershed characteristics are missing showed a high contribution of basef ow and groundwater a key water source. To account for this additional source, discharge to total streamf ow. In other words, we picked a we conservatively took the lowest runoff depth in the watershed where we expected our model to perform poorly. driest period of the observed time series as the basef ow The basin is midsized (3240 km2 ), and is located in the and groundwater discharge depth generated outside the subtropical monsoon eco-region with an annual rainfall modeled watershed. Adding this value to our model average of 1172 mm. Rainfall is distributed over the year, estimate of annual average yield gave us 92% of the but there is an intense rainy season from May to September observed annual average streamf ow. (with f ooding from July to August). The basin is rather These results suggest that our simple, tier 1 model can steep (avg. slope 32°), with elevation ranging from 750 to be useful in regions of high groundwater–surface water 5328 m. Land cover and land use is relatively diverse, with interaction, but only if time series data are available for more than 47% of the area in forest, 21% in natural calibration. In the USA, the US Geological Survey has TIER 1 WATER SUPPLY MODEL 55

classif ed all watersheds into 24 regions that can help Dismal River, NE; Duckabush River, WA; Dry Frio River, TX; identify when the use of our tier 1 model would require Brushy Creek, CA; Sturgeon River, MI; and Ammonoosuc calibration. Each region represents an area of similar River, NH. In the other four regions, less than 50% of physiography, climate, and ground water–surface water streamf ow comes from ground water and tier 1 interactions (Winter et al. 1998 ). In six of ten regions applications could proceed more readily. Rivers character- analyzed for groundwater contribution to streamf ow, istic of those regions are Homochitto River, MS; Santa more than 50% of streamf ow comes from ground water, Cruz River, AZ; Orestimba Creek, CA; and Forest River, ND. and we would strongly recommend using our tier 1 model Similar kinds of classif cations, or a close look at time only under careful scrutiny and calibration. The following series data, will help identify where the tier 1 water yield rivers are representative of conditions in those six regions: model can be applied most readily around the globe.

the local interaction of f uctuating precipitation and climatic and soil conditions (Milly 1994; Potter et al. potential evapotranspiration given the water stor- 2005; Donohue et al . 2007), and is given by age properties of the soil (Milly 1994). The relation- ⎛⎞ AWCx ship between potential and actual evapotranspiration w xj = Z⎜⎟, (4.3) P is described by the Budyko curve, which is based on ⎝⎠xj over 2000 water balance observations representing where AWC x is a volumetric (mm) measure of the catchments of different climates and eco-regions water content (mm) in the soil available to plants, worldwide (Budyko and Zubenok 1961; Zhang et al . and Z is a parameter applied to each homogeneous 2001). In particular, we determine the annual basin in the landscape and is found with calibra- amount of precipitation that does not evapotran- tion. Generally, ω x , varies between 0.5 and 2 (Zhang spire, more simply called water yield, for each par- et al. 2001), with the 0.5 typical of pasture biomes cel on the landscape (indexed by x = 1, 2, . . . , X ): and the 2 typical of forest biomes. The above equations depict how the Budyko dry- ⎛⎞AET YPA1,xj (4.1) ness index ( R ) and the ratio of water available to jx= ∑⎜⎟−⋅⋅ xj x xj j ⎝⎠Pxj plants relative to annual precipitation ( AWCx /P xj )

where AET xj is the annual actual evapotraspiration affect the annual water balance. However, the model on parcel x with LULC category j , P xj is the annual does not explicitly incorporate the impact of the fre- precipitation on parcel x w i t h L U L C j , and A xj is the quency of annual events, the sub-parcel spatial var- area of x in LULC j . Annual precipitation can be iability of soil water storage capacity, and modif ed upward for j that have signif cant fog drip synchronicity of the energy-precipitation cycles on contributions ( Bruijnzeel 2000). the water balance, which all inf uence the water bal- The evapotranspiration portion of the water balance (see Milly 1994). In order to adjust for these ance, AETxj / P xj, is an approximation of the Budyko neglected effects, Z is used as a calibration constant. curve developed by Zhang et al . (2001): To determine Z , the user must have information on the annual water balance partition, which is gener- AET 1 + w R xj = xj xj , (4.2) ally obtained as the difference of observed annual P 1 xj 1 ++w R precipitation and observed annual streamf ow, cor- xj xj R xj rected for groundwater recharge and important

where Rxj is the dimensionless ratio of potential eva- consumptive losses. The parameter Z is adjusted potranspiration to precipitation, known as the until results from Eq. (4.2) reasonably correspond to Budyko dryness index (Budyko 1974), on parcel x observed water balance partitions. with LULC j , and ωjx is a dimensionless ratio of Finally, we def ne the Budyko dryness index as plant accessible water storage to expected precipita- kEToPET⋅ tion during the year. ω characterizes water balance R ==jx xj, (4.4) jx xj PP in distinctive plant communities, given prevailing xj xj 56 WATER SUPPLY AS AN ECOSYSTEM SERVICE FOR HYDROPOWER AND IRRIGATION

where R xj > 1 denote parcels that are potentially arid and use the ratio of saturated hydrologic conductiv-

(Budyko 1974), ETo x is the reference evapotranspira- ity to precipitation as a proxy for the relative effects tion on parcel x , and kj is the plant evapotranspira- of rainfall intensity and frequency. tion coeff cient associated with the LULC j on parcel We include these variables in the water retention

x . ETo x is an index of climatic demand with k j largely index ( v jx ) as: determined by j ’s vegetative characteristics ( Allen et al . 1998 ). ⎧ ⎪ ⎧⎫rjj⎧⎫ ksat x ()1min1,min1,−⋅HSSx ⎨⎬ ⋅ ⎨ ⎬ ⎪ rp 4.2.2 Water retention index ⎪ ⎩⎭for ⎩⎭d ⎪ vjx = ⎨ , (4.5) When the water retention properties of the land- ⎪1 when the water yield regulation scape are reduced, such as due to deforestation, ⎪ properties of the landscape are water yield is increased but without a means for ⎪ ⎪not applicable storage; larger portions can f ow to streams too ⎩ quickly for benef cial use. Our annual average yield estimation does not consider the exchange between w h e r e HSSx ε [0,1] is a normalized topographic index surface water and groundwater via inf ltration. from Lyon et al . 2 0 0 4 o f x that provides an index of

However, inf ltration is an important process linked hydraulic connectivity of a parcel to a stream, rj to the timing of water f ows and the availability of denotes j ’ s r o u g h n e s s c o e f f cient (Kent 1972), r for is a groundwater. We do not attempt to provide a simple normalizing roughness coeff cient that represents the model of quantitative groundwater recharge rates. roughness of a natural forest coverage (Kent 1972),

Instead, we provide a way to rank the landscape to ksatjx is the saturated hydraulic conductivity, a stand- identify areas where water can inf ltrate or leak into ard soil property, of x w i t h L U L C j scaled to a daily groundwater and thus be drawn on for extended time step, and p d is the mean daily rainfall depth at periods of time from basef ow or wells. Our water the area of interest. For most applications pd is likely retention index does not adjust the availability of to be constant across parcels. HSS ix i s b a s e d o n t o p o - surface water based on groundwater recharge rates, graphic wetness index (Beven and Kirkby 1979 ) that nor does it estimate the amount of groundwater predicts areas in the landscape prone to saturation recharge provided by a landscape. It does provide a and associated runoff due to drainage area, slope, simple approach for identifying high and low areas depth of soil, and permeability ( Steenhuis et al. 1995 ; of groundwater recharge, allowing managers to see Lyon et al. 2004). We argue that areas prone to satura- how management options will affect the location tion will tend to have higher soil moisture and the and relative magnitude of recharge. least favorable characteristics for water retention. Although many factors contribute to retention, The water retention index varies between 0 and 1. we consider three primary elements. First, because If v jx tends to 1, parcel x with LULC j has high water of geomorphology or artif cial drainage systems, retention characteristics, whereas if v jx tends to 0 it some parcels will be more hydrologically connected implies that parcel x has poor water retention prop- to streams. Secondly, vegetation and other surface erties and any precipitation will likely runoff imme- features can facilitate the inf ltration of surface diately. If data are lacking to def ne any of the water into the soil. Thirdly, if soil inf ltration capac- product terms in Eq. (4.5) that def ne v jx, one can ity is high relative to rainfall intensity or snowmelt, simply use the terms for which data are available, water will be more likely to inf ltrate the soil than to realizing that the characterization will be less accu- runoff to streams. Soil moisture, effective soil depth, rate. Obviously, water retention properties of the and mean rainfall depth and frequency interact to landscape are largely irrelevant if annual water determine runoff and leakage (Porporato et al . 2002). yield is regulated by large reservoir storage. Using We use the formulation of the topographic index as the index of water retention, we identify parcels x a proxy for soil moisture and unsaturated soil depth, that contribute to the landscape regulating function TIER 1 WATER SUPPLY MODEL 57 as those that exceed a threshold of water retention months that account for about 5% of annual water index score as Ỹ xj , where yield. βp can be determined by using studies of regional streamf ow distribution or by accounting ! YIxj=>()v xj α Yxj (4.6) for water (Eq. (4.7)) at a representative monitored catchment. When there are signif cant interbasin w h e r e I is an indicator function equal to 1 when groundwater transfers, the value of β can be greater v > α and equal to 0 otherwise. The constant α is p jx than 1, and the user will have to make corrections. the threshold value of v that establishes whether jx In the following sections, except the section on val- parcel x is a water-regulating parcel. uation for irrigation, we will assume an annual Our predictions of annual water yield using Eqs. period p . (4.1)–(4.4), and of identifying high inf ltration areas By contributing surface water to parcels down- using Eqs. (4.5) and (4.6) are admittedly highly sim- stream for consumptive use, “source” parcels (the v plif ed approximations. The basic idea is to use spa- parcels in Eq. (4.7)) provide an ecosystem service. In tial hydrology and land-use or land cover this section we calculate how much a source parcel information to reveal impacts of changes in land contributes to the benef ts derived from surface use and land cover. Our simple models predict water use downstream. First, we def ne B as the higher annual water yields but lower retention Dp benef t obtained from productive use of surface properties for urban areas than for forests, espe- water at demand point D in period p. Second, we cially if the forests lie on permeable soils and have a def ne B as the portion of benef t B provided by low hydraulic connectivity to a water body. This MDp Dp parcel x ’s surface water yield or the yield of a collec- agrees with f eld and more complicated modeling tion of parcels def ned by x Ì M (x o r M ’s contribu- studies. Later in the chapter we compare this tier 1 tive value will be 0 by def nition if x or M does not approach to more traditional detailed models such drain into demand point D ). as SWAT. ! βpxjpYSC⎛⎞ yp− yjp BBMDp= ∑ ⎜⎟∏ Dp , (4.8) 4.2.3 Water allocation in tier 1 xM⊂ SSDp⎝⎠yD⊂ yp Land use and land cover on a given parcel are not where y Ì D is the set of parcels along the f ow path the only determinants of how much water arrives between x or M and D , S and S is the supply of downstream for use. Consumptive uses also play a Dp yp surface water to D and parcel y in period p as role in determining downstream water supply. To def ned by Eq. (4.7), and C is the volume of water quantify the water available for use in irrigation or yjp consumed at y in year p . If there are no consump- the production of electricity, the model tracks water tion parcels along x and D ’s f ow path then consumption and water use along f ow paths. Let v be the set of parcels that drain to parcel x . The water ! βYxjp available for use during period p at parcel x is BBMDp= ∑ Dp . (4.9) xM⊂ SDp def ned as Sxp , The management unit M will likely have a wide range ! SYCxp= β p⋅−∑∑ jv jvp , (4.7) of sizes, ranging across the size of sub-watersheds, vx∈∈ vx protected areas, riparian buffers, and the smallest size

where S xp ³ 0, Cjvp is consumption at x with LULC j of landholdings. At the smaller size of management ! during demand period p , βpjv.∑Y is the yielded parcel, M , our approach likely generates greater water available for consumption from v given LULC errors because non-hydraulic boundaries neglect the

j during p , and βp is a constant, less than 1, to account effects of hydrologic interdependence within the for the fraction of annual water yield available dur- landscape and small scales remove the benef ts of ing p. For example, an irrigation district without a averaging over heterogeneous landscape properties. reservoir may require water during the dry season Our experience indicates that tier 1 annual yield 58 WATER SUPPLY AS AN ECOSYSTEM SERVICE FOR HYDROPOWER AND IRRIGATION

models are most useful when applied to evaluate In the next sections we propose formulations of BD trade-offs between management units of similar size for non contingent valuation. and located within similar watershed zones. For the sake of simplicity in def ning the equa- 4.2.4 Linking water supply to hydropower tions in the subsequent valuation sections, we production def ne the contribution of water from x to support an ecosystem service at D as FjxD : We modify Eq. (4.7) to def ne the average rate of water f ow available to generate hydropower at ⎛⎞SC− point H in year p as ! yjy FYjxD= b ⋅⋅ jx ⎜⎟ . (4.10) ∏ S ⎝⎠yD∈ y !! b YCYvjp − vjp vjp S = ∑∑vH⊂⊂vH , (4.12) Hp n The user can def ne benef ts BD as the power generated by a hydropower station, as crop yields by where H indexes the hydropower generating sta- an irrigated plot, or simply as the water contribution at demand point H , v Ì H is the set of parcels tion from x potentially supplied for productive use that drains to demand point H , and n represents the at D. In each case a biophysical map of ecosystem time steps to def ne f ow rate. services can be derived using Eq. (4.8) whereby the Dams are not only used for electricity production. contribution of parcel x in achieving the power gen- In the USA and Central America dam use is almost erated, crop yield, or water supply at D can be equally divided between irrigation, hydropower, mapped. water supply, f ood control, and recreation (see In many cases, conservation planners or manag- Table 4.1 ). To value water used for electricity gen- ers for payments of ecosystem services schemes eration, the user must know the average releases of simply want rules to effectively allocate a f xed water that go through the generating turbines, source of revenue from royalties or user fees, such which can depend on seasonal multi-use demands as in hydropower royalties for watershed protec- and the capacity of diversion infrastructure and of tion in Nepal (Winrock 2004), watershed protec- turbines. This model assumes that a constant fraction fees from utility bills in Quito, or sugar cane tion, γ , of S ¯ is released through the turbines to gen- growers in the Valle del Cauca of Colombia H erate energy. (Pagiola et al. 2002 ). Equation (4.8), which can be At hydropower station H , practically available modif ed for water quality protection or soil reten- power generated is calculated using the following tion practices, provides a tool for helping spatially equation, slightly modif ed from standard form allocate revenue in proportion to the services pro- (e.g., see Edwards 2003 for typical formulation): vided by different sections of a watershed. In this case of proportional allocation, Eq. (4.8) becomes emHH= ⋅⋅k (), g ⋅ShH ⋅ (4.13)

F jxD where ε is power generated in kilowatt-hours, h is BBMB= ∑ ⋅ , (4.11) H H xM⊂ SD the effective average head, μ is the turbine eff ciency (generally varying between 0.75 and 0.95), and κ is w h e r e t h e b e n e f ts from x d u e t o L U L C j a r e n o t a constant determined by the product of water den- dependent on the production benef ts at a water sity, gravity, and a conversion factor. The product, ¯ use point D , b u t r a t h e r a f xed value or budget, B B , γ· S H , represents the average water released from that in this case corresponds to willingness to pay turbines, which must be established from hydro-

(contingent valuation) a proportion of royalties or power operators. ε H represents the benef ts from the utility fees for conservation, or a total “score” value ecosystem functions for the generation of electricity to help prioritize the landscape. B M m i g h t c o r r e - due to water provision and regulation. The contri- spond to an allocation of the f xed budget for bution of each pixel upstream of H to εH is given by

protection, conservation, or restoration activities. substituting εH for B D in Eq. (4.8). TIER 1 VALUATION 59

Table 4.1. Global breakdown (%) of large dams by purpose (McMahon and Mein 1986)

Europe Asia North and South America Africa Australia Central America

Flood control 3 2 13 18 1 2 Hydropower 33 7 11 24 6 20 Irrigation 19 63 11 15 50 13 Multipurpose 25 26 40 26 21 14 Recreation 0 0 9 0 0 0 Water supply 17 2 10 13 20 49 Other 3 0 6 4 2 2

4.2.5 Linking water supply to irrigation w h e r e b Y! is the sum of surface water yield p ∑ vD⊂ vjp Predicting the provision of water supply for irriga- during p from the set of parcels v t h a t d r a i n t o D , s AW is the surface irrigation water (in tion poses several modeling challenges. Spatially ∑∑vD⊂ k kvp kvp kvp explicit modeling of irrigation supply and use volumetric units) consumed in the set of parcels v ,

and Gvjp is the sum of other water consumption in requires information on watershed characteristics ∑ vD⊂ as well as data on diversion systems and wells that the set of parcels v that drain into D. In other words, might be used to obtain water from neighboring drainage catchments or aquifers. Unlike the demand ∑∑∑CAvjp =+s kvp kvpWG kvp ∑vjp . (4.15) for hydropower, irrigation water is usually required vD⊂⊂ vDk vD⊂ at specif c periods within a year—when water is The irrigation water consumed in any parcel v is a scarce. Therefore, the landscape water yield regula- function of the water needs per hectare of crop k in tion properties discussed above are most critical v in period p not met by precipitation in period p when reservoir storage options on the landscape ( W in depth units), the area devoted to crop k in v are limited and the water needs of crops are not kvp during period p ( A ), and a management decision being met fully by precipitation (high water yields k v p of how much water needed by k in v ̱ during p is have no value for irrigation if this water is delivered actually delivered (σ Î [0,1]). when crop growth is not limited by water). kvp Therefore, detailed crop and water scheduling information is required for an accurate depiction of 4.3 Tier 1 valuation irrigation demand and use. Each parcel on the landscape is assigned a value Equation (4.7) provides the framework to deter- due to its contribution to each of the services pro- mine surface water used for irrigation. Unlike f led in this chapter. In the case of hydropower, a hydropower, the user must correctly account for parcel’s value is related to the value of the energy water diversions that may not be apparent in a produced. In the case of irrigation, value is deter- land-use map. The surface water available for irri- mined from additional crop productivity due to the gation at use point D during irrigation period p (as irrigation at point D . A parcel that yields water used opposed to year p) is S given by a modif ed ver- Dp downstream for hydropower or irrigation is sion of Eq. (4.7), assigned a share of the downstream production values according to its relative contribution to the utilized water f ow. By taking this approach we can ⎧⎫! SYADp = max⎨⎬bpvjp∑∑∑−−s kvp kvpWG kvp ∑vjp , 0 understand both the total value of each service ⎩⎭vD⊂⊂ vDk vD⊂ delivered to users on a landscape and the location (4.14) of high-value service supply regions. 60 WATER SUPPLY AS AN ECOSYSTEM SERVICE FOR HYDROPOWER AND IRRIGATION

4.3.1 The value of water f ow for hydropower power, NPVHCxH derived from management unit, M , is obtained as Albery ( 1968 ) estimates the maximum willingness to pay for water by comparing the cost of electricity F production from hydropower with that of the xHj NPVHCHM= ∑ NPVH H , (4.17) cheapest alternative source of electricity. The differ- xM⊂ SH ence between these two costs can be interpreted as the economic rent to the water resource. In order to where S H represents the total water used in electric- estimate this value, the user must have information ity generation at point H . about the cost of hydropower production as well as the next cheapest alternative cost of power genera- A p p l i c a t i o n e x a m p l e . Let us consider the value of tion. While this approach is theoretically accurate, it water available for electricity generation in f ve is also data intensive, requiring that users have hydropower plants of the Willamette river basin information about the costs of hydropower and ( Figure 4.1 ; see Plate 1). Given specif c power other sources of energy production. For this reason, generation ratings, and a LULC pattern from 1990, we choose to value water in hydropower genera- Eq. (4.1) simulates annual water yield at a 30 × 30 m tion based on the price of hydropower alone. While parcel resolution. In this example, the estimate of the this approach will not fully capture the social value net present value of water for hydropower is based of hydropower production (such as reduced green- on a price of electricity of $0.01 per kilowatt, with a house gas emissions compared to coal-f red power discount rate of 5% for a 100-year productive plants), it will provide a lower bound estimate of lifespan. Our model assigns the highest landscape this value. values to the Detroit and Green Peter hydropower Under this framework, the net present value of plant catchments because they have high power hydropower production is given by ratings with respect to their drainage areas (Figure 4.1a). The landscape draining into Lookout Point has lower per-parcel values despite its high energy T −1 ()pce − eH H (4.16) rating because the contributing area is large. To NPVHH = ∑ t t=0 (1+ r) examine how LULC change might impact the quantity and value of water available for hydropower

where pe is the market price of electricity (per unit of production, we developed a scenario in which all energy) provided by hydropower plant at dam d , ε H the forested area below 1000 m in elevation was represents the annual energy generated by hydro- converted into pasture. Using our tier 1 surface power station H , c H represents the average annual water yield model, the increase in water yield of our cost of operating hydropower station H a n d s h o u l d deforested landscape enhanced annual runoff by up include the external environmental damages of to 105 mm per parcel (Figure 4.1b). Since we were dam construction, T indicates the number of years only evaluating total annual yield in this example, we expect present landscape conditions to persist or our tier 1 approach assigned greater landscape gains the expected remaining lifetime of the station at H for hydropower provision to the deforested parcels (set T to the smallest value if the two time values (Figure 4.1c). Notice, that while deforestation is differ), and r is the market discount rate. likely to negatively impact biodiversity and carbon We rely on a parcel’s relative contribution of emissions, in this situation it enhances potential water used for hydropower to distribute the above hydropower production thru increased water yield. value to the parcels upon which the electricity- If deforestation also enhances sediment discharge generating water is generated. Let FxH j represent the into the reservoir, the net impact on hydropower amount of water that originates on parcel x in LULC generation might be reduced (see Chapter 6 ). This j that is available for use for hydropower produc- example is a good illustration of the trade-offs tion at demand point H (see Eqs. (4.8) and (4.10)). inherent in most land-use decisions, but glossed Then, the contribution of water values for hydro- over when considering one service at a time. TIER 1 VALUATION 61

(a) (b) North Fork North Fork

Value of Land for Hydropower Production ($) Change in Water Yield (mm) 21305 105 Detroit Detroit

1482 0 Dam Watersheds Dams Dam Watersheds Green Peter Dams Green Peter

Fall Creek Fall Creek Lookout Point Lookout Point

01020Kilometers 01020Kilometers

0 Detroit

–1106 Dam Watersheds Dams Green Peter

Fall Creek Lookout Point

01020Kilometers

Figure 4.1 Hypothetical example application of tier 1 model of water provisioning for hydropower generation in the Willamette river watershed. The example evaluates f ve sub-catchments of hydropower stations at North Fork (41MW), Detroit (115MW), Green Peter (92MW), Fall Creek (6.4MW), and Lookout Point (138MW). (a) The net present value of landscape water provision services for hydropower; (b) changes in water yield as a result of hypothetical deforestation of all land below 1000 m above sea level; and (c) the changes in landscape value for hydropower under the deforestation scenario. (See Plate 1.) 62 WATER SUPPLY AS AN ECOSYSTEM SERVICE FOR HYDROPOWER AND IRRIGATION

4.3.2 The value of water f ow used for 4.5 Tier 2 water supply model irrigation at the crop f eld Our tier 2 water models simulate hydrologic variabil- The monetary value of the amount of water used ity, incorporate the temporal nature of management, at D for irrigation purposes is given by B D . The and model additional hydrologic processes that inf u- water available for irrigation, SDp , at D is given by ence the ecosystem services discussed in this chapter.

Eq. (4.14). Let Irr Dp indicate the annual amount of In addition, tier 2 provides the tools for characterizing surface water required for consumption by annual water resources infrastructure. Irrigation water is crop irrigation demand at D that is required dur- diverted and often stored to support or permit agri- ing crop cycle with period, p , in an average year cultural production when water is lacking at crop

(Irr Dp ≤ S Dp ) . I n t h i s c a s e BD is equal to the net value f elds. Similarly, infrastructure for hydropower plants of agriculture production in D w i t h Irr Dp . See diverts, stores, or enhances energy production.

C h a p t e r 9 f o r d e t a i l s o n d e t e r m i n i n g B D a t i r r i - The soil matrix, vegetation roots, and shallow gated points D . The contribution of each manage- aquifers of a watershed provide a regulatory func- ment parcel in the landscape to BD is approximated tion by enhancing opportunities for basef ow or using Eq. (4.8). groundwater recharge. In effect, this provides a storage value to the watershed landscape. Tier 2 provides the means to simulate daily hydrology 4.4 Limitations of the tier 1 water yield and climate variability, water resources infrastruc- models ture, and all the rules of operation, management, As a general rule, the tier 1 water yield models are and rights. This additional functionality provides most physically realistic at watershed scales, spe- greater accuracy in absolute and temporal terms. cif cally, for areas that are hydrologically coherent. However, the tier 2 approach requires far more A hydrologically coherent area maintains the effort in compiling necessary data, and calibrating integrity of surface and subsurface water f ow the models, and may not be feasible unless one has paths or drainage to a point of discharge, such as formal hydrological and modeling training and rel- up to rivers, streams, creeks, or springs. When a atively long time windows for analysis. watershed is subdivided without accounting for The hydrology in the tier 2 models is driven by a the potential f ows between them, our tier 1 mod- modular system that allows users to replace or add eling approach presented can estimate relative different hydrology modules based on site-specif c contributions of different parts of the landscape to conditions, existing data, and hydrologic under- water f ow, retention, and consumption but will standing. In addition, the modular approach allows likely misrepresent the integrated watershed one to adjust model complexity as needed to explain response. Second, we do not explicitly model site-specif c phenomena (Farmer et al. 2003). Our groundwater, especially in terms of pumping from tier 2 approach uses the Precipitation Runoff shallow aquifers for irrigation. A third major short- Modeling System (PRMS) developed by the US coming of tier 1 water yield models is its annual Geological Survey (Leavesley et al. 1983) to estimate timeframe. Many critical hydrological events or water supply. Supply then becomes an input to a demands for water use occur on the timescale of water resources systems model that incorporates days, and representing hydrology as an annual demand, or use. Outputs of this model are then output can lead to large errors of interpretation. linked back to the watershed, such that the service One way around this is to build seasonal models, provisioning is a function of hydrology, infrastruc- so that as a compromise between annual and daily, ture, and management. one breaks the year up into different seasons, within which seasonal averages provide approxi- 4.5.1 Tier 2 modeling of physical hydrology mations that can get much closer to the degree of realism one seeks. Tier 2 addresses this shortcom- Our tier 2 hydrology model evaluates the impacts ing in that it is a daily time step model. of various combinations of precipitation, climate, TIER 2 WATER SUPPLY MODEL 63 and LULC on streamf ow, sediment yields, and with the data available and the user’s understand- general basin hydrology. In PRMS the spatial mod- ing and needs. D u ( t , σu ) represents water withdraw- eling units are def ned as Hydrologic Response als during t to satisfy demand u , w h i c h , l i k e x , a r e Units (HRUs), which represent watershed parcels part of the watershed landscape benef tting d . of homogenous geomorphology, vegetation, man- Different demands for water u will have different agement, and hydraulic connectivity. We simulate priorities, or management policies, σ u, t h a t c a n b e the hydrologic regime on a daily time step and per function of t a n d Q(t) . R u,STO ( t , σu ) a r e f ow releases storm event using historical records or probabi- from reservoirs STO , o r r e t u r n f ows from demands listic hydrology based on stochastic methods. By u d u r i n g t. R e t u r n f ows vary depending on the type incorporating inter- and intra-annual variability, we and eff ciency of water use and transport. They usu- can effectively examine the nonlinear response of ally represent the difference between water diverted hydrology to climate or management practices (i.e., to satisfy a demand and the amount of water actu- dry years often lead to disproportionate scarcity). ally consumed, which is often 80 to 90% of demand Within each HRU modeling parcel, the model keeps for municipal and industrial uses (Loucks and Van track of critical hydrological variables such as inf l- Beek 2005). V sto ( t) is the water stored during t i n a tration, soil moisture, and canopy interception of lake or reservoir, or unavailable water being routed water—all of which will impact water supply. In the through the stream channel network. Irr u ( t , σ u) i s t h a t tier 2 equations that follow, parcels or landscape irrigation water required at u , a n d WRR u i s t h e w a t e r units, x , are HRUs. requirement ratio, a fraction to account for transmission losses or other irrigation requirements for excess demand of irrigation water. See Chapter 9 for 4.5.2 Modeling of water resource systems: information on Irr ( t , σ ) . water accounting ud u The temporal dimension of Eq. (4.18) denoted by Once water supply is estimated with the PRMS t introduces seasonality and interannual variability, model, the tier 2 approach accounts for water such as dry and wet seasons or wet and dry years, resource systems and their impacts on f ows. In par- which are often key to meeting demand for ecosys- ticular, we recommend using the Water Evaluation tem services. In addition, in tier 2, the last four terms and Planning System (WEAP), which is a readily of Eq. (4.18) incorporate human effects on water available tool that links hydrology to water resources availability, such as storage infrastructure, diver- systems (Siebert and Purkey 2007). In tier 2, the sions, management strategies, and water rights water from parcel x available for use, S d, at demand policies. As might be evident, the greater is the point d is described in its most generic form for any alteration of f ow regimes by human inf uence dur- time, t , as ing t ; the lower is the relative importance of landscape hydrology. The temporal and engineering Stdu( )= ∑∑ Qt( )− Dt(),,ssu + ∑ Ru,STO () t u system management dimensions of tier 2 provide xd∈∈ ud ud ∈ (4.18) the tools for broader examination of ecosystem Irruu() t,s −−Vtsto(),s u services and trade-off options. The following sec- ∑∑WRR sto∈∈ d ud u tions describe how we link the ecosystem provision Qt S w h e r e ∑ xd∈ ( ) r e p r e s e n t s s t r e a m f ow (including of water, d ( t), to hydropower production and how both storm runoff and shallow groundwater dis- we perform the valuation of the landscape. See charge components) from parcels x i n t h e w a t e r s h e d Chapter 9 for the approach used in valuing the available for use at d d u r i n g t i m e s t e p t . ∑Qt( ) r e s u l t s landscape for irrigation. xd∈ from interaction of each hydrology module, such as canopy interception, inf ltration and percolation, 4.5.3 Linking water supply to hydropower transpiration and evaporation, groundwater proc- production esses, and vegetation growth. Each of the biophysical processes represented by the modules can be Our tier 2 biophysical model allows us to calcu- represented at a level of complexity commensurate late the daily water yield and use on a parcel. This 64 WATER SUPPLY AS AN ECOSYSTEM SERVICE FOR HYDROPOWER AND IRRIGATION f ne-scale temporal resolution lets the user model f rst constraint states that the release of water the effects of time-sensitive management decisions through dam d in t (given by r dp ( t) + r do ( t )) cannot be on energy production. This ability should add real- greater than the amount of water in the reservoir at ism to the model predictions associated with tem- the beginning of t (given by Vd (t)) less some mini- poral variability in electricity production and price mum amount of water that the manager wants to – throughout the year. Societal demand for electricity maintain in the reservoir at all times (given by V ). – d can be decomposed into two segments: peak and V d is unique to each impoundment system and is off-peak. In temperate zones, peak demand for elec- assumed to be a function of the manager’s risk pref- tricity occurs during the summer when electricity erences, d ’s other uses (e.g., recreation, source of use increases to cope with higher temperatures. drinking water, source of irrigation water), and Because the price of electricity is higher during other reservoir management policies. The second periods of peak demand, we expect managers of and third constraints limit the rate of release during impoundment systems (reservoirs and dams) to the peak and off-peak periods to what is physically maximize production during peak periods. To do possible, given by r¯ dp ( t) and r¯do ( t), which are func- this, system managers store water in the reservoir tions of the dam’s capacity and the duration of peak during off-peak periods. and off-peak times in period t . The f nal equation is The impoundment system manager must choose used to track time-specif c reservoir water levels releases through generating turbines r dp (t) and r do (t) where Sd (t) gives the recharge rate of the impound- for each time step t , where there are T time steps in ment d . See Eq. (4.18) for the calculation of S d (t) . a year. r dp (t) and r do (t) represent peak and off-peak release volumes, respectively. The net revenue- 4.5.4 The challenge of calibrating tier 2 water maximizing manager will choose r (t) and r (t) for dp do models each time t according to an objective function based on Eq. (4.12) similar to Calibration of physical hydrology parameters is usually part of a three-step process of calibration,

T verif cation, and validation of a model. Calibration ⎛⎞prthVtpdpdde ⋅⋅() ( ())+ max km⋅⋅ −cdt (4.19) establishes model parameters using optimizing rr,..., ∑ ⎜⎟ dp1 dpT t=1 ⎝⎠prthVtododde ⋅⋅() ( ()) rrdo1 ,..., doT objective functions to simulate specif c hydrologic processes. Verif cation is generally associated with Subject to evaluating performance of a calibrated model using a different data set or time series at the same cali-

rtdp( )+ rt do ( )≤− VtV d ( )d tT= 1,..., brated site. Validation of the model occurs when

rtdp( )≤≤ rt dp ( ) and rtdo ( ) rt do ( ) t= 1,..., T , performance of the model continues to be adequate

Vtd (1)+= Vddpdod−− r ()()(), t r t+ St using optimized parameters and setup during calibration and verif cation at different yet representa- where the objective function indicates the annual tive sites. Having access to representative data (Yew net revenue generated at d , p pε and p o ε are the prices et al. 1997) correctly conf guring the processes of electricity during peak and off-peak periods, (Farmer et al . 2003) and optimizing parameters, respectively, p pε > po ε , and cdt indicates the marginal such that they can be considered “validated”, are cost of maintaining impoundment system d (we the biggest limitations to using tier 2 hydrology assume cdt is the same across each t ). h d ( V d ( t)) is the models in ecosystem service prediction. generating hydraulic head that is a function of vol- The purpose of model calibration is to better iden- ume of water, V d ( t ), in reservoir d during each time tify model parameter values that cannot be accu- step t determined by accounting for all inf ows from rately determined based on physical data. The the watershed, S d (t), and direct precipitation, and traditional approach to calibration of a distributed- subtracting all releases and other losses. hydrologic model has been comparison of measured The last four equations place constraints on the and simulated runoff at the outlet of the basin. choices of {r dp (1), . . . , r dp (T)} and { rdo (1), . . . , r do (T)}. The Farmer et al. (2003) recommend to begin a calibration SENSITIVITY ANALYSES AND TESTING OF TIER 1 WATER SUPPLY MODELS 65

process by f rst adequately simulating the general TT * ⎛⎞1 annual water balance, that is, calibrate for interan- NPVHd = ∑∑Ψ⋅⎜⎟t tt==11⎝⎠(1+ g ) nual yields of a basin to adequately account for the . (4.20) impact of dry and wet years on ecosystem services. ⎛⎞prthVt⋅⋅* () ( ())+ Ψ* = km⋅⋅pe dp d d −c Then, the modeler must adequately simulate the sea- ⎜⎟* dt ⎝⎠prthVtoe⋅⋅ do() d ( d ()) sonal variability of the water balance (intra-annual performance). Once the inter- and intra-annual com- We also provide a model for predicting the contri- ponents of the simulated water balance correspond bution of water yield to run of the river hydropower within acceptable ranges of observed values, the cali- production. Managers of diversion systems (run-of- bration process should continue to match f ow dura- river dams) do not have limited capability to store tion curves. Finally, the modeler can make water. The total annual value of hydroelectric pro- adjustments to generate a suitable time series of pre- duction at diversion system dam z with negligible dicted hydrology. Clearly, how one assesses model storage is determined by its unmanaged f ow’s tem- performance should be dictated by the ecosystem poral relationship with peak and off-peak periods. services of interest. For example, if ecosystem serv- A diversion system powered by f ows that tend to ices for irrigation depend on drought f ows during be higher during peak hours than off-peak will gen- the months before a rainy season, then calibration erate higher revenues than other diversion systems efforts should focus on adequately modeling the rel- with the opposite f ow–peak relationship, all else evant basef ow processes and reporting performance equal. The net present value of energy produced by statistics during those dry months. dam d (point of use value) until time T is given by

NPVH z ,

4.6 Tier 2 valuation model TT⎛⎞1 NPVHz = ∑∑Ψ⋅⎜⎟t The tier 2 valuation models are implemented at tt==11⎝⎠(1+ g ) , (4.21) applicable management time steps instead of aggre- ⎛⎞prthVtpe⋅⋅ zp() z ( d ())+ gating on an annual basis. Water yield in the tier 2 Ψ = km⋅⋅⎜⎟ −czt ⎝⎠prthVtoe⋅⋅ zo() z ( d ()) supply and use models is measured at a time scale that matches the temporal scale of water use where r (t) indicates the amount of water that f ows decisions. zp through z during the peak period of t , r zo (t) indicates the amount of water that f ows through z during the off-peak period of t , and c indicates the average 4.6.1 The value of water f ow for hydropower zt cost of maintaining dam z (assume c zt is the same In order to assess the value of hydropower genera- across each t ). tion, we assume the net revenue has been maximized over some time period. In other words, given operation rules def ned in Eq. (4.19) that maximize 4.7 Sensitivity analyses and testing of net revenue subject to policy constraints, we esti- tier 1 water supply models mate expected annual energy production, ε ¯ , hyd 4.7.1 Sensitivity analysis of tier 1 models which is used to determine the net present value of energy. Let T indicate the number of years we expect Because of the many simplifying assumptions of present landscape conditions to persist or the the tier 1 water models, it is essential to gain an expected lifetime of d (set T to the smallest value if understanding of their sensitivity to different the two time values differ). parameters. To accomplish this, we conducted a for- * * * * Let {r d p ( 1 ) , . . . , rd p ( T ) } and {r d o ( 1 ) , . . . , rd o ( T ) } indicate mal multivariate sensitivity analysis (Nearing et al . the peak and off-peak releases that maximize the 1989) that calculates global sensitivity using the net present value of water used in the production of maximum, minimum, and average values for each hydropower, NPVH d . Then NPVHd is given by parameter, i , in the model as 66 WATER SUPPLY AS AN ECOSYSTEM SERVICE FOR HYDROPOWER AND IRRIGATION

We do not present sensitivity analysis for the ()OUTiii,max − OUT,min IN ,aver SIi = ⋅ , (4.22) water retention index because it contains a hydrau- ()INii− IN OUT i ,max ,min ,aver lic connectivity parameter that is dependent on specif c watershed geomorphology, shape, and land-use

where INi , max , INi , min , and IN i, aver are respectively the conf gurations. Thus, these factors are somewhat maximum, minimum, and average values for model invariant for each watershed, although there may input parameter i , OUT i, min and OUTi , max are the be issues associated with the scale of available data model outputs corresponding to input parameter or how “parcels” are def ned. A valuable tool for

INi , min and INi , max respectively while holding all other managing uncertainty is to explore several simula- parameters at their average value, and OUT i, aver is tions using the reasonable ranges of possible param- the average of OUT i, max and OUT i , min . eter values and examining how rankings of A sensitivity index of 1 indicates that the model ecosystem services provision or regulation values average output varies with the same magnitude differ. and direction as the model average input. A negative value means that the input and output change inversely. Sensitivity analysis for the tier 1 water 4.7.2 Testing of tier 1 models yield model is presented by Table 4.2 , which sum- Given the major simpli f cations we have made to marizes the sensitivity index values for different provide tier 1 models that can be run with mini- parameters of the water yield model. The parame- mal data inputs, it is important to know how well ters in Table 4.2 are listed in descending order of our simple models agree with more complex, sensitivity. These values are all negative, meaning widely accepted models. However, there are no that any increase in parameter value corresponds to (or very few) models that predict a landscape’s a decrease in model output. The most sensitive contribution to particular uses such as hydro- parameters of our tier 1 model were associated with power and irrigation. Much more common are vegetation properties: the evapotranspiration coef- models for the f rst step of our modeling process: f cient and root depth. This has two important the biophysical supply step. As such we focused implications. First, it may indicate that vegetation is on verifying this biophysical modeling approach extremely important in regulating ecosystem serv- by testing it against the popular hydrology model, ices. Secondly, it means that users should pay spe- SWAT, in climatically diverse eco-regions of the cial attention to uncertainty associated with USA: California, Texas Gulf, Tennessee, the vegetation properties. Willamette river basin, and the lower Colorado Solving Eq. (4.2) for all dryness index ratios pro- basin (Figure 4.2 ). We aggregate our model results vides insight on how sensitive the model is to cli- into catchment areas within each eco-region, and mate, i.e., the energy–rainfall relationship. In compare spatial variation in water yields pro- general, Eq (4.2) has a greater sensitivity to land-use duced by our tier 1 model to spatial variation in parameters the closer the PET–rainfall ratio is equal water yield produced by the much more detailed to 1. In particular, very dry regions are least sensi- and data-intensive SWAT model (Arnold et al. tive to climate factors in the tier 1 model. 1998 ) .

Table 4.2 Sensitivity index for the tier 1 yield model and the principal drivers that describes parameters

Parameter, i Parameter description Driver SIi

k xj Evapotranspiration coeff cient Land use -97184 RD Root depth Land use -7324 Z Water balance calibration constant Climate -45

AWC x Available water capacity Soil property -0.161 SENSITIVITY ANALYSES AND TESTING OF TIER 1 WATER SUPPLY MODELS 67

Willamette

California

Tennessee Lower Colorado

Texas Gulf

Figure 4.2 Basins across the continental USA representing different eco-regions used in testing tier 1 water yield model rankings with respect to SWAT.

Our measure of model concordance is the non- Results show that our limited data models can parametric Kendall tau test (e.g., Helsel and Hirsch predict trends and rankings fairly well for annual 2002), which quantif es the degree of agreement or water yield. With respect to absolute values we disagreement between two non-normal ranked sets overestimate higher and underestimate lower of data; a tau coeff cient value (τ) of 1 means perfect results when compared with SWAT (Figure 4.4 ); in agreement, while a τ of 0 means the sets are com- other words, our tier 1 model somewhat exagger- pletely independent. The Kendall tau correlation ates water yield at both the high end (overestimates) statistics for all tests are shown in Table 4.3 . We and low end (underestimates). In simulating water evaluate performance for annual water yield, and yield, we found the best agreement in model results compare our water retention index with SWAT in Tennessee (τ = 0.89, p < 0.01). The lower agree- annual groundwater percolation outputs. Figure 4.3 ment between models in the Willamette river basin provides a basin scale inspection of comparative is likely due to the smaller scale. model outputs for water yield and water recharge Moreover, when our tier 1 average annual water index in the Texas Gulf and Willamette river basins, yield model was calibrated and then compared with respectively. Figure 4.4 provides scatter plot com- observed data in f ve watersheds in the Hainan parison between outputs from our non-calibrated Island of China, performance is greatly improved in tier 1 water yield model and that from SWAT model absolute terms (Figure 4.5 ). However, it is impor- applied to average annual water yield in the Texas tant to note that research is needed to evaluate per- Gulf as presented in Table 4.3 . formance within management units that do not

Table 4.3 Kendall tau ranking statistics for non-calibrated tier 1 water yield and water retention index outputs with respect to annual yield and groundwater percolation simulation by SWAT at sub-catchments in f ve eco-regions/river basins in the USA

Eco-region N Water yield Water retention index

R2 Kendall tau (τ) p value R2 Kendall tau (τ) p value

Willamette 111 nl 0.38 < 0.01 0.59 0.50 < 0.01 Texas 122 0.89 0.59 < 0.01 nl 0.25 < 0.01 Tennessee 32 0.96 0.89 < 0.01 nl 0.15 0.25 Lower Colorado 85 nl 0.59 < 0.01 nl 0.23 < 0.01 California 135 0.95 0.42 < 0.01 nl 0.18 0.003

n = Number of sub-catchments per eco-region. nl = nonlinear 68 WATER SUPPLY AS AN ECOSYSTEM SERVICE FOR HYDROPOWER AND IRRIGATION

Figure 4.3 Comparison of InVEST and SWAT spatial patterns in annual water yield. Five quantile rankings of mean annual water yields for the Texas Gulf basin between our tier 1 water yield approach (top left) and the SWAT model (top right), and our tier 1 water retention index for the Willamette River basin (bottom left) and the SWAT model’s mean annual groundwater percolation rate (bottom right).

necessarily have distinct hydrologic boundaries, percolation rate estimated by SWAT (Eq. (4.5)). since trade-off analysis often occurs within the Comparisons between our water retention index boundaries of a hydrologic region. and groundwater f uxes in SWAT yielded signif - The performance statistics of our tier 1 water cantly positive, but often relatively low measures of retention index model were not as good. In this concordance ( Table 4.3 ). case, we compared our water retention index, which To preliminarily test our water retention index is based on surface roughness, hydraulic connectiv- against observed data, rather than another model, ity, and surface permeability, with the groundwater we compared streamf ow recession coeff cients SENSITIVITY ANALYSES AND TESTING OF TIER 1 WATER SUPPLY MODELS 69

1000

800

600

400

SWAT Annual Water Yield (mm) 200

0 0 200 400 600 800 1000 Tier 1 Annual Water Yield (mm)

Figure 4.4 Scatter plot comparison of our tier 1 annual water yield and annual SWAT yield in the Texas Gulf. A non-calibrated average water yield tier 1 model is compared to a calibrated SWAT study to illustrate model application with limited data can provide useful guidance for landscape ranking. The dashed line corresponds to the 1:1 line.

100 ) 3 75

25 Observed annual yield (100 Mm Observed annual

0 0255075100 Modeled annual yield (100 Mm3)

Figure 4.5 Scatter plot comparison of a calibrated tier 1 annual water yield model to observed annual water yield for 1980, 1985, 1990, 1995, and 2000, in f ve watersheds of Hainan Island, China. Solid circles correspond to 1980, empty circles to 1985, solid squares to 1990, empty squares to 1995, and triangles to 2000. The solid bounding lines correspond to a 40% uncertainty range and the dashed line corresponds to the 1:1 line. Units are in 100 000 m3 . 70 WATER SUPPLY AS AN ECOSYSTEM SERVICE FOR HYDROPOWER AND IRRIGATION

0.8

0.6

0.4

0.2

Average Streamflow Recession coef. Streamflow Average 0 0 0.2 0.4 0.6 0.8 1 Water Retention Index

Figure 4.6 Scatter plot of water retention index and average streamf ow recession coeff cient for 180 unregulated stream gauges across the continental USA. Average streamf ow recession coeff cients are derived from unregulated streamf ow (selected for minimal anthropogenic effects in the HCDN data set by Slack et al. 1993). Receding streamf ow is sampled from data time series with at least 4 consecutive days of decrease in f ow, from sub-catchments between 200 and 600 square miles, occurring between June and September, after 1970, and with aridity index greater than 1 (see Eq. (4.4)). For each of the 180 stream gauge stations, on average, 122 time series were used to approximate a recession f ow coeff cient ( RFC ) by a simplifying assumption of an ⎛⎞Q ln t=0 ⎝⎠⎜⎟Q exponential decay in f ow, i.e., RFC = t ). An average water retention index was estimated by applying Eq. (4.5) for the sub-catchments t corresponding to the stream gauges. The streamf ow recession coeff cient is usually inversely proportional to the water regulation function of the landscape. The higher the RFC, the lower the water regulation properties of the landscape, and vice versa.

from streamf ow data from across the continental 4.8 Next steps USA to respective values of the water retention index. The recession coeff cient index (similar to an Watersheds can support the provision of water-re- index of the rate of decay) is an indicator of the rate lated ecosystem services by regulating the amount that basef ow decreases after a rainfall period. A and timing of water availability. However, more sci- high value for the recession coeff cient signif es that entif c guidance is needed on how to support con- the rate of decrease of stream basef ow is high, servation practitioners in developing and managing which might be expected to result from a landscape watersheds to preserve or enhance these services, with poor water retention properties, and vice and how to tailor the approaches to different biomes versa. Figure 4.6 compares streamf ow recession (Rodriguez-Iturbe 2000). We have presented simple coeff cient data with water retention index. We (tier 1) and more complex (tier 2) approaches for found that, in general, watersheds with higher valuing the general landscape properties for water streamf ow recession coeff cient values were more quantity and timing as applied to hydropower gen- likely to have lower water retention index values. eration and irrigation. These tools may not necessar- That is, stream basef ow that decrease the fastest ily provide def nitive answers but they do provide a can be associated with landscape properties that structured framework for linking hydrologic science score lowest on our water retention index. This with decision-makers and planners. These tools f nding suggests that it is reasonable to def ne a need to be continuously improved, used cautiously threshold value of water retention, such as in Eq. and transparently, while presenting uncertainty (4.6), to evaluate whether the landscape effectively ranges as a starting point for informed decision- regulates streamf ow. making in the spirit of adaptive management. NEXT STEPS 71

Improvements to our models will include the can be more reliably applied; tier 1 hydrologic algo- incorporation of uncertainty using two possible rithms can be improved and catered to specif c approaches. The f rst involves drawing parameter needs, and ultimately lead to improved valuation values from a distribution for a given LULC across with the use of tier 2 models if desired. a landscape for parameters whose values might vary with LULC type. This approach is insightful because it can provide an uncertainty range of aver- References age service provision by LULC characteristics. The Albery, A. C. (1968). Forecasting demand for instream second approach would assign conf dence intervals uses. In W. R. D. Sewell and B. T. Bower, Eds. Forecasting directly to non-landscape specif c parameters whose the Demands for Water . Department of Engineering, values might vary between regions and affect the Mines, and Resources, Ottawa, Canada. impact of LULC characteristics. Allen, R. G., Pereira, L. S., Raes, D., et al. (1998). Crop eva- Hydrologic improvements to tier 1 will include potranspiration guidelines for computing crop water require- improvements to our water regulation function; addi- ments . FAO Irrigation and drainage Paper 56. Food and tion of soil moisture dynamics for improving esti- Agriculture Organization of the United Nations. Rome, mates of likely inf ltration and leakage, and reducing Italy. errors due to modeling scales used; a proper def ni- Arnold, J. G., Srinivasan, R., Muttiah, R. S., et al . (1998). tion of water partition parameter Z in Eq. (4.3) that is Large area hydrologic modeling and assessment, Part I: most likely a nonlinear function; improvements to Model development. Journal of the American Water Resources Association , 34 (1), 73–89. hydropower generation by def ning an average Beven, K. J., and Kirkby, M J. (1979). A physically based hydraulic head that is dependent on inf ows, such as variable contributing area model of basin hydrology. by adapting Gould-Dincer solutions (see McMahon Hydrologic Science Bulletin , 24 (1), 43–69. et al. 2007); and a stochastic function to relate the tim- Bosch, J. M., and Hewlett, J. D. 1982, A review of catching of irrigation requirement with the timing of water ment experiments to determine the effect of vegetation availability. In addition, our water regulation and changes on water yield and evapotranspiration. Journal provision functions can be modif ed and coupled dif- of Hydrology , 55 , 3–23. ferently to support other services, such as fresh water Bruijnzeel, L.A. 2000, Hydrology of tropical montane cloud f sheries or drinking water. Although tier 2 can better forests: A reassessment . Tropical Environmental account for the complexity of natural systems at any Hydrology Programme, Amsterdam. time step, it needs to be better integrated with the Bruijnzeel, L. A. (2001). Hydrological functions of tropical forests: not seeing the soil for the trees? Agriculture, spatial sophistication of tier 1. Ecosystystems and Environment , 104 (1), 185–228. Finally, we need to learn from f eld application of Budyko, M. I. (1974). Climate and Life, Academic Press, San our models in collaboration with planners and serv- Diego, CA. ice providers. Adaptive management, which recog- Budyko, M. I., and Zubenok, L. I. (1961). The determina- nizes that humans do not know enough to eff ciently tion of evaporation from the land surface. Izvestiya manage ecosystems and that decision-makers can- Akademii Nauk. SSR Seriya Geograf ya , 6 , 6–17. not afford to consequently postpone action (Lee Chomitz, K. M., and Kumari, K. (1998). The domestic ben- 1999), is an important next step for integrated model ef ts of tropical forest preservation: A critical review improvement. Adaptive management, often emphasizing hydrological functions. World Bank described as a process of learning by doing (Lee Research Observer , 13 (1), 13–35. 1999), provides a framework where our tier 1 mod- Donigian, A. S., Jr., Imhoff, J. C., Bicknell, B. R., et al . (1984). Application guide for Hydrological Simulation Program— els become incredibly valuable for helping early Fortran (HSPF) . U.S. Environmental Protection Agency, decision-making and dialogue, but can also be iter- Environmental Research Laboratory, Athens, GA, EPA- atively improved. At the onset of a project, tier 1 can 600/3-84-065. be used to rank and prioritize the landscape to eff - Donohue, R. J., Roderick, M. L., and McVicar, T. R. (2007). ciently allocate scarce management resources. As a On the importance of including vegetation dynamics in knowledge base increases through hydrologic mon- Budyko’s hydrological model. Hydrology and Earth itoring and analysis, tier 1 market-based valuation System Sciences , 11 , 983–95. 72 WATER SUPPLY AS AN ECOSYSTEM SERVICE FOR HYDROPOWER AND IRRIGATION

Edwards, B. K. (2003). The economics of hydroelectric power . Oyebande, L. (1998). Effects of tropical forest on water Edward Elgar, Northampton, MA. yield. In E. R. C. Reynolds and F. Thompson, Eds., Farmer, D., Sivalapan, M., and Jothityangkoon, C. (2003). Forests, climate, and hydrology: regional impacts, pp. 16–50. Climate, soil, and vegetation controls upon the variabil- United Nations University; Kefford Press, Singapore. ity of water balance in temperate and semi-arid land- Pagiola, S., Bishop, J. and Landell-Mills, N. (2002). Selling scapes: Downward approach to water balance analysis. forest environmental services: Market-based mecha- Water Resources Research , 39 (2), 1–21. nisms for conservation and development. Earthscan, FAO. (2003). The state of food insecurity in the world , Food London. and Agriculture Organization of the United Nations, Porporato, A., D’Odorico, P., Laio, F., et al . (2002). Rome. Ecohydrology of water-controlled ecosystems. Advances Gleick, P. H. (1996). Water resources. In S. H. Schneider, in Water Resources , 25 , 1335–48. Ed., Encyclopedia of Climate and Weather . Oxford Potter, N. J., Zhang, L., Milly, P. C. D., et al . (2005). Effects University Press, New York. of rainfall seasonality and soil moisture capacity on Helsel, D. R., and Hirsch, R. M. (2002). Statistical methods mean annual water balance for Australian catchments. in water resources . US Geological Survey, Washington, Water Resources Research , 41 (6). DC. Rodriguez-Iturbe, I. (2000). Ecohydrology: A hydrologic IWMI (2001). Water for rural development: draft background perspective of climate-soil-vegetation dynamics. Water paper on water for rural development prepared for the World Resources Research , 36 (1), 3–9. Bank . International Water Management Institute, Siebert, J., and Purkey, D. (2007). WEAP: water evaluation Colombo, Sri Lanka. and planning system—users guide . Stockholm Environment Kent, K. (1972). Section 4: Hydrology. In National Institute, Sommerville, MA. Engineering Handbook. United States Department of Slack, J. R., Lumb, A. M., and Landwehr, J. M. (1993). Agriculture, Washington, DC. Hydro-Climatic Data Network (HCDN): Streamf ow data Leavesley, G. H., Lichty, R. W., Troutman, B. M., et al . set, 1874–1988. US Geological Survey Water-Resources (1983). Precipitation-runoff modeling system: user’s manual . Investigation Report 93-4076 (CD). US Geological Survey, Washington, DC. Steenhuis, T. S., Winchell, M., Rossing, J., et al . (1995). SCS Lee, K. N. (1999). Appraising adaptive management. runoff equation revisited for variable-source runoff Conservation Ecology , 3 (2), article 3. areas. Journal of Irrigation and Drainage Engineering , Loucks, D. P., and Van Beek, E. (2005). Water resources sys- 121 (3), 234–8. tems planning and management: an introduction to methods, Tallis, H., Kareiva, P., Marvier, M., et al. (2008). An ecosys- models and applications. Studies and reports in hydrol- tem services framework to support both practical con- ogy. UNESCO, Paris. servation and economic development. Proceedings of the Lyimo, B. M. (2005). Energy and sustainable development in National Academy of Sciences , 105 (28), 9457–64. Tanzania , Helio International, Paris. Winrock International (2004). Financial incentives to com- Lyon, S. W., Gerard-Marcant, P., Walter, M. T., et al . (2004). munities for stewardship of environmental resources . Using a topographic index to distribute variable source Feasibility study: LAG-A-00-99-00037-00. USAID, area runoff predicted with the SCS–Curve Number Washington, DC. equation. Hydrology Proceedings. 18 (15), 2757–71. Winter, T. C., Harvey, J. W., Frankey, O. L., et al . (1998). McMahon, T. A., and Mein, R. G. (1986). River and reservoir Ground water and surface water a single resource. US yield . Water Resources Publications, Littleton, CO. Geological Survey Curcular 1139 . US Government Printing McMahon, T. A., Pegram, G. G. S., Vogel, R. M., et al . (2007). Off ce, Denver, CO. Review of Gould-Dincer reservoir storage-yield-relia- Yew, D. T., Dlamini, E. M., and Biftu, G. F. (1997). Effects of bility estimates. Advances in Water Resources , 30 , model complexity and structure, data quality, and objec- 1873–82. tive functions on hydrologic modeling. Journal of Milly, P. C. D. (1994). Climate, soil water storage, and the Hydrology , 192 , 81–103. average annual water balance. Water Resources Research , Zektser, I. S., and Loaiciga, H. A. (1993). Groundwater 3 (7), 2143–56. f uxes in the global hydrologic cycle: past, present and Nearing, M. A., Ascough, L. D., and Chaves, H. M. L. future. Journal of Hydrology , 144 , 405–27. (1989). WEPP model sensitivity analysis. In L. J. Lane Zhang, L., Dawes, W. R., and Walker, G. R. (2001). Response and M. A. Nearing, Eds., USDA-Water Erosion Prediction of mean annual evapotranspiration to vegetation Project: Hillside prof le model documentation, NSERL Report changes at catchment scale. Water Resources Research , 37 , No. 2 . USDA-ARS-NSERL, West Lafayette, IN. 701–8. CHAPTER 5 Valuing land cover impact on storm peak mitigation

Driss Ennaanay, Marc Conte, Kenneth Brooks, John Nieber, Manu Sharma, Stacie Wolny, and Guillermo Mendoza

5.1 Introduction permeable soils often reduce runoff as the result of enhanced soil infiltration and soil water stor- Images of floods displacing or even killing peo- age capacity. Conversion of forests and wetlands ple provide a constant reminder of the power of to agricultural or developed land covers will nature and human vulnerability to natural disas- tend to increase the volume of runoff and the ters. Although storms and storm events are flooding associated with storm events for highly unpredictable, it is possible to use hydro- medium and small return period events logical models to predict the magnitude of a par- ( Ennaanay 2006 ). However, forests have limited ticular flood, given information on the local ability to mitigate flooding associated with large geology, soil properties, vegetation, and manage- return period storm events because enhanced ment practices. We have developed approaches soil infiltration only captures a small fraction of for quantifying the link between changes in land total precipitation depth for such storms (FAO use and land cover (LULC), and flood risk. In and CIFOR 2005 ). We develop models in this flood management, risk has three ingredients: chapter that can help decision-makers take the hydrological response to a storm, the possi- advantage of nature-based mitigation of floods ble failure of flood protection infrastructure and storm damage from medium and small (such as a levee breaking), and the value of what return period events to avoid unnecessary flood might be destroyed by a flood. We focus on the risk due to poor land management. hydrology and economic value of what may be This chapter presents two different types of mod- destroyed, leaving structural integrity to be els for quantifying the impact of LULC on storm addressed by civil engineers. Given a well-de- outcomes. The data-sparse tier 1 model quantif es fined storm, we estimate the severity of flooding the reduction in storm peak volume due to LULC in terms of water volumes and flow rates, and relative to bare soil on a parcel-by-parcel basis and corresponding damages from the storm. values this reduction based on each parcel’s relative In general, the combination of meteorological contribution to mitigation. In the present formula- (e.g., rainfall intensity, extent and duration of the tion the tier 1 model is not set up to predict the event) and geophysical (e.g., basin size, basin extent of downstream f ooded area associated with geomorphology, soil characteristics, and land a storm peak. The more robust, data-intensive tier 2 use) characteristics are the main factors influenc- model provides probabilistic output for f ood mag- ing major flooding following large rainfall events nitudes as affected by incremental changes in the ( Hamilton and King 1983 ; Kattelmann 1987 ; landscape mosaic and quantif es the incremental B r u i j n z e e l 1 9 9 0 , 2 0 0 4 ) . I n s o m e s i t u a t i o n s , n a t u - changes in risks associated with a specif c f ood vol- ral landscapes and vegetation can offer storm ume, where risk is associated with a cost. In tier 2, peak mitigation. For example, forests and deep the extent of f ooded area is determined using the

73 74 VALUING LAND COVER IMPACT ON STORM PEAK MITIGATION

Hydrologic Engineering Centers River Analysis (HRUs), which are homogeneous with regard to System software (HEC-RAS) and streamf ow time LULC, soil, and slope. It should be noted that while series from the Precipitation Runoff Modeling the LULC category is a key determinant of an area’s System (PRMS) model. ability to intercept rainfall, the equations presented below will not include direct references to LULC categories, as they are captured by the parcel and 5.1.1 Storm peak mitigation modeling theory HRU indices. It should also be noted that while the For small to medium storms vegetation may analyses in tiers 1 and 2 occur at the parcel and retain water as it falls and f ows through the land- HRU level, respectively, the model results can be scape (through canopy interception, enhanced reported at other scales more relevant to manage- inf ltration, soil water storage) and thus reduce ment decisions such as individually owned parcels peak f ow. In a modeling study using the or counties. Hydrologic Simulation Program Fortran (HSPF) The key parameters linking LULC to storm peak in the Cottonwood River watershed within the mitigation are canopy interception, soil inf ltra- Minnesota River Basin, Ennaanay (2006 ) showed tion, LULC type, soil water storage, and land-use that conversion of different percent acreage (60, positioning on the landscape. Peak f ows will 75, and 86% of watershed area) of annual crop- increase as soil inf ltration, interception by canopy, ping systems (soy and corn) to perennial vegeta- or soil water storage is reduced. However, if the tion over a 50-year simulation period showed a magnitude of storm depth becomes large, the decrease in annual instantaneous peak f ows for impact that soil and plants have on storm f ow small event storms, but not for larger event peaks is small relative to water inputs ( Bruijnzeel storms. Similarly, studies from small paired 1 9 9 0 ; B r o o k s et al. 2003 ; Ennaanay 2006 ) and thus experimental basins showed that clear-cutting of reduced value in terms of f ood risk reduction. and road building increased only some peak The impacts of land use on f ooding also depend storm discharges (Wright et al. 1990 ). Indeed, in on the size of the area being examined. In particu- the Pacif c Northwest, increases in the peaks were lar, land-use impacts on f ooding are most evident greater for small early wet season storms but in watersheds less than 1000 km2 ( Kiersch 2001 ) there was no signif cant increase in peak f ows for because the sheer length of stream channels and the largest storms ( Rothacher 1970). extent of f oodplains in larger basins provide storm Wetlands in both up- and downstream areas, peak mitigation, swamping the signal of land use. and floodplains have a significant role in miti- For example, forest harvesting has produced gating floods and storm peaks. Both land-use detectable changes in peak discharges in basins up types have storage capacities higher than many to 600 km 2 in size. Increases are a result of changes other LULC types. Ennaanay (2006 ) showed that in f ow routing (due to roads) rather than to mere conversion of 27% of the annual cropping sys- changes in water storage due to vegetation removal tems to wetlands could significantly reduce peak in small basins. flows for small and moderate storm events. Wetlands not only reduce the peak flows but also significantly delay time to peak flows, and alter 5.2. Tier 1 biophysical model the inflow–discharge relationship and rough- 5.2.1 Modeling storm mitigation properties ness. Floodplains have impacts similar to those of wetlands on flood mitigation and storm peak The tier 1 model for storm peak mitigation focuses attenuation. on a storm event of a specif c size def ned by the The spatial resolution of our models differs user. Our approach estimates the impact of land use between tier 1 and tier 2. In tier 1, the analysis takes on f ood mitigation at the parcel level by determin- place at the parcel level, where parcel size is def ned ing the amount of on-parcel storm runoff retained by the spatial resolution of the input data. In tier 2, by each parcel following a rainfall event. We use the analysis is based on hydrologic response units GIS capabilities to generate a synthetic hydrograph TIER 1 BIOPHYSICAL MODEL 75 for the def ned size of storm ( Martinez et al. 2002 ; In order to evaluate the damage-mitigating prop- Melesse et al. 2003 ) ( Figure 5.1 a). To do this, we esti- erties of a landscape we apply the tier 1 model for mate storm runoff volumes for each parcel given a storm peak mitigation for a known f ood return LULC using the SCS–curve number (CN) method period, and the expected damages of such a storm. (Mockus 1972) and keep track of potential travel Prior to applying the model in the watershed of times from parcel to the watershed point of drain- interest, one must use regional rainfall data or rain- age. The synthetic hydrograph is formed by aggre- fall-runoff models coupled to f ood routing models gating runoff of parcels with similar travel time to characterize the storm event Ps related to f ood class. Factors affecting travel time in the model event s of probability π s. This tier 1 model assumes include LULC surface roughness and slope. By that the storm rainfall depth is constant in time dur- identifying landscape units with equal travel times ing the storm event and spatially uniform over the to the watershed outlet and summing the storm watershed of interest. water that reaches the stream from these units, we The tier 1 storm peak mitigation model calcu- create a hydrograph of a uniform storm depth, and lates the direct runoff generated by each parcel on associated duration. For model output to be useful the landscape using the SCS-CN equations for the for land managers, we generate a map of landscape user-def ned storm event ( Kent 1972 ). The SCS-CN contributions to these peak f ows. method is a simple, widely used, and eff cient The tier 1 storm peak model runs for a single method for determining the approximate amount specif ed storm event and predicts the magnitude of direct runoff from a rainfall event within any and the timing of the peak f ow. The tier 1 storm particular parcel. We use this method to compute peak is mainly based on the SCS-CN method, which the rainfall excess as the remainder of precipita- is known to work well for the design of culverts and tion after on-parcel inf ltration loss. Direct runoff civil works storm f ow infrastructure. Admittedly is generated by a wide variety of surface and sub- this is a very simple approach, but rainfall-runoff surface f ow processes, of which the most relevant modeling always demands balancing model com- ones are the Hortonian overland f ow, saturation plexity versus available data. Interestingly, it has overland f ow, shallow sub-surface f ow, and been found that more complex models are not nec- through-f ow ( Ponce and Hawkins 1996 ). The essarily more accurate than their simpler alterna- Hortonian overland f ow occurs when rainfall tives (Branson et al. 1962 , 1981 ; Loague and Freeze exceeds inf ltration capacity. It is a characteristic of 1985 ). dry to semi-dry regions and areas where vegeta-

Storm depth (b) (a) 3000

2500

2000 Lag time 1500 Time of 1000 Flow rate concentration 500 Storm Runoff Volume (m3) Volume Storm Runoff Base flow 0 0 12345678910 Time Classes Time to Peak

Figure 5.1 (a) Storm hydrograph terminology and (b) synthetic storm hydrograph generated by the storm peak mitigation model. 76 VALUING LAND COVER IMPACT ON STORM PEAK MITIGATION tion is sparse and the soil surface is highly dis- should be most wary of heterogeneous landscapes, turbed. The saturation overland f ow occurs after such as heavily urbanizing encroachments into the soil prof le has become saturated because of forests. either high antecedent soil moisture or high rain- The SCS-CN equation determines storm direct fall depth that f lls in the soil prof le; this mecha- runoff depth at parcel x for a f ood event s that nism is referred to as the Dunne mechanism. occurs with probability π s generated by rainfall

Shallow sub-surface f ow describes the process depth P s as that takes place when water f ows downslope in the shallow soil prof le quickly and contributes to 25400 Sxj = − 254 storm f ow. Kirkby and Chorley ( 1967 ) show that CN′′xj the shallow sub-surface f ow can be similar to the ⎧QPSsxj= 0 if s ≤ 0.2 xj , (5.1) through-f ow that occurs in heavily vegetated ⎪ 2 PS− 0.2 landscapes with thick soil covers and less perme- ⎨ ()sxj ⎪QPSsxj=> if s 0.2 xj able soil prof les atop impermeable bedrock ⎩⎪ PSsxj+ 0.8 (Kirkby and Chorley 1967 ). In our approach we use the curve number to quantify runoff under the where CN′′xj is the medium soil moisture condition assumption that the storm event occurs uniformly curve number adjusted for slope, 0.2 S x i s t h e i n i t i a l throughout the watershed and that Hortonian run- abstraction, which accounts for the amount of pre- off is the dominant process that generates the cipitation occurring before runoff, or the rainfall storm peak. Nonetheless, the CN method can be interception by vegetation. The value has been set used in landscapes dominated by saturation over- to 0.2 S x through developmental history and docu- land f ow as demonstrated by Boughton (1987 ), mentation; however, Hawkins ( 1979 ) showed that

S t e e n h u i s et al. ( 1 9 9 5 ) , a n d L y o n et al. ( 2004 ), but using 0.2 S x did not result in good runoff prediction would require some adjustments in how we unless Sx was dependent on rainfall amounts. Q sx is develop the synthetic hydrograph. Although ini- the direct runoff or quick-f ow that is potentially tially designed for watershed catchment runoff generated by P s at x . estimates, a distributed CN method has been The SCS-CN method uses the CN values that applied effectively to large parcels (from 900 m2 up were developed and assumed to be appropriate for to several hectares) def ned as HRUs in the Soil slopes of 5%. Our model uses the CN adjustment and Water Assessment Tool (SWAT) model ( Arnold recommended by Williams ( 1995 ) to slopes differ- et al. 1 9 9 8 , a n d s e e H a w k i n s et al. 2009 ). There is no ent than 5%, published guidance on an acceptable lower limit to parcel size for application of the CN approach. ⎛⎞100 − CNxj −13.86θ CN′′ = *(1−⋅ 2 exx )+ CN , (5.2) The parcel size in our model is def ned by the xj⎝⎠⎜⎟3 xj user and can be as small as is commensurate with input data. A common input data resolution is 900 w h e r e CN xj is the CN value associated with LULC m 2 , corresponding with the resolution of widely j i n x applied for soil conditions of medium wet- available digital elevation models. Our approach of ness type II (USDA 1986) with a 5% slope. applying the CN model on a parcel-by-parcel basis However, the user can specify one of three wet- that is then aggregated (distributed approach) ness conditions (dry, medium, or wet): in temper- rather than on a watershed catchment (lumped ate conditions, generally forested lands are drier approach) is not standard. We argue that runoff dif- than annual crops after long dry periods, and qx is ferences between distributed and lumped the parcel’s slope. approaches are minimized in evaluations of mid- to Once direct runoff is calculated, we estimate larger sized storm events since the curve number travel time of that water to the point of interest. becomes more linear at greater volumes. In addi- When def ning the drainage area to evaluate the tion, since differences are smallest in homogeneous storm-mitigating properties of the landscape, it is landscapes, such as natural landscapes, the user important to exclude parcels that drain into inter- TIER 1 BIOPHYSICAL MODEL 77 mediate f ood control reservoirs. Thus, we def ne the storm runoff generated by the storm event that

XD as the set of all parcels that f ow into point of reaches f ood risk point D arises from parcels in interest D that are not routed through a f ood con- time class 3. In effect, our tier 1 model develops a trol reservoir. The following equations provide a lumped synthetic storm runoff hydrograph to iden- total time of travel for excess rainfall originating on tify the sources of storm peaks in the watershed. a parcel in the landscape to a drainage point of The synthetic hydrograph allows the model to interest, D , select parcels that contribute to mitigating the storm νθ= 1 . % peak. The parcels, x', most likely to contribute to xxc x mitigating the storm peak are those that lie between y T = x , (5.3) the peak and the demand point D . XD ∑ ν ∀∈xXD x The parcels x' are found by determining the peak * QiD* of the synthetic hydrograph. Q iD , def ned as

where x indexes parcels in the landscape; v x is an aggregated runoff at each isochrones i, is def ned % estimate for overland f ow velocity; q x is the mean as percent slope of parcel x ; c 1/cx is a roughness coef- f cient for each LULC type based on the National xii'∈≤τ iD′ ' * * Engineering Handbook ( Kent 1972 ; ASCE 1996) that QQiD* ≥∀ iD i , (5.5)

QQiD= sxj relates slope and surface vegetation to velocities; y x ∑ is the distance travelled on parcel x equal to the x∈τiD width of the parcel if f ow direction is north, south, where, i* is the isochrone number that corresponds east, or west and equal to the hypotenuse of the par- to the peak of the hydrograph and QiD** contains the cel if f ow direction is otherwise; and T represents highest volume of runoff among all isochrones. Q XD iD the potential travel time that storm runoff from paris the parcel runoff summed for each isochrone i . x ' cel x is routed to point of interest D . There may be are all the parcels that lie on the f ow path between several drainage points of interest that might denote the point of demand D and potential parcels that an area of high importance for f ood mitigation. If are likely in synchronicity with the storm peak. so, the model can be applied iteratively for each point of interest. 5.2.2 Modeling the landscape benef ts of Once travel times are calculated for each parcel, storm peak mitigation isochrones (sets of parcels with the same travel time class to the point of interest) are def ned. Let us The service we want to represent on each parcel x is introduce τiD to represent isochrone i, where τiD is the reduction in storm f ow volume provided by def ned as follows for demand point D , vegetation. The tier 1 model does not account for inf ltration or storage of storm f ow reaching a given iI= 1,.., parcel x from upslope. This means we give a con- (5.4) ⎡⎤(iT−⋅1) xD→ max iT⋅ xD→ max servative estimate of the minimum amount of storm τ iD ∈⎢⎥,, ⎣⎦II peak f ow mitigation provided by each parcel. The model calculates mitigated runoff by each parcel on

where τ iD is a set of parcels x with a similar class of the landscape and the contribution of each parcel to travel time to point D ; D is the f ood risk reduction the storm peak at the point of interest. demand point and is the reference for estimates of The amount of storm peak mitigation associated f ow travel times; i is an isochrone identif er; I is the with extant vegetation, here called direct mitiga- number of isochrones (time classes) in the analysis tion, DMsx , is determined by the runoff depth at that is input by the user; and TxD→ max is the maxi- each parcel x retained by vegetation on that mum time of travel to the outlet. For example, parcel, Figure 5.1b depicts how the tier 1 storm peak mitigation model aggregates storm runoff of parcels ⎧PQssx− when xx= ' DMsx = ⎨ , (5.6) within the same time class. In this example, most of ⎩0 otherwise 78 VALUING LAND COVER IMPACT ON STORM PEAK MITIGATION

where P s is the storm depth and Q sx is the direct tier 1 model for a 30-year return period storm event storm runoff at parcel x for storm s, as def ned in of 150 mm. The results show areas where storm Eq. (5.1). It is important to note that as a storm event, f ow is being generated ( Figure 5.2a ). One can also

P s , gets higher such as in 50-year return period identify separate travel time classes, and see which storms or greater, the amount of potential storm class contributes the most f ow to the peak f ood mitigation at x , DM sx , tends to become smaller than (Figure 5.2b). The darker zones contain parcels that the storm depth. In other words, large storms satu- contribute to the peak volume while the lighter are rate the soil and make inf ltration and storm peak arriving either early or later to the watershed outlet, mitigation negligible relative to the total storm which is the point of interest. The tier 1 model gen- depth. This is consistent with earlier assertions and erates output at the parcel level; however, a user can observations that LULC properties to mitigate aggregate these outputs at larger scale to respond to storm peak is reduced for larger storms. specif c needs (e.g., values could be aggregated to a We estimate the marginal storm runoff mitigation map of individually owned parcels). Finally, we provided by land cover, ∆xsD, as a function of the show the expected pattern of storm peak mitigation marginal change in runoff with respect to total run- that occurred under this specif c storm and this spe- off for peak-contributing parcels, cif c LULC scenario (Figure 5.3 ). These maps can help managers interested in stabilizing or improving natural f ood control to identify two important ⎧ DMxjs − DM'xs ⎪ ; xx = ' , (5.7) parts of the landscape; (1) areas to protect because ∆xjsD = ⎨ ()DMxjs − DM'xs ∑ ∀∈xX of their current high contribution to the reduction of ⎪0 ; o t h e r w i s e ⎩ storm peak f ows, and (2) areas that currently contribute high f ow to the storm peak itself. Managers where ∆ is the parcel x ’s contribution in the xjsD may focus restoration or other improved manage- overall storm peak attenuation at the point of inter- ment practices in the latter parts of the landscape est D under storm s, and DM and DM ' are the xjs xs to help reduce f ood risk and damage (see also mitigation by parcel x for the storm s with the cur- Box 5.1 ). rent LULC j under the current scenario and a bare Since antecedent soil moisture is one crucial ele- soil scenario, respectively. One can use Eq. (5.7) to ment that def nes how much runoff is generated fol- map relative scores of storm peak mitigation as a lowing a rainfall event, the user may need to modify means of identifying those watersheds with the the condition of expected antecedent soil moisture. highest priority for management. However, this As default we use medium antecedent soil moisture method only counts f ood mitigation by parcels that CN for given LULC. However, this can be adjusted contribute to the peak f ow. We do this because upward or downward according to what is known f ooding damage occurs before and during the peak about soil moisture conditions when the storms of f ood. Waters arriving after peak f ood seldom cause interest arrive. additional damage. Given this timing–damage relationship, we do not want to assign social value to mitigation of f ows that do not cause damage to 5.3 Tier 1 valuation humans. Again, this is a conservative estimate of Storm peaks with longer return periods (i.e., lower the service provided. probabilities of occurrence) are associated with larger expected areas of f ooding. To mitigate the risk of f ooding, society can (1) better manage the Example 1: Determining cell runoff, sources of natural landscape to reduce the volume of water the storm peak, and storm peak mitigation coming out of each parcel and to delay this water as For illustrative purposes, we model a watershed much as possible so to spread the volume over a located in South-West Tanzania that f ows into the longer time period, (2) invest in man-made infra- town of Ifakara. The watershed has a drainage area structure such as levees and reservoirs to stop and of 32 km 2 with a diverse land cover. We applied the store f ood waters, and (3) manage people’s behav- TIER 1 VALUATION 79

(a) Flow per Cell (m^3)

11.2

Ifakara 0

0 25 50 Kilometers

(b) Time Class Flow (m^3) 4523844

Ifakara 1092377

02550 Kilometers

Figure 5.2 Storm volume upstream of the city of Ifakara, Tanzania. Areas with high CN, bare soil, and urban areas, presented by dark color, generate high volumes after the modeled storm event (a). Time classes of the storm hydrograph (b) show areas of the landscape that yield water that arrives in Ifakara during the storm peak f ow (dark zones) or before or after the peak f ow (light zones). 80 VALUING LAND COVER IMPACT ON STORM PEAK MITIGATION

Runoff Mitigation (mm)

150

Ifakara 29.3

02550 Kilometers

Figure 5.3 Storm peak f ow mitigation map. The land uses in this watershed have the capability to reduce and attenuate the storm volume from 29.3 to 150 mm for a storm event of 150 mm and antecedent moisture content (AMC) II.

Box 5.1 Integrated f ood risk management: gaining ecosystem services and increasing revenue

David Harrison frequency to one in 40, the annual costs may be reduced by one-fourth. However, the exposure to serious human It has become commonplace around the world to plan and disaster—loss of life—is potentially increased as society construct dams for multiple purposes. If a dam is primarily relaxes vigilance with a false sense of security and neglects conceived for hydroelectric generation, for example, it has emergency preparedness. seemed opportune to include other purposes, most commonly M o r e o v e r , f ood control imposed on a hydropower f ood control. After all, if a large public works project is being reservoir will generally cause substantial reduction of its built, it is only logical that it should provide as many benef ts revenue-generating potential. Full reservoirs produce more as possible. However, this simple proposition has serious energy than reservoirs held partly empty, which is especially downsides and has often led to perverse effects. important in today’s energy situation. Making things worse, Flood control operations of a dam are achieved by the trade-off of hydropower for f ood control often occurs lowering the reservoir water level during seasons of higher at inopportune times. In many monsoonal systems, the f ood risk to maintain space to receive and hold f ood waters f ood control season coincides with the high energy for subsequent release at lower f ow rates. The goal is to demand season. The time that reservoir levels are lowered reduce the peak of the f ood, or reduce the frequency of a in anticipation of monsoonal f oods is the very time of f ood of a given peak. Reducing the frequency of a f ood highest energy demand in the system—the hot summer may indeed produce calculable economic benef ts. Water season with high demand for air conditioning and industrial damage to property and disruption of economic activity is cooling. spread over more years and thus the annual cost is reduced. There is an alternative perspective in providing for However, small increments of f ood control can produce multiple benef ts from hydropower projects. Suppose the unintended consequences. For example, if the risk of a f ood control reservoir function were shifted out of the certain area of f oodplain is reduced from a one in 10-year reservoirs, thereby allowing increased energy generation. TIER 1 VALUATION 81

This is feasible if f ood risk management were to be 3. Use funds for regular assessment of the condition of accomplished by management in the f ood plain itself. f ood plain infrastructure—inspection and maintenance of Increased hydropower revenues deriving from fuller those levees and detention facilities—to reduce risk of reservoirs could be directed to a specif c fund to provide infrastructure failure. revenues for this management. 4. Develop new f nancial instruments for f ood risk The management of f ood risks in the f ood plain can be coverage. Reinvent f ood insurance applications to be based accomplished through the following measures: on incremental hydropower revenues. Recognize that inevitably water will occupy the land at some times, but that 1. Develop and maintain a comprehensive and effective productive uses of the land will otherwise continue. emergency preparedness plan—early warning, orderly evacuation, equipped refuge locations, and orderly system The fundamental idea is that integrated planning for f ood for reoccupation and recuperation of property. risk management, hydropower production, and ecosystem 2. Restore f ood plain ecosystems that provide not only protection has much to offer over simply imposing multiple ecological values but also f ood attenuation values. purposes on planned infrastructure.

ior in f ood prone areas—i.e., f ood plain manage- 5.3.2 Determining D ment, through Federal Flood insurance in the ks USA—limiting development in f ood prone areas is Flooding can affect crops, infrastructure, and the often the most feasible and economically viable production of valuable ecosystem services. Under option. None of these approaches is foolproof, and certain conditions, the magnitude of f ood events is both have to be designed in reference to the particu- impacted by the pattern of land use in a watershed. lar severity of storm events. When severe storms The value of storm peak mitigation increases with occur, f ooding is likely to ensue, particularly if the the ability of a LULC scheme to reduce the peak mitigation efforts are focused on less extreme f ow after a signif cant rainfall event in the events. watershed. In this model, we determine each parcel’s contribution to a f ooding event at the watershed’s base. 5.3.1 Calculating f ood damage as a function Next, we identify each parcel’s contribution to the of storm peak and LULC pattern economic damages in the f ood area. Finally, we Our models are not designed to predict the f ooded determine the potential savings or additional losses area. This means that if one wants to calculate dam- in total f ood damage when a parcel changes ages they need to obtain f ood area maps. In the LULC.

USA, f ood footprint maps can be obtained from Let Vxt indicate the total economic value of parcel

Federal Emergency Management Agency (FEMA) x i n y e a r t. If possible, Vxt should include all market or f ood insurance companies. In the developing values and monetized ecosystem service values. world local knowledge ( Tran et al. 2009 ) can help The total present value of damage due to a f ood in piece together scant information. It may often be area k with storm peak s is given by necessary to develop f ood prints directly using river channel hydraulic software such as HEC-RAS T α VC D xh xh xT+1 , ks =+∑∑ ht− T+1 (5.8) (USACE, 2002; ESRI-HEC 2004) applied to each xksht∈ = (1+ r) (1+ r ) storm of interest. Once f oodmaps have been obtained, using economic valuation techniques, we where x Î ks indicates the parcels in k that are f ooded attach a damage value to each of the mapped f ood given storm peak s . T indicates the number of years events. To isolate the value of storm peak mitigation for which the value of parcel x cannot be fully real- provided by LULC, we must compare the damages ized due to storm peak s . αxt is an approximation of associated with events on bare soil with events on x's portion of the value damaged by the f ood in extant vegetation. each year (i.e., if a f ood covers land upon which a 82 VALUING LAND COVER IMPACT ON STORM PEAK MITIGATION

20-story skyscraper is built, the entire value of that model, the area of the landscape inundated by building may not be lost as a result of the storm storm s is not impacted by the LULC pattern. As a event, although the ability of workers to go to their result, we are unable to identify how the presence of off ce might be). C represents the costs that must be native vegetation mitigates the damages caused by incurred in order to return parcel x to its productive f ooding. We provide an estimate of the true value capacity (e.g., construction costs to rebuild struc- of this impact as tures). r represents the annual discount rate. While ω the linear damage function (area-discharge) might BD = (1− fs ) , (5.10) ks ks ω seem arbitrary, Dutta et al. ( 2003 ) specify depth– bs damage curves for urban damage estimation that where ω fs represents the volume of post-peak water are linear until 3 m of depth above the f oor level, delivered to the damage point during storm s when the depth at which the damage function plateaus at the landscape is covered by the current LULC pat-

60% of total structure value. tern and ωbs represents the volume of post-peak A s n o t e d i n M e r z et al. ( 2004 ), the majority of water delivered to the damage point when the land- f ood damage estimation methodologies are based scape is covered by bare soil. on direct tangible damage caused to structures in The model then aggregates sources of runoff from the f ood zone. However, the cost of inundating parcels into isochrones. Finally, we distribute B ks land may be greater than the structural damage across the parcels that are in isochrones that precede caused by the storm event. The true costs imposed the peak. These groups include all parcels that con- by the f ooding are the stream of prof t and serv- tribute water to the storm f ow that impacts area k . ice values foregone during the period of inunda- B ks is distributed among these parcels according to tion and drainage as well as any expenses that each parcel’s relative contribution of water volume. must be incurred to return the parcel to produc- Let B xks represent parcel x's share of Bks , where X tive use. BB= ks , x indexes the parcels contributing to ks∑ x=1 xks Information on the stream of expected pro f t the storm volume, and Xks is the set of all such par- and service values at the parcel level may be dif- cels. Then B xks can be def ned as f cult to gather. In many cases, especially if the f ood event in question affects urban or rural-res- BBxks = ks * ∆ xjsD , (5.11) idential areas, the value of real estate can proxy for the discounted stream of all future prof ts where D xks = 0 if x lies within a time class that pre- emanating from a parcel (Polasky et al. 2008 ). cedes the storm peak, or the storm return period is Note that such an approximation is only reason- greater than 100 years, assuming that at 100 years, able in areas with fully functioning land markets. LULC has no signif cant impact on reducing storm Because the value of real estate in functioning peak. markets ref ects the inf nite prof t stream associated with optimal use of a parcel, but a parcel will 5.3.4 Accounting for the distribution of storm only be impaired for a period of time (t t o T ) , V xt peak return periods could be set equal to Whether or not a particular return period storm will T RVxt (5.9) occur on a landscape in any given year (i.e., the ∑ ht− , ht= (1+ r) probability that a particular storm peak will occur on the landscape in any given year) is uncertain.

where RV xt is the annual rental value of real estate in The return period indicates on average whether a parcel x in year t . f ood of a particular size will occur at least once The value of storm peak mitigation is captured by during that time period. the difference in expected damages due to the pres- Accurate valuation of storm peak mitigation ence of vegetation in the landscape. In the tier 1 requires calculation of the expected damages TIER 1 VALUATION 83 avoided due to extant vegetation. To calculate Example 2: Economic valuation of the landscape expected damages, the user must determine the for its storm peak mitigation damages associated with several different return In our example catchment in Tanzania ( Figure 5.4 ), periods. To determine damages and f ooding extent we applied the tier 1 storm peak mitigation model for each it may be necessary to run a f ood routing for the watershed f owing into the city of Ifakara model, such as HEC-RAS, for different storm events for a 150-mm storm depth and hypothetical because f ood hazard maps are often tied to a few US$10 000 000.00 f ood damage. Damage values are specif c probability events—often targeting the distributed on the upstream parcels that have travel longer return periods (50–100) not affected consid- time within the time classes that are less than or equal erably by land use. to time to peak. The assignment of value is propor- Let π indicate the probability that a storm that ks tional to the amount of storm peak reduction caused will produce storm peak s at k will occur on the by the parcel’s vegetation, relative to bare, saturated landscape in any given year. Then, soil. The value of a parcel is a function of its capability to reduce the excess-rain volume since the travel DEDkksksks≡ [ ]= ∑π D , (5.12) s time is mainly a function of the distance to the outlet, slope, and a small impact of the parcel roughness where D = 0 for any s that has a return period of ks associated with the land-use type. This map shows 100 years or more. In order to identify the expected values ranging from US$0 to US$1022 per hectare, value of storm peak mitigation by vegetative cover representing the wide range of f ood mitigation pro- on parcel x , we can modify Eq. (5.12) above to vided by the wide range of land-use categories develop the expected value of mitigation on cell x present in this watershed. However, these results do with LULC j as not ref ect the full natural system characteristics such as storm depth variability within the watershed, and BEBD==[] *∆ . (5.13) xk xks k xjsD the storage capacity of different land-use categories.

Landscape Value ($) 92.8

Ifakara 0

02550 Kilometers

Figure 5.4 Economic valuation of the landscape for avoided f ood damages. 84 VALUING LAND COVER IMPACT ON STORM PEAK MITIGATION

5.4 Tier 2 supply and use model independently evaluate canopy interception, root- ing effects, soil litter, and many other functions 5.4.1 Peak f ow model and f ood analysis attributed to mitigating runoff. Moreover, the model We are in the process of developing a tier 2 storm will incorporate hydraulic routing impacts on miti- peak f ow mitigation model, which will be based on gating storm peaks, such that the user is informed assessing incremental changes in risk given changes about the relative impact of vegetation versus land- in land use. This model under development com- scape geomorphology and scale. bines several components of existing models in As in tier 1, the tier 2 model will generate maps of order to bring attention to the role of extant vegeta- a parcel’s contribution to each peak f ow of interest tion in regulating storm peak f ows. The tier 2 and the parcel’s mitigation of the storm that caused approach will use PRMS (Leavesley et al. 1983 ) or a that peak. Specif cally, we will assess the incremen- similar model to evaluate specif c characteristics of tal changes in risk given a change in landscape or landscape vegetation in mitigating storm peaks. climate using the f ood frequency analysis methods Currently, the CN approach lumps all known LULC by running different LULC scenarios with incre- inf ltration and storage functions into one number. mental storm intensities ( Figure 5.5 ). This f gure is In contrast, our tier 2 approach is being designed to known as the f ood frequency analysis curve. We

105 Land Use A - Climate A Land Use B - Climate A Land Use A - Climate B

104

103 Annual Peak Discharge Cubic Feet Per Second Annual Peak Discharge

102 99.5 98 95 90 80 70 50 30 20 10 5 2 1 0.5 0.2 Annual Exceedance Probability, Percent

1 2 10 20 50 100 200 500 Return period

Figure 5.5 Flood frequency analysis for several scenarios of land use and climate. TIER 2 VALUATION 85

1400

1200

1000

800

600

400 Flooded Area (Hectares) Area Flooded

200

0 0 1000 2000 3000 4000 5000 Peak Flow (m3/s)

Figure 5.6 Peak f ow–f ooded area relationship. The HEC-GeoRAS model runs for different peak f ows at steady state modeling or the whole time series at unsteady-dynamic state modeling to draw a relationship between peak f ows and f ooded area at each point of interest. This relationship will be used to determine the f ood damage every time there is a f ooding. However, studies show that the channel geomorphology is constantly changing such that the river cross sections will be different between separate f ood events. Changes in river cross section will change this relationship; therefore the modeler should include the new cross section or modify and adjust the existing one to match the stream reality.

will develop such curves using the simulated or damage from drowning or soil saturation. The observed annual peak f ows in a statistical method, modular f ow estimation component of tier 2 will such as that used in USGS-PeakFQ (Flynn et al. include a module that tracks the soil moisture in 2006 ) software, to construct frequency distributions every HRU. The module will contain a component for different recurrence intervals. Different land-use to def ne the duration of soil moisture greater than scenarios and/or different climate scenarios will be the f eld capacity and create a time series identify- used to generate different annual peak time series ing the duration of excess moisture conditions. that reveal the impacts of land-use changes and cli- This information will be coupled with the tier 2 mate changes on the peak f ow and extent of f ooded agricultural production model’s information on area. crop growth stage and crop f ood to estimate the Using the annual peak time series generated agricultural crop damage occurring after storm above, one can run f ood frequency analyses to events. determine the peak f ow associated with each return period. Then, for each return period, HEC-RAS can 5.5 Tier 2 valuation be used to generate f ooded area prof les at river cross sections of interest. Valuation will require the The economic valuation for tier 2 will be similar to input of an area-value index map that, when over- that in tier 1, except that increased spatial and tem- laid with the surface area prof les maps, will allow poral resolution will allow for the extent of f ooding estimation of approximate damages in the f ooded to vary across LULC scenarios. Given this enhanced area around the stream ( Figure 5.6 ). realism, we replace our proxy for the value of native Area-value index maps usually focus on dam- vegetation on the landscape as described in Eq. ages associated with loss of property and infra- (5.12) above, with the true value, structure. We will provide the additional capacity in our tier 2 model to estimate agricultural crop BDDxks = kbs− kfs , (5.14) 86 VALUING LAND COVER IMPACT ON STORM PEAK MITIGATION

where D kbs represents the damages when the land- authors noted that for larger design storms (> 50- to scape is covered by bare soil, and D kfs represents the 100-year return period) the difference in the runoff damages with the current LULC pattern. computed using composite and distributed CNs is minimal. 5.6 Limitations and next steps There is no explicit provision for the appropriate spatial extent at which to apply the CN approach. The tier 1 storm peak mitigation model uses the CN By design the method is assumed to apply to small model to only evaluate landscape parcels for their and mid-sized catchments ( Ponce and Hawkins potential to store water from precipitation and keep 1996 ). Simanton et al. ( 1973 , 1996 ) found that CN it from becoming direct runoff. Runoff from snow- varied inversely with drainage area and noted a CN melt, ice, sleet, or rain on frozen ground is not esti- decrease of 2.2 units/100 ha of drainage area ref ect- mated with this model because under these ing the substantial role of transmission losses. White conditions vegetation plays a negligible role in miti- (1986 ) showed the CN approach effectively pre- gating storm peaks. Our current tier 1 approach dicted stream f ow for a large (421 km 2 ) watershed also does not value the benef ts from vegetation to in eastern Pennsylvania. Several studies have also extend f ow routing times (i.e., expand the base of shown that rainfall heterogeneity within larger the synthetic hydrograph resulting in lower storm watersheds is an issue when applying the CN peak). This can be resolved in the next steps. method ( Van Mullem et al. 2002 ). Our model requires Additionally, our model does not include the in- one uniform storm depth where in reality storms stream channel and f ood plain attenuation due to can hit watersheds with different rainfall depths. friction and expansion–contraction energy losses in Furthermore, with the CN approach rainfall inten- routing. This means that f oodplain management, sity and duration are not considered, only total which is one of the most effective strategies for storm depth. However, Hawkins ( 1975 ) noted that reducing damages associated with storm peaks, is for a considerable range of precipitation values, not accounted for by our tier 1 model (it is address- accurate CNs are more important than accurate able with a tier 2 model). When running a tier 1 rainfall estimates. analysis that does not include the value of f ood- Given these limitations, one should not use out- plains, one is assessing a complementary strategy to comes from these tier 1 models to argue for the f ood plain management or restoration—a strategy replacement of a f ood reservoir. However, they can that places value on the benef ts that a forested be used to value portions of the landscape that are landscape might have to mitigate storm peaks. often neglected and that contribute to f ood mitiga- The main limitation of our approach is that we tion. We will continue to develop these models, and are applying a tool created for lumped watershed are working to incorporate a new approach where analysis at the parcel scale. Curve numbers are the CN is used to compute runoff from variably dependent on land cover, hydrologic position, soil saturated source areas (i.e., saturation overland type, and moisture content, each of which vary con- f ow process of runoff generation). This new siderably spatially. In most models, an area- approach began with Boughton (1987 ) and has been weighted average CN approach is used to assign a followed by Steenhuis et al. ( 1995 ), Lyon et al. ( 2004 ), single value for a region or for a small subgrouping and Nachabe ( 2006 ). The latter studies have shown of areas considered homogenous (SWAT, HEC-1). that by incorporating surface topographic charac- Recently, however, work has been completed to teristics into quantifying CNs for an area the accu- assess whether improvements can be attained from racy of runoff prediction improves signif cantly. using a more distributed CN approach. In compar- Further development of this new approach should ing the effects of composite versus distributed CNs improve the accuracy of the CN method for predic- on estimates of storm runoff depths, Grove et al. tion of runoff from agricultural, forested, and range- ( 1998 ) showed that distributed CNs provide closer land areas. estimates particularly for wide CN ranges, low CN We are also exploring the addition of a method values, and low precipitation depths. However, the for including inf ltration of runon from upslope LIMITATIONS AND NEXT STEPS 87 parcels. This would provide a more full accounting Dutta, D., Herath, S., and Musiake, K. (2003). A mathemat- of the role parcels play in drawing down f ood peak ical model for f ood loss estimation. Journal of Hydrology , f ows and give a more realistic (rather than conserv- 277 , 24–49. ative) estimate of the storm f ow peak mitigation Ennaanay, D. (2006). Impacts of land use changes on the hydrologic regime in the Minnesota River Basin benef ts provided by vegetation and soils. . PhD Thesis, Graduate School, University of Minnesota. Lastly, it should be noted that although f ooding Environmental Systems Research Institute Hydrologic has many adverse impacts, some benef cial impacts Engineering Center (ESRI-HEC). (2004). HEC GeoRAS do exist. Flood waters inundate f oodplains, leaving Tools Overview Manual . US Army Corps of Engineers, the soil moisture content and soil fertility high, Davis, CA. which can prove benef cial for agriculture, depend- FAO and CIFOR. (2005). Forests and f oods: Downing in ing on the crop cycle (World Bank 1990 ). In many f ction or thriving on facts? RAP Publication 2005/03. parts of the world, f oodwaters replenish ground- United Nations Food and Agricultural Organization, water aquifers, allowing them to fully recover. Regional Off ce for Asia and the Pacif c, Bangkok. These represent additional ecosystem services that Flynn, K. M., Kirby, W. H., and Hummel, P. R. (2006). will need to be represented with other models. It User’s manual for program peak FQ, annual f ood frequency analysis using bulletin 17B guidelines. may be important to consider these positive fea- US Geological Survey Techniques and Methods Book 4, Chapter B4. tures of f oods when comparing the cost-effective- Grove, M., Harbor, J., and Engel, B. (1998). Composite ver- ness of nature-based as opposed to concrete-based sus distributed curve numbers: Effects on estimates of approaches to mitigating f ood risk. storm runoff depths. Journal of the American Water Resources Association , 34 , 1015–23. References Hamilton, L. S. and King, P. N. (1983). Tropical forested watersheds: hydrologic and soils response to major uses or ASCE. (1996). Hydrology handbook . American Society of conversions . Westview Press, Boulder, CO. Civil Engineers, New York. Hawkins, R. H. (1975). “The importance of accurate curve Arnold, J. G., Srinivasan, R., Muttiah, R. S., et al . (1998). numbers in the evaluation of storm runoff.” Water Large-area hydrologic modeling and assessment: Part I. Resources Bull etin, 11 (5), 887–91. Model development. Journal of the American Water Hawkins, R. H. (1979). Runoff curve numbers from partial Resources Association , 34 , 73–89. area watersheds. Journal of the Irrigation and Drainage Boughton, W. C. (1987). Evaluating partial areas of water- Division , 105 , 375–89. shed runoff. Journal of Irrigation and Drainage Engineering , Hawkins, R. H., Ward, T. J., Woodward, D. E., et al . (2009). 113 , 356–66. Curve number hydrology: state of the practice . American Branson, F. A., Gifford, G. F., Renard, K. G., et al . (1981). Society of Civil Engineers, Reston, VA. Rangeland hydrology . Society for Range Management, Kattelmann, R. (1987). Uncertainty in assessing Himalayan Range Science Series No. 1. Kendall/Hunt, Dubuque, water resources. Mountain Research and Development , 7 , IA. 279–86. Branson, F. A., Miller, R. F., and Queen, I. S. (1962). Kent, K. (1972). Travel time, time of concentration, and lag. Effects of contour furrowing, grazing intensities, and In National Engineering Handbook; Section 4: Hydrology. soils on inf ltration rates, soil moisture and vegetation US Department of Agriculture, Washington, DC. near Fort Peck, Montana. Journal of Range Management , Kiersch, B. (2001). Land use impacts on water resources: a lit- 15 , 151–8. erature review. Food and Agriculture Organization of the Brooks, K. N., Folliott, P. F., Gregersen, H. M., et al . (2003). United Nations, Rome. Hydrology and the management of watersheds . 3rd edn. Kirkby, M. J. and Chorley, R. J. (1967). Through f ow, overland Iowa State Press, Ames. f ow and erosion. Hydrological Sciences Journal , 12 , 5–21. Bruijnzeel, L. A. (1990). Hydrology of moist tropical forests Leavesley, G. H., Lichty, R. W., Troutman, B. M., et al . and effects of conversion: a state of knowledge review . Humid (1983). Precipitation-Runoff Modeling System: User’s Tropics Programme, UNESCO International Hydro- Manual. US Geological Survey Water Resource logical Programme, UNESCO, Paris. Investigations Report 83-4238. Bruijnzeel, L. A. (2004). Hydrological functions of tropical Loague, K. M., and Freeze, R. A. (1985). A comparison of forests: not seeing the soil for the trees? Agriculture rainfall runoff modeling techniques on small upland Ecosystems and Environment , 104 , 185–228. catchments. Water Resources Research , 21 , 229–48. 88 VALUING LAND COVER IMPACT ON STORM PEAK MITIGATION

Lyon, S. W., Walter, M. T., Garard-Marchant, P., et al . (2004). Simanton, J. R., Hawkins, R. H., Mohseni-Saravi, M., et al . Using a topographic index to distribute variable source (1996). Runoff curve number variation with drainage area runoff predicted with the SCS–curve number equa- area, Walnut Gulch, Arizona. Transactions of the American tion, Hydrological Processes , 18 , 2757–71. Society of Agricultural Engineers , 39 , 1391–4. Martinez, V., Garcia, A. I., and Ayuga, F. (2002). Distributed Steenhuis, T. S., Winchell, M., Rossing, J., et al . (1995). SCS routing techniques developed on GIS for generating runoff equation revisited for variable-source runoff synthetic hydrographs. Transactions of the American areas. Journal of Irrigation and Drainage Engineering , 121 , Society of Agricultural Engineers , 45 , 1825–34. 234–8. Melesse, A. M., Graham, W. D., and Jordan, J. D. (2003). Tran, P., Shaw, R., Chantry, G., and Norton, J. (2009). GIS Spatially distributed watershed mapping and modeling: and local knowledge in disaster management: a case GIS based storm runoff response and hydrograph anal- study of f ood risk mapping in Viet Nam. Disasters , ysis. Part 2. Journal of Spatial Hydrology , 3 , 2–28. 33 (1), 152–69. Merz, B., Kreibich, H., Thieken, A., et al. (2004). Estimation United States Army Corps of Engineers (USACE). (2002). uncertainty of direct monetary f ood damage to build- HEC-RAS River Analysis System. US Army Corps of ings. Natural Hazards and Earth System Sciences , 4 , Engineers, Davis, CA. Available at http://furat.eng.uci. 153–63. edu/wsmodeling/Software/USACE_HEC/HEC- Mockus, V. (1972). Estimation of direct runoff from storm RAS/V4Beta/docs/HEC-RAS_Reference_Manual.pdf rainfall. In National Engineering Handbook; Section 4: United States Department of Agriculture (USDA). (1986). Hydrology . US Department of Agriculture, Washington, Urban Hydrology for Small Watersheds. S o i l C o n s e r - DC. vation Service, Engineering Division, Technical Release Nachabe, M. H. (2006). Equivalence between topmodel 55 (TR-55). and the NRCS curve number method in predicting vari- Van Mullem, J. A., Woodward, D. E., Hawkins, R. H., et al . able runoff source areas. Journal of the American Water (2002). Runoff Curve Number Method: Beyond the Handbook . Resources Association , 42 (1), 225–35. US Geological Survey Advisory Committee on Water Polasky, S., Nelson, E., Camm, J., et al . (2008). Where to put Information—Second Federal Interagency Hydrologic things? Spatial land management to sustain biodiver- Modeling Conference. July 28—August 1, Las Vegas, NV. sity and economic returns. Biological Conservation , 141 , White, D. (1986). Synoptic-Scale Assessment of Surface 1505–24. Runoff: An Analysis of the Soil Conservation Service Ponce, V. M., and Hawkins, R. H. (1996). Runoff curve Runoff Curve Number. Proceedings of the Annual number: Has it reached maturity? Journal of Hydrologic Pittsburgh Conference , 17 , 159–63. Engineering , 1 , 11–18. Williams, J. R. (1995). The EPIC model. In V. P. Singh, Ed., Rothacher, J. (1970). Increases in water yield following Computer Models of Watershed Hydrology, pp. 909–1000. clear-cut logging in the Pacif c Northwest. Water Water Resources Publications, Highlands Ranch, CO. Resources Research , 6 , 653–8. World Bank. (1990). Flood Control in Bangladesh: A Plan for Simanton, J. R., Renard, K. G., and Sutter, N. G. (1973). Action . World Bank, Washington, DC. Procedures for identifying parameters affecting storm Wright, K.A., Sendek, K.H., Rice, R.M., and Thomas, R.B. runoff volumes in a semiarid environment. USDA-ARS (1990). Logging effects on streamf ow: Storm runoff at Agricultural Reviews and Manuals ARM-W-1. USDA- Casper Creek in Northwestern California. Water ARS, Washington, DC. Resources Research, 26 , 1657–67.

CHAPTER 6 Retention of nutrients and sediment by vegetation

Marc Conte, Driss Ennaanay, Guillermo Mendoza, Michael Todd Walter, Stacie Wolny, David Freyberg, Erik Nelson, and Luis Solorzano

6.1 Introduction Also, vegetated areas with low slopes tend to provide higher levels of nutrient and sediment retention As water f ows across the land, its physical and bio- than steeply-sloped areas. Pulling these ideas chemical characteristics are shaped by human activ- together, we present two methods of landscape anal- ities and the vegetative cover on the landscape. Of ysis regarding sediment and nutrient retention. The particular importance to society is the transport of simple, tier 1 models are based on average annual nutrients and sediment from upstream locations to precipitation levels as well as data related to land- downstream water bodies. For example, the trans- scape topography and nutrient loading. The outputs port of nitrogen from upstream agricultural lands is from these models include annual average maps of partially responsible for the extreme eutrophication nutrient export and retention, sediment export and of coastal waters that has led to the dead zone in the retention, and the value of different parcels of land Gulf of Mexico. From the perspective of sediment based on their nutrient and sediment retention. pollution, more than 0.5% of global reservoir stor- Because the rate of water f owing across the land- age is lost annually due to sedimentation (White scape inf uences nutrient and sediment retention, 2001) at possible costs of US$13 billion per year and can vary enormously from day to day, annual (Palmieri et al. 2003). The degree to which surface water volumes may not capture the differential f ows remove and deliver nutrients and sediment retention properties of assorted parcels of land. In from their sources to downstream locations is highly particular, analysis conducted at a f ne temporal dependent on the pattern of land use and land cover resolution (e.g., daily) will be able to provide more (LULC) on the landscape. By retaining portions of accurate estimates of nutrient and sediment reten- the nutrients and sediment released and trans- tion. The tier 2 models capture these temporal ported by surface f ows, vegetation can help to miti- dynamics by using the Agriculture Non-Point gate damages downstream. Source (AnnGNPS) model to estimate nutrient The quantities of nutrients and sediment retained export and retention and the Precipitation Runoff by vegetation are largely functions of slope, vegeta- Modeling System (PRMS) to predict the amount of tion type, volume of water f owing across the land, sediment eroded and delivered to focal water bod- and the vegetation’s location in the watershed. Lands ies. Both models can provide daily estimates. While with intact natural vegetation will tend to be net models in tier 2 provide more realistic estimates of retainers of both nutrients and sediment, whereas nutrient and sediment retention in the landscape, lands used intensively for agricultural production they require data that are costly to collect and a will tend to be sources of both nutrients and sedi- detailed understanding of local hydrology. ment. The likelihood of retaining nutrients and sedi- The models here focus solely on the nutrients and ments increases as proximity to source areas increases. sediment that move with surface water f ows across

89 90 RETENTION OF NUTRIENTS AND SEDIMENT BY VEGETATION the landscape from non-point sources. There are to represent the lower bound of the social value of several other damaging non-point source particu- nutrient retention. For sediment retention, the social lates that move with surface water, including heavy value is approximated using the dredging costs that metals and pesticides; our modeling approach are avoided by having sediment-retaining vegetation could be adapted for these other pollutants as long on the landscape, which should be interpreted as a as the reactions of these pollutants with soil and the lower bound of the value of this ecosystem service. ambient environment could be described. In areas with more-detailed information, planners can apply our tier 2 models that extend the hydrologic principles of tier 1 to include the tempo- 6.1.1 Modeling f ows relevant to nutrient ral dynamics of hydrology and nutrient and and sediment retention sediment loading on the landscape. The setup of The tier 1 and tier 2 models that we develop rely on tier 2 models requires additional data and expertise the principle of saturation excess runoff to generate for calibration, verif cation, and validation. Our tier nutrient and sediment runoff estimates. These mod- 2 approach to nutrient retention has the capability els use a runoff index obtained from a topographic of distinguishing between inf ltration and satura- index that contains a permeability function. By com- tion excess runoff transportation pathways. It simu- bining the concept of hydraulic connectivity with lates the timing of nutrient application activities pixel-specif c runoff indices, these models allow us with respect to the generation of runoff. to estimate the amount of nutrient and sediment The tier 2 framework for sediment retention uses retention on individual pixels based on their loca- methods from PRMS to simulate erosion and sedi- tion in the watershed and their position along the ment processes. PRMS simulates sediment detach- f ow path to the downstream body of water. ment using a revised form of the universal soil loss evaporation (USLE) method (Leavesley et al. 1 9 8 3 ) . The detachment rate of sediment in PRMS is depend- 6.1.2 Implementing theory to model ent on rainfall intensity, geomorphology, and runoff ecosystem-service provision volumes (Hjelmfelt et al. 1975), and the movement of Many processes drive the impairment of water bod- sediment in the stream channel is coupled with the ies by nutrient and sediment loading. The magni- energy in simulated f ows. PRMS accounts for both tude and timing of precipitation events, irrigation sheetwash and streambank erosion, by incorporat- scheduling, stream and soil cycling, soil characteris- ing stream scouring and deposition processes, which tics, and landscape geomorphology and topography improves the model’s ability to assign value to sedi- are some of the variables that affect the impact of ment retention on vegetated parcels. nutrient pollutant loading and sediment transport to a water body. Detailed information about all of 6.2 Tier 1 biophysical models the above processes may not be available in all regions struggling to cope with the effects of nutri- The models presented in this chapter use LULC cat- ent and sediment deposition in water bodies. Our egory as a principle driver of the processes being tier 1 model of nutrient retention uses as key inputs modeled, including water yield, pollutant loading, a water yield index (as a proxy for contributing area, and sediment retention. The unit of analysis in these precipitation, soil type, and slope), LULC, export models is the parcel, whose size is based on the spa- coeff cients, and downslope retention. Our tier 1 tial resolution of the available data in the study area. model of sediment retention uses as key inputs The model assumes that each parcel is homogene- LULC, downslope retention, slope, rainfall erosivity, ous with respect to LULC, meaning that there is a and soil erodibility. The processes chosen as inputs single LULC category associated with each parcel. for our models tend to be associated with more eas- For this reason, the parcel index captures the impact ily-accessible data, making them more practical. of LULC on each of the processes being modeled, The economic valuation for nutrient retention is and we do not include LULC indices in the chapter based on the cost of water treatment, which is meant equations unless this assumption is relaxed. TIER 1 BIOPHYSICAL MODELS 91

6.2.1 Nutrient retention Table 6.1 Example phosphorus and nitrogen export coeff cients (Reckhow et al . 1980) There are two key components to nutrient retention: (1) the ability of vegetation and soils to avoid LULC Nitrogen export Phosphorus export coeff cient coeff cient initial nutrient loss (on-parcel retention) and (2) the (kg/ha/yr) (kg/ha/yr) ability of vegetation and soils to take up nutrients Feedlot or dairy 2900 220 exported to the parcel from upstream parcels (corss- Business 13.8 3 parcel retention). To capture these processes, we Soybeans 12.5 4.6 f rst need to know how much nutrient is exported Corn 11.1 2 from each parcel on the landscape. We do this using Cotton 10 4.3 export coeff cients derived from previously pub- Residential 7.5 1.2 lished f eld measures of how much nutrient leaves a Small grain 5.3 1.5 farmf eld under average conditions of nutrient (fer- Industrial 4.4 3.8 tilizer, plant-driven nitrogen f xation) inputs, f eld Idle 3.4 0.1 management, slope, rainfall, and soil type. We use Pasture 3.1 0.1 export coeff cients in our tier 1 approach because Forest 1.8 0.011 they are widely available and relatively easily applied around the globe. However, we cannot rep- model is the management of non-point sources of resent on-parcel retention because export coeff - nutrient pollutants, it is still critical to map point cients bundle the effects of vegetation and soils on sources using relevant lookup tables to help manag- nutrient retention with the effects of other factors. ers determine the impact of non-point source man- As such, our tier 1 model only accounts for the role agement with respect to eff uent sources. Table 6.1 of vegetation and soils in cross-parcel nutrient displays the USEPA export coeff cient values for retention. nitrogen and phosphorous for different LULC categories. The table illustrates that intensive manage- 6.2.1.1 Modeling non-point source nutrient loads ment is associated with dramatically more nitrogen The nutrient retention model is based on the theory and phosphorous loading than native forests; that the damages caused by excess nutrient applica- however, loading of these nutrients from forests is tion on runoff-generating surfaces (i.e., non-point non-zero. Both nitrogen and phosphorous are natu- sources) can be mitigated by intercepting vegetated rally present in aquatic systems. f lters (Baker et al. 2006; Schneiderman et al. 2007). T h e e x p o r t c o e f f cients, developed by Reckhow The f rst step in the tier 1 model is to characterize et al. (1980), are annual averages of pollutant load- the landscape based on nutrient pollutant loadings. ings derived from f eld studies that measured Pollutant loading at parcel x with a given LULC cat- export from representative, or average, agricul- egory, polx , is based on export coeff cients directly tural parcels in the USA. Since we want to apply derived by associating a specif c LULC category this model in other regions, we include a proxy with a lookup table of corresponding export values. factor that helps offset differences between the The US Environmental Protection Agency (USEPA) f elds where the measures were developed and the and other environmental agencies in the USA and locations at which the user is applying the model. abroad publish nutrient pollutant export coeff cient Factors likely to vary from site to site and inf u- tables (Gotaas 1956; Reckhow et al. 1980; Athayde ence how much nutrient is exported from a given et al. 1983), which can be applied to different regions parcel include slope, soil type, and precipitation. and replaced by local estimation and local knowl- These same factors are key drivers of water yield, edge when possible. Other pollutant proxies, such so we use a relative surface yield index to correct as manure or fertilizer application, can be used with for non-average conditions in application parcels. plant nutrient uptake to estimate the potential We do this by f rst estimating a runoff index based export values per LULC if export coeff cients are on the amount of water yield that will accrue to not readily available. Although the goal of this parcel x , 92 RETENTION OF NUTRIENTS AND SEDIMENT BY VEGETATION

Land Use Residential/Urban Crop Grass/Pasture Forest/Shrub Water/Wetland Streams

N 0 2.5 5 Kilometers

Figure 6.1 The Willamette Valley, Oregon ( USA). This catchment is characterized by upstream forested areas and more intensive agricultural, residential, and urban land uses downstream. Streams run from the southeast to the northwest. Most of the cropland is concentrated in the northwest portion of the catchment, which is an area without many forests or shrublands.

ALV= EAF⋅ pol , (6.3) ⎛⎞ xxx lxu= Log⎜⎟∑Y , (6.1) ⎝⎠U where pol x is the export coeff cient at parcel x. where Y is the sum of the water yield of parcels ∑U u To demonstrate the predictive capacity of the meth- along the f ow path above parcel x (including the odology described above, we modeled phosphorous water yield of parcel x ). This value can be derived retention in a catchment in the Willamette Valley of from outputs of the surface water yield model Oregon (USA) ( Figure 6.1 ). In this example, which we described in Chapter 4 , or from any other water will use to illustrate the complete set of outputs from yield model. Then, because we want to represent the nutrient retention model, parcels are 900 m2 . increased nutrient export in areas with above- average The tier 1 model shows that the majority of the factors associated with nutrient export and water sources of phosphorous in this catchment are crop- yield, and decreased nutrient export in areas with lands located in the northwest portion of the catch- below-average values, we develop the export ment (Figure 6.2a). Note that the rivers in this adjustment factor, sub-catchment f ow from the southeast to the northwest. Figures 6.2a and 6.2b provide maps of ALV at lx EAFx = , (6.2) different spatial resolutions. lW

where lW is the mean runoff index in the watershed 6.2.1.2 Modeling vegetative f ltration and its effects of interest. Finally, we combine the export adjust- on non-point source pollution ment factor with the export coeff cient to estimate To model cross-parcel nutrient retention on the an adjusted loading value ( ALVx), landscape, we must know the f ltering capacity of (a)

Adjusted Load (kg)

13.61

0 Streams

N 0 2.5 5 Kilometers

(b)

Total Adjusted Load (kg) 20023.5

0.136 Streams

N 0 2.5 5 Kilometers

Figure 6.2 Phosphorous adjusted loading values at the parcel (a) and sub-catchment (b) scale in the Willamette Valley, Oregon (USA). Note that, due to the positions of the sub-catchments, the sub-catchment with the highest total loading in the catchment is not one that borders the focal body of water. This example illustrates the importance of spatial scale in identifying priority areas for the regulation of nutrient application. 94 RETENTION OF NUTRIENTS AND SEDIMENT BY VEGETATION

different LULC categories as well as the f ow path 6.2.2 Modeling soil erosion to quantify along which nutrient-carrying surface water will sediment retention by vegetation f ow. We f rst identify all landscape parcels that For sediment retention, the tier 1 approach starts by include f ltering vegetation. Next, we route the ALV x calculating the potential sheetwash erosion on the through the downslope vegetated parcels, which landscape of interest. This model uses the USLE will each retain some of this loading based on their method to incorporate the geomorphological, climatic, f ltration eff ciencies and export the remainder of and land management characteristics of the landscape the nutrient load to the next downslope parcel. Let (Wischmeier and Smith 1978). Modif ed forms of the E represent the f ltering eff ciency of parcel x ’s x USLE continue to be used, at a minimum, to provide LULC, where E represents the percentage of nutri- x the relative potential of a landscape parcel for sheet- ents reaching parcel x that will be retained on the wash erosion (excluding gully and streambank ero- parcel. This means that (1 – E ) of the nutrients x sion) (Reid and Dunne 1996 ). Hydraulic connectivity reaching parcel x will be exported to the next is used to account for the location of sediment genera- downslope parcel. We assume that the uptake tion, retention, and transport in the landscape. Outputs mechanisms acting on the pollutants are never of the model include maps of estimated sediment saturated. retention and the cumulative amount of sediment To calculate the nutrient retention provided by exported to downstream bodies of water. parcel x , we index all parcels along a given f ow Unlike the nutrient retention model, the sediment path based on their distance to the destination water retention model allows parcels to retain some of the body, with x = 1, 2, . . . , X . Parcel 1 is the parcel fur- sediment for which it is a source. In other words, this thest from the water body along the f ow path, model accounts for both on-parcel and cross-parcel while parcel X is adjacent to the body of water. In retention. As such, to calculate the total amount of this system, the nutrient retention on parcel x in a sediment retention taking place on a parcel, we must given f ow path is given by calculate both the amount of erosion avoided from x−1 x−1 parcel x and the sediment reaching parcel x from Nuretxx= E.(∑ ALV y∏ 1)− E z, (6.4) y=1 zy=+1 upslope parcels that is retained by parcel x . First, we

let USLEx represent the amount of sediment originat- where Nuret x is the nutrient retention in kilograms ing from parcel x , w h e r e USLEx i s d e f ned as on parcel x based on its f ltration eff ciency ( E x ),

ALV y is the adjusted loading value of parcels higher USLE R K LS C P , (6.6) up the f ow path than parcel x , and E z is the f ltra- xxxxxx= ⋅⋅ ⋅⋅ tion eff ciency of parcels higher up the f ow path than parcel x . where R x is the rainfall erosivity, which represents We also estimate the amount of nutrients origi- the ability of rainfall to move and erode soil and is a nating on parcel x that reach the downstream function of average regional rainfall intensity and duration; K is the soil erodibility, which represents water body. Let Exp x , which represents this por- x tion of the nutrients originating on parcel x , be the soil’s susceptibility to erosion and is a function given by of soil texture and characteristics; and LS x is a slope-

X length index that characterizes the potential energy associated with the uninterrupted slope leading up Expxx= ALV∏ (1− E y ), (6.5) yx=+1 to parcel x. Breaks in slope length are based on

where all of the variables in Eq. (6.5) are as previ- Renard et al. (1997) and the algorithm for LS x comes ously def ned. The amount of nutrient export from from Stone and Hilborn (2000). C x is a dimension- a parcel will increase as loading on that parcel less ground cover variable that varies from 1 on increases and as the number of downstream f lter- bare soil to 0.001 for forest. Finally, Px is a manage- ing parcels, or their f ltering eff ciencies, decrease. ment factor that accounts for specif c erosion con- Figure 6.3 illustrates the retention provided by par- trol practices such as contour tilling or mounding, cels downstream of nutrient loading sources. or contour ridging. P x varies from 1 on bare soil TIER 1 BIOPHYSICAL MODELS 95

Nutrient Retention (kg) 235.99

0 Streams

N 0 2.5 5 Kilometers

Figure 6.3 Average phosphorous retention values in the Willamette Valley, Oregon (USA). This map underscores the point that our model formulations make it such that native vegetation must be downstream of loading sources to provide nutrient retention of value to society. As the large area of contiguous forest in the eastern part of the catchment lies upstream of anthropogenic phosphorous sources, it does not provide much, if any, cross-parcel nutrient retention in the eastern part (light zones). However, in buffer areas, forested lands around streams provide much retention. Loadings in the western part of the watershed are retained and removed by forested lands abutting streams (dark zones close to streams).

with no erosion control to about 0.1 with tiered sediment originating on parcels higher up the f ow ridging on a gentle slope (Roose 1996). path, is given by

The tier 1 model of sediment retention determines x−1 x−1 the potential sediment release on each parcel on SEDRxx= SE∑ USLE y∏ (1− SE z ), (6.7) the landscape of interest. These soil particles are y=1 zy=+1 detached and move through the watershed with where SE x is the sediment retention eff ciency of runoff along f ow paths. As in the nutrient retention parcel x, USLE y is the sediment generated on model, we index all parcels along a given f ow path upstream parcel y , and SE z is the sediment retention based on their distance to the destination water eff ciency of upstream parcel z . body, with x= 1, 2, . . . , X . Parcel 1 is the parcel fur- The potential amount of sheetwash sediment thest from the water body along the f ow path, trapped by landscape vegetation or best manage- while parcel X is adjacent to the body of water. The ment practices in soil conservation upstream of model routes the sediment originating on parcel x , reservoir D , SEDRETxD , can be estimated by the dif-

USLE x , along the f ow path, with vegetated parcels ference between the geomorphological characteris- retaining some of this sediment based on their tics of x that might promote soil loss, and the sediment retention eff ciency and exporting the retention properties of the parcel’s LULC that help remaining loading to the next parcel in the f ow contain sediment released by on-site erosion (C x path. SEDR x, the retention by parcel x' s LULC of and P x ) and upstream transport (SEDR x ): 96 RETENTION OF NUTRIENTS AND SEDIMENT BY VEGETATION

Sediment Retention (kg) 3462.72

0 Streams

N 0 2.5 5 Kilometers

Figure 6.4 The map of annual average sediment retention emphasizes the retention capacity of forested pixels. Streams throughout the sub-catchment have forested pixels nearby, which provide substantial sediment retention that should serve to extend the productive life of downstream reservoirs. Interestingly, the forested pixels in the southeastern section exhibit signif cant retention, which illustrates the inf uence of slope in the sediment retention model.

SEDRETxD= R x⋅⋅ K x SL x ⋅−⋅()1. Cxx P+ SEDRx (6.8) f ltration on the landscape in achieving a certain nutrient load. It should be noted that this value will T h e f rst term on the right-hand side of Eq (6.8) is the represent the lower bound of the social value of amount of sediment originating on parcel x r e t a i n e d maintaining water below the specif ed concentra- by that parcel, while the second term represents the tion level: if the community is willing to incur these amount of sediment originating on upslope parcels costs, then the value of achieving this threshold retained by parcel x . SEDRET xD , c a n b e m a p p e d t o must be at least as great as the treatment costs. display the supply of sheetwash sediment retention While we include the avoided treatment cost meth- by vegetated parcels on the landscape. Figure 6.4 odology in the model, the dollar value assigned to illustrates sediment retention at the parcel level in achieving a certain standard is an input to the the Willamette basin sub-catchment where we pre- model, so the user is free to choose any of alterna- viously considered nutrient retention. tive valuation methodology such as contingent valuation or one of the reaveled-preference approaches (Desvousges et al. 1987; Englin and Cameron 1996; 6.3 Tier 1 economic valuation Leggett and Bockstael 2000; Hanley et al. 2003). Let s represent a standard regarding the concen- 6.3.1 Valuation of nutrient retention Dh tration of nutrient pollutant h in focal water body D .

It is possible to use the avoided cost of cleanup at We assume that s Dh represents the minimum annual water treatment plants as a proxy for the value of loading of nutrient h below which there are no TIER 1 ECONOMIC VALUATION 97

damages associated with the presence of h in water the present value of the stream of costs incurred body D . Let L Dh represent the actual total pollutant over a pre-specif ed length of time. The present load of h in D. The total cost of cleaning the water to value of a stream of treatment costs is given by the desired threshold concentration level, C(L ) , T Dh ˆˆ will be a function of the desired threshold, the CL()DtD""= ∑ C () L t=1 (6.12) incoming concentration, and the technology T AC()( L×− L s ) tDh D"" D. employed in treatment: = ∑ t t=1 (1+ r )

CL()Dh = f (,, LDh s Dh tech D ). (6.9) Note that in the above equation, we allow the average treatment cost, the pollutant load, and the water Utilities are generally considered natural monopolies quality standard criterion to vary across years, because their operations exhibit decreasing average though it is possible for them to be constant over and marginal costs across their range of output. Water time. r represents the market interest rate used to treatment utilities are no exception. In this situation, discount future payments into present value. there is no producer surplus, and the welfare impacts This model focuses on a single contaminant that of increased environmental quality accrue solely to impacts the value of water for downstream use. The consumers. Furthermore, marginal cost pricing valuation methodology described above is predi- would lead to f nancial losses for the utility, requiring cated on the user’s ability to identify the cost of treat- subsidies for ongoing operation, meaning that the ing water for removal of a focal contaminant supply curve is the average cost curve (Foster and (nitrogen, sediment, fecal coliform, mercury, etc.). If Beattie 1981 ). Assuming perfectly inelastic demand isolation of individual treatment costs is possible, the allows us to def ne the social value of pollutant regu- above methodology can be used to identify the value lation as the rectangle representing the change in of contaminant load reduction for all contaminants price and the quantity produced (Holmes 1988). present in a given watershed. However, the isolation of water cleanup costs on a per-contaminant basis is The average treatment cost, AC (L Dh ), can be def ned as often not possible. In these cases, we suggest two alternative methodologies. One approach would be CL()Dh fL(,,tech)Dh s D"" AC() LDh == , (6.10) to use the average cost of treating water for all its con- Volume Volume tained contaminants as the basis for valuing avoided where Volume represents the total volume of water treatment costs for a single contaminant. Note that treated. The estimated average cost of treatment can this approach will necessarily result in double-count- be obtained through discussion with operators of ing if applied to multiple contaminants. An alterna- the appropriate treatment plant. The average cost tive approach is to rely on the development of a can then be used to assign value to water of any weighting matrix to allocate the total cost of treat- quality reaching the focal water body given certain ment across the vector of contaminants removed assumptions. These assumptions are that there is a during treatment. The allocation of treatment costs constant average cost of treatment and that the across pollutants might be based on existing expert treatment technology will remain constant through opinion or estimates of such per-pollutant costs in time. Thus, the proxy for the cost of treatment is the literature. Applying this approach to estimate contaminant-specif c costs of water treatment relies ˆ CLtD()""= ACL ()(Dh×− L D s D" ). (6.11) on the assumption that the average treatment cost is constant across pollutants. Each contaminant’s con- It is important to note that in Eq. (6.11), we only tribution to total treatment costs could be estimated assign costs to pollutant loading that exists in excess by its contribution to turbidity, a measure of the sedi- of the target water quality criterion. The above ment load in the water, which is frequently used to equation provides the cost of achieving the target determine treatment costs (Dearmont et al. 1998). criterion in a given year, t . In order to accurately We can think of a parcel’s value in the context of capture the costs of treatment, we need to identify nutrient loading in two ways. First, we can consider 98 RETENTION OF NUTRIENTS AND SEDIMENT BY VEGETATION how land that exists in native vegetation is able to where all variables in Eq. (6.14) are as previously restrain the f ow of pollutant h into the focal body of def ned. Note that a parcel with f ltration capacity water, D . Filtration by vegetation represents an eco- will be assigned zero value if there is no pollutant to system service that provides value to society. We can f lter—namely, vegetation with f ltering properties also think of the conversion of land that is currently in a pristine watershed has no value. This is a result an anthropogenic source of pollutant h i n t o a n a l t e r - of our inability to capture on-parcel f ltration, and native land use that would no longer be an anthro- the fact that vegetation in landscapes with higher pogenic source of pollution or of how the f ltration nutrient inputs do f lter out more nutrients because currently offered by a parcel covered in extant veg- more nutrients are moving across the landscape etation would change if the parcel were converted to and available for f ltration. development. These two approaches come at the We applied this valuation approach to phosphorous issue of a parcel’s impact on nutrient loading from retention in the Willamette Basin. We assumed that different angles, and together they offer the com- the basin’s land use will stay stable for 15 years, and plete picture of how a planner might think of the that average treatment costs are US$68 per one million regulation of nutrient pollution in a watershed. gallons of water. We used a discount rate of 5%. Figure To determine the treatment costs that might be 6.5 illustrates the value of cross-parcel nutrient reten- avoided if parcel x were converted from pollutant- tion by vegetation based on these values. contributing land use j t o l a n d u s e j' not associated with pollutant h , we must identify the portion of load- 6.3.2 Valuation of sediment retention ing from each parcel that is costly. We attribute costs to a constant portion of the loading coming from each The more sediment a landscape upstream of a reser- parcel, based on the difference between actual load- voir can retain, the longer the life expectancy of the ing and the desired maximum threshold in the water reservoir, or the less a reservoir manager has to body. Given this assumption, the cost of water quality spend on sediment removal. In this model, we impairment induced by source parcel x i s approximate the value of this benef t using the avoided cost of sediment removal. It should be Yx ()Ls− AC() L××∑ X Dh Dh × Exp noted that not all sediment that reaches a reservoir T−1 tDhLs x c = Dh Dh , (6.13) is removed (e.g., dredged): a certain rate of sedi- x ∑ (1+ r )t t=0 mentation in a reservoir is usually tolerated

w h e r e c x is the contributing cost at water body D o f (Palmieri et al. 2003). For those reservoirs whose pollutant h f r o m p a r c e l x w i t h L U L C j g i v e n p o l l u t - sedimentation has caused the reservoir to reach its ant loading at the parcel and downstream f ltration dead volume, which is the point at which reservoir by natural vegetation, Σ Yx represents the cumulative function is impacted by sedimentation, we assume water yield in the watershed, and all other variables that all sediment originating from upstream parcels are as previously def ned. We include the ratio of will be removed. In all other reservoirs, we assume water yield to pollutant load to allow our biophysi- that none of the sediment reaching the reservoir cal measure of pollutant loading, measured in kilo- will be removed. The present value of retained sedi- grams, and our cost of treatment, which is measured ment on parcel x , PVSRxD , is given by in units of currency per volume of water treated, to give us a result measured in currency terms. T 1 PVSRxD= () SEDRET x×× R D MC D ×∑ t , (6.15) We assign value to each parcel based on the treat- t=0 (1 + r) ment costs avoided due to the presence of f ltering vegetation. The benef ts provided by parcel x f o r n u t r i - where the index D i n d i c a t e s t h a t x i s u p s t r e a m o f ent retention by f ltering vegetation, bx , a r e g i v e n b y reservoir D , SEDRETx is the amount of sediment removed by the LULC type on parcel x a n n u a l l y ( s e e Y X x ()LsDh− Dh Eq. (6.8)), and T indicates the number of years we AC() L××∑ × Nu _ ret T−1 tDhLs x b Dh Dh , expect present landscape conditions to persist or the x = ∑ t (6.14) t=0 (1+ r ) expected lifetime of reservoir D ( s e t T t o t h e s m a l l e s t TIER 2 BIOPHYSICAL MODELS 99

Value of Nutrient Retention ($) 59, 467

0 Streams

N 0 2.5 5 Kilometers

Figure 6.5 Avoided water treatment costs in the Willamette Valley, Oregon (USA). The most valuable pixels providing nutrient retention in this catchment are those located closest to the sources of pollution rather than those that have the highest f ltration capacity (a). When results were aggregated to the sub-catchment level (b), the maximum value is assigned to those that do not include the pixels with the highest value, again emphasizing the importance of spatial scale regarding management decisions.

–1 value if the two time values differ). R D , the retention assumed that sediment dredging costs US$1.37 ton factor of reservoir D , describes the fraction of deliv- (based on costs for removing river sediment in 1983; ered sediment that is retained in reservoir. MC D i s Moore and McCarl 1987). We assumed a productive the marginal cost of sediment removal from reser- lifetime for the reservoir of 100 years and a discount voir D and is based on the appropriate regional tech- rate of 5%. nologies available, and the size and purpose of the reservoir. For simplifying purposes, we assume a 6.4 Tier 2 biophysical models constant marginal cost of sediment removal. Sediment removal technologies have widely dif- The tier 1 models are created with the goal of pro- ferent costs and include f ushing, sluicing density viding a credible depiction of nutrient and sediment current venting, dredging, dry excavation, and retention on the landscape, while imposing mini- hydrosuction (Palmieri et al. 2003). Note that the mal data requirements on model users. Achieving specif cation of Eq. (6.15) treats sediment retention this goal required the development of novel hydro- and the marginal cost of removal as constant across logical models. The tier 2 models aim to provide time and assumes a constant marginal cost of more realistic depictions of these processes, which removal. is partially achieved by relying on existing hydro- We applied this approach to a sub-catchment of logic models that provide f ner scale temporal the Willamette River, Oregon (USA) ( Figure 6.6 ). We resolution. (a) Value of Sediment Retention ($) 37738.9

0 Streams

N 0 2.5 5 Kilometers

(b) Total Value of Sediment Retention ($) 13, 131, 600

0 Streams

N 0 2.5 5 Kilometers

Figure 6.6 Avoided reservoir dredge costs in the Willamette Valley, Oregon (USA). The highest values, up to $1667 per 30 × 30 m pixel (~$18 500 per ha), occur near the outlets into the reservoirs because sediment has few potential opportunities for re-deposition (a). Aggregation of the value to the sub-catchment level changes the view of the basin somewhat (b). TIER 2 BIOPHYSICAL MODELS 101

(saturation excess) landscapes (Steenhuis et al. 1995; 6.4.1 Tier 2 nutrient retention Schneiderman et al. 2007; Easton et al. 2008b) and In tier 2 we simulate non-point source nutrient run- Hortonian f ows (Horton 1940) in dry (arid rainfall off using the Annualized Agricultural Non-Point excess) landscapes. In order to apply this approach Source (AnnAGNPS) model (Bingner and Theurer one must def ne Hydrologic Response Units (HRUs) 2007) to simulate non-point source loadings. that incorporate, at a minimum, parcel hydraulic AnnAGNPS is a joint USDA–Agricultural Research and topographic index characteristics but that Service (ARS) and Natural Resource Conservation might also consider management boundaries, such Service (NRCS) system of computer models devel- as tenure or zoning units (Fig. 6.7). oped to predict non-point source pollutant loadings In most water quality models (e.g., Soil Water within agricultural watersheds. The AnnAGNPS Assessment Tool (SWAT)) HRUs are def ned by the model consists of several different programs for coincidence of land use and soil inf ltration capac- realistically capturing pollutant loading, the f ow of ity. In landscapes where variable source areas con- pollution across the landscape, and the in-stream trol runoff generation, it is more appropriate to impacts of this pollution (Bingner and Theurer def ne HRUs by land use and a soil wetness index 2007). The model contains a continuous simulation (e.g. SWAT-VSA; Easton et al. 2008a). Wetness index surface runoff model designed to assist in identify- classes are determined by dividing a watershed into ing best management practices, setting Total several (at least ten) equal areas delineated by lines Maximum Daily Loads (TMDLs), and incorporat- of equal topographic index. Any additional infor- ing risk and cost–benef t analyses (NRCS 2008). We mation needed to estimate pollutant loading other structure the tier 2 nutrient retention model with than LULC, such as topography (e.g., slope position the needs of watershed managers in mind. Linking and length) and soil chemical properties, are aver- AnnAGNPS to a process of economic valuation aged within each HRU. This is generally acceptable allows landscape managers to incorporate ecosys- because there is evidence that soil variability tem services into their management practices. roughly correlates with topographic features, which There are several fundamental differences are captured by the topographic index (Page et al. between the tier 1 and tier 2 nutrient retention mod- 2005; Sharma et al. 2006; Thompson et al. 2006). els. First, the tier 2 model allows the user to incorpo- The tier 2 nutrient retention model can character- rate management decisions made on sub-annual ize each loading and retaining parcel. Furthermore, timeframes. Second, the Tier 2 model recognizes the model can evaluate to what extent parcels that nutrient pollution export is highly dependent impact overall loads and how incremental changes on dynamic hydrology. Finally, the tier 2 model at each of these parcels affect nutrient loading at incorporates nutrient-specif c characteristics when the demand point. In tier 2, we keep track of nutri- tracking the transport of the pollutant across the ent application, pollution runoff, and vegetative landscape. The challenge of using the more sophis- f ltering on a daily time step. This arrangement ticated tier 2 model is that model setup and inter- allows the user to evaluate not only changes in pretation require the user to possess relatively magnitude but also to see changes in nutrient sophisticated knowledge of hydrology. Useful sim- export and retention as a function of seasonal ulations will, at a minimum, require extensive cali- hydrology, vegetation growth stages, and seasonal bration, thoughtful def nition of modeling units, management changes. and an understanding of the driving processes of Tier 2 outputs are aggregated within the mode- the landscape in question. ling HRUs at any desired temporal period with less The tier 2 nutrient retention model provides than daily frequency. There are two key tier 2 nutri- estimates of non-point source pollutant loading ent retention model outputs. First, there is the at f ne temporal scales that are linked to hydrology expected non-point source contribution of nutrient via climate and land use. Runoff hydrology in the pollutant h from HRUx to D during time period I ,

AnnAGNPS model is based on a modif ed SCS- NPSP DhI (HRUx). Next, there is the expected vegeta-

Curve Number (CN) approach and is taken to tive f ltration from HRU x of nutrient pollutant h

represent variable source area hydrology in wet upstream of D during time period I , FILT DhI (HRU x ) . 102 RETENTION OF NUTRIENTS AND SEDIMENT BY VEGETATION

Wetness index class Land use

w1 w2 forest corn

HRU1

HRU4

HRU3 HRU2

Figure 6.7 Schematic showing how digital maps of wetness index and land use are combined to delineate HRUs. In this case, two wetness index classes, w1 and w2, are combined with two land uses, corn and forest, to produce a watershed with four HRUs. The HRU is an homogeneous land segment with same wetness index and land use.

Both of these outputs are modeled using the rou- As noted earlier, the detachment rate of sediment tines and functions of AnnAGNPS. is dependent on rainfall intensity, geomorphology, The largest period, I, is one year and the model is and runoff volumes (see Hjelmfelt et al. 1975), and run using hydrology forcings (rainfall events) of its movement in the stream channel is coupled with several years, such that the natural variability in the energy in simulated f ows. We def ne the stock precipitation can be accounted for as best as possi- of sediment delivered to reservoir or impoundment ble. Pollutants are aggregated in the form of aver- D at time t , zDt , as age daily load averages (mass) within each period ⎛⎞ that can be combined with water yield simulations z=+ z h,(,) sdr x D c . (6.18) Dt ⎝⎠⎜⎟∑∑ xt' D per HRU to produce effective concentrations. The VT' biophysical service processes for pollutant removal where z Dt is a function of h xT ´ , the cumulative sedi- for each modeling HRUx is simply ment detached from V , the set of upstream HRUs indexed by v = 1, 2, . . . , V across all times before t ,

FILTSERVDhI() HRU x = FILTDhI(. HRU x (6.16) given by T ', where T’ i s i n d e x e d b y t’ = 1, 2, . . . , t – 1 (i.e., t' < t). The likelihood of this sediment reaching The relative impairment of contributing pollutant is D is mitigated by a sediment delivery ratio, sdr , thus expressed as follows for each modeling HRU x : which is a function of the spatial conf guration

between x and D . Finally, χ D accounts for all the NPSPDhI() HRU x NPSPIMPDhI() HRU x = . (6.17) sediment that arises from other sources or processes NPSP HRU ∑ DhI() x not modeled, such as landslides or stream bank ero- Awxj ∈ sion, which can be assumed to be a constant value based on site- or region-specif c sediment budget 6.4.3 Tier 2 sediment retention studies. The tier 2 model simulates sediment detachment and transport for storm events using the modules in 6.5 Tier 2 economic valuation models the PRMS developed by the US Geological Survey (Leavesley et al. 1983). In order to identify retention The tier 2 economic valuation models do not dif- by LULC j on HRU x, we must run PRMS twice, fer fundamentally from the tier 1 models. The f rst with bare soil on each HRU and then again value of nutrient and sediment retention is still with current LULC on each HRU, with the differ- estimated using the costs avoided by vegetative ence representing retention on each HRU attributa- retention of nutrients and sediment on the ble to its LULC. landscape. The main advance in the tier 2 models TIER 2 ECONOMIC VALUATION MODELS 103 is that management decisions can be incorpo- revenue streams per hectare of LULC j* , meaning rated into the valuation of the focal ecosystem that T V services. d = xj* t , xj* ∑ t−1 t=1 (1+ r ) (6.20) 6.5.1 Tier 2 valuation of nutrient retention where V xj*t represents the private per-hectare value Simple maps of valuation are informative but are by accruing on parcel x in year t from LULC j* . Solution no means an end unto themselves. For many land- of this problem leads to the outcome that land will use or resource management decisions, the real only be placed in nutrient-loading LULC j* so long question is what mixes of land use and activities as T dxj* 1()Ls− achieve some goals at minimum costs. For example, ≥ AC() L Dh Dh . Exp∑ (1+ r )t tDh s there are two ways to achieve reduced nutrient xj* t=1 Dh (6.21) loading in a focal body of water, namely reduced nutrient pollutant loading or increased vegetative Equation (6.21) shows that the net benef t of LULC f ltration in the landscape. The question then is: j* per unit of nutrient exported must be greater than how might the targeted level of nutrient loading be or equal to the net cost of nutrient loading in water achieved in the least-cost manner? The tier 2 model body D for the planner to allow nutrient-loading provides users with several frameworks to answer LULC j* on parcel x . this question. In areas where estimates of the cost of water treat-

Let us assume a specif c load threshold, s Dh, and a ment are not available, we might restate the objec- single entity, a benevolent social planner, that has tive function presented in Eq. (6.19) as the opportunity to make land-use decisions across XX the landscape. There are several different objectives maxdxj* I x+ λ ( s D" − Exp x I x ), (6.22) I ∑∑ such a land-use planner might pursue regarding x xx==11 development and nutrient loading. One possibility where dxj * , Exp x , s Dh , and Ix are as def ned above, and is that the social planner will attempt to maximize λ represents the additional development value that the development value of the landscape (i.e., the might be realized by increasing the amount of load- sum of the development value across developed ing in water body D . This specif cation allows users parcels), while accounting for the social costs of to evaluate the trade-offs between intensive land nutrient loading associated with different LULC use and pollutant loading in water body D without categories. Typically, we think of minimizing the necessitating the identif cation of the damages cost of regulation, but in the case of pollution from caused by such loading. land use, the costs of regulation are opportunity In Eqs. (6.19) and (6.22), the unit of management costs rather than costs of technology adoption. is assumed to be the parcel. In reality, decision units Under certain assumed conditions, this objective is may have different spatial resolution, with bounda- represented in the expression ries based on property lines or sub-catchments. In

X this case, there may be several LULC categories on

max (dcIxj**− xj ) x , (6.19) a single decision unit. To accommodate this possi- I ∑ x x=1 bility, we can rewrite Eq. (6.22) as

Z J where c xj* is as def ned in Eq. (6.13), dxj * represents max dAzj zj + the net present value of parcel x in LULC j* , where A ∑∑ zj zj==11 j* represents the highest value LULC on parcel x Z J (6.23) λ(),s− Exp A and is taken as exogenous, and Ix is a binary ran- Dzjzj" ∑∑ dom variable, where 0 indicates that the land is not zj==11 in use (i.e., not acting as an anthropogenic source of where A zj represents the area of decision unit z that pollutant h ) and 1 indicates that the parcel is in is in LULC j , and all other variables are as previ- LULC j*. In landscapes with functioning land mar- ously def ned. In order to obtain the eff cient pat- kets, d xj * represents the net present value of future tern of land use on a decision unit, the planner must 104 RETENTION OF NUTRIENTS AND SEDIMENT BY VEGETATION have precise information about how revenues asso- storage loss and inf ux of sediment stock. See the ciated with LULC j vary with the area of land supplementary online materials (SOM) for techni- engaged in the given land use. Note that in addition cal details on f nding the path of y dt for t = 1 to T that to identifying the least-cost means of achieving a minimizes the negative impact of sediment loading targeted threshold loading, the frameworks pre- on the reservoir. Let this path by given by y* dt and sented above can also be used to target cost-effective let the net present value of sedimentation at system restoration sites, which would entail identifying d until time T be given by those parcels whose value of nutrient retention is T−1 −−By()* cy* greatest per lost value of being converted from pro- NPVS = ddtytdt, (6.26) d ∑ (1+ γ )t duction to native vegetation. t=0 where T indicates the number of years we expect present landscape conditions to persist or the 6.5.2 Tier 2 valuation of sediment retention expected lifetime of reservoir d (set T to the smallest

L e t gdt represent a mass of sediment stock in impound- value if the two time values differ). NPVSd ≤ 0 for all ment system d a t t i m e t . N e x t , l e t y dt represent the solutions. By comparing NPVSd scores from a landmass of sediment removed from d i n t i m e p e r i o d t , scape covered by bare soil with those from a land- while zdt represents the rate at which sediment is scape with a different LULC pattern, positive delivered to d i n p e r i o d t via landscape runoff. The ecosystem service values will be generated. See calculation of zdt is a function of the LULC pattern and Box 6.1 for a discussion of efforts in China to pro- rainfall during period t on the landscape. Therefore, vide incentives for land management associated with increased nutrient and sediment retention.

ggyzdt+1 = dt− dt+ dt (6.24) 6.6 Constraints and limitations gives the stock of sediment in d at time t +1. When deciding how much sediment to remove One major simplif cation made in these models is from the reservoir in time period t , a manager is that we ignore in-stream processes in quantifying interested in minimizing the sum of all future costs the amounts of nutrient and sediment delivery, of sediment removal and the dis-benef ts associated retention, and value. In doing so, we attach artif - with any future remaining sediment stocks. The cially high signif cance to the contribution of decision-rule on how much sediment to remove upstream parcels regarding total nutrient and sedi- from d in time period t (i.e., what level to set ydt for ment loads in the focal downstream body of water. each year t ) is given by The nutrient retention model is based on surface and subsurface f ows in saturation excess regions. T −−Bg() cy minddtytdt , (6.25) In watersheds where the interaction between sur- ∑ t ydt (1+ r ) t=1 face water and ground water is signif cant (i.e.,

subject to gdt +1 = g dt – y dt + z dt where B d ( gdt ) represents shallow aquifers), nutrient modeling using tier 1 the dis-benef t associated with a sediment stock of will likely misrepresent export and retention, again

gdt , c yt is the per unit cost of sediment removal in overestimating these values in most cases. time period t and is a function of the volume of the Application to such systems should be undertaken reservoir and regional technical capabilities (see cautiously, and other factors such as travel time Palmieri et al. (2003) for costs and appropriate tech- between the source parcel and point of interest nologies for sediment removal), and r is the dis- should be analyzed. count rate. Bd ( g dt ) is estimated by applying the tier 2 Not surprisingly, the scales of the watershed and water allocation model and determining the differ- the river basin being modeled play a signif cant role ence in benef ts associated with reservoirs with in determining the extent to which LULC changes lower storage capacity. However, to do this requires impact water quality. In large basins, landscape f eld data, such as bathymetry observations, to complexity can become important in mitigating the establish general relationships between reservoir impacts of land use on the hydrologic regime and TESTING TIER 1 MODELS 105 water quality by providing storage capacity in that def ne the formation of an ephemeral stream) the watershed and in stream beds. For the same and the eff ciency of pollutant removal per unit of percentage of LULC change, impacts will be much downhill vegetation are uncertain variables that greater in smaller watersheds. can affect outputs in nonlinear ways. In addition to the scale impact on overall water To understand the impact of uncertainty in these quality processes, the tier 1 nutrient retention model two variables, we performed a preliminary sensitiv- assumes that saturation excess hydrology and prox- ity analysis of these two variables in the Baoxing imity to water bodies dominate water impairment. County watershed within the Upper Yangtze River Also, this model measures all factors at an annual Basin in China. The selected catchment exhibits time step including rainfall-runoff dynamics and highly varied soil, LULC, and management types in processes, fertilizers or pollutant applications, and a humid monsoon landscape. It is of interest because plant growth dynamics. In the real world, water it suffers from a signif cant nutrient loading prob- quality components and processes depend upon a lem due to livestock and intensive agricultural pro- variety and complexity of sub-annual processes not duction. Furthermore, tens of millions of people simulated in this tier 1 model. downstream of this watershed rely on this water for The tier 2 nutrient retention model does not sim- consumptive uses, so ensuring adequate water ulate in-stream processes and/or any biochemical quality is essential. transformations. However, temporal dynamics are Prior to evaluating the sensitivity of modeled more fully incorporated as all calculations are done loading with regard to various parameters, we com- at the HRU level on daily time step. pared modeled water yield totals (used as one of The tier 1 and tier 2 sediment retention models the input parameters) with annual average observed are both based on the USLE, which is best suited for water yield from a ten-year time series and, follow- agricultural land and moderate slopes. Furthermore, ing adjustments for groundwater sources outside of these models only incorporate sediment from sheet- the watershed, found that modeled water yield rep- wash sources, thus conceptually ref ecting a lower resented 92% of the observed annual average yield. limit of the true relationship of total sediment yield We performed the sensitivity analysis for both nitro- and landscape practices. The tier 2 model, PRMS, is gen and phosphorous pollutant loadings. We f nd based on USLE and revised to include the temporal that, for both nitrogen and phosphorous, the model dynamics of precipitation and their impacts on is more sensitive to changes in export coeff cients erosivity. than retention eff ciency.

6.7 Testing tier 1 models 6.7.2 Validation testing One risk associated with simple models that have The theoretical foundations of our contaminant limited data requirements is the possibility that model have been successfully applied in previous they will misrepresent the processes being mod- studies dealing with saturation excess runoff water- eled. To ensure that our models provide reasonable sheds (Endreny 2002; Walter et al. 2003; Schneiderman outputs, we have started testing them in watersheds et al. 2007). Nevertheless, it is important to conf rm around the world. Here we report the results of that our model predictions align with observed out- these efforts for the nutrient retention model. comes in as many different systems as possible. However, testing our predicted outcomes relative to observed nutrient loading and sediment deposi- 6.7.1 Sensitivity tion, particularly at the sub-catchment level, is quite The application of a pollutant on the landscape is a diff cult due to the lack of such spatially explicit direct and linear function—uncertainty in these val- information in most watersheds. As a proxy for ues will be directly proportional to errors in pollut- observed data, we compare our predicted outcomes ant loading modeling. However, stream threshold with those of a well-respected model that has been (the aggregate number of analysis drainage parcels calibrated to a specif c region. 106 RETENTION OF NUTRIENTS AND SEDIMENT BY VEGETATION

Here we present a comparison of our tier 1 attached to soil particles may be transported by predictions to the predictions of the much more surface runoff. Unlike nitrogen, phosphorus solu- data-intensive SWAT model (Arnold and Fohrer bility is very limited. Soluble P is leached only from 2005). In particular, we compare the total phospho- the top 10 mm of the soil. Surface runoff is the pri- rus load scores from our tier 1 nutrient retention mary mechanism by which phosphorus is removed model with those of simulated soluble phosphorus from the catchment; organic and mineral P attached by the SWAT model in 111 sub-catchments in the to soil particles may be carried off land by erosion Willamette River Basin as displayed in Figure 6.8 processes. The SWAT model is satisfactorily cali- (see Plate 2). Before describing the details of the brated to observed water yield at several different comparison, we offer a brief description of the sub-catchments within the Willamette River Basin, method applied by SWAT to describe nutrient load- with an R2 value of 0.17 at the daily unit of observa- ing and retention. tion. Given its correlation with observed data, The nitrogen and phosphorus cycles are dynamic which is not available at the desired spatial resolu- systems inf uenced by atmospheric, hydrologic, tion, we use the SWAT output as a proxy for plant, and soil conditions. These two nutrients are observed conditions. modeled in SWAT separately in very specif c meth- SWAT and our model differ signif cantly with ods and algorithms. They are added to the soil by regard to methodology. SWAT models specif c proc- fertilizers, manures or residue applications, f xation esses individually while making restrictive assump- (N), and rain (N). They are removed from the soil tions about uniform land use within HRUs summed by plant uptake, leaching, volatilization, de-nitrif - to the sub-catchment scale. Our model allows for cation, and erosion. SWAT monitors different pools f ne LULC detail across the landscape with a coarser of nutrients in the soil, generally categorized in model of the hydrology (annual average water mineral and organic forms. It calculates nitrogen yield) behind the focal processes. That said, we leaching as it calculates surface runoff and lateral might expect our modeled outputs to correlate f ow when it uses an exponential decay weighting closely with observed conditions, as proxied for by function in groundwater response. Organic N calibrated SWAT outputs ( Hernandez et al. 2008 ),

Normalized SWAT and tier 1 Phosphorous export 3.5

3 tier 1 = 0.621 × SWAT 2 2.5 R = 0.385

1.5

0.5

–0.5

Normalized tier 1 Phosphorous export Correlation Groups –1 1 2 –1.5 3 N –2 –1 0 1234 4 Normalized SWAT Phosphorous export 5 6 7 Group 1 Group 2 Group 3 Group 4 Group 5 Group 6 Group 7

Figure 6.8 Aggregated sub-catchment phosphorous export comparison between our model and SWAT (graph) and agreement of spatial phosphorous export patterns predicted by the two models (map) in the Williamette Valley, Oregon (USA). The graph on the left depicts a correlation between the normalized tier 1 model outputs and the normalized SWAT outputs. The groupings in the graph depict sub-catchments whose tier 1 outputs lie within a threshold distance of SWAT outputs given the correlation between normalized tier 1 and SWAT outputs. Note that Groups 4 and 6 represent sub-catchments in which tier 1 outputs are unexpectedly low and Groups 5 and 7 represent sub-catchments in which tier 1 outputs are unexpectedly high. The map illustrates the sub-catchment groups identif ed in the graph. (See Plate 2.) TESTING TIER 1 MODELS 107

Box 6.1 China forestry programs take aim at more than f oods

among the world’s largest payments for ecological services Christine Tam schemes (Liu et al . 2008). In the wake of the devastating Yangtze River f oods of 1998 These two forestry programs were initiated with and unprecedented extent of the drying of the Yellow River substantial funding from the central government for soil in 1997, the Government of China (GOC) initiated a and water retention benef ts but rapidly gained further number of forestry programs aimed at water and soil political favor for addressing an additional broader suite of conservation to mitigate these impacts in the future. Since government priorities and objectives. Besides reducing soil these initial forestry programs began, provincial- and erosion, the SLCP also aims to alleviate poverty, especially national-level government investment has continued and in remote, mountainous areas with low productivity lands, even expanded to protect watershed services more widely. reducing urban–rural development imbalances. In addition, Often with little quantif cation of actual service delivery, the SLCP promotes local economic development (Xu and these programs experienced relatively rapid approval and Cao 2002), facilitates shifts from farming to more initiation and generally sustain widespread support and sustainable production structures (SFA 2003), and bolstered compliance, in part because of a strong belief in the overall a then-troubled State Grain Bureau (Bennett 2008). benef ts of forests and in part for their ability to address a Subscription to the program was highly successful given the variety of associated objectives. relatively high rates of payments such that local and For the past two decades, China has experienced a regional implementation almost always met or exceeded consistent level of f ood damage despite heavy investment targets. in structural solutions to f ood control. These damages The NFPP additionally supported restructuring of rural culminated in 1998 with the Yangtze f oods. A follow-up economies, retraining and relocating over 700 000 former UNEP study (1999) attributed the effects of upper forestry workers (ADB 2006). Economies once solely watershed deforestation and overgrazing in reducing water dependent on timber production were now more reliant on storage capacity as a key factor that aggravated the multiple industries, including forest management, impacts of prolonged and substantial rainfall. Not only plantation farming, and tourism (Liu et al . 2008). were the degraded soils less able to retain water, but Compliance with the NFPP has been relatively successful, eroded soils washed downstream f lling river channels, with little illegal commercial logging. Villagers, even for lakes, and reservoirs, reducing their ability to mitigate f ood subsistence use, must apply to their local forestry bureaus waters (Zong and Chen 2000). for fuelwood or domestic construction timber allocations. A direct response to this disaster was consideration and State-run logging companies were eliminated virtually development of a more comprehensive system of f ood overnight, and new economies, such as that of Jiuzhaigou management that addressed both structural and non- in Sichuan or Zhongdian, renamed “Shangri-La,” in Yunnan structural components. Among the non-structural have successfully embraced tourism as their alternative components that included f ood forecasting, resettlement income source. from high f ood risk areas, and f ood detention basin The success of these two programs in compliance and compensation, the Chinese government also participation coupled with belief in their ecological benef ts enthusiastically initiated the Natural Forest Protection has led the GOC to expand its forestry programs to Program (NFPP) and Sloping Lands Conversion Program conserve further watershed services. More recently, the (SLCP) to address upper watershed impacts. These Forest Ecological Benef t Compensation Fund (FEBCF) was programs, f rst piloted then approved for the period 2000 approved. First conceived of in the late 1980s, the FEBCF to 2010, banned logging of natural forests on 30 million ha achieved legal backing in April 1998, initiated pilot work in and supported reforestation of 15 million ha of sloping 2001, and was expanded across China in 2004, providing cropland, respectively, along the upper Yangtze and upper payments of 5 rmb (7 rmb = US$) per mu to maintain and middle Yellow rivers. The original investment of the natural forest land (Sun and Chen 2000). Land demarcated SLCP reached nearly US$40 billion through annual grain as “state ecological forests” based on a suite of mostly and cash payments to farmers for retiring and reforesting ecological criteria are subject to use restrictions and cropland, while the GOC has allocated over $11 billion for associated payment, spanning both state- and collective- the NFPP over the same timeframe, numbering these owned forests. Provincial governments are encouraged to

continues 108 RETENTION OF NUTRIENTS AND SEDIMENT BY VEGETATION

Box 6.1 continued supplement the basic subsidy payments, and wealthy management (i.e., enforcement patrols and management provinces such as Guangdong have structured payment bureau administration). Indeed, Jinxiu Count y obtains systems above and beyond the central government roughly 50% of its annual f scal revenue from these supported amounts. watershed payments (Lu, pers. comm.). In Jinxiu County, Guanxi, which spans the biologically Thus, the GOC has reacted swiftly to address the f ood diverse Dayaoshan region in the upper watershed of the mitigation issue within critical basins largely because of a Pearl River basin and eventually drains to the economic fundamental belief in the myriad of ecosystem services powerhouses of Guangzhou, Shenzhen, Macao, and Hong forest systems provide. By expanding the benef ts beyond Kong, seven surrounding counties paid Jinxiu County a the ecological, these programs have been able to get a total of over 2 280 000 rmb in 2009 for water and soil strong toehold, enabling them to play an increasing role in conservation supplied by its forests with a potential 10% China’s current socio-economic transformation. However, increase per year. With little scientif c information of with limited information of service provisioning and an quantity or quality of water provisioning based on land use unclear understanding of trade-offs especially under or management actions, these counties have continued complex hydrological conditions, it is increasingly diff cult payments for over 20 years. The payments, combined with to ensure the continuation or expansion of payments and Central Government FEBCF investment, provide important programs. This, in fact, may ultimately impact achievement funds for local poverty alleviation and nature reserve of China’s broader goals and objectives.

when considering ranked or normalized sub-catch- nutrient processes as a one-dimensional, well- ment nutrient export. This, in fact, is what we see. mixed stream at the daily time step. This could be a Several conclusions emerge from the map of these major reason for this dissemblance since soluble ranking groups. First, there is general agreement phosphorus could be used by riparian vegetation, between the calibrated SWAT results and our model algae, and other biota in the stream, which could outputs in much of the Willamette River Basin, reduce the amount of soluble phosphorus reaching which is characterized by relative f at terrain on the the outlet of the watershed. The source of the devia- valley f oor and forest or shrubland. Next, the dif- tion between the calibrated SWAT outputs and our ferences between the model outputs are most dra- outputs in the Cascades is less obvious, though it is matic along the banks of the Willamette River (in not a result of the different assumed LULC patterns the middle of the map) and on the slopes of the across the models as these areas are dominated by Coastal Mountains (on the left edge of the map). forest. Along the Willamette River (groups 5 and 7), which includes sub-catchments with signif cant cropping 6.8 Next Steps systems and agricultural land, our model predicts much higher phosphorus loading than that pre- We are taking several steps to strengthen the models dicted by calibrated SWAT. This dissemblance could described in this chapter to ensure that they can be due to a number of factors, including SWAT’s adequately serve users’ need. First, we continue to assumption of homogeneous LULC within HRUs, validate the outputs of our nutrient and sediment which could deviate signif cantly from the actual retention models through comparison with observed LULC pattern in the sub-catchment, which is well data. We are running the model in different water- represented in our model. Indeed, our model uses sheds in different eco-regions to test the patterns of the LULC pattern on the landscape without any nutrient and sediment transport with different lumping and routes the pollutant loading from each LULC types, distributions, and patterns under dif- LULC to the downslope parcels. This latter point is ferent slopes. This will allow us to characterize not represented in SWAT, which directly links all model strengths and weaknesses vis-à-vis these dif- HRUs to streams. SWAT simulates in-stream ferent conditions and factors. We are also adjusting NEXT STEPS 109 the model to allow for the inclusion of biological Dearmont, D., McCarl, B. A., and Tolman, D. A. (1998). pathogens (e.g., fecal coliform). To achieve this goal, Costs of water treatment due to diminished water qual- we have included the option of adding the pathogen ity: A case study in Texas. Water Resources Research , 34 (4), source map, pathogen concentration (intensity) and 849–54. vegetation f ltration eff ciency for this pathogen to Desvousges, W. H., Smith, V. K., and Fischer, A. (1987). Option price estimates for water quality improvements: the nutrient pollutant model. A contingent valuation study for the Monongahela Finally, the models described above are functions River. Journal of Environmental Economics and Management , of several parameters whose values are drawn from 14 (3), 248–67. a distribution in reality although they have been Easton, Z. M., Fuka, D. R., Walter, M. T., et al . (2008a). described as deterministic throughout the chapter. Re-conceptualizing the Soil and Water Assessment Tool We are currently taking steps to incorporate uncer- (SWAT) model to predict saturation excess runoff from tainty into our models, which will allow the users to variable source areas. Journal of Hydrology , 348 (3–4), understand the impacts of this uncertainty on model 279–91. outputs. For a limited set of input parameters, we Easton, Z. M., Walter, M. T., and Steenhuis, T. S. (2008b). will allow the user to explore the impacts of param- Combined monitoring and modeling indicate the most Journal eter uncertainty on model outputs in one of two effective agricultural best management practices. of Environmental Quality , 37 , 1798–1809. ways. The f rst methodology involves treating each Endreny, T. A. (2002). Forest buffer strips: Mapping the parcel in the landscape as a realization of an under- water quality benef ts. Journal of Forestry , (January/ lying distribution. The second involves providing a February), 35–40. conf dence interval as well as a point estimate for Englin, J., and Cameron, T. A. (1996). Augmenting travel use in supply estimation. cost models with contingent behavior data. Environmental and Resource Economics , 7 (2), 133–47. References Foster, H. S. J., and Beattie, B. R. (1981). On the specif cation of price in studies of consumer demand under Agnew, L. J., Lyon, S., Gérard-Marchant, P., et al . (2006). block price scheduling. Land Economics , 57 (2), 624–9. Identifying hydrologically sensitive areas: bridging the Gotaas, H. B. (1956). Composting: Sanitary disposal and rec- gap between science and application. Journal of lamation of organic wastes, Colombia Ubiversity Press, Environmental Management , 78 (1), 63–76. New York. Arnold, J. G., and Fohrer, N. (2005). SWAT2000: Current Hanley, N., Bell, D., and Alvarez-Farizo, B. (2003). Valuing capabilities and research opportunities in applied the benef ts of coastal water quality improvements watershed modeling. Hydrological Processes , 19 (3), using contingent and real behavior. Environmental and 563–72. Resource Economics , 24 (3), 273–85. Asia Development Bank (ADB). May 2006. Poverty Hernandez, M., Kepnerb, W. G., Goodrich, D. G., et al . Revolution in Key Forestry Conservation Programs. (2008). The use of scenario analysis to assess water ecosystem Final Draft Report. services in response to future land use change in the Athayde, D. N., Shelley, P. E., Driscol, E. D., et al . (1983). Willamette River Basin, Oregon. IOS Press, Amsterdam, Results of the nationwide urban runoff program: Final report . The Netherlands. US Environmental Protection Agency, Water Plannning Hjelmfelt, A. T., Piest, R. P., and Saxon, K. E. (1975). Division, Washington, DC. Mathematical modeling of erosion on upland areas. In Baker, M., Weller, D., and Jordan, T. (2006). Improved Congress of the 16th International Association for Hydraulic methods for quantifying potential nutrient interception Research , pp. 40. by riparian buffers. Landscape Ecology , 21 (8), 1327–45. Holmes, T. (1988). Soil erosion and water treatment. Land Bennett, Michael. (2008). China’s sloping land conversion Economics , 64 (3), 356–66. program: institutional innovation or business as usual. Horton, R. E. (1940). An approach toward a physical inter- Ecological Economics 65: 699–711. pretation of inf ltration-capacity. Soil Science Society of Bingner, R. L., and Theurer, F. D. (2007). Research: AGNPS. America Proceedings , 5 , 399–417. US Department of Agriculture–Agricultural Research Service . Leavesley, G. H., Lichty, R. W., Troutman, B. M., et al . [Online]. Available: http://www.ars.usda.gov/Research/ (1983). Precipitation-runoff modeling system: user’s manual . docs.htm?docid=5199 Accessed 10 October 2008. US Geological Survey, Washington, DC. 110 RETENTION OF NUTRIENTS AND SEDIMENT BY VEGETATION

Leggett, C. G., and Bockstael, N. E. (2000). Evidence of the Schneiderman, E. M., Steenhuis, T. S., Thongs, D. J., et al . effects of water quality on residential land price. Journal (2007). Incorporating variable source area hydrology of Environmental Economics and Management , 39 (2), into a curve-number-based watershed model. 121–44. Hydrological Processes , 21 , 3420–30. Liu, Jianguo, Shuxin Li, Zhiyun Ouyang, Christine Tam, Sharma, S. K., Mohanty, B. P., and Zhu, J. T. (2006). and Xiaodong Chen. (2008). Ecological and Socio- Including topography and vegetation attributes for economic effects of China’s policies for ecosystem serv- developing pedotransfer functions. Soil Science Society of ices. Proceedings of the National Academy of Sciences. America Journal , 70 , 1430–40. 105 (28): 9477–9482. State Forestry Administration (SFA). (2003). Sloping Land Moore, W. B., and McCarl, B. A. (1987). Off-site costs of soil Conversion Program Plan (2001–2010). (In Chinese.) erosion: a case study in the Willamette valley. Western Steenhuis, T. S., Winchell, M., Rossing, J., et al . (1995). SCS Journal of Agricultural Economics , 12 (1), 42–9. runoff equation revisited for variable-source runoff Page, T., Haygarth, P. M., Beven, K. J., et al . (2005). Spatial areas. Journal of Irrigation and Drainage Engineering , variability of soil phosphorus in relation to the topo- 121 (3), 234–8. graphic index and critical source areas: Sampling for Stone, R. P., and Hilborn, D. (2000). Universal soil loss equa- assessing risk to water quality. Journal of Environmental tion (USLE): Factsheet. Government of Ontario, Ministry of Quality , 34 , 2263–77. Agriculture, Food and Rural Affairs. [Online]. Available at: Palmieri, A., Shah, F., Annandale, G., et al. (2003). Reservoir http://www.omafra.gov.on.ca/english/engineer/ conservation , vol. 1: The RESCON approach . World Bank, facts/00-001.htm#tab3a Accessed May 2008. Washington, DC. Thompson, J. A., Pena-Yewtukhiw, E. M. and Grove, J. H. Reckhow, K. H., Beaulac, M. N., and Simpson, J. T. (1980). (2006). Soil-landscape modeling across a physiographic Modeling phosphorus loading and lake response under uncer- region: topographic patterns and model transportabil- tainty: a manual and compilation of export coeff cients . US ity. Geoderma , 133, 57–70. UNEP Assessment of 1998 Environmental Protection Agency, Washington, DC. Yangtze Floods. 1999. Reid, L. M., and Dunne, T. (1996). Rapid evaluation of sediment White, W. R. (2001). Evacuation of sediments from reservoirs . budgets . Catena Verlag GMBH, Reiskirchen, Germany. Thomas Telford, London. Renard, K. G., Foster, G. R., Weesies, G. A., et al . (1997). Wischmeier, W. H., and Smith, D. (1978). Predicting rainfall Predicting soil erosion by water: a guide to conservation erosion losses: a guide to conservation planning . USDA-ARS planning with the revised universal soil loss equation Agriculture Handbook, Washington, DC. (RUSLE). USDA Agriculture Handbook 703. USDA, Xu, J. and Cao Y. (2002). On sustainability a of converting Washington, DC. farmland to forests/grasslands. International Economics Roose, E. (1996). Land husbandry—components and strategy , Review 22: 56–60 (in Chinese), Zong, Yongqiang and 70 FAO Soils Bulletin. Food and Agriculture Organization Xiqing Chen. 2000. The 1998 Flood on the Yangtze, of the UN, Rome, Italy. China. Natural Hazards 22: 165–184. CHAPTER 7 Terrestrial carbon sequestration and storage

Marc Conte, Erik Nelson, Karen Carney, Cinzia Fissore, Nasser Olwero, Andrew J. Plantinga, Bill Stanley, and Taylor Ricketts

7.1 Introduction (REDD; Ebeling and Yasue 2008 ), and replacing annual crops with perennials ( Fargione et al . 2008 ). Ecosystems help regulate Earth’s climate by adding Concerns about climate change have led to both and removing greenhouse gases (GHGs) such as regulated and voluntary emissions reduction. Regu- carbon dioxide (CO2 ) from the atmosphere (IPCC lated markets such as the European Union Emissions 2006). Terrestrial ecosystems currently store four Trading System (EU ETS; Victor et al. 2005 ) and the times more carbon than is found in the atmosphere Regional Greenhouse Gas Initiative (Burtraw et al . (3 060 versus 760 gigatons (Gt); see Lal 2004 ). 2006 ), and voluntary offset markets like the Chicago Changes in land use and land cover (LULC) due to Climate Exchange provide a forum in which land- timber harvesting, land-clearing for agriculture, owners can generate carbon credits, through incre- and f re can release substantial amounts of terrestri- ased carbon sequestration on their lands, that can ally stored carbon. For example, tropical deforesta- be bought by entities looking to offset their own tion was responsible for 15–25% of the globe’s total emissions (Marechal and Hecq 2006). Further, the GHG emissions in the 1990s (including fossil-fuel Kyoto Protocol allows signatory nations to claim use and other land-use change; see Gibbs et al. GHG emission reductions by funding afforestation (2007 ) and the Technical Summary in IPCC (2007a)). and reforestation projects in developing nations In Africa, deforestation accounted for nearly 70% of (e.g., Pfaff et al. 2000 ). Finally, discussion at the 2007 the continent’s total GHG emissions at the end of United Nations Framework Convention on Climate the twentieth century ( Gibbs et al. 2007 ). Change Conference focused on developing f nan- We can mitigate the expected economic damages cial incentives to reward developing nations for due to climate change by slowing down GHG accu- reducing emissions by avoiding deforestation and mulation in the atmosphere ( Stern 2007 ). This fact forest degradation (e.g., Mollicone et al. 2007 ; has intensif ed global interest in enlarging or at least Ebeling and Yasue 2008 ). maintaining the size of the terrestrial carbon pool Given the growing interest in carbon markets and (e.g., IPCC 2007a, Lehmann 2007 ). A variety of land offset programs, decision-makers need a way to management techniques can be used to achieve this simply and quickly understand how much carbon end, including lengthening harvest rotation time in is held in landscapes today, and how storage and plantation forests (e.g., Sohngen and Brown 2008 ), sequestration will change under different manage- planting trees to restore forests (reforestation; e.g., ment options. Such information will help managers Canadell and Raupach 2008 ), planting trees in aban- f nd and create opportunities for additional carbon doned croplands (afforestation; e.g., Nilsson and sequestration and associated payments in their Schopfhauser 1995 ), improving soil management landscapes. We present two tiers of relatively sim- (e.g., Schuman et al. 2002 , Lal 2004 ), reducing forest ple models that address these needs by using mini- f re and forest disease risk (e.g., Brown et al. 2002 ), mal data to estimate carbon sequestration and reducing deforestation and forest degradation storage on a landscape.

111 112 TERRESTRIAL CARBON SEQUESTRATION AND STORAGE

Unlike other ecosystem services, there is no dif- woody understory). Belowground biomass is ference between the supply and use of carbon the root system of the aboveground biomass. sequestration and storage (see Chapter 3 for def ni- Sequestration and storage accounting in the soil tions of “supply” and “use” used throughout the pool is generally concerned with soil organic car- book); every unit of sequestered GHG emission will bon (SOC) in mineral soils; in certain land cover allow us to avoid some economic damage that types, however, SOC in organic soils (e.g., wet- would have occured otherwise (although current lands, peatlands, rice paddies) is the dominant soil and proposed sequestration and storage markets carbon (Post and Kwon 2000 ). The other organic only pay for a portion of the avoided damage; see matter pool includes plant litter and dead wood. below). Further, because GHGs uniformly mix in The HWPs pool includes all carbon stored in prod- the atmosphere the whole world benef ts from a ucts made with wood removed from the landscape unit of sequestration regardless of where it occurs. (e.g., furniture, paper, charcoal; Harmon et al. 1 9 9 0 ; After presenting our ecosystem service supply S m i t h et al. 2 0 0 6 ) . model, we present an approach for estimating the L e t Caj , C bj , C sj , a n d C oj indicate the metric tons of economic value of this supply. We also discuss how carbon stored per hectare (Mg of C ha –1 ) , i n t h e we might approximate carbon offsets and avoided aboveground, belowground, soil, and other organic emission credits with our model. matter pools of LULC j respectively, where j = 1 , Gathering the data needed to estimate total car- 2, . . . , J indexes all LULC found on the landscape. bon sequestration and storage on a landscape, LULC types can simply indicate land use and or including species-specif c growth rates, associated cover (e.g., conifer forest, cropland) or include carbon storage, and species compositions, can be a other landscape details that affect pool-storage val- time-intensive effort. The IPCC has developed an ues such as time since disturbance (e.g., conifer for- approach that assigns carbon stocks to different est 120 years or more since a clear-cut, cropland LULC categories based on a meta-analysis of f eld that is plowed annually), extent of disturbance storage studies (IPCC 2006). The models presented (e.g., heavily disturbed conifer forest due to illegal in this chapter are grounded in this approach. In timber harvest), soil properties (cropland on clay our simplest storage and sequestration model, soil; e.g., Torn et al. 1 9 9 7 ) , a n d c l i m a t i c c o n d i t i o n s tier 1, we associate carbon storage with different (conifer forest below 1000 m, where elevation bands LULC categories and each LULC category is proxy for precipitation gradients; see McGuire et al. assumed to be in storage equilibrium at any point in 2 0 0 1 ; R a i c h et al. 2 0 0 6 ) . A d d i n g s p e c i a l i z e d L U L C time ( Section 7.2.1 ). In tier 1, the amount of carbon classes such as these can help add more reality to sequestered in a parcel is the difference in steady- the estimates of this steady-state model. For exam- state storage at two points in time (Section 7.2.2 ). In ple, using age-specif c forest classes with f ne this simple model, carbon sequestration is not regis- enough temporal resolution can help represent tered in a parcel unless the parcel’s LULC changes additional carbon sequestration as forests mature or it is subject to harvest of wood. and change LULC classes.

The carbon stored in a parcel’s HWPs pool, Cp , is the sum of the carbon in the woody material 7.2 Tier 1 supply model removed from the parcel in the past (even if the removed material eventually leaves the parcel) less 7.2.1 Carbon stored on a landscape the decay over time in the products that were made We disaggregate terrestrial carbon storage into f ve from this woody material (which releases carbon in pools: (1) aboveground biomass, (2) belowground the form of CO2 ) and the carbon-equivalent emis- biomass, (3) soil, (4) other organic matter, and sions from manufacturing and transporting the (5) harvested wood products (HWPs). Aboveground f nal products. Let biomass is composed of all living plant material Wx −1 above the soil (e.g., grass, herbaceous material, CCfqpx= ∑ qqs hqx hqx() qx,,− qx (7.1) bark, tree trunks, branches, leaves, and other q=0 TIER 1 SUPPLY MODEL 113

Annual amount of carbon in wood qhqChq removed from parcel that reaches HWPs processing stage (Mg of C)

Cp = Aggregate height of all bars (Mg of C) Carbon remaining in timber Carbon remaining plantation (Mg of C) from removed q0 q1 q2q3 q4 q5 q6 q7 q8 q9 q10q11q12

First Year Today of Harvest Time

W = 13

Figure 7.1 Calculating the amount of carbon stored in a parcel’s harvested wood products (HWPs) pool.

Wood harvested from a parcel is converted into various products (e.g., furniture, f rewood), which burn or decay at a given rate over time, returning CO 2 to the atmosphere. In the graph, the carbon remaining in the HWPs made from wood harvested 13 years ago is given by the height of q 12 ’s bar, and so on. The sum of the bars’ heights indicates the amount of carbon still stored in HWPs made from that parcel’s wood removed over the past 13 years. The rate at which the height of the bars decays over time is a function of the decay rates associated with the product mix made with the harvested wood.

where C px is measured in metric tons of C in parcel sum of the carbon stored in each pool in the parcel x , where x = 1, 2, . . . , X indexes all unique parcels on at time t ,

J a landscape, Chqx gives the metric tons of C in the woody biomass when it was removed from parcel x CCxt=+ pxt∑ ACCCC xjt() aj +++ bj sj oj , (7.2) j=1 q years ago, θ hqx is the portion of the harvested woody biomass removed q years ago from x that where C pxt is parcel x ’s HWPs pool-storage level at makes it to the initial wood processing stage, f ( ω , qx time t , A xjt is the area of LULC j in parcel x a t t i m e t , J q) gives the fraction of θ hqx C hqx still stored in wood and AA= (parcel area does not change over xx∑ j=1 jt products q years after removal from x , ω qx is the time). If data or interest in some pools is lacking, the decay rate of the products made with wood removed model can be used with any subset of the f ve car- from x q years ago, and σ qx is the carbon-equivalent bon pools. To determine the metric tons of C stored emissions in metric tons associated with the pro- across the whole landscape at time t , symbolized by duction and distribution of the product made from Ct , we sum all parcel-level carbon storage values, C hqx . The variable W x indicates how many years in X the past we want to account for harvest on parcel x . CCtx= ∑ t . (7.3) An illustration of Eq. (7.1) is given in Figure 7.1 and x=1 Section 7.2.5 . Appropriate inclusion of the HWP See Glenday ( 2006 ) and Ruesch and Gibbs ( 2008 ) for pool requires thoughtful accounting, given its sig- applications of Eq. (7.2). nif cant storage potential and the emissions associated with a product’s life cycle (Niles and Schwarze 7.2.2 Carbon sequestered by a landscape 2001 ; USEPA 2009). The carbon stored on a parcel at time t , given by Carbon sequestration represents an increase in car-

C xt and measured in metric tons of C, is equal to the bon storage over time. In the tier 1 model, carbon 114 TERRESTRIAL CARBON SEQUESTRATION AND STORAGE sequestration or loss in a parcel is registered when tracting this from the parcel’s sequestration from the parcel’s LULC mix changes, its management or time t to T under an alternative, offset program- Off, A production of HWPs changes, or its HWPs pool is inf uenced LULC mix as of time T , ∆CxtT : not in equilibrium. Otherwise, for the sake of mode- Off Off,, A Off B ling simplicity, in tier 1 we assume storage in a parcel ∆CCCxtT = max{} 0, ∆−∆xtT xtT . (7.6) is in equilibrium or steady state and will not change over time despite any evidence to the contrary. Such new, program-inf uenced scenarios are not To determine the amount of carbon sequestered generated by our models, but rather can be assessed in a parcel from year t to year T (t < T), given by by our models to reveal the additional carbon ben-

∆C xtT , we use Eq. (7.2) to calculate the amount of car- ef ts associated with the offset program. bon stored in a parcel in year t , C xt , and in year T , Two other carbon market issues are permanence

C xT, which accounts for any changes in the parcel’s and leakage, and both are topics of intense policy LULC or HWP management or production between debate (e.g., Brown 2002 ; Vohringer et al. 2 0 0 6 ;

t and T , and then subtract C xt from C xT , M u r r a y et al. 2 0 0 7 ) ( B o x 7 . 1 ) . P e r m a n e n c e i s a c o n - cern because trees and other biomass eventually

∆CCCxtT= xT− xt , (7.4) decay; f res and illegal logging can occur regardless of management efforts; soil can be disturbed, releas- where ing trapped carbon back into the atmosphere; and

X offset providers may decide to convert their forest to ∆CCtT= ∑ ∆ xtT (7.5) another use even with an existing offset contract. x=1 Therefore, when a landowner is compensated for gives the change in carbon storage from t to T over additional sequestration, we might say that they are the entire landscape. renting a temporary benef t to society. How much

I n t h i s c a s e , Cxt is the point of reference or base- society should pay for this temporary benef t and at line for determining whether net sequestration has what point temporary sequestration becomes occurred from t t o T. I f ∆ CxtT o r ∆ CtT is positive, permanent sequestration in a climatic sense is con- then sequestration has occurred from t t o T in the tested (e.g., Chomitz 2002 ; Marechal and Hecq 2006). parcel or landscape, respectively. If they are nega- Leakage is an additional concern with offset, rather tive, carbon has been lost between t a n d T . See than regulatory, programs because when a land- Cairns et al. ( 2000 ) and Glenday ( 2006 ) for applica- owner decides to manage her land for additional tions of Eq. (7.4). sequestration instead of clearing it for agriculture or urban development the economic pressure to clear the land for other uses will not disappear. Instead it 7.2.3 Additional sequestration may be “leaked” or displaced to other parts of the in offset markets landscape or globe where subsequent clearing will Markets and programs that give carbon credits to release stored carbon, decreasing the amount of landowners in exchange for enhancing terrestrial additional sequestration. Leakage is accounted for carbon sequestration on their lands currently only in this model if it occurs within the study landscape recognize additional sequestration in market-eligi- and between the years t a n d T , so broadening both ble pools (see Box 7.1 for an illustration). Additional spatial and temporal scales can reduce the issue. sequestration is the amount of sequestration above and beyond the sequestration that would have 7.2.4 Preventing emissions from deforestation occurred in the absence of a carbon offset market or and degradation program (i.e., baseline sequestration). Additional sequestration in a parcel’s eligible pools from t to T , REDD is a policy mechanism, currently under Off ∆CxtT , can be approximated in our model by calcu- debate, that rewards efforts to reduce deforestation, lating a parcel’s sequestration from time t to T under forest degradation, and their associated carbon Off, B its baseline LULC mix as of time T , ∆CxtT , and sub- emissions in developing countries (e.g., Ebeling and TIER 1 SUPPLY MODEL 115

Box 7.1 Noel Kempff case study: capturing carbon f nance

Bill Stanley and Nicole Virgilio environmental changes (IPCC 2000). In the case of NKCAP, the project provided carbon f nancing to stop logging in the Policies that constrain carbon dioxide (CO ) emissions, and 2 park and deforestation around communities. Without this include provisions for trading carbon sequestered in forests funding, these activities would have continued, leading to and other ecosystems, are anticipated to be the largest the loss of forest cover and release of carbon dioxide. driver of payments for carbon capture and storage and are a major topic of discussion in the development of national and international climate change policies and legislative Key factor 2: baseline language. If successful, these programs could generate The business-as-usual, or “without-project” scenario, is also billions of dollars of annual funding for forest carbon called the baseline , and its development should be among projects. the f rst steps to assessing the carbon benef ts of a project. The Noel Kempff Mercado Climate Action Project Baselines are essentially predictions, or future projections (NKCAP) in Bolivia, borne of a partnership between The based empirically upon historical information or a Nature Conservancy and Fundación Amigos de la performance benchmark, of what may have happened had Naturaleza with funding from industry and contributions the project not been put into place. The success of any forest by the Bolivian Government, is one example of such a carbon emissions reduction project will depend on the project, designed to simultaneously address climate estimated carbon storage in the baseline or benchmark, change, conserve biodiversity, and bring sustainable subtracted from the performance of the project itself. Most development benef ts to local communities by avoiding recently, in the context of programs being considered within logging and agricultural land conversion. The 1.5 the context of the United Nations Framework Convention on million-acre project, which began in 1997 and expanded Climate Change for activities to reduce emissions from the Noel Kempff Mercado National Park, is the largest deforestation and degradation (REDD), several national- effort of its kind. It is expected to prevent the release of up scale baseline emission scenarios have been proposed. to 5.8 million tons of CO into the atmosphere over 30 2 These proposals vary widely, and the baseline approach years. As with any carbon sequestration project, there were established as policy will have a signif cant impact on the a number of issues that had to be addressed to ensure emissions reductions that could be claimed. To determine high-quality carbon offsets. the total emission reductions that were additional to business-as-usual activities at Noel Kempff, baseline rates of deforestation were developed by GIS analysis of a time Key factor 1: additionality series of satellite photos from 1986, 1992, and 1996–7 and Nearly all voluntary frameworks, as well as the few then were applied in a spatially explicit land-use change compliance programs in place or proposed, require that model to project future deforestation. The rates and models project-based emissions reductions and removals must be are to be reassessed periodically. Also, a national-scale beyond “business as usual” to be credited (e.g., Alig and economic model of Bolivian timber markets was created to Butler 2004 , UNFCCC 1995, UNCCCS 1997, and several develop baseline rates of logging and f eld work was US Senate bills including the Forest Resources for the undertaken to determine the amount of emissions Environment and the Economy Act (S. 1547), the associated with the logging. International Carbon Sequestration Incentive Act (S 2540), and the Lieberman–Warner Climate Security Act (S.3036)). Key factor 3: leakage The guidelines suggest that a scenario of what would have happened without the project must be developed for Leakage has been def ned as “the unanticipated decrease comparison against project management. The difference in or increase in GHG benef ts outside of the project’s carbon storage and other greenhouse gas emissions accounting boundary . . . as a result of project activities” (GHGs) between the two represents the GHG impacts that ( Chapter 5 of IPCC 2000). A straightforward example of are truly additional rather than simply the result of leakage from conservation is where a farmer who is incidental or non-project factors such as recent market or seeking to clear a plot of uninhabited forest for conversion

continues 116 TERRESTRIAL CARBON SEQUESTRATION AND STORAGE

Box 7.1 continued to agricultural f elds is told that the land has been made pests, f re, and political turnover all have the potential to part of a carbon project and is now protected. Leakage release carbon sequestered in the forest (IPCC 2000). Thus, occurs if the farmer moves to the next available land and permanence must be taken into consideration during the clears that forest instead. The project would simply displace project planning stage, targeting areas where land is likely the farmer’s activity and emissions, and not result in any to remain intact indef nitely and using approaches like

real reductions of atmospheric CO2 . The risk of leakage can permanent conservation easements and sustainable be minimized through thoughtful project design, taking forestry, which will reduce likelihood that the carbon local conditions into consideration and addressing the storage will be lost, or that those losses will be sustained, underlying drivers of deforestation ( Schwarze et al. 2002 ). over time. Carbon offset policies generally establish For example, NKCAP attempted to limit leakage just liability for losses. To ensure against these losses one tool outside the project area by working with the bordering is to set aside “buffer” credits that can be drawn upon community to develop a sustainable management plan, should the stored carbon be unexpectedly lost. To ensure with which they applied for legal land title through the the permanence of NKCAP, the national park was Bolivian government. This reduced the risk of uncontrolled expanded to include the project area and a permanent forest conversion ( Aukland et al . 2002 ). Market leakage endowment was established to fund protection activities that was not avoided was estimated using an economic into the future. model ( Sohngen and Brown 2004 ) and deducted from the claimed carbon savings. Conclusion Key factor 4: permanence Obviously, there are many factors that must be considered in the successful employment of forest carbon emissions O n e o f t h e f rst issues that critics of forest carbon crediting reductions programs. Additionality, baseline, leakage, and will cite is the possibility for the reversal of the carbon permanence are challenges, but pilot projects such as sequestration. Project success is dependent upon keeping NKCAP have shown that it is possible to address and the forest healthy and standing. Illegal logging, invasive overcome them.

Yasue 2008 ). In this case, if deforestation and degra- it can remain as is or be converted to a LULC with Loss dation that would have occurred under a business- lower carbon storage.) Let ∆CxtT indicate the case as-usual LULC trajectory is prevented, then where ∆ C xtT < 0. ecosystems retain additional carbon, thereby avoid- The expected loss in a parcel’s carbon stock due ing climate change-related economic damages that to conversion from LULC j ' to some other LULC by were originally expected. time T is the product of the difference in carbon

Here we illustrate how we can use our models to stock ∆C xtT and the overall probability that the forest approximate the carbon credits that could be gener- in parcel x will be converted at some point between ated under REDD or similar programs to compen- t and T , π xtT , under the business-as-usual baseline. sate avoided forest loss and degradation. This Let there be i = 1, 2, . . . , I events that can cause forest illustration relies on several assumptions. First, we type j' in parcel x to be converted by T. If we assume assume that forest type j ' ’s carbon stock in parcel x that the probability of each event occurring, given is in steady state at time t . Next, we assume that if by πxtT , i , under the baseline is independent and addi- forest type j' is cleared from x in order to establish tive, then LULC j, the carbon storage at time T is lower than it ⎧⎫I was at time t (i.e., C > C ). Let the change in stor- min 1, . (7.7) xt xT ππxtT= ⎨⎬∑ xtT, i ⎩⎭i=1 age on x be given by ∆ C xtT , where ∆C xtT = 0 if forest type j' remains on the parcel as of time T and ∆C xtT < The expected amount of aboveground biomass car- 0 if the forest is cleared. (We assume that there are bon storage in parcel x at time T under the baseline only two possible future states for parcel x at time T : is given by TIER 1 SUPPLY MODEL 117

Loss EC[]xT=+ C xtp xtT∆ C xtT . (7.8) Eastern Arc Mountains of Tanzania and their watersheds. The mountains contain over 1000 endemic In order to achieve a net reduction in expected car- species and are widely considered a global conser- bon emissions at time T , the parcel’s manager must vation priority (Burgess et al. 2007 ). They also pro- intervene such that expected storage in x a t t i m e T vide a range of ecosystem services, including water ˆ ˆ is greater than E [C xT ] . L e t p xtT a n d p xtT, i r e p r e s e n t for drinking, agriculture, and hydropower, carbon the overall probability and specif c event probabil- sequestration, and non-timber forest products ity, respectively, that parcel x w i l l b e c o n v e r t e d ( Ndangalasi et al. 2007 ; Mwakalila et al. 2009 ). The from j ' by time T under alternative management map used for this analysis estimates land cover as ˆ where p xtT is calculated similarly to π xtT ( E q . ( 7 . 7 ) ) . of 1995 ( Valuing the Arc 2008 ) and consists of 590 Finally, the net reduction in expected emissions in LULC categories (each LULC type is a unique com- parcel x i s g i v e n b y bination of land cover, elevation range, and terrestrial ecoregion ( Olson et al. 2001 )). ∆CECECAvoid = [][]ˆ − xtT xT xT (7.9) For mean aboveground ( C ), belowground ( C ), ˆ Loss a b = min{} 0, (ppxtT−×∆ xtT )C xtT , and soil (C s ) pool values, we average across reported where values for each LULC type from various sources (see the chapter’s SOM for values and their sources ˆ Loss EC[]xT=+ C xtpˆ xtT∆ C xtT . (7.10) and details on the 1995 land cover map). Low and high pool values for each LULC are set equal to the In the exposition above we assume that data on lowest and highest values observed in the literature. LULC conversion probabilities at the parcel level We calculated the HWP pool only for the forest are available. In practice, only historical conversion plantation LULC type. We used data from Makundi rates at the country level are available globally. For ( 2001 ) and IPCC (2006) and an assumption of even- example, national estimates of annual deforestation age rotation forestry to calculate mean, high, and rates are available for the periods 1990–2000 and low C p , C a , C b , and C s values for the plantation LULC 2000–5 for most nations ( FAO 2005 ). However, we (see the chapter’s SOM for details). may be able to predict conversion rates at the parcel The tier 1 models produce maps of per-hectare level using basic economic principles. Under such a mean, lower, and upper bound estimates of aggre- framework whether forested parcel x will be con- gate storage in the biomass, soil, and HWPs pools verted to another LULC in the future will be a func- as of 1995 (we ignore the other organic material tion of the net value of conversion, the availability pool) ( Figure 7.2 ; Plate 3). The highest densities of of substitutable parcels for conversion, the enforce- stored carbon occur in the tropical montane forests ment of property rights, labor and capital con- found at the highest elevations of the Arc Mountains. straints, and other socio-economic variables. In this These areas remain forested today, amid continued chapter’s supplementary online material (SOM), agricultural conversion in the lower elevations. we introduce one method for estimating conversion Many of the areas of high carbon storage are already ˆ probabilities π xtT and p xtT as a function of predicted protected by national parks or forest reserves, but gross value of production following deforestation unprotected forests on the edges of protected areas in x and its surrounding parcels, x ’s distance to the remain. Several aggressive reforestation projects— nearest road, and x’s distance to the nearest source undertaken mostly to garner carbon offset pay- of labor and capital (e.g., a population center). ments—have begun, which could increase stored carbon values in parts of this landscape over time. Our assumption that forest plantations are the 7.2.5 Tier 1 example: terrestrial carbon storage only LULC types that produce HWPs is the greatest in the Eastern Arcs Mountain Watershed, shortcoming of this illustrative example. Non- Tanzania plantation forests in this watershed provide fuel We illustrate our tier 1 carbon storage model as wood, charcoal, and timber for many households described in Eq. (7.2) on a landscape def ned by the on the landscape (e.g., Luoga et al. 2000; ECCM 118 TERRESTRIAL CARBON SEQUESTRATION AND STORAGE

K e n y a T a n z a n i a

Dar Es Salaam Lower bound

Carbon stock (Mg/ha)

0–88 150–313 783–1,122 Upper bound 88–150 313–783

Figure 7.2 Tier 1 carbon storage estimates for 1995 in Tanzania’s Eastern Arcs Mountains and their watersheds. The polygons formed with the dark lines represent Eastern Arc Mountain blocks, which rise from the surrounding woodlands and savannas. These blocks were once largely forested, but now consist of a mixture of agriculture, forest, and woodlands. Gray lines are major rivers. Black squares represent major cities. Timber plantations cover approximately 0.3% of the study landscape. Spatially explicit land cover and other landscape data are from the Valuing the Arc project (2008; Mwakalila 2009). See the chapter’s SOM for details on data used in the maps. (See Plate 3.)

2007; Ndangalasi et al. 2007 ). If we could obtain bet- temperature increases to expected global or regional ter estimates on these harvests we could include the net economic damage over time (e.g., Nordhaus

Cp pool for more forest types and adjust their C a and 1992 ; Mastrandrea and Schneider 2004 ). Damages

C b values accordingly. In addition, plot level meas- in these models include property damage due to ures of standing carbon stocks being taken in this rising sea levels and storm occurrence, the net region will be used to groundtruth our model changes in crop yields due to drier and hotter cli- estimates and improve input values for the above- mates, f shery damage due to ocean acidif cation, ground carbon pool (see Box 7.2 ). and the net gain in human mortality and morbidity due to greater disease prevalence in a warmer world 7.3 Tier 1 valuation model: an avoided ( Stern 2007 ). economic damage approach The social cost of carbon (SCC) is the marginal cost, manifested in the economic damages due to By slowing GHG emissions, terrestrial carbon the resultant higher atmospheric GHG concentra- sequestration and storage decrease the severity of tion, of emitting an additional metric ton of C, all future climate change and its associated damages. else equal. Global IAMs provide SCC estimates. Integrated assessment models (IAMs) estimate Mathematically, the societal value of sequestra- these potential damages by relating rising GHG tion on parcel x across the years t to T is given by concentrations in the atmosphere and expected multiplying x ’s supply of sequestration (positive or TIER 1 VALUATION MODEL: AN AVOIDED ECONOMIC DAMAGE APPROACH 119

Box 7.2 Valuing the Arc: measuring and monitoring forest carbon for offsetting

Andrew R. Marshall and P. K. T. Munishi Tree volume Carbon offsetting programs and associated policy tools The most important measurement for estimating tree (REDD, MAC, CERs, etc.) require accurate quantif cation of volume is the diameter at breast height (dbh, typically terrestrial carbon stocks. While there are continuing efforts 1.3 m) (e.g., Kuebler 2003 ). Where environmental to determine vegetation structure remotely, permanent conditions are thought to cause the dbh-to-height ratio to sample plots (PSPs) of woody vegetation continue to give deviate from observed norms, it is also important to the most reliable estimates ( Brown 2002 ; Gibbs et al . measure tree height. This can be time-consuming and it is 2007 ). These estimates can both calibrate remotely sensed advisable to simply measure a representative sample (e.g., measures and provide inputs to models like those described 10 from each size class; 10–20, 20–30, 30–40, 40–50, and in this book. The many environmental and human impacts > 50 cm dbh). The dbh-to-height relationship in the sample on carbon storage levels in forests must be considered can then be determined from regression analysis and when interpreting PSP data. Outputs of PSPs can also be extrapolated to the remaining stems in each PSP. combined with data on other ecosystem services produced The above measures are used to estimate the volume of a by the forest to determine the overall economic value of a tree bole, but other components of the tree are more forest ecosystem. complicated. Calculating the volume of branches and roots The Valuing the Arc project, funded by the Leverhulme requires destruction of the trees, so that they can then be Trust, is doing just that in the Eastern Arc Mountains of placed in water to observe the volume displaced. Estimating Tanzania, one of the world’s most important areas for the volume of foliage and dead wood requires litterfall traps, species richness and endemism ( http://www.valuingthearc. but litterfall is highly variable among species, elevations, and org ). The combined importance of carbon and biodiversity seasons. Because of these complications, most studies either production in a forest is likely to be far more appealing to focus on the bole carbon alone or apply a simple expansion policy-makers than biodiversity alone. factor to adjust the f nal carbon estimate (see below).

Plot location and size Calculating tree biomass and carbon content There are many considerations for planning the location of PSPs and the methods to use for estimating its carbon The next step is to calculate the biomass of a tree, given its content. First, PSP locations need to be representative of estimated volume. Many equations have been employed for the area. This will require prior knowledge of the variation this purpose, and these vary according to habitat and in environmental and human inf uences. In montane location, due to variations in environmental conditions and habitats such as those found in the Eastern Arc Mountains, tree allometry ( Chave et al. 2005 ). All approaches, however, elevation is the primary predictor of vegetation involve multiplying tree volume by its wood density composition and tree allometry, particularly the height to (g cm –3 ). The standard measure of wood density is wood diameter ratio (Marshall et al. in prep.). For this reason a specif c gravity (WSG), which is the oven-dry mass divided stratif ed random design for locating plots is advisable. The by green volume. WSG is typically determined by taking a size and number of plots should provide a balance core from a live tree, or a section of a freshly dead tree. between time and resources to sample the maximum WSG for many species is unknown. Studies therefore often range of environmental conditions. One-hectare plots have use estimates from the most closely related species, genus, been adopted by many international projects, but may be or family. impractical for short-term projects or for sampling broad Around 50% of the biomass in woody vegetation is environmental gradients. composed of carbon; therefore, the f nal amount of The majority of aboveground carbon in forest habitats is elemental carbon is calculated by simply multiplying contained within the large trees and lianas (as much as biomass by 0.5. If this calculation is limited to the tree bole, 95% in some tropical forests; Chave et al . 2003 ). A huge IPCC recommends expansion factors of 0.24, 0.25, and amount of measurement and taxonomic identif cation time 0.05 for the carbon in branches, roots, and foliage, can be saved in limiting surveys to these stems. respectively (IPCC 2003). These have mostly been derived

continues 120 TERRESTRIAL CARBON SEQUESTRATION AND STORAGE

Box 7.2 continued from studies carried out in association with commercial sinks by woody biomass, sinks by soil, and releases by logging, as felled trees can be physically measured. human activity. Accurate calculation is also important for There are many sources of error from the various stages the development of a market that pays landowners for of calculation. In terms of allometry some studies have reduced emissions from deforestation and degradation incorporated measurements from the bole to buttress (REDD). Interest in REDD markets is driven by the fact that

edge (e.g., Glenday 2006 ). Lianas and stranglers also 20% of global CO2 emissions come from forest destruction pose problems due to their unpredictable shape. and degradation, and that much of this is occurring in Estimates of wood density are also highly dependent on developing countries that have limited funds and capacity the availability of data, and the reliability of using to address it. A key question is whether the developing estimates from closely related species is debated. world is already, or can be assisted to become, ready for Furthermore the impact of human disturbance on forest REDD. In some countries this answer is likely to be no due ecosystems is particularly hard to quantify. Despite the to inadequate policy and environmental law enforcement. many sources of error, published estimates for intact More positively, Tanzania may be among the developing tropical forests are usually relatively consistent between countries with highest potential for carbon trading and studies (typically 100–250 t ha–1 ) and precision within entering the REDD market in coming years. Tanzania is 10% of the mean is usually possible (Brown 2002 ). politically stable, with an advanced forest policy and Remote methods including optical, radar, or laser sensors legislation, has huge forest resources, and has for many are becoming more widely used, but do not have the years been attractive for international development accuracy and consistency of inventories using vegetation funding. It also has one of the world’s most extensive plots (e.g., Gibbs et al. 2 0 0 7 ) . protected area networks, with nearly 40% of land area covered by National Parks and various other forms of reserved land ( Chape et al. 2008 ). These combined factors Implementation of offset programs have led the Norwegian government to approve a US$100 million grant for the implementation of REDD in Tanzania. Accurate calculation of carbon emissions is paramount to There is also considerable interest from other nations, the determining national terrestrial carbon budgets, including United Nations, and the World Bank.

negative) over that time period by the SCC, dis- X VAD= V . (7.12) counting all annual values to year t values, and then tT∑ xtT x=1 summing across all discounted values: In Eqs. (7.11) and (7.12) we assume every unit of T −1 ∆CSCC VAD xzz,,++ 1 z 1, (7.11) sequestration after time z , even if it is just tempo- xtT = ∑ zt− zt= (1+ r ) rary, is valuable to society because it reduces the stock of carbon in the atmosphere, thereby (margin- w h e r e VA D xtT (value of avoided damage) is the present value of all economic damage avoided (or ally) mitigating or delaying climate change and additional damage caused if negative) due to car- related damages associated with the continuation of C levels into the future (i.e., C is the storage bon sequestration on x ( e m i t t e d f r o m x) from time xz xz baseline in year z ). However, if we only want to t t o T . I n E q . ( 7 . 1 1 ) ∆C x , z , z + 1 measures the metric tons of C sequestered (emitted) on x between year value the supply of additional sequestration or avoided emissions then ∆ C in Eq. (7.11) could z a n d z + 1 (if we only have estimated ∆C or x, z , z +1 xtT Off sequestration for several time steps in between t be replaced by an annualized ∆Cx (see Eq. (7.6)) or Avoid and T , we can approximate annual sequestration ∆Cx (see Eq. (7.9)). Finally, by indexing the SCC with z in Eq. (7.11), we allow it to change over time. with linear interpolation), SCC z +1 is the SCC in year z + 1 , a n d r i s t h e r e a l ( i n f ation-adjusted) There is some expectation that the SCC will grow discount rate. The landscape-level analog of over time as the marginal impact of avoided emissions becomes even more valuable in a world deal- VA DxtT i s g i v e n b y TIER 2 SUPPLY MODEL 121 ing with a rapidly changing climate (e.g., the IPCC per annum in Stern (2007 ) and 2–4% in Weitzman assumes that the SCC will grow at a rate of 2.4% per ( 2007 ) versus a “typical” cost–benef t analysis rate annum; see Chapter 20 of IPCC 2007b). of 7%) whether due to intergenerational equity con- It is not clear whether the real (in f ation-adjusted) cerns (Stern 2007 ) or concerns over the highly discount rate in the denominator of Eq. (7.11), unlikely but particularly disastrous scenarios of r, should match the discount rate embedded in catastrophic change ( Weitzman 2007). Other econo- our chosen SCC (see below). The denominator of mists argue that in order to avoid a potentially mas- Eq. (7.11) discounts all future costs and benef ts from sive misallocation of monetary resources, climate

∆ Cx , z , z + 1 o v e r z Î [ t , T ] t o t ’s value of money, allowing change mitigation activities should be judged with for the aggregation of costs and benef ts incurred the discount rate we use to judge all other policies over time. The real discount rate used in Eq. (7.11) (i.e., 5 to 10% per annum; e.g., Nordhaus 2007 ). measures our preference for more immediate con- Further, the SCC is a marginal cost—that is, it meas- sumption over investment and our expectations ures the economic damage avoided for very small regarding economic growth. For short time spans, and changes in carbon emissions relative to total global when comparing VA D tT t o o t h e r b e n e f ts and costs real- stock. While signif cant amounts of carbon are at stake ized in the near term, we may want to use a real dis- at landscape scales, most applications of this model count rate closer to an observed market discount rate. will only involve small changes in carbon stocks rela- There remains much uncertainty and debate over tive to any emission baseline. For our purposes, then, the appropriate value of the SCC (e.g., Nordhaus SCC is a legitimate and useful estimate of social cost. 2007 ; Stern 2007 ; Weitzman 2007 ). Any SCC esti- Finally, the SCC will invariably differ from the mate is a function of the assumptions in its source market price for carbon offsets as the market-clear- IAM, including the trajectory of global GHG emis- ing price of an offset (the price that sellers and buy- sions over time and the impact that climate change ers agree on) has no functional relationship to the will have on the social welfare of people around the social value of carbon sequestration. In regulated world today and in the future. As a result, estimates markets, buyers and sellers will settle on a price of SCC vary widely. By using an average or median that is a function of the offset provider’s cost of par- SCC estimate, however, we can avoid assuming ticipating in the offset program and the buyer’s some of the more idiosyncratic predictions of global willingness to pay for an offset (a function of GHG emission, economic, and demographic trends. Tol abatement costs in the industrial, electrical, and (2009 ) surveyed the peer-reviewed SCC literature transportation sectors and emission caps). In volun- and found representative current estimates that tary markets, several other factors may impact range from $46 to $91 Mg –1 of C but with a large price, as the motivation of offset buyers is not solely variance (values are given in 1995 dollars). to achieve least-cost emission reduction (see Conte Another cause of divergence in SCC estimates is and Kotchen (2009 ) for further discussion of deter- disagreement over the appropriate discount rate to minants of voluntary offset prices). use when determining the SCC in an IAM. The discount rate, which combines expectations about 7.4 Tier 2 supply model future economic growth and our preferences for present consumption over future consumption, In the tier 1 supply model, sequestration between equates future climate change-related damages to time t and T only registers in parcel x if x ’s LULC more immediate damages. A lower discount rate mix changes between t and T , x’s wood harvest will lead to higher estimates of the SCC, as future rates or harvest management changes between t events, including the risk of catastrophic change, and T, or x’s HWP carbon pool is not in a steady are more heavily weighted with lower interest rates state at time t . However, terrestrial carbon storage (Chapter 20 of IPCC 2007b). Some economists argue levels, especially on recently disturbed land, tend to that the discount rates used in climate change anal- change continuously due to vegetation growth and ysis should be lower than what is typically used for decay and organic matter accumulation in the soil. cost–benef t analyses (e.g., a discount rate of 1.4% In the tier 2 terrestrial carbon supply model we 122 TERRESTRIAL CARBON SEQUESTRATION AND STORAGE account for any continual changes with carbon ered by LULC type j that was established 11 years sequestration functions. prior to time m on land formally in LULC j' of age

Similar to tier 1, carbon storage in a parcel in the i' a t t h e t i m e o f c o n v e r s i o n t h e n α j´ i´ j 1 1 C a j gives the tier 2 supply model is a function of the carbon aboveground carbon storage levels as of time m . stored in the aboveground biomass, belowground See Figure 7.3 and the SOM for an illustration on biomass, soil, other organic matter, and HWPs the use of carbon sequestration coeff cients (in pools. Unlike tier 1, the tier 2 supply model incor- general all coeff cients behave in the same man- porates a series of terms that account for the sub- ner). See USEPA (2009) for an example of a tier 2 sequent change in terrestrially stored carbon in x approach. after a disturbance or land-use change. Formally, If we are to use tier 2 modeling in conjunction the metric tons of C stored on parcel x in year with an avoided emission analysis (Section 7.2.4 ) m Î [ t , T ] is then the exact timing of deforestation will matter. In other words, we will have to define the CCxm= pxm probability of deforestation in each parcel that is JJ ⎛⎞abCC++ (7.13) forested as of time t for each time period m (i.e., + A jiji′′ aj jiji′′ bj , ∑∑ xj′ jm ⎜⎟ ˆ jj'1== 1 ⎝⎠ghjiji′′ CC sj+ jiji′′ oj πxtm a n d p xtm will need to be defined for each m Î [t , T]) as well as biomass carbon storage at where each pool-speci f c coeff cient (the Greek each time period m in case deforestation is letters) gives the fraction of the pool’s maximum avoided. storage capacity achieved as of time m in an area that transitioned to LULC j i years ago from LULC 7.5 Tier 2 valuation: an application of j' that was i' years old at the time of transition to j the avoided economic damage approach (a LULC is i ' years old if it has been i ' years since the last major disturbance in the area occupied by The valuation approach used for tier 1 is also the LULC). Further, Axj ´jm is the area of parcel x applied to tier 2 carbon sequestration estimates. We that is in LULC j a s o f y e a r m but was previously provide an example of tier 2 carbon sequestration in LULC j ' and the HWPs pool variable Cpxm is the and valuation modeling in a 22 × 20 km landscape same as the tier 1 HWPs pool variable (see Eq. located in northwest Minnesota, USA ( Figure 7.4 ; (7.1)). Assuming that a pool-storage variable in Plate 4). The eastern half of the modeled landscape Eq. (7.13) gives the pool’s maximum storage is dominated by early succession tree stands of capacity, then a value of 1 for its associated aspen, white birch, maple, basswood, and oak. The sequestration coeff cient means that the pool has western half is primarily in row crops and pasture, reached its maximum storage value as of year m . with a smattering of Conservation Reserve Program If for some reason the pool-storage variables in (CRP) perennial grasslands. The developed area in Eq. (7.13) give another reference value (e.g., the the northwest corner of the landscape is the town of average storage value), then the value of the asso- Mahnomen. ciated sequestration coeff cient will need to be I n F i g u r e 7 . 4 w e g i v e t h e y e a r 2 0 0 0 L U L C p a t t e r n recalibrated such that the maximum value of the on the landscape (USDA-FSA 2000; Minnesota coeff cient multiplied by the pool’s reference stor- DNR—Division of Forestry 2000; USDA/NRCS age value equals the pool’s maximum storage 2008). We also generate two visions of land use by capacity. 2050 (not pictured). The Carbon Sequestration In our approach a series of pool-speci f c coeff - Scenario includes the restoration of 6 km2 o f p r a i r i e cients approximates the pool’s sequestration potholes on cropland, 6 km2 of afforestation on crop- function after a change to LULC j. For example, land, 17 km2 of cropland conversion to pasture, and 2 α j´ i´ j 1 , α j´ i´ j 6 , α j´ i´ j 1 1 , . . . describes the relative change in 39 km of cropland to perennial grassland by 2050 aboveground carbon storage levels every 5 years (we convert the least valuable croplands to these beginning the year after LULC j ' o f a g e i ' t r a n s i - new uses and assume all other parcels retained their tioned to LULC j . I f p a r c e l x i s c o m p l e t e l y c o v - year 2000 LULC). Conversely, the CRP Loss Scenario TIER 2 VALUATION: AN APPLICATION OF THE AVOIDED ECONOMIC DAMAGE APPROACH 123

(a) (b) Soil carbon storage 70 years after grassland establishment 1.75 1.00 1.00 (gcrop, 3, grass, 5Cs,grass) Soil carbon Soil Carbon Storage sequestration ) ) 1.50 between t and T 0.75 0.75 ( g grass,i’,crop,i

s,grass Soil carbon storage s,grass Soil carbon storage Soil carbon storage 5 C 50 years after grassland C 50 years after grassland years after conversion to 1.25 0.50 establishment 0.50 establishment cropland (g C ) (g C ) (g C ) crop, 3, grass, 4 s,grass crop, 3, grass, 4 s,grass crop, 5, grass, 1 s,crop ) crop,i’, grass,i crop,i’, crop,i’, grass,i crop,i’, 1.00 g g ( 0.25 Soil carbon storage ( 0.25 Soil carbon storage Soil carbon storage when grassland Soil Carbon Storage when grassland was established was established 0.75 0.00 0.00 50 yrs 50 yrs t Time t t + 20T Time 25 yrs

Figure 7.3 A tier 2 model illustration: carbon sequestration in soil from years t to T . Assume a parcel is in grassland at time t . Assume the grassland was established 50 years ago on land that had been in row crops. Over time the carbon in the parcel’s soil accumulates according to the sequestration curve in (a) (in this illustration soil sequestration rates are given for discrete steps in time). By time t the carbon stored in the soil of each hectare of the parcel has reached γ crop, 3, grass, 4 C s, grass = 0.9 × Cs, grass , where Cs, grass is grassland’s per hectare maximum storage potential, and the carbon sequestration coeff cient γ crop, 3, grass, 4 indicates the portion of the storage maximum that has been reached in a grassland that is in its fourth age-class bin since conversion from cropland in its third age-class bin at the time of conversion. Now suppose the parcel converts back to cropland 20 years after time t , or 5 years before T (i.e., T = t + 25). Between t and t + 20 soil carbon levels continue to increase according to the grassland sequestration curve (the light gray curve in (b)). The conversion to cropland at time t + 20 causes an immediate and signif cant loss of soil carbon (the initial vertical portion of cropland’s sequestration curve given by the black curve in (b)). At time T the soil carbon level in the parcel is given by γ pg, 5, crop, 1 C s, crop = 1.45 × Cs, crop , where Cs, crop is cropland’s per hectare maximum storage capacity and γ pg, 5, crop, 1 is the soil carbon sequestration coeff cient for cropland in its f rst age bin at time T that was in grassland’s f fth age bin at the time of conversion from grassland to cropland. Finally, sequestration in a hectare of this parcel from time t to T is given by 1.45 × C s, crop – 0.9 × Cs, grass (a negative value).

assumed 877 ha of the perennial grassland (CRP sequestration is $6 036 494 (SD $5 606 984) or $892 land) that existed in 2000 converts to either row ha− 1 . In contrast, mean present value under the CRP crops or hay production by 2050 (we assume all Loss Scenario is $19 043 572 (SD$17 634 696). other parcels retained their year 2000 LULC). Including only the parcels that experienced LULC We use tier 2 models to evaluate the consequences change (1.94% of the landscape), the mean mone- of the two land-use change scenarios on the soil car- tary value of soil sequestration is $210 952 −1 bon pool (C s). Because estimates of C sj and γ j´ i´ j i are (SD$219 300) or $241 ha . Either way it is summed, uncertain, we estimate distributions for each of the Carbon Sequestration Scenario results in more these model inputs using data from Smith et al. carbon sequestration, worth roughly $6 million in ( 2006 ), Anderson et al. ( 2008 ), and Nelson et al. avoided economic damages, than the CRP Loss ( 2009 ). We also use distributions (instead of point Scenario. See this chapter’s SOM for more details estimates) for the year of LULC conversion on the on scenario creation and tier 2 model variables. parcels that changed LULC, the discount rate ( r ), In this example, we use variable distributions to and the SCC ( Tol 2009 ). For each scenario, we simu- account for some of the uncertainty in storage val- late VADxtT for all x on the landscape and associated ues, sequestration rates, SCC, the market discount

VAD tT 1000 times, each time drawing a unique value rate, and dates of LULC transitions or disturbances from our distributions for model parameters that in a scenario. However, we do not address several are uncertain. potential biases in our modeling. First, the dynam- The mean present value of economic damage ics of carbon storage and sequestration are complex avoided because of soil carbon sequestration under and greatly simplif ed, even in the tier 2 models; the Carbon Sequestration Scenario is $25 114 953 whether or not output produced with a simplif ed (SD$23 244 250). If we only consider the parcels that model is systematically biased when compared to experience LULC change under this scenario (15% output from more detailed carbon sequestration of the landscape), the mean monetary value of soil models is an issue that warrants further investigation 124 TERRESTRIAL CARBON SEQUESTRATION AND STORAGE

(a) 22 km 20 Km

Row crops Oak Alfalfa Maple/basswood Grass, pasture, Aspen/White Birch and range (inc. CRP) Lowland Shrub deciduous Wetlands/Marsh Water Roads/urban/ barren

(b) Carbon Sequestration Scenario CRP Loss Scenario Mean Values

Present value per hectare < 0 0 1–441 442–602 603–928

Sample Draw > 928

Figure 7.4 The value of carbon sequestered in soil across two alternative LULC scenarios. (a) The year 2000 landscape. (b) The per-hectare monetary value of carbon sequestration in soil from 2000 to 2050 for each LULC scenario. The top row of maps gives mean results across all model simulations. The bottom rows of maps give the results from one particular run of the model. The black outlines on the parcels indicate parcels that experience LULC change in some portion of its area at some point between 2000 and 2050. The Carbon Sequestration Scenario map ref ects a program of afforestation, restoring prairie pothole, and converting row crops to pasture and perennial grassland. In the CRP Loss Scenario any parcel that was primarily in CRP in 2000 was converted to row crops or a hayf eld by 2050. (See Plate 4.)

(see Chapter 15 ). Second, measurement error and 7.6 Limitations and next steps non-standardized sampling methods of storage and sequestration rates in the f eld may introduce sys- 7.6.1 Limitations tematic bias in storage and sequestration rate data Our analysis of carbon sequestration and storage used in our models ( Brown 2002 ). Third, because across a landscape is limited by several constraints. the map used in this analysis and others like it rep- First, our models are driven by patterns of LULC resents a simplif cation of the actual landscape, and change in these patterns over time. While LULC additional error and potentially systematic bias are change is likely the dominant factor in determining introduced into model results. terrestrial carbon sequestration and storage (IPCC LIMITATIONS AND NEXT STEPS 125

2006), the nature and rate of disturbance events are sophisticated models. For scenarios of LULC change also important (e.g., see IPCC 2006; Bond-Lamberty in California that consider climate change ( Shaw et al. 2007 ; Kurz et al. 2008 ). Prairie and forest f res, et al . 2009 ), we are comparing our results to those forest disease outbreak, and exotic species invasion derived from a more detailed, process-based model can have signif cant impacts on storage and seques- of carbon storage used in Shaw et al . ( 2009 ). We will tration, but may not be ref ected in typical LULC compare our simple results in Tanzania (see Section maps. Our models do not explicitly consider such 7.2.5 ) to more detailed models developed by the events, but using more nuanced LULC classif ca- Valuing the Arc Programme ( Mwakalila et al . 2009 ). tions may begin to capture them (e.g., diseased These comparisons will allow us to estimate the conifer forest, disease-free conifer forest). accuracy of our simple approach and the limitations Second, our models largely ignore local varia- and biases it presents. tions in climate, which also have a signif cant impact on carbon storage and sequestration rates 7.6.2 Next steps (e.g., McGuire et al. 2 0 0 1 ) . R a i n f a l l a n d t e m p e r a t u r e patterns determine such ecosystem processes as The models presented in this chapter can be used net primary productivity (NPP) and soil erosion, for many purposes. We can layer maps of sequestra- the primary inputs in terrestrial carbon storage tion and storage and their value with other ecosys- capacity and sequestration rates. The best strategy tem service maps to identify areas of ecosystem for minimizing this limitation is to use “local” car- service synergies and trade-offs on the landscape bon sequestration and storage data when possible (see Chapter 14 ). We can also compare sequestra- and to limit the extent of the modeled landscape in tion supplies and values to those of other ecosystem order to minimize over-extrapolation of carbon service across a landscape to determine the oppor- estimates. Another way to deal with this limitation tunity costs of increasing carbon sequestration in is to stratify LULC categories by landscape features both biophysical and economic terms (e.g., Jackson that may ref ect microclimate. This requires, how- et al. 2005 ; Nelson et al. 2008 ). ever, observations of model parameters for each of Further, we can estimate the additional carbon the climate-related strata created within each broad sequestration on a landscape by comparing a baseline LULC type. Further, as a climate changes, vegeta- scenario of LULC with an alternative projection that tion patterns may shift on the landscape. To account ref ects efforts to sequester carbon or reduce emis- for this we can modify LULC types on future sions (see the SOM for further details). By overlaying scenario maps according to climate–vegetation these maps with an opportunity cost map we can models. See Chapter 17 for such an exercise. begin to predict market prices for offsets and avoided Finally, land-use management practices that do emissions credits ( Kindermann et al. 2008 ), and to not have a large impact on carbon storage in plant construct additional sequestration supply curves for biomass and soils can still be signif cant sources of different policies (e.g., Lubowski et al. 2006 ). Please other GHG emissions. For example, crop residue see the SOM for descriptions of our efforts to validate burning (e.g., IPCC 2006), livestock use (e.g., IPCC the biophysical and deforestation risk models. 2006), and various soil management practices on Finally, these models can serve as the foundation farms (e.g., Bouwman et al. 2002 ) can signif cantly for a tier 3 approach to carbon storage and seques- increase the rate of methane and nitrous oxide emis- tration modeling. A tier 3 sequestration and storage sions, two powerful GHGs (see Chapter 9 ). In addi- model would not only consider how land use and tion, all of the GHGs emitted by fossil fuel-burning conversion decisions effects storage, but it would machinery used to support farms and other LULC also simulate the affect of climate and landscape- are not reported except for the production and level disturbance stochasticity on biomass growth delivery of HWPs (see Eq. (7.1)). and carbon formation in and release from soils. In part to understand the consequences of our Models such as CENTURY (Parton et al. 1992 ) and simple model and its corresponding constraints, we the vegetation model LPJ ( Sitch et al. 2003 ) are exam- have begun to validate its results against more ples of tier 3 models. 126 TERRESTRIAL CARBON SEQUESTRATION AND STORAGE

References Chave, J., Condit, R., Lao, S., et al . (2003). Spatial and temporal variation of biomass in a tropical forest: results Alig, R. J., and Butler, B. J. (2004). Projecting large-scale from a large census plot in Panama. Journal of Ecology, area changes in land use and land cover for terrestrial 91 , 240–52. carbon analyses. Environmental Management, 33 , 443–56. Chomitz, K. M. (2002). Baseline, leakage and measure- Anderson, J., Beduhn, R., Current, D., et al. (2008). The ment issues: how do forestry and energy projects com- potential for terrestrial carbon sequestration in Minnesota: a pare? Climate Policy, 2 , 35–49. report to the Department of Natural Resources from the Conte, M. N., and Kotchen, M. J. (2009). Explaining the Minnesota Terrestrial Carbon Sequestration Initiative . price of voluntary carbon offsets. NBER Working Paper University of Minnesota, St. Paul. 15294. Aukland, L., Sohngen, B., Hall, M., et al. (2002). 2001 Ebeling, J., and Yasue, M. (2008). Generating carbon Analysis of leakage, baselines, and carbon benef ts for the f nance through avoided deforestation and its potential Noel Kempff Climate Action Project . Winrock International, to create climatic, conservation and human develop- Arlington, VA. ment benef ts. Philosophical Transactions of the Royal Bond-Lamberty, B., Peckham, S. D., Ahl, D. E., et al . (2007). Society B: Biological Sciences, 363 , 1917–24. Fire as the dominant driver of central Canadian boreal Edinburgh Centre for Carbon Management (ECCM). forest carbon balance. Nature, 450 , 89–92. (2007). Establishing Mechanisms for Payments for Carbon Bouwman, A. F., Boumans, L. J. M. and Batjes, N. H. (2002). Environmental Services in the Eastern Arc Mountains,

Modeling global annual N2 O and NO emissions from Tanzania . fertilized f elds, Global Biogeochemical Cycles, 16 , 1080. FAO. 2005. State of the World’s Forests 2005 . FAO, Rome. Brown, S. (2002). Measuring, monitoring, and verif cation Also available at http://www.fao.org/docrep/011/ of carbon benef ts for forest-based projects. Philosophical i0350e/i0350e00.HTM . Transactions of the Royal Society of London, Series Fargione, J., Hill, J., Tilman, D., et al . (2008). Land clearing A-Mathematical Physical and Engineering Sciences, 360, and the biofuel carbon debt. Science , 319 , 1235–8. 1669–83. Gibbs, H. K., Brown, S., Niles, J. O., et al. (2007). Monitoring Brown, S., Swingland, I. R., Hanbury-Tenison, R., et al . and estimating tropical forest carbon stocks: making (2002). Changes in the use and management of forests REDD a reality. Environmental Research Letters, 2 , 0 4 5 0 2 3 . for abating carbon emissions: issues and challenges Glenday, J. (2006). Carbon storage and emissions offset under the Kyoto Protocol. Philosophical Transactions of potential in an East African tropical rainforest. Forest the Royal Society of London Series A-Mathematical Physical Ecology and Management , 235 , 72–83. and Engineering Sciences, 360 , 1593–605. Harmon, M. E., Ferrell, W. K., and Franklin, J. F. (1990). Burgess, N. D., Butynski, T. M., Cordeiro, N. J., et al . (2007). Effects on carbon storage of conversion of old-growth The biological importance of the Eastern Arc Mountains forests to young forests. Science, 247 , 699–702. of Tanzania and Kenya. Biological Conservation, 134 , Intergovernmental Panel on Climate Change (IPCC). 209–31. (2000). IPCC special report on land use land-use change, and Burtraw, D., Kahn, D., and Palmer, K. (2006). CO2 allow- forestry . Cambridge University Press, Cambridge, UK. ance allocation in the regional greenhouse gas initiative Intergovernmental Panel on Climate Change (IPCC). and the effect on electricity investors. Electricity Journal, (2003). Good practice guidance for land use, land-use change 19 , 79–90. and forestry . Institute for Global Environmental Strategies Cairns, M. A., Haggerty, P. K., Alvarez, R., et al . (2000). (IGES), Hayama, Kanagawa, Japan. Tropical Mexico’s recent land-use change: A region’s Intergovernmental Panel on Climate Change (IPCC). contribution to the global carbon cycle. Ecological (2006). 2006 IPCC guidelines for national greenhouse gas Applications, 10 , 1426–41. inventories. Institute for Global Environmental Strategies, Canadell, J. G., and Raupach, M. R. (2008). Managing for- Hayama, Kanagawa, Japan. ests for climate change mitigation. Science, 320 (5882), Intergovernmental Panel on Climate Change (IPCC). 1456–7. (2007a). Climate change 2007: mitigation of climate change: Chape, S., Spalding, M., and Jenkins, M., Eds. (2008). The Contribution of Working Group III to the Fourth Assessment worlds protected areas: status, values and prospects in the Report of the Intergovernmental Panel on Climate Change, 21st century . University of California Press, Berkeley. Cambridge University Press, New York. Chave, J., Andalo, C., Brown, S., et al. (2005). Tree allome- Intergovernmental Panel on Climate Change (IPCC). try and improved estimation of carbon stocks and bal- (2007b). Climate change 2007: impacts, adaptation and vul- ance in tropical forests. Oecologia, 145 , no. 1, 87–99. nerability: contribution of Working Group II to the Fourth LIMITATIONS AND NEXT STEPS 127

Assessment Report of the Intergovernmental Panel on version of intact and non-intact forests. Climatic Change, Climate Change, Cambridge University Press, New 83 , 477–93. York. Murray, B. C., Sohngen, B., and Ross, M. (2007). Economic Jackson, R. B., Jobbagy, E. G., Avissar, R., et al . (2005). consequences of consideration of permanence, leakage Trading water for carbon with biological sequestration. and additionality for soil carbon sequestration projects. Science, 310 , 1944–7. Climatic Change, 80 (1), 127–43. Kindermann, G., Obersteiner, M., Sohngen, B., et al . (2008). Mwakalila, S., Burgess, N. D., Ricketts, T., et al . (2009). Global cost estimates of reducing carbon emissions Valuing the Arc: linking science with stakeholders to through avoided deforestation. Proceedings of the National sustain natural capital. Arc Journal, 23, 25–30. Academy of Sciences, 105 , 10302–7. Nelson, E., Mendoza, G., Regetz, J., et al. (2009). Modeling Kuebler, C. (2003). Standardized vegetation monitoring proto- multiple ecosystem services, biodiversity conserva- col . Centre for Applied Biodiversity Science, Conser- tion, commodity production, and tradeoffs at land- vation International, Washington, DC. scape scales. Frontiers in Ecology and the Environment, Kurz, W. A., Dymond, C. C., Stinson, G., et al . (2008). 7 , 4–11. Mountain pine beetle and forest carbon feedback to cli- Nelson, E., Polasky, S., Lewis, D. J., et al . (2008). Eff ciency mate change. Nature, 452 , 987–90 . of incentives to jointly increase carbon sequestration Lal, R. (2004). Soil carbon sequestration impacts on global and species conservation on a landscape. Proceedings of climate change and food security. Science, 304 , 1623–7. the National Academy of Sciences, 105 , 9471–6. Lehmann, J. (2007). A handful of carbon. Nature, 447 , Ndangalasi, H. J., Bitariho, R. and Dovie, D. B. K. (2007). 143–4. Harvesting of non-timber forest products and implica- Lubowski, R. N., Plantinga, A. J., and Stavins, R. N. (2006). tions for conservation in two montane forests of East Land-use change and carbon sinks: econometric estima- Africa. Biological Conservation, 134 , 242–50. tion of the carbon sequestration supply function. Journal Niles, J. O., and Schwarze, R. (2001). The value of careful of Environmental Economics and Management, 51 , 135–52. carbon accounting in wood products. Climatic Change , Luoga, E. J., Witkowski, E. T., F., and Balkwill, K. (2000). 49 , 371–6. Economics of charcoal production in miombo wood- Nilsson, S. and Schopfhauser, W. (1995). The carbon-se- lands of eastern Tanzania: some hidden costs associated questration potential of a global afforestation program. with commercialization of the resources. Ecological Climatic Change, 30 , 267–93. Economics 35 , 243–57. Nordhaus, W. D. (1992). An optimal transition path for McGuire, A. D., Sitch, S., Clein, J. S., et al . (2001). Carbon controlling greenhouse gases. Science, 258 , 1315–19. balance of the terrestrial biosphere in the twentieth cen- Nordhaus, W. D. (2007). Critical assumptions in the Stern tury: Analyses of CO2, climate and land use effects with review on climate change. Science, 317 , 201–2. four process-based ecosystem models. Global Biogeo- Olson, D. M., Dinerstein, E., Wikramanayake, E. D., et al . chemical Cycles, 15 , 183–206. (2001). Terrestrial ecoregions of the worlds: A new map Makundi, W. R. (2001). Carbon mitigation potential and of life on Earth. Bioscience, 51 , 933–8. costs in the forest sector in Tanzania. Mitigation and Parton, W. J., McKeown, B., Kirchner, V., et al . (1992). Adaptation Strategies for Global Change, 6 , 335–53. CENTURY users manual . NREL, Colorado State Maréchal, K., and Hecq, W. (2006). Temporary credits: University, Fort Collins, Colorado. A solution to the potential non-permanence of carbon Pfaff, A. S. P., Kerr, S., Hughes, R. F., et al. (2000). The Kyoto sequestration in forests? Ecological Economics, 58 , 699–716. protocol and payments for tropical forest: An interdisci- Marshall, A. R., Lewis, S., Lovett, J. C., Burgess, N., et al . plinary method for estimating carbon-offset supply and (In preparation). Variation in carbon storage and tree increasing the feasibility of a carbon market under the allometry with elevation in the eastern arc. CDM. Ecological Economics, 35 , 203–21. Mastrandrea, M. D., and Schneider, S. H. (2004). Post, W. M., and Kwon, K. C. (2000). Soil carbon sequestra- Probabilistic integrated assessment of dangerous tion and land-use change: processes and potential. climate change. Science, 304 , 571–5. Global Change Biology, 6 , 317–27. Minnesota Department of Natural Resources (DNR)— Raich, J. W., Russell, A. E., Kitayama, K., et al . (2006). Division of Forestry (2000). GAP Land Cover Map of Temperature inf uences carbon accumulation in moist Minnesota. Minnesota Department of Natural Resources, tropical forests. Ecology, 87 , 76–87. St. Paul, MN. Ruesch, A. S., and Gibbs, H. K. (2008). New IPCC tier-1 Mollicone, D., Achard, F., Federici, S., et al . (2007). An global biomass carbon map for the year 2000. Available incentive mechanism for reducing emissions from con- online from the Carbon Dioxide Information Analysis 128 TERRESTRIAL CARBON SEQUESTRATION AND STORAGE

Center (http://cdiac.ornl.gov ), Oak Ridge National Torn, M. S., Trumbore, S. E., Chadwick, O. A., Vitousek, Laboratory, Oak Ridge, Tennessee. et al . (1997). Mineral control of soil organic carbon stor- Schuman, G. E., Janzen, H. H., and Herrick, J. E. age and turnover. Nature, 389 , 170–3. (2002). Soil carbon dynamics and potential carbon United Nations Climate Change Convention Secretariat sequestration by rangelands. Environmental Pollution, (UNCCCS). (1997). UNFCCC AIJ Methodological Issues . 116 , 391–6. UNCCCS, Bonn, Germany. Schwarze, R., Niles, J. O., and Olander, J. (2002). United Nations Framework Convention on Climate Understanding and managing leakage in forest-based green- Change (UNFCCC). (1995). Decision 5/CP.1 from Report of house gas mitigation projects . Institute for Environmental the Conference of the Parties on its f rst session, held at Berlin Economics at the Technische Univer sität Berlin, Berlin. from 28 March to 7 April 1995. Addendum. Part two: Action Shaw, R., Pendleton, L., Cameron, R., et al . (2009). The taken by the Conference of the Parties at its f rst session . impact of climate change on California’s ecosystem services . UNFCCC, Berlin, Germany. California Climate Change Center. Draft Paper. United States Department of Agriculture-Natural Sitch, S., Smith, B., Prentice, I. C., et al. (2003). Evaluation Resource Conservation Service (USDA/NRCS). of ecosystem dynamics, plant geography and terrestrial (2008). USDA-NASS Cropland Data Layer . 2006 . carbon cycling in the LPJ dynamic global vegetation USDA/NRCS, Washington, DC. model. Global Change Biology, 9 , 161–85. United States Department of Agriculture-Farm Service Smith, J. E., Heath, L. S., Skog, K. E., and et al . (2006). Agency (USDA-FSA). (2000). Conservation reserve pro- Methods for calculating forest ecosystem and harvested car- gram map of Minnesota. 1997 . USDA-FSA, Washington, bon with standard estimates for forest types of the United DC. States. General Technical Report NE-343, US Department United States Environmental Protection Agency (USEPA). of Agriculture, Forest Service, Northeastern Research (2009). Inventory of US greenhouse gas emissions and sinks: Station, Newtown Square, PA. 1990–2007 . USEPA, Washington, DC. Sohngen, B., and Brown, S. (2004). Measuring leakage Valuing the Arc (2008). Land use and land cover map of from carbon projects in open economies: a stop timber the Eastern Arc Mountains and surrounding watersheds . harvesting project in Bolivia as a case study. Canadian Valuing the Arc, Cambridge, UK. Journal of Forest Research-Revue Canadienne De Recherche Victor, D. G., House, J. C., and Joy, S. (2005). A Forestiere, 34 , 829–39. Madisonian approach to climate policy. Science, Sohngen, B., and Brown, S. (2008). Extending timber rota- 309 , 1820–1. tions: carbon and cost implications. Climate Policy, 8 , Vohringer, F., Kuosmanen, T., and Dellink, R. (2006). How 435–51. to attribute market leakage to CDM projects. Climate Stern, N. (2007). The economics of climate change: the Stern Policy, 5 , 503–16. review . Cambridge University Press, Cambridge, UK. Weitzman, M. L. (2007). A review of the Stern review on Tol, Richard S. J. (2009). The economic effects of climate the economics of climate change. Journal of Economic change. Journal of Economic Perspectives, 23, 29–51. Literature, 45 , 703–24. CHAPTER 8 The provisioning value of timber and non-timber forest products

Erik Nelson, Claire Montgomery, Marc Conte, and Stephen Polasky

8.1 Introduction rights, where only a limited number of individuals or entities have the legal right to harvest (Feder and Forests play an iconic role in environmental conser- Feeny 1991 ). An individual or entity with an exclu- vation campaigns and are habitat to much of the sive right to harvest forest stocks has an incentive to world’s known terrestrial biodiversity ( Repetto and maximize the net present value of economic returns Gillis 1988 ). Forests also provide and regulate many to harvest over time, whereas open-access harvest important ecosystem services (e.g., Williams 2003 ; tends to be characterized by a race to exploit the R i c k e t t s 2 0 0 4 ; T i c k t i n 2 0 0 4; M a a s s et al. 2 0 0 5 ) , i n c l u d - resource. Therefore, forests with well-enforced ing carbon sequestration (e.g., Scholes 1996 ; Sohngen property rights will tend to have lower harvest rates and Brown 2006 ), potable water supply (e.g., Núñez and greater biological stocks at any point in time et al. 2006 ), and a stock of plants and animals that can than open-access forests (e.g., Luoga et al . 2005 ; be used to meet human needs for food, materials, Birdyshaw and Ellis 2007 ; for a discussion on excep- and medicine (e.g., Milner-Gulland and Clayton tions to this general rule see Larson and Bromley 2 0 0 2 ; B e l c h e r et al. 2 0 0 5 ; D a m a n i a et al. 2005 ; 1990 ). N d a n g a l a s i et al. 2007 ). In fact, the legal and illegal As with other models presented in this book we harvest of timber and non-timber forest products present two tiers of analysis. Tier 1 involves (NTFPs) supports and supplements the livelihoods approaches that are analytically simpler and require of millions of families around the world (e.g., Justice less data than the approaches in tier 2. In tier 1 mod- et al. 2 0 0 1 ; P a t t a n a y a k a n d S i l l s 2 0 0 1 ; Ve d e l d et al. els, we assume that the rate of harvest equals the 2004 ; Box 8.1 ). The ability of relatively intact forests harvested stock’s natural regeneration rates, leav- to provide habitat and valuable ecosystem services ing the biological stock unchanged or in steady state year in and year out is the main argument given for over a harvest period. Steady-state harvest implies forest conservation (e.g., Peters et al. 1989 ; Boot and that the forest is being sustainably managed and Gullison 1995 ; Bawa and Seidler 1998 ; Arnold and future harvests of stocks can continue indef nitely. Pérez 2001 ; Silvertown 2004; Sinha and Brault 2005 ). However, in many forested landscapes around the In this chapter we present approaches for modeling world stock levels are declining (e.g., FAO 2006). To the quantity and value of timber and NTFP harvest accommodate the reality of forest degradation, we from forested parcels across a landscape (the role that develop a tier 2 dynamic approach that allows for forests play in the regulation and provision of other an estimation of harvest volume and value when ecosystem services is covered in other chapters of this harvest rates do not necessarily equal stock growth book). Timber and NTFP harvest levels and patterns rates and forest stocks can degrade (or improve) are largely determined by forest ecology and property over time. rights structure. Open-access forests are those in which The focus of this chapter is on forest and forest anyone can harvest forest stocks. At the other extreme, stocks but our approach can be used to model other a forest may be regulated with exclusive harvest land-cover types that provide extractive resources

129 130 THE PROVISIONING VALUE OF TIMBER AND NTFP

Box 8.1 Wildlife conservation, corridor restoration, and community incentives: a paradigm from the Terai Arc Landscape

inbreeding, and breakdown of natural ecological dynamics Eric Wikramanayake, Rajendra Gurung, and behavioral interactions that structure their populations and Eric Dinerstein and communities. The conservation challenge in the TAL is to The Terai Arc Landscape (TAL) extends from Nepal’s Chitwan conserve these species as ecologically, demographically, and National Park to Rajaji Tiger Reserve, India ( Figure 8.A.1 ), genetically viable populations. The strategy is to create a and includes the forests and grasslands along the base and conservation landscape that links twelve protected areas inner valleys of the Himalayas. The landscape was designed with corridors to facilitate dispersal. to conserve metapopulations of Asia’s largest mammals, We used a GIS-based habitat analysis to identify the especially the tiger (Panthera tigris), Greater one-horned potential corridors ( Wikramanayake et al . 2004 ). The rhinoceros (Rhinoceros unicornis), and Asian elephant analysis revealed several bottlenecks in the potential (Elephas maximus ). These grasslands and forests are the network, including six restoration priorities and three most fragmented and converted Himalayan ecosystems transboundary corridors between protected areas in Nepal ( Wikramanayake et al. 2 0 0 1 ) ; t h u s , t h e e n d a n g e r e d s p e c i e s and India ( Figure 8.A.1 ). are mostly conf ned to protected areas where they face an There are several challenges to restoring and uncertain future from the inevitable consequences of genetic maintaining corridors. Over 7 million people live in the TAL

80'00'E 81'00'E 82'00'E 83'00'E 84'00'E 85'00'E

Basanta CHINA (TAR) SUKLAPHANTA WR 29'00'N 29'00'N Laljhadi

Lagga Bagga

BARDIA NP NEPAL

Mahadevpuri KISHANPUR WSDUDHWA NP Khata Lamahi 28'00'N 28'00'N KATARNIGHAT WS Dovan

CHITWAN NP Butwal SOHELWA WS

INDIA SOHAGABARWA WS PARSA WR VALMIKINAGAR TR 27'00'N 27'00'N Legend

CHINA Protected Areas

Foresst Corridors

NEPAL Bottlenecks N 26'00'N 26'00'N International Boundary Chitwan INDIA River

0 50 100 WWF Kilometers WWF Nepal. 2007

80'00'E 81'00'E 82'00'E 83'00'E 84'00'E 85'00'E

Figure 8.A.1 The Terai Arc Landscape, Nepal. INTRODUCTION 131

and tend over 4.5 million head of livestock (MFSC 2004). opportunity for carbon trading in the voluntary market. As Immigration to the Terai still continues, and contributes of 2005, the biogas program qualif ed as a Gold Standard signif cantly to the 2.86% population growth rate (WWF CDM-VER project. 2006). Most immigrants encroach into state forests, clear The transboundary Khata corridor represents an land, and begin to cultivate. After a couple of years, some interesting case. It links Bardia National Park with sell the land and move on to encroach and occupy another Katerniaghat Tiger Reserve, and tigers harbored in both forest patch, creating a chronic process of forest parks now use the corridor. Bardia also supports an degradation. important rhinoceros population. Several rhinos now use Most of the livestock are “scrub” cattle which are the corridor, and f ve rhinoceros from Bardia have begun to allowed to free-range in state forests. As a result, these reside in Katerniaghat. Elephants have also begun to use forests are overgrazed. Controlling cattle grazing in state the corridor. Thus Khata is now a functional wildlife corridor. forest land is diff cult because the forest department is In Khata, the TAL program has promoted community- under-resourced. based enterprises that use forest products. For instance, Bel To restore bottlenecks the TAL program identi f ed and fruits (Aegel marmelos ) are used to produce juice under a facilitated the conversion of strategic corridor areas to community-based project. During the f scal year 2005–6 over community forests, a forest management strategy that has 25 metric tons of fruit were harvested and sold to the worked well to restore forests in Nepal (Nagendra et al. cooperative. Over 17 400 bottles of Bel juice were produced 2005 ). Under community forestry, degraded state forests and marketed, which earned a net prof t of about are assigned to local forest user groups to manage, based US$6000, a considerable sum for rural communities. on plans approved by the Forest Department (Nagendra The TAL program has also promoted mentha (Mentha 2002 ). Because user groups receive management and piperita ) farming, and provided a distillery to extract oil. usufruct rights they have a vested interest in sustainably Mentha was originally promoted as a “live fence” to managing the forests. prevent crop damage by elephants. But the f nancial Even within the context of community forestry, however, benef ts from mentha oil proved to be so lucrative that user groups still require fuelwood and other forest resources. more farmers adopted it as a primary cash crop and the Therefore, alternatives to reduce this demand were necessary. output exceeded the carrying capacity of the distillation The fuelwood demand was eased by providing subsidies to plant, and three additional distillation plants had to be built the communities for biogas plants, which use cattle dung to to accommodate the demand. In f scal year 2008, the produce methane for cooking and lighting. Better cattle income from 80 ha of mentha production was US$59 000. breeds promote stall-feeding, instead of free-grazing, with The TAL program also established a community-based the added incentive of making it easier to collect cattle dung rattan furniture-making enterprise that earns the user for biogas plants. groups over US$3000 annually. The rattan is sustainably Five years after bottleneck restoration began the “big harvested from the community forests. three” species—tigers, rhinoceros, and elephants—have Over 160 men and women from the user groups are begun to use the corridors. Tigers are present in four also involved in community-led anti-poaching operations corridors, rhinoceros in two, and the frequency of elephant in Khata. This group patrols the corridor to safeguard movement has increased in f ve. In some corridors the forests and wildlife from illegal activities such as elephants have returned after a lapse of 50 years, encroachment, logging and collection of forest products, according to local residents. and poaching. Even though several rhinos have been Since 2002, the TAL program has facilitated the poached from within the core of Bardia National Park, conversion of over 193 km2 to community forests in no poaching incidents have occurred in the corridor, strategic areas of six bottlenecks to 196 forest users under the vigilance of the community anti-poaching groups. These forests are being used by over 24 500 units. households. The biogas plants and fuel-eff cient cooking The experiences from the TAL show that restoration and stoves provided from 2001 to 2006 have saved an conservation of wildlife habitat in a larger landscape is estimated 21 000 metric tons of fuelwood annually, possible through local stewardship, as long as the local representing an equivalent of over 1.6 km 2 of clear-cut communities benef t f nancially from their natural resources forests. Biogas as an alternative energy also represents an management. 132 THE PROVISIONING VALUE OF TIMBER AND NTFP under a range of property right structures. For we can generate an estimate of H f r o m E a n d V ; example, open-access grasslands can provide food H = E × V . and fuel for local households as well as a food and The cost of harvesting a stock is comprised of water source for their livestock (e.g., Swallow two components, the wages paid to labor and the and Bromley 1995 ; Thwaites et al . 1998 ; Adger and capital costs associated with harvesting equipment. Luttrell 2000 ). Let the total cost of harvesting the stock in the forest parcel over a harvest period be given by C . I n rare cases, information on costs will be directly 8.2 The supply, use, and value of available. More commonly, cost estimates can be forests’ provisioning service in tier 1 generated by multiplying harvest effort by the sum of hourly average wage rate of labor, W , a n d t h e In our ecosystem service taxonomy only the por- amortized cost of capital equipment used per unit tion of a forest’s stock that is harvested represents of harvest effort, Y (in this cost calculation we the forest’s provisioning service. We place a value ignore the opportunity costs associated with har- on the use of a forest’s provisioning services when vesters traveling to and back from forest parcels we convert the forest’s harvest volume into a mon- and any costs associated with transporting har- etary value. Accurately quantifying timber and vested stocks to processing centers). NTFP harvest levels and values requires socio-eco- The price or value of the product made from the nomic and ecological data. Because data availabil- harvested stock, denoted by p , i s t h e f nal data element ity and quality will vary, we provide several needed to compute harvest value. If the harvested alternative approaches in tier 1 to quantify harvest product is traded in a market then market prices can volume and value. We f rst introduce a method that be used. Otherwise, if the product is generally con- uses current or historical harvest volume or harvest sumed directly by the harvester and not traded in a effort, harvest costs, and stock price data to meas- market, then other methods, such as non-market valu- ure a current or baseline period’s volume and value ation methods, will need to be used to estimate p . of harvest in a forest parcel (Section 8.2.1 ). We then The net value of a product’s harvest from the for- provide a method for estimating steady-state har- est parcel over the given time period is equal to the vest volume and value in a forest parcel when data revenue from that harvest minus the costs incurred on harvest volume or effort are not available in the process, (Section 8.2.2 ). NV= ( p×− H) C= E××−−()( p V) W Y , (8.1) 8.2.1 Calculating harvest value when harvest where harvest volume can be observed, H , or calcu- volume or harvest effort is observable lated with E × V , and costs can be observed, C , or In some cases, such as with commercial timber calculated with E × (W + Y ). harvests in developed countries, relatively good In Eq. (8.1) increasing harvest effort in a time data exist on harvest volumes, harvest costs, and period increases both revenue and costs in a linear product prices. This data can come from one of fashion (each additional unit of E generates p × V in several sources, including f eld monitoring of revenue but costs W + Y ). However, large jumps in harvests, surveys of harvesters, government sta- effort may lead to nonlinear changes in net returns. tistics, or, if the stocks are sold in the market, This may occur because a large pulse in effort over market data. We denote harvest volume of some the course of a harvest period could drastically product from a forest over a time period (typi- reduce the stock’s level, making the stock harder to cally a year) with H . I f w e c a n n o t f nd data on H f nd as the time period progresses. Such a dynamic but instead can f nd data on harvesting “effort” will reduce the harvest rate V as the harvester has to (measured in hours) in the forest over the time use more time to search for the increasingly scarcer period, denoted by E , and the harvest volume col- stock. If V does fall as effort increases, then net value lected per hour by a typical harvester, given by V , will fall as well because per unit effort costs remain THE SUPPLY, USE, AND VALUE OF FORESTS’ PROVISIONING SERVICE IN TIER 1 133 the same no matter the effort level. Therefore, cau- can specify the biological stock levels and their tion should be exercised in using the same V value growth functions in every forest parcel. across very different effort levels. However, even if we lack observations of biological We can map all harvest volumes and values on stock, we can still estimate harvest volumes and values the landscape by expanding the net value equation under steady-state conditions if we can specify V (S ) over all harvested biological stocks and forest par- and G (S ) and ascribe certain behavior to harvesters. In cels on the landscape, the next sub-sections we describe several methods for f nding harvest volumes and values without stock

NVxz= p z×− H xz C xz= E xz()( p x×−− V xz ) W Y , (8.2) data. The method we use to f nd harvest volumes and values will depend on the presence and protection of w h e r e x = 1 , 2 , . . . , X indicates distinct forest parcels harvest rights and our assumptions about harvest on the landscape and z = 1 , 2 , . . . , Z indexes biological behavior. We begin with a situation of open-access har- stocks available in the landscape’s forest parcels. The vesting, followed by exclusive harvest rights, and con- sum of H xz a n d NV xz o v e r a l l z represents the use and clude by discussing a case of intermediate access. net value of forest parcel x ’s provisioning service over the given time period, respectively, and the sum 8.2.2.1 Open-access steady-state harvest volume of Hxz a n d NV xz o v e r a l l z a n d x represents the use and value and net value of the landscape’s forest provisioning In many parts of the world, especially in develop- service over the given time period, respectively. ing countries, households harvest stocks from for- If we do not observe harvest volumes or harvest ests to supplement their incomes (e.g., Monela et al . effort then we cannot calculate the current or baseline 1993 ). Household harvesting is most active in forest period’s net value of harvest with Eq. (8.1) and we parcels where harvest rights are not established or need to apply alternative methods described below. not enforced, so-called “open-access” parcels (Feder and Feeny 1991 ; Hyde 2003). When using such 8.2.2 Calculating steady-state harvest volume open- access forests, households typically do not and value consider the value of leaving biological stock in situ to mature and harvest at a later date because other L e t G ( S ) indicate the growth of a biological stock harvesters are likely to harvest it in the meantime. over a time period in a forest parcel where S i n d i - As long as household labor is not in short supply on cates the level of the stock. For many stocks growth the landscape, we can expect households to con- slows down as stock increases due to resource tinue to enter the forest and harvest resources until limitations (density-dependent growth rates; see the last entrant cannot make an economic prof t Boot and Gullison 1995 ). In steady-state harvest- from further harvesting (where foregone wages in ing, the per period harvest of a stock from a forest the market and any harvesting equipment costs parcel is equal to its biological growth rate in the represents household harvesting costs; see Gordon parcel, 1954 ; Conrad and Clark 1987 ; Clark 1990 ; Lopez- Feldman and Wilen 2008 ). HEVSGS= × ( )= ( ), (8.3) This means that if all households earn the same return from a unit of harvest effort, then in open- where we now explicitly indicate that a stock’s har- access equilibrium no household will make an eco- vest rate, V, is a function of its biological stock level. nomic prof t from harvesting (i.e., the revenues from If we can observe the level of the biological stock harvesting equal the wages a household foregoes to from a forest inventory and can specify the stock’s harvest plus any equipment costs). However, when growth function G (S ) then we can use Eqs. (8.3) and households are differentiated, either because they (8.1) to estimate a period’s steady-state harvest and have different harvesting skills, different travel dis- its net value (assuming we have cost and price tances to harvest sites, or have different transport data). We can map all steady-state biological stocks, access to harvest sites (e.g., bicycle versus truck), harvest volumes, and values on the landscape if we then some households may earn positive net returns 134 THE PROVISIONING VALUE OF TIMBER AND NTFP in open-access equilibrium. With differentiation, interest (e.g., e indicates daily effort decisions and E there will be a critical distance or skill level at which is the annual sum of all daily decisions). a household choosing to harvest will just break even On each harvest trip, a household will earn (the value of harvest will just equal the household’s ( W + Y ) ( d n – d h ) (see the supplementary online opportunity cost of harvesting plus any equipment material (SOM) for a proof). This represents the costs). In open-access equilibrium, those households difference in travel time to the forest parcel closer to the forest or with higher skill level will har- between household h a n d h o u s e h o l d n m u l t i - vest while other households will not. plied by the value of an hour to a household. Here we illustrate the case of open access where Because household n earns zero net returns, the households are identical in all ways except for their value added to household h from harvest is sim- proximity to the forest. Suppose that a standard har- ply given by the value of time that h does not vesting trip to a forest parcel for a certain stock involves have to spend traveling. All household net e hours of harvesting in the forest as well as the travel returns from harvesting over the time period is time to and from the forest. Let dh be the hours spent given by, traveling to and from the forest for household h . We n can either model the actual locations of all households, NV=+∑ F()(). W Y dnh− d (8.7) or if spatially explicit census data is coarser, we can h=1 divide the landscape into regions and treat all house- Implementing this model requires specif cation of holds in a region as having the same distance. travel time to the harvest site as a function of dis- The household most distant from the forest that tance, location of households relative to the site, engages in harvest, labeled household n , will earn average productivity of harvest in the site, V ( S ), zero economic prof t from a harvesting trip, wage and equipment costs per unit of time, W +Y , average amount of effort per trip for households pH==++ pV() S e ( W Y )( e dn ). (8.4) that harvest from the site, e , and average number of harvest trips in a time period, F . We can map all We can use Eq. (8.4) to solve for the distance (meas- open-access steady-state harvest volumes and val- ured in hours) that household n travels to harvest, ues on the landscape by expanding the harvest volume Eq. (8.6) and the net value Eq. (8.7) over all [()(pV S− W+ Y )] e dn = . (8.5) relevant biological stocks and open-access forest ()WY+ parcels on the landscape. If we order households from closest to furthest from the forest, then all households with a travel time 8.2.2.2 Exclusive harvest-right steady-state harvest less than or equal to dn , given by h = 1, 2, . . . , n , will volume and value enter the forest to harvest. Household harvest over At the other extreme from open access, harvests some time period is in a forest can be completely limited to those with a property right to harvest. Exclusive harvest HVSnFeo = ()× , (8.6) rights can be granted to private companies, communities, or governments ( Feder and Feeny 1991 ; where H o indicates aggregate household harvest of E n g e l a n d L o p e z 2 0 0 8 ; G u a r i g u a t a et al. 2008 ). In the stock from the forest parcel over the time period, most cases exclusive harvest rights pertain to tim- F is the average number of trips that the household ber harvest. takes during the time period, and Eo = nFe is the If there are no restrictions on forest structure then household’s aggregate harvest effort in the forest we assume the holder of a timber concession will over the given time period (the superscript “ o ” sig- create a monoculture forest where a targeted vol- nif es household behavior versus exclusive harvest- ume of wood can be harvested at regular periods right holder behavior). Hereinafter the lower case (rotational forestry; see Tahvonen and Salo 1999 ; “ e” indicates sub-time period harvest effort and E Dauber et al. 2005 ). Further, we would expect rota- indicates the effort over the entire time period of tional forest operators to structure their holdings THE SUPPLY, USE, AND VALUE OF FORESTS’ PROVISIONING SERVICE IN TIER 1 135 such that operation-wide annual harvest levels are marginal cost is the sum of the direct cost of effort roughly equal to annual biological growth. One rea- plus the loss in future value from depleting the bio- son for such behavior is that it creates a smoother logical stock. Combining Eqs. (8.8) and (8.9) yields f ow of revenue and consumption over time, some- an expression for optimal harvest-right holder thing many businesses and communities seek effort, ( Browning and Crossley 2001 ). Because the holder of an exclusive harvest right ∂V ⎡⎤(WY+ ) Er∂S ∂G ⎢⎥+=∂S . (8.11) can manipulate stock levels in a forest, and thus m ⎣⎦VS() stock growth rates, the holder can theoretically choose any steady-state harvest level they wish. In words this condition means that the harvest-right Here we assume the operator of a rotational forest holder will hold a stock in situ up to the point will choose steady-state harvests that maximize the where its marginal growth rate plus the marginal net present value (NPV) of economic returns from benef t in harvest productivity due to the additional harvest over time. To do this, the operator will have stock equals the economy’s interest rate r . Finally, to consider how the current harvest affects harvest Eq. (8.10) enforces the steady-state assumption. potential in the future. For example, by harvesting Assuming G ( S) and V ( S) are def ned and p , α , d , r , most of the biological stock in the current period the W, and Y are observed we can use Eqs. (8.8) through owner will not be able to capture the additional rev- (8.10) to f nd harvest-right holder optimal effort in a enue that could be gained by leaving stock in situ to forest, optimal stock levels in the forest, and the grow larger and be harvested in the next period. optimal value of the stock, or E c , S c , and µ c , respec- Therefore, the solution to the owner’s problem tively. Optimal steady-state harvest by the exclusive involves solving a dynamic optimization model that harvest-right holder, H c , is determined by evaluat- incorporates trade-offs between the current and ing the harvest function at Ec and S c while the net future harvest. This type of dynamic harvest model value of the optimal steady-state harvest each time has been extensively analyzed (Clark 1990 ). period, NV c , is given by evaluating Eq. (8.1) at H c , The steady-state solution to the harvest-right E c , and S c (however, we now have to add transpor- holder’s net revenue maximization problem is char- tation costs to the net revenue equation). If we acterized by the following system of equations, determine exclusive harvest-right volumes H c and net values NV c across all relevant biological stocks ()()()()pdVSWYVS− am=++ (8.8) and parcels with exclusive harvest rights then we can create a map of these values. While this approach is theoretically consistent, it ⎛⎞ ()pdE− am∂∂∂VGV= ⎜⎟ r− + E (8.9) may be diff cult to def ne the functional relation- ∂∂∂SSS⎝⎠ ships and f nd the necessary data. In such cases we can fall back on simpler methods to f nd H c, such as GS()== H E× VS (), (8.10) looking for published harvest volumes, biological growth rates, and harvest costs in government doc- where α represents the cost to haul one unit of har- uments, forestry management journals, or other vest volume one kilometer, d is distance in kilom- industry documents (this approach to f nding har- eters from the forest parcel to the harvested stock’s vest-right holder harvest volumes and costs was processing site, and μ is the monetary value of the used in Polasky et al. 2008 ). biological stock (the “shadow” value in economics If the holder of an exclusive harvest right is a vernacular; see Clark (1990 ) or the SOM for more community or government, their objective may not information about the derivation of these condi- be to maximize the NPV of harvest but to maximize tions). According to Eq. (8.8) optimal harvest is an objective function that includes spiritual, biodi- found by equating the net value of one more unit of versity, and regulating ecosystem service values harvest effort (the left hand side) with the marginal associated with the forest. The approach outlined cost of additional unit of harvest effort, where this here can be modif ed to include these different 136 THE PROVISIONING VALUE OF TIMBER AND NTFP objectives. Alternatively these considerations can growth. However, if we observe harvest by harvest be included as constraints on the type of practices right-holders or it does not exist in a forest parcel that harvest right-holders are allowed to engage in. then we can drop the steady-state assumption in If a government or community is interested in main- this particular tier 1 model. taining a semblance of a forest’s natural state then T h e f rst step in this intermediate access model is to timber harvest rights typically only allow selective calculate a household’s expected net revenues gener- logging (Repetto and Gillis 1988 ; Pinard and Putz ated from harvest, no matter whether it is illegal or not. 1996 ). For example, in New Zealand only certain Like all other tier 1 models this is a function of effort. old-growth forests can be logged and the total bio- Let D be the hours in a household’s working day if we mass removed annually from these forests is lim- are modeling day trips or working hours a harvester ited to 20% of annual biomass growth (see http:// expects to devote to a harvesting trip if we choose to www.insights.co.nz/Natural_Forests_r.aspx#a). In include multi-day trips in our model, let d b e t h e h o u s e - such systems the most mature trees are generally hold’s travel time in hours to and from the forest par- selected for logging in order to keep harvest costs cel, and let e z be the hours that the household spends per unit of volume removed low. When a forest par- harvesting stock z in the forest on the trip such that cel includes selective logging concessions we can Z use Eq. (8.1) along with information on the selective Dd=+∑ ez , (8.12) logging restrictions and costs (including transporta- z=1 tion costs) to estimate the net value of harvest by Z Z where e = 0 if d ³ D. We can calculate e the harvest-right holders. ∑ z=1 z ∑ z=1 z by using GIS software and travel speed assumptions to determine d and subtracting this from D . 8.2.2.3 Intermediate access steady-state harvest In this intermediate access model a household volume and value incurs two costs when harvesting, wages lost by not The open-access and exclusive harvest models rep- working in the labor market and expected f nes from resent extreme cases of harvesting activity. In many being caught harvesting illegally (we will ignore forests around the world households (illegally) har- equipment costs for now). Let ρ Î [ 0 , ρ¯ ] i n d i c a t e t h e vest stocks that are subject to exclusive harvest z intensity of efforts to prevent the illegal harvest of rights (Hyde 2003 ). In response to the threat of ille- stock z w h e r e ρ = ¯ ρ i n d i c a t e s m a x i m u m i n t e n s i t y gal harvest, harvest-right holders often expend z and delta (p ) is the expected daily or trip monetary some effort to protect their property rights (Feder z f ne from harvesting z illegally. We assume delta is and Feeny 1991 ). In addition, the models above do increasing in p. Efforts to reduce illegal harvest will not consider the inf uence that alternative harvest affect household harvest in two ways. First, they sites would have on household harvest behavior, may reduce hourly harvest rates as households have that households may harvest multiple stocks in a to spend time avoiding detection. Second, forests forest, and the unobserved household attributes with less intensive protection of harvest rights will that inf uence its harvest effort choices (e.g., house- be subject to more household harvest, all else equal. hold harvesting skills or preferences). The household’s daily or trip net revenue from Here we describe an approach that incorporates harvest in a forest parcel is given by, these factors in a steady-state harvest model. In this approach we model household harvesting decisions Z by comparing a household’s potential harvesting ⎛⎞ nr= ⎜⎟∑( pzz e V z(,rdr z S z ))− () z revenues to their opportunity costs where costs ⎝⎠z=1 (8.13) include time spent avoiding any forest guards and −(),WYDu++ potential illegal harvest f nes (Barbier and Burgess

2001 ). In this approach we do not model by harvest where V z ( ρz , S z ) is decreasing in ρ z and increasing in right-holders but instead recover or estimate their S z , d(ρz ) is the expected daily or trip monetary f ne harvest levels by enforcing steady-state equilibrium from harvesting z, which is increasing in ρz , and u is and comparing household harvest volume to stock a household-specif c random variable that includes THE SUPPLY, USE, AND VALUE OF FORESTS’ PROVISIONING SERVICE IN TIER 1 137 unobserved household characteristics that affect we have divided the landscape into 1) three dis- the net value calculation, and all other variables are tinct regions, 2) assume two classes of harvesting as before. If biological stock data is missing then we skill, low and high, and 3) assume two harvesting estimate V z ( ρz ) instead. If harvest rights for stock preferences, gathering timber to sell in the market z in parcel x do not exist or are not enforced then versus gathering NTFPs for home consumption,

V z ( ρz = 0, S z) = V z ( Sz ), δ(ρz =0) = 0, and Eq. (8.13) then there are 12 representative households. Let v reduces to the open-access net revenue function. = 1, 2, . . . , V index all household types on the land- We assume the household will choose to spend the scape. If Nv is the number of households of type v day or trip harvesting in the forest parcel that is and β v i s t h e n u m b e r o f v ’s working days in the expected to generate the highest net return when time period if we are modeling daily decisions or compared to all other working options ( Parker et al. the number of trips considered over the course of a 2003 ). Each forest’s probability for generating the time period if we are considering multi-day trips highest household net revenues for a day or trip time then the expected aggregate household harvest of period is given by the joint probability that harvest- stock z i n p a r c e l x over the time period is the sum ing from the parcel generates both positive expected of expected daily or trip household harvests over net returns (i.e., the probability that harvesting in the all households. parcel is better then working in paid labor) and gen- V erates the highest expected net returns when com- ovvvvv HNeVSxz= ∑ bg x xz xz(, r xz xz ), (8.14) pared to all other potential harvest sites. Let this joint v=1 probability for forest parcel x b e g i v e n b y γ . The Z v x where we have allocated e a c r o s s a l l z i n p a r - ∑ z=1 xz vv vv parameter γx will only lie between 0 and 1 (and not be cel x and bgxxzNe is equivalent to aggregate illegal equal to 0 or 1 exclusively) if we include the random harvest effort for z in x over a harvest period by variable u in Eq. (8.13) or assume that the price, wage, v households of type v , or E xz. For example, if we are f ne, effort, or harvest-right variables in Eq. (8.13) are modeling two stocks, timber and food NTFPs, then random variables. the representative households that prefer timber The value of γ decreases as enforcement efforts x could be assigned an effort value of 0 for food in parcel x increase (higher ρ ), as the distance z NTFPs and vice-versa. between the household and parcel x increases Z Finally, the aggregate harvest of stock z in forest (higher d and lower e ), and as household wage ∑ z=1 z parcel x over the time period assuming steady-state rates increase. The value of γ will increase with an x harvest is given by H , increase in biological stocks on parcel x or with the xz

co price of the harvested stock. If the sum of γ x across HHHGSxz=+= xz xz xz ( xz ) . (8.15) all forest parcels on the landscape is less than 1 (i.e., X g x < 1 ) then there is some possibility that a c ∑ x=1 where Hxz is the harvest level of z in forest parcel x household will not harvest at all on a given day or by z’s harvest-right holder in x and is recovered by over a potential trip period and instead work as o subtracting Hxz from G xz ( S xz ). In this model we do paid labor in another economic sector. not estimate harvest-right holder behavior. Instead N e x t w e u s e γ x values over all households to cal- we are reliant on observations of ρ xz and G xz ( Sxz ) to culate expected aggregate household harvest effort c estimate Hxz . in parcel x over the modeled time period. Rather The net value of harvest in parcel x over the time than model each individual household, which period is given by, could be a computationally diff cult task on large landscape, we can use a limited number of repre- ZZ sentative households to account for all household oc NVx =+∑∑ pzxz H() pzzxzxz− a d H harvesting behavior on the landscape. The repre- zz==11 V sentative households should delineate the greatest vvvvvv (8.16) −∑()WY+ bgx ND , differences in household locations, harvesting skill v=1 c sets, and harvesting preferences. For example, if −()WYE+ xz− wr xz () xz 138 THE PROVISIONING VALUE OF TIMBER AND NTFP

w h e r e d xz is the distance from parcel x t o s t o c k z ’s Figure 8.1 . The area within the longer-dashed line is processing site, the superscript “v ” o n W a n d Y i n d i - much better protected by government off cials than cate that wages can vary across household types areas within the shorter-dashed lines (data on rela- c (versus average wages levels on the landscape), Exz tive stock levels and protected areas come from a is the effort level needed by z ’s harvest-right holder panel of Tanzanian experts consulted at a conference c in x t o a c h i e v e Hxz a n d ω xz ( ρxz ) i s t h e c o s t t o z ’ s h a r - in Morogoro, Tanzania in February, 2008). vest-right holder in x to achieve illegal harvest pre- In this example, we assume that there are two types vention effort ρxz . We do not include the revenues of harvesting households in each urban area: high- generated from f nes because they are a transfer of skilled (HS) and low-skilled (LS; for simplicity we household wealth to the enforcement authority assume all household harvest originates from these (either the harvest-right holder or the government); seven urban areas). We index the urban area-house- no additional societal value is generated in these hold type combination on the landscape by v (seven transfers. By indexing harvest volumes and values urban areas and two household types means v = 1, by biological stock and parcels we can easily transfer 2, . . . , 14). We assume highly skilled households travel harvest volumes H xz a n d n e t v a l u e s NV x to maps. to forested-parcel access points by truck and then While this model incorporates many realistic walk from the road to the parcel’s centroid. We aspects of harvesting, it requires extensive informa- assume that low-skilled households travel to a for- tion to implement. It requires information on illegal ested parcel access points by bike and then walk from harvest prevention effort, expected illegal harvest the road to the parcel centroid. In addition, we assume f nes, household locations, distance to forest parcels highly skilled households have higher harvest rates and speed of travel, wages and equipment costs, than their low-skilled counterparts for both mush- product prices, and legal harvest rates or biological rooms and poles, all else equal (i.e., VHS (S ) > V LS (S ); growth rates. Obtaining all of the data inputs see the SOM for more information on the values of required for use of this approach will most likely V (S ) for each representative household type). We require independent research as well as consultation assume all harvesting trips are daily trips. with local experts and modeling with GIS software. We use a modif ed version of Eq. (8.13) to determine the daily net revenue value of v ’s harvest in forest parcel x , given by, 8.2.3 Tier 1 intermediate access model example ⎛⎞2 To illustrate portions of the household harvest model vvvv (8.17) nrxxx= ⎜⎟(1−−×r )∑()epVSz z xz (xz ) ( W D ), described in Section 8.2.2 we focus on a small region ⎝⎠z=1 in the Eastern Arc Mountains watershed of Tanzania with seven urban areas that vary in size from small where efforts to prevent illegal harvest anti-poaching villages to major urban areas and 32 forest parcels efforts are not distinguished by stock type, ρ x = 0 i n f o r - (see Figure 8.1 ). In this example we model the har- est parcels with not conservation status (open-access vest of wooden poles (poles are used in small-scale parcels), ρx > 0 in conserved forest parcels, and D m e a s - construction and building) and mushrooms. The ures hours in a work day. Here we set δ(ρ x ) = 0 because maps in Figure 8.1 indicate in which forest patches there is no f ne if a household is caught illegally har- mushroom and pole stocks are either high or very vesting; instead, when caught, the harvest is conf s- v high. In this example we assume that it is only cated. Daily household effort in forest parcel x ( i . e . , exz ) worthwhile for households to harvest poles mush- is found by calculating the time needed by representa- rooms in forests where stocks are high or very high. tive household v to travel to parcel x on the landscape’s There are no exclusive right harvests in this land- road and path network given v ’s transportation mode, scape (H c = 0 for all parcels). Several areas on the multiplying this by two (there and back), and subtract- landscape have conservation status and technically ing this total travel time from D . forbid harvest of biological stocks. The conserved To account for uncertainty in model variables we v areas are indicated by the dashed lines on the map in treat ρ x for all conserved parcels, p z for all z , V xz()S xz THE SUPPLY, USE, AND VALUE OF FORESTS’ PROVISIONING SERVICE IN TIER 1 139

18 22 16 17 21 24 5 19 20 23 25 6 1

3 Morogoro (228,863) 7 2 Kilosa (8,408) 8 4 Mkuyuni (19,366)

91 km 26 14 15 9 Msongozi (9,778) Ulaya (13,123) 27 12 28 Mbongwa (14,064) 10 31 30 13 29 Kikeo (13,571) 11 32 Mushrom stocks Pole stocks Protected status 138 km High High Moderatly protected Very high Very high Well protected

Road Urban area

Figure 8.1 Distribution of urban areas, mushroom and pole sources, and protected areas on a portion of the Eastern Arc Mountains watershed in Tanzania. In map (a) each dark-shaded parcel is a distinct forest stand. Parcels with the same border (either white or black) have similar stock levels of mushrooms and poles. The polygons formed with dashed lines represent areas imperfectly protected by a government agency for conservation. The polygon formed by the longer-dashed lines indicates a very well protected area (although not necessarily strong enough to prevent all illegal household harvest). The black lines are roads and the black dots are urban centers, scaled by population size. Map (b) gives parcel IDs ( x = 1, 2, . . . , 32), urban area names, and urban area populations.

for all unique combinations of x , z , and v , W v for all vv of times out of 1000 that n rnrxj> max (see the SOM v D xj≠ , and as random variables. To account for unob- for details). served harvesting preferences across household Next we use Eq. (8.14) to calculate aggregate u types (the random variable from Eq. (8.13)), we household effort and harvest levels across a harvest v treat ’s allocation of daily effort across pole and period for each x and z combination 1000 times. In x mushroom harvesting in each forest parcel as a each iteration, values for N v for each household type random variable. v, the number of days in a harvest period, given by As noted above, the probability that a household β, e v for all unique combinations of x , z , and v , and v x x z of type w o u l d h a r v e s t i n in any given day is the V v ( S ) for all unique combinations of x , z , and v are joint probability that the daily net revenue of v ’s x z x z randomly drawn from their distributions ( γ v for all x x harvest in forest parcel is greater than 0 and that x and v combinations, calculated in the f rst step of v the daily net revenue from ’s harvest in forest this illustrative example, remain constant in each x parcel is greater than all other parcel-level daily run). See the SOM for details. harvests. In this example we assume these two The range in aggregate household effort, harvest, probabilities are statistically independent and that and daily net revenues values for each x and z com- vv v v bination over the harvest period in each parcel x are g xx=>Pr(nr 0)Pr( nr x > max nr j ). (8.18) xj≠ reported in Tables 8.1 through 8.3 . The most aggre- v We f nd each probability by calculating nrx , as gate effort is expended in parcel 30, the forest patch given by Eq. (8.17), 1,000 times for each x and v right next to the town of Kikeo and within one day’s combination where we randomly draw variable distance of 3 other towns, including Morogoro, by values from each random variable’s assumed distri- far the largest urban area on the landscape. This v bution for each iteration. Pr()nrx > 0 is given by parcel attracts a lot of effort because it is right on the the number of times out of 1,000 that highway and some of the other forest patches in the v ⎛⎞vv nrx > 0 and Prnrxj> max nr is given by the number area have protected status. Parcels 10 and 14 also ⎝⎠xj≠ 140 THE PROVISIONING VALUE OF TIMBER AND NTFP

o Table 8.1 Low, mean, and high E xz (hours per year) values for each x and z combination (a parcel not represented has a 0 value in every column)

Parcel ID Low Mean High

Mushrooms Poles Mushrooms Poles Mushrooms Poles

5 1 2 923 953 3 757 4 336 6 0 0 13 13 65 56 7 0 0 3 3 16 16 8 3 4 1 389 1 339 6 203 5 452 9 23 16 2 957 3 040 10 821 12 423 10 33 52 7 251 7 466 32 575 28 131 11 2 3 419 389 1 653 1 333 14 182 606 9 316 9 366 31 101 28 615 20 19 31 1 953 2 027 7 786 6 500 21 114 40 2 588 2 476 9 203 8 633 22 119 150 3 019 2 921 11 627 10 636 23 0 0 1 1 6 5 28 0 0 282 294 1 768 2 094 30 57 131 11 570 11 209 39 340 41 506 32 0 0 0 0 5 5

o Table 8.2 Low, mean, and high Hxz (Mg per year) values for each x and z combination (a parcel not represented has a 0 value in every column)

Parcel ID Low Mean High

Mushrooms Poles Mushrooms Poles Mushrooms Poles

5 0.0 0.1 9.3 28.5 41.0 124.3 6 0.0 0.0 0.1 0.4 0.7 1.9 7 0.0 0.0 0.0 0.1 0.1 0.5 8 0.0 0.1 13.9 40.2 65.0 181.2 9 0.2 0.4 29.5 90.9 109.5 418.7 10 0.1 1.8 28.8 224.7 130.5 785.9 11 0.0 0.1 2.1 15.4 8.8 59.7 14 1.8 13.8 84.0 236.8 301.5 779.9 20 0.2 0.9 19.7 60.9 75.4 216.6 21 1.1 1.1 25.8 74.5 91.1 262.1 22 1.1 4.5 30.1 87.4 99.8 336.8 23 0.0 0.0 0.0 0.0 0.1 0.2 28 0.0 0.0 2.8 8.8 18.9 64.3 29 0.0 0.0 0.0 0.0 0.0 0.0 30 0.6 3.6 115.4 336.9 431.7 1,239.0 32 0.0 0.0 0.0 0.0 0.1 0.1

attract substantial effort. Not surprisingly, this household harvest in most parcels is approximately example illustrates that patches that are right next 0. This indicates that in general households are to highways (traveling speeds off the main roads indifferent between harvesting in the given parcel are very slow), are found between towns, and are and working in paid labor. This calculus would not protected attract the most effort (e.g., Luoga change if wages fell, unemployment in other eco- et al . 2002 ). Finally, the mean value of aggregate nomic sectors became a problem (here we do not THE SUPPLY, USE, AND VALUE OF FORESTS’ PROVISIONING SERVICE IN TIER 2 141

Table 8.3 Low, mean, and high aggregate household net economic returns from harvesting (NV ) for each x and z combination in thousands $US

Parcel ID Low Mean High

5 –25 –5 38 6 0 0 1 8 –26 –7 37 9 –71 –13 74 10 –97 –23 131 11 –6 –2 15 14 –78 –11 253 20 –36 –13 164 21 –77 –39 75 22 –100 –46 21 23 –1 0 0 25 –57 –23 –6 26 –24 –10 –3 27 –7 –2 –1 28 –52 –27 –6 30 –713 –336 –86 formally model unemployment possibilities), or determined within the model to be consistent with conservation status was given to and enforced in market equilibrium. more forest parcels. This intermediate access model was introduced above as a method for estimating both households 8.3.1 Tier 2 model when harvest volume or and harvest-right holder harvest in a matter that bet- harvest effort is observable ter incorporated harvesting realities. The steady- In this initial tier 2 model we assume that data on har- state restriction is applied in most applications of vest volume of biological stock z or the total effort this method in order to recover harvest-right holder used to harvest the stock in a forest parcel is available. harvest volumes. However, because there is no Let H zt represent harvest volume of stock z i n p e r i o d t property right harvest to recover in this example the in a forest parcel and E zt represent total harvest effort steady-state assumption is irrelevant in this case. for stock z i n p e r i o d t in a forest parcel. If effort data are observable and harvest volume data are not, then stock z ’s harvest rate function, given by V (S ), must 8.3 The supply, use, and value of z z forests’ provisioning service in tier 2 also be def ned. Just as in tier 1, the net value of stock z’s harvest in period t from a forest parcel is given by Unlike tier 1, in tier 2 we do not assume a constant the basic harvest net revenue equation, harvest volume per harvest period but instead NV = p × H-C=E allow the trajectory of harvest volume and value to zt zt zt zt zt (8.19) change through time. By tracking biological stocks ⎛⎞p ×VS -W-Y, ⎝⎠( zt z() zt) t t and harvest though time we can model biological degradation or growth dynamics. Just as in tier 1, which is equivalent to Eq. (8.2) except that we have the solution to these models will depend on the included time subscripts (again, in this cost calcula- property right structure in the forest and the behav- tion we ignore the opportunity costs associated ior ascribed to harvesters. In all cases, prices of with harvesters traveling to and back from forest products made from biological stocks and wages parcels and any costs associated with transporting can either be given as part of a scenario or can be harvested stocks to processing centers). The NPV of 142 THE PROVISIONING VALUE OF TIMBER AND NTFP harvest from periods t = 1, 2, . . . , T on a forest parcel equation of motion. Specif cally, the stock of z in a is equal to forest parcel in harvest period t + 1 is equal to period

T t ’s stock level plus the growth in stock from t to NVzt NV = ∑ t=1 . (8.20) t + 1 less the parcel harvest of the stock in period t . z (1+ r )t−1

We can either def ne prices and wages for each time SSGSSSHzt,1+ =+ zt z(1 t , ..., zt ,..., Zt )− zt , (8.22) period t or we can determine them within our model.

We can use several methods to def ne price and where Gz ( S1 t , . . . , S z t , . . . , S Z t ) gives the growth of z ’s wage trajectories. Most simply, we can extrapolate stock in the parcel from time t to t + 1. The inclusion past price and wage trends into the future. Or, if of other stock levels in the growth function means available, we can use published estimates of future that the growth of stock z from one period to the timber and NTFP prices and wage rates on the study next can be affected by the contemporaneous stock landscape. Otherwise we can endogenously deter- levels of one or more other stocks in the parcel. For mine prices by gathering information about demand example, the growth rate of a bush meat species for the stock and solving for the price that equates may be dependent on its own stock and the stocks of demand for timber and NTFP stocks with local sup- predators and prey. ply. Let the function that describes local demand for If the model ever calculates a harvest of z larger a stock harvested from the study landscape at time t than its stock, H zt ³ Szt , then we set H zt = Szt and Hz,t+1 be given by F zt ( pzt ). The market clearing price for = S z,t+1 = 0 for time periods t + 1 to T unless the forest stock z a t t i m e t is by solving the following for p zt , parcel is re-colonized later by biological stock z .

X HFp= (8.21) ∑ x=1 xzt zt( zt ) 8.3.2 Calculating harvest volume and value in tier 2 when such data is missing where Hxzt indicates z ’s harvest in parcel x at time t . In Eq. (8.21) we assume no imports or exports of In tier 2, when harvest volumes or efforts are not stock z in the study landscape. If the demand func- observable, we analyze a unif ed model of interme- tion F zt ( pzt ) describes local demand for stock regard- diate access that includes the possibilities of exclu- less of point of origin we can assume some portion sive harvest and complete open access as special of F zt ( pzt ) is satisf ed by imports of z and some por- cases. We begin by describing household harvest, tion of z ’s harvest is exported before we solve for which will be a function of household characteris- local equilibrium prices. tics, biological stocks, and illegal harvest preven- We can solve for wages in a similar manner if tion efforts. Next we describe how the holder of X we have data on harvest effort. Let DE− harvest rights should choose harvest and enforce- txzt∑ x=1 indicate the supply of labor available after ment efforts to maximize the NPV of harvest returns accounting for harvest effort where D t is the over time. Because households react to efforts to aggregate work time of all harvesters in period t enforce harvest rights, harvest-right holders make X and E is the amount of time all harvesters their decisions f rst in each time period. We com- ∑ x=1 xzt spend in all forest parcels x = 1 , . . . ,X i n p e r i o d t . bine both harvest by households and harvest-right Further, let local demand for labor at time t be owners to describe total harvest and evolution of X given by B ( W ) . F i n a l l y , w e e q u a t e DE− the biological stock and harvests through time. t t txzt∑ x=1 with Bt ( Wt ) a n d s o l v e f o r W t . Because endogenous solutions require def ning F zt ( p zt) f o r e a c h s t o c k z 8.3.2.1 Household harvest across all time periods and B t ( Wt ) across all time Just as in tier 1, in tier 2 a household allocates their periods, this approach may be diff cult to labor in a time period between paid labor and harvest- implement. ing activities in order to maximize their net revenues. Finally, we have to verify that the given (or calcu- Further, just as in tier 1, because an individual house- lated) harvest volumes are possible. In tier 2 we hold does not consider the affect of their harvest on track biological stock in a parcel over time with an biological stock levels their optimization procedure is THE SUPPLY, USE, AND VALUE OF FORESTS’ PROVISIONING SERVICE IN TIER 2 143 static; they make harvest decisions without consider- f ne for harvesting z f r o m p a r c e l x i n p e r i o d t ing the ramif cations of such harvest effort on future where δ increases in enforcement effort, ρ xzt , and stock levels. In the tier 2 framework, a household allo- effort. If the forest parcel x is completely open to cates their entire period’s labor budget simultaneously households for the harvest of z , ρ xzt = 0, and h h (in tier 1 a household made multiple daily or trip-level δρ()xzt,E xzt = 0 f o r a l l Exzt . L e t h o u s e h o l d h ’s wage h decisions). We index households on the landscape with in period t b e g i v e n b y W t (different households h = 1 , 2 , . . . , N , w h e r e h o u s e h o l d h is located at point h can command different wages) and the per effort h on the landscape. Let Dt indicate the total work hours unit cost of harvesting equipment and accesso- h in time period t f o r h o u s e h o l d h , Exzt i n d i c a t e t h e t o t a l ries, such as trucks and saws, in period t be given h time (in hours) that h uses to harvest stock z in forest by Yt . h parcel x during time period t , dxt indicate the one-way In the static optimization framework the household travel time between household h and forest parcel x in allocates their harvest effort over each stock and parcel h hours in period t , a n d Ixt indicate the number of trips combination and the number of trips they make to h household h t a k e s t o p a r c e l x i n p e r i o d t. F i n a l l y, Lt each parcel for each time period t . S p e c i f cally, the indicates household h ’s time available for work in household maximizes their net revenues in time wage labor during the harvest period. period t by choosing the effort levels that solve the following for each z and x combination in time period t , hh⎛⎞ h hh LDtt= −−⎜⎟∑∑ E xztxtxt ∑2, dI (8.23) ⎝⎠xz x ⎛⎞⎛⎞ ⎜⎟⎜⎟∂∂y h d pYhh−− where the term hh converts trips to parcel x in zt⎜⎟⎜⎟∂∂EE t 2dIxt xt ⎝⎠⎝⎠xzt xzt period t into the number of hours household h uses (8.25) to travel to and back from forest parcel x in period t . ⎛⎞ hh∂σ To simply matters we replace the number of trips in =+Wd⎜⎟12 t⎜⎟ xt ∂Eh Eq. (8.23) with a function that relates the number of ⎝⎠xzt trips to parcel x in period t to total harvest effort in Z If household h spends all of their working time har- the parcel during period t , IEhh= σ , where xt(∑ z=1 xzt ) vesting and never works in the paid labor market then σ is increasing in effort in parcel x . h Wt in Eq. (8.25) should be replaced by h ’s marginal Household harvest volume is related to effort with value of harvesting. The left hand side of Eq. (8.25) a generic version of the tier 1 production function, measures the revenue generated by the last unit of effort used to harvest stock z i n p a r c e l x less the per hh HyESxzt= () xzt,, xztσ xzt , (8.24) effort unit cost of capital costs and the marginal expected f ne for that less unit of harvest effort. The h where Hxzt is household h ’s harvest of stock z from right hand side of Eq. (8.25) is the opportunity cost of lost wages due the last unit of effort to harvest z in parcel x in period t , Sxzt represents the level of bio- parcel x in period t (including travel time). If the parcel logical stock z in parcel x in period t, and qxzt represents a vector of parcel-level biophysical variables. x is completely open to household harvest then the marginal expected f ne term in Eq. (8.25) is dropped. For example, the vector q xzt could include information on the elevation or slope of parcel x where The solution to Eq. (8.25) gives the optimal harvest greater elevations and slopes reduce the productiv- effort for stock z i n p a r c e l x b y h o u s e h o l d h i n p e r i o d t , h* ity of harvest effort, all else equal. We assume that given by Exzt . Optimal effort levels in period t are a h h h h* harvest of z in x increases in effort and that harvest function of pzt , S xzt , d xt , q xzt , ρ xzt , W t , a n d Y t . L e t Hxzt i n d i c a t e h ’s optimal harvest of z f r o m x i n p e r i o d t per unit of effort increases in S xzt . h* In choosing how much to harvest, the house- when evaluated at Exzt f(see Eq. (8.24)) and let the opti- hold considers the opportunity cost of harvesting mal aggregate household harvest of stock z i n p a r c e l x in period t i s g i v e n b y HHoh* = *. B y s o l v i n g Eh* (lost wages), and harvest-related capital costs, xzt∑ h xzt xzt plus the costs associated with illegal harvesting. over all h , x , z , a n d t we specify the complete solution h Let δρ()xzt,E xzt indicate the expected monetary to the tier 2 household model problem. 144 THE PROVISIONING VALUE OF TIMBER AND NTFP

8.3.2.2 Exclusive harvest-right harvest Here we assume that harvest-right holders make volume and value their harvest and property right protection choices Just as in tier 1, we assume the holder of a harvest f rst and then households react accordingly as right for stock z in parcel x chooses harvest and described in optimality condition (8.25). Therefore o enforcement effort over time such that the NPV of the harvest-right holder will replace Hxzt in equation harvest returns are maximized. In tier 2, unlike tier of motion (8.28) with their expectation for aggregate

1, we do not assume that biological stocks and har- household harvest as a function of their choice of ρxzt . vest are constant through time. According to the previous section, the harvest-right o Let the harvest of biological stock z from parcel holder can expect Hxzt to be given by the function o* x during harvest period t by its harvest-right HSxzt() xzt,,r xztb xzt w h e r e b xzt is the vector of all other c holder be given by Hxzt . Harvest of a stock by har- variables and parameters that def ne the optimal vest-right holders is a function of harvest labor, household harvest of stock z i n p a r c e l x i n p e r i o d t c E xzt , capital equipment used, Q xzt , s t o c k i n p e r i o d t , (e.g., prices, wage). If the tier 2 version of o* and parcel-level biophysical variables, represented HSxzt() xzt,,r xztb xzt i s t o o d i f f cult to solve because we c by vector qxzt . The relationship between H xzt and cannot def ne the relationship between effort in par- these input variables is def ned by production cel x and the number of trips to parcel x or between function f , a generic version of the tier 1 produc- effort in parcel x and expected trespassing f nes we tion function, can use a tier 1 approach to f nd household harvest in the parcel. All that is required of the household har- cc HfEQSxzt= (, xzt xzt ,,). xztθ xzt (8.26) vest expectation function is that it is explained by biological stock and effort to prevent illegal harvest. We assume harvest increases in labor and capital The conditions that satisfy the optimization prob- equipment and that the rate of harvest per unit of lem for the entity that has right to harvest z in x include effort increases in S xzt . In tier 2, the harvest-right holder considers three c ∂∂ff types of costs: the cost of harvest effort, which is a ()pdzt− am zt xztcc=+ W t xzt (8.29) ∂∂EExzt xzt function of the wage labor rate W t and capital wage rate M t; the net cost of illegal harvest pre- c ∂∂ff pd M ()zt− am zt xztcc=+ t xzt (8.30) vention efforts, φ(ρxzt ) (the costs of prevention ∂∂QQxzt xzt efforts less the revenue from f nes); and the cost of transporting harvest to a processing site or o ∂j ∂Hxzt market. = mxzt (8.31) ∂∂rrxzt zxt The harvest-right owner’s NPV of revenues from harvesting stock z in parcel x is maximized by c ∂f ()pdzt− amm zt xzt=+ xz,1 t− (1 r ) −− xzt choosing harvest and enforcement over time such ∂Sxzt (8.32) o ∂∂GHzxzt∂f that satisfy, mmmxzt++ xzt xzt ∂∂SSxzt xzt ∂ S xzt

T c c max pd− α fE(,xzt Q xzt ,,) S xztq xzt c c ∑(()zt zt xzt t−1 + ESQxzt,, xzt xzt ,ρ xzt S=S+GS-fE,Q,S,()( q ) t=1 (1 + r) xz, t+1 xzt xz xzt xzt xzt xzt xzt (8.27) o (8.33) c ⎞ j ()rxzt−−WE t xzt MQ t xzt −HS,r,xzt() xzt xztb xzt t−1 , (1 + r) ⎠⎟

where μxzt is the value of the biological stock z in subject to the stock constraint parcel x in period t. The f rst two conditions immediately above are no different from optimality con- co SSGSfEQSHxz,1 t+ =+ xzt xz() xzt− (, xzt xzt ,,) xztq xzt− xzt (8.28) dition (8.8) from tier 1. Further, conditions (8.29) and (8.32) can be combined as they were in tier 1 to

w h e r e α zt represent the cost to haul one unit of stock z solve for optimal effort by the harvest-right holder. c one kilometer in period t a n d dxzt is distance in kilom- Condition (8.31) means that illegal harvest preven- eters from parcel x t o z ’s processing site in period t . tion efforts should be increased until the net cost of THE SUPPLY, USE, AND VALUE OF FORESTS’ PROVISIONING SERVICE IN TIER 2 145 one more unit of effort equals the value of the last The net value of z ’s total harvest from parcel x at unit of stock z saved from illegal harvest. Equation time t is given by, (8.33) keeps track of the evolution of biological stock co** cc* (stock in the next period equals stock this period NRxzt=+ p zt() H xzt H xzt− a zt d xt H xzt plus growth minus harvest). c* * −−WEtxzt MQ t xzt −w xztxzt()ρ* (8.35) c* * * L e t Exzt , Qxzt , S xzt , r*xzt , a n d mxzt* f o r a l l t indicate the H hhhh* h⎛⎞ h* variable values that solve optimality conditions (8.29) − ()WYEWttxzttxt+ − 2 ds Exzt ∑∑∑⎝⎠⎜⎟ through (8.33). These optimal actions are functions of hx=1 z c the model parameters pzt , αzt , dxzt , q xzt , W t , M t , and r w h e r e ω xzt ( ρ xzt ), as in tier 1, is the cost to z ’ s h a r v e s t - c* across all t beginning at time t = 1 . L e t Hxzt indicate right holder in x to achieve the anti-poaching effort the value of harvest production function f w h e n e v a l - ρ xzt. As in tier 1 we do not include the revenues gener- c* * uated at Exzt , Q xzt* , a n d Sxzt (see Eq. (8.26)). We can f nd ated from f nes because they are a transfer of house- the NPV of timber harvest for the harvest-right hold wealth to the enforcement authority (either the holder of stock z i n p a r c e l x by plugging in their opti- harvest-right holder or the government); no additional mal choices into the objective function (8.27). societal value is generated in these transfers. We sum As in tier 1, a harvest-right holder’s objective may across stocks, parcels, and time periods t = 1 t o T (in a not be to maximize the NPV of harvest returns but to discounted fashion) to generate the NPV of revenues maximize an objective function that includes spirit- generated by the landscape’s forests from t = 1 t o T . ual, biodiversity, and regulating ecosystem service As discussed in Section 8.3.1, we can exogenously values associated with the forest. In these cases we def ne prices and wages in Eq. (8.33) over the time can change the objective function (8.27) accordingly. period span of t = 1 to T. Otherwise, if we solve for For example, assume a harvest-right holder’s utility them endogenously assuming market clearing then is maximized when anthropogenic disturbance in a we would set local supply of stock z in period t forest parcel, including the extraction of timber and equal to local demand for stock z in period t , NTFPs by households, is minimized over time. In this co** case, the harvest-right holder’s problem is to allocate ∑ Hpxzt() zt+= Hp xzt () zt Fp zt (), zt (8.36) her illegal harvest prevention budget over time such x that aggregate household harvests of stocks from the where we explicitly note that optimal harvest vol- forest parcel over time are minimized. umes are a function of stock prices (this assumes that harvest product z is not exported from the local area; if it is the market demand curve for z would 8.3.2.3 Total harvest volume and value have to include appropriate external demand). Once we have solved the harvest-right holder of Finally, we would set local household labor supply stock z’s problem and each household’s harvest in time period t equal to a demand function for it to effort problem over stock z in forest parcel x the solve for market clearing wage levels, aggregate level of stock z ’s harvest in parcel x in * H period t is given by H , hh h h h h xzt DEW−−*() 2 ds EW* ()= BW (), (8.37) ∑∑txztt xtxztt( ) tt hx=1 * co* * HHHxzt=+ xzt xzt (8.34) where we explicitly note that optimal harvest effort levels are a function of wages. * subject to the constraint that HSxzt≤ xzt. I f t h e s o l u - tions to the household and harvest-right holder problems generate a collective harvest level that * 8.3.3 Solving the tier 2 model is greater than or equal to a parcel’s stock (i.e., HSxzt≥ xzt * ) t h e n w e s e t HSxzt= xzt and all subsequent collective To solve the tier 2 timber and NTFP model given by harvests of z f r o m x are equal to 0 (i.e., equations 8.23 to 8.37 we need at a minimum the * * HHxz,1 t++=== xz ,2 t ... 0 ) u n l e s s t h e f o r e s t p a r c e l i s r e - following data: a map of forest parcels as of period colonized by individuals of stock z at a later date. t = 1, stock levels of all relevant stocks in all forest 146 THE PROVISIONING VALUE OF TIMBER AND NTFP parcels as of period t = 1, growth relations for each diff cult and assuming rational choice is a more biological stock, harvest production functions, a tractable method for approximating household spatial database of harvest rights as of period t = 1, decision-making ( Manson and Evans 2007 ). net cost of efforts to prevent illegal harvesting in In tier 1 we can only estimate harvest value over each forest with property rights, information on the time if we assume the baseline or current period effectiveness of efforts to prevent illegal harvest, a conditions hold into the future. Steady-state solu- map of the household locations on the landscape as tions will not accurately predict long-run values if of period t = 1, and a vector of (exogenously or the landscape’s demographic, socio-economic, or endogenously def ned) stock prices and wages, ecological conditions are in f ux. For example, from periods t = 1 to T. Household density across increasing unemployment in cities or increasing the landscape and the spatial distribution of harvest timber and NTFP prices to wage ratios may encour- rights for periods t > 1 can remain f xed at t = 1 lev- age greater harvest effort in the future, leading to els or can be changed according to some scenario. harvest rates that exceed growth rates and forests The location of forest parcels should remain f xed at with rapidly declining resource stocks. Further a period t = 1 levels unless a scenario of LULC change steady-state analysis ignores the various distur- indicates a parcel will be afforested, reforested, or bances or shocks to the ecosystem (e.g., f res or pest clear-cut in some period t > 1. outbreaks) that could abruptly change stock levels Running the tier 2 model will be computationally and biomass growth rates. Therefore, the applica- challenging. We can use several simplif cations to tion of a steady-state framework on a landscape make the solution process a bit easier. First, we could with rapidly changing conditions or subject to peri- divide the landscape into regions and def ne several odic disturbance should only be used for short-term household types where each region-household type calculations. combination is also indexed by h = 1 , 2 , . . . , N . I n t h i s The tier 2 model estimates the rate and value of case, total household harvest of z i n x a t t i m e t w i l l timber and NTFP harvest over time assuming that N w hhH h be given by ∑ h=1 txzt w h e r e ωt i s t h e n u m b e r o f households and harvest-right holders are motivated households of type h on the landscape at time t . by net revenue maximization. Unlike tier 1, stock Second, we could create a few very large forest par- harvest and stock levels in forest parcels can change cels by combining many unique forest parcels. Third, over time. However, this added level of realism also we could ignore capital inputs and costs. places an increased data collection and modeling complexity burden on the user. A tier 2 approach to 8.4 Limitations and next steps valuing the provisioning service provided by forests on a landscape will take considerable time and The tier 1 models assume that households and har- effort to implement. Further, as in tier 1, in tier 2 we vest-right holders make decisions that are net reve- assume households and harvest-right holders make nue maximizing with full information of options. In decisions that are net revenue maximizing with full reality, most harvesters have limited information information of options. In reality actors on a land- and use “bounded rationality” when making deci- scape make many sub-optimal decisions due to sions. Because of information and computation imperfect information and in many cases the inabil- limitations, agents use “rules of thumb” to make ity to actually determine what is optimal. decisions that may not be optimal but are accepta- In this chapter we have presented methods for ble ( Manson and Evans 2007 ; Brown 2008 ). In addi- estimating the volume and value of timber and tion, households or harvest-right holders may care NTFPs extracted from forested parcels on a land- about other things, such as environmental steward- scape. We have neglected uncertainty. The simplest ship or the provision of other ecosystem services, way to add more robustness to modeled tier 1 or tier besides a strict focus on returns from harvesting. 2 results is to simulate the response of the model to Incorporating different preferences into the models variation in input values. For example, we can run is possible by changing the objective function. the tier 1 exclusive harvest-right model with a range Incorporating bounded rationality, however, can be of typical biological stock prices, wage rates, and the LIMITATIONS AND NEXT STEPS 147 interest rates to approximate the range in potential the most part, remain in forest from time t = 1 to T steady-state exclusive harvest volumes and values (if it is completely drained of all of its stocks by time (see Section 8.2.4). We can also explore how sensitive T then it is no longer a forest); these are not models tier 1 and 2 modeled results are to changes in func- of land-use change. To model such changes we rely tional relationships, such as harvest rates, growth on LULC change scenarios that are created with rates, and expected f nes for illegal harvesting. For land-use change models. These models can provide example, climate change could change stock growth output that would be useful to such LULC change rates in forests over time ( Aber et al. 2 0 0 1 ) . W e c o u l d models. Specif cally, estimates of the NPV of forest explore the ramif cations of such changes on harvest harvest from forests across a landscape along with volumes and values ( Irland et al. 2 0 0 1 ) b y v a r y i n g estimates of the NPV of returns from alternative G ( S ) for all relevant stocks in a manner that is con- uses of forested areas could be used by LULC sistent with climate change estimates. change models to determine which forest parcels We can also add structural uncertainty and risk to would most likely remain forested. the model. For example, we could assume a household harvester will not always f nd wage employ- References ment while not harvesting. In this case we could make the prospect of f nding wage work probabilis- Aber, J., Neilson, R. P., McNulty, S., et al . (2001). Forest tic. And we could add the risk of forest disturbances, processes and global environmental change: Predicting such as f re or disease ( Dale et al. 2001 ), into stocks’ the effects of individual and multiple stressors. Bioscience, 51 , 735–51. equations of motion (Sohngen and Sedjo 1996 ). It Adger, W. N., and Luttrell, C. (2000). Property rights and would be useful then to map out how the inclusion the utilisation of wetlands. Ecological Economics, 35 , of risk parameters in the model causes it to deviate 75–89. from deterministic solutions. Arnold, J. E. M., and Pérez, M. R. (2001). Can non-timber In addition to examining uncertainty, it is impor- forest products match tropical forest conservation and tant to consider the negative impacts that stock har- development objectives? Ecological Economics, 39 , vests can have on other ecosystem processes and 437–47. services (e.g., Luoga et al. 2000 ). If we can modify Barbier, E. B., and Burgess, J. C. (2001). The economics of LULC def nitions according to the harvest volumes tropical deforestation. Journal of Economic Surveys, 15 , and stocks calculated in the models above then the 413–33. ramif cations of predicted harvests can be incorpo- Bawa, K. S., and Seidler, R. (1998). Natural forest management and conservation of biodiversity in tropical for- rated into almost every other ecosystem service ests. Conservation Biology, 12 , 46–55. model described in this book. For example, if the Belcher, B., Ruíz-Pérez, M., and Achdiawan, R. (2005). water quality function is a function of forest type Global patterns and trends in the use and management (degraded versus intact) then harvest volumes and of commercial NTFPs: Implications for livelihoods and stocks as calculated in the models above will affect conservation. World Development, 33 , 1435–52. water quality model output. Alternatively, we can Birdyshaw, E., and Ellis, C. (2007). Privatizing an open- modify other ecosystem service model outputs ex post access resource and environmental degradation. with results from this model. For example, suppose Ecological Economics, 61 , 469–77. we have data on the carbon storage potential of a rela- Boot, R. G. A., and Gullison, R. E. (1995). Approaches to tively intact forest type. If modeled harvest volumes developing sustainable extraction systems for tropical in a parcel with this forest type are high then the par- forest products. Ecological Applications, 5 , 896–903. Brown, D. R. (2008). A spatiotemporal model of shifting cel’s predicted carbon storage in the aboveground cultivation and forest cover dynamics. Environment and and belowground biomass pools could be scaled Development Economics, 13 , 643–71. downward to ref ect modeled stock loss in the forest. Browning, M., and Crossley, T. F. (2001). The life-cycle Finally, a major source of global land-use change model of consumption and saving. Journal of is the clearing of forests to open land for agricul- Environmental Perspectives, 15 , 3–22. tural or residential use ( Repetto and Gillis 1988 ). Clark, C. W. (1990). Mathematical bioeconomics: the optimal Here we value harvests in forests that begin and, for management of renewable resources . Wiley, New York. 148 THE PROVISIONING VALUE OF TIMBER AND NTFP

Conrad, J. M., and Clark, C. W. (1987). Natural resource eco- communal miombo woodlands of eastern Tanzania. nomics: notes and problems. Cambridge University Press, Forest Ecology and Management, 164 , 15–30. Cambridge. Luoga, E. J., Witkowski, E. T. F., and Balkwill, K. (2005). Dale, V. H., Joycel, L. A., McNulty, S., et al . (2001). Climate Land cover and use changes in relation to the institu- change and forest disturbances. Bioscience, 51 , 723–34. tional framework and tenure of land and resources in Damania, R., Milner-Gulland, E. J., and Crookes, D. J. (2005). Eastern Tanzania miombo woodlands. Environment, A bioeconomic analysis of bushmeat hunting. Proceedings Development and Sustainability, 7 , 71–93. of the Royal Society B-Biological Sciences, 272 , 259–66. Maass, J., Balvanera, P., Castillo, A., et al . (2005). Ecosystem Dauber, E., Fredericksen, T.S. and Peña, M. (2005). services of tropical dry forests: Insights from long-term Sustainability of timber harvesting in Bolivian tropical ecological and social research on the Pacif c Coast of forests, Forest Ecology and Management, 214 , 294–304. Mexico. Ecology and Society, 10 , 17. Engel, S., and Lopez, R. (2008). Exploiting common Manson, S. M., and Evans, T. (2007). Agent-based mode- resources with capital-intensive technologies: The role ling of deforestation in southern Yucatan, Mexico, and of external forces. Environment and Development reforestation in the Midwest United States. Proceedings Economics, 13 , 565–89. of the National Academy of Sciences of the United States of FAO. (2006). Global forest resources assessment 2005: progress America, 104 , 20678–83. toward sustainable forest management . United Nations Milner-Gulland, E. J., and Clayton, L. (2002). The trade in Food and Agricultural Organization, Rome. babirusas and wild pigs in North Sulawesi, Indonesia. Feder, G., and Feeny, D. (1991). Land tenure and property Ecological Economics, 42 , 165–83. rights: Theory and implications for development policy. Monela, G. C., O’Kting’ati, A., and Kiwele, P. M. (1993). World Bank Economic Review, 5 , 135–53. Socio-economic aspects of charcoal consumption and Gordon, H. S. (1954). The economic theory of a common- environmental consequences along the Dar-es-Salaam– property resource: The f shery. Journal of Political Morogoro highway, Tanzania. Forest Ecology and Economy, 62 , 124. Management, 58 , 249–58. Guariguata, M. R., Cronkleton, P., Shanley, P., et al . (2008). Nagendra, H. (2002). Tenure and forest conditions: The compatibility of timber and non-timber forest prod- Community forestry in the Nepal Terai. Environmental uct extraction and management. Forest Ecology and Conservation, 29 , 530–9. Management, 256 , 1477–81. Nagendra, H., Karmacharya, M., and Karna, B. (2005). Hyde, W. F. (2003). Economic considerations on instru- Evaluating forest management in Nepal: Views across ments and institutions. In Y. Dube and F. Schmithusen, space and time. Ecology and Society , 10 , 24. Eds., Cross-sectoral policy impacts between forestry and Ndangalasi, H. J., Bitariho, R., and Dovie, D. B. K. (2007). other sectors , FAO Forestry Paper 142. Rome. Harvesting of non-timber forest products and implica- Irland, L.C., Adams, D., Alig, R., et al. (2001). Assessing tions for conservation in two montane forests of East socioeconomic impacts of climate change on US forests, Africa. Biological Conservation, 134 , 242–50. wood-product markets, and forest recreation. Bioscience, Núñez, D., Nahuelhual, L., and Oyarzún, C. (2006). Forests 51 , 753–64. and water: The value of native temperate forests in sup- Justice, C., Wilkie, D., Zhang, Q., et al . (2001). Central plying water for human consumption. Ecological African forests, carbon and climate change. Climate Economics, 58 , 606–16. Research, 17 , 229–46. Parker, D. C., Manson, S. M., Janssen, M. A., et al . (2003). Larson, B., and Bromley, D. (1990). Property rights, exter- Multi-agent systems for the simulation of land-use and nalities, and resource degradation: Locating the tragedy. land-cover change: A review. Annals of the Association of Journal of Development Economics, 33 , 235–62. American Geographers, 93 , 314–37. Lopez-Feldman, A., and Wilen, J. E. (2008). Poverty and Pattanayak, S. K., and Sills, E. O. (2001). Do tropical forests spatial dimensions of non-timber forest extraction. provide natural insurance? The microeconomics of non- Environment and Development Economics, 13 , 621–42. timber forest product collection in the Brazilian Amazon. Luoga, E. J., Witkowski, E. T. F., and Balkwill, K. (2000). Land Economics, 77 , 595–612. Economics of charcoal production in miombo wood- Peters, C. M., Gentry, A. H., and Mendelsohn, R. O. (1989). lands of eastern Tanzania: Some hidden costs associated Valuation of an Amazonian rainforest. Nature, 339 , with commercialization of the resources. Ecological 655–6. Economics, 35 , 243–57. Pinard, M. A. and Putz, F. E. (1996). Retaining forest bio- Luoga, E. J., Witkowski, E. T. F. and Balkwill, K. (2002). mass by reducing logging damage. Biotropica, 28 , Harvested and standing wood stocks in protected and 278–95. LIMITATIONS AND NEXT STEPS 149

Polasky, S., Nelson, E., Camm, J., et al . (2008). Where to put for African rangelands. Environmental and Resource things? Spatial land management to sustain biodiver- Economics, 6 , 99–118. sity and economic returns. Biological Conservation, 141 , Tahvonen, O., and Salo, S. (1999). Optimal forest rotation 1505–24. with in situ preferences. Journal of Environmental Repetto, R. C. and Gillis, M., Eds. (1988). Public policies and Economics and Management , 37 , 106–28. the misuse of forest resources. Cambridge University Press, Thwaites, R., De Lacy, T., Li, Y. H., et al. (1998). Property New York. rights, social change, and grassland degradation in Ricketts, T. H. (2004). Tropical forest fragments enhance Xilingol Biosphere Reserve, Inner Mongolia, China. pollinator activity in nearby coffee crops. Conservation Society and Natural Resources, 11 , 319–38. Biology, 18 , 1262–71. Ticktin, T. (2004). The ecological implications of harvest- Scholes, R. J. (1996). Miombo woodlands and carbon seques- ing non-timber forest products. Journal of Applied Ecology, tration. In B.M. Campbell, Ed., The Miombo in transition: 41 , 11–21. woodlands and welfare in Africa. Centre for International Vedeld, P., Angelsen, A., Bojö, J., et al . (2004). Forest envi- Forestry Research (CIFOR), Bogor, Indonesia. ronmental incomes and the rural poor. Forest Policy and Silvertown, J. (2004). Sustainability in a nutshell. Trends in Economics, 9 , 869–79. Ecology and Evolution, 19 , 276–8. Wikramanayake, E., Dinerstein, E., Loucks, C., et al . (2001). Sinha, A., and Brault, S. (2005). Assessing sustainability of Terrestrial ecoregions of the Indo-Pacif c: a conservation nontimber forest product extractions: How f re affects assessment . Island Press, Washington, DC. sustainability. Biodiversity and Conservation, 14 , 3537–63. Wikramanayake, E., McKnight, M., Dinerstein, E., et al . Sohngen, B. L., and Sedjo, R. (1996). A comparison of timber (2004). Designing a conservation landscape for tigers in models for use in public policy analysis. Resources for the human-dominated environments. Conservation Biology , Future, Washington, DC. 18 , 839–44. Sohngen, B., and Brown, S. (2006). The inf uence of con- Williams, M. (2003). Deforesting the earth: from prehis- version of forest types on carbon sequestration and tory to global crisis . University of Chicago Press, other ecosystem services in the South Central United Chicago. States. Ecological Economics, 57 , 698–708. World Wildlife Fund. (2006). Demographic analysis: Terai Swallow, B. M., and Bromley, D. W. (1995). Institutions, Arc Landscape—Nepal. World Wildlife Fund, Nepal governance and incentives in common property regimes Program, Kathmandu, Nepal. CHAPTER 9 Provisioning and regulatory ecosystem service values in agriculture

Erik Nelson, Stanley Wood, Jawoo Koo, and Stephen Polasky

9.1 Introduction yield as a function of climate, soil type, input use intensity, and, sometimes, basin-wide water Ecological processes combine with human labor and resource availability (e.g., Fischer et al. 2002 ; inputs such as fertilizer and irrigation to produce B r u i n s m a 2 0 0 3 ; R o s e g r a n t et al. 2 0 0 5 ; N e l s o n et al. agricultural goods used for food, fodder, f ber, and 2009 ). These broader-scale models do not provide fuel. The ecosystem processes that inf uence agricul- an explicit framework for examining changes in tural production include soil retention, pest control, farmer behavior and prof ts due to policy, price, or nutrient recycling in the soil, water capture, and ani- environmental changes. Nor, in general, do they mal pollination (e.g., Wood et al. 2 0 0 5 ; S w i n t o n et al. assess the ecological consequences of agricultural 2007 ). By contributing to agricultural productivity, production (e.g., Naidoo and Ricketts 2006 ; these processes become ecosystem services. The Cassman and Wood 2005 ). value of these services can be proxied by their con- Conversely, in farm-level models, farmers are tributions to the monetary value of commercial agri- usually represented as making crop and production cultural production or the utility value of subsistence method choices that are expected to maximize their agricultural production. Because of the agricultural economic prof ts. In farm-level models, the value of sector’s importance in regional economic develop- a change in an ecosystem service input is equal to ment, especially in developing countries (Byerlee the change in prof t it induces. For example, the et al. 2009 ), estimating the potential agricultural out- value of an increased supply of irrigation water on put on as yet unconverted natural habitat may also a farm can be estimated by the change in farm prof- be of interest because it represents production value its due to the additional water volume. Farm-scale foregone. models can be applied to understand how farmers There is a range of approaches for estimating the might respond to taxes or subsidies (e.g., Just and value of agricultural production and the contribu- Antle 1990 ; Wu et al. 2004 ; Wossink and Swinton tion that particular ecosystem services make to it. 2007 ) and—if the models relate agriculture produc- These approaches are largely differentiated by their tion to environmental impact—how they might geographical scale, the degree to which they incor- respond to payments for ecosystem services (e.g., porate behavioral responses by farmers to changes Holden 2005 ; Antle and Valdivia 2006 ; Antle et al. in biophysical and market variables, and the extent 2007 ). Further, farm-level models can be used in to which they track the impacts of agricultural pro- conjunction with commodity demand models to duction on ecological processes. assess the impact of agricultural policies and Landscape-scale, regional, or global agricultural changes in ecosystem service inputs on commodity models typically assess expected or potential crop supply and food prices (e.g., Gillig et al. 2004 ;

150 DEFINING AGRICULTURAL SCENARIOS 151

Zilberman et al. 2008 ). When appropriate, farm divided into parcels, indexed by x = 1 , 2 , . . . , X , models can assume alternative optimization strate- where a parcel can be any geographic unit, includ- gies among farmers, such as risk minimizing behav- ing a polygon, grid cell, or hexagon. Let the har- ior by subsistence households in developing vested area (in hectares) of crop c (c = 1 , 2 , . . . , C ) countries. In many cases subsistence farmers adopt grown under production system k ( k = 1 , 2 , . . . , K ) production systems that minimize food insecurity in parcel x over some time period (e.g., a growing risk rather than those that maximize expected prof- season, the whole year) be given by A ckx. In this its (e.g., Dercon 1996 ; Kinsey et al. 1998 ; Kandlikar formulation a crop can represent a single crop and Risbey 2000 ; Luckert et al. 2000 ). (e.g., maize), a crop mix (e.g., an intercrop of maize In this chapter we propose an approach to assess and beans), a crop rotation (e.g., maize-soybeans), and map the expected value of agricultural produc- or a livestock type (e.g., cattle). Let pc b e t h e p e r - –1 tion and the value added by ecosystem services at unit output price of crop c . L e t s ckx b e t h e h a costs the landscape scale. As in other chapters, we present incurred by the farmers in parcel x w h e n p r o d u c - the approach in two tiers that differ by information ing and delivering crop c to market under produc- needs, data availability, model sophistication, out- tion system k . put scope, and treatment of time. In tier 1, we use In subsistence agriculture crops may be pro- broad-scale crop production, yield, and cost maps duced and consumed within a household and not along with crop prices to assess the value of agricul- sold in a market, in which case we may not observe tural production across coarser time steps. When a market price. In these cases we will need to use possible, we also assess the contributions of sup- other proxies for value. For example, subsistence porting and regulating services to output value. agriculture prices could be based on measures of The tier 1 approach can be implemented with exist- the caloric, protein, or micronutrient value of crops. ing maps of agricultural production or productivity Costs in subsistence agriculture could be expressed (e.g. crop yield) or those we create. In tier 1, our in terms of total wages foregone from working in ability to model farmer response to market or envi- another economic sector. From an ecosystem serv- ronmental change is limited. ice perspective, it is also possible to express the In tier 2, we model agricultural production from “price” of farm outputs in terms of appropriated a more detailed perspective and consider the deci- ecosystem services (e.g., water consumed or nutri- sions that farmers make over f ner time steps. ent extracted; see Section 9.5 ) rather than in mone- Unlike tier 1, optimum spatial patterns of agricul- tary terms. tural production across the landscape and changes Production systems can be differentiated on the in agricultural output and input prices can be deter- basis of production outputs, technologies and prac- mined within the tier 2 model. We also discuss tices utilized, and use of inputs (e.g., f ood irriga- methods for evaluating landscape-scale ecological tion, drip irrigation, rainfed commercial, or consequences of agricultural land use, including slash-and-burn subsistence using, say, inorganic the emissions of greenhouse gases (GHGs) as well fertilizer and genetically modif ed seeds). as water and nutrient uptake. Production system def nitions can be general or quite specif c. Examples of general production sys- 9.2 Def ning agricultural scenarios tems come from the Global Agro-Ecological Zone (GAEZ) datasets where a combination of a technol- Our starting point is to generate alternative sce- ogy/input use category (low, medium, and high; narios of land use/land cover (LULC) pertinent to see Fisher et al. 2002) and a binary variable that indi- agricultural productivity. For each LULC scenario cates whether a system is irrigated or rainfed def nes we need to specify the crops grown and the a production system. Alternatively, a production method of production on each agricultural parcel system can be def ned with an exact specif cation of in the landscape, as well as crop prices and crop- the fertilizer, pesticide, irrigation water, and other production costs. As with all other ecosystem inputs applied over the course of a growing season service models in this book, the landscape is and a precise def nition of production methods and 152 PROVISIONING AND REGULATORY ECOSYSTEM SERVICE VALUES IN AGRICULTURE

technology used. In tier 2 the timing of input use is q x, j k . This would allow us to estimate the change in also part of a production system def nition. yield with respect to any change in biophysical If the agricultural production map being evalu- characteristics or managed inputs, including many ated is from the current or a past year then current ecosystem services. However, if such yield func- or past commodity prices and production costs can tions cannot be found we suggest two methods for be used. If we are using a published agricultural deriving information about yield across the land- production forecast we can use the forecast's scape: f rst, using existing yield “lookup” tables or accompanying prices (e.g., the USDA Long-term maps and, second, estimating a yield function with Projections, the OECD-FAO Agricultural Outlooks); observed data. ̂ otherwise we will have to use some other method to Prepared maps of Y c k x can be found for major determine prices. In tier 2 we can exogenously crops for many areas in the world. For example, the def ne all scenario components, as in tier 1, or with USDA-NRCS (2001) has published observed yields additional data and analytical effort we can deter- of various crops in each US county as a function of mine crop-production system choices endogenously soil capability class (a parameter in q ) under typical by assuming prof t maximizing or other farmer management practices. These county-level lookup behavior under a given set of prices. We can extend tables of yield as a function of soil class have been the scenario creation process even further in tier 2 used in conjunction with the behavioral assump- by determining all prices endogenously under the tion of prof t maximization to predict land-use deci- assumption of market equilibrium. sions across the USA in response to a hypothetical carbon sequestration payment policy ( Lubowski et al. 2006 ) and to measure the opportunity costs of 9.3 Tier 1 biodiversity conservation in the Willamette Basin, Oregon, U.S ( Polasky et al. 2008 ). Similarly, Vera- 9.3.1 The supply and use of agricultural Diaz (2008) has calculated expected soybean yields production (and net economic returns) across a gridded map of First we need to gather information on the potential Brazil. For the GAEZ project ( Fischer et al. 2002 ), ̂ or expected yield for crop c grown with production global maps of Y ck x were generated with yield mod- ̂ system k on parcel x , denoted as Yc k x . In our ecosys- els explained by general production management ̂ tem service taxonomy, Y c k x represents the measure categories, ecological characteristics and processes, of the land’s ability to supply the ecosystem service and climate. Naidoo and Iwamura (2007 ) used of agricultural production at point x . For agricul- expected yield maps from GAEZ to plot the expected tural production, ecosystem service supply and use opportunity costs of conservation in forest ecosys- (as def ned in Chapter 3 ) are the same. Expected tems and across the globe. yield is a function of a set of biophysical character- While these ready-made yield maps may be suf f - istics such as soil quality, weather, and animal pol- cient for the purposes of some applications, they will linator abundance, and a set of managed inputs not allow us to answer several questions of interest. such as fertilizer and irrigation water, as well as the For one, static maps of yield are not sensitive to ̂ production system itself. Formally, Y ckx can be repre- changes in ecological processes or climate. Second, sented as, these yield maps may ignore the impacts of atypical conditions on yields, such as droughts, f oods, major ˆ Yckx = gcxk()qj, . (9.1) pest and pathogen infestations, and future climate change. Third, some existing yield maps are com-

where g c is crop c ’s yield function and is explained posed of large spatial units, masking the heterogene- by qx , the vector of biophysical characteristics or ity in local farming systems and ecological ecological processes in parcel x , and j k, the vector of characteristics and processes (e.g., a parcel in a GAEZ managed input levels used in production system k . map covers approximately 10 000 hectares). Ideally we will be able to f nd completely speci- As an alternative, we can use statistical techniques f ed functions g c ( q x, j k ) for all potential values of to estimate a relationship between expected yields TIER 1 153 and input management, ecological processes, and harvested area of each crop-production system in environmental conditions. This estimation process the parcel over the course of the growing season by –1 –1 requires contemporaneous data on crop yield, Y c , its expected ha revenue less its ha production agricultural input use, j , and biophysical data, q , and transportation costs, from enough sites on a landscape that statistically KC ˆ robust yield functions can be estimated with regres- AVxckxcckxckx= ∑∑ A(). p Y− s (9.2) sion techniques (see Pender 2005 for a review of kc==11 regression techniques used to build crop yield func- If there is a price premium for crops grown under tions). It is also useful to have data from a number of certain conditions (e.g., organic apples versus con- growing seasons so variations in climate and atypi- ventionally grown apples) then such price effects cal events such as drought or pest outbreaks can be can be included in Eq. (9.2) by stratifying crop price ref ected in estimated yield functions. by production system, pck . Prices and costs should The advantage of the statistical approach is that it not include production subsidies or cost share pay- allows us to build yield functions that are explained ments from a government even if they are payments by input and biophysical data we can collect. If for ecosystem services. We are interested in assess- these inputs include ecological processes then we ing the net social value of agricultural production in can predict changes in agriculture output due to a this model; taxes and subsidy payments are trans- change in such inputs. For example, if we build fers of wealth from one group in society to another, yield functions that are explained by crop access to and not a creation of value (the societal losses cre- water during the growing season then we could ated by taxation ineff ciencies are beyond the scope estimate the impact of a crop-production system of this chapter). The benef ts of subsidized environ- only receiving half of its prescribed irrigation water mentally friendly production choices, for example, needs. Drawbacks associated with the statistical cleaner water, less soil carbon emissions, are technique include the need for a large data set and accounted for when we model these specif c the necessary time and expertise to assemble and services. analyze the data. In addition, the conf dence in pre- Easily accessible and reliable estimates of pro- dicted yields decreases when estimated yield func- duction costs for most crop-production systems tions are applied with explanatory variable levels grown around the world do not exist (Naidoo and that differ signif cantly from those used to estimate Iwamura 2007 ). USDA-ERS (2009) estimates typi- the yield function (e.g., predicting yield on a land- cal ha–1 production costs for most major commodi- scape signif cantly affected by future climate change ties grown under best management practices in with functions estimated with data from historic the USA. Cost estimates can also be generated climate patterns). using published budget sheets that itemize the cost of most or all inputs used in a production process (e.g., Polasky et al. 2 0 0 8 ) . T h e c o s t o f t r a n s - 9.3.2 The value of agriculture’s provisioning porting produce to a market can be calculated with services road network data and information on per-unit Maps of crop-production areas, expected yields, transportation costs. However, in many cases cost and expected costs can be combined with crop price data will need to be gathered from local experts. If databases to produce a map of expected net agricul- the cost term is dropped from Eq. (9.2), then AV x tural production value. In standard parlance used will measure the gross agricultural production in this book (see Chapter 3 ), crop production in a value in parcel x rather than agriculture’s provi- parcel represents the supply and use of agriculture’s sioning value. Only if production costs are fairly provisioning service in that parcel, while the net uniform over every combination of c , k a n d x on value of this production represents the use value the landscape will gross agricultural production ( Bockstael et al. 2000 ). value give a reliable picture of the spatial variation Expected net production value in parcel x for a in the net value of agriculture across the growing season, AV x, is found by multiplying the landscape. 154 PROVISIONING AND REGULATORY ECOSYSTEM SERVICE VALUES IN AGRICULTURE

9.3.3 Examples of tier 1 modeling The map in Figure 9.1 does not fully indicate the use value of agricultural production in Tanzania in 9.3.3.1 Using ready-made yield maps 2000 for two reasons. First, we have not included W e u s e d E q . ( 9 . 2 ) t o m a p g r o s s a g r i c u l t u r a l p r o d u c - production cost data. Second, the four modeled tion values in Tanzania for the year 2000. We crops comprised only 54% of Tanzania’s total focused on 12 crop-production system combina- cropped area in 2000 (FAO 2008a) and the map tions, formed by combining 4 crops: maize, sor- does not include livestock, egg, or dairy produc- ghum, sweet potatoes, and groundnuts, with each tion. Futhermore, the size of the individual parcels of 3 production systems: high inputs and irrigated (roughly 10 000 ha) limits this map’s relevance for (HI ), high inputs and rainfed (HR ) , a n d l o w i n p u t s informing land-use decisions at more local scales. and rainfed ( LR). The yield and year 2000 harvested area maps used to create the gross agricultural map in Figure 9.1 are from You and Wood (2006 ) and 9.3.3.2 Using yield functions to generate yield maps IFPRI (2008). Commodity prices for all four crops In general, we need to use yield functions versus were calculated using 2000 market prices from sev- yield lookup tables or maps if we want to assess eral east African countries (Tanzanian prices were the impact of input management and ecological not reported; FAO 2008a). See the chapter’s supple- processes on yield. Here we demonstrate a statisti- mentary online material (SOM) for more details on cal method for estimating a maize yield function this illustrative example. for a landscape in east-central Africa. The modeled

Arusha

Dodoma Dar es Salaam Morogoro

Iringa

1–50 51–100 101–150 151–200 201–250 251–300 301–3,229

Figure 9.1 Gross agriculture production value in Tanzania. Parcel (5 minute grid cells) values are measured in thousands of year 2000 US dollars. Data on yield and area devoted to each of the 12 modeled crop-production system combinations come from IFPRI (2008). Year 2000 crop prices are the average market prices observed in ten east African counties in 2000 (Tanzanian prices were not available; FAO 2008a). A white parcel indicates that none of the 12 modeled production systems were produced in that parcel in 2000. See the SOM for more modeling details. TIER 1 155 landscape is a 300 × 450 km region made up of 1 function of fertilizer use and a set of biophysical 551 parcels (5 arc minute, roughly 10 km2 grid explanatory variables. Finally, we used the esti- cells) that straddles f ve countries ( Fig. 9.2a ). mated yield function to predict maize yields on all 1 Ideally, we would use observed yields, input use, 551 parcels on the landscape. and ecological process levels to estimate yield func- We simulated maize yields with data from 31 locations. However, because maize yield data were not tions within the landscape that coincided with the readily available for this region, we f rst used a crop availability of detailed soil prof le information (Batjes growth model to simulate maize yields for the few 2002 ). Daily weather data for the select locations locations on the landscape where detailed input from 1997 to 2003 were obtained from the NASA data were available. We then estimated a linear Langley Research Center Atmospheric Sciences Data regression model to explain simulated yields as a Center POWER Project (http://power.larc.nasa.gov).

0 45 90 180 km (a) (b) (c)

Uganda

D.R.Congo

Tanzania

Rwanda

Tanzania

Burundi

Tanzania Clay Loam Sand < 500 501–600 601–700 701–800 > 800

(d) (e) (f)

< 1.00 1.01–2.00 2.01–3.00 3.01–4.00 > 4.00 < 0.25 0.26– 0.51– 0.76–1.01–1.26– 1.51–1.76– 2.01–2.26– 2.51– > 2.76 0.26–0.51– 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50 2.75 0.50 0.75

Figure 9.2 Expected maize yield in central Africa. The modeled landscape straddles f ve countries and includes 1 551 parcels ((a) 5 minute grid cells). The triangles indicate the locations of maize yield simulations. Explanatory variables used to explain yields include (b) soil texture ( TXTC ), (c) rainfall data in mm season–1 in 2000, and (d) soil organic carbon ( SOC ) percentage in the top 100 cm of soil. Expected yield of maize in Mg ha–1 using the estimated yield function (Eq. (9.3)) is given in (e). We also show the expected maize yield increase in Mg ha–1 i f SOC percentage in the top 100 cm of soil increased by 1% (f). See Table 9.1 and the SOM for more details on the estimated yield function used in this example. 156 PROVISIONING AND REGULATORY ECOSYSTEM SERVICE VALUES IN AGRICULTURE

For each of the 31 locations we assumed that maize To predict yield in each of the landscape's 1 551 was cultivated once a year in six separate but iden- parcels, Eq. (9.3) was applied using a parcel’s data tical plots differentiated only by application rates of on TXTC , RAIN in the year 2000, SOC , and the inorganic nitrogen (N) fertilizer (0, 20, 40, 60, 80 100 assumption that 5 kg of N ha–1 would be applied kg N ha–1 ). Using this set of input data and the ( Morris et al. 2007 ; see Figure 9.2a –d). The resulting Decision Support System for Agrotechnology map depicts the expected maize yield on each par- Transfer (DSSAT) Crop Systems Maize Model v4.02 cel on the landscape (Figure 9.2e ) and can be used in ( Jones et al. 2003 ), growth of a widely adopted, long conjunction with a maize harvested area map, maturity maize variety (measured in Mg ha-1) was maize prices, maize production costs, and Eq. (9.2) simulated annually at all 31 locations under the 6 to determine the expected net value or use value of different N application for the years 1997 through maize production across the landscape. 2003, creating 1 302 simulated maize yield observa- T h e p o s i t i v e c o e f f cients on the N variables in the tions (31 locations × 6 plots × 7 years). linear yield Eq. (9.3) imply that each increment of N Next we used simple regression techniques to esti- fertilizer application will generate a proportionate mate the functional relationship between the 1 302 increase in yield. At higher levels of N fertilizer appli- yield observations and several biophysical variables cation (or any other managed input use), the yield that are available everywhere on the landscape. response declines ( Tilman et al. 2 0 0 2 ) . S u c h d i m i n i s h - These variables describe soil texture ( TXTC , clay = 1 ing returns to inputs are rarely encountered in Africa and non-clay = 0; see FAO/IIASA/ISRIC/ISS-CAS/ because of limited use of inputs like fertilizer (e.g., 5 JRC 2008), growing season precipitation ( RAIN , mm; kg of N ha–1 is very low) and irrigation water ( Crawford NASA POWER Project), soil organic carbon percent- et al. 2003 ). However, if we are predicting yield with a age in the top 100 cm of soil ( SOC ; Batjes 2002 ), and linear function in an area where managed input use the managed input of inorganic nitrogen (N) ferti- can be high then the value range of the explanatory lizer application over the growing season (NFRT , kg variables should be limited to the range used to esti- N ha–1 ). The ordinary least squares estimate of a lin- mate the function. Alternatively, we can use a non- ear specif cation of the yield function is linear model to estimate yield in such systems. Besides using the map of estimated maize yield Yˆ =+0.02 (0.034NFRT ) + (0.012 NFRT × to determine net values of agricultural production maize (9.3) TXTC)+ (0.001 RAIN× SOC) on the landscape, we can also use Eq. (9.3) to deter- ̂ mine the change in estimated maize yield, D Ym a i z e , ̂ –1 w h e r e Ym a i z e is maize yield in Mg ha . See Table 9.1 and due to small changes in TXTC , RAIN , SOC , or NRFT . the chapter’s SOM for regression equation details. For example, the organic carbon content of soil,

Table 9.1 Regression results for maize yield in an African landscape

Variables Estimated Eq. (9.3) Descriptive statistics of variables

Regression Standard p value Variable Variable Variable Variable coeff cient error mean std. dev. min. max.

Independent Maize yield (Mg ha-1) 2.8 1.9 0 8.0 Dependent Intercept 0.020 0.068 0.767 NFRT 0.034 0.001 0.000 50 34 0 100 NFRT × TXTC 0.012 0.001 0.000 27 35 0 100 RAIN × SOC 0.001 4.03 × 10–5 0.000 1 038 743 163 3 374 N 1,301 R2 0.67 Adjusted R2 0.67 TIER 1 157

SOC , ref ects an ecological process that inf uences et al. 2005 ; Swinton 2005 ). In this analysis we allow yield and is controlled to some degree by LULC for the possibility that changes in ecosystem service and land management choices on the landscape. inputs can lead to changes in other input use, crop- 1 Therefore, if the land could be managed such that production system combination choices (A c k x ) , p r o - 1 1 SOC increased by 1 percentage point in a parcel duction costs (s c k x ) , a n d c r o p p r i c e s (p c ) . F o r e x a m p l e , then the expected change in maize yield in the par- in response to honeybee declines, farmers of animal- ∂Y ̂ cel would be equal to 0.001 × Rain (i.e., m a i z e ; pollinated crops are experimenting with ways to ΋∂ S O C see Figure 9.2f ) In the next section we discuss in attract native pollinators to their f elds ( Kremen et al. more depth a method for f nding the service value 2008 ). Such experimentation will change input costs provided by changes in ecological processes and and may change cropping practices. In tier 1 we conditions mediated by land-use and management do not include a formal method for determining choices. changes in A ckx , pc , o r s ckx in response to changes in ecological processes (we do in tier 2). As presented here, we gain the greatest f exibility 9.3.4 Measuring the value of regulatory and in measuring the incremental value of regulatory or supporting ecosystem service inputs to supporting services when we use yield functions agriculture that are continuous in their levels of provision. A change in the supply of one or more ecological However, in some cases we can still approximate processes that is an input into agricultural pro- the impact of a change in the provision of a service duction (i.e., an ecosystem service) can change the on production values with expected yield maps, as expected value of crop production. For example, we illustrate next. the net value generated by pollinator-dependent crop can decline if the number of animal pollina- 9.3.4.1 An example of supporting service valuation: tors in the parcel declines. We can measure the the case of surface water supply in Tanzania value of the change in the regulating or supporting Surface water that f ows across a landscape can con- ecosystem service by comparing the expected net tribute to agricultural production if it is used for agricultural value before and after the change in irrigation. We used the map and datasets discussed 0 0 the service. Let qi x a n d q j x b e t h e i n i t i a l l e v e l o f e c o - in Section 9.3.3 and a version of Eq. (9.4) to estimate 1 1 system services i a n d j i n p a r c e l x a n d l e t q i x a n d qj x the decline in gross agricultural production value be the levels after the change. Using the tier 1 yield from a small reduction in surface water availability. function, we can def ne the pre- and post- Specif cally, we estimated the percentage decline in ̂0 0 0 change expected yields as Y c kx =g c ( jk ,q i x ,q j x ,q ~x ) and year 2000 aggregate gross production values across ̂1 1 1 Y c k x =g c (j k ,q ix ,q j x ,q ~x ) , r e s p e c t i v e l y , w h e r e q~ x d e f nes Tanzania if 5% of the area of high input irrigated all ecological process inputs other than i a n d j . I f ( HI ) maize and sorghum found in each parcel could ̂0 ̂1 appropriate, Y c k x a n d Y c k x c a n i n c l u d e e v e n m o r e no longer be irrigated due to reduced surface water changes in ecological processes and/or changes in f ow (see Figure 9.3 ). In this analysis we ignored managed inputs, such as water use for irrigation, irrigation via groundwater (such practice is rare in 0 1 def ned with ji x , ji x , a n d j ~x . Tanzania, FAO 2008b) and we ignored sweet potato The change in the net value of agricultural output and groundnut systems because they are not typi- due to the change in the supply of ecosystem serv- cally irrigated. ices i and j is given by ∆ AVx , The maize and sorghum yield maps do not explicitly include surface water supply for irrigation as an KC explanatory variable. Instead, we assumed that ∆ AV= ⎡⎤ A1111()(). p Yˆˆ−− s A 0000 p Y − s xckxcckxckxckxcckxckx∑∑⎣⎦ (9.4) kc==11 farmers of the hectares formerly in HI maize or sorghum production, in reaction to the loss in irriga-

w h e r e ∆ AVx can also be interpreted as the value of tion water, grew the same crops but under rainfed the joint change in ecosystem services i a n d j a s conditions with less intensive input use (LR pro- expressed through agricultural production ( Shiferaw duction). In many cases farmers without access to 158 PROVISIONING AND REGULATORY ECOSYSTEM SERVICE VALUES IN AGRICULTURE

Arusha

Dodoma Morogoro Dar es Salaam

Iringa

0.1%–0.5% 0..6%–1.0% 1.1%–1.5% 1.6%–2.0% 2.1%–4.1%

Figure 9.3 Decline in gross agricultural production value due to reduced irrigation in Tanzania. The map gives the percentage decline in a parcel’s 2000 agriculture gross production value ( Figure 9.1 ) when 5% of each parcel’s area in irrigated production (HI ) of maize and sorghum were replaced with low input and rain-fed (LR ) production. Parcels shown in white do not contain any of the 12 modeled production systems in 2000. Price data are the same as those used in Figure 9.1 . See the SOM for more modeling details. irrigation water will reduce the use of costly inputs sons. Third, we propose a framework for modeling due to the lower yields and greater crop failure risk farmer behavior given government policy, prices, associated with rainfed versus irrigated agriculture technology and environmental conditions. Finally, (Feder et al. 1985). we can use the tier 2 model to solve for product out- Assuming prices remain at their 2000 levels our put and input prices by assuming that product analysis indicates that small losses in irrigation water output and input markets clear (i.e., aggregate would have caused a range in parcel-level gross value quantities supplied balance aggregate demand). losses of 0 to 4.1%. In aggregate, this decline in irrigation capacity would have caused a 0.02% decrease in 9.4.1 The supply, use, and value of agriculture’s the country-wide gross production value represented provisioning ecosystem services in Figure 9.1. These results, while simple and illustrative, would indicate that the ecosystem service of irri- The tier 2 yield function considers the temporal pat- gation may currently add little to the net value of tern of input use across a growing season. Let z maize and sorghum production in Tanzania. index growing seasons and let b = 1, 2, . . . , B index the sub-periods within season z . The net value of 9.4 Tier 2 agricultural production in parcel x i n t i m e p e r i o d z , AV xz , is given by Tier 2 extends the tier 1 approach in several ways. ⎛⎞⎡⎤C First, we account for the timing of input use during ⎜⎟K ⎢⎥ˆ ∑ pAcz ckxz g cz()jjqq11 kxz,...., Bkxz ; xz ,..., Bxz AV = ⎢⎥c 1 the growing season. Second, we can track agricul- xz ⎜⎟∑ ⎣⎦= (9.5) ⎜⎟k=1 ⎜⎟ ture production value over multiple growing sea- ⎝⎠−Akxz()qqq11 zjjj kxz ++...bz bkxz ++ ... Bz Bkxz TIER 2 159

w h e r e p cz is the market price of crop c i n s e a s o n z (or was in the previous season—and all other input use

pckz if there is a price premium for crop c grown under and prices remain the same across growing sea- production system k ) , A ckxz i s t h e a r e a o f x i n c r o p c sons—then ∆AV x = AV xz+1 – AV xz represents the value under production system k i n s e a s o n z , g ĉ z is the esti- of the vegetation change in parcel x a s e x p r e s s e d mated yield function of crop c i n s e a s o n z and is through agricultural production. explained by an intraseasonal series of inputs associated with production system k (j 1 k x z , . . . , j B k x z ) a n d b i o - 9.4.2 Tier 2 scenarios physical properties and processes in x (q , . . . , q ), 1 x z B x z We propose two approaches to determining agri- A is the area of production system k i n x in season kxz cultural production scenarios in tier 2. One approach z, a n d q is a vector of per-unit input prices during bz is similar to that used in tier 1 in which we def ne all sub-period b i n s e a s o n z , including any harvested relevant model parameters and variables exoge- crop transportation costs (the vector product q j b z b k x z nously, including crop-production system combina- gives the total ha–1 input use costs incurred in sub- tions, managed input use levels, and prices for all period b o f s e a s o n z under production system k ). For sub-period and parcel combinations for growing simplicity's sake, we assume that the production sys- seasons z = t , t + 1, . . . , T . tem determines production costs and not the crop- Alternatively, we can assume that farmer behav- production system (i.e., costs in equation (9.5) are a ior will be guided by objectives such as prof t maxi- function of k a n d n o t c). However, the modeler could mization. To make this approach more tractable we differentiate costs across each crop-production sys- f rst def ne those parcels that will be used for agri- tem choice (i.e., A in equation (9.5) could become kxz culture in each growing period. Then farmers will A and the managed input vectors in equation (9.5) ckxz choose amounts of different crops to cultivate (A ), could be indexed by c as well). Because a crop’s ckxz inputs (j j ), and to the extent possible bio- response to inputs over time can change due to tech- 1kxz, . . . , Bkxz physical properties (q , . . . ,q ), over each growing nological growth (e.g., Rosegrant et al. 2001 ; Alston 1 xz Bxz season z such that the expected net present value et al. 2009 ) or climatic shifts, we index the yield func- (NPV) of their agricultural production returns over tion g ̂ by seasons. For example, from 1966–2005, c z time are maximized, maize yield in the USA improved 100% largely due to farmer adoption of improved maize production KC ⎛ T ⎡⎤pA gˆ (,...,;,...,)jjqq+ eA technology ( Cassman and Liska 2007 ). ∑∑kc==11cz ckxz cz11 kxz Bkxz xz Bxz ckxz ckxz max⎜ ⎣⎦− ⎜∑ (1+ r )zt− As in tier 1, we can link tier 2 yield equations with ⎝ zt= models of ecosystem service input supply. For exam- K T A (...)qqjj++ ⎞ ∑ k=1 kxz11 t kxz Bz Bkxz ple, we can determine the amount of surface water ∑ zt− ⎟ z=1 (1+ r ) ⎠⎟ f ow available for the irrigation of crop c u n d e r p r o - duction system k i n s u b - p e r i o d b o f s e a s o n z u s i n g (9.6) the tier 2 surface water f ow model (Chapter 4 ). In addition the vectors q , . . . ,q can include several 1 xz Bxz K ABqqjj++... ≤ other tier 2 outputs from models noted in this book, subject to ∑ k =1 kxz()11 z kxz Bz Bkxz xz for –1 including pollinator abundance (Chapter 10 ), soil all z where eckxz is the ha subsidy (or tax if negative) erosion ( Crosson et al. 1 9 9 5 ) , a n d t e m p o r a r y f ood- for producing crop c under production system k , r is ing ( Chapter 5 ). As in tier 1, the change in the net the relevant interest rate, and B xz is the budget con- value of agricultural output on parcel x d u e t o t h e straint for parcel x during season z . In Eq. (9.6) we change in the supply of an ecosystem service input assume parcel x is in agricultural land use from is given by evaluating Eq. (9.5) pre- and post-change growing season t to T. If the farmers in parcel x have assuming all input use and prices not affected by the access to capital (e.g., smallholder f nancial serv- change in the ecosystem service input remain con- ices) then the budget constraint can be loosened or stant. For example, if vegetation changes are made even dropped. We explicitly include any subsidies on the landscape such that temporary f ooding in (or taxes) in Eq. (9.6) because, while they do not parcel x during growing season z is much less than it affect the net value of agricultural production 160 PROVISIONING AND REGULATORY ECOSYSTEM SERVICE VALUES IN AGRICULTURE explicitly, they do affect farmers’ private decisions those that equate supply and demand for all agricul- (and the provision of other ecosystem services; e.g., ture products and inputs (e.g., Zilberman et al. 2008 ). Antle and Valdivia 2006 ). If the subsidy is a function To establish the trajectory of output market-clearing of the amount of an ecological process or good prices we need to def ne functional relationships

produced then the A ckxz after eckxz can be replaced by between the demand for each modeled crop and an appropriate ecological production function. managed inputs and their prices. Specif cally, let

L e t Ac k x z (p z , ez , q 1 z , . . . ,q B z ) , j1 k x z ( pz , e z , q 1 z ) , . . . , jB k x z ( pz , D cz (p cz ) represent the demand in season z f o r c r o p c

e z , q B z ) , a n d t o t h e e x t e n t t h e y a r e m a n a g e a b l e , q 1 x z ( pz , produced on the study landscape (we can further

e z , q 1z ) , . . . , q B x z (p z , e z , q B z ) , f o r a l l z f r o m g r o w i n g s e a - segment by production system if it is relevant). sons t t o T indicate the choices made in parcel x t h a t Further, let F1 z (q 1z ), . . . , F Bz (q Bz ) represent the supply are expected to solve the maximization problem of inputs available on the landscape in the sub- described by Eq. (9.6). All solutions are functions periods of time z . We estimate market equilibrium that are explained by prices p z , e z , a n d q bz (p z a n d e z prices by equating the optimal supply of commodity are vectors of all p cz a n d eckxz a t t i m e z) . I n m a n y c a s e s , c to its demand and, similarly, the demand for inputs choices for growing season z will need to be made equal to their supply and then solve for all prices, before some biophysical values for that season, such XK ⎡A (,pe , q )× as rainfall or irrigation water availability (e.g., Perry ∑∑⎣ ckxz cz ckz bz xk==11 and Narayanamurthy 1998 ), are known. In these ⎛⎞qq(,pe , q ),..., ⎤ (9.7) gDpˆ 11kxz cz ckz z Bkxz = () cases, we hypothesize that farmers will maximize cz⎜⎟⎥ cz cz ⎝⎠(,pecz ckz , q Bz );,...,qq1 xz Bxz ⎦ over expectations for biophysical values. Evaluating Eq. (9.5) with the choices that solve Eq. (9.6) gives and A V * , . . . , A V * , t h e s e r i e s o f m a x i m i z e d p r o f t levels XK x t x T A peq Fq in parcel x from growing season t t o T . ∑∑ kxzj bkxz(,, z z bz )= bz ( bz ) (9.8) xk==11 As with tier 1 models we can use alternatives to prof t maximizing assumptions when considering for b = 1, 2, . . . , B . In general, as the supply of an farmer behavior. Subsistence farmers may be prima- agricultural output or input increases, its price rily concerned with reducing the chance of crop fail- tends to fall, all else equal, because it has become ure in any given growing season rather than less scarce in the market. In many cases demand maximizing the expected value of yield or prof t (e.g., functions for agricultural output, whether describ- Dercon 1996 ). For example, if a landscape is exposed ing demand at regional or global scales, are indiffer- to an extended drought period (rainfall levels in ent to point of output origin. In such cases we can assume a certain portion of demand will be met by j 1 k x z , . . . , j B k x z a r e e x p e c t e d t o b e l o w o r u n r e l i a b l e d u r - ing the growing season) subsistence farmers are likely agricultural production outside of the study land- to make very different crop-production system choices scape and let D cz ( pcz ) represent the residual demand than a prof t maximizing farmer; for example, mini- to be met locally. Further, the supply of inputs avail- mizing the potential severity of food shortages during able to farmers, and thus relevant supply functions, the drought period. In such cases we can replace the varies greatly across the world. For example, farm- expected net revenue function in Eq. (9.6) with a util- ers in the USA can generally buy fertilizer and other ity function described by a high degree of risk aver- inputs from a global supply market. Farmers in sion (see Bardhan and Udry 1999 for more details). Africa may be limited to regional market with less To fully specify farmer behavior in a scenario, choice and higher prices (e.g., Morris et al. 2007 ). whether modeled as maximizing the NPV of revenue Solving Eq. (9.6) can be computationally diff cult, streams or, more generally, of utility streams, we even if 1) we set prices exogenously, 2) assume that need to establish output and input prices. We can farmers in x have no control over q1 xz , . . . , q Bxz , and 3) assume that j , . . . , j and q , . . . , q are not simply assume some p z a n d q bz for z = t > , . . . , T (see 1 kxz Bkxz 1 xz Bxz Section 9.2 ) or, to maintain theoretical consistency dependent on current and past management choices with typical economic equilibrium analyses, we can on neighboring parcels (i.e., spatial production endogenously determine market-clearing prices, i.e., externalities). When we include production MAPPING THE IMPACTS OF AGRICULTURE ON IMPORTANT ECOLOGICAL PROCESSES 161

externalities (e.g., water use upstream affects water ⎛ KC ⎛⎞25YCH4 + 298 FluxGHG= 0.2727 A Yˆ ckx availability downstream) and simultaneously deter- xz ⎜ ∑∑ ckxz ckx ⎜⎟ ⎝ kc==11 ⎝⎠()YDN2Ockx + YIN2Ockx mine prices with Eqs. (9.7) and (9.8) then solving KC ⎛⎞25CH4 + 298 ⎞ + A ckx Eq. (9.6) can be exceedingly diff cult (see Conrad ∑∑ ckxz ⎜⎟⎟ kc==11 ⎝⎠()DN2Ockx + IN2Ockx ⎠ and Clark ( 1987 ) and Holden ( 2005 ) for a review of modeling methods that have been used to solve (9.9) Eqs. (9.6) through (9.8)). We can reduce the degree of complexity by modeling a minimum number of w h e r e FLuxGHG xz i s t h e f ux of CH 4 a n d N 2O o n p a r - crops, production systems, and inputs; keeping the cel x in growing season z in Mg of carbon equiva- number of time steps between t and T small; and lents, YCH4ckx , YDN2O ckx , a n d YIN2O ckx r e p r e s e n t t h e using coarsely def ned maps to keep the number of direct emissions of CH 4 , direct emissions of N2 O , agricultural parcels low. and the indirect emissions of N2O, respectively, per unit yield of crop c o n p a r c e l x grown under production system k , a n d CH4 , DN2O , a n d IN2O r e p - 9.5 Mapping the impacts of agriculture ckx ckx ckx resent the direct emissions of CH , t h e d i r e c t on important ecological processes 4 emissions of N 2 O, and the indirect emissions of N2 O , Agricultural land use and management practices respectively, per hectare of crop c on parcel x grown can have profound effects on the supply and condi- under production system k . The constants 25 and tion of ecosystem services and biodiversity on the 298 in Eq. (9.9) convert CH 4 a n d N2 O, respectively, to landscape (Swinton et al. 2007 ). We can measure CO2 -equivalents (IPCC 2007). The constant 0.2727 in many of these effects using the models presented Eq. (9.9) converts CO2 measurements to elemental elsewhere in this book. Below we present a simple carbon equivalents. By indexing each coeff cient by system for tracking three biophysical processes x we allow for the possibility that emission rates can affected by agriculture: greenhouse gas (GHG) f ux also be explained by features that vary across the other than carbon dioxide (CO2 ) and water and landscape (soil types, slope, etc). nitrogen use on agricultural f elds. We can use esti- Values for the coeff cients in Eq. (9.9) can come mates of GHG release and water and nutrient use to from IPCC (2006), can be estimated using f eld data, summarize some of the landscape-wide “external” or come from other relevant sources. For example, if costs of an agricultural scenario. we have CH 4 emissions data for multiple f elds of c Agricultural production can be the source of under production system k we could regress the methane (CH4 ) and nitrous oxide (N2 O) emissions, emissions data against corresponding yield esti- two powerful GHG gases (see Chapter 7 for the mates to derive coeff cients YCH4 ck (the estimated accounting of CO2 emissions from agricultural land slope and) and CH4 ck (the estimated intercept). If use). Both livestock and f ooded agriculture sys- some coeff cients are not relevant we can set their tems (e.g., rice paddies) produce CH4 (e.g., Neue values to zero. For example, if CH4 and N 2 O emis- 1993 ; IPCC 2007). A fraction of the N produced in sion data are only given per hectare of crop c grown and applied to f elds, including the N inorganic fer- under production system k then CH4 ckx , DN2O ckx , tilizer and manure, contributes to the formation of and IN2O ckx will be greater than 0 and all other emis-

N 2 O (e.g., Bouwman et al. 2002 ; Dalal et al. 2003 ). sion coeff cients will be equal to 0. Finally, if we

Further, the portions of the N in soil, manure, and multiply FLuxGHGxz by the social cost of carbon biomass residue that do not contribute to the direct (SCC) we estimate the economic damage generated formation of N 2O may eventually contribute to the by GHG f ux (see Chapter 7 for a discussion of the formation of N 2O via other ecological processes SCC). (IPCC 2006). Agricultural systems also modify water and

We can model agriculture-derived CH 4 and N 2 O nutrient cycles on the landscape ( Lesschen et al. f ux by tabulating data on per-area or per-yield 2007 ). For example, by quantifying water use by all f uxes, converting them to CO2 -equivalents, and crops on a landscape (i.e., total crop transpiration) then to elemental carbon. under different scenarios of agricultural production 162 PROVISIONING AND REGULATORY ECOSYSTEM SERVICE VALUES IN AGRICULTURE

we can determine which scenarios result in lower z t h e n Ωxz measures the net decrease in N levels (if consumptive use by crops. Or we may be interested positive) or excess N (if negative). in determining if agricultural practices on a landscape in period z remove more water or nutrients 9.6 Uncertainty than the amount replenished (from natural or human mediated sources), i.e., the net impact. If the water In general the models described above ignore the volumes or nutrient amounts removed are greater issues of stochasticity in agricultural production than those replenished over a growing season then and value. For example, yields may differ from landscapes run a temporary water or nutrient def cit expectations due to variability in weather (e.g., that might, given successive growing season def - droughts, storm damage, and f oods) and pest cits, lead to a water or nutrient supply crisis. and pathogen outbreaks. Further, problems in We can modify Eq. (9.9) to model a water or nutri- transporting crops between farm and processing ent cycle within parcel x in growing season z , centers or markets, post harvest losses, and mar-

KC ket related instability (e.g., volatile prices for ˆ Ωxz= −w xz + ∑∑AYYUUckxz(), ckxz ckx+ ckx (9.10) either crops or inputs) means revenues and costs kc==11 can deviate, perhaps signif cantly, from

w h e r e ωxz is the volume of water or mass of nutrient expectations. added to parcel x during growing season z and the One approach to dealing with biophysical and double summation term represents the amount of market uncertainty in our models is to generate water or nutrient consumed by crops in parcel x in bounds on the range of potential production val- period z. For example, suppose we are tracking water ues by analyzing farmer behavior and production use by crops. In this case YUckx is the volume of water results under a range of climatic and market con- used per unit output of crop c grown under produc- ditions, including various output and input price tion system k o n p a r c e l x w h i l e U ckx is the per hectare scenarios. For example, what agricultural produc- volume of water used when producing crop c with tion values would be generated on a landscape production system k o n p a r c e l x and is independent with high levels and a timely pattern of rainfall of c ’s yield. As with the GHG f ux accounting, the over a growing season, making typical irrigation uptake coeff cient YU is greater than 0 if the use data practices unnecessary, versus the values gener- are a function of yield and U is greater than 0 if some ated during a drought that causes many rainfed or the entire use is explained by area. If we set ωxz = 0 crops to fail and creates high demand for irriga- then Ω xz measures the water used in parcel x by its tion? Or we could look for published relationships crops or, in other words, the amount of water no between climate and crop-production system pat- longer available to the rest of the landscape for use in terns on the landscape and probabilities of pest or period z . I f w e s e t ω xt equal to water production in pathogen outbreak. Then we could adjust maps of parcel x d u r i n g p e r i o d z (i.e., rainfall and irrigation) expected net agricultural production and value then Ω xz measures the net import (if positive) or appropriately ex post, reducing output and value excess water produced on x (if negative). Crop water in areas where disease and pathogen risks are models and databases such as CropWat (FAO 1992), judged to be high. AquaCrop ( S t e d u t o et al. 2 0 0 9 ) , a n d AquaStat (FAO Another approach to incorporating uncertainty is

2008b) can provide data on YW ckx a n d W ckx for many to perform a Monte Carlo analysis of production crop-production systems across the globe. value where we introduce random variation around If we are tracking N use instead of water use, expected values of environmental conditions,

YUckx is the use of N per unit output of crop c under prices, input availability, and farmer behavior to production system k i n x a n d U ckx is N use per unit generate histograms of model outputs. Or, if we are of area. If we set ωxz = 0 t h e n Ω xz measures the gross using the tier 2 framework to model farmer behav- use of N in x d u r i n g t i m e z in (measured in kg). If ior we might compare prof t maximizing to risk we set ωxt equal to N (natural and man-made) pro- minimizing (or utility maximizing) assessments of duction and application in parcel x during period production values. LIMITATIONS AND NEXT STEPS 163

agement decisions in the future (e.g., Mendelsohn et 9.7 Limitations and next steps al. 1 9 9 4 ; S c h l e n k e r et al. 2 0 0 5 ; L u e d e l i n g et al. 2 0 0 9 ; Mendelsohn and Dinar 2009 ). If we have yield func- 9.7.1 Limitations tions that include climate variables then we may be In this chapter we propose two tiers of agricul- able to predict some of the changes in the agricultural modeling, aiming for a range of conceptually tural sector from climate change (see Chapter 18 ). In sound and tractable approaches to estimating and addition, we can model some of the expected effects mapping values of agricultural production and of climate change on agriculture production via the ecosystem services that support agricultural changes in ecosystem services that are inputs to production. However, the app roaches outlined do agricultural production (see Chapters 4 and 10 ). have several limitations that are important to However, major climate change might cause unan- acknowledge. First, we did not explicitly model ticipated changes in crop and animal growth pat- livestock operations. We propose modeling live- terns that cannot be predicted with statistical models stock production systems as if they are a crop with based on current or historic relationships between expected yield (e.g., kg of meat ha –1 , animal units yields and agricultural inputs (crop simulation mod- supported ha–1 y r –1 ) a n d i n p u t u s e d e f ned by a els would appear better able to model yields in production system k . However, the physical size response to climate change; e.g., Parry et al. 1 9 9 9 ) . of intensive and conf ned livestock production Further, predicting commodity supply and demand enterprises may be too small to be represented on in a climate change-affected world will also be very the LULC maps that form the basis of the agricul- diff cult, making the endogenous price determina- tural production models described in this chapter. tion method discussed in the tier 2 approach partic- In such cases we may need to add point data to ularly challenging to implement for future scenarios the LULC maps to account for these systems, or under climate change. account for them outside the mapping framework. A better understanding of the links between live- 9.7.2 Next steps stock production and water and air pollution, ecosystem processes, and biodiversity conservation The output from the agriculture models described is becoming more important as meat demand here can be used for several types of analyses. We increases throughout the world (Delgado et al . can use expected net production value maps to 1 9 9 9 ; S t e i n f e l d et al. 2006 ). determine the opportunity costs of setting aside Second, many agricultural landscapes include a agriculture land for conservation, or conversely, the diversity of crops and systems of input use. Most value added to society from bringing land into applications of the approaches described here agricultural production. We can use maps of agri- focus on major crops, both due to data limitations cultural production values, the value added to agri- and time constraints. This restriction may under- culture production by regulating and supporting estimate production and ecosystem service value ecosystem services, and agriculture-related GHG and distort the spatial distribution of that value— f ux maps along with maps of other ecosystem serv- especially if the modeled systems are likely to be ice use to create a series of service trade-off analyses allocated across the landscape differently than (see Chapter 14 ). non-modeled systems. Further, by ignoring minor Other models presented in this book are able to crops we might miss important niche products on use outputs from the agriculture models described the landscape and overlook important crop alter- here. For example, when we assign fertilizer and natives that could improve environmental condi- pesticide use to each crop-production system and tions on the land while maintaining the value of then map these systems we have also mapped ferti- agricultural output. lizer and pesticide use. The water pollution regula- Climate change is likely to have a large impact on tion model (see Chapter 6 ) can use these data to agricultural yield functions, crop-production system estimate agriculture’s impact on water quality patterns, crops prices, production costs and man- across the landscape. Surface water withdrawals for 164 PROVISIONING AND REGULATORY ECOSYSTEM SERVICE VALUES IN AGRICULTURE irrigation will affect water use in other sectors of the approach. Australian Journal of Agricultural and Resource economy (see Chapter 4 ). We could also use the out- Economics, 50 , 1–15. puts from this model to make predictions about the Antle, J. M., Capalbo, S., Paustian, K., et al . (2007). number of rural households in developing coun- Estimating the economic potential for agricultural soil tries that are involved in timber and non-timber for- carbon sequestration in the Central United States using an aggregate econometric-process simulation model. est product harvesting. In general, the lower the Climatic Change, 80 , 145–71. predicted net values from agriculture the more that Bardhan, P. K., and Udry, C. (1999). Development rural households will look to supplement their Microeconomics, Oxford University Press, New York. incomes with products generated by nearby forests Batjes, N. H. (2002). A homogenized soil prof le data set for (see Chapter 8 ). global and regional environmental research (WISE, version Moving beyond tier 2 to gain even more mode- 1.1). International Soil Reference and Information ling realism implies the use of more elaborate Centre, Wageningen. yield, biophysical, and economic simulation mod- Bockstael, N. E., Freeman, A. M., Kopp, R. J., et al . (2000). els. For example, we can better model crop yields On measuring economic values for nature. Journal of by using one of more specialized crop growth sim- Environmental Science and Technology, 34 , 1384–9. ulation models such as DSSAT, APSIM, or EPIC Bouwman, A. F., Boumans, L. J. M., and Batjes, N. H. (2002). Modeling global annual N O and NO emissions (e.g., Bryant et al. 1992 ). These models, as partly 2 from fertilized f elds. Global Biogeochemical Cycles, 16 , demonstrated in Section 9.3.3 where we used 1080. DSSAT to generate “observed” maize yields, are Bouwman, A. F., Van Der Hoek, K. W., and Van Drecht, G. data intensive and usually require detailed data- (2006). Modelling livestock-crop-land use interactions sets and specialized training to use. Complex mod- in global agricultural production systems. In A. F. els that track nutrient cycles in soils (e.g., Bouwman, T. Kram, and K. K. Goldewijk, Eds., Integrated CENTURY, RothC-26.3, or SCUAF), water balance modelling of global environmental change: An overview of and quality, and other environmental impacts of IMAGE 2.4. Netherlands Environmental Assessment farming (e.g., WATBAL, PERFECT, or SWAT; see Agency (MNP), Bilthoven. W a n i et al. 2005 for a review of all of these and Bruinsma, J. (2003). World agriculture: towards 2015/2030 . other simulation models) are also available. In: J. Bruinsma, ed., An FAO perspective . Earthscan, London. However, these models often require signif cant Bryant, K. J., Benson, V. W., Kiniry, J. R., et al . (1992). calibration effort and data not readily available in Simulating corn yield response to irrigation timings: developing countries. Finally, far more elaborate Validation of the EPIC model. Journal of Production agricultural production models that embed micro- Agriculture, 5, 237–42. economic principles within macroeconomic and Byerlee, D., de Janvry, A., and Sadoulet, E. (2009). general equilibrium analytical components are Agriculture for development: Toward a new paradigm. available to evaluate trajectories of land-use deci- Annual Review of Resource Economics, 1 , 15–31. sions and ecosystem service provision over time Cassman, K. G., and Liska, A. J. (2007). Food and fuel for (e.g., Eickhout et al. 2 0 0 6 ; B o u w m a n et al. 2006 ; all: realistic or foolish? Biofuels, Bioproducts and N e l s o n et al. 2009 ). Once again, however, these Bioref ning, 1 , 23. models are complex and are often only accessible Conrad, J. M., and Clark, C. W. (1987). Natural resource economics: Notes and problems. Cambridge University Press, to the research groups that developed and main- New York. tain the models. Crawford, E., Kelly, V., Jayne, T. S., et al . (2003). Input use and market development in Sub-Saharan Africa: an References overview. Food Policy, 28 , 277–92. Crosson, P., Pimentel, D., Harvey, C., et al. (1995). Soil ero- Alston, J. M., Beddow, J. M., and Pardey, P. G. (2009). sion estimates and costs. Science, 269 , 461–5. Agricultural research, productivity, and food prices in Dalal, R. C., Wang, W., Robertson, G. P., et al . (2003). the long run. Science, 325 , 1209–10. Nitrous oxide emission from Australian agricultural Antle, J. M., and Valdivia, R. (2006). Modelling the supply lands and mitigation options: a review. Australian of ecosystem services from agriculture: a minimum-data Journal of Soil Research, 41 , 165–95. LIMITATIONS AND NEXT STEPS 165

Delgado, C., Rosegrant, M., Steinfeld, H., et al . (1999). Economic and Environmental Impacts . CABI Publishing, Livestock to 2020. The Next Food Revolution . International Cambridge. Food Policy Research Institute (IFPRI), Washington, Intergovernmental Panel on Climate Change (IPCC). DC. (2007). Climate change 2007: the physical science basis. Dercon, S. (1996). Risk, crop choice, and savings: Evidence Contribution of Working Group I to the Fourth Assessment from Tanzania. Economic Development and Cultural Report of the Intergovernmental Panel on Climate Change. Change, 44 , 485–513. Cambridge University Press, New York. Eickhout, B., van Meijl, H., and Tabeau, A. (2006). Intergovernmental Panel on Climate Change (IPCC). Modelling agricultural trade and food production (2006). 2006 IPCC Guidelines for National Greenhouse under different trade policies. In A. F. Bouwman, T. Gas Inventories. In H. S. Eggleston, L. Buendia, K. Miwa, Kram, and K. K. Goldewijk, eds., Integrated modeling of et al. , Eds., Prepared by the National Greenhouse Gas global environmental change: An overview of IMAGE 2.4 . Inventories Programme. Institute for Global Environmental Netherlands Environmental Assessment Agency Strategies, Hayama, Kanagawa. (MNP), Bilthoven. International Food Policy Research Institute, The (IFPRI). Feder, G. Just, R. E., and Zilberman, D. (1985). Adoption of (2008). Spatial allocation model (SPAM) . International agricultural innovations in developing countries: A sur- Food Policy Research Institute (IFPRI), Washington, vey. Economic Development and Cultural Change, 33 , 255–98. DC. Accessed at http://www.mapspam.info Fischer, G., van Velthuizen, H., Medow, S. and Jones, J. W., Hoogenboom, G., Porter, C. H., et al . (2003). Nachtergaele, F. (2002). Global agro-ecological assessment The DSSAT cropping system model. European Journal of for agriculture in the 21st century, Food and Agricultural Agronomy, 18 , 235–65. Organization/International Institute for Applied Just, R. E., and Antle, J. M. (1990). Interactions between Systems Analysis (FAO/IIASA), Laxenburg. agricultural and environmental policies: A concep- Food and Agriculture Organization of the United Nations tual framework. American Economic Review , 80 , (FAO) (1992). CROPWAT, a computer program for irriga- 197–202. tion planning and management by M. Smith, FAO Irrigation Kandlikar, M., and Risbey, J. (2000). Agricultural impacts and Drainage Paper 26, Food and Agriculture of climate change: If adaptation is the answer, what is Organization of the United Nations (FAO), Rome. the question? Climatic Change, 45 , 529–39. Food and Agriculture Organization of the United Nations Kinsey, B., Burger, K., and Gunning, J. W. (1998). Coping (FAO). (2008a). [Online]. Available: http://faostat.fao. with drought in Zimbabwe: Survey evidence on org/default.aspx . responses of rural households to risk. World Development, Food and Agriculture Organization of the United Nations 26 , 89–110. (FAO). (2008b). Aquastat, [Online]. Available: http:// Kremen, C., Daily, G. C., Klein, A., et al . (2008). Inadequate www.fao.org/nr/water/aquastat/main/index.stm . assessment of the ecosystem service rationale for con- Food and Agriculture Organization of the United Nations/ servation: Reply to Ghazoul. Conservation Biology, 22 , International Institute for Applied Systems Analysis/ 795–8. International Soil Reference and Information Centre/ Lesschen, J., Stoorvogel, J., Smaling, E., et al . (2007). A spa- Institute of Soil Science—Chinese Academy of Sciences/ tially explicit methodology to quantify soil nutrient Joint Research Centre of the European Commission balances and their uncertainties at the national level. (FAO/IIASA/ISRIC/ISS-CAS/JRC). (2008). Harmonized Nutrient Cycling in Agroecosystems, 78 , 111–31. world soil database (version 1.1), Food and Agriculture Lubowski, R. N., Plantinga, A. J., and Stavins, R. N. (2006). Organization of the United Nations (FAO) and Land-use change and carbon sinks: Econometric esti- International Institute for Applied Systems Analysis mation of the carbon sequestration supply function. (IIASA), Rome. Journal of Environmental Economics and Management, 51, Gillig, D., McCarl, B. A., and Sands, R. D. (2004). Integrating 135–52. agricultural and forestry GHG mitigation response into Luckert, M. K., Wilson, J., Adamowicz, V., et al . (2000). general economy frameworks: Developing a family of Household resource allocations in response to risks and response functions. Mitigation and Adaptation Strategies returns in a communal area of western Zimbabwe. for Global Change, 9 , 241–59. Ecological Economics, 33 , 383–94. Holden, S. T. (2005). Bioeconomic modelling for natural Luedeling, E., Zhang, M., and Girvetz, E. H. (2009). resource management impact assessment. In B. Shiferaw, Climatic changes lead to declining winter chill for fruit H. A. Freeman, and S. M. Swinton, Eds., Natural Resource and nut trees in California during 1950–2099. PLoS ONE, Management in Agriculture: Methods for Assessing 4 , e6166. 166 PROVISIONING AND REGULATORY ECOSYSTEM SERVICE VALUES IN AGRICULTURE

Mendelsohn, R., Nordhaus, W. D., and Shaw, D. (1994). Schlenker, W., Hanemann, W. M., and Fisher, A. C. (2005). The impact of global warming on agriculture: A Ricardian Will US agriculture really benef t from global warming? analysis. American Economic Review, 84 , 7 5 3 – 7 1 . Accounting for irrigation in the hedonic approach. Mendelsohn, R., and Dinar, A. (2009). Land use and cli- American Economic Review, 95 , 395–406. mate change interactions. Annual Review of Resource Shiferaw, H., Freeman, H. A., and Navrud, S. (2005). Economics, 1 , 309–32. Valuation methods and approaches for assessing natural Cassman, K. G., Wood S. (2005). Cultivated Systems. In resource management impacts. In B. Shiferaw, Millennium Ecosystem Assessment: Global Ecosystem H. A. Freeman, and S. M. Swinton, Eds., Natural resource Assessment Report on Conditions and Trends. Island Press, management in agriculture: methods for assessing economic Washington, DC. and environmental impacts . CABI Publishing, Cambridge. Morris, M., Kelly, V., Kopicki, R., et al . (2007). Fertilizer use Steduto, P., Hsiao, T. C., Raes, D., et al . (2009). AquaCrop— in African agriculture: lessons learned and good practice the FAO crop model to simulate yield response to water: guidelines . World Bank, Washington, DC . I. Concepts and underlying principles. Agronomy Journal, Naidoo, R., and Iwamura, T. (2007). Global-scale mapping 101 , 426–37. of economic benef ts from agricultural lands: Implications Steinfeld, H., Gerber, P., Wassenaar, T., et al . (2006). for conservation priorities. Biological Conservation, 140 , Livestock’s long shadow — environmental issues and options . 40–9. Food and Agriculture Organization of the United Naidoo, R., and Ricketts, T. H. (2006). Mapping the eco- Nations (FAO), Rome. nomic costs and benef ts of conservation. PLoS Biology, Swinton, S. M. (2005). Assessing economic impacts of nat- 4, 2153–64. ural resource management using economic surplus. In Nelson, G. C., Rosegrant, M. W., Koo, J., et al . (2009). B. Shiferaw, H. A. Freeman, and S. M. Swinton, Eds., Climate change. impact on agriculture and costs of adapta- Natural resource management in agriculture: methods for tion. International Food Policy Research Institute assessing economic and environmental impacts . CABI (IFPRI), Washington, DC. Publishing, Cambridge. Neue, H. (1993). Methane emission from rice f elds. Swinton, S. M., Lupi, F., Robertson, G. P., et al . (2007). Bioscience, 43 , 466–74. Ecosystem services and agriculture: Cultivating Parry, M., Rosenzweig, C., Iglesias, A., et al . (1999). Climate agricultural ecosystems for diverse benef ts. Ecological change and world food security: a new assessment. Economics, 64 , 245–52. Global Environmental Change, 9 , S51–S67. Tilman, D., Cassman, K. G., Matson, P. A., et al . (2002). Pender, J. (2005). Econometric methods for measuring Agricultural sustainability and intensive production natural resource management impacts: Theoretical practices. Nature, 418 , 671–7. issues and illustrations from Uganda. In: B. Shiferaw, United States Department of Agriculture-Economic H. A. Freeman, and S.M. Swinton, Eds., Natural resource Research Service (USDA-ERS). (2009). Commodity Costs management in agriculture: methods for assessing economic and Returns. Accessed at http://www.ers.usda.gov/ and environmental impacts. CABI Publishing, Data/CostsAndReturns/ . Cambridge. United States Department of Agriculture-Natural Perry, C. J., and Narayanamurthy, S. G. (1998). Farmer Resources Conservation Service (USDA-NRCS). (2001). response to rationed and uncertain irrigation supplies . National SSURGO (Soil Survey Geographic) Database . International Water Management Institue, Colombo. Vera-Diaz, M. D. C., Kaufmann, R. K., Nepstad, D. C., Polasky, S., Nelson, E., Camm, J., et al . (2008). Where to put et al. (2008). An interdisciplinary model of soybean things? Spatial land management to sustain biodiver- yield in the Amazon Basin: The climatic, edaphic, sity and economic returns. Biological Conservation, 141 , and economic determinants. Ecological Economics , 65, 1505–24. 420–31. Rosegrant, M. W., Paisner, M. S., Meijer, S., and Witcover, Wani, S. P., Singh, P., Dwivedi, R. S., et al . (2005). Biophysical J. (2001). Global food projections to 2020: emerging trends indicators of agro-ecosystem services and methods for and alternative futures . International Food Policy Research monitoring the impacts of NRM technologies at differ- Institute, Washington, DC. ent scales. In B. Shiferaw, H. A. Freeman, and S. M. Rosegrant, M. W., Ringler, C., Msangi, S., et al . (2005). Swinton, eds., Natural resource management in agriculture: International Model for Policy Analysis of Agricultural methods for assessing economic and environmental impacts . Commodities and Trade (IMPACT-WATER): Model Des- CABI Publishing, Cambridge. cription . International Food Policy Research Institute, Wood, S., Ehui, S., Alder, J., et al. (2005). Food. In Ecosystems Washington, DC. and human well-being , v o l . 1 : Current state and trends: LIMITATIONS AND NEXT STEPS 167

Millennium Ecosystem Assessment . Island Press, Washington, ment of agricultural conservation policies. American DC. Journal of Agricultural Economics, 86 , 26–41. Wossink, A., and Swinton, S. M. (2007). Jointness in pro- You, L., and Wood, S. (2006). An entropy approach to duction and farmers’ willingness to supply non-mar- spatial disaggregation of agricultural production. keted ecosystem services. Ecological Economics, 64 , Agricultural Systems, 90 , 329–47. 297–304. Zilberman, D., Lipper, L., and McCarthy, N. (2008). When Wu, J. J., Adams, R. M., Kling, C. L., et al. (2004). From could payments for environmental services benef t the microlevel decisions to landscape changes: An assess- poor? Environment and Development Economics, 13 , 255–78. CHAPTER 10 Crop pollination services

Eric Lonsdorf, Taylor Ricketts, Claire Kremen, Rachel Winfree, Sarah Greenleaf, and Neal Williams

Maintaining pollinator habitats in agricultural land- 10.1 Introduction scapes, therefore, can help ensure food production, 10.1.1 Importance to agriculture and policy quality, and security. While other pollinators (e.g., bats and moths) also pollinate crops, bees are the Crop pollination by bees and other animals is an most important crop pollinators (Free 1993) and are ecosystem service of enormous economic value thus the focus of our models. (Losey andVaughan 2006; Allsopp et al . 2008). The pollination models aim to quantify and map Pollination can increase the yield, quality, and sta- scores for relative pollinator abundance across an bility of crops as diverse as almond, cacao, canola, entire landscape, including farms requiring pollina- coffee, sunf ower, tomato and watermelon. Indeed, tion. The models use these results to indicate areas Klein et al. (in press) found that 75% of globally supplying pollinators that increase crop yields. important crops benef t from animal pollination. Alternatively, intermediate results can be integrated The value of this service, while diff cult to quantify into our agricultural model (see Chapter 9 ) to esti- properly, has been estimated several times over the mate the economic value of pollination services as past decade (Southwick and Southwick 1992; an input to crop yields in a more sophisticated man- Costanza et al. 1997; Losey and Vaughan 2006; ner. Either way, these models can inform agricultural Allsopp et al . 2008) with a recent estimate of 195 bil- and land management policies in several ways. First, lion Euros (~$200 billion) worldwide ( Box 10.1 ). land-use planners could predict consequences of While much research and policy attention has different policies on pollination services and income focused on managed bees (especially the honey bee, to farmers (Priess et al. 2007). Second, farmers could Apis mellifera), wild bees and other insect species use these tools to locate crops in places where their also contribute importantly to crop pollination. Our pollination needs are most likely to be met. Third, models focus on wild bees, because the pollination conservation organizations that guide land manage- they deliver represents an ecosystem service from ment and restoration could use the tool to optimize natural systems. In fact, for some crops (e.g., blue- conservation investments for both biodiversity and berry), wild bees are more eff cient and effective crop productivity. Finally, governments or others pollinators than honey bees (Cane 1997). Diverse bee proposing payment schemes for ecosystem services communities potentially provide more stable polli- could incorporate the results into plans for who nation services over time, compared to single (man- should pay whom, and how much. aged) species (Greenleaf and Kremen 2006; Hoehn et al. . 2008). Finally, if alarming regional declines in 10.1.2 Scientif c foundations and context honeybee populations continue (National Research Council of the National Academies 2006; Stokstad Our pollination models are founded on an increas- 2007; Klein et al ., in press), wild pollinators may ing number of studies that have investigated the become increasingly important to farmers. impacts of landscape structure and habitat-quality

168 INTRODUCTION 169

Box 10.1 Assessing the monetary value of global crop pollination services

Nicola Gallai, Bernard E. Vaissière, calculated based upon the f ve dependency levels def ned Simon G. Potts, and Jean-Michel Salles in the Appendix 2 of Klein et al. (2007). For each crop, we calculated an average dependence ratio based on the Most major crop species are pollinated by bees or other reported range of dependence on animal-mediated insect groups (Klein et al . 2007). As the abundance and pollination. Based on this, the 2005 worldwide economic diversity of bees are now declining in many parts of the value of pollinators was US$190 billion compared to world, there is a growing need for improved methods to: US$2,013 billion for the overall crop production value (1) adequately assess the potential loss in terms of (Gallai et al. 2009). economic value that may result from pollination shortfalls, Vulnerability is a function of three elements: exposure, and (2) link this value to the vulnerability of agriculture sensitivity and adaptive capacity. For crops, the agricultural confronted with pollination shortages. vulnerability to pollinator decline depends upon the crop To evaluate the monetary value of crop pollination dependency on pollinators and the capacity of farmers to services worldwide, we used the FAO global crop adapt to pollinator decline. In this context, we used the production statistics (http://www.fao.org) coupled with the ratio of the economic value of pollinators (EVP ) to the total reported degree of dependency of each crop on biotic economic crop production value ( EV ) to calculate a level of pollination (Klein et al. 2007). FAO statistics are available vulnerability, which provides a measure of the potential for direct crops (production data available) and commodity relative production loss attributable solely to the lack of crops (individual crop production data is aggregated for insect pollination. We evaluated the vulnerability in term of each commodity). Although these aggregations of crop the proportion of the agricultural production value that production may represent a signif cant part of the depends on insect pollination (Gallai et al. 2009): agricultural output of a given country or region, and some of these species depend heavily on biotic pollination (Klein et al. 2007), we excluded all aggregated crop complexes IX EVP PQDix ix i VR∑∑ix==11 % from our analysis as prices and production f gures were not ==IX (10.A.2) EV PQ available for each individual crop. ∑∑ix==1 1 ix ix Following Gallai et al. (2009), we def ned the economic value of pollinators (EVP) as the value of the pollinator The vulnerability ratio of global agricultural production contribution to the total economic crop production value. used for human food in 2005 was 9.5% (Gallai et al . This contribution was calculated based upon the 2009). The ratio varied considerably among different dependency ratio of crop production on pollinators, def ned geographical areas, for example, at a national level, the as the proportion of the yield attributable to insect vulnerability of European countries varied between 1% in pollinators. The economic value of pollinators was thus Ireland to 19.5% in Austria (Figure 10.A.1). In Europe, calculated as there was a positive correlation between the vulnerability IX to pollinators of a crop category and its value per EVP = P Q D , ∑∑ ix ix i (10.A.1) production unit (r = 0.729, n = 10, P = 0.017), indicating ix==11 that the more a crop is dependent on insect pollination, the where P is the producer price per production unit, Q is higher its value per production unit. the quantity produced for each crop i Î [1; I ] and for each However, our approach provides an incomplete picture country x Î [1; X ], and D is the dependence ratio for each of the value of insect pollinators to society because we did crop i Î [1; I ]. For Qix we used 2005 FAO production data not take into account agricultural production not used expressed in metric tons. Producer prices, P ix , for 2005 were directly for human food (e.g., fodder crops), seeds produced obtained using data from f nancial markets, USDA (http:// for plant breeding, and perhaps most importantly, natural www.fas.usda.gov) and Eurostat (http://epp.eurostat. vegetation and all its associated ecosystem services which ec.europa.eu), and actualization of FAO data and expressed would almost certainly be strongly impacted by pollinator in US$per metric ton. The dependence ratios D i were decline. Our estimates are therefore conservative. continues 170 CROP POLLINATION SERVICES

Box 10.1 continued

Distribution of the vulnerability ratio across Europe

Vulnerability ratio (%):

1.0–5.7 5.7–8.0 8.0–9.4 9.4–12.1 12.1–19.5

Figure 10.A.1 Distribution of crop vulnerability to pollination service across Europe.

on pollinator populations (reviewed by Kremen strongly by proximity to nesting habitats (Morandin et al . 2007). They have found that the availability of and Winston 2006). nesting substrates (e.g., suitable soils, tree cavities; For example, Ricketts and colleagues (Ricketts Potts et al. 2005) as well as f oral resources (i.e., both 2004; Ricketts et al. 2004) found that bee diversity, nectar and pollen) in both natural and semi-natural visitation rate, pollen deposition rate, and fruit set habitats can strongly inf uence the diversity (Hines are all signif cantly greater in coffee f elds near and Hendrix 2005), abundance (Williams and forest than in f elds further away. On the other Kremen 2007), and distributions of pollinators hand, other studies have found little effect of across a landscape (Tepedino and Stanton 1981; landscape pattern on pollinator visitation, such as Potts et al. 2003). In addition, because bees forage Winfree et al. ’s (2008) study of pollination services from f xed nest sites with limited foraging ranges, to vegetable crops in the northeastern USA. their abundance and diversity on a farm, as well as Despite this variation among studies, Ricketts their effect on crop pollination, can be inf uenced et al. (2008) synthesize 23 case studies (including INTRODUCTION 171 many of those cited above) and f nd a general 10.1.3 Model intuition and difference “consensus” decline in pollination services with between tiers increasing isolation from natural or semi-natural 10.1.3.1 Overview of data requirements habitat. Pollinators require two basic types of resources to Building from these and many other studies, persist on a landscape: nesting substrates and f oral Kremen et al. (2007) have proposed a general frame- resources (Westrich 1996; Kremen et al. 2007). The work for understanding how pollination services model therefore requires estimates of availability of are delivered across landscapes, and how these both of these resource types for each land-use and services are affected by land-use change in agricul- land cover type (LULC) in the map. These data can tural regions (Figure 10.1 ). Here, we develop a sim- be derived from quantitative f eld estimates or from plif ed version of this general model (indicated in expert opinion. Pollinators move between nesting Figure 10.1 ), which uses simple landscape indices, habitats and foraging habitats (Westrich 1996; governed by a few key parameters that can be esti- Williams and Kremen 2007), and their foraging dis- mated from f eld data or expert opinion, to predict tances, in combination with arrangement of differ- relative pollinator abundances across a landscape ent habitats, affects their persistence, their and agricultural f elds. Moreover, we use the frame- abundance, and the level of service they deliver to work that predicts abundance at crop f eld to farms. Our model therefore also requires a typical attribute the pollinator-dependent gains in yield foraging distance for pollinators. These data can be and crop value to the parcels supplying the supplied, e.g., from quantitative f eld estimates pollinators.

Biotic and abiotic factors 4a

Economics & Policy 3b 4b Pollinators Pollinator of target 2b community 8 plant Local Site 2a c Land Use & 3a Management Pollination 1 6b Landscape 6a service Structure b value f Disturbance 3c a 2c 5b 2d Plant Target community 3d plant d Geographic context 5a Biotic and abiotic factors e

Figure 10.1 General conceptual model describing pollination services and their delivery across an agricultural landscape (full framework, reproduced from Kremen et al. (2007)). Land-use practices (Box a) determine the pattern of habitats and management on the landscape (Box b). The quality and arrangement of these habitats affect both pollinator and plant communities (Boxes c and d). The value of pollination services (Box f) depends on the interaction between specif c plants (e.g., crops) and their specif c pollinators. Our pollination model is a simplif ed version of this full model, capturing the following arrows only: 3a, 3c, 6a, and with economic model 6b. 172 CROP POLLINATION SERVICES

(Knight et al. 2005), from proxies such as body size parcel as the analytical spatial unit on the land- (Gathmann and Tscharntke 2002; Greenleaf et al. scape. In our case, it is a 90 meter by 90 meter grid 2007), or from expert opinion. cell that may have more than one land cover The ultimate level of pollination service pro- within it. In calculating f oral resources, nearby vided to a farm depends on the crops grown, the parcels are given more weight than distant par- ability of each modeled species to pollinate them cels, based on the species’ average foraging range. effectively, the crop’s response to animal pollina- The result is a map of relative abundance scores tion and the abundance of pollinators at the crop. (0–1) for each species in the model (the “supply The model therefore incorporates data on location map”). of farms of interest, the crops grown there, and Given this pattern of pollinator supply, the model how effective each species is as a pollinator for a then estimates the relative abundance of foraging given crop. bees arriving at each farm (“farm abundance”). It sums the relative bee abundances in neighboring parcels, again giving more weight to nearby parcels, 10.1.4 Model intuition based on average foraging ranges. This weighted Using these data, the model f rst estimates a rela- sum is our relative index (0–1) of abundance for each tive abundance score of each pollinator species in pollinator in the farm. If the crop type at each parcel each parcel (hereafter, pollinator “supply” to fol- and its pollinators are known, the model will limit low the conventions in Chapter 3 ), based on the the weighted sum only to relevant pollinators. available nesting resources in that parcel and the We use a very simple yield function to translate f oral resources in surrounding parcels. We def ne farm abundance into relative crop yields. Alternatively,

Table 10.1 Comparison of model complexity and parameters used in tier 1 versus tier 2

Parameter Description Tier 1 Tier 2

HN Habitat suitability for nesting x x HF Habitat suitability for foraging x x J Number of land cover types (each indexed by j ) x x

Nj Compatibility of habitat j for nesting x x

Fj Compatibility of habitat j for foraging x x M Number of parcels in landscape (indexed by m ) x x D Distance between parcels x x α Expected pollinator foraging distance x x P Pollinator abundance score x x O Number of farm parcels (indexed by o ) x x X Pollinator source parcel x x ψ Farms’ average change in normalized scores (used for sensitivity analysis) x x Y Crop yield (indexed by o ) x x V Crop value based on agricultural production function (indexed by o ) x x ν Proportion of a crop’s yield attributed only to wild pollination (indexed by c ) x x Κ Pollinator abundance to achieve ½ of pollinator-dependent yield (indexed by c ) x x PS Pollinator service provided to crops (indexed by m ) x x S Number of species (indexed by s ) x I Number of nesting types (indexed by i ) x W Weight describing importance of f oral season for pollinator x K Number of f oral seasons (indexed by k ) x C Crops’ pollinator requirement x ε Relative abundance of pollinator in landscape x

The last seven parameters are unique to tier 2. The model can be run with a mixture of tier 1 and tier 2 parameters, allowing a continuum of model complexity to match data availability. TIER 1 SUPPLY MODEL 173

one can use farm abundance as an input in the more w h e r e N j Î [0,1] represents compatibility of LULC sophisticated agricultural production model (Chapter j f o r n e s t i n g a n d p jx is the proportion of parcel x 9 ) to determine the crop yield and value on each farm that is covered by LULC j . This provides a land- parcel. Finally, our model redistributes crop value scape map of nesting suitability where HN x Î [0,1] back onto the landscape to estimate the service value ( Figure 10.2b ; Plate 5b). A score of 1 would indicate provided from each parcel to surrounding farms that the entire area of the parcel provides habitat (equivalent to the “use” and “value” in the parlance suitable for nesting (e.g., forest habitat in of this book, Chapter 3 ). It does so using the same Supplemental Online Material (SOM) Table 10.S1 ) foraging ranges, so that parcels that are sources of while a score of 0.2 would indicate 20% of the par- abundant pollinators and near to farms tend to have cel’s area provides suitable nesting habitat (e.g., relatively high service value. coffee/pasture habitat). The pollination model has two tiers permitting We calculate the proportion of suitable foraging use of different amounts of information. The tier 1 habitat surrounding a parcel x, g i v e n b y HF x Î [0,1]. models are nested within tier 2 (i.e., tier 1 is a sim- We assume that foraging frequency in parcel m pler version of tier 2). Both tiers are based on a declines exponentially with distance (Cresswell LULC map, showing both natural and managed et al. 2000), and that pollinators forage in all direc- land types. Onto this landscape, tier 1 models a sin- tions with equal probability. Therefore, parcels far- gle pollinator that represents the overall pollinator ther away from nest parcel x contribute less to total community, while tier 2 considers multiple pollina- resource availability than parcels nearby, and leads tor species or guilds, allowing them to differ in to the following prediction for the potential f oral f ight season, resource requirements, and foraging resources available to pollinators nesting in parcel distance. Because the models are nested, tier 1 and x , HF x: tier 2 are actually endpoints of a continuum of potential model complexity ( Table 10.1 ). For exam- −Dmx MJ Fp e a ple, one can recognize multiple species and use a ∑∑mj==11jjm HF = (10.2) x −Dmx different foraging radius for each (Eq. (10.2)), but M e a model only a single f owering season and nesting ∑ m=1 guild. This nestedness also allows easy comparisons of model outputs between tiers, as we explore w h e r e p jm is the proportion of parcels m i n L U L C j , in Chapter 15 . D mx is the Euclidean distance between parcels m and x , α is the expected foraging distance for the pollinator (Greenleaf et al. 2 0 0 7 ) a n d F Î [0,1] rep- 10.2 Tier 1 supply model j resents relative amount of foraging resource in Pollinators require habitat for nesting and within LULC j . The numerator is the distance-weighted some foraging distance they require f oral resources resource summed across all M parcels. The for food. The f rst step of our model is to translate a denominator represents the maximum possible LULC map (Figure 10.2a ; Plate 5a) into a nesting amount of forage in the landscape. This equation suitability map and a f oral resource availability generates a distance-weighted proportion of habi- map. Then based on the amount and location of tat providing f oral resources within foraging nesting and f oral resources, we calculate the polli- range, normalized by the total forage available nator supply map. within that range (Winfree et al. 2005) ( Figure The f rst step in calculating the pollinator source 10.2c ; Plate 5c). score at each parcel is identifying the proportion of Supply map : Since pollinator abundance is limited suitable pollinator nesting habitat in a parcel x as a by both nesting and f oral resources, the pollinator function of LULC j , HNx : abundance score on parcel x is simply the product of foraging and nesting such that P = HF HN Î J x x x [0,1].This score represents the location and supply HNxjjx= ∑ N p , (10.1) j=1 of pollinators available for crop pollination from 174 CROP POLLINATION SERVICES

(a) Land Cover (b) Nest Suitability

Ditchside edge High: 1 Farm parcel edge Residential edge Low: 0 Roadside edge Agriculture Organic Ag Pasture Riparian Scrub Forest Unclassified Water

High: 1 High: 1

Low: 0 Low: 0

(e) Farm Abundance (f) Service Value Map

High: 0.25 High: 1

Low: 0 Low: 0

W E Kilometers 0 4.5 9 18 27 36 S

Figure 10.2 Example results of pollination model for watermelon in Yolo County, California. The model uses (a) land cover data as input and derives maps of (b) nesting habitat and (c) f oral resources. From this, it generates (d) a pollinator supply map that describes an index of pollinator abundance on the landscape. Based on the supply map, the model generates (e) a pollinator abundance map on farm parcels (i.e., “farm abundance”). After using a simple yield function to translate farm abundance into relative yield the model distributes yield or economic value back onto the surrounding landscape to generate (f) the value map. All steps are the same for tier 1 and tier 2 models; results here are tier 2, based on data supplied in supplemental online appendix. (See Plate 5.) TIER 1 VALUATION MODEL 175

parcel x and results in the supply map. This map 10.4 Tier 1 valuation model does not account for the location or type of crops present in the landscape, and as such has not Pollination has economic value as an ecosystem adjusted pollination to show the actual service sup- service because it is an input to agriculture, from plied to people, but rather all potential pollination which people derive food and income. In formal on the landscape. terms, pollination can be an important factor in agricultural production functions, which relate yields of a given crop to the quantity and quality of various inputs (e.g., water, soil fertility, labor, chemicals). 10.3 Tier 1 farm abundance map Production functions (or “yield functions”) are a For pollinators’ actions to provide crop pollination well-established econometric technique used widely benef ts to people, pollination must take place on a in agriculture and product manufacturing (Polasky farm growing a crop that requires insect pollina- et al. 2008). The agricultural models described in tion. In the next modeling step, we identify farms Chapter 9 take exactly this approach, so we do not on the landscape, and the relative abundance of repeat it here; instead, we offer an extremely simple wild pollinators on ech farm. alternative. Our “farm abundance” results above Pollinators leave their nesting sites to forage in can be used as inputs to either one. surrounding parcels, so farms surrounded by a Using production functions in this way will result higher abundance of nesting pollinators should in an estimate of the economic value of pollinators experience higher abundances of pollinating visi- at each farm. It is most likely of interest, however, to tors to their crops. We use the foraging framework estimate the value of the habitats in the landscape described in Eq. (10.2) to determine the contribution that support these pollinators. For this we can use to pollinator abundance from a single nest parcel m the ecological models described here, which model to forage on a crop in farm o : movement of pollinators from source parcels to farms, to attribute economic value realized on farms back to the pollinator-supporting habitats.

−Dom Pe a P = m , (10.3a) om −Dom M 10.4.1 Estimating crop yield and value e a ∑ m=1 The calculated pollinator abundance from Eq. (10.3b) will be an input into the agricultural produc-

where Pm is the relative supply of pollinators on tion function to determine the crop yield and crop map unit m , Dom is distance between source parcel m value on each parcel. In lieu of a more detailed agri- and farm o, and α is species’ average foraging dis- cultural production model ( Chapter 9 ), we use a tance. O can be used to index specif c farms of inter- simple saturating yield function to translate the est or every agricultural parcel on the landscape. abundance of pollinators on farms into an expected The numerator of Eq. (10.3a) represents the dis- yield. Yield should increase as pollinator abundance tance-weighted proportion of the pollinators sup- and diversity increase (Greenleaf and Kremen 2006), plied by parcel m that forage within farm o and the but crops vary in their dependence on pollinators, numerator is a scalar that normalizes this contribu- i.e., some crop species are self-compatible and yield tion by the total area within foraging distance to is less dependent on pollination while other species farm (Winfree et al . 2005). The total pollinator abun- obligately require pollination to generate any yield dance on farm o , P o , is simply the sum over all M (Allsopp et al. 2 0 0 8 ; R i c k e t t s et al. 2 0 0 8 ) . W e a c c o u n t parcels, for both observations, and thus calculate the

expected yield of a crop c on farm o , Y o , as M P Y = 1,− nn+ o PPoom= ∑ . (10.3b) occ (10.4) m=1 Poc+k 176 CROP POLLINATION SERVICES

where ν c represents the proportion of total crop c ’s or species-specif c information to be incorporated. yield attributed only to wild pollination (e.g., ν c While we refer to species throughout, these same would be equal to 1 if a crop is an obligately out- models could also be applied to guilds. Specif cally, crossing species and equal to 0 if the crop species we model multiple pollinators, incorporate multi- were wind-pollinated). .In the denominator of the ple nesting types per habitat and allow for multiple third term, κ c is a half-saturation constant and repre- seasons of foraging. sents the abundance of pollinators required to reach As in tier 1, the f rst step in calculating the polli- 50% of pollinator-dependent yield. The monetary nator score is identifying compatible nesting habitat value of the crop on farm o, Vo , is simply the prod- for each pollinator species across the landscape, uct of yield per hectare, Yo , the number of hectares given by HN sx Î [0,1]. In tier 2, we account for spe- of the crop and the price of the crop (Gallai et al. cies (or guild) differences in habitat suitability so 2009). that the proportion of suitable nesting habitat in a parcel x for pollinator species s as a function LULC

j , HNsx is 10.4.2 Assigning value back to pollination sources: service value J HNsx= ∑ N js p jx , (10.6) We use the pollinator model here to redistribute j=1 each farm o ’s value back onto the landscape based where N Î [0,1] represents compatibility of LULC j on the actual level of service supplied by each par- js for nesting by species s . cel m . Recall Eqs. (10.3a) and (10.3b) that determined Some LULC classes can provide habitat suitable the total abundance on farm o by summing across for multiple nesting types. For example, in all M supply parcels the proportion of pollinators California, we scored oak woodland habitat as pro- foraging from each supply parcel to farm o . Here, viding good habitat for wood-nesting, ground-nest- we instead attribute the pollinator-generated value ing and cavity-nesting bees, but scored agricultural from the O farms back to the M supply parcels. For habitat as providing poorer habitat for ground- each supply parcel m , we sum across all O parcels, nesting bees, and non-habitat for wood or cavity weighting the contribution from each farm o to nesters (see Section 10.4.1 ). parcel m by their proximity. Thus, supply parcels For bee species or groups that span nesting types close by crops should provide a greater service than (e.g., species that nest in the ground and in hollow parcels far from any crops. Formally, we calculate stems) we assigned the habitat type according to pollinator service provided to O f a r m s f r o m e a c h m the nest type that maximized its suitability for that parcel, PS , as m bee species. In other words, if there are I nesting O Pom types, then N js = max[ NS si N ji , . . . , NS sI N jI], where NS si PSmc= n ∑ V o , (10.5) o=1 Po is the nesting suitability of nesting type i for species

s and N ji is the suitability of LULC j for nesting type w h e r e Vo represents the crop value in farm o . This i . This analysis provides a map of nesting suitability score generates the pollinator service map (Figure ( Figure 10.2b ; Plate 5B). 10.2f ; Plate 5F) and represents the location and As in Tier 1, we calculate the proportion of suit- value based on supply of pollinators that provide able foraging habitat for pollinator species s n e s t - crop pollination to surrounding farms (i.e., ing in parcel x g i v e n b y HF sx Î [0,1]. In tier 2, equivalent to “value” results in the parlance of though, we allow for production of f oral resources Chapter 3 ). to vary among K seasons. We also use data or expert opinion to assess f ight period and account 10.5 Tier 2 supply model for variation among pollinators in their K f ight seasons, e.g., some are present in summer only, The tier 2 model follows the same logic as tier 1, but while others are present in multiple seasons. We each step allows for more detailed, season-specif c calculate the overall f oral resources available as a TIER 2 VALUATION MODEL 177

weighted sum across K seasons where the weight pollinator abundance of species s o n f a r m o , P os , i s s i m -

( wsk ) Î [0,1] represents the relative importance of ply the sum of Posm o v e r a l l M parcels at each farm o , f oral production in season k f o r s p e c i e s s. W e c o n - M K strain each w v a l u e s u c h t h a t w = 1. This PPos= osm . (10.8b) sk ∑ k =1 sk ∑ leads to the following prediction for the potential m=1 f oral resources available to species s p a r c e l x This score represents the relative abundance of pol- across K s e a s o n s , HF xs , linators visiting farm and results in the farm abundance map ( Figure 10.2e ; Plate 5E). M J −Dmx Fpeas To calculate the total pollinator score for farm o from K ∑∑ js, k jm mj==11 all pollinators, P , we calculate the normalized pollina- HF= w , (10.7) o sx∑ sk M −Dmx k=1 tor score for all pollinator guilds or species, such that ∑ e as m=1 S CP ∑ s=1 sos (10.9a) where p jm is the proportion of parcel m in LULC j , Po = S , Cs D mx is the Euclidean distance between parcels m and ∑ s=1

x , α s is the typical foraging distance for species s and

Fjs, k Î [0,1] represents suitability for foraging of where C s Î [0,1] if the crop requires pollinator s and

LULC j f o r s p e c i e s s d u r i n g s e a s o n k . The use of F js, k 0 otherwise. This unweighted summation assumes permits attributing different resource levels to the that all pollinators are equally abundant. However, same LULC type for different bee species or guilds if some pollinators have higher background abun- by season—for example, in California, riparian hab- dance than others, then a weighted average may be itat produces important early spring resources but more appropriate such that many pollinator species are not yet f ying at this time. By contrast, riparian habitat produces almost S e CP no summer resources. Using the normalized pro- ∑ s=1 ssos (10.9b) Po = S , portion controls for differences among pollinators C ∑ s=1 s that vary in their foraging radii, and allows us to estimate total pollinator abundances in subsequent w h e r e εs represents the abundance of pollinator s model steps. in the landscape, relative to other pollinator species As in tier 1, we calculate a supply score for each or guilds. The weights for each species can be species on parcel x as the product of foraging and determined by expert opinion or with observed nesting: P sx = HF sx HN sx ( Figure 10.2d ; Plate 5D). data. Additional species-by-crop weights could be

added to Eq. (10.9b) in the same fashion as ε s to 10.6 Tier 2 farm abundance map account for differences among pollinators in their effectiveness on a given crop (Greenleaf and To calculate the abundance of each pollinator spe- Kremen 2006). cies on a crop in parcel o , we use the framework described in Eqs. (10.3a) and (10.3b). First to calcu- 10.7 Tier 2 valuation model late pollinator visitation by species s from nest parcel m to farm parcel o , P osm , As in tier 1, the calculated tier 2 abundance from Eq. (10.9b) will be an input into a simplif ed agricultural −D om production model to determine the crop yield and Pe s P = sm , (10.8a) osm −Dom crop value on each farm parcel. The description in M e s Section 10.4.1 and Eq. (10.4) are the same for tier 2. ∑ m=1 It follows we again use the ecological model to

w h e r e P sm represents the supply of pollinator s on map redistribute the value from all O farms onto each unit m , D om is distance between map unit m and farm supply parcel m for each species s . The resulting

o a n d α s i s s p e c i e s s ’ typical foraging distance. The total score represents the available supply weighted by 178 CROP POLLINATION SERVICES the relevant demand, each species’ relative abun- tropical/premontane moist forests (Janzen 1983). dance in the landscape, effectiveness (ε s ) and the Studies were conducted on 12 sites in a large coffee crop value within foraging distance. Thus we calcu- farm (approx. 1100 ha) in the center of this late pollinator service value from parcel m to other landscape.

O parcels, PS m , as High-resolution (1 m) aerial photos, supplied SO by CATIE (Centro Agronómico Tropical de Investi- Poms PSmc= ne sso C V . (10.10) gación y Enseñanza), were classif ed into six ∑∑ P so==11 o major classes of LULC and resampled to 30m spa- This score generates the tier 2 pollinator service tial resolution. These classes were then assigned value map ( Figure 10.2f ; Plate 5F) and represents values of nesting and f oral resources (assuming a the location and pollination service value based on single f owering season) based on expert opinion relative abundance of pollinators that provide crop (see SOM Table 10.S1 )), informed by f eld work in pollination from parcel m to farm o . the area (Ricketts 2004; Brosi et al. 2008). The most common visitors to coffee in this region are 11 10.8 Sensitivity analysis and model species of native stingless bees (Meliponini ) and validation the introduced, feral honey bee, Apis mellifera . For the model, these 12 species were assigned to two We f rst compare model predictions against f eld nesting guilds based on expert opinion (SOM data in two contrasting landscapes in California, Table 10.S2 ). All 11 species were observed during USA, and San Isidro, Costa Rica. We then illustrate the period of study, but sampling did not continue a sensitivity analysis with the Costa Rican data to year-round. Lacking this information on season- determine the extent to which our results depend ality, we assumed a single f ight season for all on the precision and accuracy of our parameter species. To estimate typical foraging ranges for estimates. each species (Table 10.2 ), we used intertegular spans for 10 museum specimens and the statistical relationship presented by Greenleaf et al. 10.8.1 Model validation (2007). To validate our model, we compare its predictions During the f owering seasons of 2001 and 2002, of total (community-wide) abundance against total Ricketts and colleagues (Ricketts 2004) measured observed abundance in farms of crops in two land- bee activity, pollen deposition, and pollen limitation scapes: coffee in Costa Rica and watermelon in in 12 sites, varying from 10 to 1600 m from the near- California. The Californian and Costa Rican exam- est major forest patch. We used these observations ples use different levels of model complexity and to compare against our model. Our models predict differing mixes of f eld- and expert-derived param- at least 80% of the variance in observed pollinator eters. In all cases, model parameters were derived abundance ( Figure 10.3a ). independently of f eld validation data (e.g., esti- The model’s predictions for farm abundance mates of typical foraging ranges (α ), were derived scores were not as strongly related to f eld measure- from bee body size; f oral availability was estimated ments of pollen deposition on coffee stigmas ( Figure through expert assessment based on other studies, 10.3b ), which is a closer correlate to actual pollina- not from f eld measurements taken simultaneously tion service (Ricketts 2004). Modeled abundance with pollinator abundances). scores do not predict pollen limitation of coffee well ( Figure 10.3c ). Pollen limitation is the degree to 10.8.1.1 Costa Rica which coffee production (seed number and mass) is We applied the model to an agricultural landscape reduced due to insuff cient pollination, and is a in the Valle del General, Costa Rica, one of that close measure of actual pollination services. Pollen country’s major agriculture regions. The landscape limitation does decline with increased modeled is dominated by coffee, sugar cane, and cattle pas- service scores, but the f t is weak. Variation in pol- ture, all of which surround hundreds of remnants of len deposition and pollination limitation also SENSITIVITY ANALYSIS AND MODEL VALIDATION 179

(a) (b)

(e)

Figure 10.3 Comparison of predicted and observed pollination scores at two study sites. In each site we compared the model’s predicted abundance to pollinator abundance (a: Costa Rica, b: California), pollen deposition (c: Costa Rica, d: California), and in Costa Rica, we also compared the model to pollen limitation as measured by seed mass (e). depends on pollination eff ciency of each bee spe- 10.8.1.2 California cies, on resource limitation of the coffee plant itself, We applied the model to an agricultural landscape and other factors not captured in a prediction of in the Central Valley of California, across a strong pollinator abundance, which likely contributes to gradient in isolation of farms from large tracts of the poor f t. natural habitats (oak woodland, chaparral scrub 180 CROP POLLINATION SERVICES and riparian deciduous forest). Studies were con- 10.8.2 Sensitivity analysis ducted on watermelon (Kremen et al. 2002b, Kremen Sensitivity analysis should identify the model et al. 2004) across this landscape. parameters that have the greatest inf uence on The LULC data were simpli f ed from a 13-class model results. This allows the scientist to focus on supervised classif cation of Landsat TM data at 30 × improving accuracy and precision of parameters to 30 m resolution (described in detail in Kremen et al. which the model is most sensitive, and allows man- 2004) into six classes. Four additional cover classes agers to determine the major sources of uncertainty were hand drawn on the landscape using ArcGIS to affecting model predictions. In our case, we are account for nesting and f oral resources that come interested in how estimates of nesting suitability, from edges of roads, agricultural parcels, residential f oral resource availability and bee dispersal dis- areas and irrigation ditches (Figure 10.2a ; Plate 5A). tance inf uence our predicted pollinator abundance These classes were then assigned values of nesting scores. If we f nd them to be quite sensitive, then and f oral resources based on expert opinion values further research is required to reduce this uncer- (SOM Table 10.S3 ), informed by studies of bee-plant tainty before the model can be used with networks (Kremen et al. 2002a, Williams and Kremen conf dence. 2007, Kremen et al ., unpublished; Williams and Our model predicts a parcel’s pollinator abun- Kremen, unpublished) and bee- nesting densities dance relative to other parcels on a landscape, so our (Kim et al. 2006) in the same landscape. sensitivity analysis focuses on these relative scores. D u r i n g t h e f owering season of 2001, bee visits We let P ̂ represent a normalized pollinator score on were recorded at 12 sites, and median species- o farm o (from Eq. (10.3) or (10.9)) based on the origi- specif c pollen deposition per visit was estimated nal parameter estimates such that (Kremen et al. 2002b). Each bee species in the study was characterized by its nesting habit based on ˆ PPo − min expert opinion and the length of its f ight period, Po = (10.11) PPmax− min based on over 12 000 bee specimens collected from 1999 to 2004 by pan-trapping and netting at f owers where Pmin and Pmax are the minimum and maximum in this landscape (Kremen and Thorp, unpublished; pollinator service scores for all farms on the land- Williams et al . , u n p u b l i s h e d ) ( S O M T a b l e 1 0 . S 4 ) . ̂ scape. We let P o , c represent the analogous normal- Typical foraging distances were calculated from ized score on farm o resulting from modif ed measurements of intertegular span, using the regres- ̂ parameter combination c, and let ψ c be the average sion in Greenleaf et al. (2007). For nearly all bee spe- change in normalized scores from combination c cies, at least f ve individuals were measured but for such that a few species, only one measurement was used. Data O on Apis mellifera , which are managed for pollination PPˆˆ− ∑ o=1 ooc, (10.12) y# = in this landscape, were removed prior to analysis. c O The model provided a reasonable f t to the observed data on total abundance of native bees on where O is the number of farms in the analysis. watermelon, although with considerable scatter We use regression analysis to determine sensitiv- (Figure 10.3b ). Model predictions were strongly ity, similar to McCarthy et al.’s (1995) logistic regres- related to estimated pollen deposition from native sion approach used in population viability analyses. bees (Figure 10.3d ), a more direct measure of polli- Our goal is to calculate how variation in each nation services that has been used to assess the con- parameter affects estimates of a parcels’ pollinator tributions of wild bees to pollination services abundance, independent of all other parameters in (Kremen et al. 2002b; Kremen et al. 2004; Winfree the model. Given the number of parameters, explor- et al. 2007). However, we caution against interpreting every combination is impractical. Instead, we ing the model’s ability to predict pollen deposition create a sample of parameter combinations by since pollen deposition is calculated from visitation selecting parameter values randomly from a uni- data, not direct observations (Kremen et al. 2002b). form distribution, each within its range of uncer- SENSITIVITY ANALYSIS AND MODEL VALIDATION 181 tainty and then generate an estimated pollinator values and drew a random number from a uniform ̂ score Po , c for each parcel. distribution with this range (Table 10.2 ). For f oral To generate parameter combinations, we set a and nesting resources we set the range as ±0.1 minimum and maximum for the range of parameter around the estimate, and we did not allow the max-

Table 10.2 Results of sensitivity analysis for Costa Rican study

Parameter Estimate Max Min δ SE δ Standardized (Slope) regression coeff cient (t -value)

Forest 1 1 0.9 4.550 1.303 3.493*

) Coffee 0.5 0.6 0.4 7.758 0.666 11.643* F j F Cane 0 0.1 0 0.027 1.356 0.020 Pasture/grass 0.2 0.3 0.1 0.144 0.652 0.221 Scrub 0.3 0.4 0.2 0.553 0.657 0.842 availability ( Forage resource Bare 0.1 0.2 0 0.300 0.663 0.453 Built-up 0.3 0.4 0.2 0.450 0.676 0.666 Forest 1 1 0.9 0.046 1.300 0.035

) j Coffee 0.2 0.3 0.1 0.975 0.648 1.505 N Cane 0 0.1 0 1.950 1.292 1.509

nesting Pasture/grass 0.2 0.3 0.1 0.535 0.664 0.805 Apis

suitability ( Scrub 0.3 0.4 0.2 0.784 0.674 1.162 Bare 0 0.1 0 2.896 1.302 2.224* Built-up 0.2 0.3 0.1 0.739 0.672 1.101 Forest 1 1 0.9 3.381 1.287 2.626*

) Coffee 0.1 0.2 0 0.067 0.650 0.102 N j N Cane 0 0.1 0 0.382 1.318 0.290 Pasture/grass 0.1 0.2 0 0.259 0.683 0.379 Scrub 0.2 0.3 0.1 0.458 0.658 0.696 suitability ( Native nesting Bare 0.1 0.2 0 0.247 0.656 0.376 Built-up 0.1 0.2 0 0.640 0.659 0.972 Apis mellifera 663 776 562 0.001 0.001 1.467

Huge Black 2002** 214 239 191 0.007 0.003 2.376* Melipona fasciata 578 634 525 0.001 0.001 0.653 Nannotrigona mellaria 70 79 61 0.008 0.008 1.037 Partamona cupira/Trigona 87 110 69 0.007 0.003 2.134* fussipennis/Trigona corvina ***

) for each species (m) Plebeia jatiformis 28 30 25 0.027 0.024 1.131 s α Plebia frontalis 34 36 33 0.005 0.051 0.096 Trigona (Tetragona) clavipes 55 63 48 0.004 0.009 0.490 Trigona (tetragonisca) angustula 22 24 20 0.013 0.029 0.453 Trigona dorsalis 60 66 54 0.006 0.011 0.544

Foraging range( Trigona fulviventris 77 82 73 0.046 0.015 3.158* Trigonisca sp. 21 23 20 0.043 0.051 0.829

* p < 0.05. Unidentif ed species. *** These species were indistinguishable during f eld observations and lumped together. The strength of the model’s sensitivity is given by the standardized regression coeff cients in the f nal column. These coeff cients result from a multiple regression of the parameter value combinations on the average change in normalized pollination score, ψ ĉ . 182 CROP POLLINATION SERVICES imum to exceed 1 or the minimum to drop below 0. 10.9 Limitations and next steps For foraging ranges, we set the range using the minimum and maximum of the 10+ measurements of 10.9.1 Limitations intertegular span. Despite the promising results, there are several lim- By iterating this parameter draw process 1000 itations to our model. First, our models estimate the ̂ times, and then regressing the change in scores, ψ c benef ts of wild pollinators to agricultural crop pro- against randomly varying parameters, we can esti- duction, but pollinators contribute to a much mate sensitivity to each parameter while account- broader set of social benef ts that need to be mod- ing for variation in the others. The sensitivity of eled separately ( Box 10.2 ). Second, our models are each predictor variable is indicated by its stand- limited to predicting relative pollinator abundance, ardized regression coeff cient ( t -value), calculated which is only one of many potential contributors to from the best f t of a multiple linear regression crop yield (see Chapter 9 ). Translating from pollina- # model: yddc =+011xx ++... dnn , w h e r e xn a r e p r e - tor abundance to pollinator inf uence on crop yield dictor variables (foraging distance, nesting suita- will be limited in many cases by gaps in our under- bility values, etc) and δ n are the regression standing of pollinator-yield effects. First, we often coeff cients. The standardized regression coeff - do not know the functional form of the relationship cient is the regression coeff cient (slope of a line) between increased number or quality of pollen divided by its standard error (Cross and Beissinger grains deposited and yield, and the functional form 2001). This is a unitless quantity that allows one to may further vary with crop variety as well as water directly compare the sensitivity among parame- and nutrient availability. In addition to the relation- ters, and because our parameter combinations ship between pollinator abundance and the amount were created randomly, also accounts for potential and quality of pollen delivered, pollination is inf u- interactions among model parameters (Cross and enced by pollinator foraging behavior and effective- Beissinger 2001). The standard error for one model ness, across scales from within f ower, inf orescence, ̂ parameter is caused by the dependence of ψ c on patch and landscape (Klein et al . 2007; Kremen et al . other parameters and the signif cance of the slope 2007; Ricketts et al . 2008). is calculated using a two-tailed t -distribution (a The uncertainty in the relationship between the t-value greater than 1.9 or less than -1.9 is signif - model’s output, a relative score, and quantitative cant at p < 0.05). pollinator abundance currently limits the models We illustrate the sensitivity analysis using our application to land-use planning. Without a quanti- Costa Rica data set ( Table 10.2 ). The results indicate f ed relationship between the model score and ̂ that a farm’s normalized pollinator score, Po , is most abundance, it is diff cult to determine the precise sensitive to foraging resources present in coffee yield, crop value and subsequent service value of ( t-value = 11.65; p < 0.05) and forest (t -value = 3.50; supply parcels. And without these precise values, ̂ p < 0.05) habitats. Interestingly, P o is also sensitive to decisions about land management, often based on a uncertainty in a group of species’ foraging dis- cost-benef t analysis, would be diff cult because the tances, which ranged between 77 and 214 m. benef ts are thus uncertain. In other words, the Pollinator service scores were not sensitive to spe- model can determine that one landscape will cies with smaller or greater estimated foraging provide qualitatively more pollinators to a farm, ranges. These sensitivities are likely due to the vari- but it cannot determine if the cost of management ation in forest composition surrounding farms sites or habitat restoration is outweighed by the benef ts. at these moderate scales. The implication for con- Parameterizing the model to facilitate this type of servation is that additional effort to estimate and cost-benef t analysis is an obvious next research manage the f oral resources within coffee farms, priority. and bee-pollinated crops in general, would be of LULC data are often only available at resolutions highest priority for understanding the response of coarser than the scale at which they inf uence polli- pollination to landscape change. nator behavior. Thus, while our model predicts the LIMITATIONS AND NEXT STEPS 183

Box 10.2 Pollination services: beyond agriculture

Berry Brosi to the bulk of other ecosystem services and that is essential to life on Earth. The role of pollination In the middle of the vast Amazon River, a f sherman strains interactions in the functioning of ecosystems is to pull a huge tambaqui f sh ( Colossoma macropomum ) particularly important because it is a limiting factor to into his homemade wooden boat. This scene seems about reproduction in more than two-thirds of plant species as removed from pollination as you can get—but that (Burd 1994). couldn’t be further from the truth. The roles that pollination plays in the production of This chapter has focused its valuation approach on crop non-agricultural ecosystem services are diverse. In terms of pollination, but pollinators are perhaps even more understanding the value of pollination services in this important in providing a huge range of non-agricultural context, a central issue is the ecological distance from pollination services, many of which are quite surprising. For pollination to the service being considered (Figure10.B.1). example, a sizable proportion of the f sh species harvested At one end of the spectrum are ecosystem services that are in the Amazonian freshwater f shery—including the ecologically proximal to pollination (left side of Figure tambaqui —eat fruits that drop into the waters of 10.B.1), such as the pollination of non-agricultural products seasonally f ooded forests and which have evolved to be derived from fruits or seeds. These services are f sh-dispersed (Correa et al. 2007). The bulk of these characterized by: fruit-producing trees rely on animals to pollinate their f owers as an essential step in producing fruit. Thus, • A direct dependence on one or a few discrete pollination pollination disruptions in f ooded forests would have severe events (f oral visits) to provide a tangible product economic and nutritional consequences for the people of • Pollination’s role in the value of the service is large the Amazon and their multi-million dollar f shery. relative to that of other ecological interactions over short Animal-mediated pollination is necessary for the scales of space and time reproduction of the great majority of f owering plant • Relatively low resilience of the value of the service to species globally, providing a service that is foundational pollination shortfalls over short scales of space and time.

greater direct dependence less direct dependence on pollination; on pollination; less resilience to pollination greater resilience to losses over short pollination losses over short spatiotemporal scales spatiotemporal scales

proximalecological distance from pollination distal

Brazil nut mahogany maintenance primary production timber of tropical productivity production biodiversity

almond Amazonian soil nitrogen oxygen production freshwater fishery maintenance production (agricultural) production (through legumes) (fruit-feeding fish)

Figure 10.B.1 Spectrum of ecosystem service reliance on pollination.

At the other end of the spectrum, services that are • The role of pollination for the value of the service in any ecologically distal to pollination have the opposite set of given small scale of space and time is relatively minor characteristics (right side of Figure 10.B.1): • There is relatively high resilience of the service to • Many pollination events, integrated over large scales of pollination losses over short scales of space and time; but if space and time, are needed to support the service pollination losses were to be sustained over larger 184 CROP POLLINATION SERVICES

Box 10.2 continued spatiotemporal scales, these functions and services could family (legumes) are animal-pollinated; this family is critical greatly suffer for its f xation of atmospheric nitrogen to the soil. If legumes were to suffer pollination reductions, even plant Animal-mediated pollination is ultimately derived from the species that are wind-pollinated or self-pollinated would be actions of single pollinators moving between a few plants greatly affected by reductions in available soil nitrogen over over small spatial scales. Thus, the services that are most timescales as short as a few years. proximal to pollination are typically tangible, plant-derived Flowering plants are also central to a host of climate products, while those more distal to pollination are regulation functions (oxygen production, carbon sequestra- produced by the aggregate actions of countless pollinators tion, etc.) and hydrological functions (water f ltration and at scales larger than that of individual plants (from several f ow regulation) that yield vital services. As with supporting square meters to the globe). services, pollination is important, but ecologically distal, to There is considerable middle ground in this spectrum. many of these regulating functions. The aforementioned Amazon freshwater f shery is relatively The benef ts of pollination are most tangible in the proximal to pollination (disruptions in f ooded forest production of provisioning services, such as products from pollination would have major consequences for the f shery non-managed ecosystems, including wild food (e.g., Brazil over short timescales). The pollination of mahogany nuts), f ber (e.g., rattan), and fuelwood resources. Many ( Swietenia macrophylla ), which provides valuable tropical animals hunted as food for people (not just the tambaqui timber, has fewer ecological linkages than the Amazon f sh) in turn feed on pollination-dependent fruits and other f shery example. Yet mahogany timber production could be plant parts. Such products can have a high economic value, considered more distal to pollination, because pollination particularly when considered in the aggregate (Peters et al. disruptions in any given year would be unlikely to have a 1989). Pollination is key for the population persistence of a strong effect on the value of the mahogany harvest that number of valuable timber trees as well, not just year. Continued pollination disruptions, however, would mahogany. eventually damage the mahogany timber industry since the Pollination interactions are invaluable in the varied roles trees could no longer reproduce in the absence of they play in providing ecosystem services beyond pollination. agriculture. Yet we still know little about how ongoing Pollination affects ecosystem services in interdependent anthropogenic environmental changes will affect ways. For example, the weevil-mediated pollination of communities of pollinators or the pollination functions Bactris gasipaes , the peach palm, is a regulating service. they perform. For example, there is serious concern in the But that service allows for the production of peach palm scientif c community that global climate change will lead fruits (a provisioning service), which in turn are a cultural to changes in the timing of f owering and of pollinator necessity in parts of Latin America—providing important foraging behavior, disrupting pollination interactions cultural services. worldwide (e.g., Memmott e t a l . 2007). Such disruptions Because plants are central to all of the primary would have major impacts on global ecological supporting services in the terrestrial biosphere—such as functioning and thus on a huge range of non-agricultural primary production, nutrient cycling, and preservation of pollination services. This makes the need for options (e.g., genetic diversity for future use in understanding and ameliorating the effects of pharmaceuticals)—this is perhaps the most important environmental change on pollination all the more functional role of non-agricultural pollination services. Just pressing. as one example, a large proportion of plants in the bean

likelihood that a pollinator could reach a given 90 Finally, our model, while quantitative, is essen- meter parcel, pollination delivery may be inf u- tially a statistical evaluation of the landscape so it enced by plant composition within 90 meters cannot project pollinator abundance over time. (Morandin and Winston 2006; Kremen et al . 2007). Rather it assumes population stasis given a This was not as much of a limitation in the land- particular landscape con f guration. In other words, scapes in this chapter but has been in other land- our model does not provide an estimate of pollina- scapes (Winfree et al. 2008). tor population viability or predict pollinator tem- LIMITATIONS AND NEXT STEPS 185 poral dynamics or interaction of time and space lination ; PI’s C.K. and N.M.W. Conservation priorities: through meta-population dynamics. As such, it Can we have our biodiversity and ecosystem services does not incorporate stochastic events, which may too? PI’s P. Kareiva, T. Ricketts, G. Daily) supported inf uence long-term population dynamics and by NSF grant DEB-00–72909, the University of yield. California at Santa Barbara, and the State of California. Kirsten Almberg helped with f gures and Jaime Florez provided measurements of inter- 10.9.2 Next steps tegular spans for Costa Rican bees. Berry Brosi pro- While new ecological data are needed to gain a vided expert assessment of nesting and f oral better understanding of the relationship between resources for the Costa Rica validation. Peter crop pollination and yield, we can use this current Kareiva, Erik Nelson, Berry Brosi, and Kai Chan all model framework to advance our understanding provided input in early development of the model. in a number of ways. First, we can apply this Saul Cunningham and Alexandra Klein provided model to a much larger set of crop studies con- valuable comments and corrections that improved ducted at the landscape scale (viz, studies in the chapter. Ricketts et al. 2008). Second, using statistical techniques, we can relate the landscape-level outputs of the model (pollinator supply) to the observed References measure of pollination services in each study (e.g., Allsopp, M. H., de Lange, W. J., and Veldtman, R. (2008). pollen deposition, pollen limitation) to attempt to Valuing insect pollination services with cost of replace- develop a direct relationship between landscape ment. PLoS One 3, e3128. and yield effects via pollinator abundances. Third, Brosi, B. J., Daily, G. C., Shih, T. M., et al. (2008). The effects by manipulating modeled landscapes (e.g., by of forest fragmentation on bee communities in tropical increasing f oral or nesting resources in different countryside. Journal of Applied Ecology 45 , 773–83. spatial conf gurations), we can estimate the effects Burd, M. (1994). Bateman’s principle and plant reproduc- on pollinator abundances and pollination services tion—the role of pollen limitation in fruit and seed set. Botanical Review 60 , 83–139. across a range of changes in resources, and look Cane, J. H. (1997). Lifetime monetary value of individual for generalities across landscapes in the density pollinators: the bee Habropoda laboriosa at rabbiteye blue- and arrangement of resources needed to provide berry (Vaccinium ashei Reade). Acta Horticulturae 446 , adequate pollinators and pollination services. 67–70. This would inform efforts to preserve existing Correa, S. B., Winemiller, K. O., Lopez-Fernandez, H., et al . habitats within degraded landscape and also (2007). Evolutionary perspectives on seed consumption guide planning of habitat restoration. Similar to and dispersal by f shes. Bioscience 57 , 748–56. our sensitivity analysis of model parameters, we Costanza, R., dArge, R., deGroot, R., et al. (1997). The value also envision analyses exploring the sensitivity of of the world’s ecosystem services and natural capital. modeled pollination services to resource patchi- Nature 387 , 253–60. ness at different grain sizes or to different land- Cresswell, J. E., Osborne, J. L., and Goulson, D. (2000). An economic model of the limits to foraging range in cen- scape conf gurations. tral place foragers with numerical solutions for bumble- bees. Ecological Entomology 25 , 249–55. Cross, P. C., and Beissinger, S. R. (2001). Using logistic Acknowledgments regression to analyze the sensitivity of PVA models: A comparison of methods based on African wild dog This work was facilitated by McDonnell Foundation models. Conservation Biology 15 , 1335–46. 21st Century and University of California Free, J. B. (1993). Insect pollination of crops . Academic Press, Chancellor’s Partnership awards to C.K., and by San Diego. two National Center for Ecological Analysis and Gallai, N., Salles, J. M., Settele, J., et al. 2009. Economic valu- Synthesis working groups ( Restoring an ecosystem ation of the vulnerability of world agriculture confronted service to degraded landscapes: native bees and crop pol- with pollinator decline. Ecological Economics 68 , 810–21. 186 CROP POLLINATION SERVICES

Gathmann, A., and Tscharntke, T. (2002). Foraging ranges McCarthy, M. A., Burgman, M. A., and Ferson, S. (1995). of solitary bees. Journal of Animal Ecology 71 , 757–64. Sensitivity analysis for models of population viability. Greenleaf, S., Williams, N., Winfree, R., et al. (2007). Bee Biological Conservation 73 , 93–100. foraging ranges and their relationships to body size. Memmott, J., Craze, P. G., Waser, N. M., et al . (2007). Global Oecologia 153 , 589–96. warming and the disruption of plant-pollinator interac- Greenleaf, S. S., and Kremen, C. (2006). Wild bee species tions. Ecology Letters 10 , 710–717. increase tomato production and respond differently to Morandin, L. A., and Winston, M. L. (2006). Pollinators surrounding land use in Northern California. Biological provide economic incentive to preserve natural land in Conservation 133 , 81–7. agroecosystems. Agriculture, Ecosystems & Environment Hines, H. M., and Hendrix, S. D. (2005). Bumble bee 116 , 289–92. (Hymenoptera apidae) diversity and abundance in tall- National Research Council of the National Academies. grass prairie patches: Effects of local and landscape f o- 2006. Status of pollinators in North America . National ral resources. Environmental Entomology 34 , 1477–84. Academy Press, Washington, DC. Hoehn, P., Tscharntke, T., Tylianakis, J. M., and Steffan- Peters, C. M., Gentry, A. H., and Mendelsohn, R. O. (1989). Dewenter, I. 2008. Functional group diversity of bee pol- Valuation of an Amazonian rainforest. Nature 339 , 655–6. linators increases crop yield. Proceedings of the Royal Polasky, S., Nelson, E., Camm, J., et al . (2008). Where to put Society B: Biological Sciences 275 , 2283–91. things? Spatial land management to sustain biodiver- Janzen, D. H., Ed. (1983). Costa Rican natural history . sity and economic returns. Biological Conservation 141 , University of Chicago Press, Chicago. 1505–24. Kim, J., Williams, N., and Kremen, C. 2006. Effects of culti- Potts, S. G., Vulliamy, B., Dafni, A., et al . (2003). Linking vation and proximity to natural habitat on ground- bees and f owers: how do f oral communities structure nesting native bees in California sunf ower f elds. pollinator communities? Ecology 84 , 2628–42. Journal of the Kansas Entomological Society 79 , 309–20. Potts, S. G., Vulliamy, B., Roberts, S., et al. (2005). Role of Klein, A. M., Mueller, C. M., Hoehn, P., et al . (in press). nesting resources in organising diverse bee communi- Understanding the role of species richness for pollina- ties in a Mediterranean landscape. Ecological Entemology tion services. In: D. Bunker, A. Hector, M. Loreau, et al ., 30 , 78–85. Eds., The consequences of changing biodiversity—solutions Priess, J. A., Mimler, M., Klein, A. M., et al. (2007). Linking and scenarios . Oxford University Press, Oxford. deforestation scenarios to pollination services and eco- Klein, A. M., Vaissière, B. E., Cane, J. H., et al . (2007). nomic returns in coffee agroforestry systems. Ecological Importance of pollinators in changing landscapes for Applications 17 , 407–17. world crops. Proceedings of the Royal Society 274 , 303–13. Ricketts, T. H. (2004). Tropical forest fragments enhance Knight, M. E., Martin, A. P., Bishop, S., et al . (2005). An pollinator activity in nearby coffee crops. Conservation interspecif c comparison of foraging range and nest Biology 18 , 1262–71. density of four bumblebee (Bombus ) species. Molecular Ricketts, T. H., Daily, G. C., Ehrlich, P. R., et al . (2004). Ecology 14 , 1811–20. Economic value of tropical forest to coffee production. Kremen, C., Bugg, R. L., Nicola, N., et al . (2002a). Native Proceedings of the National Academy of Sciences of the USA bees, native plants and crop pollination in California. 101 , 12579–82. Fremontia 30 , 41–9. Ricketts, T. H., Regetz, J., Steffan-Dewenter, I., et al . (2008). Kremen, C., Williams, N. M., and Thorp, R. W. (2002b). Landscape effects on crop pollination services: are there Crop pollination from native bees at risk from agricul- general patterns? Ecology Letters 11 , 499–515. tural intensif cation. Proceedings of the National Academy Southwick, E. E., and Southwick, L. (1992). Estimating the of Sciences 99 , 16812–16. economic value of honey-bees (hymenoptera, apidae) as Kremen, C., Williams, N. M., Bugg, R. L., et al . (2004). The agricultural pollinators in the United States. Journal of area requirements of an ecosystem service: crop pollina- Ecological Entomology 85 , 621–33. tion by native bee communities in California. Ecology Stokstad, E. (2007). The case of the empty hives. Science Letters 7 , 1109–19. 316 , 970–2. Kremen, C., Williams, N. M., Aizen, M. A., et al . (2007). Tepedino, V. J., and Stanton, N. L. (1981). Diversity and Pollination and other ecosystem services produced by competition in bee-plant communities on short-grass mobile organisms: a conceptual framework for the prairie. Oikos 36 , 35–44. effects of land-use change. Ecology Letters 10 , 299–314. Westrich, P. (1996). Habitat requirements of central Losey, J. E., and Vaughan, M. (2006). The economic value European bees and the problems of partial habitats. In: of ecological services provided by insects. Bioscience 56 , A. Matheson, S. L. Buchmann, C. O’Toole, et al. , E d s . , The 311–23. conservation of bees . Academic Press, London, pp. 1–16. LIMITATIONS AND NEXT STEPS 187

Williams, N., and Kremen, C. (2007). Floral resource distri- Winfree, R., Williams, N. M., Dushoff, J., et al . (2007). bution among habitats determines productivity of a Native bees provide insurance against ongoing honey solitary bee, Osmia lignaria , in a mosaic agricultural bee losses. Ecology Letters 10 , 1105–13. landscape. Ecological Applications 17 , 910–21. Winfree, R., Williams, N. M., Gaines, H., et al . (2008). Wild Winfree, R., Dushoff, J., Crone, E., et al . (2005). Testing sim- bee pollinators provide the majority of crop visitation ple indices of habitat proximity. American Naturalist 165 , across land-use gradients in New Jersey and Pennsylvania, 707–17. USA. Journal of Applied Ecology 45 , 793–802. CHAPTER 11 Nature-based tourism and recreation

W. L. (Vic) Adamowicz, Robin Naidoo, Erik Nelson, Stephen Polasky, and Jing Zhang

11.1 Nature-based tourism and In this chapter a site’s environmental attributes recreation values in context include its quantity and quality of ecological processes such as water delivery and cleanliness, biodi- International tourism and recreation generated over versity, quality and diversity of habitat, net primary $1 trillion in receipts in 2007 (roughly equivalent to productivity, etc. Tourism dynamics, however, are South Korea’s 2007 gross domestic product; World not explained solely by environmental attribute Tourism Organization 2008 ). Environmental supply. Tourism valuation models are essentially attributes of tourism sites are important in deter- models of human behavior applied to the decisions mining visitation rates and the value of tourism and of where, when, and how to engage in tourism. As recreation. For example, the earliest writings on such, the value that a tourist places on a particular tourism emphasize the environment quality at sea- site will depend on his/her personal characteristics side resorts, parks, and wilderness areas ( Towner (including past behavior and social interactions), and Wall 1991 ). National parks are often located in the local geography (including distance and cost of areas with photogenic biodiversity (e.g., Serengeti accessing sites as well as the presence of substitute National Park in Tanzania and Krueger National tourism sites) and the individual’s perception of Park in South Africa) or areas of scenic beauty congestion, environmental quality, and other site- (mountains, coasts, etc.). Many forms of recreation level attributes. require natural amenities (clean water for swim- The value of nature-based tourism in various ming, species diversity for birdwatching). By pro- parks, landscapes, or regions has been estimated in viding the natural features that attract tourists, a large number of economic studies ( Phaneuf and ecosystems provide a tourism and recreation Smith 2005 ). Most of these studies have involved service. primary data collection. However, if a tourism site Tourism generally refers to travel for pleasure or landscape has not been the focus of a detailed and typically involves overnight stays away from economic analysis, which represents the majority of home. Recreation typically refers to activities that cases, we have to rely on secondary data (e.g., data occur over part or all of a single day (e.g., hiking or collected by government agencies) to assess tour- f shing) that may take place during a tourism trip or ism values. Unfortunately, these secondary datasets as a day trip from home. For simplicity in this chap- often leave out key variables required for under- ter we use “tourism” as the general category for standing the linkages between visitation rates and both recreation and tourism. characteristics of sites such as their environmental Economists have developed a variety of tech- attributes. niques for assessing the economic value of tourism Here, we outline the conceptual basis for assess- ( Champ et al. 2003 ; Phaneuf and Smith 2005 ; ing the values people place on engaging in tour- Bockstael and McConnell 2007 ) and how the value ism activities across a landscape and develop of tourism is affected by changes in the supply of methods that increase in sophistication with environmental attributes ( Phaneuf and Smith 2005 ). increased data availability. Each potential tourism

188 NATURE-BASED TOURISM AND RECREATION VALUES IN CONTEXT 189 site has environmental characteristics that inf u- improved water quality. Figure 11.1 outlines the ence the attractiveness of tourism at the site. relationships between individual tourists, tourism Tourism tends to increase with improvement in destinations, and four factors that form the main environmental characteristics but is also inf u- components of the linkages between tourism, the enced by the distance of sites from tourists’ start- environment, and value: environmental attributes, ing location, availability of substitute tourism tourism site infrastructure, costs of visiting sites sites and other factors. To isolate the effect of the (illustrated by travel distance) and the availability environment on tourism use and value, we need of substitutes. Figure 11.1 provides an example with to construct models that control for these other two cities where individual tourists live, and two inf uences. (We use the term “tourism use” instead tourism sites. At each tourism site environmental of “tourism demand,” the term of choice in the attributes and infrastructure affect the attractive- valuation literature, to remain consistent with the ness of the site and thereby inf uence the number of taxonomy of ecosystem services presented in this trips taken and/or the choice to visit site A versus book; see Chapter 3 .) site B. The costs of visiting a site, captured in the We present three methods for assessing tourism travel distances, and the availability and impact of use and values in a landscape. In tier 1 we present a substitute sites complete the characterization. mapping methodology for spatially representing In assessing the value of a change in environmen- important tourism areas. Overlaying these site tal attributes, consider a base case in which all of the maps on maps of environmental attributes displays tourists visit the closest site, site A for city 1 tourists spatial correlations between tourism use and envi- and site B for city 2 tourists. Suppose the quality of ronmental attributes. The tier 2 model provides a the water in the river that f ows through site A more theoretically appropriate mechanism for improves. The improved water quality increases measuring the change in tourism use and value the attractiveness of site A, thereby increasing the given marginal changes in environmental attributes total number of trips from residents of city 1. In on the landscape. Tier 2 models can be used to addition, some residents from city 2 may now be approximate the change in tourism values under willing to travel further to enjoy the improved sup- future scenarios of land use/land cover (LULC) ply of water quality at site A. Both the increase in vis-à-vis the current landscape. We conclude with a number of visits from residents of city 1 and the discussion of the state-of-the-art tourism valuation change in site choice by residents of city 2 are ref ec- models (tier 3). In these models individuals exam- tions of the value of the improvement in an environ- ine the attributes of alternative destinations and mental attribute at site A. choose the destination that generates the highest Figure 11.1 also presents some of the complexities utility ( Train 2003 ; Phaneuf and Smith 2005 ). In associated with the measurement of tourism values. principle this approach can capture linkages among The assessment of value depends on modeling the environmental attributes, substitute sites, substitute choice of sites and/or number of trips by residents activities, demographic factors affecting value and of cities 1 and 2. Predicting trips requires informa- other aspects of tourism valuation, though practical tion on the characteristics of these residents (income, complexities and data requirements are both high. perceptions of site attributes, etc.), environmental and infrastructure attributes, cost of trips (travel distances, time costs) and substitute sites. In Figure 11.1.1 Major social and environmental 11.1 the set of substitutes is def ned as the two sites processes that affect tourism values A and B. In reality, there may be hundreds of substi- In general, the value of a tourism site will increase tute sites. Information about the spatial location as the quantity or quality of environmental attributes and costs associated with travel to each of these at the site increases. For example, the value of a site sites from the relevant residence zones is necessary visit for a bird watcher increases in the abundance for predicting trips. Finally, the model of tourism and diversity of species, for an angler with cleaner should be able to translate changes in environmen- water and f sh stocking, and for a beachgoer with tal attributes (e.g., water quality/quantity, species 190 NATURE-BASED TOURISM AND RECREATION

Tourism Site B Environmental Attributes: Fishing Quality, Wildlife Populations River/Watershed Scenery, Congestion Infrastructure Attributes: City 1 Residents: Facilities/Cultural/Spiritual Attractions Perceptions of Site Features Income, family size, etc. Time available Experience Equipment 75 km Substitute activities

150 km

200 km City 2 Residents: Perceptions of Site Features Income, family size, etc. Time available Experience 50 km Equipment Substitute activites Tourism Site A Environmental Attributes: Fishing Quality, Wildlife Populations River/Watershed Scenery, Congestion Infrastructure Attributes Facilities/Cultural/Spiritual Attractions

Figure 11.1 Tourism linkages to environmental attributes.

abundance and composition) into impacts on site recreational f shing particularly popular at sites attractiveness and the subsequent impact on with the cleanest water in a landscape? number of trips or site choice. The main strength of tier 1 is that it gives a picture of the current state of nature-based tourism on the 11.2 Tier 1 tourism supply and use landscape and only requires relatively easily col- model lected data to implement. The tier 1 data can also form the basis of the data collection efforts needed Creating a complete model of tourism that incorpo- to run a tier 2 analysis. In interpreting tier 1 results, rates all the dimensions discussed above requires it is important to note that spatial correlation in more data than are typically available. In the tier 1 tourism and environmental attributes does not tourism model we start with a more modest aim imply causality. Further investigation with more that requires much less data. We use maps to plot detailed data and analytical techniques, such as and characterize the current spatial pattern of those developed in tier 2 and tier 3 approaches, are nature-based tourism across the landscape. We needed to fully disentangle the effects of environ- measure use of a site for nature-based tourism with mental attributes from other factors such as site the number of visits to the site. We compare tourism accessibility and availability of substitute sites. use with several categories of site features and characteristics, including environmental attributes, site 11.2.1 Developing tier 1 maps infrastructure (e.g., campgrounds), and site accessibility (e.g., proximity to population centers, roads, The tier 1 approach involves compiling and over- airports). By overlaying these data layers we can laying f ve categories of data to investigate the spa- correlate use of a site for tourism activities with its tial relationship between use of sites for tourism, supply of environmental attributes. For example, is environmental attributes, and landscape features. TIER 1 TOURISM SUPPLY AND USE MODEL 191

The f rst data layer includes the sites that provide can calculate distance and travel time from urban tourism opportunities. Tourism sites can include areas to each tourism site. national, provincial, state, county, or privately held parks and recreation areas. In some cases, tourism 11.2.2 Example application of tier 1: activities are not restricted to easily def ned places Willamette Basin, Oregon, USA on the landscape. Some tourism activities take place across a broad region or landscape. For example, We applied the tier 1 approach to tourism in the duck hunting occurs on private and public land state parks of the Willamette Basin, Oregon, USA. throughout the Prairie Pothole Region of the The Willamette River and the Basin’s major high- Midwest USA. To account for this type of activity ways and cities are located in the Basin’s valley we may subdivide the landscape into zones to iden- f oor (Figure 11.2A ). Most state parks lie near cities tify the portions of the landscape where diffuse and are on the main stem of the Willamette River or tourism activity is more and less popular. one of its tributaries; a few parks are in the Cascades The second data layer maps visitation data. mountain range on the eastern side of the Basin Visitation data can be measured a number of ways: ( Figure 11.2A ). site visits per unit time (where one person can reg- Figures 11.2b through 11.2d present maps of ister multiple visits over the course of the time environmental attributes and other features in the period), number of visitors to the site per unit time, Basin (ODFW 2005; OGEO 2008; PNW-ERC 2008). visitor days per unit time (the sum of all visitors’ The state parks in the Basin with the most day length of stay in days), number of visitors purchas- visitors (not deconstructed by activity) are located ing entrance permits per unit time, etc. Stratifying in the Cascade Mountains and offer outstanding visit data by visitor place of origin, reason for visit, scenic attractions or recreation opportunities and time allows more detailed and comprehensive (Figure 11.3a ). Silver Falls State Park, which con- analysis. For example, stratifying visitors by inter- tains many large waterfalls, had the most visits in national and domestic origin can show differences 2004 even though its aggregate distance to the in visitation rates by cost (on average, international Basin’s cities was more than most other state visitors pay much more to tour than domestic tour- parks. ists). We discuss a way of placing a monetary value Detroit Lake State Park, higher up in the Cascades on the annual number of visits in Section 11.4 than Silver Falls and of greater distance from cities, below. was the most popular destination in the Basin for The third data layer maps information on envi- overnight camping (an activity that is generally ronmental attributes and other landscape features more costly than a day visit, given the time require- at sites. For example, does the site include a lake ments and the price of camping permits; see that would be a draw for swimming or f shing, Figure 11.3b ). Many sites in the northeast corner of major changes in elevation good for hiking or the Basin (and just outside the Basin along the dramatic views, habitat for charismatic species Columbia River) were popular tourism destinations that are a draw for wildlife viewing? High water both because they are close to the largest urban quality, suff cient water f ow, and abundant game center in the Basin (Portland) and feature the spec- f sh are vital for certain stretches of rivers if they tacular scenery around Mount Hood and the are to provide tourism value via recreational Columbia River Gorge. f shing. The greater use of nature-based tourism in the The fourth data layer maps information on impor- northeast corner of the Basin is also ref ected in tant infrastructure at each site. Infrastructure impor- hunting data. The Santiam hunting region (the tant for determining tourism site visits includes region that includes Mount Hood National Forest, roads, hiking trails, lodges, camping sites, and inter- Silver Falls State Park, Detroit Lake State Park, and pretative facilities. part of the Columbia River Gorge) is the most pop- The f fth data layer maps major transportation ular region for big game hunting, with 127 446 infrastructure and urban areas. With these data we hunter days in 2004 (see Figure 11.4a ). 192 NATURE-BASED TOURISM AND RECREATION

(a) (b)

Toursim site Willamette Greenway site Major road Urban area Highway River

Agriculture Closed forest Open forest Other Urban

Figure 11.2 Willamette Basin state parks and landscape features, characteristics, and environmental attributes. Major landscape features and characteristics and state parks in the Basin (a; OGEO 2008, PNW-ERC 2008, site data provided by Terry Bergerson, Oregon Parks and Recreation Department). Some access sites to the Willamette Greenway, a bicycle path, are considered state parks. The map of hillshade (b) in the Basin (PNW-ERC 2008) indicates areas of signif cant elevation changes. The map of landcover in the Basin circa 2000 (c; PNW-ERC 2008). The gray gradient in (d) represents an area’s marginal biodiversity value ( MBV ) score for 24 at-risk vertebrates, a tier 2 measure of biodiversity supply (see Chapter 8 ): the darker the parcel the greater the share of the 24 species’ total habitat on the landscape found in the parcel (see Hulse and Baker 2002 and Nelson et al. 2009 for details). TIER 2 TOURISM SUPPLY AND USE MODEL 193

(a) (b)

Figure 11.3 State park use in the Willamette Basin. Total number of day visits in a state park in 2004 (points proportional to number of visits). Silver Falls State Park (the black circle) had the greatest number of day visits in 2004 (981 680). The Lincoln Access of the Willamette Greenway (the white circle just to the northwest of the black circle) had the lowest use with 5 440 day visits. Total number of overnight camping visits in a state park in 2004 (b; the points in (a) and (b) are on the same scale). Detroit Lake State Park (the black circle) had the greatest number of campers in 2004 (84 137). Willamette Mission State Park (the white circle to the northwest of the black circle) had the lowest use with 2 312 campers. Most state parks in the Basin do not have camping facilities. Visit data provided by Terry Bergerson, Oregon Parks and Recreation Department.

All of this suggests that any changes in the sup- change, however, as the roster of species analyzed ply of environmental attributes or transportation changes. Similar analyses can be performed for infrastructure in the northeast corner of the Basin other environmental attributes of interest. may have the greatest impact on tourism use and value in the Basin. Interestingly, this area is expected 11.3 Tier 2 tourism supply and use to experience major forest cover transitions over model the next 100 years due to climate change (see C h a p t e r 1 8 ) . The tier 1 tourism methodology has two major One way to relate use of a site for tourism to envi- shortcomings. First, we cannot quantify how the ronmental attributes at state parks is to overlay the supply of environmental attributes at a site affects site map with maps of these attributes. To illustrate the overall tourism experience and its value at the this point, we construct a biodiversity supply map site. In addition, the tier 1 analysis cannot explicitly using the tier 2 biodiversity model (see Chapter 8 ), estimate how changes on the landscape could based on the habitat preferences of 24 terrestrial change tourism activity or value. For example, does vertebrates that are habitat-limited in the Basin a future LULC and land management scenario ( Nelson et al. 2009 ). Many of the state parks on the reduce environmental attribute supply across por- Basin f oor align spatially with some of the most tions of the landscape, and would this change valuable habitat areas in the Basin (see Figure 11.2d ). impact tourism at particular sites? Or, are new roads However, only camping visit rates are correlated or airports being built to facilitate access to a tour- with areas that supply the greatest share of habitat ism site? An additional challenge comes from eval- for these 24 at-risk species, whereas day visit rates uating the addition of a new tourism site. A new are not (see Figure 11.5 ). These correlates may tourist site is a substitute that could siphon some 194 NATURE-BASED TOURISM AND RECREATION

Figure 11.4 Diffuse tourism demand in the Willamette Basin. The size of a point indicates the total number of big game hunter days in a hunting region (deer and elk with bow or rif e, all seasons) in 2004 (OFWD 2005). Each hunting region is given by a distinct polygon. The Santiam hunting region, the region with the black circle, was the most popular region for big game hunting in 2004 (127 446 hunter days). The Metolius hunting region (to the southeast of the Santiam hunting region), the region with the white circle, was the least popular region for big game hunting in 2004 (8,952 hunter days). use from existing sites but it could also make tour- in the annual number of visits to a site due to a ism on its host landscape more attractive overall change on the landscape. and increase the landscape’s overall tourism use and value. 11.3.1 The visits model In tier 2 we develop a relatively simple model that predicts annual visitor days at each tourism The tier 2 model estimates the annual number of site (or region) as a function of the site’s (1) environ- visitor days or visits (hereafter, “visits”), to tourism mental attributes; (2) infrastructure; (3) amenities; sites as a function of several landscape variables. (4) distance to relevant population centers; and Environmental attribute and infrastructure varia- (5) spatial distribution of potential substitutes. We bles are site specif c and directly affect visitation. can use this model to predict the expected changes The costs of visiting a site depend on the location of in visitor days at each tourist site on the landscape the tourists relative to the sites and the activities due to expected changes in any of these f ve explan- participated in at the site. To capture costs we exam- atory landscape variables. In Section 11.4 , we dis- ine the proximity of population centers to the site to cuss a way to place a monetary value on (1) the develop an index of visitation cost for each type of annual number of visits to a site and (2) the change tourism activity. The availability and impact of TIER 2 TOURISM SUPPLY AND USE MODEL 195

(a) (b) 1.6 1.6 10–3 10–3 Lowell

1.2 Detroit Lake 1.2 Detroit Lake

0.8 0.8

0.4 0.4 Willamette Mission

Champoeg Champoeg Marginal Biodiversity Score Marginal Biodiversity Score Silver Falls Silver Falls 0.0 0.0 0246810 0246810 Total Day Visits (100,000s) Total Camping Visits (100,000s)

Figure 11.5 State park use versus biodiversity supply. The y -axes of both graphs indicate the marginal biodiversity value (MBV ) score at a state park’s location for 24 at–risk vertebrate terrestrial species (a tier 2 biodiversity supply measure; see Chapter 8 and Nelson et al . 2009 ). The x -axes indicate total visits in 2004, either day visits (a) or camping visits (b). Each circle represents a state park; the outliers are labeled. Most state parks do not have overnight camping.

substitutes depends on the spatial location of the stratify tourism visits or activities. Doing so involves site and potential substitute sites, as well as the sup- dropping subscripts a and/or ω and modeling the ply of environmental attributes, infrastructure, and total number of annual visits to each site q (i.e., qω , activity opportunities of the competing sites. We qa , or simply q ). illustrate how an index can be constructed to assess The vector of environmental attributes ( EAqa ) the impact of substitutes. includes biodiversity, scenic overlooks, boating In tier 2 we model the number of annual visits to opportunities, etc., while the vector of infrastruc- each site q to participate in a combination of activi- ture variables ( X qa ) can include campgrounds, bath- ties a by tourist type ω, represented as Tqa ω , as a rooms, hiking trails, etc. The costs of visiting site q function of a vector of environmental attributes at q to participate in activity combination a ( G qa ) will that could impact participation in a (EA qa ), a vector depend on the distance of the site to population of infrastructure variables at q that could impact centers and the costs of participating in activity participation in a (X qa ), the relative cost of visiting combination a . An index that uses the distance from the site q to participate in a (G qa), and an index of all relevant population centers to q is one way to substitute sites that provide at opportunity to par- determine the relative cost of visiting a site to participate in a ( Sqa ), ticipate in activity combination a ,

I T f E A X GgChd= (1− bbs ) + (), q a ω = ( q a , q a , G q a , Sq a ) , (11.1) qa qa qa qa∑ ia iq (11.2) i=1

where a = 1,2, . . . , A indexes any combination of where g qa is a f xed cost of participating in activity activities, e.g., a = { f shing; camping; hiking; f shing combination a at site q (e.g., an entrance or partici- and camping; f shing and hiking; . . . ; f shing, camp- pation fee), σ is a scalar, C ia is the number of people ing and hiking}. We deconstruct visits by activities that might participate in activity combination a at because explanatory variables can affect participa- site q that are from population center i , h ( diq ) is an tion for each activity differently. Further, different increasing function of the distance from population types of tourists can react to site attributes and costs center i to site q , d iq , and βqa Î [0, 1] is used to weight differently. Here and throughout, these models can the relative importance of f xed participation costs be simplif ed if we do not have suff cient data to versus relative travel costs in the cost index of 196 NATURE-BASED TOURISM AND RECREATION

participating in activity combination a at site q . A commonly used form for h ( d ) is erdiq with 0 < ρ iq Site q’ ≤ 1. In general, as the aggregate distance to site q increases (weighed by C ia ) the greater the average cost of visiting site q . In Eq. (11.2), population cent- City A ers that have a greater number of people that might participate in a are given more weight in the deter- k1 mination of G qa . The use and value of a tourism site for activity combination a will tend to be lower if there are nearby substitute sites for a . We illustrate a method for measuring the impact of substitute sites on the k2 use and value of a given site q for activity combination a . Our proposed index, S , is a function of envi- qa City B ronmental attributes and infrastructure at nearby sites weighted by the difference in distance that Site q these sites are from population centers,

k3 −−()ddik iq g SCyzeqa= ∑∑ ia×××()()EA ka X ka , (11.3) kN∈∈qa iN qa

where q is a site suitable for activity combination a ,

N qa is the set of tourism sites in q ’s “neighborhood” Figure 11.6 Calculating the impact of local tourism substitutes. The neighborhood of site q for stream f shing includes alternate tourist sites k , that provide opportunities to participate in activity 1 k , and k and city B (all sites within the large black circle centered on city B). In combination a (neighborhood sites are indexed by 2 3 other words, people from city B typically choose among q , k 1 , k 2 , and k3 when k ), population center i is in N qa if population center i planning a day of f shing. The value of Sq in this case will depend on the supply could supply people for a in the area formed by N , qa of environmental and infrastructure attributes at sites k 1 , k 2 , and k 3 and the

y ( EA ka) is an increasing function in k ’s environmen- distances between city B and the alternate tourism sites (given along road highways, as indicated by dashed lines). The neighborhood of more northern site tal attributes that affect a , z ( X ka) is an increasing qʹ includes alternate tourist site k and city A (both within the black circle function in k ’s infrastructure that affects a , d is as 1 iq centered on city A). above, d ik is the distance from population center i t o site k , and γ is a scalar. The distance to alternate sites is a proxy for the cost of using these sites. other sites in the world that provide safari tourism

A higher Sqa indicates more competition from and all of the world’s major urban centers. On the rival sites for use of site q for activity combination a . other hand, if we are modeling recreational stream

All else equal, Sqa increases with (1) the number of f shing (typically a one day activity) then N qa should alternate sites k in site q ’s neighborhood N qa, (2) the only include alternative stream sites a few hours number of population centers i in N qa , (3) the envi- from q ’s stream and urban centers that are a few ronmental attributes and infrastructure in alternate hours from q . Def ning N qa for such a local recrea- sites k in N qa , and (4) increasing distance from i in Nqa tion market is illustrated in Figure 11.6 . to q vis-à-vis the distance from i in Nqa to the set of rivals k . 11.3.2 Estimating a visits model The extent of site q ’s neighborhood N qa should be given by the furthest someone would travel to par- Assuming we have data on tourist visits to sites and ticipate in activity a . For example, if a is a wildlife information on the independent variables described safari, an activity that people will travel around the above, we can estimate Eq. (11.1) across all potential globe to participate in, then q should be a site that sites for each combination of activity and tourist provides safari tourism and N qa should include all type, using standard regression techniques. Or, TIER 1 AND 2 USE VALUE 197 given that all activities in a site will be affected by remains constant. If the landscape has changed such similar unobserved factors (e.g., the macroeconomic that environmental attributes in q have changed, conditions in the demonstration site’s country, the represented by ∆EA q, then the expected change in macroeconomic conditions in the countries that predicted T qa ω is given by supply the tourists, country stability), it may be ˆ most appropriate to pool all activity and tourist ∆Tfqaww= a(,,,),∆∆EA q X q GS qa qa (11.5) type models. A promising technique for this approach is the seemingly unrelated regression where fa ω indicates the regression-estimated Eq. framework (see Greene (2003 ) for details). In Section (11.1) for combination a , ω, and ∆ S qa indicates any 11.4.1 we f t Eq. (11.1) to data for the Willamette change in the substitute index due to the change in

Valley in Oregon, USA. ∆ EA q (a change on the landscape that changes EA at

q could also change EA at some other q ). ∆ EA q can 11.4 Tier 1 and 2 use value include just one change (e.g., the change in water quality at q ) or multiple changes (e.g., the change in A monetary value of tourism at each site can be gen- water delivery and quality at q ). We can also meas- ˆ erated by multiplying the number of observed (tier ure ∆Tqaw due to changes in Xq and G qa . 1) or estimated (tier 2) visits by an average value per The expected change in monetary value at site q ˆ visit (or visitor day). The value of a tourist visit to a due to the change in ∆Tqaw is, site equals the benef t generated by this visit over A Ω ˆˆ and above the costs of this visit. This measure of ∆AVTqqaqaqaqaqa= ∑∑[( Vwww∆ T+ T )][− V ww T ], (11.6) value is known as consumer surplus ( Loomis 2005 ). a==11w ˆ In general, consumer surplus can vary by site, type where Vqaw represents the new value per trip after of tourist, and tourist activity. By aggregating the the change in environmental attributes on the consumer surplus generated by each visit to a site landscape. over the course of a year we calculate a site’s annual The landscape-level annual value of tourism and value (effectively the value lost if access to the site is annual change in the value of tourism due to removed). In addition, using the tier 2 methodol- changes in environmental quality is given by sum- ogy, an estimate of the monetary effect of a change ming AVT q and ∆ AVT q , respectively, across all q on in environmental attributes can be developed by the landscape. comparing the predicted value (trips multiplied by Finally, we can replace V qaω in the equations above value per trip) at the site under the baseline condi- with other values of interest, for example, expendi- tions against the predicted value at the site after a tures in the region by tourists participating in activ- change in environmental attribute. ity combination a at site q , to generate other Let the annual value of tourism at site q be given tourism-related economic information (e.g., the eco- by AVTq where nomic impacts on local businesses). See this chap-

A Ω ter’s text box ( Box 11.1 ) for a discussion on tourism AVTqq= ∑∑ Vaww Tqa , (11.4) revenues generated by visits to Tambopata, Peru. a==11w If we do not have consumer surplus estimates for

where T qaω is trips as above, Vqa ω is the average con- the site we are studying, we may be able to use a sumer surplus generated by a visit to site q to par- consumer surplus estimate from a valuation study ticipate in activity combination a by tourist type ω of a similar region as a proxy (for example, using an (drop all a and/or ω subscripts if we do not have estimate of recreation values per day generated by a data by activity combination or type). study in a nearby state). As noted in Chapter 3 there

We can measure the change in AVT q due to a are limitations to benef ts transfer approaches, thus change in environmental attributes at q by using the primary studies are preferable. In this case we only regression-estimated Eq. (11.1) to predict changes in transfer the value per unit of recreation activity visitation with a change in environmental attributes, (days, trips). For behavioral response (visits) to and assuming that consumer surplus per trip changes on the landscape, we use a locally calibrated 198 NATURE-BASED TOURISM AND RECREATION

Box 11.1 How the economics of tourism justif es forest protection in Amazonian Peru

Christopher Kirkby, Renzo Giudice, figure. High levels of profitability, and the expectation of Brett Day, Kerry Turner, Britaldo Silveira future profits, have created the incentive and the means Soares-Filho, Hermann Oliveira-Rodrigues, for lodge owners to protect their businesses by and Douglas W. Yu protecting forest cover. Many lodges have taken advantage of government legislation, passed in 2002, The province of Madre de Dios in south-east Peru, an that lets private businesses lease public forest outside of Amazonian frontier region bordering Bolivia and Brazil, protected areas as concessions, for renewable 40-year is renowned amongst scientists for its biologically and terms. By 2005, lodges had taken control of 32 477 ha, culturally rich landscape. One area of this region in with 90% of this area acquired by the four most particular, known as Tambopata, is now also firmly profitable operators, together managing 8 lodges. entrenched in the minds of international travellers as the Another 16 159 ha have been provisionally awarded as quintessential Amazon rainforest destination, attracting of early 2008, totalling 48 636 ha ( Figure 11.A.1 ). Eight 39 565 ecotourists in 2005. Tambopata’s growing other lodges own and manage <100 ha each. In 2005, popularity is largely the result of (i) ease of access, only the pre-tax profit value of lodge-controlled lands a 30-minute flight from Cusco (the gateway city to corresponding to 12 fully operational lodges (for which Machu Picchu) to Puerto Maldonado (the gateway town economic and land-use data were available) was to Tambopata) followed by a few hours of river travel in US$38.9 ha− 1 [US$1 238 002/31 807 ha]. This figure motorised canoe; (ii) the proximity to two large exceeds the 2005 pre-tax profit value of titled lands, protected areas, the Tambopata National Reserve (TNR, covering 10 2511 ha, that were managed by 200 274 690 ha) and the Bahuaja-Sonene National Park generalist rural households for agriculture, fruit (BSNP, 1 091 416 ha); (iii) a healthy natural ecosystem production, animal husbandry (cattle ranching, chickens, showcasing primates, giant otters, macaws, and other and pigs), and timber extraction, which was calculated charismatic fauna; (iv) intact oxbow lakes and clay-licks at US$27.1 ha −1 [US$277 472/10 2511 ha]. However, that concentrate fauna in predictable ways; (v) a choice when it comes to those households that specialize in a of 37 ecotourist establishments (lodges), from 100-bed given land use (> 50% of household revenues), operations to small research stations and village unsustainable cattle ranchers (with stocking rates >3.5 guesthouses; and (vi) spending on international animals ha− 1 , N = 4), who owned 262 ha, extracted a marketing by the larger lodges. In 2005, ecotourists pre-tax profit value of US$39.0 ha− 1 , whilst sustainable spent a total of US$11.6 million to visit Tambopata, of cattle ranchers (stocking <3.5 animals ha −1 , N = 1 5 ) , which US$5.9 million were lodge revenues, US$5.2 who owned 1 154 ha, extracted US$35.8 ha −1 . W e n o t e , million were airline revenues for airfares between Cusco though, that cattle ranchers are generally located within and Puerto Maldonado, and US$0.5 million were TNR 2 km of a road, for ease of transport, which limits the entrance fees and airport taxes. US$3.8 million (32.5%) land area where this activity would compete with of these revenues were in turn spent locally, in that the tourism. For households specialized in timber extraction first-order transaction took place in Tambopata. This or annual agriculture (i.e., rice, maize, manioc, and local spending could be further divided into low (12.2%, bananas), the pre-tax profit value of land is US$35.7 e.g., local staff, produce) and high leakage (20.3%, e.g., ha −1 a n d U S $ 2 1 . 3 h a− 1 , respectively. We expect that gasoline, boat motors) to the national economy. The TNR specialist producers, especially unsustainably stocked entrance fees exceeded the local park management ranchers, will suffer reduced profits in the future as budget, allowing US$172 530 to be transferred to the productivity drops, which should lower the present value national budget. In 2005, the lodges earned a of their profit stream below that expected for tourism. combined, after-tax profit of US$844 472. Additional Our comparisons are also conservative in that we do not (but only partly quantified) profits were distributed to include the above-market-rate wages paid to lodge some of the lodge owners in the form of above-market owners. One stated motive for lodges acquiring forest wages, possibly as much as doubling the above profit concessions is to exclude competitors from primary TIER 1 AND 2 USE VALUE 199

Figure 11.A.1 The location of lodge-controlled lands, made up of a mixture of designated ecotourism, conservation and Brazil-nut concessions, and concessions awaiting f nal approval, as of 1Q 2008, in relation to protected areas (TNR, Tambopata National Reserve; BSNP, Bahuaja-Sonene National Park), and areas of deforestation associated with Puerto Maldonado and the Interoceanica Highway. “A” depicts deforestation within the TNR, which is associated with the colonist communities of Jorge Chavez and Loero, corresponding to a 20-km wide gap between the two clusters of lodges. “B” and “C” are proposed ecotourism concessions and “D” is an ecotourism concession granted to a mestizo community that has historically mined alluvial gold deposits. “E” is a portion of forested land, located within the Native Community of Inf erno, which though not controlled by a lodge has been set aside by the community for their ecotourism joint venture with a Lima-based tour company. forest with valuable touristic features such as trail Interoceanica Sur Highway, a westerly extension of the networks, oxbow lakes, and clay-licks. Another motive is Trans-Amazon Highway that will connect Brazil to the that titles and concessions provide lodges with Pacific Ocean. The highway will be completed in 2009 state-legitimized claims to forest parcels that they can and will encourage deforestation along its length, thus defend via the legal system or direct action. Lodges have directly threatening the lodges, which are on average successfully sued and evicted loggers and miners from only 18 km (8–62 km) from the highway, within the their concessions and have entered into benefit-sharing 50-km deforestation zone associated with paved roads agreements with neighboring communities to cease in the Brazilian Amazon. Based on current behavior, extraction and hunting. Two lodges have entered into a Tambopata’s ecotourism industry has the incentive and joint venture with a community-based ecotourism the means to continue protecting and even expand their project in return for monitored agreements to maintain concession holdings, but computer simulations indicate forest cover and limit hunting in and around the that even if lodges successfully maintain their proposed ecotourism concession ( Figure 11.A.1 , “B”). In concessions, deforestation will proceed through the one notable episode in 2007, the ecotourism industry unprotected gap between the two ecotourism clusters added its weight to lobbying by Peru’s conservation ( Figure 11.A.1 , “A”), ultimately threatening many of the community and successfully reversed a government lodges by degrading connectivity to the TNR. The gap proposal to de-gazette a portion of the BSNP for oil area is less suitable for tourism, as it lacks oxbow lakes, exploration. The most serious threat to Tambopata’s so conservation in this section will require public biodiversity is yet to come, however. In 2005, the investment, perhaps bolstered by collective action government of Peru secured US$892 million to pave the among the lodges. 200 NATURE-BASED TOURISM AND RECREATION production function (Eq. (11.1)). Loomis (2005 ) pro- points and may support more visits before becom- vides estimates of consumer surplus for US nature- ing congested), and the presence/absence of his- based recreation and tourism, and Shrestha and toric sites. For G , we approximate the overall cost Loomis ( 2001 ) provide estimates in other parts of of accessing each site with a variable that meas- the world. A roster of tourism consumer surplus ures the population living within 20 miles of the databases is provided at http://recvaluation. park (i.e., Cia in Eq. (11.2) is replaced by C q , the forestry.oregonstate.edu/brief_history.html. population living within 20 miles of q , σ = 1, e−diq = 1

Defenders of Wildlife and Colorado State University for all i , q , a n d βa = 1 f o r a l l q ). have constructed a “Toolkit” for benef t transfer We estimate the visitation model with ordinary that includes databases, meta-analyses and visitor least squares regression (Table 11.1 ). Fishing and use models: http://dare.colostate.edu/tools/bene- wildlife opportunities are highly related (i.e., f ttransfer.aspx. Finally, a database of values from parks tend to have both or neither), so we estimate around the world, called the Environmental the model twice, once with each of these variables Valuation Reference Inventory, can be found at included. We found that canyon features increase http://www.EVRI.ca. The reader should be aware, park visitation. Somewhat surprisingly, f shing however, and as we emphasize in Chapter 3 , the and wildlife viewing opportunities do not explain evidence on the extent to which environmental ben- visitation in a statistically signif cant manner. ef t measures are transferable is mixed at best Parks that are less costly for more people to access ( Navrud and Ready 2007 ). (as indicated by population around the park) are used more. Based on these results one might assume that important environmental attributes 11.4.1 Example of tier 2 using data from the do not contribute to tourism use. What is more Willamette Basin likely is that there is relatively little variation in In this example we illustrate the estimation of a tier these variables over the sites, particularly when 2 visitation model and the value of a change in envi- they are measured using coarse indicators like ronmental attributes in state parks in the Willamette presence/absence. This lack of variation in eco- Basin. This model is very similar to the approach system-based variables is a common challenge in that Hill and Courtney ( 2006 ) use to estimate the econometric studies of this type (e.g., Adamowicz effect of changes in landscape variables on visits to et al. 1994 ). Improved information on levels of public and private forests in Great Britain. environmental attributes will help alleviate these First, we use Eq. (11.1) and data from 41 problems. Willamette Basin state parks to assess the role of Though this illustrative model is fairly simple, it environmental attributes, infrastructure, and park reveals a great deal about the marginal value of location, on visitation rates. The model used in environmental attributes at state parks in the this example is a simplif ed form of Eq. (11.1) for Basin. For example, if a site that currently supports two reasons: (1) visitation data are not stratif ed boating experienced a change in water delivery by activity combination a or tourist type, and (2) and/or quality to the degree that boating could no we did not model a substitution effect (S ). We longer be supported, then, all else equal, the aver- explain the number of visits to state park q in 2004, age site is predicted to lose between 65 000 (model

T q , with seven factors. For the EA vector, we 1) and 97 000 (model 2) visitors a year (the average include presence/absence for four environmental ∆ T ̂across all sites given the loss of boating). Given attributes: (1) f shing possibilities, (2) wildlife that the average visitation rate at the parks is about viewing possibilities, (3) canyon features, and (4) 200 000 visitors per year, this is a substantial loss. If boating facilities. We assume that a park with can- we could obtain a measure of the consumer sur- yon features has dramatic views and good hiking plus in the Basin before and after a potential loss of ̂ opportunities. For the X vector, we include two boating in a state park, given by V q a n d V q , r e s p e c - infrastructure variables: the area of the park tively, we could derive coarse estimates of lost (assuming that larger sites have more access value using Eq. (11.6). STATE-OF-THE-ART TOURISM MODEL 201

Table 11.1 Ordinary least squares regression model of annual visits to Oregon State Parks

Variable Coeff cient (t -stat)

Model 1 Model 2

Constant 36 870.47 59 669.76 (1.34) 1.93 *** Area of park (in km 2 ) 14 328.77 16 504.51 (3.72) * (4.31) * Population (in thousands) within 20 miles of park 109.68 109.19 (2.13) ** (2.15) ** The presence of a historical site at the park (HISTST, 0/1 indicator variable) 155 545.83 157 442.03 (2.66) * (2.73) * The presence of canyons as a natural feature of the park (CANYONS, 0/1 indicator variable) 222 907.70 203 305.19 (3.24) * (3.05) * The opportunity for boating (BOATING, 0/1 indicator variable) 64 352.87 96 905.34 (1.58) *** (1.96) ** The opportunities for wildlife viewing including birdwatching and other wildlife 29 581.59 (WILDLFW, 0/1 indicator variable) (0.78) The opportunities for f shing (FISHING, 0/1 indicator variable) –43 391.51 (–0.98) R 2 0.78 0.78 R 2 -adjusted 0.74 0.75 N 41 41

* Signif cant at a 0.01 level. ** Signif cant at a 0.05 level. *** Signif cant at a 0.15 level.

11.5 State-of-the-art tourism model The basis for these choices is a preference function that is def ned on characteristics of the individ- The limitations associated with the simple overlay ual (income, family size, etc.) as well as the attributes approach in tier 1 or regression model approach in of the available options (distance to the parks, tier 2 point to the utility of the development of attributes at the parks, etc.). Two types of models state-of-the-art or tier 3 models. Tier 3 begins with have been examined in the literature: the Kuhn– the individual tourist as the core unit of analysis, Tucker model ( Phaneuf 1999 ; Phaneuf et al. 2000 ) and models human behavior from the standpoint and the random utility model, or RUM (see Phaneuf of the decision on where and when to visit, rather and Smith 2005 ). than as a statistical analysis of aggregate visitation The Kuhn–Tucker model assumes that the indi- rates at parks. Note that the notation used in this vidual n obtains utility from a set of trips to rec- section deviates somewhat from that above in order reation sites where q = 1 , . . . , Q indexes the set of to be consistent with the published literature. sites available and ( x1 , . . . , x Q ) indicates the number Consider an individual living in the Willamette of visits to each site. Associated with these sites is

Basin. This individual has a set of available choices a set of attributes (b 1 , . . . , b Q ) where each element b of state parks, as well as other tourism and recrea- contains a number of characteristics of the sites, tion options within and outside the Basin. The indi- including those related to environmental vidual also has other non-tourism options competing attributes. The individual’s preference function for her time. The individual is assumed to make includes a set of “other” activities that generate choices of how many tourism and recreation trips to utility (all other activities are accumulated into a make (in a season or a year) as well as the trip single element z ). This results in a preference func- destination. tion U (x 1 , . . . , x Q , b 1 , . . . , bQ , z ) . T h e i n d i v i d u a l n is 202 NATURE-BASED TOURISM AND RECREATION

assumed to maximize utility (U n) subject to income a desirable environmental attribute at site q declines ( m ) over the period and the prices of accessing the then the probability of choosing site q for tourism recreation alternatives (p 1 , . . . , p Q ; usually travel declines. According to Eq. (11.9), as quality declines cost plus the opportunity cost of time plus site at one site, visitation will increase at other sites. This access costs) as well as the price of the “good” z addresses one of the major limitations of the tier 2 (which is normalized to unity). The individual model: the diff culty with satisfactorily modeling maximizes utility by choosing the number of trips substitutability when explaining tourism visits (the to each site q , g i v e n b y x1 , . . . x Q where some or all index Sqa ). Further, because this model incorporates x c a n b e 0 : the preferences of the individuals in the conditional indirect utility function (V ) it can also be used to

maxUx (11 , ..., xQQ , b , ..., b , z ) (11.7) calculate the economic value of the change directly. xxz,..., , 1 Q In this case the value of the decline at site q i s g i v e n Subject to pʹ x + z = m . by the amount of money it would take to make per- The model provides estimates of the relationship son n as well off as they were before the change at q ; between travel costs, attributes, income (and other this can be calculated using the utility expression in demographic characteristics) and the frequency of (11.8) over the entire set of sites available. The rela- visits to each of the def ned alternatives. Typically, tively simple model can be combined with ecologi-

xq*, the number of visits to q that solves problem cal models that describe the linkages between (11.7), will decrease as the environmental attributes attributes to create an integrated ecological— at q decreases, all else equal. economic model. An example of such a model is The random utility model ( Phaneuf and Smith Naidoo and Adamowicz (2005 ) in which a behavio- 2005 ) examines the individual’s choice on a particu- ral model of tourism site choice is integrated with lar occasion (a single trip). Rather than assessing the an ecological model of landscape change and bird number of trips, the utility or satisfaction associated diversity. with visiting an alternative is described as a func- There have been many signif cant advances in the tion of the travel costs and site environmental and modeling framework outlined above. Some of the infrastructure attributes (and potentially individual most signif cant include: (1) accounting for unob- specif c characteristics such as income, demograph- served attribute effects in the model (Murdock ics, etc.). The utility (or preference) function for the 2006 ); (2) incorporation of congestion into the model choice to visit to site q (from individual n ’s view- as an example of interdependence between tourists point) takes the form: ( Timmins and Murdock 2007 ); (3) the incorporation of time (habits, variety seeking) into the framework (e.g., Swait et al. 2004 ); (4) inclusion of preference VVmpbnq = (,).− qq (11.8) heterogeneity among tourists (Scarpa et al. , forth- A common approach is to model the probability of coming; Boxall and Adamowicz 2002 ; Train 1998 , individual n visiting a particular alternative q (or 2003 ); and (5) the development of models that 0 ≤ π ≤ 1) as account for different alternatives in the choice set or nq set of sites that a tourist considers ( Haab and Hicks eVnq 1997 ; von Haefen 2008) as well as a host of other p nq = , (11.9) eVnk advances in the modeling of choice data. ∑ kQ=1, ...

where the V function would have to be estimated nq 11.6 Limitations and next steps using observed data in a regression analysis ( Train 2003 ). We have outlined three approaches (tiers 1, 2, and 3) Equation (11.9) expresses the probability that an for assessing the value of nature-based tourism and individual visits a particular alternative (on a par- recreation on the landscape and the changes in ticular choice occasion—e.g., a day, a week) as a value that could be expected with a change in envi- function of the attributes and prices. If the supply of ronmental attributes on the landscape. LIMITATIONS AND NEXT STEPS 203

Tier 1 provides an assessment of observed visi- Finally, the models described here are static and tation rates to sites. If information on the value of do not ref ect trends in preferences, demographics a unit of visitation is available from other studies or other factors that might inf uence visitation rates. (benef ts transfer) then these visitation rates can For example, Pergams and Zaradic ( 2006 , 2008 ) be used to approximate measures of economic argue that there have been widespread declines in value. The tier 1 methodology does not specify a nature-based recreation visits. Meanwhile Balmford relationship between changes in environmental et al. (2009) f nd that visits to protected areas in most attributes and changes in tourism value nor does parts of the world are in fact increasing. These are it disentangle effects on visitation rates from envi- clearly areas for further research. ronmental attributes, infrastructure, distance to Tier 3 generates value estimates based on the population centers, and availability of substitute behavior of the individuals. It is a fully integrated sites. model of tourism and the environment. However, it Tier 2 approximates visitation behavior by devel- is also very demanding in terms of data require- oping a statistical relationship between visitation ments and familiarity with sophisticated modeling and attributes at the sites. This approach can pro- techniques. As the model is individual based it vide additional insight into the changes that may requires information on the individual’s residential be expected in visitation rates and values if envi- location (for the determination of travel and time ronmental attributes change on the landscape. costs), the set of sites the individual considers when Tier 2 approaches, however, suffer from several planning trips, as well as information on the limitations. attributes of the sites. There are several “scale” First, the required data are usually highly corre- issues including assessment of the relevant geo- lated, and attributes often do not vary strongly graphical scale (how large is the area that is relevant among sites. Sites with good f shing quality also for the demand for recreation tourism at a particu- tend to have boating, picnicking, and other facili- lar site) and the relevant time scale (is an annual ties. And sites often share many features, preclud- time scale appropriate for decision making or are ing the opportunity to identify the impact that the trips more seasonal or perhaps a one- every-5- these features have on visitation rates. Identifying years type of trip; is there a broad trend of declining the impact of the change in environmental participation in recreation and preferences for attributes in such cases will be diff cult. Increasing nature?). Only a few regions will have the data the number of parks or expanding the spatial available for such analysis. extent of the range of parks may help, but these Ideally the data for tier 3 models would be col- actions will also increase the complexity of the lected from general population surveys. Data col- research task. lection of this type could provide excellent sampling Second, there are few linkages to information properties and would provide information on the about the tourists. Only the potential tourism popu- total number of visits taken by a population as well lation is included in the model. Factors such as spe- as the sites selected. However, such data are rarely cif c travel and time costs of visiting (instead of an collected. An alternative is to sample at the recrea- index), incomes, experience levels of the tourists, tion sites and collect information on the partici- the substitution between tourism and other uses pants. This is known as choice-based sampling of time, and other demographic features are not ( Ben-Akiva and Lerman 1985 ) and is commonplace included. in the transportation demand literature. This Third, tier 2 models will generally rely on benef ts approach may provide the most practical solution transfer procedures to provide estimates of eco- for the development of tier 3 models. nomic value rather than estimating the values from An additional area that has not been investigated the population of interest. Evidence on the applica- to any great extent is the feedback between visita- bility of values from one site to others is decidedly tion rates, congestion and environmental quality. mixed, but may be the only option (Navrud and There has been some assessment of the role of con- Ready 2007 ). gestion in nature-based tourism, and examination 204 NATURE-BASED TOURISM AND RECREATION of the impact of tourism on environmental quality, Technical Report PNW-GTR-658. US Department of but little examination of the three elements jointly. Agriculture, Forest Service, Pacif c Northwest Research These issues as well as the continuing evolution of Station, Portland, OR. niche markets for tourism and emerging trends Murdock, J. (2006). Handling unobserved site characteris- form the basis for a rich research agenda. tics in random utility models of recreation demand. Journal of Environmental Economics and Management, 51 , 1–25. Acknowledgments Naidoo, R., and Adamowicz, W. L. (2005). Economic benef ts of biodiversity exceed costs of conservation at an Thanks to Isla Fishburn, University of Sheff eld, for African rainforest reserve. Proceedings of the National collecting much of the Willamette data used in this Academy of Sciences of the USA, 102 , 16712–16. chapter. Navrud, S., and Ready, R. (2007). Environmental value transfer: issues and methods . Springer, Dordrecht, The Netherlands. References Nelson, E., Mendoza, G., Regetz, J., et al . (2009). Modeling multiple ecosystem services, biodiversity conservation, Adamowicz, W. L., Louviere, J., and Williams, M. (1994). commodity production, and tradeoffs at landscape Combining stated and revealed preference methods for scales. Frontiers in Ecology and the Environment, 7 , 4–11. valuing environmental amenities. Journal of Oregon Fish and Wildlife Department (OFWD). (2005). Environmental Economics and Management , 26 , 271–92. 2005 Big Game Statistics. Balmford, A., Beresford, J., Green, J., et al . (2009). A global Oregon Geospatial Enterprise Off ce (OGEO). (2008). perspective on trends in nature-based tourism. PloS Oregon geospatial data clearinghouse. Accessed at Biology , 7 (6), e1000144. doi:10.1371/journal.pbio.1000144. http://www.oregon.gov/DAS/EISPD/GEO/alphalist. Ben-Akiva, M., and Lerman, S. R. (1985). Discrete choice shtml in 2008 . analysis: theory and application to predict travel demand . Pacif c Northwest Ecological Research Consortium (PNW- MIT Press, Cambridge. ERC). (2008). The datasets. Accessed at http://www.fsl. Bockstael, N. E., and McConnell, K. E. (2007). Environmental orst.edu/pnwerc/wrb/access.html . and resource valuation with revealed preferences: a theoretical Pergams, O. R. W., and Zaradic, P. A. (2006). Is love of guide to empirical models . Springer, Dordrecht, The nature in the US becoming love of electronic media? Netherlands. 16-year downtrend in national park visits explained by Boxall, P. C., and Adamowicz, W. L. (2002). Understanding watching movies, playing video games, internet use, heterogeneous preferences in random utility models: a and oil prices. Journal of Environmental Management, 80 , latent class approach. Environmental and Resource 387–93. Economics, 23 , 421–46. Pergams, O. R. W., and Zaradic, P. A. (2008). Evidence for Champ, P. A., Boyle, K. J., and Brown, T. C. (2003). A primer a fundamental and pervasive shift away from nature- on nonmarket valuation . Kluwer, Dordrecht, The based recreation. Proceedings of the National Academy of Netherlands. Sciences of the USA, 105 , 2295–300. Greene, W. H. (2003). Econometric analysis, 5 t h e d n . P r e n t i c e - Phaneuf, D. (1999). A dual approach to modeling corner Hall, Upper Saddle River, NJ. solutions in recreation demand. Journal of Environmental Haab, T. C., and Hicks, R. L. (1997). Accounting for choice Economics and Management , 37 , 85–105. set endogeneity in random utility models of recreation Phaneuf, D., Kling, C., and Herriges, J. (2000). Estimation demand. Journal of Environmental Economics and and welfare calculations in a generalized corner solu- Management, 34 , 127–47. tion model with an application to recreation demand. Hill, G. W., and Courtney, P. R. (2006). Demand analysis Review of Economics and Statistics, 82 , 83–92. projections for recreational visits to countryside wood- Phaneuf, D., and Smith, V. K. (2005). Recreation demand lands in Great Britain. Forestry, 79 , 185–200. models. Handbook of Environmental Economics, 2 , Hulse D., Gregory, S., and Baker, J., Eds. (2002). Willamette 671–751. River Basin planning atlas: trajectories of environmental and Scarpa, R., Thiene, M., and Train, K. (forthcoming). Utility ecological change. Oregon State University Press, in willingness to pay space: a tool to address the con- Corvallis. founding random scale effects in destination choice to Loomis, J. (2005). Updated outdoor recreation use values the Alps. American Journal of Agricultural Economics , on national forests and other public lands. General Appendices. LIMITATIONS AND NEXT STEPS 205

Shrestha, R. K., and Loomis, J. B. (2001). Testing a Towner, J., and Wall, G. (1991). History and tourism. meta-analysis model for benef t transfer in interna- Annals of Tourism Research, 18 , 71–84. tional outdoor recreation. Ecological Economics, 39 , Train, K. (1998). Recreation demand models with taste 67–83. variation. Land Economics, 74 , 230–9. Swait, J., Adamowicz, W., and Van Bueren, M. (2004). Train, K. (2003). Discrete choice methods with simulation . Choice and temporal welfare impacts: Incorporating Cambridge University Press, Cambridge. history into discrete choice models. Journal of Von Haefen, R. H. (2008). Latent consideration sets and Environmental Economics and Management, 47 , 94–116. continuous demand systems. Environmental and Resource Timmins, C., and Murdock, J. (2007). A revealed prefer- Economics , 41 , 363–79. ence approach to the measurement of congestion in World Tourism Organization. (2008). UNWTO World travel cost models. Journal of Environmental Economics Tourism Barometer , 6 (2, June). Accessed at http://www. and Management, 53 , 230–49. unwto.org/facts/eng/barometer.htm . CHAPTER 12 Cultural services and non-use values

Kai M. A. Chan, Joshua Goldstein, Terre Satterf eld, Neil Hannahs, Kekuewa Kikiloi, Robin Naidoo, Nathan Vadeboncoeur, and Ulalia Woodside

dimensions of natural capital as scientif c, cultural, 12.1 Introduction historical, religious, and artistic educational bene- 12.1.1 Def ning cultural ecosystem services f ts or f ows. and non-use values More recently, the “cultural” class of value has come to encompass the concerns of indigenous In the ongoing effort to better def ne ecosystem peoples who might have a political and moral services (MA 2005; Boyd and Banzhaf 2007 ; Costanza right to a natural area, such as a Treaty-based right 2008 ; Fisher and Turner 2008; Wallace 2008 ) and to traditional territory. Hence, a number of studies their valuation, few classes of value have been more in the values literature have begun to def ne such diff cult to identify and measure than those con- intangible things as place value ( Basso 1996 ; nected with the cultural and non-use dimensions of N o r t o n a n d H a n n o n 1 9 9 7 ; B r o w n et al. 2 0 0 2 ) , s p i r - ecosystems. Whereas other ecosystem services itual value ( Milton 2002 ), heritage value ( Throsby make life possible in biophysical terms, cultural 2001 ), and social-relational value ( Lin 2001 ; Sable ecosystem services and non-use values inspire deep and Kling 2001 ). attachment in human communities. Accordingly, H e r e w e d e f ne cultural services inclusively as they need to be integrated into conservation and ecosystems’ contribution to the nonmaterial benef ts broader policy if societies are to sustain meaningful (e.g., capabilities and experiences) that arise from links between people and nature, and indeed many human-ecosystem relationships. F o r e x a m p l e , c u l - say if societies are to sustain themselves at all. tural services include the contribution of ecosys- Cultural services and their connected values tems to recreational experiences, to sense of place, have come to represent nonmaterial benef ts and to the knowledge that a valued nonhuman derived by those who use, might use, or share an species exists or will exist for future generations attachment to a [natural] area. Costanza et al. ( 1 9 9 7 ) (knowledge that has existence value or bequest def ned cultural values-cum-service as “aesthetic, value, respectively). For clarity, and to align this artistic, educational, spiritual and/or scientif c chapter with other chapters in this book, we adopt values of ecosystems” (p. 254). The Millennium the economics convention of differentiating values Ecosystem Assessment (2005 , p. 894) expanded this into use and non-use values. Use value refers to def nition to include the “nonmaterial benef ts peo- the direct (consumptive and non-consumptive) ple obtain from ecosystems through spiritual and indirect uses of ecosystem goods and services, enrichment, cognitive development, ref ection, rec- while non-use encompasses all values separate reation, and aesthetic experience, including, e.g., from use (Goulder and Kennedy 1997 ). Our treat- knowledge systems, social relations, and aesthetic ment of cultural ecosystem services includes both values.” Others focus more fully on the educational use and non-use values. Indeed, for all cultures, benef ts of cultural services. Chiesura and de Groot maintenance of cultural values requires the active (2003 ), for example, delineate key socio-cultural use of biological resources and nonmaterial

206 INTRODUCTION 207

interaction with important sites, among other tions such as naming and gifting ceremonies and direct and indirect uses. feasts wherein keystone foods are central (see Non-use values can be categorized into compo- also Turner et al. 2008 ). nent values, which often include existence value, The non-market nature of many cultural serv- bequest value, and option value. Krutilla (1967 ), an ices and non-use values creates another impor- economist, f rst introduced the concept of “exist- tant characteristic. Whereas many other services ence value” to capture the provision of “satisfac- are the production of things o r conditions that have tion from mere knowledge that part of wilderness value in relation to some market (supporting North America [ sic ] remains” regardless of any services do not produce marketed commodities, intention the valuing agent has of ever visiting but they may be essential steps in the production such destinations. More broadly, existence value of such commodities, such as the pollination of can be designated geographically as well as in ref- agricultural crops), cultural services and non-use erence to specif c biophysical units such as a par- values generally involve the production of experi- ticular species or ecosystem. While existence value ences that are valued without entering markets. In is specif c to an individual’s satisfaction, bequest such cases, the production and valuation are inti- value relates to knowing that an environmental mately linked, both occurring in part in the valu- amenity will be available for future generations. er’s mind. These services are therefore Option value represents “the premium that people co-produced by ecosystems and people in a are willing to pay to preserve an environmental deeper sense than other services. For example, if amenity, over and above the mean value (or we consider the contribution to aesthetic experi- expected value) of the use values anticipated from ences, there is no metric of production that does the amenity” (Goulder and Kennedy 1997 ; not involve the valuer. In other words, potential Chapter 2 ). While option value has a stronger con- supply and realized supply cannot be differenti- nection to use values, non-use values is the broad ated as they can be for market services. There is category capturing non-users’ preferences for the no objective way to claim that a site provides a continued existence of ecosystem goods and serv- great aesthetic service except by appealing to ices (Cicchetti and Wilde 1992 ). people’s behaviors and stated preferences regard- T h e p r o p e r t y o f intangibility i s c e n t r a l t o c u l - ing the object, which must generally be assessed tural ecosystem services and non-use values and for each site in turn. often renders them diff cult to classify and meas- While many assume that what matters gets ure. Classif cation and measurement of these val- measured (Meadows 2001 ), this is not so with cul- ues and services are nonetheless necessary as the tural services. Their consideration is long overdue. risks associated with their loss are central to pub- In the following sections, we lay a conceptual back- lic thinking and discontent with land manage- ground from which we identify components of ment decisions (Satter f eld and Roberts 2008 ). An cultural and non-use values and then link these analogous “intangibility” case in point for many values to ecosystem services. First Nations communities in Canada is enshrined in the concept “cultural keystone species” 12.1.2 From values to valuation: ( Garibaldi and Turner 2004 ). The term refers to methodological conundrums species, for example, salmon in the Pacif c Northwest, which, if diminished, result in not To the extent that the de f nition of cultural values just material loss of a provisioning service (the has been controversial, the same may be said for salmon), but a much larger suite of linked cul- methodological efforts to value them. While eco- tural losses or impacts including enduring prac- nomics continues to provide important conceptual tices. These may include food sharing and the and methodological approaches for ecosystem serv- social cohesion and alliances engendered by such ice valuation (e.g., Champ 2003; National Research exchanges; the transmission of traditional knowl- Council (US) 2005), there is another dialogue raising edge; and the maintenance of key cultural institu- concerns that the common approach of expressing 208 CULTURAL SERVICES AND NON-USE VALUES values in dollar metrics will result in the cultural of one lake versus f ve, in some cases because they and intangible dimensions of land management feel that protecting lakes is the “kind of thing they being improperly considered, or even left out alto- should support.” As such, these respondents offer a gether ( Rees 1998 ; Gowdy 2001 ; Wunder and Vargas dollar amount that is in fact a proxy for a donation 2 0 0 5 ; M c C a u l e y 2 0 0 6 ) . M o s t s c h o l a r s a g r e e t h a t a l l and not an expression of market value per se. dimensions of benef t should be recognized and val- In order to match relevant values to appropriate ued, but they differ sharply on the accuracy and eff - methods and to differentiate values pertaining to cacy of using a single currency, the dollar, to cultural services from those not related to services, appropriately measure multiple kinds of value we distinguish eight dimensions of cultural and ( Brown 1984 ; Lockwood 1998 ; Martinez-Alier et al. non-use values (Chan et al. , in prep.): 1 9 9 8 ; S a g o f f 1 9 9 8 ) . 1. Preferences versus principles versus virtues The need to match valuation methods to relevant (concern for ends versus means versus intent) kinds of value is best illustrated through the contro- 2. Market-mediated versus non-market-mediated versy over the validity of “stated preference” (derived from contribution to a marketed commod- approaches to valuation of ecosystem services ity versus to something valued intrinsically) (e.g., willingness to pay, WTP). For example, prob- 3. Self-oriented versus other-oriented (for one’s lems of validity have arisen with stated-preference own versus for others’ enjoyment) approaches aimed at dollar valuation, because often 4. Individual versus holistic/group (held by indi- survey respondents are expressing not a willing- viduals versus groups or larger wholes; an example ness to pay, but rather a willingness to contribute to of the latter is the community value of cultural a moral cause (Kahneman and Knetsch 2005 ). integrity) Opposing kinds of values are in operation in these 5. Physical versus metaphysical (stemming from judgments. Some kinds of value (virtues or princi- concrete physical experiences versus conceptual ples) violate the assumptions of evaluation meth- experiences; an example of the latter is existence ods for preference value. By virtues and principles, value) we mean moral values associated with a person’s 6 . S u p p o r t i n g v e r s u s f nal (valued for its contri- intent, duties, and rights—notions of rightness or bution to another value versus valued intrinsi- wrongness of people or actions themselves as cally; e.g., education might be valued for its opposed to their resulting consequences. Social sci- contribution to other values, while spiritual value entists have provided evidence for this claim by is intrinsically valuable; this distinction is also demonstrating that expressions of value are rooted broadly relevant to biophysical ecosystem servic- both in preferences (which address what a person es—for example, pollination services support the values or will “pay for” because it produces out- production of agricultural products, a f nal comes that benef t him or her) and in virtues and service) principles (which address what a person believes to 7. Transformative versus non-transformative (valu- be right for nature and society) ( Sagoff 1998 ). Two able because of its contribution to changes in values kinds of problems with WTP studies can be traced and perspectives versus valuable in relation to to the inf uence of principle-based values on peo- unchanging values/perspectives) ple’s responses to questions: “protest zeros” (which 8. Anthropocentric versus bio/ecocentric (for are often a rejection of dollars as the appropriate human beings versus for all living organisms/biotic metric) and what are known as scaling or scoping communities and ecological processes) problems (wherein respondents in a valuation survey are insensitive to quantity of the thing valued). Not all kinds of values above pertain to cultural In the former case, respondents resist the survey ecosystem services. From an ecosystem service format by entering a 0, often stating that they f nd perspective, values are a measure of the impor- the question inappropriate. In the latter case, tance of a thing or experience; but some values respondents do not distinguish between the value cannot be understood as such because they are INTRODUCTION 209 underlying ideals (virtues and principles, which one-to-one mapping from services to values is overlap closely with held values—Brown 1984 ). challenging, and sometimes inappropriate. Certain Others are not considered to be products of ecosys- ecosystem services, such as the provision of water, tem services because the services framework is can be distinguished in such a way that each service focused on people (so bio- and ecocentric values produces one value: e.g., one might distinguish the are excluded). Whereas the eight dimensions of provision of water for irrigation from the provision kinds of values are crucial for determining appro- of water for household use, calling each a separate priate kinds of valuation, it is also worth distin- ecosystem service with a separate use. With many guishing categories of values associated with cultural ecosystem services, a one-to-one mapping cultural ecosystem services and non-use dimen- is awkward because component values of activi- sions of ecosystems. These ten proposed general ties/objects generally cannot be properly separated categories of values are a way of organizing the in people’s minds. For example, the physical-suste- nature of reason that a thing or experience is valued nance value of salmon to First Nations cultures of (e.g., because it provides knowledge, or social cap- British Columbia cannot be effectively decoupled ital). Some of the values are neither strictly non- from the social and cultural values of the harvest use nor really “cultural” values (e.g., activity and the ceremonies that depend upon this cultural value); they are included here because they are keystone species ( Garibaldi and Turner 2004 ). crucial components of services in these categories Perhaps more importantly, disentangling co-occur- (e.g., recreation, subsistence). ring values can be deeply offensive to people and antithetical to their cultures. Another example of the dif f culty of separating 12.1.3 Distinguishing values from services values can be drawn from the cultural world of We distinguish values from services (the produc- Hawaiians. Kalo (taro), is referenced in the ancient tion of things of value) for two principal reasons cosmogonic history of the Hawaiian Islands not ( Table 12.2 ). First, values do not exist as entities only as a prominent staple crop, but also as emerg- for probing or characterizing as separate from ing from the body of Haloa-naka-lau-kapalili objects and activities; rather, they are merely one (Haloa-of-the-quivering-leaf), the f rst-born son of way in which we organize our ideas about moral- Wakea and Ho‘ohokukalani. Stillborn, his body ity and preference. Accordingly, they generally was returned to the earth and grew into the f rst require some concrete thing or activity to enable kalo plant in Hawai’i. Their second-born son, their elicitation (as in valuation): any attempt to named Haloa in honor of his older sibling, became ask people how much they value a landscape/ the progenitor from whom all kanaka maoli seascape for its contribution to knowledge is likely (indigenous Hawaiians) descend. These are not to be far too abstract to get useful and meaningful just cosmological narratives, but are equally ref ec- answers. In our typology, we propose that serv- tive of what, for lack of appropriate terms in ices be the contribution to those concrete activities English, ref ect Hawaiians’ genealogical or kin- that ground values and provide a forum for their like connections to both the human and nonhu- expression (or non-activities like contemplation, man world. Such meanings—inscribed in and in the case of metaphysical services). So why not inextricably linked to physical places or goods of simply stop at services—why do we need to iden- importance (in this case, kalo)—are crucial prod- tify values at all? This brings us to our second ucts of metaphysical ecosystem services. These reason. may be as important to Hawaiian people today as Second, individual cultural ecosystem services they were to ancestors who planted the f rst f elds simultaneously produce multiple intertwined val- of kalo over a hundred generations ago. They also ues, so any attempt to valuate or characterize serv- represent an important sense of responsibility to ices must account for these interdependencies (see all of one’s ancestors: cosmological (Haloa), Section 12.2.1 ). With many cultural services, a human, and nonhuman (e.g., kalo), and thus a 210 CULTURAL SERVICES AND NON-USE VALUES

Table 12.1 Categories of cultural ecosystem services and associated benef ts, and the site substitutability of each—with one possible mapping of ecosystem services to benef ts

Benef ts

Service Contribution to . . . (experiences) Place/heritage Place/heritage Activity Spiritual Inspiration Knowledge Existence, Bequest Option Social capital and cohesion Aesthetic Employment Identity Site Substitutability

Subsistence x x x x x x x x x varies Recreation x x x x x x x x x depends Education & research x x x x x x x x x x depends Artistic x x x x x x x varies ‘Ceremonial’ x x x x x x x varies Site substitutability low high varies depends high low high varies

S i t e s u b s t i t u t a b i l i t y i s low if the service or value is linked directly to particular places, and high i f n o t ; i t depends on whether there are clearly identi f able and under- standable qualitative differences in instances within a category (e.g., existence value may be site-substitutable for valued species, but not for sacred sites); or it varies if the logic of the variation is more complex (e.g., the provision of subsistence opportunities is not site-specif c for activity values, but it may be for place values). The contributions to employment and subsistence experiences are directly linked to provisioning services but the experiential portions are included here as a benef t that has not been (and likely would not be) captured effectively by strict market valuation. For example, the value of employment transcends its contribution to aggregate economic values: consider the intangible value of employment that allows a person to care for and sustain a relationship to a place or resources of cultural or spiritual signif cance.

need to care for this heritage plant as one would cause of the spiritual experience and the inspiration care for family. Reducing this complex, spiritual of photography). Valuation exercises must account kinship to an economic currency is tantamount to for these multiple benef ts. pricing one’s great-grandparent. Not only is the An example may help illustrate the point. In a measure diff cult to derive, the effort to derive it Hawaiian forest, trees are regarded as manifesta- could be construed as deeply offensive. tions or the embodiment of deities, such that they Table 12.1 illustrates that an individual service simultaneously provide use values and spiritual can provide multiple benef ts. While we agree that values. One source for this spiritual belief arises it would be simplest to map one service to one ben- from a perceptive grasp of the role the forest plays ef t, we do not see this being possible with cultural in sustaining hydrologic ecosystem services. services. Spiritual, inspiration, and place values are Hawaiians say, “ Hahai no ka ua i ka ulula`au.” Rains not products of single kinds of experiences; rather always follow the forest. Knowing this and honor- these benef ts are products of all manner of experi- ing the deities’ embodiments, Hawaiians hewed ences associated with ecosystems (including meta- only the trees that were needed and ensured that a physical contemplation). Rather than expecting that suff cient stand was left in place to perform a key we can partition separate “spiritual” and “inspira- function in the hydrological cycle. Sacred sites in tion” categories of ecosystem service (de Groot et al. Tibet also seem to be revered and protected in part 2005 ) and value each category separately, we must because of their contribution to ecosystem services recognize that the production of these benef ts is a such as f ood mitigation ( Box 12.1 ). deeply complex function of many activities and We distinguish between categories of benef ts components. For any service, each kind of benef t and services based on the potential for one site to denoted by an “x” may contribute to the values that substitute for another in production of the benef t/ people assign to the service’s pertinent activities service (Table 12.1 ). There is a difference between and components (e.g., a person may enjoy hiking the variation across sites and the possibility for sub- and berry-picking partly—and inseparably—be- stitution: for example, the provision of employment INTRODUCTION 211

Box 12.1 The sacred geography of Kawagebo

of ecosystem service provisioning for communities, such as Jianzhong Ma and Christine Tam prevention of mountain slope erosion or protection of Kawagebo Mountain, rising 6740 m along the eastern water resources. Indeed the ties between the belief system Tibetan Plateau, lies at the heart of a region dominated by and the provisioning of ecosystem services are strong. sacred geography at the headwaters of three of the great Ubiquitous local stories cite the importance of sacred sites rivers of Asia—the Yangtze, the Mekong, and the Salween. in securing human well-being. Adong village is one such Sacred natural sites abound across this 1600 km2 area, example. The f oodwaters of Zhili-Rongqu of Adenggong from springs to lakes, rocks to caves, trees to whole forests. inundated a large stretch of farmland and its households Inhabited for over 2000 years, the region gained stature on the lower reaches of the mountain during the 1990s, primarily in the thirteenth century when the living Buddhas while the main channel near Adong experienced three Karma Pakshi and Karma Rinpo Dorje paid formal respect f oods in the 1980s. The community organized all its to Kawagebo, leaving behind offerings of scriptures that resources for f ood prevention, but was unsuccessful. It was elevated the mountain to one of the most sacred in all the only when the sacred mountain was established and sealed Tibetan world ( Ma 2005 ). More recently, the surrounding from direct human activity that the f oods stopped. In mountains and valleys, dotted with small Tibetan villages, another example, the people in and around Waha Village, were included in the area declared a World Heritage site in who traditionally harvested the forest around their 2002 as the Three Parallel Rivers. This strength of culture headwater springs, suffered from a decrease in water based on sacred geography, in fact, lays the foundation and supply. In the end, it was after inviting the Living Buddha framework for protection of the extraordinarily rich Da-De to establish a sealed sacred mountain area biodiversity, critical life-support functions, stunning encompassing the springs that the water supply returned to landscapes, and cultural traditions that attracted UNESCO normal (KCA 2004). Biologically, sacred sites have also here in the f rst place. been found to encompass greater biodiversity than similar Tibetan sacred geography is an embodiment of the habitats at similar elevations that are not sacred ( Anderson integration of cultural sites and natural ecosystems, beliefs et al . 2005 ), suggesting the special protection powers that arising from a mixture of Tibetan Buddhism and the local fears of divine punishment can offer. Bon folk religion that attributed all natural things with Other sacred natural sites may include temple forests, spirit. In the Kawagebo region, sacred geography is lakes, waterfalls, and forests or grasslands near sacred expressed in four main forms: sacred mountains, “rigua,” marks, and number over 300 in the Kawagebo region. other sacred natural sites, and pilgrimage routes. Sacred Lastly, two pilgrimage routes circle Kawagebo, and host mountains, or “rida” in Tibetan, are the main form of thousands of devotees each year from across the Tibetan sacred geography, numbering over 70 in the region, and region to pray for good fortune as they circumambulate the are venerated by local clans, specif c communities, the sacred mountain God. broader region, and in some cases by all of Tibetan culture. Culturally, Tibetans draw three main functions from The sacred mountain of Kawagebo is actually comprised of these types of sites. They solidify their religious core, f ve major Gods that are the f ve highest peaks: Miancimu structure moral thinking and behavior, and strengthen (Goddess, 6 054 m), Jiewarena (Crests of Five Buddhas, 5 group identity (Ma 2005 ). Veneration of sacred mountains 470 m), Buqiongsongjiewuxie (Prince), Mabingzalawangdui originated from the awe and mystery surrounding the (Fight God for Subduing the Devil, 6 400 m), and ever-changing natural environment of Tibetans who “live Kawagebo itself (Major God, 6 740 m) (Xiroa 2001). at the roof of the world,” spawning a psychology of fear Reverence for these mountains includes restricting use of and respect for nature. These early mountain gods were resources on their slopes. elevated under Buddhism to symbols of Buddhist doctrine. “Rigua” are traditional zones established to seal off At the same time, sacred natural sites play a moral role by mountain areas especially for sustainable resource shaping f xed sets of rules to be obeyed related to the utilization, strictly controlling normal production activities. landscape and its resources, thereby stabilizing community Use restrictions range from prohibitions on tree felling and action and thought. Hillside zones or mountain tops have hunting, to limitations on herding and non-timber product designated rules of use, and those who violate the rules (NTP) collection. These zones often coincide with protection will be punished by divinities. Lastly, these sites strengthen continues 212 CULTURAL SERVICES AND NON-USE VALUES

Box 12.1 continued community cultural identity through common religion, modern economy. The key to successful development, values, beliefs, and living habits. Community activities and however, most likely lies in the sacred geography of the rituals surround these sites, resulting ultimately even in region, for it is this cultural and religious belief system institutionalization of commonly held boundaries, such as that sets out a protected area network much stronger in the delineation between national and community land in many ways than the formal nature reserve networks China, or the international borders between China-India or established by government entities. The framework for China-Nepal (Ma 2005 ). sustainable utilization through zoned restrictions and Most recently, vineyard conversion, infrastructure strict protection directed by sacred geography also helps construction, tourism expansion, and other associated to ensure the continued provisioning of ecosystem development have been intensifying in this area. With services to the millions of people living downstream of the declaration of the UNESCO World Heritage site, this area, not only in China, but in the neighboring plans have become even more ambitious to transform countries of Myanmar, Thailand, Vietnam, Laos, and this remote mountainous landscape into a thriving Cambodia.

opportunities might be high (temporarily) for for- by ecosystem change. Second, the principal purpose estry in one woodlot due to the old-growth forests of identifying the categories and kinds of values at that occur there; but there is high potential for sub- stake is to determine appropriate kinds of valuation stitution in theory because—for the purposes of and decision-making methods, and appropriate employment—the site could be replaced by other ways to apply and combine these ( Figure 12.1 ). sites. In the case of place value, in contrast, particu- Some of these methods cover only a subset of lar places are of value, so site substitutability is low. the kinds of values (e.g., market valuation and eco- In some cases, site substitutability depends on the nomic non-market valuation), while others address kind of opportunity provided: e.g., it depends on many or all kinds of values (e.g., structured deci- whether recreation is place-based because of some sion-making). site-specif c (e.g., historical) signif cance (low substitutability), or mainly because of a site’s aesthetic 12.1.4 Environmental and social processes beauty (high substitutability). Variation in site sub- that affect service production stitutability is less easily pinned down in cases such as the provision of artistic or ceremonial opportuni- The production of cultural ecosystem services is ties, which are often complex functions of site-spe- affected by many of the environmental processes cif c and other attributes. that affect other ecosystem services and biodiver- While researchers might benef t from an under- sity. Accordingly, principal environmental threats standing of the categories of ecosystem services at to service production are land-use change (includ- play, and the attendant categories and kinds of val- ing habitat loss and fragmentation), pollution, cli- ues, we do not recommend belaboring the applica- mate change, overharvesting, invasive species, and tion of any typology. Several points follow and disease. explain this position. First, these typologies are Because cultural ecosystem services are co-pro- intended for researchers and analysts, not for con- duced by ecosystems and people, they are deeply stituents and stakeholders. Those who are inter- affected by social processes—at both the production viewed do not need to know how their values are and valuation stages. For example, if social proc- being interpreted and organized in terms that have esses have led people away from nature-based rec- meaning principally in relation to the academic lit- reation ( Pergams and Zaradic 2006 ; Louv 2008 ), erature; what matters is that we as researchers can then these processes also affect the ecosystem con- organize them in ways that will help us characterize tribution to recreational experiences and the corre- the various ways that human values may be affected sponding value. METHODS: INTEGRATING CULTURAL SERVICES AND NON-USE VALUES INTO DECISIONS 213

Categories of Categories of Kinds of Values Ecosystem Services Benefits

Provisioning: Material Preferences Principles Virtues market goods, etc. Aesthetic Market Non-market Regulating Place/heritage

Activity Other- Supporting: Self-oriented processes; oriented organisms; Spiritual sites, habitats Inspiration Individual Holistic/group Subsistence Knowledge Outdoor Physical Metaphysical recreation Existence/bequest

Education & Option Final Supporting research

(nature-based) application of these notes for Social capital & cohesion Non- Artistic (nature- Transformative based) transformative Identity

‘Ceremonial’ and decision-making methods, Kinds of valuation Anthropo- (place-based) Bio-/Ecocentric Employment centric

Figure 12.1 The suggested use of the typologies of ecosystem services and values. First, identify the relevant categories of ecosystem services; map these services onto categories of benef ts, and the benef t categories onto kinds of values; use the kinds of values at stake to inform choice and application of valuation and decision-making methods, to ensure appropriate representation of the full range of relevant values and to avoid double-counting. The arrows linking subsistence to categories of values are only one example of a mapping of one service onto values (other mappings are certainly possible).

The role of social processes in the production of spatially across a landscape. Our approach for tier cultural ecosystem services allows positive feedback 1 and tier 2 models is to provide an array of options cycles in which service production fosters habits and for integrating these values into a comprehensive transfer of knowledge that in turn enhance service ecosystem services analysis with attention to link- production. The negative f ip side of this “positive” ages with models described in other chapters of this feedback is that external shocks that undermine the book, or their counterparts in the literature. social side of service production can cause “vicious Capturing the complex nature of cultural services cycles” that fuel long-term loss of knowledge and (or any service) is challenging. We focus on gen- practice (Turner and Turner 2008 ). For ecosystem erating summary information that decision-makers services in which human use enhances—rather than can use, with place-specif c knowledge, to support degrades—biological production, service produc- improved decision-making about changes in tion is vulnerable to an even more pernicious socio- ecosystems that affect cultural and non-use values. ecological spiral of lost use ( Figure 12.2 ). We hope that the approaches presented below will contribute to vibrant discussion about effective 12.2 Methods: integrating cultural mapping and valuation tools, while recognizing services and non-use values into that substantial work still lies ahead to cover the decisions full array of values described in 12.1.2 , as well as the needs of decision-makers in diverse socio-eco- In this section, we discuss methods for mapping logical contexts. Before discussing tier 1 and tier 2 and valuing cultural services and non-use values methods, we f rst examine a set of issues that 214 CULTURAL SERVICES AND NON-USE VALUES

First Nations depopulation

Í use Í TEK transfer

Í use Í ecosystem prodution

Í use Í interest (demographic) Í use Ecosystem Service Provision

Figure 12.2 The downward spiraling of ecosystem service production for cultural ecosystem services co-produced by ecosystems and people, and for which use practices actually enhance future ecosystem production. An example is the harvest of edible seaweed Porphyra abbottiae in coastal British Columbia, a practice initially reduced by the great depopulation of First Nations people by European colonization. While Porphyra provides a provisioning service, it is integrally linked to subsistence and ceremonial activities, among other cultural ecosystem services. Porphyra seems to be enhanced by harvest practices, which involve clearing competitors and herbivores, and spreading the reproductive gametes (although often not consciously) (White and Chan, in preparation). The harvest—as practice—depends on transmission of traditional ecological knowledge (TEK) from generation to generation, and TEK transfer in turn depends on use ( Turner and Turner 2008 ). There are similar feedback cycles with use and both ecosystem production and loss of interest, and there are also feedback cycles between loss of knowledge, ecosystem production, and interest (White and Chan, in preparation). Other external inf uences can be positive (increased access by speedboats) or negative (pulp mill and domestic sewage pollution, new regularly scheduled commitments like school and wage jobs that interfere with the harvest cycle) ( Turner and Turner 2008 ). cut across all projects seeking to map and value For example, both hiking (contribution to recrea- cultural services and non-use values. tional experiences) and certain jobs (contribution to employment) may have value partly for their contribution to physical activity. In this context, these serv- 12.2.1 Cross-cutting issues: interdependency, ices are partly substitutable: an individual’s need for double-counting, and trade-offs physical activity from recreation would be lessened T h e o v e r l a p o f b e n e f ts across services raises the by a job that entails considerable physical activity. In specter of interdependency and the double- counting either A or B, we must deal with the fact that some that can result from interdependency. These issues benef ts are supporting and some are f nal, and the have crucial implications for quantif cation and valu- same is true for services. For example, if we value the ation methods. If we wish to assess the impact of a contribution to subsistence activities, we must be management decision on cultural values through its aware that this service is important partly because it impacts on cultural ecosystem services, we have a supports the contribution to “ceremonial” activities choice. (A) If we attempt to evaluate impacts on each (e.g., by providing shellf sh, salmon, and herbs to benef t separately (e.g., spiritual value, identity First Nations people in coastal British Columbia— value), we are likely to encounter cognitive overload Garibaldi and Turner 2004 ). It would be double- or dissonance (by which the sheer number and com- counting to add the value of ceremonial activities to plexity of benef ts affected by multiple services will an inclusive value of subsistence activities, unless the overwhelm people’s abilities to consider each benef t benef ts are carefully parsed out, which may be separately, or the artif cial separation of apparently impossible in practice. linked benef ts will seem inappropriate or artif cial) These problems can be solved, though not easily. or other resistance (see earlier Section 12.1.3 ). (B) On For intangible values, more inclusive valuation the other hand, if we attempt to evaluate impacts by approaches should be favored over unconnected service, we encounter diff culties associated with valuation of separate services. Separate valuation the interdependency of services in producing value. will only be possible when experiences provided by METHODS: INTEGRATING CULTURAL SERVICES AND NON-USE VALUES INTO DECISIONS 215

N Site A

Island of O’ahu

Site B

Site C 0 1.25 2.5 5 Kilometers

Figure 12.3 Land use/land cover (LULC) map from the Hawai’i Gap Analysis Program (2006) for the planning region on the north shore of the Island of O’ahu, Hawai’i. See text for discussion of this planning region as an example of calculating the relative total landscape score of site quality. LULC categories are aggregated into three general classes: built areas (black), unbuilt areas dominated by non-native vegetation (dark gray), and unbuilt areas dominated by native vegetation (light gray). Dashed triangles denote location of the three hypothetical culturally important sites included in the calculation of the total landscape score. a service stem from entirely distinct activities (e.g., and spiritual values as separate end products will in the provision of experiences for photography, likely generate considerable resistance from subsist- exclusive photography trips are separate from hik- ence harvesters for whom the value of subsistence ing trips during which photos are taken). But we can gathering arises through the concurrent co-produc- expect interdependencies to be so pervasive that tion of multiple benef ts. Second, if end products are activities will rarely be truly separate from a value considered at a high level (e.g., overall well-being) perspective. Many ecosystem services act simulta- to avoid the inseparability problem, we are likely to neously as both a supporting and a f nal service. At encounter cognitive overload in people participat- the supporting level, they are almost never inde- ing in value elicitation exercises. However, if we can pendent of other services. For example, the provi- use information from value elicitation to relate peo- sion of experiences for subsistence gathering is a ple’s responses to high-level categories like overall supporting service in its contribution to ceremonial, well-being, then the Boyd and Banzhaf (2007 ) educational, and identity values, but it is a f nal approach is promising. service in its provision of medicine and nutrition. In addition to concerns about interdependency These values are interdependent. Boyd and Banzhaf and double-counting, some model users and stake- ( 2007 ) propose a framework for ecosystem services holders may be uncomfortable with the explicit that would solve double-counting problems by manner in which our simple model deals with recording only values of end products. Unfortunately, trade-offs in ecosystem service production. In par- this useful framework does not apply easily to the ticular, this concern may arise with communities for valuation stage for values that cannot be easily mon- which all aspects of the cultural landscape are etized, for two reasons. First is the problem of insep- sacred. For example, stakeholders may be uncom- arable end products. For instance, treating nutritional fortable with or unwilling to evaluate alternative 216 CULTURAL SERVICES AND NON-USE VALUES land-use change scenarios in which one cultural important to link the location being mapped to the value increases at the expense of another, or in specif c service(s) of interest. which carbon stock increases at the expense of Our own experiences suggest that there will be cultural values (e.g., through reforestation with wide variation in availability of even the most basic non-native species versus native species used for information to inform a spatial mapping exercise. ceremonial purposes). At the same time, being This poses a practical challenge but also strong aware of trade-offs and synergies is critical to under- motivation to compile this information into a GIS. standing how our management of ecosystems Doing so will require synthesizing information affects all values of concern. Being sensitive to this from diverse sources such as historical documents, issue is essential for ensuring that our model is used interviews with community elders, and public in a productive decision-support environment. agencies. In some cases, the location of sites may be sensitive information (e.g., sacred burial grounds), and care must be taken to respect the private nature 12.2.2 Tier 1 methods of such information. The f rst step in any modeling effort is to prioritize While mapping is a good f rst step, the ability to the cultural ecosystem services and non-use values detect and represent change in sites across scenarios for a project’s analysis. Any given modeling exer- is limited to discrete addition or deletion of features. cise is unlikely to have suff cient data or resources Extensions to mapping, as discussed below, and to consider all, or even a large fraction, of these where feasible given data and resource constraints, services and the benef ts at stake. It is critical to pri- could greatly improve the quality of information oritize those cultural dimensions most important to supporting tier 1 decision-making. people and likely to be impacted by ecosystem changes considered in modeled scenarios. 12.2.2.2 Relative scoring of site quality across In this context, we present approaches for map- the landscape ping and valuing from which users can choose The value of culturally important sites is often based upon project goals and data availability. We affected by ecosystem conditions and other site note again that nature-based recreation and tourism characteristics, such as the surrounding land use/ activities are covered separately in Chapter 11 . land cover (LULC), proximity to population centers, and legal rights to access, among others. For a 12.2.2.1 Mapping culturally important sites given cultural service, relevant factors could be The ecosystem services framework presented in integrated into a quantitative analysis that provides this book is inherently spatial, meaning that cultural information on the relative improvement or services and non-use values must be derived from deterioration of site quality in the context of land- spatial data and presented spatially to be effectively use change scenarios. This approach would require integrated with other service model outputs. The that stakeholders identify which characteristics most basic piece of information of possible use to inf uence site quality and by how much. For exam- decision-makers is a map of the location of sites or ple, in some cases, enhancing access to a site may be objects of cultural importance that are linked to eco- desirable (e.g., plant gathering area), while in other system features. There is a wide range of features cases limiting access would be desirable (e.g., that could be mapped such as sites of subsistence restricted ceremonial site). activities (e.g., hunting and f shing grounds), recre- To illustrate one approach for relative scoring of ational experiences (e.g., hiking or kayaking routes; site quality across a landscape, we consider a case these may already be covered by models in in which a landscape score of site quality (L ) is Chapter 11 ), educational and research experiences, (1) increasing in the number of sites (with f exibility, ceremonial signif cance, and signif cance for iden- if appropriate, to assign weightings of relative tity value (e.g., historical trails such as the Grease importance across sites), and (2) affected by the Trail in Canada or uplands and ocean access trails LULC classes located in a buffer area surrounding in Hawaii) or existence value. In each case, it is the site. In certain cases, such as gathering areas, it METHODS: INTEGRATING CULTURAL SERVICES AND NON-USE VALUES INTO DECISIONS 217 may also be important to consider (3) the physical primary landowner in this region, as well as the and legal accessibility of each site to users. State of Hawai’i’s largest private landowner (see A general functional form to assess site condi- Box in Chapter 14 for background information on tions across the landscape could be expressed as Kamehameha Schools). Kamehameha Schools follows: manages its diverse portfolio of lands to derive an overall balance of economic, educational, cultural,

Lfv= {}1 ,..., vN , (12.1) environmental, and community returns for current and future generations.

where L is a landscape score of site quality and vi is The region shown in Figure 12.3 covers approxi- a relative score of site i calculated as follows: mately 10,500 ha from mountain top to the sea, including approximately 800 ha of developed rural

vwSBgiiiii= ⋅⋅{}1 ,..., g in , (12.2) community lands along the coast, 3,600 hectares of agricultural lands further inland, and 6,100 ha of

where wi is the weighted importance of each site i ; S i rugged forested lands in the upper part of the is an indicator variable (0/1) representing the pres- watershed. Within this planning region, we have ence or absence of site i (in the current landscape, all identif ed three culturally important sites for incor- existing sites would be given a value of “1.” In sce- poration into the overall landscape score (L ): Site A narios, sites that are deleted would get a “0” and sites is near the coast, Site B is in the agricultural f elds that are added would be given a “1”); B i is the overall along a riparian corridor, and Site C is in the upper appropriateness score of surrounding land uses in forested conservation lands (Figure 12.3 ). These the buffer around site i , most simply calculated as sites are hypothetical but representative locations for the illustrative purpose of this example, though Bgii= ∑ j , (12.3) the approach would be identical with real sites. We j assume equal weights ( w i ) in this example suggest-

where g ij is the contribution of LULC class j to B i , ing that all sites are of equal importance from a calculated as decision-making perspective. Evaluating the impacts of land-use change

garij= ij(),⋅a j (12.4) through scenarios is an important feature of our modeling approach ( Chapter 3 ). The analysis

where a ij is the fractional area of each LULC class j described above would need to be rerun for each in the buffer area around site i ; r is the radius of the scenario, holding input values constant. If a site is buffer area (assuming a circular buffer; other shapes removed in a scenario, then this should be ref ected are possible); and α j is the appropriateness scores through a change in the indicator variable, S i . for each LULC class j to ref ect the cultural appro- Calculating the difference in the landscape score ( L ) priateness/value of the LULC class being located in for each scenario relative to the current landscape the site’s buffer area. provides model users with a quantitative indication All told, with the simplest versions of f and B of the impacts of land-use change on sites of cul-

(summations of v i and g ij respectively), we have the tural importance. In this example, a scenario that following: expands residential development within the buffer region of site A would decrease the relative score for Nn⎛⎞ LwSar() (12.5) this site (B ), as well as the overall landscape score = ∑∑ii⋅⋅⎜⎟ ij ⋅a j i ij==11⎝⎠ (L ). Conversely, investments in native plant restora- As a descriptive illustration of this approach, let’s tion around sites B and C would increase each site’s consider an example of computing a relative total score and the landscape score. landscape score (L ) for culturally important sites in This approach requires stakeholders to make a region on the north shore of the island of O’ahu, judgments regarding the appropriate size and Hawai’i (Figure 12.3 ). Kamehameha Schools, an shape of the buffer region, the relative appropriate- educational trust for Native Hawaiians, is the ness of different LULC classes contained in the 218 CULTURAL SERVICES AND NON-USE VALUES buffer, and potentially other factors. To provide number of people who can sustainably partici- more robust information to decision-makers, model pate in harvesting. users should perform analyses to determine the Hydrologic services including providing water sensitivity of the landscape score to different speci- for consumptive uses, regulation of water pollu- f cations of model inputs. Sensitivity analysis could tion, and mitigation of storm peak f ow (Chapters be straightforward, such as testing different buffer 5 – 7 ) a l s o i m p a c t c u l t u r a l s e r v i c e s a n d n o n - u s e v a l - radius distances. Where useful, it could also con- ues. Water f ows and their corresponding quality sider different ways of quantifying the impacts of affect the condition of aquatic cultural resources, land-use change in the buffer or redef ning the such as wild f sh or f sh cultivated using traditional nature of the buffer. For example, an alternative practices in f shponds. As such, changes in water buffer could contain two parts: a core area near the f ows could be used to quantify effects on cultural site and a doughnut-shaped area immediately sur- resources. Related to f sh consumption, such an rounding this core. analysis could evaluate the fraction of a target population that could be served by the local f sh 12.2.2.3 Linking cultural services with models for harvest (wild or cultivated) under different other ecosystem services and biodiversity management scenarios. In the context of mitigation Cultural services are produced, in part, by land- of storm peak f ow, model users could def ne scape features and processes quantif ed in other regions of interest based upon culturally signif cant models described in this book for biophysical landscape features. By narrowing their analysis to services and biodiversity. Agricultural models these regions, they could specif cally understand (Chapter 9 ) can provide information on the loca- changes in f ood risk to cultural assets across alter- tion, yield, and market value of culturally impor- native landscape scenarios. These examples illus- tant agricultural crops (e.g., taro in Hawai’i). trate some of the many linkages with cultural Pollination services (Chapter 10 ) support cultiva- services that can improve information supporting tion of insect-pollinated crops, and there may be decision-making. overlap between habitat supporting wild pollinators and areas of cultural signif cance (e.g., plant 12.2.2.4 Integrating demographic gathering areas). Identifying this overlap is impor- and socio-economic data tant for effective decision-making, as is an evalua- Information related to demographic and socioeco- tion of how the relative benef ts from pollination nomic conditions is another source of data for input and cultural services would change given alterna- in tier 1 models. It is obvious that different ethnic tive land-use change scenarios. groups, even when not part of an indigenous Timber and non-timber forest products culture, bring very different social and cultural (Chapter 8 ) are often connected with the contri- backgrounds to their relationships with nature bution to artistic and employment experiences, ( Box 12.2 ). Census data are likely to be broadly and with other cultural services. Output from available, and because this information is linked to these models could inform decisions about geographic location, it is readily applicable to spa- resource management and harvest levels for cul- tial ecosystem service analyses. The units of infor- turally important species. For example, decision- mation (e.g., household, census block, county, state), makers could use the non-timber forest products however, will limit analysis to specif c scales. As an model to obtain information on the location, illustration of integrating demographic and socioe- stock, and growth rate of species harvested for conomic data, we consider the value of subsistence cultural purposes (e.g., plants used for traditional activities to communities in Alaska. hula activities in Hawai’i) under different land- At the tier 1 level, it will often be diff cult or scape scenarios and management regimes. This impossible to distinguish between the various val- output could inform an assessment of the sustain- ues discussed in this chapter that result from cul- ability of varying levels of harvest in different tural activities such as subsistence harvest. At this gathering areas, which may in turn affect the level, it is therefore often desirable to calculate METHODS: INTEGRATING CULTURAL SERVICES AND NON-USE VALUES INTO DECISIONS 219 lower bound estimates of the importance of services degree to which people will be affected by changes to people. in subsistence harvest. For example, based on per- Subsistence hunting and gathering is of great capita economic values as a percentage of income, it importance to Alaskans, but it is of much greater is clear that the predominantly Alaskan Native importance to some than to others. Those who ben- community Hydaburg will be more severely ef t most from subsistence activities are often those affected by loss of subsistence harvests than would who earn the least in monetary terms, which other communities. Similarly, within each commu- implies that averaging the surveyed value of sub- nity, subsistence permit holders will be affected sistence activities across people will often under- much more than would others in the communities. represent the contribution to well-being. An Furthermore, assuming that the well-being impact example from f ve communities in Southeast of a foregone dollar of income is inversely propor- Alaska, demonstrates this. tional to a person’s income, the relative value of the Vadeboncoeur and Chan (in preparation) calcu- subsistence harvest to Hydaburg is 2.7 times greater lated a lower bound estimate of the economic (34% of local inclusive income) than it would have value of subsistence goods using publicly availa- seemed if we had used nation- or state-wide income ble harvest data for 48 species or groups of species statistics. in each community. The economic values for har- This analysis clearly underestimates the value of vested species were derived from market prices the subsistence harvest to Alaskans, but it does for the species in question or the cheapest availa- illustrate (i) how readily available data can be used ble local substitute meat. In each case, the lower to calculate lower bound estimates of these values bound valuation is supported by the logic that and (ii) how presenting these values in relation to a there is considerable trade and barter of such spe- group’s income can better track distributional con- cies, so if a harvest is not traded, this is likely cerns and provide a starkly different perspective on because the harvester values it more than the pos- the potential well-being impacts of losses of subsist- sible exchange value. Income and population data ence harvest. This analysis also serves as an exam- were obtained from the 2000 US Census ( US ple of the interrelatedness of the models presented Census Bureau 2003 ). throughout this book; the service of subsistence Even with a tier 1 model drawing upon existing harvest may also be relevant in analyses of the information sources, spatially resolved data paired relationship of ecosystem services to poverty (see with income data can provide insight into the Chapter 15 ).

Table 12.2 The economic value of subsistence harvest in f ve sites in Southeast Alaska relative to income

Jurisdiction Population Adults in Seasonal Per-capita Per-captia Economic Economic values of labor force occupations income value value of subs. harvest/income—for (%) (%) (thousands) of subs. harvest/ subs. permit holders harvest income (%) (%)

Sitka 8835 73.6 15.9 $23.9 $246 1.0 10 Petersburg 3224 70.8 24.6 $26.2 $368 1.4 7 Haines 2392 61.6 18.9 $22.3 $176 0.8 9 Yakutat 808 77.8 33.2 $23.0 $392 1.7 33 Hydaburg 382 49.1 21.1 $12.0 $602 5.0 34 United states 301 139 947 $30.5 Alaska 670 153 $32.0

Here, adults are considered to be all people over 16 years of age. Seasonal occupations include farming, f shing, forestry, resource extraction, construction, and maintenance. Economic values were calculated based on lower bound estimates of the values that could be obtained through reciprocal non-market exchange. Percentages of total income (including these subsistence values) are provided for average members of each community and—using more detailed harvest data—for those holding special subsistence permits. The low fraction of the population in the labor force demonstrates that in these communities time is not limiting; we assumed that time spent on subsistence harvesting is generally not detracting from time spent in gainful employment. 220 CULTURAL SERVICES AND NON-USE VALUES

12.2.2.5 General landscape composition that could be integrated into tier 2 analyses for cul- and conf guration tural services and non-use values. In most cases, Certain cultural services and intangible values may consideration of non-use values will need to occur be linked to general landscape characteristics, such at a tier 2 level, since collection of new, site-specif c as the composition or conf guration of LULC types data will be necessary. Collecting data for each serv- within a planning region. Aesthetic values in rural ice and value will require specif c methods. We areas, for example, may be linked with the amount describe below the process of a tier 2 analysis with or conf guration of open space in agricultural or for- emphasis on one value as an example. ested LULC types relative to built areas. This type of information can be quantif ed in a GIS using spa- 12.2.3.1 Mapping and quantifying existence value tial analytical tools applied to the LULC layer and We discuss a possible tier 2 approach to mapping other relevant datasets. With this example, if aggre- non-use values using existence value as an exam- gate values are suff cient for decision-makers, then ple. The few studies that have attempted to meas- several general values could be computed such as ure existence value have done so using stated the fractional area in different LULC types or the preference survey techniques such as contingent degree of fragmentation of open space. A more valuation or choice experiments ( Mitchell and detailed and information-rich approach would be Carson 1989 ; Louviere et al. 2000 ) to estimate will- to map viewsheds using three-dimensional projec- ingness-to-pay (WTP) for conservation programs tions in a GIS for locations important to stakehold- ( Kramer and Mercer 1997 ; Rolfe et al. 2000 ; Horton ers ( Bishop and Karadaglis 1997 ; Lim and Honjo et al. 2003 ). 2003 ; Grêt-Regamey et al. 2007 ), assessing how the Contingent valuation exercises ask survey composition of viewsheds changes under different respondents to express their WTP for a change in an scenarios of development. environmental good or service that is described in a This approach of quantifying general landscape hypothetical scenario; resultant WTPs are “contin- characteristics must be partnered with considera- gent” on the accurate description of the good being tion of cultural services that are highly site-specif c. valued. Choice experiments present scenarios While the site substitutability of forest cover in a def ned by quantitative or qualitative attributes region may be high with regards to aesthetic values, (generally including a price variable), and respond- it may be low with regards to its contribution to the ents are asked to choose the scenario they prefer. heritage, spiritual, identity, and other values associ- Both contingent valuation ( Kramer and Mercer ated with a sacred site. As such, loss of forest cover 1997 ; Horton et al. 2003 ) and choice experiments around the site would substantially reduce these ( Rolfe et al. 2000 ) have been used to estimate WTP site values, but this conversion, all else equal, would for the existence of tropical rainforests. have a lesser impact on the landscape’s aesthetic Evaluating existence value at f ne spatial scales value. involves eliciting values for the existence of species or ecosystems at particular geographic locations. This is problematic for several reasons: (1) At with- 12.2.3 Tier 2 methods in-country scales, individuals’ WTP for a number Model users may have resources available to sup- of environmental non-use values declines with port new, on-site data collection. This is important household distance from site of provision (Hanley for modeling cultural services and non-use values et al. 2 0 0 3 ; B a t e m a n et al. 2 0 0 6 ) ( F i g u r e 1 2 . 4 d ) ; ( 2 ) because existing data are likely to be scarce for cul- Individuals more familiar with a site will be better tural services in many locations. Indeed, one impor- able to disaggregate their existence value across tant advance at the tier 2 level will be to generate space, resulting in a more heterogeneous value sur- more ref ned, site-specif c data inputs for the gen- face than would come from individuals who are eral methods described above. In addition, using less familiar (Figure 12.4b); and (3) when values are tier 2 models for biophysical services and biodiver- elicited from an international group of respond- sity will also provide more sophisticated output ents, individuals in richer countries are less income- METHODS: INTEGRATING CULTURAL SERVICES AND NON-USE VALUES INTO DECISIONS 221

Box 12.2 People of color and love of nature

Like many individuals, I am multifaceted and my Hazel Wong appreciation of nature competes with other activities and For the f rst 10 years of my life, the outdoors was my passions. For example, I have driven to mountains, seeking playground. Growing up in the Seychelles, without nature’s solitude and the comfort of the vast open space to television, my sisters and I spent our vacation days and gain clarity and strength for major decisions in my life. Yet, Saturdays immersed in creative play outdoors. Sundays I have been camping only once in all my years, and I am were reserved for either a beach outing or a family lunch not an avid hiker—I would rather do a two-hour martini feast followed by seven people piling into a car to tour the lunch with friends than hike on most days. (I hope this island, making pit stops along the way for my parents to confession is not grounds for revoking my conservationist say hello to their friends and our extended family. I have card.) However, that does not negate my devotion and memories of my paternal grandmother watching fourteen passion to protect the natural world. grand children at once. The house, a modest four-bedroom Working in the conservation f eld for six years, I quickly home, was by no means big enough to accommodate that learned that the Anglo Saxon culture def nes the meaning many rambunctious children. Looking back, grand-mere of the word conservationist and narrowly interprets what had help, the outdoors. The only reasons to set foot inside constitutes the correct way of valuing and recreating in the the house were to eat lunch and use the bathroom. natural world. I have concluded that the problem is not On September 14, 1980, I woke up in the desert of Las that people of color do not value and care about Vegas, Nevada after my family left the Seychelles following a conservation issues. They are under represented in the coup d’état and ensuing unrest. At ten years old, I remember conservation movement because of conservation’s history feeling a tremendous loss for the beautiful island I left behind. of class, privilege, and a homogenous culture that has Slowly, the desert became my new playground. Back then, rendered a myopic view, perhaps a benign neglect and at there was plenty of desert in Las Vegas, the no-end-in-sight worst actual resistance to conservation becoming a more development boom that started in the late eighties had not inclusive movement. taken place. Our introduction to television, even in a limited Conservation’s past is a history of white male privilege, and supervised capacity, replaced playing outdoors, especially conserving nature for the pleasure and enjoyment of the during the winter months, a new and not so welcome season well-to-do class. In fact, conservation history books are for those born on an island with a year round average f lled with stories of how John Muir worked to save the temperature of 80°F. redwoods and later founded the Sierra Club, or how Life in the States was very different from the Seychelles; Theodore Roosevelt established our national park system. my parents had to adjust to a new culture, a lack of family While there is much to be proud of in the early support and raising f ve girls in an unfamiliar setting. Their conservation movement, there is also some disturbing number one priority at all times was our education, and history which rarely gets documented. When our national they worked and saved to ensure we had a head start in parks were founded, Native Americans were removed life. This left little time and resources to explore the natural from their homes and displaced as contaminators of world in America. Yet, my father had a voracious appetite pristine nature. Cities were held with disdain as the dark for nature books and television programs—fueling my polluted homes of industry, as well as society’s knowledge, curiosity, and sense of amazement. undesirable poor minorities. John Muir, long heralded as Years later, I am now a self identif ed conservationist and the father of the conservation movement, was also known one of the 10% of people of color working in the for his disparaging remarks about Native Americans and conservation f eld. I also served on the board of the Nevada African Americans. Native Americans were “dirty,” Conservation League for f ve years, serving as chair for over “savage,” and needed the wilderness to cure “the a year. My childhood exposure to the natural world, grossness of their lives.” African Americans “made a great embedded in my psyche a love and passion for the deal of noise and did little work.” While today’s environment, instilling in me a value system that f rmly conservationists do not hold these views of Native believes in the rights of nature. Americans and African Americans, the legacy is that

continues 222 CULTURAL SERVICES AND NON-USE VALUES

Box 12.2 continued conservation organizations’ appeal beyond the Anglo strategy just as ecosystem services is a strategy to connect Saxon culture has been limited. and drive home a key message with a key audience that If you look at the surface statistics, reasonable people nature has an economic value. Those of us who work in the may draw the conclusion that people of color do not value conservation arena value nature for its intrinsic worth, and conservation. Indeed, 90% of people working in that is not the case for the vast majority of Americans conservation are white. To many, this could be interpreted including many of our decision-makers. The existence of to mean that people of color have no interest in this f eld. ecosystem services as a f eld substantiates that fact. Furthermore, over 90% of members and donors to Ecosystem services is a strategy to get people to connect conservation organizations are white, again, leading many with nature who do not value nature for nature’s sake but to conclude that people of color do not support may value it for its economic benef ts. The purist in me conservation because they do not f nancially contribute. abhors the idea that the natural world must have a dollar Lastly, people of color do not value nature because they do value in order for us to protect it. However, given the lay of not recreate in nature as much, or in the same manner as land, where economics reigns supreme, ecosystem services the dominant culture. is pragmatic, strategic and will resonate with key The paradox, and what most conservationists are inf uencers in the political arena. More importantly it will unaware of, is that people of color consistently vote help us achieve our conservation goals. overwhelmingly in favor of conservation measures placed Along that same line, conservationists must take a on the ballot and often times they are taxing themselves to pro-active and strategic approach to ensure protecting our support public funding for conservation. Furthermore, natural world is an investment that everyone is on board qualitative and quantitative analysis through focus groups with politically, culturally and economically. It starts with and public opinion research surveys show that communities researching and developing strategic messages and outreach of color value the natural world just as much (and in some plans to target communities of color and bring them into the cases more) than whites and that they support policies to membership and donor base. Although all humans depend address global climate change. Protecting land, air and on the benef ts of a healthy ecosystem, delivering that water is a core value of communities of color. The challenge message with an emphasis on biodiversity and wilderness is to get these communities actively engaged in f nancing protection to communities of color simply will not work. and advocating for conservation. Van Jones, an advocate for environmental and social Today’s conservation organizations’ marketing and justice, points out that the standard gloom and doom outreach efforts continue to target the dominant group in messages about species extinction, polar bears and melting American society, a group of donors getting older and glaciers do not resonate with a wide swath of the narrower in terms of overall population numbers. In the population. As an alternative message, he is touting next three decades, communities of color will comprise ecosystem services as it relates to green jobs and the new 50% of the US population and the lack of cultivation of green economy. Yet, he is not calling it ecosystem diverse groups means potential new donors and members services—a term that only has resonance and meaning to are being neglected while nonprof ts compete with each a small number of insiders. other for limited private resources. Furthermore, a lack of In conclusion, ecosystem services can play a pivotal role in diversity in the work force continues to send a message of the outreach to communities of color. For conservation to be disconnect with communities of color, and ultimately leads relevant in the twenty-f rst century, we must frame our to conservation losses as the talents of diverse people are message beyond biodiversity protection to include nature’s not being harnessed. Unfortunately, the insular mode of benef ts to people such as green jobs, children’s health, hiring makes it that “who you know” is still the prevalent economic solutions, spirituality, quality of life issues such as hiring practice. When you start with a 90% homogenous clean air and clean water, and protecting places for families to culture—that “who you know” way of hiring contributes to gather. We need conservation messages to be more relevant a lack of meaningful outreach for diverse candidates. to all demographics, and more directly connected to people’s The rapid and profound demographic changes means everyday lives. We also need to ensure that all people “see” we, the conservation community, are already behind in our themselves and those like them working in conservation. outreach efforts to ensure a more diverse constituency base When we do that, we will f nd the common ground to build a is working to protect our natural resources. Therefore, movement that truly includes everyone. And the good news is, reaching out to communities to color should be a priority we do not have to change our mission. METHODS: INTEGRATING CULTURAL SERVICES AND NON-USE VALUES INTO DECISIONS 223 constrained than those in developing countries; preferences are formed through a process of repeti- therefore their monetized values for the existence tion and experience with unfamiliar goods (Bateman of biodiversity will be higher, all else equal et al. 2008 ), such as ecosystem services. Recent work (Figure 12.4c). Designing surveys to elicit compara- has also demonstrated that the results of stated ble WTP estimates across large distance gradients preference surveys can vary depending on whether that span multiple countries, languages, and individual valuation or deliberative, citizen-jury cultures is challenging. Yet without such surveys, valuation techniques are used (Alvarez-Farizo and the empirical data necessary to produce spatially- Hanley 2006 ), which we would expect given that explicit existence value estimates will remain the latter emphasizes different kinds of value (e.g., elusive. principle-based, group). Choice experiments and contingent valuation This latter distinction may be particularly rele- have well known limitations, especially for ecosys- vant when trying to estimate existence values across tem services that are unfamiliar to respondents and geographic scales. For example, conservation of that vary spatially. Recent research on “learning threatened Amazonian forests may be important design” contingent valuation may provide an to both local and international non-users. While appropriate methodology for situations in which international non-users are likely to hold existence

(a) (b)

Dist from landscape

Figure 12.4 Variation in existence value across landscapes and stakeholder groups. (a) Schematic of a conservation landscape in a developing country; dark gray units indicate tropical forest patches, light gray indicates anthropogenic land use. (b) Mean (proportional to height) WTP for conservation of patches from respondents living near the landscape. Variance in WTP is high as respondents are knowledgeable regarding the landscape and can discriminate among patches. (c) Mean WTP for conservation of patches from respondents living far from the landscape, in a developed country. Mean WTP is higher due to lower income constraints, and variance is lower due to less detailed knowledge of the landscape, as compared to (b). (d) Relationship between WTP for patches in the landscape (mean and variance) and distance of respondents to the landscape. 224 CULTURAL SERVICES AND NON-USE VALUES values independent of other members of their com- surveys drawn from the social sciences. Especially munity, locals may see forests as a community asset. important, are techniques drawn from sociology As such, a deliberative group valuation approach ( Dunlap et al. 2000 ), human ecology ( Dietz et al. may be the most appropriate to use when multiple 2005 ), psychology ( Gregory et al. 1993 ), anthro- perspectives like these exist, rather than an individ- pology (Kempton et al. 1995 ; Satterf eld 2001 ), ual survey ( Spash 2007 ). multi-attribute utility theory (Russell et al. 2001 ), Bearing in mind all of these issues and constraints, and legal methods designed for the valuation of a program to evaluate the existence value of a par- loss ( Rutherford et al. 1998 ). ticular biophysical feature across a landscape (e.g., Ultimately newer practices must address above- tropical forests, endangered species) might follow mentioned problems of human judgment (e.g., the the following steps: instability in WTP judgments based on framing effects), and yet also achieve a balance between the 1. Identify the group or groups of people whose benef ts of group deliberation (collective and demo- existence values one wishes to elicit. cratic conversations about what matters); the recog- 2. Consider the service being valued and use a vari- nition and accommodation of political factors (justice, ety of methods (focus groups, literature reviews, equity, land tenure, Treaty rights, etc.); and the kinds expert knowledge, etc.) to identify the features of of quantitative and qualitative analyses necessary to the good that people care about. arrive at robust and defensible valuations. Good 3. Using similar methods, identify appropriate units practice with regard to analysis ensures that the val- of the service on the landscape. These units are liable uations assigned ref ect the full range of values perti- to be larger/coarser for individuals who have little nent to the case and are thereafter expressed as knowledge of the area. ordinal or cardinal scales amenable to trade-offs. It 4. Characterize each of the landscape units with also requires that the constituents and/or the rights regard to the important features identified in of different stakeholders are well def ned thereby step 2. signaling important political concerns. 5. Using maps and detailed descriptions, construct For these reasons, many valuation methods are a survey that will elicit people’s WTP for the exist- moving away from the aggregation of willingness- ence of the feature across the landscape. We do not to-pay judgments and toward deliberative proc- dwell here on the multitude of issues associated esses whereby such values are arrived at collectively. with constructing a stated preference survey; good Often know as “deliberative monetary valuation” references are Mitchell and Carson (1989 ) and (DMV), the intent is to arrive at a “social willingness Louviere et al. ( 2000 ). to pay.” In so doing, new classif cations of value and 6. Summarize WTP values for each particular concomitant valuation practices have arisen (Spash unit of the service being valued across the land- 2 0 0 7, 2 0 0 8) . T h e s e i n c l u d e v a l u e s a n d v a l u a t i o n scape; inference can then be made about WTP based on exchange value, charitable contributions, across the population that the sampled group and prices negotiated in reference to equity or represents. linked social values. Others have moved toward the 7. Modeling approaches ( Mitchell and Carson 1989 ; valuation of loss (for services since extinguished or Louviere et al. 2000 ) can be used to explain variation degraded) by combining conjoint analyses (simple in WTP across both individuals and sites; this may paired comparisons) and damage schedules (see, be useful for predictive purposes in other contexts. for example, Chuenpagdee et al. 2 0 0 1 ) . A l s o p r o m i s - ing is the application of multi-attribute utility theory 12.2.3.2 Mapping and quantifying other values to non-market goods including the use of swing Many of the problems addressed in assigning meas- weighting (McDaniels and Trousdale 2005 ). ures to existence value apply, equally, to the larger Finally, practices derived from structured decision- class of cultural ecosystem services and values making are particularly likely to be useful to the articulated here. While WTP methods continue to more intangible dimensions of cultural services be employed, so too are expressed preference (Gregory et al. 1 9 9 3 ; G r e g o r y 1 9 9 9 ) . T h i s i s d u e t o LIMITATIONS AND NEXT STEPS 225 the practice of constructive scaling, which assumes produced must be integrated into broader decision- that a locally def ned valuation metric (achieved making processes considered legitimate by its par- through deliberation) is optimal when natural or ticipants and stakeholders. proxy scales do not exist ( Keeney and Gregory Citizen juries and multi-criteria decision-making 2005 ) and when there are many different sub- (MCDM) processes implicitly or explicitly include components of a singularly important value. Com- such process-oriented values, so they are possible pelling examples of this method can be found in alternatives (Chee 2004 ) or addenda to valuation of work with First Nation/Native American commu- ecosystem services. Both approaches involve gath- nities in Canada ( Gregory et al. 2008 ) and in ering groups of citizens intended to represent the environmental decision-making more broadly spectrum of stakeholders affected by an issue, to ( O’Neill and Spash 2000 ). carefully consider a body of evidence and collectively make decisions. They differ in that citizen 12.3 Limitations and next steps juries involve competing teams of experts presenting alternative viewpoints ( Shrader-Frechette 1985 ; We have provided a framework for understanding, Coote and Lenaghan 1997 ), and MCDM involves categorizing, mapping, and valuing cultural ecosys- explicit structuring of the multiple dimensions rel- tem services and non-use values. This contribution evant to the decision ( Saaty 1996; Munda 2004 ). supports the notion that it is possible to quantify Both approaches could be undertaken based on many of these services and their corresponding val- (1) solely biophysical attributes, (2) biophysical ues through spatial landscape analysis that facili- attributes and a subset of valuation metrics (e.g., tates integration with output from other biophysical from market values), or (3) biophysical attributes and economic service models. Such integration is and a full set of valuation results. essential for ensuring that cultural services and non- There are pros and cons to these various possibili- use values are considered on equal footing with ties. On one hand, defensible valuation of the full other services and biodiversity in land management set of ecosystem services at stake is likely to be time- and policy decisions. At the same time, although we consuming and diff cult (e.g., to avoid double- have presented a typology and categorization of val- counting and account for interdependencies). Not ues and services, the methods discussed above do only is value elicitation time-consuming in its own not cover all of these values and services (Table 12.2 ). right, but in order to have accurate results, these Expanding the set of mapping and valuation tools valuation exercises should be performed using real- represents an important frontier for research and istic scenarios of outcomes from decision options— conservation practice; but as we have discussed, ideally, therefore, valuation will follow biophysical there are several obstacles to comprehensive valua- analysis and not run in parallel. Furthermore, tion including interdependency between values and explicit measurement of some intangible or sacred services and associated double-counting. values will always trigger discomfort, whereas this Modeling efforts explore ways to assign value to discomfort may be lessened considerably if the the cultural and non-use dimensions of ecosystems. decision is made without prices or similar metrics For many stakeholders, an equally important (as in citizen juries and MCDM). consideration will be people’s preferences regard- On the other hand, citizen juries and MCDM are ing the processes by which decisions are made. This also time-consuming and diff cult, and they, too may be particularly true in the context of cultural must follow biophysical analysis. Furthermore, services given the personal and community ties that they require that participants integrate in their they evoke. People’s preferences or principles stem minds the multitude of diverse values that will be from their underlying or held values ( Brown 1984 ), affected by management scenarios, often indi- such as fairness, responsibility, autonomy, and rectly—a considerable cognitive challenge. Finally, sovereignty. For our proposed models to be effec- they involve giving a certain authority to the small tive tools guiding land-use decisions, the subset of stakeholders invited to the decision- biophysical, economic, and non-economic values making table, authority likely to be questioned by 226 CULTURAL SERVICES AND NON-USE VALUES parties unhappy with the resulting decision, per- Chee, Y. E. (2004). An ecological perspective on the valua- haps even if considerable time is invested in choos- tion of ecosystem services. Biological Conservation , 120 , ing and vetting stakeholders involved. 549–65. The most thorough option (3), which can capture Chiesura, A., and de Groot, R. (2003). Critical natural Ecological Economics, the full set of values and consider them in appropri- capital: a socio-culural perspective. 44 , 219–31. ate context through democratic, deliberative proc- Chuenpagdee, R., Knetsch, J. L., and Brown, T. C. (2001). esses, is likely the most time-consuming and Coastal management using public judgments, impor- expensive. Fair, inclusive, and enlightened environ- tance scales, and predetermined schedule. Coastal mental decision-making is the Holy Grail. We hope Management , 29 , 253–70. this chapter has provided insights and approaches Cicchetti, C. J., and Wilde, L. L. (1992). Uniqueness, irre- that move us partway toward this goal. versibility, and the theory of nonuse values. American Journal of Agricultural Economics, 74 , 1121–5. Coote, A., and Lenaghan, J. (1997). Citizens’ juries: theory References into practice. Institute for Public Policy Research, London. Alvarez-Farizo, B., and Hanley, N. (2006). Improving the Costanza, R. (2008). Ecosystem services: Multiple classif - process of valuing non-market benef ts: Combining citi- cation systems are needed. Biological Conservation, 141 , zens’ juries with choice modelling. Land Economics , 82 , 350–2. 465–78. Costanza, R., d’Arge, R., de Groot, R., et al. (1997). The Anderson, D., Salick, J., Moseley, R., et al . (2005). value of the world’s ecosystem services and natural Conserving the Sacred Medicine Mountains: a vegeta- capital. Nature , 387 , 253–60. tion analysis of Tibetan sacred sites in Northwest de Groot, R., Ramakrishnan, P. S., v. d. Berg, A., et al . (2005). Yunnan. Biodiversity and Conservation, 14 , 3065–91. Cultural and amenity services. in Millennium Ecosystem Basso, K. H. (1996). Wisdom sits in places: landscape and lan- Assessment, Ed., Ecosystems and human well-being: guage among the Western Apache. University of New current status and trends. Island Press, Washington, DC, Mexico Press, Albuquerque. pp. 455–76. Bateman, I. J., Burgess, D., Hutchinson, W. G., et al . (2008). Dietz, T., Fitzgerald, A., and Shwom, R. (2005). Learning design contingent valuation (LDCV): NOAA Environmental values. Annual Review of Environment guidelines, preference learning and coherent arbitrari- and Resources , 30 , 335–72. ness. Journal of Environmental Economics and Management, Dunlap, R. E., Van Liere, K. D., Mertig, A. G., et al . (2000). 55 , 127–41. Measuring endorsement of the new ecological paradigm: Bateman, I. J., Day, B. H., Georgiou, S., et al . (2006). The A revised NEP scale. Journal of Social Issues 56 , 425–42. aggregation of environmental benef t values: Welfare Fisher, B., and Turner, R. K. (2008). Ecosystem services: measures, distance decay and total WTP. Ecological classif cation for valuation. Biological Conservation , 141 , Economics , 60 , 450–60. 1167–9. Bishop, I. D., and Karadaglis, C. (1997). Linking modelling Garibaldi, A., and Turner, N. (2004). Cultural keystone and visualisation for natural resources management. species: Implications for ecological conservation and Environment and Planning B: Planning and Design , 24 , restoration. Ecology and Society , 9 (3), article 1. 345–58. Goulder, L. H., and Kennedy, D. (1997). Valuing ecosystem Boyd, J., and Banzhaf, S. (2007). What are ecosystem services: philosophical bases and empirical methods. services? The need for standardized environmental In G. C. Daily, Ed., Nature’s services: societal dependence accounting units. Ecological Economics, 63 , 616–26. on natural ecosystems . Island Press, Washington, DC, Brown, G. G., Reed, P., and Harris, C. C. (2002). Testing a pp. 23–47. place-based theory for environmental evaluation: an Gowdy, J. M. (2001). The monetary valuation of biodiver- Alaska case study. Applied Geography , 22 , 49–76. sity: Promises, pitfalls, and rays of hope. In V. C. Brown, T. C. (1984). The concept of value in resource allo- Hollowell, Ed., Managing human-dominated ecosystems: cation. Land Economics, 60 , 231–46. proceedings of the symposium at the Missouri Botanical Champ, P. A., Boyle, K. J., and Brown, T. C., Ed. (2003). Garden, St. Louis, Missouri, 26–29 March 1998, pp. 141–9. A primer on nonmarket valuation: the economics of non- Missouri Botanical Garden Press, St. Louis. market goods and resources. Kluwer Academic, Dordrecht, Gregory, R. (1999). Identifying environmental values. In The Netherlands. V. H. Dale and M. R. English, Eds., Tools to aid environ- LIMITATIONS AND NEXT STEPS 227

mental decision making. Springer-Verlag, New York, Ma, J. (2005). Sacred natural sites and conservation in the pp. 32–58. Meili area. In J. Ma and J. Chen, Eds., Tibetan Culture and Gregory, R., Lichtenstein, S. and Slovic, P. (1993). Valuing Biodiversity Conservation. Yunnan Nationality Press, environmental resources: A constructive approach. China, pp. 33–40. Journal of Risk and Uncertainty , 7 , 177–97. McCauley, D. J. (2006). Selling out on nature. Nature, 443 , Gregory, R., Failing, L., and Harstone, M. (2008). 27–8. Meaningful resource consultations with f rst peoples: McDaniels, T. L., and Trousdale, W. (2005). Resource com- notes from British Columbia. Environment , 50 , 34–45. pensation and negotiation support in an aboriginal con- Grêt-Regamey, A., Bishop, I. D., and Bebi, P. (2007). Predicting text: Using community-based multi-attribute analysis the scenic beauty value of mapped landscape changes in a to evaluate non-market losses. Ecological Economics , 55 , mountainous region through the use of GIS. Environment 173–86. and Planning B: Planning and Design , 34 , 50–67. Martinez-Alier, J., Munda, G., and O’Neill, J. (1998). Weak Hanley, N., Schlapfer, F., and Spurgeon, J. (2003). Agg- comparability of values as a foundation for ecological regating the benef ts of environmental improvements: economics. Ecological Economics , 26 , 277–86. distance-decay functions for use and non-use values. Meadows, D. H. (2001). Dancing with systems. Whole Earth Journal of Environmental Management, 68 , 297–304. (Winter). Horton, B., Colarullo, G., Bateman, I. J., et al . (2003). Millennium Ecosystem Assessment. (2005). Ecosystems and Evaluating non-user willingness to pay for a large-scale human well-being: synthesis . Island Press, Washington, conservation programme in Amazonia: a UK/Italian DC. contingent valuation study. Environmental Conservation, Milton, K. (2002). Loving nature: towards an ecology of emo- 30 , 139–46. tion . Routledge, London. Kahneman, D. and Knetsch, J. L. (2005). Valuing public Mitchell, R. C., and Carson, R. T. (1989). Using surveys to goods: The purchase of moral satisfaction. In L. Kalof value public goods: the contingent valuation method , 3rd and T. Satterf eld, Eds., The Earthscan Reader in environ- edn. Resources for the Future, Washington, DC. mental values. Earthscan, Sterling, pp. 229–43. Munda, G. (2004). Social multi-criteria evaluation: Kawagebo Culture Association (KCA). (2004). Report on Methodological foundations and operational conse- Sacred Sites in the Kawagebo Area. Submitted to The quences. European Journal of Operational Research , 158 , Nature Conservancy China Program. 662–77. Keeney, R. L., and Gregory, R. S. (2005). Selecting attributes National Research Council (US). Committee on Assessing to measure the achievement of objectives. Operations and Valuing the Services of Aquatic and Related Research, 53 , 1–11. Terrestrial Ecosystems. (2005). Valuing ecosystem services: Kempton, W., Boster, J. S., and Hartley, J. A. (1995). toward better environmental decision-making . National Environmental values in American culture. MIT Press, Research Council, Washington, DC. Cambridge, MA. Norton, B. G., and Hannon, B. (1997). Environmental Kramer, R. A., and Mercer, D. E. (1997). Valuing a global values: A place-based theory. Environmental Ethics , 19 , environmental good: US residents’ willingness to pay to 227–45. protect tropical rain forests. Land Economics , 73 , 1 9 6 – 2 1 0 . O’Neill, J., and Spash, C. L. (2000). Conceptions of value in Krutilla, J. V. (1967). Conservation reconsidered. American environmental decision-making. Environmental Values , Economic Review , 57 , 777–86. 9 , 521–35. Lim, E.-M., and Honjo, T. (2003). Three-dimensional visu- Pergams, O. R. W., and Zaradic, P. A. (2006). Is love of alization forest of landscapes by VRML. Landscape and nature in the US becoming love of electronic media? Urban Planning , 63 , 175–86. 16-year downtrend in national park visits explained by Lin, N. (2001). Social capital: a theory of social structure and watching movies, playing video games, internet use, action . Cambridge University Press, Cambridge. and oil prices. Journal of Environmental Management , 80 , Lockwood, M. (1998). Integrated value assessment using 387–93. paired comparisons. Ecological Economics, 25 , 73–87. Rees, W. E. (1998). How should a parasite value its host? Louv, R. (2008). Last child in the woods: saving our children Ecological Economics, 25 , 49–52. from nature-def cit disorder . Updated and expanded edition . Rolfe, J., Bennett, J. and Louviere, J. (2000). Choice model- Algonquin Books, Chapel Hill. ling and its potential application to tropical rainforest Louviere, J. J., Hensher, D. A., and Swait, J. D. (2000). Stated preservation. Ecological Economics , 35 , 289–302. choice methods: analysis and applications. Cambridge Russell, C., Dale, V., Lee, J. S., et al. (2001). Experimenting University Press, Cambridge. with multi-attribute utility survey methods in a multi- 228 CULTURAL SERVICES AND NON-USE VALUES

dimensional valuation problem. Ecological Economics , technology assessment and environmental impact analysis. 36 . 87–108. Springer, Dordrecht, The Netherlands. Rutherford, M. B., Knetsch, J. L., and Brown, T. C. (1998). Spash, C. L. (2007). Deliberative monetary valuation Assessing environmental losses: judgments of impor- (DMV): Issues in combining economic and political tance and damage schedules. Harvard Environmental processes to value environmental change. Ecological Law Review, 22 , 51–101. Economics , 63 , 690–9. Saaty, T. L. (1996). Multicriteria decision making: the analytic Spash, C. L. (2008). Deliberative monetary valuation and hierarchy process: planning, priority setting, resource alloca- the evidence for a new value theory. Land Economics , 84 , tion , 2nd edn. RWS Publications, Pittsburgh. 469–88. Sable, K., and Kling, R. (2001). The double public good: Throsby, D. (2001). Economics and culture. Cambridge A conceptual framework for “shared experience” values University Press, Cambridge. associated with heritage conservation. Journal of Cultural Turner, N. J., Gregory, R., Brooks, C., et al . (2008). From Economics, 25 , 77–89. invisibility to transparency: identifying the implica- Sagoff, M. (1998). Aggregation and deliberation in valuing tions. Ecology and Society , 13 (2), 7. environmental public goods: A look beyond contingent Turner, N. J., and Turner, K. L. (2008). “Where our women pricing. Ecological Economics , 24 , 213–30. used to get the food”: cumulative effects and loss of eth- Satterf eld, T. (2001). In search of value literacy: sugges- nobotanical knowledge and practice; case study from tions for the elicitation of environmental values. coastal British Columbia. Botany-Botanique , 86 , 103–15. Environmental Values , 10 , 331–59. US Census Bureau. (2003). U.S. Census 2000 . Satterf eld, T., and Roberts, M. (2008). Incommensurate Wallace, K. J. (2008). Ecosystem services: Multiple clas- risks and the regulator’s dilemma: considering culture sif cations or confusion? Biological Conservation , 141 , in the governance of genetically modif ed organisms. 353–4. New Genetics and Society , 27 , 201–16. Wunder, S., and Vargas, M. T. (2005). Beyond “markets”: Shrader-Frechette, K. S. (1985). Science policy, ethics, and Why terminology matters. Ecosystem Marketplace. The economic methodology of social science: some problems of Katoomba Group. CHAPTER 13 Terrestrial biodiversity

Erik Nelson, D. Richard Cameron, James Regetz, Stephen Polasky, and Gretchen C. Daily

13.1 Introduction Because the tier 1 biodiversity model uses data that are available virtually everywhere in the world and Biodiversity is the variety of life at all levels of empirical data on the status of rare, endemic, and organization, from genetically distinct populations other species of conservation concern are unavailable to species, habitats, ecosystems, and biomes (Leopold for many places, a habitat analysis, instead of a 1 9 4 9 ; W i l s o n 1 9 9 2 ) . W h i l e b i o d i v e r s i t y i n f uences species-based approach is commonly implemented as the provision of all ecosystem services, and is the the f rst phase of a conservation assessment. basis for many (Sekercioglu et al . 2 0 0 4 ; D í a z et al . In tier 2 we assume data on potential distribu- 2005), it also inspires conservation for its own sake tions or ranges of species and on habitat suitability (e.g., Ehrlich and Ehrlich 1982 , Chapter 2 ). To protect (as gauged by breeding and foraging activity) are what remains of declining biodiversity ( Hughes available to us. We present two tier 2 models that et al . 1 9 9 7 ; D i r z o a n d R a v e n 2 0 0 3 ; W o r m et al . 2 0 0 6 ) rely on this species-specif c data. The f rst model conservationists seek to identify habitat conserva- combines these data to calculate the relative contri- tion networks that maximize habitat or species per- bution of a parcel’s habitat to the overall quantity of sistence. The design of these networks typically uses suitable habitat across a landscape or region. This species distribution maps (e.g., Ceballos et al . 2 0 0 5 ) approach is similar to deductive species distribu- and an understanding of the factors that affect spe- tion modeling (Stoms et al . 1992 ), the rarity-weighted cies presence and persistence on the landscape (e.g., richness methodology (Williams et al . 1996 ), and the Sekercioglu et al. 2007). More recently, the design of Biological Intactness Index (BII) system (Scholes conservation networks have considered the f nan- and Biggs 2005 ). The second tier 2 model aggregates cial and opportunity cost of network conservation information on species distributions and habitat and implementation ( Ando et al . 1 9 9 8 ; W i l s o n et al . suitability into a single landscape score. This score 2 0 0 7 ; P o l a s k y et al . 2 0 0 8 ) . S y s t e m a t i c c o n s e r v a t i o n is derived using species–area relationships to trans- planning (SCP) ( Margules and Pressey 2000 ) mar- late habitat area into a measure of landscape-wide ries many of these network design principles and biodiversity status, based on work by Sala et al . conservation organizations implement SCP princi- ( 2005 ) and Pereira and Daily ( 2006 ). In addition, we ples in their work ( Groves et al. 2 0 0 2 ) . can incorporate tier 1 output into tier 2 modeling in We present three relatively simple biodiversity order to weight suitable habitat area by quality. models that spring from this conservation planning Such quality weighting of habitat is common in literature. The models are straightforward to imple- algorithms that are used to select networks of areas ment and are designed for analysis at landscape scales. for conservation (e.g., Schill and Raber 2008 ). Our simplest model, tier 1, combines basic information about land cover and threats to biodiversity to 13.2 Tier 1: habitat-quality and produce habitat-quality and habitat rarity maps. In rarity model tier 1 we assume that protection of a variety of high- quality habitats will confer protection to their compo- Habitat-quality depends on its proximity to human nent species and populations ( Groves et al . 2002 ). land uses and the intensity of these land uses.

229 230 TERRESTRIAL BIODIVERSITY

Generally, habitat-quality is degraded as the intensity reducing the habitat’s ability to contribute to species of nearby human land use increases ( Nelleman et al. persistence. Our def nition of habitat-quality is simi- 2 0 0 1 ; F o r m a n et al 2003 ). For example, a forest near a lar to the notion of habitat integrity as used by many city in a developing country may be stripped of much conservation organizations. Habitat with high integ- of its timber and other non-timber forest products rity, like high-quality habitat, is relatively intact and while forests isolated from people will tend to be less has structure and function within the range of historic disturbed (see Chapter 8 ). Or a wetland near agricul- variability. tural lands may have greater water quality issues This tier 1 model assumes that habitat areas with then a wetland surrounded by other wetlands. These higher quality scores are better able to maintain their are both examples of the “edges” that human land full complement of biodiversity over time than those use creates on the boundaries of near-by habitat. In areas with lower scores. This does not mean, how- general, edges facilitate entry of various degraders ever, that areas with lower quality scores are bereft of into habitat including predators, competitors, invasive rare species or are not important sources of biodiver- species, toxic chemicals, and humans. In addition, a sity (see Box 13.1 ). For example, in the U.S. some of high density of human land use in an area means that the last remaining populations of the most threat- any near-by habitat will tend to be isolated, further ened species are on or are immediately adjacent to

Box 13.1 Integrating biodiversity and agriculture: a success story in South Asia

Jai Ranganathan and Gretchen C. Daily plantations and Minor Forest together harbored a distinct bird community, including 90% of the forest-aff liated What is the long-term prospect for harmonizing food species of most conservation concern, such as the Great production and biodiversity conservation? Recent work in Hornbill and the Malabar Grey Hornbill. the Neotropics shows that native species across a wide Arecanut is consumed by over 10% of people, concentrated range of taxa can persist in farming countryside, decades in south and south-east Asia. In traditional cultivation practices after land clearing, if critical landscape features are of the area, arecanut is intercropped (with pepper, vanilla, maintained (e.g., Medellín et al. 2000; Daily et al . 2003 ; coffee, banana, cacao, etc.), increasing the economic return to Horner-Devine et al . 2003 ; Ricketts 2004 ; Mayf eld and farmers and the structural complexity so critical for forest birds. Daily 2005 ; Şekercioğlu et al . 2007 ). Further, as arecanut palm plantations have high water But will these agricultural landscapes continue to demands, they displace rice production, in effect trading a low support native species over centuries to millennia? We conservation value land cover with a much higher one. There is surveyed bird diversity in an ancient agricultural a strong economic incentive to maintain the Minor Forests of landscape, cultivated continuously for over 2000 years the region as forest, since they provide a critical component for and inhabited by people for at least 20 000 years traditional arecanut cultivation: leaf litter, used as mulch in (Ranganathan et al . 2008 ). On the fringes of the Western plantations ( Figure 13.A.1 ). Ghats in southwest India, the area retains many habitat Arecanut may be key to conservation in south and features known to be important for biodiversity southeast Asia, a region with critical conservation (landscape heterogeneity, vegetative structural complexity, challenges on the planet. This example shows that and native vegetation). Most of the land covers—rice, agricultural landscapes can sustain high levels of peanut, cashew, arecanut palm, extremely degraded biodiversity over centuries to millennia, and offers hope shrublands, and native forest—have been present for well that other such production systems can be found. More over 200 years. The native forests are designated either as generally, a likely key to sustainable protection of Reserve Forest (relatively intact, no extraction off cially biodiversity is harmonizing its protection with the delivery allowed) and Minor Forest (extraction of non-timber of as many other ecosystem services as possible, so that products permitted). people reap rewards far beyond the iconic species and We found a rich bird fauna, of which only 4% of species endemic species that have been the more traditional focus were restricted to Reserve Forest. Arecanut palm of conservationists. TIER 1: HABITAT-QUALITY AND RARITY MODEL 231

Figure 13.A.1 Ox-cart loads of leaf litter, collected from Minor Forest, bound for arecanut palm plantations. Photo credit: Jai Ranganathan.

heavily modif ed landscapes (this proximity to such as high vegetation density forests due to f re human activity may be why they are rare in the f rst suppression, different classes of roads (e.g., primary place; Scott et al . 2006 ). versus secondary), or different densities of development. The maps of land uses R do not have to be comprised of the same parcel units as the parcel map but 13.2.1 Calculating a parcel’s habitat-quality all maps need to overlap in space. We use generic spa- score tial units y = 1 , 2 , . . . , Y to allocate land uses that can

The tier 1 model builds a habitat-quality score for impact habitat-quality across the landscape. Let Dyr each parcel x on the landscape ( x = 1 , 2 , . . . , X w h e r e a indicate the amount or density of land-use type r in parcel can be any user-def ned land unit, including a spatial unit y . F o r e x a m p l e , D yr can measure kilom- grid cell, a hexagon, a polygon, etc.) by mapping the eters of road ha-1 i n g r i d c e l l y , p e o p l e h a-1 i n p a r c e l y , location and intensity of all human land uses in the or the hectares of cropland in hexagon y . neighborhood of the parcel and then estimating the The impact of Dyr on habitat-quality in parcel x impact of this human land use on the parcel’s habitat. depends on several factors. First, some human land- We index land-use types that can have a major impact use types have more impact per unit than others. on habitat by r = 1 , 2 , . . . , R. These land-use types can Let wr be the relative impact weighting for r . For be coarsely def ned, such as roads, built areas, crop- example, if built areas have been estimated to have lands, and so forth, and can be supplemented as war- twice the impact of roads on habitat-quality, then ranted by much more ref ned land-use categories, wbuilt /w roads = 2. An equal weight can be used across 232 TERRESTRIAL BIODIVERSITY all land uses R if information on the relative impact equally susceptible to sources of disturbance. Let j = of each source on habitat-quality is not known. 1, . . . , J index habitat types on the landscape where

Second, D yr ’s inf uence on habitat in parcel x is habitat can include land covers and uses highly affected by x’s institutional and structural features. modif ed from their natural state, e.g., urban areas, For example, a fence along the edge of a protected high-intensity croplands, roads. In general, not every parcel might reduce the impact of nearby human land use/land cover type found on the map has to be land uses on habitat in the area. Or extreme slopes included in the set of J; the modeler is free to choose may prevent the entry of predator and competitor the subset of land use/land cover types in set J based species (the protection accorded by institutional on their def nition of habitat. We scale habitat type j ’s and/or structural features in x will vary by land- resistance to human land uses by L j Î [0,1], where use r). Let the resistance to r in parcel x due to the higher values of Lj m e a n s j is more resistant and L j is parcel’s institutional and structural features be close to or equal to 0 for j that are highly modi f ed given by the parameter βxr Î [0,1] with βxr = 1 indi- from their natural state. The values of L j f o r a l l j need cating maximum resistance to inf uence. to be gauged empirically, ideally based on biodiver-

T h i r d , Dyr ’ s i n f uence on habitat in parcel x is sity response. The habitat- quality score for each habi- affected by the distance between x a n d y ; generally, tat type j i n a p a r c e l , Qxj , and the parcel’s aggregate

D yr’s affect on habitat in x declines with distance. Let habitat-quality score, Qx , i s g i v e n b y

d xyr represent the distance between parcel x a n d s p a - tial unit y on the map of land-use type r as measured QqLDxj= () j, x (13.3) by Euclidean distance, road network, or any other relevant distance measure and let α r Î [0,1] be the param- and eter that determines how quickly r ’ s i n f uence on J ⎛⎞⎛⎞ALxj j habitat-quality decays with distance, all else equal. Qqxx= ⎜⎟,, D (13.4) ∑⎝⎠⎜⎟A To determine the potential impact of all land-use ⎝⎠j=1 x types R on habitat in parcel x , given by D x , we con- w h e r e q is any function that is increasing in Lj a n d sider all three factors that affect Dyr ’s relationship decreasing in D x , Q xj = 0 when L j = 0, A xj i s t h e a r e a o f with habitat in parcel x , parcel x i n h a b i t a t t y p e j , a n d Ax is the area of parcel RY x (land use/land cover types that are not part of the DwfdDxrrxrrxyryr= ∑∑ ××()ba,, , (13.1) habitat set j are not given habitat-quality scores). A ry==11 parcel with a Q x score near 0 includes little habitat

where w r × f r ( βx r ,α r , dx y r ) translates the value of D yr into area and/or contains habitat that is severely an impact on parcel x . The higher D x is, the greater degraded within the context of the landscape while max Qx the potential impact of human land uses on the a Q x s c o r e n e a r xX=1,..., {}i n d i c a t e s t h a t a p a r c e l i s quality of habitat in x . A standard way to model replete with habitat and is approximately of the

f r ( βx r ,α r , d x y r) , but by no means the only way, is with highest quality within the context of the landscape. an exponential decay function,

−abrxrxyrd 13.2.2 Calculating a parcel’s rarity score fderxrrxyr()ba,, = . (13.2) While mapping habitat-quality can help identify the

In Eq. (13.2), f r declines, and therefore, the impact of areas where biodiversity is likely to be more or less

D yr on habitat in parcel x declines, as α r , βxr , and/or threatened on the landscape, it is also important to

dxyr increases. A parcel with a D x score near 0 is rela- prioritize habitat types based on their relative rarity. tively unaffected by human land use within the We def ne a habitat type’s rarity as the amount of the

max Dx context of the landscape while a Dx near xX=1,..., {} habitat type currently found on the landscape relative indicates that a parcel is greatly affected by human to the amount of that habitat type that existed on the land use within the context of the landscape. landscape at some reference time. The ideal reference

W h e n t r a n s l a t i n g D x into a habitat-quality score in landscape would be from a period prior to substantial parcel x we assume that not all habitat types are anthropogenic conversion of land (Scholes and Biggs TIER 2 MODELS OF TERRESTRIAL BIODIVERSITY 233

2005 ). For example, a description of the distribution J ⎛⎞Axj of major habitat types in the Willamette Basin of YYxj= , (13.7) ∑ ⎝⎠⎜⎟A Oregon, USA, from the year 1850 ( Christy et al . 2000 ) j=1 x would meet this criterion (even though Native where the more rare the habitat area in x the closer

Americans shaped their landscapes in many ways as Y x is to 1. well, see Mann 2005 ). If a habitat type that was rela- We can combine data on habitat-quality and rarity tively abundant on the reference landscape is now to provide a measure that can be used to prioritize con- rare on the modern landscape, then species depend- servation efforts. Rare habitats with high quality could ent on that habitat type will likely have declined. represent one conservation priority. Such areas are

Assuming we have a reference landscape, we identif ed by parcels with high V x Î [0,1] scores where calculate the relative rarity of habitat type j on a QY V = xx× , (13.8) modern landscape as x maxQYmm max mM==1,..., {} mM1,..., {} ⎛⎞X Axj YI= ⎜−1,∑ x=1 ⎟ (13.5) where m = 1, 2, . . . , M also index parcels. Another jj⎜⎟A ⎝⎠jb simple composite habitat-quality and rarity score is

given by VW x Î [0,1], where Yj = 0 if A jb = 0, I j is equal to 1 if j is a natural land cover (as opposed to a j that is signif cantly QYxx VWx =+gg(1,− ) (13.9) managed) and is equal to 0 otherwise, and A i s t h e max QYmmmax jb mM==1,..., {} mM1,..., {} area of habitat type j on the reference landscape. where γ Î [0,1] determines the weight given to qual- The closer Yj is to 1, the rarer the habitat type j is on the modern landscape vis-à-vis the reference ity versus rarity when tracking the conservation landscape. value of each parcel. If appropriate reference maps are not available (and as a useful check even when they are), another option is to weight the relative scarcity of habitat 13.3 Tier 2 models of terrestrial types on the modern landscape according to a rarity biodiversity metric from NatureServe. NatureServe measures the The major drawback with the tier 1 approach is that status of habitat types with a metric that ranges from it does not necessarily indicate how well the land- 1 through 5 where a 1 indicates critical imperilment scape is meeting the specif c needs of species of con- across a region or the globe, and a 5 indicates that cern. For example, if the species of concern are the habitat type is abundant and secure ( NatureServe generally located in areas of low-quality habitat then 2008 ). In this approach, the relative rarity of habitat conserving the remaining patches of high-quality type j on the modern landscape is given by habitat may not generate as great a conservation X 6 − NS ⎛⎞A return as restoring valuable degraded habitat. In tier ⎛⎞j ∑ x=1 xj Y = ⎜−1, ⎟ (13.6) 2 we complement tier 1 results by assessing how j ⎝⎠⎜⎟5 ⎜⎟A ⎝⎠ species react to and use the landscape and their spatial relationship with habitat-quality patterns. w h e r e NS is NatureServe’s conservation status of j To this end we consider two alternative formula- habitat type j a n d A is the area of the landscape. Set tions of tier 2 models that measure biodiversity pat- NS = 6 for all j t h a t a r e s i g n i f cantly managed. If pos- j terns and status using species-specif c data. The f rst sible, we use NatureServe’s regional scores for NS j tier 2 model measures the marginal contribution of instead of their global scores because a habitat type each parcel to biodiversity on the entire landscape. that is relatively secure from a global perspective The model can also track the change in a parcel’s mar- may be relatively scarce in the particular landscape. ginal contribution to biodiversity as the land use/land Once we have calculated Y for each habitat type, j cover (LULC) pattern on the landscape changes over we quantify the overall rarity of habitat types in time. The second tier 2 model summarizes an entire parcel x on the modern landscape by taking the landscape’s ability to support a suite of species. area-weighted average of x ’s Yj scores, 234 TERRESTRIAL BIODIVERSITY

13.3.1 Parcel-level contribution to biodiversity area marginal value of each LULC type j for species conservation on the landscape s on the landscape is, C For each parcel x on the landscape we calculate a ˆ sj Csj = XK (13.10) marginal biodiversity value (MBV x ) that measures CAH ∑∑xk==11sk xk xs the proportion of the landscape’s total modeled bio- where k indexes LULC diversity supplied by that parcel. The MBV of a par- types as well, A is the area of LULC k in parcel x , cel is a function of: (1) the number of modeled kx and the denominator gives species s ’ total suitable species whose potential ranges o v e r l a p with that habitat area on the landscape. parcel, and (2) the fraction of each species’ suitable Next, we use C ̂ to calculate the MBV score on habitat area that the parcel contains. We deem a s j parcel x , habitat type suitable for a species if the species has S J been observed intermittingly or consistently using ⎛⎞ˆ MBVxsxsxjsj= ∑∑ w H⎜⎟ A C , (13.11) the habitat for breeding, foraging, migration, or sj==11⎝⎠ other life-sustaining purposes. To understand MBV , w h e r e MBV is an estimate of each parcel’s contri- consider a simple example for a single species. Five x bution to the landscape’s total supply of biodi- equally sized parcels on a landscape of 100 parcels versity, biodiversity consists of the S modeled comprise the species’ geographic range. Each of the species, and w is the weight assigned to species s . f ve parcels in the species’ range is completely cov- s If we disregard w temporarily, a parcel will score ered by equally suitable habitat. Then, in our MBV s highly on the MBV metric if it contains suitable model, each of these f ve parcels has an MBV of 0.2, habitat for the species that have little suitable and all other parcels have an MBV o f 0 . C o m p u t i n g habitat elsewhere on the landscape or if it con- MBV on a large landscape with multiple species is a tains reasonable shares of suitable habitat for straightforward generalization of this simple exam- many species. Not all species need be weighted ple that allows for different parcel areas, species- equally. For example, threatened and endangered specif c weights, multiple habitat types within a species may be given greater weight in conserva- S parcel, and incorporation of tier 1 habitat-quality w tion planning and implementation. If ∑ s=1 s = 1 model scores. X then MBV = 1. The calculation of MBV requires a potential range ∑ x=1 x Alternative formulations of the per-unit-area map for each modeled species and a compatibility marginal value that incorporates habitat-quality score for each species/LULC combination that indi- scores from tier 1 are cates the degree of suitability of LULC j f o r s p e c i e s s (s C = 1, 2, . . ., S ) ( i n t i e r 1 j indexed the more narrowly Cˆ = sj sj XJ (13.12) def ned habitat types). We set H = 1 i f p a r c e l x is in the QCAH xs ∑∑xj==11xj sj xj xs potential range of species s , and equals 0 otherwise. Ideally, each species’ potential range map is based on if we have habitat-quality scores for each LULC its estimated geographic range in the pre-modern era. (habitat) type j in each parcel x or C However, such data are only available for a limited ˆ sj Csj = XJ subset of species and may not be particularly reliable. QCAH (13.13) ∑∑xj==11xsjxjxs Therefore, we most often def ne H xs with recently observed patterns of species distribution (however, if we only have or prefer to use parcel-level habitat- see Rondinini et al. (2006 ) for a discussion of the biases quality scores. We assume quality affects suitable introduced in biodiversity modeling and mapping habitat in a linear manner; alternative rates of suit- when using maps of recently observed ranges). able habitat modif cation can be used if we have the We incorporate the degree to which LULC j sup- data to support such relationships. If we use a ver- ̂ ports species s by setting C sj = 0 for unsuitable habi- sion of C s j given by Eq. (13.12) or (13.13) we calcu- tat and C sj > 0 for suitable habitat where C sj = 1 late habitat-quality-adjusted MBVx with a modif ed indicates the most preferred habitat. The per-unit- version of Eq. (13.11), TIER 2 MODELS OF TERRESTRIAL BIODIVERSITY 235

S J S J ⎛⎞ˆ ⎛⎞ˆ (13.17) MBVxsxsxjxjsj= ∑∑ w H⎜⎟ Q A C (13.14) RMBVxt= ∑∑ w s H xst⎜⎟ Q xjt A xjt C sjb sj==11⎝⎠ sj==11⎝⎠

or or S ⎛⎞J MBV= Q w H A Cˆ , S ⎛⎞J xx∑∑ sxsxjsj⎜⎟ (13.15) RMBV Q w H A Cˆ sj==11⎝⎠ xt= xt∑∑ s xst⎜⎟ xjt sjb (13.18) sj==11⎝⎠ where Eq. (13.14) uses the C ̂ from Eq. (13.12) and s j where we use Eq. (13.16) if C ̂ was calculated with Eq. (13.15) uses the C ̂ from Eq. (13.13). s j b s j ̂ Eq. (13.10), Eq. (13.17) if Cs j b was calculated with Eq. (13.12), Eq. (13.18) if C ̂ was calculated with 13.3.1.1 Tracking changes in parcel-level marginal s j b Eq. (13.13), indexing H by t H acknowledges that biodiversity value xs xst species potential range can change over time due to T h e MBV score compares the biodiversity value of climate change or other landscape-level distur- parcels on a landscape for a given point in time, but it bances, and A is the area of LULC j in a parcel x at can be misleading to compare MBV scores of a partic- xjt time t . ular parcel across time. Recall the simple example Finally, for each time t , we calculate the ratio above for one species with f ve equally sized parcels in RMBV /MBV for each parcel x. This so-called its range, each entirely covered with habitat preferred xt x RMBV ratio is greater than 1 if parcel x ’s LULC by the species, the MBV score for each parcel is 0.2. xt and habitat-quality composition (if included in Consider a scenario in which all suitable habitat for RMBV) has changed vis-à-vis the base landscape in the species in four of these parcels is lost, while half of a manner that produces a net increase in the parcel’s the f fth parcel’s area is converted to unsuitable habi- suitable habitat across all species S . tat. In this new landscape, the MBV s c o r e o f t h e f fth parcel would increase to 1 despite the loss of half of its suitable habitat. This change in score may provide 13.3.2 Landscape level biodiversity model useful information in terms of the parcel’s contribu- T h e MBV a n d RMBV models described above are tion to remaining habitat, but it obscures determina- intended primarily for evaluating and comparing tion of whether the amount of biodiversity supported the biodiversity supplied by individual parcels, by a parcel has increased or decreased over time. either within a landscape (the MBV m o d e l ) o r a c r o s s We therefore use an alternative biodiversity statis- time on the landscape (the RMBV m o d e l ) . T h e a b o v e tic, the relative marginal biodiversity value (RMBV ), methodologies do not yield a single, satisfactory to track the contribution of a parcel to the landscape’s landscape score, however, that can be used for level of biodiversity through time. Calculation of assessing the trade-offs between landscape-level RMBV scores and associated RMBV ratios use the measures of biodiversity and ecosystem services same input data as MBV . The calculation of RMBV under different landscape scenarios. To remedy this uses the per-unit-area value of each LULC type j as shortcoming of the MBV model, we use species–area evaluated in the base landscape. This quantity is then relationships (SAR) to develop a landscape-level applied to a future specif cation of the landscape. biodiversity score that requires many of the same Using a base landscape map (i.e., the f rst map in a data inputs used in the MBV o r RMBV m o d e l s . chronological series of maps for the landcape), we The SAR of biogeography ( MacArthur and Wilson calculate C ̂ as in Eq. (13.10), (13.12), or (13.13). This s j 1 9 6 7 ) s p e c i f es the following relationship between statistic is denoted as C ̂ w h e r e t h e b s u b s c r i p t i n d i - s j b total habitat area (A ) and species richness ( S ) : cates the year associated with the baseline landscape. To obtain RMBV scores for parcel x a t t i m e t ( w h e r e t ̂ ScA= z , (13.19) > b ) , w e p l u g C s j b a n d x ’s LULC mix from year t i n t o , S ⎛⎞J ˆ (13.16) where c is a constant and z indicates the rate of spe- RMBVxt= ∑∑ w s H xst⎜⎟ A xjt C sjb sj==11⎝⎠ cies accumulation as A increases and is typically 236 TERRESTRIAL BIODIVERSITY between 0.1 and 0.7. Ideally, the values of the param- and Brooks 2000 ; Parks and Harcourt 2002 ; Cardillo et eters c a n d z are calibrated to observed patterns on al . 2 0 0 6 ) . SAR st r e f ects this tendency as a one-unit the studied landscape. This relationship between decrease in suitable habitat for a species (the numera- richness and habitat area has proven to be one of the tor of Eq. (13.20)) with small range (a small denomina- most empirically robust patterns in all of ecology, tor value) reduces SAR st more than a similar one-unit though with important nuances ( Rosenzweig 1995 ). decrease for a geographically widespread species. While the SAR is typically used to predict species We can modify Eq. (13.20) to include the habitat- richness, we use it to determine how well a land- quality as calculated in tier 1, scape supports species. Each species receives a SAR z XJ st score based on the proportion of area in that spe- QAHC (∑∑xj==11xjt xjt xst sjt ) SARst= γ st , (13.21) cies’ potential geographic range that is suitable X zst AH habitat. The SAR score for species s at time t is ()∑ x=1 xxst

z XJ st z AHC XJ st ∑∑xj==11xjt xst sjt ( ) (13.20) QAHCxt xjt xst sjt (13.22) SARst= g st z , (∑∑xj==11 ) X st SARst= γ st , AH X zst ()∑ x=1 xxst AH ()∑ x=1 xxst w h e r e γ i s a s p e c i e s - s p e c i f c constant (the c i n st where again habitat-quality modif es suitable habi- Eq. (13.19)), z is the species–area function power st tat in a linear manner and the reference species parameter for species s , and all other variables are as range (the denominators of Eqs. (13.21) and (13.22)) before. We index γ, H , C , a n d z w i t h t t o a l l o w f o r assumes habitat of the highest quality is uniformly changes in these parameters and variables over time. present across s ’ range (Q = 1 for all x , j , and t com- In Eq. (13.20) we normalize the observed species- xjt binations and Q = 1 for all x and t combinations). area relationship value (the numerator) with the xt These equations are such that, all else equal, a species–area relationship value that would hold if smaller area of high-quality suitable habitat can the species’ entire potential range were in perfectly generate a higher SAR score than a larger area of suitable habitat (the denominator). Therefore, SAR st st low-quality suitable habitat. is a measure of the fraction of total potential support We can generate a single SAR score for the collec- that the landscape provides for species s a t t i m e t t tion of modeled biodiversity on the landscape by where complete support on the landscape at time t i s taking a weighted sum of all SAR st scores, given by SAR st = 1 ( a s s u m i n g γst = 1 ) . Species that have lost a greater proportion of their S SARtsst= ∑ w SAR , (13.23) suitable habitat across their reference-era range are s=1 at a greater risk of extinction than species that have where w is the weight attached to species s . lost a smaller proportion, regardless of the absolute s size of habitat loss (Channell and Lomolino 2000 ; Abbitt and Scott 2001 ; Scholes and Biggs 2005 ). 13.4 Tier 1 and 2 examples with Therefore, just like our MBV and RMBV models, sensitivity analysis our SAR model may be more accurate if each species’ potential range map (i.e., H ) is based on its esti- We illustrate the tier 1 and 2 models with projected mated spatial distribution in the pre-modern era in LULC change in the region covered by the Sierra lieu of recently observed ranges (though reliable Nevada Conservancy (a California state agency) estimates of pre-modern ranges are not available (see Figure 13.1 ). The 101 000 km2 region encom- for most species and places). passes the Sierra Nevada ecoregion and includes Regardless of the potential range map used, research portions of six other ecoregions, with elevation has also shown that, all else equal, range-restricted ranging from 100 to 4421 m, including the highest species tend to have higher extinction risks than large- peak in the contiguous United States. The region range species ( Newmark 1995 ; Purvis et al . 2000 ; Pimm has approxiamtely 3500 species of vascular plants, TIER 1 AND 2 EXAMPLES WITH SENSITIVITY ANALYSIS 237 including more than 400 endemics (Shevock 1996 ); referred to as “urban,” though this will overstate 293 birds, 135 mammals, 46 reptiles, 37 amphibians, the impact of low density development as exempli- and 61 f sh (CDFG-CIWTG 2007). f ed by the Growth scenario on species that can tol- We used Sierra Nevada housing-density projec- erate lower density residential development. tions for 2030 from the Spatially Explicit Regional Further, we assume for simplicity’s sake that no Growth Model (Theobald 2005 ) to create two 2030 other land-use changes occur in either scenario, landscape scenarios. The Conservation scenario including no expansion of the region’s transporta- assumes that new development will have a mini- tion network. Obviously, no matter the actual mum housing density of one unit per 0.6 ha. The future, new roads will be built in the region as its developed land footprint increases 140% (from population and urban footprint expands. 78 389 ha in 2000 to 187 769 ha in 2030) under this scenario. The Growth scenario includes lower den- 13.4.1 Tier 1 sity housing development options (a minimum of one unit per 4 ha). The developed land footprint We conducted twelve separate tier 1 habitat-quality increases 927% (from 78 389 ha in 2000 to 805 362 ha mapping analyses; in each case we used a different in 2030) under this scenario. The region’s LULC combination of w r a n d Lj values to map parcel-level map in 2000 is the base landscape. (See Davis et al . habitat-quality scores, Q x (the set of 1 habitat types (2006 ) for a much more complex and thorough map- remain constant across all analyses). In this illustra- ping of conservation priority areas in the Sierra tive example, human land uses that affect habitat- given the spatial distribution of habitats, species, quality are roads, urban areas, and agricultural f elds. land tenure, and the predicted spatial pattern of In Figure 13.2 (Plate 6) we present tier 1 results for development in the region, i.e., a tier 3 approach.) two such unique combinations of wr a n d L j values. In For the purposes of the examples below, the areas general the greatest difference between the two is the that develop housing in these scenarios are always relative weight assigned to the human land uses of roads, urban areas, and agriculture. In the “Roads” 350 km parameter combination we assume roads are more deleterious to habitat-quality than urban areas and agriculture and in the “Urban” parameter combination we assume urban areas are more disruptive then the other two land uses. See the chapter’s supplementary online material (SOM) for model details. The “Roads” and “Urban” parameter combinations

775 km produce the two most extreme distributions of habi- Sacramento tat-quality scores on the baseline and two future maps. Specif cally, of the twelve parameter combinations, the Roads parameter combination produces a distribution of Q values that is most skewed to the left on San Francisco/ x Oakland the unit scale (low values) under both future scenarios whereas the Urban parameter combination produces

a distribution of Q x values that is the most skewed to the right under both future scenarios (high values). Not surprisingly, regardless of the parameter Los Angeles combination used, the Growth scenario landscape consistently, because of its larger footprint change, San Diego produces lower Q scores when compared to the Conservation scenario landscape. Only a handful of parcels have lower Q scores under the Conservation Figure 13.1 The Sierra Nevadas Conservancy, California, USA in dark gray. scenario landscape than they do under the Growth 238 TERRESTRIAL BIODIVERSITY

Current Conservation Growth that the habitat-quality trade-off between the two Landscape LULC Scenario LULC Scenario future scenarios seems starker when using the Urban parameter combination rather than the Road parameter combination (see Figure 13.S2 in the SOM). Spatially, the Urban parameter combination leaves many more large patches of relatively high- quality habitat (the darkest green on the maps) than the Roads parameter combination (Figure 13.2 ; Plate 6). Given our uncertainty regarding the direction of development in the future and the relative impact of these sources of human land use on habi- Roads parameter combination tat-quality, it is appropriate to conclude that the areas that are the darkest green (high-habitat-quality) on all four future scenario-parameter combination variations in Figure 13.2 (Plate 6) are the areas most likely to contain high-quality habitat in the future.

13.4.2 Tier 2 analyses

W e c a l c u l a t e d MBV , RMBV, a n d RMBV ratio scores (without considering habitat-quality) for various subgroups of herpetofauna—federally endangered herpetofauna (FE), federally threatened herpeto-

Urban parameter combination fauna (FT), California herpetofauna of special concern (CSC), amphibians (A), and reptiles (R)—that Parcel Q scores have at least part of their range in the Sierra Nevada Conservancy. In Figure 13.3 histograms of >0– 0.91– 0.93– 0.95– 0.97– 0.99– the distribution of parcel RMBV ratio values across <0.9 0.92 0.94 0.96 0.98 1.00 the landscape (i.e., the relative change in parcels’ MBV scores from 2000 to 2030) under each scenario Figure 13.2 Maps of parcel habitat-quality scores when the “Roads” and “Urban” parameter combinations are used in the Sierra Nevada illustrative for each subgroup of herpetofauna using mean Csj example. We ran the tier 1 model on a grid map with a cellular resolution of values for each s a n d j combination are given. The 400 m × 400 m (16-ha grid cells). In these maps we present the mean FT (N = 5 ) , F E ( N = 3 ) a n d C S C ( N = 25) subgroups habitat-quality score (Q ) of all grid cells within 500 hectare hexagons. There experience severe reductions in effective habitat are 23 042 500-ha hexagons in the Sierra Nevada Conservancy. In both area in many parcels under the Growth scenario future LULC scenarios the majority of residential development is centered on Sacramento, and generally along the western foothills. In the Growth (recall that a RMBV ratio value near 0 indicates a scenario, montane hardwood is the land cover type that loses the most area signif cant loss of habitat in the parcel over time to development (158 268 ha). In the Conservation scenario, annual for the modeled species). These histograms also grassland is the land cover type that loses the most area (17 798 ha). See show that the reduction in habitat under the the chapter’s SOM for all tier 1 model details. (See Plate 6.) Conservation scenario most acutely affects the same subgroups. We map the FT subgroup species year 2000 MBV landscape scenario (see Figure 13.S1 in the chapter’s and RMBV ratio values under both scenarios

SOM) using either set of parameter combinations. using minimum and maximum C sj values in Figure Further, because the only LULC change on the land- 13.4 (Plate 7). The MBV maps ( Figure 13.4 ; Plate 7) scape is development expansion, it is not surprising indicate that the most important habitat for FT TIER 1 AND 2 EXAMPLES WITH SENSITIVITY ANALYSIS 239

A&R A CSC FE FT R 400

300

200

100

Conservation LULC Scenario 0 400

300

200

100 Growth LULC Scenario 0 0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1 RMBV ratio

Figure 13.3 Frequency of RMBV ratio parcel scores using mean species-habitat suitability scores for all herpetofauna subgroups in the Sierra Nevada illustrative example. The frequency axis of each histogram has been truncated at 400. The California herpetofauna of special concern (CSC), federally endangered herpetofauna (FE), and federally threatened herpetofauna (FT) subgroups experience the greatest drop in effective habitat area relative to the 2000 landscape. The amphibian (A) herpetofauna subgroup has the least number of parcels with signif cant reductions in effective habitat area under both future LULC scenarios. The other herpetofauna subgroups include all (A&R) and reptiles (R). See the chapter’s SOM for all tier 2 model details. species in 2000 were in the western foothills. It is in zone in which most of the projected development this area that most urbanization is expected to is expected to occur. At what point a decline in occur by 2030 under both scenarios, albeit to a SAR indicates an immediate and imminent threat much greater extent in the Growth scenario. The to the persistence of a subgroup is an empirical maps appear to be fairly insensitive to the range in question. See the chapter’s SOM for tier 2 model

Csj . Finally, these maps explain why the A (N = 35) details. subgroup has the most right-skewed histogram of parcel RMBV ratio scores: species in subgroup A 13.4.3 Incorporating tier 1 results in a are more likely than any other subgroup of species tier 2 analysis to be found in the higher elevation areas, the areas experiencing the least development. In many cases we will not be able to perform a Data in Figure 13.5 corroborates the decrease in tier 2 analysis due to a lack of species-specif c FT and CSC subgroup suitable habitat as sug- data. However, if we do have the wherewithal to gested by the RMBV ratio histograms. These two perform a tier 2 analysis, as we do in this exam- subgroups experience the greatest relative ple, if would be judicious on our part to perform decrease in their SAR scores (without considering it twice, once with tier 1 output incorporated and habitat-quality) under both scenarios no matter another time without (e.g., the results above). which C sj and z values we use (low, mean, or high Such an analysis will indicate some of the spatial for the Csj s c o r e s f o r e a c h s a n d j combination and relationships between habitat-quality, species 0.11, 0.25, 0.64, and 1 for the z scores). Again, this habitat, and the pattern of habitat loss on the is due to the fact that many of the taxa included in landscape over time. these subgroups are primarily found in the foot- By def nition, MVB xt and SAR st scores calculated hill areas and the Central Valley to the west, the with habitat-quality scores will always be equal to 240 TERRESTRIAL BIODIVERSITY

MBV RMBV ratios Minimum C Average C Maximum C Current Conservation Growth 0% Landscape LULC Scenario LULC Scenario R A&R FER A&R FE A RA&R FE A CSCA CSC CSC

–2% FT FT FT Conservatio LULC Scenario –4% FE FE FE

R R R A&R –6% A&R A&R A A A Minimum C Growth CSC

–8% LULC Scenario s Countryside SAR Score

¢ CSC CSC

–10%

–12%

–14%

FT FT FT –16% Maximum C Percent Change from Base Landscape Change from Percent –18%

Figure 13.5 Percent change in a herpetofauna subgroup’s SAR scores from 2000 to 2030 under both future LULC scenarios in the Sierra Federally Threatened (FT) Herpetofauna Nevadas illustrative example. The subgroups in the graph include all MBV herpetofauna (A&R), amphibians (A), California herpetofauna of special 0 1 RMBV concern (CSC), federally endangered herpetofauna (FE), federally ratios 0 1 threatened herpetofauna (FT), and reptiles (R). We conduct the analysis

using minimum, average, and maximum Csj values for all s ,j combinations. Figure 13.4 The spatial distribution of MBV and RMBV ratio scores for See the chapter’s SOM for all tier 2 model details. federally threatened herpetofauna (FT) using minimum and maximum species-habitat suitability scores in the Sierra Nevada illustrative example. The Growth scenario creates a much greater loss in FT subgroup species when and when not incorporating Q x s c o r e s f r o m effective habitat area in the foothills of the Sierra Nevada than the the Roads and Urban parameter combinations Conservation scenario does. See the chapter’s SOM for all tier 2 model (using mean C v a l u e s f o r a l l s a n d j c o m b i n a t i o n s ) . details. (See Plate 7.) sj The inclusion of Q x scores in the SAR calculations suggests: (1) that habitat conversion under the or lower than those without. However, habitat- Growth scenario tends to occur more frequently on quality-modif ed RMBV xt ratio scores and the higher quality habitat than is does under the change in habitat-quality-modif ed SARst scores Conservation scenario, (2) if we use the Urban over time may be more or less than their unmodi- parameter combination to measure the relative f ed counterparts. In general, habitat-quality- impact of human land uses on habitat-quality then a modif ed RMBV ratio scores will be higher and greater proportion of higher quality habitat is devel- decreases in habitat-quality-modif ed SAR scores oped by 2030 under either scenario than if we used will be less severe than their unmodif ed counter- the Roads parameter combination, and (3) a greater parts if the habitat that is lost over time tends to be fraction of CSC subgroup’s high-quality habitat is of low quality on the base landscape. developed than that of the FT subgroup’s. I n T a b l e 1 3 . 1 w e p r e s e n t t h e c h a n g e i n t h e C S C While we do not perform the analysis for this and FT herpetofauna subgroups’ SAR s t a t i s t i c s illustrative example, we expect a comparison of the Table 13.1 Percent change in SAR statistic from 2000 to 2030 where z = 1

2030 LULC scenario Conservation Growth

Q scores from tier 1 are not Q scores from the tier 1 Q scores from the tier 1 Q scores from tier 1 are not Q scores from the tier 1 Q scores from the tier 1 included (%) “Roads” parameter “Urban” parameter included (%) “Roads” parameter “Urban” parameter combination are included combination are included combination are included combination are included (%) (%) (%) (%)

Herpetofauna subgroup

California herpetofauna −1.11 −0.88 −1.11 − 8.26 − 9.95 − 10.14 of special concern (CSC) Federally threatened −2.35 −1.76 −2.15 − 15.58 − 17.91 −18.16 herpetofauna (FT) 242 TERRESTRIAL BIODIVERSITY

RMBV ratio scores with and without habitat-quality the Monte Carlo method, one or more variable scores to corroborate our initial f nding that the and parameter values are randomly drawn from Growth scenario converts a greater portion of high- distributions that describe the variables and quality habitat and that under either scenario a parameters range of possible values. In addition, greater portion of high-quality CSC subgroup habi- functional forms can also be varied in a random tat than high-quality FT subgroup habitat is devel- manner. The process of randomly drawing varia- oped between 2000 and 2030. ble and parameter values or functional forms and then running the model is repeated many times. Thus, this method calculates a distribution of 13.4.4 Sensitivity analyses and analysis potential solutions, including, if the process is limitations iterated enough times, approximations of the A parcel’s habitat-quality scores are explained by worst and best case scenarios. The more variables, variables wr , αr , β, and L j, that are uncertain and parameters, and functions that are simultaneously functions, fr and q, that are simplistic and in many varied, the more robust the uncertainty analysis. cases, unverif able representations of complicated See the SOM for all data used in this illustrative ecological processes. In tier 2 analyses, the data example. used in the H and C matrices and to set the z expo- nents are often derived from a limited number of f eld-based studies. 13.5 Limitations and next steps In this illustrative example we attempt to quan- 13.5.1 Limitations tify some of the ramif cations of this uncertainty with a limited data sensitivity analysis. First, for the Tier 1 and 2 models should be viewed as comple- baseline and the two scenario maps, we used 12 dif- ments and not as alternatives. Because of data limitations, we presume that tier 1 modeling will be ferent combinations of w r and Lj values to f nd the combinations that produced the most left-skewed much more widely implemented. If species-level data on habitat compatibility and potential ranges and right-skewed distributions of Q x across all 3 maps, respectively ( Figure 13.2 ; Plate 6). We did not are available, however, we believe it is important to determine if both tiers suggest the same trends and vary the values of α r or βxr nor did we experiment spatial patterns of biodiversity. If tier 1 and 2 analy- with the structures of f r a n d q ; otherwise, we would have calculated an even greater range in habitat- ses produce spatially correlated results (e.g., areas quality results. with high Q scores also have high MBV scores; the In our tier 2 illustration we found MBV , RMBV , parcels that exhibit the greatest decline in habitat- and RMBV ratio scores for all parcels and SAR val- quality over time have the lowest RMBV ratio scores; the relative change in SAR scores over time ues on all three landscapes using a range of Csj values for each s ,j combination. For example, in Figures are the same with and without habitat-quality 13.S3 and 13.S4 in this chapter’s SOM we illustrate included; etc.), then the quality of habitat can be which subgroup’s distribution of RMBV ratios is used as a proxy for the status of species of concern on the landscape. most sensitive to uncertainty in C sj scores under both the Growth and Conservation scenarios. In The sources of anthropogenic threat that we have Figures 13.4 and 13.5 the ramif cations of some of presented here as candidates for use in the tier 1 habi- this uncertainty is mapped and graphed, tat-quality model are all land use related, but many respectively. factors affect biodiversity that are diff cult to map, Placing reasonable bounds on model outputs is such as the presence of exotic species or an altered dis- the simplest uncertainty analysis we can perform. turbance regime. Further, our tier 1 model could be

A more thorough method for determining how extended to allow habitat resistance Lj to vary across sensitive a model’s output is to variable and the R sources of human land uses (i.e., we could use Lj parameter value and functional form uncertainty instead of L jr). For example, some forested areas may is to run a Monte Carlo simulation analysis. Under be resistant to the chemicals applied to nearby farm LIMITATIONS AND NEXT STEPS 243 f elds but particularly affected by a nearby road’s pro- abilities of some specif c objective of biodiversity vision of easy access to gatherers of timber and non- conservation; for example, a species guild that timber forest products. In addition, the source of shares habitat or ecological roles. We would then be human land use and its ecological impact may be calculating guild-specif c Q xj and Qx scores. For thousands of miles apart. This is most strikingly dem- example, if mapping the habitat-quality of interior onstrated by the pronounced effect of fertilizer use on forest-dependent species, we would construct the farm f elds of the upper Midwest US on water quality set of habitat types J and Lj values according to their in the Gulf of Mexico. Such broad-scale impacts will needs, suitabilities, and reaction to different sources not be characterized by our tier 1 model. of disturbance. Then a species MBV , RMBV , and Integrating landscape structure and connectivity RMBV ratio scores for these interior forest-depend- analyses in the tier 2 model would allow for more ent species, as well as the countryside SAR score for explicit population viability modeling (for examples the entire landscape, could be modif ed by Q xj or Q x of such modeling see Hanski and Ovaskainen 2000 ; scores that are more descriptive of the habitat limi- V o s et al . 2 0 0 1 , S c h u m a k e r et al . 2004 ). Other than the tations facing this particular guild. spatial relationship between potential range space Both tiers of the biodiversity model can be easily and habitat type, the tier 2 biodiversity models pre- integrated with systematic conservation planning sented here do not consider how the spatial conf gu- processes, such as ecoregional assessments or biodi- ration of suitable habitat on a landscape may affect versity visions. The habitat-quality model in tier 1 species. Further, the models do not consider the size could be used to ensure that the portfolio of selected of habitat patches or the ability of animals to move sites includes those features with the highest quality. from patch to patch (see Polasky et al . 2 0 0 5 ; W i n f r e e In addition, ecosystem service planning processes et al . 2 0 0 5 ; N e l s o n et al . 2 0 0 8 ; P o l a s k y et al . 2 0 0 8 f o r can use these models to calculate the biodiversity examples of species conservation models that do “costs” or trade-offs, if any exist at all, associated with consider species movement among patches). In LULC changes designed to enhance or conserve eco- addition, issues of minimum viable population size, system service provision where landscape-level bio- population stochasticity, competition and other spe- diversity costs are given by changes in SAR values cies interactions, are not addressed in our models. and spatial biodiversity costs by RMBV ratio maps By treating a landscape as an island surrounded (e.g., Chan et al . 2 0 0 6 ; N e l s o n et al . 2 0 0 8 ) . I n c o m p a r - by terra incognito we may under- or overestimate ing spatial patterns of biodiversity and ecosystem species status in the broader region. For example, services we may identify areas where conservation a species with a small amount of suitable habitat investments can benef t species and i n c r e a s e t h e s u p - in its potential range on the studied landscape ply and value of ecosystem services ( Balvanera et al . will have a low SAR st score. However, if the spe- 2 0 0 1 ; N a i d o o a n d R i c k e t t s 2 0 0 6; T u r n e r et al . 2007 ). cies has effective habitat in a signif cant portion of Finally, the output from either tier could be used as its potential range outside of the landscape in input into a site-selection algorithm, such as Marxan, question then a low SAR score may not be indica- to design alternative conservation area networks tive of the species’ overall status. On the other ( Ball and Possingham 2000 ). For example, a conser- hand, a species can have a high SAR score in the vation objective could be to f nd LULC changes that studied landscape yet only have a small portion of increase as many RMBV ratio parcel scores, or alter- its regional or global potential range space in suit- natively minimize the number of RMBV ratio parcel able habitat. scores below 1, subject to some conservation budget constraint (see Nelson et al. 2 0 0 8 f o r d e t a i l s o n a o p t i - mization model that maximized the gain in SAR 13.5.2 Next steps across a suite of species for a given conservation All of the models presented in this chapter can be budget). Then, once a network of conservation areas extended in fruitful ways. For example, the set of is selected to meet targets, the habitat-quality model habitat types J and Lj values in a tier 1 analysis can could be used to conduct a threats assessment on pro- be def ned according to the habitat needs and suit- posed or existing protected areas. 244 TERRESTRIAL BIODIVERSITY

References Hanski, I., and Ovaskainen, O. (2000). The metapopula- tion capacity of a fragmented landscape. Nature , 404 , Abbitt, R., and Scott, J. (2001). Examining differences 755–8. between recovered and declining endangered species. Horner-Devine, M. C., Daily, G. C., Ehrlich, P. R., et al . Conservation Biology , 15 , 1274–84. (2003). Countryside biogeography of tropical butter- Ando, A., Camm, J., Polasky, S., et al. (1998). Species distri- f ies. Conservation Biology, 17 , 168–77. butions, land values, and eff cient conservation. Science , Hughes, J. B., Daily, G. C., and Ehrlich, P. R. (1997). 279 , 2126–8. Population diversity: Its extent and extinction. Science , Ball, I., and Possingham, H. (2000). MARXAN (v1.8.2): 278 , 689–92. Marine reserve design using spatially explicit annealing . Leopold, A. (1949). Sand county almanac, and sketches here Balvanera, P., Daily, G., Ehrlich, P. et al. (2001). Conserving and there . Oxford University Press, New York. biodiversity and ecosystem services. Science, 291 , 2047. MacArthur, R. H., and Wilson, E. O. (1967). The theory of California Department of Fish and Game. California island biogeography . Princeton University Press, Interagency Wildlife Task Group (CDFG-CIWTG). Princeton, NJ. (2007). CWHR personal computer program . Mann, C. C. (2005). 1491: new revelations of the Americas Cardillo, M., Mace, G., Gittleman, J. et al. (2006). Latent before Columbus. Vintage Books, New York. extinction risk and the future battlegrounds of mammal Margules C. R., and Pressey, R. L. (2000). Systematic con- conservation. Proceedings of the National Academy of servation planning. Nature , 405 , 243–53. Sciences, 103 , 4157–61. Mayf eld, M. M., and Daily, G. C. (2005). Countryside bio- Ceballos, G., Ehrlich, P., Soberon, J., et al . (2005). Global geography of neotropical herbaceous and shrubby mammal conservation: What must we manage? Science, plants. Ecological Applications , 15 , 423–39. 309 , 603–7. Medellin, R. A., Equihua, M. and Amin, M. A. (2000). Bat Chan, K., Shaw, M., Cameron, D., et al . (2006). Conservation diversity and abundance as indicators of disturbance in planning for ecosystem services. Plos Biology, 4 , 2 1 3 8 – 5 2 . neotropical rainforests. Conservation Biology , 14 , Channell, R. and Lomolino, M. (2000). Dynamic biogeog- 1666–75. raphy and conservation of endangered species. Nature, Naidoo, R., and Ricketts, T. H. (2006). Mapping eco- 403 , 84–6. nomic costs and benef ts of conservation. PLoS Biology, Christy, J., Alverson, E., Dougherty, M., et al . (2000). 4 , e360. Presettlement vegetation for the Willamette Valley, Oregon, NatureServe. (2008). NatureServe Explorer: Ecological version 4.0, compiled from records of the General Land Off ce Communities and Sytems. Accessed at http:/ /www. Surveyors ( c.1850) . natureserve.org Daily, G., Ceballos, G., Pacheco, J., et al . (2003). Countryside Nelleman, C., Kullered, L., Vistnes, I., et al. (2001). GLOBIO. biogeography of neotropical mammals. Conservation Global methodology for mapping human impacts on Biology, 17 , 1814–26. the biosphere. United Nations Environment Program Davis, F., Costello, C., and Stoms, D. (2006). Eff cient con- (UNEP); Report UNEP/DEWA/TR.01–3. servation in a utility-maximization framework. Ecology Nelson, E., Polasky, S., Lewis, D., et al. (2008). Eff ciency of and Society, 11, 33. incentives to jointly increase carbon sequestration and Diaz, S., Tilman, D., and Fargione, J. (2005). Biodiversity species conservation on a landscape. Pr oceedings of the regulation of ecosystem services. In: R. Hassan, National Academy of Sciences . 105 , 9471–6. R. Scholes, and N. Ash, Eds., Ecosystems and human well- Newmark, W. (1995). Extinction of mammal populations being: current state and trends. Island Press, Washington, in western American national parks. Conservation DC, pp. 297–329. Biology , 9 , 512–26. Dirzo, R., and Raven, P. (2003). Global state of biodiversity Parks, S., and Harcourt, A. (2002). Reserve size, local and loss. Annual Review of Environment and Natural human density, and mammalian extinctions in U.S. pro- Resources , 28 , 137–67. tected areas. Conservation Biology , 16 , 800–8. Ehrlich, P. R., and Ehrlich, A. H. (1982). Extinction . Pereira, H., and Daily, G. (2006). Modeling biodiversity Ballantine, New York. dynamics in countryside landscapes. Ecology, 87 , Forman, R., Sperling, D., Bissonette, J., et al. (2003). Road 1877–85. Ecology . Island Press, Washington, DC. Pimm, S., and Brooks, T. (2000). The sixth extinction: How Groves, C. R., Jensen, D. B., Valutis, L. L., et al . (2002). large, where, and when? In: T. Raven and P. H. Williams, Planning for biodiversity conservation. Bioscience , 52 , Eds., Nature and human society , pp. 46–62. National 499–512. Academy Press, Washington, DC. LIMITATIONS AND NEXT STEPS 245

Polasky, S., Nelson, E., Lonsdorf, E., et al . (2005). Conserving human-dominated landscapes , vols 1 and 2. Island Press, species in a working landscape: land use with biological Washington, DC. and economic objectives. Ecological Applications , 15 , S ̧ekercioğlu, Ç., Daily, G., and Ehrlich, P. (2004). Ecosystem 1387–1401. consequences of bird declines. Proceedings of the National Polasky, S., Nelson, E., Camm, J. et al . (2008). Where to put Academy of Sciences, 101 , 18042–7. things? Spatial land management to sustain biodiver- S ̧ekercioğlu, Ç., Loarie, S., Brenes, F., et al . (2007). sity and economic returns. Biological Conservation , 141 , Persistence of forest birds in the Costa Rican agricul- 1505–24. tural countryside. Conservation Biology , 21 , 482–94. Purvis, A., Gittleman, J., Cowlishaw, G., et al . (2000). Shevock, J. (1996). Status of rare and endemic plants. In: Predicting extinction risk in declining species. D. C. Erman, Ed., Sierra Nevada ecosystem project: f nal Proceedings of the Royal Society of London, Series B: report to Congress, vol. II, pp. 691–707. University of Biological Sciences . 267 , 1947–52. California—Davis, Centers for Water Centers for Water Ranganathan, J., Daniels, R., Chandran, S., et al . (2008). and Wildland Resources, Davis. Sustaining biodiversity in ancient tropical countryside. Stoms, D., Davis, F., and Cogan, C. (1992). Sensitivity of wild- Proceedings of the National Academy of Sciences , 105 , 17852–4. life habitat models to uncertainties in GIS data. Ricketts, T. (2004). Tropical forest fragments enhance pol- Photogrammetric Engineering and Remote Sensing , 58 , 843–50. linator activity in nearby coffee crops. Conservation Theobald, D. (2005). Landscape patterns of exurban growth Biology , 18 , 1262–71. in the USA from 1980 to 2020. Ecology and Society , 10 , 3 2 . Rondinini, C., Wilson, K., Boitani, L., et al. (2006). Tradeoffs Turner, W., Brandon, K., Brooks, T., et al. (2007). Global of different types of species occurrence data for use in conservation of biodiversity and ecosystem services. systematic conservation planning. Ecology Letters , 9 , BioScience, 57 , 868–73. 1136–45. Vos, C., Verboom, J., Opdam, P., et al. (2001). Toward eco- Rosenzweig, M. (1995). Species diversity in space and time. logically scaled landscape indices. American Naturalist , Cambridge University Press, Cambridge. 157 , 24–41. Sala, O., Vuuren, D. v., Pereira, H., et al . (2005). Williams, P. H., Gibbons, D., Margules, C., et al . (1996). A Biodiversity across scenarios. In: S. Carpenter, P. comparison of richness hotspots, rarity hotspots, and Pingali, E. Bennett, and M. Zurek, Eds., Ecosystems and complementary areas for conserving diversity of British human well-being , v o l . 2 : Scenarios , pp. 375–408. Island birds. Conservation Biology , 10 , 155–74. Press, Washington, DC. Wilson, E. O. (1992). The diversity of life . Harvard University Schill, S. and Raber, G. (2008). Protected Area Tools (PAT) Press, Cambridge, MA. for ArcGIS 9.2 (version 2.0) software. Wilson, K., Underwood, E., Morrison, S., et al . (2007). Scholes, R. and Biggs, R. (2005). A biodiversity intactness Conserving biodiversity eff ciently: What to do, where, index. Nature, 434 , 45–9. and when. PLoS Biology , 5 , e223. Schumaker, N., Ernst, T., White, D., et al. (2004). Projecting Winfree, R., Dushoff, J., Crone, E., et al . (2005). Testing sim- wildlife responses to alternative future landscapes in ple indices of habitat proximity. American Naturalist , Oregon’s Willamette Basin. Ecological Applications , 14 , 165 , 707–17. 381–400. Worm B., Barbier, E., Beaumont, N. et al . (2006). Impacts of Scott, J. M., Goble, D., and Davis, F. (2006). The Endangered biodiversity loss on ocean ecosystem services. Science , Species Act at thirty: conserving biodiversity in 314 , 787–90. This page intentionally left blank SECTION III Extensions, applications, and the next generation of ecosystem service assessments This page intentionally left blank CHAPTER 14 Putting ecosystem service models to work: conservation, management, and trade-offs

Stephen Polasky, Giorgio Caldarone, T. Ka’eo Duarte, Joshua Goldstein, Neil Hannahs, Taylor Ricketts, and Heather Tallis

14.1 Introduction ized example for two ecosystem services, carbon sequestration and water quality, under four manage- Changing land use or land management affects the ment alternatives (A, B, C, and D). For simplicity of provision and value of a range of ecosystem services the illustration, suppose that the cost of implement- as well as biodiversity. The large number of poten- ing each management is the same and that managers tially competing objectives can complicate decisions care only about water quality (measured on the verti- about landscape management. In rare cases the best cal axis) and carbon sequestration (measured on the choice among management alternatives will be obvi- horizontal axis). Alternatives B and C are preferred ous because one alternative delivers higher levels of to alternative A because both carbon sequestration all ecosystem services and biodiversity compared to and water quality scores are higher under these alter- other alternatives (“win–win” solutions). In most natives. Choosing among B, C, or D (or between A cases, however, comparing among management alter- and D), however, involves a trade-off with each alternative requires evaluating trade-offs among various native providing more of one service and less of the ecosystem services and biodiversity conservation. other. Which of these alternatives (B, C, or D) will be The models described in this book generate pre- preferred depends on the relative value of the two dictions about the provision of multiple ecosystem ecosystem services. If carbon sequestration is highly services and biodiversity for any given pattern of valued relative to water quality, then alternative D land use and management across a landscape. In this will be preferred to the other alternatives. As water chapter we illustrate how one might use these pre- quality increases suff ciently in value relative to car- dictions collectively to analyze alternative conserva- bon sequestration, alternative B or C will be the most tion and management strategies. By comparing maps preferred. Alternative A, however, will never be the of ecosystem services and biodiversity, managers can most preferred option regardless of the value judg- locate areas that, if managed correctly, can provide ment about how to weight the value of carbon high levels of both. By comparing outcomes across sequestration relative to water quality because it is different management alternatives, conservationists dominated by other alternatives (B and C). and managers can gain insight into which alterna- Here we describe four different approaches to tives may be most desirable. Further, the analyses analyzing conservation and management alterna- can be used to suggest and investigate new strategies tives that illustrate potential application of models that may improve results for key ecosystem services of the type described in this book. We start with an or biodiversity conservation objectives. example that builds from conservation planning in Trade-offs and potential win–win solutions are which the planner chooses sites to include in a illustrated in Figure 14.1 , which shows a simple styl- reserve network. We expand upon the traditional

249 250 PUTTING ECOSYSTEM SERVICE MODELS TO WORK

serve biodiversity with limited resources ( Margules and Pressey 2000 ; Sarkar et al. 2006 ). The conservation planning literature has developed a set of methods B for choosing which sites to include in a conservation reserve network in a range of applications (e.g., C K i r k p a t r i c k 1 9 8 3 ; M a r g u l e s et al. 1988 ; Cocks and Baird 1989 ; Camm et al. 1 9 9 6 ; W i l l i s et al. 1996 ;

Water Quality P o s s i n g h a m et al. 2000 ) incorporating such factors as A varying land cost (e.g., Ando et al. 1998 ; Naidoo et al. 2006 ), species persistence in reserves (e.g., Cabeza D and Moilanen 2001 ; Nicholson et al. 2 0 0 6 ) a n d s e q u e n - Carbon Sequestration tial choice and threats of habitat loss (e.g., Costello

Figure 14.1 Stylized example of output of two ecosystem services, and Polasky 2004 ; Meir et al. 2 0 0 4 ; W i l s o n et al. 2006 ). carbon sequestration and water quality, evaluated under four hypothetical Even in the well-studied context of conservation management alternatives (A, B, C, and D). Moving from A to either B or C site selection, spatially explicit models of ecosystem increases both carbon sequestration and water quality. Comparison among services and biodiversity can expand the type of all other management alternatives involves trade-offs of an increase in one information available to conservation managers and service and a decrease in the other. improve conservation decision-making. Such models can identify areas of high and low value for a variety conservation planning approach by including the of ecosystem services that can be compared spatially effect of choosing conservation reserves on the pro- to areas of high and low value for biodiversity ( Chan vision and value of ecosystem services. Second, we et al. 2006 ). In areas of high overlap, conservation evaluate the provision of multiple ecosystem serv- organizations can partner with other groups inter- ices from alternative scenarios of land use and land ested in water quality, carbon sequestration or other management, illustrating synergies and trade-offs services to affect outcomes, effectively increasing the among ecosystem services and biodiversity conser- resources available for conservation ( Goldman et al. vation. Third, we combine the models with optimi- 2008 ). Conservation organizations can then concen- zation to def ne an eff ciency frontier that shows the trate their own resources on areas of high biodiversity maximum possible combinations of ecosystem serv- value but that do not have high values for services. ices provision and biodiversity conservation that are An example of this type of analysis is shown in feasible from a landscape. Fourth, we illustrate how Naidoo and Ricketts (2006 ). They map the monetary one can include estimates of monetary value of eco- values of f ve ecosystem services (bushmeat harvest, system services to provide a benef t–cost analysis of timber harvest, bioprospecting, existence value, and management alternatives. At the end of the chapter carbon storage) in the Mbaracyau Forest Biosphere we offer some brief concluding comments on the Reserve in Paraguay (see Figure 14.2 ; Plate 8). (Some current state of the art and important next steps. conservationists express concern about putting monetary values on nature; we discuss these issues in 14.2 Applying ecosystem service and Section 14.2.4 . Also, see Chapter 2 ). Naidoo and biodiversity models in management Ricketts (2006 ) develop maps that show areas where and conservation contexts conservation benef ts are high and would more than cover the costs of conservation and other areas where 14.2.1 Site selection for conservation the converse is true. Naidoo and Ricketts (2006 ) also Conservation managers typically face a situation in use these maps to evaluate three alternative locations which they have a large number of worthwhile con- for a proposed corridor linking two protected areas, servation projects but only have resources suff cient and f nd that one corridor would provide much to fund a small fraction of these projects. The system- higher benef ts relative to costs than the other two. atic conservation planning f eld developed to provide This is an example of how such maps can help direct advice to conservation managers on how best to con- conservation efforts to high benef t areas. (a) (b)

2–19 0–1.74 20–47 1.75–3.99 48–272 4.00–6.96 273–325 6.97–13.49 326–333 13.50–15.09 334–456 15.10–16.54 457–1,045 16.55–18.50

27.60 0 2.21

(e) (f)

0 1–229 0 230–291 25 292–412 413–577 578–766 767–975 Bolivia N Brazil 20 0 20 40 60 Kilometers WE Paraguay

S Argentina

Figure 14.2 Net present values in US$ha–1 for selected ecosystem services in the Mbaracyau Forest Biosphere Reserve, Paraguay. (a) Sum of all f ve services; (b) sustainable bushmeat harvest; (c) sustainable timber harvest; (d) bioprospecting; (e) existence value; and (f) carbon storage. (See Plate 8.) Source: Naidoo and Ricketts ( 2006 ). 252 PUTTING ECOSYSTEM SERVICE MODELS TO WORK

In some cases, particularly in cases involving the outcomes considered in planning beyond biodiver- protection of municipal drinking water supply (e.g., sity conservation targets. Doing so can show areas Bogota, New York City, Quito), the value of ecosys- on the landscape that are of high priority for conser- tem services is high enough to choose management vation targets and various ecosystem services. decisions that also support conservation The main disadvantage of using spatially explicit (Chichilnisky and Heal 1998 ; NRC 2000; Echevarrıa models in this manner is that results do not neces- 2002 ). In such cases, payments for ecosystem serv- sarily indicate how the landscape should be man- ices can be more than suff cient and there is little or aged. Management to promote a particular no need for a conservation organization to spend ecosystem service might differ from management their scarce resources to accomplish conservation to promote another service or biodiversity conser- objectives. In other cases, promoting the provision vation. For example, carbon sequestration may be of ecosystem services may align with conservation, maximized by planting trees but this may decrease but the services themselves may not be valuable surface water runoff and reduce availability for enough to tip the balance toward biodiversity- downstream users ( Jackson et al . 2005 ). In classic friendly management. In this case, conservation conservation site selection the management choice organizations can usefully partner with other is simple—either protect a site or don’t—and a pro- groups interested in the provision of ecosystem tected site is assumed to benef t all species. With the services. Finally, there will be other cases where inclusion of ecosystem services, however, the choice management for ecosystem services does not align of management options is of greater interest and with conservation objectives. In these cases, conser- complexity. When different management options at vation organizations will be on their own, just as the same spatial location are best for different objec- they would be with no consideration of ecosystem tives just highlighting high priority areas on the services. landscape is not enough. What is needed in this Spatially explicit information on ecosystem serv- case is an analysis that shows outcomes for ecosys- ices can also be integrated with conservation plan- tem services and biodiversity under different types ning exercises in other ways. For example, if of management. ecosystem services are valued in monetary terms, the cost of including a particular site could be 14.2.2 Analysis of management alternatives reduced by the increase in value of ecosystem services provided if the site is chosen as a reserve. Doing The spatially explicit models de f ned in earlier chap- so would shift conservation priorities toward sites ters are designed to evaluate multiple ecosystem that generate valuable ecosystem services in a fash- services and biodiversity objectives under alternative ion similar to priority given to inexpensive sites. It conservation or management plans. In this section we is also possible to require that targets could be spec- highlight the use of these models to analyze the effect if ed for certain ecosystem services and only reserve of alternative land-use plans on the provision of eco- networks that met these targets would be consid- system services and biodiversity conservation. The ered as potential solutions in the conservation plan- f rst case study involves evaluating alternative future ning exercise. scenarios for land use in the Willamette Basin in Using spatially explicit models that incorporate Oregon. The second case study involves evaluating both ecosystem services and biodiversity is power- alternative land uses for a watershed owned by ful because conservation decisions are often inher- Kamehameha Schools on O`ahu, Hawai`i. ently spatial: Where to protect? How much area is needed? Where to allow development? In this way, 14.2.2.1 Alternative future scenarios in the adding maps of ecosystem services broadens an Willamette Basin, Oregon existing approach to conservation planning that is Nelson et al. ( 2009 ) applied several of the spatially used and understood by many conservation practi- explicit models described in previous chapters, or tioners ( Groves 2003 ). Using spatially explicit mod- their precursors. These models were used to predict els of ecosystem services expands the set of changes in ecosystem services and conservation of APPLYING ECOSYSTEM SERVICE AND BIODIVERSITY MODELS 253

165 km

Oregon Portland US

Salem

Valley FloorAlbany Ecoregion 270 km Corvallis 2050 Plan Trend

Coast Mountains Range Ecoregion

Eugene

Cascades Mountain Range Ecoregion

Orchard/Vineyard Young Conifer Grass Seed Other Forest Pasture/Hayfield Old Conifer/Oth.Natural Row Crops Dense Development/ Bare Ground Rural-Residential

1990 2050 Development

2050 Conservation

Figure 14.3 Maps of the Willamette Basin with the land-use pattern for 1990 and three land-use change scenarios for 2050. Source: Nelson et al. ( 2009 ). 254 PUTTING ECOSYSTEM SERVICE MODELS TO WORK terrestrial vertebrate species for the Willamette tration) from peer-reviewed studies (Tol 2005 ). Basin in Oregon, USA (Figure 14.3 ). Using Because there was more carbon sequestered under stakeholder-def ned land-use change scenarios for the conservation scenario, adding the carbon seques- the period 1990 to 2050, they compared outcomes tration value to the market value of commodities for the basin in terms of carbon storage, water qual- meant that the conservation scenario generated the ity (reduction of phosphorus discharge), soil con- highest monetary returns of the three scenarios servation (reduction of erosion), storm peak (Figure 14.5 ). A carbon market that rewarded carbon mitigation, terrestrial vertebrate conservation and sequestration could turn a trade-off curve with a neg- value of marketed commodities (agriculture, for- ative slope ( Figure 14.5 , circles) into one with a posi- estry and rural residential housing development). tive slope (triangles), converting a trade-off into a Basin-wide maps for the three land-use change sce- win–win. Making payments for other ecosystem narios and the 1990 land-use pattern for the services would further increase the value of the con- Willamette Basin (Figure 14.3 ) were developed by servation scenario relative to the other two scenarios. the Pacif c Northwest Ecosystem Research Consortium, an alliance of government agencies, 14.2.2.2 Kamehameha Schools, O àhu, Hawaiì non-government organizations, and universities A subset of the spatially explicit models described ( Hulse et al. 2002 ; USEPA 2002; Baker et al. 2004 ). in earlier chapters were also used to evaluate The three land-use change scenarios were: (i) “plan impacts on ecosystem services for local land-use trend” that extended current policies and trends planning in Hawaiì in collaboration with into the future, (ii) “development” that relaxed cur- Kamehameha Schools, an educational trust and the rent land-use policies and allowed greater freedom largest private landowner in the state (see Box 14.1 for market forces, and (iii) “conservation” that gave for additional information on Kamehameha greater emphasis to ecosystem protection and resto- Schools). The analysis focused on Kamehameha ration (USEPA 2002, pp. 2–3). Schools’ land holdings on the north shore of the Of the three scenarios, the conservation scenario island of Oàhu ( Figure 14.6 ; Plate 9). This region produces the best results for all ecosystem services contains approximately 26 000 acres stretching from and biodiversity conservation (Figure 14.4 ). The ocean to mountain tops, including ~2000 acres of results for the conservation scenario were signif - coastal rural community lands, ~9000 acres of agri- cantly better than for either the plan trend or devel- cultural lands in the middle section (once a sugar- opment scenarios for carbon sequestration, water cane plantation, now largely abandoned and quality, and soil conservation. Only the market invaded by exotic species), and ~15 000 forested value of commodity production was higher in the acres in the upper part. plan trend and development scenarios than in the With extensive input from Kamehameha Schools, conservation scenario. Under the plan trend and three spatially explicit scenarios were created to development scenarios, more land was devoted to explore contrasting directions that could be taken housing development and to timber production with the agricultural lands: increasing the value of market returns (but lowering the scores for biodiversity conservation and (1) Sugarcane ethanol—returning the plantation many ecosystem services). lands to sugarcane cultivation to produce ethanol The trade-off between the value of marketed com- biofuel; modities and ecosystem services changes if we (2) Diversif ed agriculture and forestry —using the expand the set of marketed commodities to include lower irrigated f elds for diversif ed agriculture, the possibility of markets in carbon credits. Nelson establishing vegetation buffers to reduce f eld run- et al. ( 2009 ) calculated the aggregate market value of off, and undertaking native forestry plantings on carbon sequestration under the three scenarios using the remaining higher elevation f elds; a price of $43 per metric ton of carbon, which is the (3) Residential subdivision —selling coastal and plan- mean of estimates of the social value of carbon reduc- tation lands for a residential housing development. tion (either from emissions reductions or from seques- These scenarios were compared in terms of effects APPLYING ECOSYSTEM SERVICE AND BIODIVERSITY MODELS 255

1.25 0.60 Agricultural, Timber, and Rural-Residential All Commodities and 1.15 Commodities Carbon Sequestration 0.59 1.05 Conservation 0.95 0.58 0.85 0.57 Water Quality Relative Reduction in Ann. Discharge of Dissolved Phosphorus (Unitless) 0.75 1.25 0.56 Development 1.15 Plan trend

1.05 in 2050 Countryside SAR Score 0.55 14.5 15.0 15.5 16.0 16.5 0.95 Net Present Market Value of 1990–2050 0.85 Commodity Production (Billions of US$) Soil Conservation Reduction in Average Annual Rate of Soil Erosion in Short Tons 0.75 Figure 14.5 Trade-offs between market values of commodity 1.25 production and biodiversity conservation on the landscape between 1990 and 2050 excluding the market value of carbon sequestration 1.15 (circles) and including the market value of carbon sequestration (triangles). 1.05 The x -axis measures the total discounted value of commodities and the y -axis measures the biodiversity conservation score. 0.95 Source: Nelson et al. ( 2009 ). 0.85 Storm Peak Management Unitless 0.75 1.25 on water quality (for nitrogen discharge), carbon storage, and income generation. 1.15

1.05 All three scenarios are projected to generate posi-

0.95 tive income streams that exceed the current negative returns ( Figure 14.7 ). The residential subdivision 0.85 scenario, not surprisingly, has the greatest net Metric Tons Carbon Sequestration 0.75 present value of income. This income boost, how- 1.25 ever, is linked with reductions in carbon stock (6.8%) 1.15 and water quality (21.1%) relative to the current 1.05 landscape. Impacts on carbon stock and water qual-

0.95 ity are even more pronounced for the sugarcane ethanol scenario with reductions of 12.6 and 44.2%, 0.85 respectively. In both cases, losses in carbon stock are Countryside Species– Area Relationship (SAR) Biodiversity Conservation 0.75 driven by clearing invasive woody vegetation on 1.25 abandoned f elds. While both scenarios lead to 1.15 reductions in carbon stock, the sugarcane ethanol 1.05 scenario has the potential to “pay off” the lost

0.95 carbon stock through use of ethanol to offset more carbon-intensive energy sources. Following the bio- 0.85 fuel carbon debt methodology of Fargione et al. Market Value of Commodity Production Constant Year 2000 Dollars 0.75 1990 2000 2010 2020 2030 2040 2050 (2008 ), the estimated payback period is approximately 10 years to return to baseline conditions. Plan Trend Development Conservation The remaining scenario, diversif ed agriculture and forestry, is projected to improve carbon stock Figure 14.4 Trends in landscape-scale ecosystem services levels, biodiversity conservation status, and market values of commodity (9.8%) and water quality (7.0%) relative to the cur- production for the three land-use change scenarios. rent landscape, while also generating positive All scores are normalized by their 1990 levels. income. These improvements are driven by plant- Source: Nelson et al. ( 2009 ). ings to restore native forest cover and establishing 256 PUTTING ECOSYSTEM SERVICE MODELS TO WORK

Box 14.1 Plight of a people

Neil Hannahs However, the commercial, residential, and agricultural land developments that brightened KS’ economic prospects Disease and change exacted a horrif c toll on the native were often conducted with insuff cient regard for cultural people of Hawaii throughout the nineteenth century. The resources, environmental impacts and community values. thriving population of more than half a million Hawaiians This tendency, coupled with rapid population growth fueled at the beginning of the century had dwindled to a mere by in-migration and the introduction of invasive exotic 40 000 by the 1880s. To address these desperate conditions species of f ora and fauna, resulted in displacement of and to assure the perpetuation of Hawaiian culture and Hawaiian communities and degradation of indigenous welfare of her people, Princess Bernice Pauahi Bishop and resources and the cultural practices that thrived upon them. her husband left over 400 000 acres of Hawaii land, as well These circumstances produced a tragic conundrum: as personal resources, in a perpetual charitable trust Hawaiians being helped by KS suffered the most from dedicated to improving the wellbeing of Hawaiian people land-use changes implemented to provide resources for through educational services offered by Kamehameha their educational programs. Schools (KS). Concern for Hawaii’s ecosystems and traditional lifestyles Since its inception in 1884, the endowment of KS’ mounted over the past four decades as Hawaiian culture founders has been managed to produce f nancial resources experienced a renaissance and natural resources became to build and maintain campuses and educational programs. increasingly stressed. Resource supply has declined in the For much of the School’s history, the trust was considered face of rising demand and ravaging impacts of invasive land rich, but cash poor. To fund construction, operation plants and ungulates. Conf icts manifested as resistance to and growth, an asset management strategy was adopted to new development and Western concepts of property rights, maximize economic productivity. This provided the means as well as advocacy for constitutional and regulatory for KS to become one of the largest private educational protections of the environment, at-risk species and institutions in the world and afforded the Schools the Hawaiian cultural practices. Consequently, KS’ efforts to opportunity to greatly expand its educational reach. apply economic maximization strategies to undeveloped

An Indigenous Worldview: Focus on Living systems

Waiwai Economics Mauli Education Malama¯ Well Aina¯ Being Environment

Ho‘oulu Kaiaulu¯ Hawaiian Community CultureCommunity Economics

Ho‘ona’auao Education

Stewardship

Ea Culture

Figure 14.A.1 Kamehameha Schools’ representation of the indigenous Hawai’i worldview. APPLYING ECOSYSTEM SERVICE AND BIODIVERSITY MODELS 257

lands faced increasing resistance in the latter twentieth NatCap’s InVEST tool, the software framework for several century. of the models described in earlier chapters, has helped to inform courses of action and land management decisions that propel a shift from one dimensional returns to a balance of Paradigm shift desirable outcomes. KS, owner of the Kawailoa lands to which The “Kamehameha Schools’ Strategic Plan 2000–2015” InVEST is now being applied, has depicted its efforts to promised an organization that would align itself to the achieve an optimal balance of multi-value returns as an image values of the founders, incorporate the views of stakehold- of over-lapping spheres. ers and set new directions. The Plan established the A risk inherent in this view, as well as in using a tool like following goals for the management of the endowment. InVEST, is that the challenge might be met by assembling Kamehameha Schools will optimize the value and use of indiscriminate and disconnected considerations in each the current f nancial and non-f nancial resources and value domain. An alternative approach is to maintain focus actively seek and develop new resources; and to practice on holistic, living systems. This is depicted in the taro (kalo) ethical, prudent and culturally appropriate stewardship of image (Figure 14.A.1 ). lands and resources. Kamehameha Schools is now monitoring several key These goals provide an opportunity to re-think the value performance metrics of sustainability to determine whether of land and each asset’s role in fulf lling the mission as part this high standard is being achieved. These include carbon of a dynamic portfolio. The emergent Integrated footprint; assessments of ecosystem services; f nancial Management Strategy has attracted the interest of cultural values and returns; and various measures of well-being stakeholders and other First Nations peoples, as well as the impact. InVEST is playing an integral role in helping KS and conservation and business communities, including the others in projecting the outcome of land-use decisions on Natural Capital Project (NatCap). many of these indicators of vitality.

f eld buffers to reduce nutrient runoff. As such, methods discussed in Chapter 12 provide a tem- the diversif ed agriculture and forestry scenario has plate for doing so. the greatest potential to provide balanced, positive returns across the modeled services. 14.2.3 Generating an eff ciency frontier Cultural values are also important to the north shore community and to Kamehameha Schools’ In the previous section, we showed how to use mul- approach to land management. While they were not tiple spatially explicit models to analyze specif c sce- assessed quantitatively in this analysis, the scenar- narios (management alternatives) of interest to ios are likely to have differing impacts. Many resi- users. Such analyses can show which of the consid- dents prize the north shore’s rural character, and ered management alternatives generates better per- maintaining active agricultural lands is one key formance in terms of provision of ecosystem services part of this. These lands provide jobs and income to or meeting biodiversity targets. Another use of these the local community, as well as contribute to a sense models is to show what is possible to achieve on the of place and connection with previous generations. landscape by considering all potential land-use sce- These benef ts would be best captured by the sugar- narios. In reality, of course, not all land-use scenarios cane ethanol and diversif ed agriculture and for- will be politically or socially acceptable. But consid- estry scenarios, and not the residential subdivision ering all possible alternatives can often identify scenario. The north shore also contains sacred burial solutions that are far superior to the narrow range of grounds and other historic remains that must be options currently being considered. Providing this considered in land-use planning. Integrating these evidence can broaden the perspective of users and and other cultural dimensions into the formal begin a dialog about what options should be on the modeling effort is an important next step, and the table. 258 PUTTING ECOSYSTEM SERVICE MODELS TO WORK

LULC Scenarios Land Use/Land Cover Classes Undefined Native Shrubland/Sparse Ohia Uluhe Shrubland Closed Ohia Forest Open Koa-Ohia Forest Open Ohia Forest Water Wetland Vegetation Agriculture ¢ Island of O ahu Sugarcane Ethanol Alien Shrubs and Grasses Alien Trees and Shrubs High Intensiyt Developed Low Intensiyt Developed Alien Grassland Alien Shrubland Christmas Berry Shrubland Koa Haole Shrubland Alien Forest Closed Kukui Forest Kiawe Forest and Shrubland Kiawe-Koa Haole Forest and Shrubland Baseline Diversified Agriculture Uncharacterized Forest LULC Map & Forestry Uncharacterized shrubland Very Sparse Vegetation to Unvegetated Koa Reforestation Field Buffer Sugarcane (irrigated) Sugarcane (unirrigated)

Residential Subdivision

Figure 14.6 Land use/land cover maps on the north shore of O’ahu. The area shown here includes all of Kamehameha Schools’ north shore land holdings, as well as small adjacent parcels that make for a continuous region. The baseline map is from the Hawai’i Gap Analysis Program’s land cover layer for O`ahu ( Hawai’i Gap Analysis Program 2006 ). (See Plate 9.)

1.10 Diversified Agriculture & Forestry

1.00 Baseline

0.90 Residential Subdivison Sugarcane

Carbon Stock (Millions of Mg C) Ethanol

–20 02040 60 Plantation Income (Present value in Millions of US$)

Figure 14.7 Projections of carbon stock and income from the plantation lands for the north shore region of O’ahu for the baseline land use/land cover map and the three planning scenarios (sugarcane ethanol, diversif ed agriculture and forestry, residential subdivision). APPLYING ECOSYSTEM SERVICE AND BIODIVERSITY MODELS 259

By combining ecosystem service models with for a given economic score. Then by repeating this optimization methods, one can determine the maxi- analysis across the full range of economic scores mum feasible combinations of ecosystem services ($0–27.6 billion) they traced out an eff ciency fron- and biodiversity that can be achieved on a land- tier ( Figure 14.8 ; Plate 10). The results show that it is scape. The results of this analysis can be presented possible to achieve both high biological and eco- with an eff ciency frontier, which is def ned as the nomic scores by thinking carefully about the spatial outcomes for which it is not possible to improve on pattern of land use in the basin. For example, the any particular objective (ecosystem service or biodi- land-use plan that for point D in Figure 14.8 (Plate versity conservation) without decreasing perform- 10) generates a biological score of 248.5 species and ance on some other objective. an economic score of $25.8 billion. This outcome is P o l a s k y et al. ( 2 0 0 8 ) e s t i m a t e d s u c h a n e f f ciency far better than the outcome generated by the current frontier for conservation of terrestrial vertebrates land use (point I in Figure 14.8 ; Plate 10), a biological and the value of marketed commodities (timber, score of 238.6 and an economic score of $17.1 billion agricultural output and housing) for the Willamette ( Polasky et al. 2 0 0 8 ) . Basin in Oregon. They developed models that used Analyses such as these can demonstrate what is a land-use plan for the basin as input and reported possible for a given region and how much output in terms of the expected number of terrestrial improvement can be made by careful planning. vertebrate species that would persist in the basin Because of political, social and economic compli- (biological score) and the value of marketed com- cations, it may not be possible to reach eff ciency modities (economic score). Using optimization frontiers. Still, knowing what it is possible can pro- methods from operations research, they searched for vide a spark to ignite efforts to improve upon cur- land-use plans that maximized the biological score rent performance.

257 H G F E

Agriculture Rural-Residential 246 Managed Forestry Conserved C UGB B I

235 Expected Number of Species A

224 0 5 10 15 20 25 30 Billions of Dollars

Figure 14.8 Eff ciency frontier showing maximum feasible combinations of economic returns and biodiversity scores. Land-use patterns associated with specif c points along the eff ciency frontier (points A–H) and the current landscape (point I). (See Plate 10.) Source: Polasky et al. ( 2008 ). 260 PUTTING ECOSYSTEM SERVICE MODELS TO WORK

Examples include the value of provisioning services 14.2.4 Benef t–cost analysis such as timber or f sh (e.g., Naidoo and Ricketts Results from ecosystem service models can be 2006 ; Barbier 2007 ; Polasky et al. 2008 ; Nelson et al. reported in biophysical units or in monetary values. 2008 , 2009 ; Chapter 8 ), or crop pollination, which is Much of the analysis on ecosystem services to date an input to a priced commodity (e.g., Ricketts et al. has been reported in biophysical units, including 2004 ; Chapters 9 , 10 ). Depending on the ecosystem most of the case studies discussed above. In some services and the decision context at hand, the user settings, such as dealing with government or private of these models can decide whether it is better to sector managers used to thinking in monetary use biophysical units or monetary values. terms, it may be advantageous to report results of Chapters in this book typically aim to monetize the analysis in terms of monetary values. Doing so the value of ecosystem services, but we do not may also make it easier to compare management attempt to translate biodiversity targets to a mone- options. Because results are reported in a single tary measure of value (Chapter 13 ). That is because metric (i.e., dollars), managers can compare apples biodiversity is a fundamental attribute of natural with apples rather than with oranges. systems, which may contribute to the provision of Economists have developed a variety of market various ecosystem services but which also has and non-market valuation methods that can be intrinsic value (i.e., value in and of itself). Even applied to estimate the monetary value of ecosys- without attempting to put monetary value on biodi- tem services (Freeman 2003 ). The estimates of mon- versity, one can still show feasible combinations of etary value can be incorporated into benef t–cost biodiversity and services, along with potential analysis to analyze the net benef ts of alternative trade-offs between them (as shown in Section 14.2.3 ). management alternatives. Naidoo and Ricketts Then managers can decide for themselves what ( 2006 ) in Section 14.2.1 and Nelson et al. ( 2009 ) in trade-offs are acceptable. Section 14.2.3 are examples of how monetizing ecosystem service values can result in benef t–cost 14.3 Extending the frontier: challenges analyses that can inform managers and potentially facing ecosystem management improve management decisions. Translating from biophysical units to monetary Integrated landscape-level analysis that tracks value units, however, is problematic for biodiver- changes across a number of dimensions of ecosys- sity targets and some types of ecosystem services. tem services and biodiversity conservation is still a In some cases trying to convert oranges into apples relatively young discipline. Models of ecosystem will result in pulp rather than a recognizable fruit. services and geographically explicit data sets are For example, trying to estimate the monetary value developing rapidly, offering the prospect of fur- of cultural and spiritual values is controversial (see ther improvements in the near future. To date, Chapters 2 and 12 ; Norton 1991 ; Sagoff 1988 ). applications in the USA, South Africa, Paraguay, Valuing the existence of species is viewed as mor- and elsewhere have demonstrated the power and ally objectionable and inherently misguided by utility of an integrated spatially explicit landscape- some (e.g., Ehrenfeld 1988 ; McCauley 2006 ), and level approach. Application of such models can even some economists who have tried to value bio- generate information for decision-makers showing diversity admit to the practical diff culties of doing the consequences of choices for a range of impor- so (e.g., Stevens et al. 1991 ). Other economists think tant ecosystem services and biodiversity conserva- that all values, including the value of biodiversity, tion objectives. In principle, putting this can be measured using economic methods as long information in the hands of decision-makers as the analysis is done properly (e.g., Loomis and should lead to improved landscape planning and White 1996 ). For some ecosystem services, estimat- management. ing monetary values using market prices or apply- To fully realize the promise of spatially explicit ing non-market valuation techniques may be integrated modeling approaches, further improve- relatively uncomplicated and uncontroversial. ments will be necessary. As discussed in Chapter 15 , EXTENDING THE FRONTIER: CHALLENGES FACING ECOSYSTEM MANAGEMENT 261 more work on improving and validating the com- total value of ecosystem services to society but also ponent models of particular services is needed. Our the distribution of benef ts to various groups in understanding of the links between management society. Such distributional analysis is important for actions and provision of ecosystem services is lim- understanding the effects of conservation and man- ited for many services. Additional empirical agement decisions on the poor (see Chapter 16 ). research on provision of services in a wide variety Distributional analysis of the people who benef t of circumstances will improve understanding and and bear the costs of alternative conservation and accuracy of models. Additional understanding of management is also important for the design of ecosystem functions and conditions that link policy approaches to ensure that those who make together provision of multiple services, such as con- decisions affecting ecosystems have incentives to nections between land cover, water availability, provide ecosystem services of value to society (see nutrient cycling and local climate, will also improve Chapter 19 ). the overall modeling effort. Perhaps the greatest It is important to recognize that estimates of need on the biophysical modeling side, however, is value, spatial priorities, trade-off analyses, and improved understanding and inclusion of system most other results reviewed in this chapter depend dynamics and feedback effects. Coupled human strongly on the choice of ecosystem services to and natural systems may exhibit threshold effects include. Ecosystem management affects a large and non-linear responses in which provision of eco- range of ecosystem services, not all of which may be system services might change suddenly as condi- feasible to model given limited time, resources, data tions in the system evolve. or scientif c understanding. In many cases, water Even with knowledge of biophysical systems, quantity and quality, carbon sequestration, and the understanding the provision of ecosystem services market value of commodities will be of great impor- also requires detailed understanding of what is of tance. Biodiversity conservation will be of primary value to people. For example, the provision of clean importance in many conservation applications. drinking water in areas without people will not However, other services may also be important in provide an ecosystem service of value while the particular applications (e.g., non-timber forest prod- same provision in a watershed providing water to a ucts, pollination services, effects on poverty, number major city will have great value. Understanding the of jobs). Early and continuous engagement with value of ecosystem services requires integration of people potentially impacted by ecosystem manage- natural and social science. Such integrated work, ment is the best approach to ensuring that the most partly in response to the focus on ecosystem serv- important ecosystem services and other policy ices, has begun to expand rapidly in recent years dimensions (e.g., number of jobs) are included in but is still limited relative to what is needed to the analysis. seamlessly integrate the supply of services (prima- Finally, the analysis of integrated spatially explicit rily the province of natural science) with the demand models is but one step in a much larger and longer for services (primarily the province of social process needed to implement real change on the science). Integrated understanding of ecosystem ground. As Knight et al. ( 2006 ) and Cowling et al. services has progressed to the point where we can ( 2008 ) emphasize, there are plenty of analyses and highlight important areas on a landscape for eco- reams of plans but far less action, and that “our system services. In many cases, however, we cannot understanding of these techniques currently far yet provide the level of certainty, either in terms of exceeds our ability to apply them effectively to biophysical or economic modeling, to underpin pragmatic conservation problems” ( Knight et al. payments for ecosystem services or other policy 2006 , p. 408). Spatially explicit integrated models approaches that require numerical estimates of can provide useful information but unless they are value (see Chapters 15 and 19 for further embedded in a larger policy process that involves discussion). those who use land and resources the information An important aspect of integrated spatially will not be utilized to improve ecosystem manage- explicit models is the ability to show not only the ment or conservation outcomes. 262 PUTTING ECOSYSTEM SERVICE MODELS TO WORK

References Hawai’i Gap Analysis Program. (2006). Land cover . US Geological Survey, Honolulu. Ando, A., Camm, J. D., Polasky, S., et al. (1998). Species Hulse, D., Gregory, S., and Baker, J., Eds. (2002). Willamette distributions, land values and eff cient conservation. River Basin Planning Atlas: trajectories of environmental Science , 279 , 2126–8. and ecological change. Oregon State University Press, Baker, J. P., Hulse, D. W., Gregory, S. V., et al . (2004). Corvallis. Alternative futures for the Willamette River Basin, Jackson, R. B., Jobbagy, E. G., Avissar, R., et al . (2005). Oregon. Ecological Applications , 14 , 313–24. Trading water for carbon with biological carbon seques- Barbier, E. B. (2007). Valuing ecosystem services as pro- tration. Science , 310 , 1944–7. ductive inputs. Economic Policy , 22 , 177–229. Kirkpatrick, J. B. (1983). An iterative method for establish- Cabeza, M., and Moilanen, A. (2001). Design of reserve ing priorities for the selection of natural reserves: an networks and the persistence of biodiversity. Trends in example from Tasmania. Biological Conservation , 25 , Ecology and Evolution , 16 , 242–8. 127–34. Camm, J., Polasky, S., Solow, A., et al . (1996). A note on Knight, A. T., Cowling, R. M., and Campbell, B. M. (2006). optimization algorithms for reserve site selection. An operational model for implementing conservation Biological Conservation , 78 , 353–5. action. Conservation Biology , 20 , 408–19. Chan K. M. A., Shaw, M. R., Cameron, D. R., et al . (2006). Loomis, J. B., and White, D. S. (1996). Economic benef ts of Conservation planning for ecosystem services. PLoS rare and endangered species: summary and meta-anal- Biology , 4 , 2138–52. ysis. Ecological Economics , 18 , 197–206. Chichilnisky, G., and Heal, G. (1998). Economic returns McCauley, D. (2006). Selling out on nature. Nature , 443 , from the biosphere. Nature , 391 , 629–30. 26–7. Cocks, K. D., and Baird, I. A. (1989). Using mathematical Margules, C. R., Nicholls, A. O., and Pressey, R. L. (1988). programming to address the multiple reserve selection Selecting networks of reserves to maximize biological problem: an example from the Eyre Peninsula, South diversity. Biological Conservation , 43 , 63–76. Australia. Biological Conservation , 78 , 113–30. Margules, C. R. and Pressey, R. L. (2000). Systematic con- Costello, C., and Polasky, S. (2004). Dynamic reserve site servation planning. Nature , 405 , 242–53. selection. Resource and Energy Economics , 26 , 157–74. Meir, E., Andelman, S., and Possingham, H. P. (2004). Does Cowling, R. M., Egoh, B., Knight, A. T., et al . (2008). An conservation planning matter in a dynamic and uncer- operational model for mainstreaming ecosystem serv- tain world? Ecology Letters , 7 , 615–22. ices for implementation. Proceedings of the National Naidoo, R., Balmford, A., Ferraro, P. J., et al . (2006). Academy of Sciences , 105, 9483–8. Integrating economic costs into conservation planning. Echevarrıa, M. (2002). Financing watershed conservation: Trends in Ecology and Evolution , 21 , 681–7. The FONAG water fund in Quito, Ecuador. In: S. Naidoo, R., and Ricketts, T.H. (2006). Mapping the eco- Pagiola, J. Bishop, and N. Landell-Mills, Eds., Selling for- nomic costs and benef ts of Conservation. PLoS Biol , 4 , est environmental services: market-based mechanisms for 2153–64. conservation and development, pp. 91–102. Earthscan, National Research Council (NRC). (2000). Watershed man- London. agement for potable water supply: assessing the New York Ehrenfeld, D. (1988). Why put a value on biodiversity? In: City strategy . National Academies Press, Washington, E. O. Wilson, Ed., Biodiversity , pp. 212–16. National DC. Academy Press, Washington, DC. Nelson, E., Polasky, S., Lewis, D. J., et al . (2008). Eff ciency Fargione, J., Hill, J., Tilman, D., et al . (2008). Land clearing of incentives to jointly increase carbon sequestration and the biofuel carbon debt. Science , 319 , 1235–8. and species conservation on a landscape. Proceedings of Freeman, A. M. III. (2003). The measurement of environmen- the National Academy of Sciences , 105 , 9471–6. tal and resource values . Resources for the Future, Nelson, E., Mendoza, G., Regetz, J., et al . (2009). Modeling Washington, DC. multiple ecosystem services, biodiversity conservation, Goldman, R. L., Tallis, H., Kareiva, P., et al . (2008). Field commodity production, and tradeoffs at landscape evidence that ecosystem service projects support biodi- scales. Frontiers in Ecology and the Environment , 7 , 4–11. versity and diversify options. Pr oceedings of the National Nicholson, E., Westphal, M. I., Frank, K., et al . (2006). A Academy of Sciences of the USA 105 (27), 9445–8. new method for conservation planning for the persist- Groves, C. R. (2003). Drafting a conservation blueprint: a ence of multiple species. Ecology Letters , 9 , 1049–60. practitioner’s guide to planning for biodiversity . Island Norton, B. G. (1991). Toward unity among environmentalists . Press, Washington, DC. Oxford University Press, New York. EXTENDING THE FRONTIER: CHALLENGES FACING ECOSYSTEM MANAGEMENT 263

Polasky, S., Nelson, E., Camm, J., et al . (2008). Where to put Stevens, T. H., Echeverria, J., Glass, R. J., et al . (1991). things? Spatial land management to sustain biodiver- Measuring the existence value of wildlife: what do CVM sity and economic returns. Biological Conservation , 141 , estimates really show? Land Economics 67 , 390–400. 1505–24. Tol, R. S. J. (2005). The marginal damage costs of carbon Possingham, H. P., Ball, I. R., and Andelman, S. (2000). dioxide emissions: an assessment of the uncertainties. Mathematical methods for identifying representative Energy Policy , 33 , 2064–74. reserve networks. In: S. Ferson and M. Burgman, Eds., US Environmental Protection Agency (USEPA). (2002) Quantitative methods for conservation biology , pp. 291–305. Willamette Basin alternative futures analysis : environmen- Springer-Verlag, New York. tal assessment approach that facilitates consensus building . Ricketts, T. H., Daily, G. C., Ehrlich, P. R., et al . (2004). EPA 600/R-02/045(a). US Environmental Protection Economic value of tropical forest to coffee production. Agency, Off ce of Research and Development, Proceedings of the National Academy of Sciences , 101 , Washington, DC. 12579–82. Willis C. K., Lombard, A. T., Cowling, R. M., et al . (1996). Sagoff, M. (1988). The economy of the earth . Cambridge Reserve systems for the limestone endemic f ora of the University Press, New York. Cape lowlands: iterative vs linear programming tech- Sarkar, S., Pressey, R. L., Faith, D. P., et al . (2006). niques. Biological Conservation , 77 , 53–62. Biodiversity conservation planning tools: present status Wilson, K. A., McBride, M. R., Bode, M., et al . (2006). and challenges for the future. Annual Review of Prioritizing global conservation efforts. Nature , 440 , Environment and Resources , 31 ,123–59. 337–40.

CHAPTER 15 How much information do managers need? The sensitivity of ecosystem service decisions to model complexity

Heather Tallis and Stephen Polasky

15.1 Introduction plexity are needed for different types of questions. For example, in water resource management, sim- Natural systems and human systems are inherently ple models have proven suitable for predicting run- complex. When they are examined as coupled sys- off in relatively dry watersheds (Gan et al. 1997 ). tems, which is necessary for the valuation of ecosys- Similarly, in f sheries and wildlife management, set- tem services, the complexity can be daunting. ting optimal f shing effort levels in systems with Because it is usually impractical to apply experi- relatively little variability (Ludwig and Walters ments to human systems, models represent the pri- 1985 ) and identifying area requirements for sustain- mary tool for studying coupled human-natural ing individual species (such as the marmot (Marmota systems. We use models in two ways: (1) to simplify marmot ; Stephens et al. 2002 )) are exercises well the complexity and extract the key dynamics, and addressed with simple models. However, more (2) to run simulations or experiments that can only complex models may be required for other manage- be done on a computer, but that are necessary for ment decisions, such as accurately predicting the anticipating the consequences of different policies consequences of harvesting when species interact and different economic activities. or predicting the effect of climate change on species Although models can never fully represent the persistence ( Stephens et al. 2002 ). intricacies of the real world, they can yield insights Understanding when simpler models can be used into how systems work. Or, as modeler George Box is critical to managers since they often function with famously put it, “All models are wrong but some short deadlines and limited information (Box 15.1 ). are useful.” Modelers of ecosystem services face the Simple models typically have reduced data require- same challenge all modelers face: deciding how ments, are easier to set up and run and are more much complexity and detail to include. The major transparent and easier to explain than more com- question is how to create models that are suff ciently plex models. If simple models give the same answers complex to represent system dynamics, yet simple as more complicated models, they would obviously enough to be understood and appropriately param- be the preferred option. Elsewhere (e.g., Chapter 4 ) eterized with often limited data (see Van Nes and we compare models to actual f eld data. Here we Scheffer 2005 for a discussion of the issue). This bal- focus on comparing models to models, asking to ancing act is a long standing dilemma in f elds as what extent simpler models are able to do the same diverse as hydrology (e.g., Gan et al. 1997 ) and pop- things and give similar answers to more data- ulation biology (e.g., Pascual et al. 1997 , Stephens hungry, complex models. To supplement our gen- et al. 2002 ). eral discussion we focus in particular on the There is no formulaic recipe for determining the correspondence between simple and complex mod- degree of complexity and detail needed. In practice, els in providing information about carbon seques- researchers have found that different levels of com- tration and crop pollination.

264 INTRODUCTION 265

Box 15.1 How much data do we need to support our models: a case study using biodiversity mapping and conservation planning

Other studies using different data have reached a similar Craig Groves and Edward Game conclusion. For example, in a widely cited study, Gaston and Governments and non-governmental organizations Rodrigues ( 2003 ) used data on bird distribution and worldwide commonly develop regional-scale conservation abundances from a South African bird atlas to evaluate plans that identify areas important for biodiversity different levels of data (abundance data, presence/absence conservation. The amount of data—both biodiversity data data, low sampling effort, no data) on the design of reserve and other important related information—is highly variable networks. Their results showed that even low sampling around the world. A few areas have been well surveyed for efforts can be reasonably effective at representing bird elements of biodiversity but most areas have limited data species in a network of reserves, suggesting that reserve sets. As a result, planners are often confronted with the designs based on methods of complementarity can be challenging issue of deciding how many data are needed effective in regions of the world with limited biological data. and whether or not to spend resources gathering additional The one disadvantage of fewer data that stands out is a information. less eff cient design of conservation areas, meaning that There are at least two reasons why additional invest- fewer options are identif ed that can achieve conservation ments in biodiversity data may not deliver a commensurate and that it often takes a greater amount of area to achieve increase in the quality of the planning effort. First, the the same conservation result. For example, a study by selection of priority areas for conservation is usually driven Grand et al. ( 2007 ) using the same data set as Grantham by many factors in addition to biodiversity information. For and colleagues reached this very conclusion. As a result, example, active use of lands and waters for production that there is a trade-off that planners must balance between a forecloses the opportunity for conservation, degraded reduction in eff ciency of the planning effort and the costs habitats or ecosystem processes of available lands and of getting better information. waters, or prohibitively high costs of conservation due to Can results from these sorts of studies be generalized escalating land prices can be overriding factors that restrict more broadly? That is diff cult to know with conf dence. The options without much need for complete biodiversity data. distribution and quantity of available data, assumptions The second and more obvious reason why additional about land-use change, the timing and amount of new data may not be worth the investment is related to the conservation areas established, the costs of implementa- time and money needed to acquire the additional tion, and a host of other factors all affect the outcomes of information as well as conservation opportunities that may conservation plans ( Grantham et al. 2008 ). At the very be lost while this information is being gathered. Several least, these efforts suggest that the amount of biodiversity studies, mostly from South Africa, have evaluated the data used in conservation plans may not be the limiting inf uence of amounts of data on conservation planning factor that we once thought it was. results. Perhaps the best example is a recent paper by A related but equally important question is whether Grantham and colleagues ( 2008 ) who took advantage of relatively simple plans with limited data can have an extensive and nearly complete data set on the conservation impact. Evidence, again from South Africa, distribution of Protea species in South Africa’s Fynbos clearly suggests that simple, straightforward plans can biome. They asked the question of whether additional indeed have conservation impact. Reyers et al. ( 2005 ) investments in surveys, mapping, and modeling would lead prepared a biodiversity prof le and priority assessment of to better planning decisions given the costs of undertaking South Africa’s grasslands, a large biome in eastern South these surveys and the fact that ongoing land clearing Africa. Like most grasslands worldwide, South Africa’s are would continue to lead to some losses while the surveys under-represented in protected areas, highly threatened, were carried out. Their results clearly showed that while and easily converted to farmland. Their assessment, which investing in minimal survey data can improve a plan aimed to integrate knowledge about this large area and substantially, especially if there is little biodiversity data use it to direct investment in mainstreaming conservation available to begin with, there are rapidly diminishing efforts in the grassland, was completed in just two months returns in additional data ( Figure 15.A.1 ). without collecting additional data (B. Reyers, personal continues 266 HOW MUCH INFORMATION DO MANAGERS NEED?

Box 15.1 continued 1

0.95

0.9

0.85

Representation in P.A.s 0.8

0.75 $0 $1,000,000 $2,000,000 $3,000,000 Initial Survey Investment

Figure 15.A.1 Representation of Protea species in protected areas in the Fynbos biome, South Africa, as a function of investments in biological surveys. Adapted from Grantham et al. ( 2008 ).

communication), but capitalizing on data that had been assessments conducted on agriculture, economics, urban gathered as part of a National Spatial Biodiversity development and the mining sector, it helped to provide a Assessment ( Driver et al. 2005 ). It was based on analyses foundation for the implementation phase, galvanizing of three different types of priority areas: terrestrial stakeholder buy-in into what needs to be done and where, biodiversity areas, river biodiversity areas, and ecosystem ensuring commitment by the national government to the service priority areas. program, and securing additional funding from national National-level data on ecosystem status, threatened agencies. To what factors can we attribute this success? species status, and critical ecological processes were used First, this was the f rst time that knowledge, data and in a summary fashion to def ne terrestrial biodiversity areas. expertise from across the biome were brought together into Simple measures of river integrity such as the percentage of a single assessment through workshops, data collation and a catchment containing natural land cover were used to interviews. Second, this assessment benef tted from the identify river biodiversity areas. Ecosystem services were lessons learned in the National Spatial Biodiversity mapped for water production, soil protection, carbon Assessment and other regional plans about the value of sequestration, grazing and services supporting harvestable using simple, visual metrics of terrestrial biodiversity, products to aid in pinpointing ecosystem service priority freshwater biodiversity, and ecosystem services which are areas. As a f nal step, priority “clusters” of catchments for easy to understand and interpret, user friendly, and grassland conservation were identif ed based largely on a compelling to their target audience ( Pierce et al. 2005 ; summing and rescaling of priority areas for terrestrial Reyers et al. 2007 ). Finally, the assessment learned from biodiversity, freshwater biodiversity, and ecosystem services. the lessons of the conservation planning community in Prof les of these clusters were developed in order to identify South Africa that data collation and assessment, although the opportunities and constraints for mainstreaming of critical importance, are very small components of the conservation into the sectors (e.g., agriculture, mining or long and complicated process of implementation, and forestry) within each cluster. should therefore not dominate the resources available This grassland assessment, based largely on existing ( Cowling et al. 2004 ). A tentative conclusion we may take data and developed over a short-time period, has resulted from these South African planning efforts is that not only in coordinated and strategic investment in conservation in are more and better data not always necessary and helpful, the grassland biome (A. Stephens, personal communica- but how data are translated and conveyed to key tion). This includes directing the selection of and investment stakeholders and decision-makers is every bit as important in demonstration project sites. Together with other as the amount of data. INTRODUCTION 267

Housing and Territorial Development is establishing 15.1.1 Predicting ecosystem service provision a system through which they will assess all major levels sectoral projects (mining, agriculture, transportation, There are two major types of ecosystem service oil and gas) for the next 5 years to identify how much model outputs that are useful to decision-makers: (1) ecosystem service and biodiversity loss is expected predictions regarding the aggregate or overall quan- and consequently what level of mitigation should be tity or value of ecosystem services provided under required. different scenarios and (2) the spatial distribution of Governments may also face choices over the kind ecosystem services across the landscape. The type of of management action to take to reach a regulatory outputs needed depends on the type of decisions goal. Knowing how much of a service each action being made and the institutional framework for the will yield can help managers choose among differ- decisions (see Table 15.1 ). In general, quantities or ent actions. For example, the city of Santiago (Chile) values of ecosystem services are useful for answer- has some of the worst particulate matter air pollu- ing the question of “how much?” Government regu- tion in Latin America. They used a production func- lators responsible for allocating permits or requiring tion model to determine how much air f ltration offsets need to know how much of each particular would be provided by urban forests. This informa- service is being generated or impacted by actions of tion was used in a cost-effectiveness analysis that regulated entities ( Table 15.1 ). For example, in one of revealed that urban forest management was a more the f rst major cap-and-trade programs, the US cost-effective pollution control approach than other Environmental Protection Agency required continu- methods such as using alternative fuels (Escobedo ous emissions monitoring that recorded the level of et al. 2008 ). Their ability to identify how much of a

SO2 emissions for all large point sources. This infor- service different management actions would yield mation was essential for knowing if f rms had the helped them make the most eff cient decision. required number of permits to cover their emissions. Conservation priorities are also shaped by know- The Colombian Ministry of the Environment, ing how much ecosystem service or biodiversity return

Table 15.1 Types of ecosystem service information used to inform management by diverse decision-makers

Decision-maker Ecosystem service trait

Spatial distribution Absolute level of production

Regulator of permits, mitigation and • Determine where offsets would be most • Determine how many permits or offsets are required offsets eff cient • Inform calculation of offset ratio • Determine how much area is required to meet offsets Regulator of subsidies or fees • Determine where to target programs for • Determine levels of payment or fee required most eff cient outcomes Conservation planner • Determine where investments would be • Design action plans to meet quantitative goals most eff cient • Identify win–win areas where services overlap with biodiversity Corporation • Determine where investments would be • Determine level of impacts that require offsets or fees or level of most eff cient services created that generate permits or payments Financial institution • Determine where investments would be • Determine likely level of return on investment most prof table and least risk prone Designer of payment for ecosystem • Determine where payments would be • Determine level of payment service program most eff cient 268 HOW MUCH INFORMATION DO MANAGERS NEED? they can be expect from different geographic patterns doubles the required mitigation ratio (e.g., from 2 of protection or restoration ( Box 15.1 ). The Nature acres restored/protected: 1 acre damaged to 4:1) Conservancy is looking at this question in the (ACOE 2004). An obvious goal for conservation is Willamette Basin in Oregon, by asking how much eff ciency which translates into identifying where ecosystem service return they are likely to get if they conservation activities should be distributed across are successful in protecting their entire portfolio of a landscape to get the greatest possible biodiversity conservation priority sites. Corporations trying to and ecosystem service returns (Box 15.1 ). Similarly, meet either internally set goals or government regula- regulators developing new subsidy or fee programs tions need to know how much offsetting their pro- can achieve eff ciency by considering where in space posed activities will cost for budgeting and strategic to target payments or levy charges (e.g., Nelson planning. BC Hydro, the third largest electric utility et al. 2008 ). in Canada, has set an internal goal to become “environment neutral” in the next 20 years, meaning that 15.2 Testing agreement between simple they intend to reduce environmental impacts as much and complex ecosystem service models as possible and then offset unavoidable damages. To achieve this goal, they need to know how much In statistical theory it is relatively straightforward impact their activities will have on biodiversity and to compare the performance of models when the ecosystem services so that they can reduce these models are nested in terms of complexity, the easi- effects or mitigate them elsewhere. est form of which is a multiple linear regression In general, when the amount of an ecosystem with a succession of possible additional predictors. service is of primary interest, then questions of However, deciding between simple versus complex appropriate model complexity should focus on models is more diff cult when the alternatives are comparing levels of ecosystem services. not nested and use different data sets (Pascual et al. 1997 ), as is the case with most of our tier 1 (simple) and tier 2 (complex) models. Here we focus on 15.1.2 Predicting ecosystem service patterns assessing how well the simple tier 1 models agree in space with more complex tier 2 models, and if models Ecosystem services are not delivered or valued uni- disagree, how using predictions from one particular formly across landscapes. Certain locations harbor model are likely to inf uence management more biodiversity, or provide more carbon storage decisions. or other ecosystem services than other locations. Delivery of some services, such as f ood control or 15.2.1 Carbon stock and carbon sequestration provision of clean drinking water, is more valuable if it takes place near large cities. In general, the The tier 1 and tier 2 models for estimating carbon effectiveness of any management action aimed at stocks both predict the amount of carbon stored in a delivering an ecosystem service will be determined land use and land cover type on the landscape in by the location of management activity. This means Mg carbon ha-1 (see Chapter 7 for details). Both tier that mapped predictions of the relative value of dif- 1 and tier 2 models can also predict the amount or ferent locations in terms of ecosystem service provi- value of carbon sequestered over time (Mg carbon sion can be very useful when implementing ha -1 yr -1 or $US (or other currency) ha-1 yr -1 ). The two investments, incentives or regulations. For exam- models differ in their treatment of carbon storage as ple, in the United States, the spatial orientation of vegetation grows and decays, and in their treatment damaged areas and corresponding mitigation areas of carbon changes with land use and cover transi- is used to determine the required mitigation ratio tions. The simpler tier 1 model assumes that all par- (specif ed as how many acres of restored or pro- cels within a specif ed land use and cover type hold tected habitat are required for each acre damaged). the same amount of carbon, regardless of the age of For example, under the Clean Water Act, selecting a vegetation. The tier 1 model also assumes that car- mitigation site outside the impacted watershed bon sequestration rates in a given land cover class TESTING AGREEMENT BETWEEN SIMPLE AND COMPLEX ECOSYSTEM SERVICE MODELS 269 are not related to previous land-use practices or values than the tier 1 model, which was expected cover types, or to the time since transition. Tier 1 is since the tier 2 model accounts for changes in car- thus most appropriate for analysis of average or bon storage with age while the tier 1 model does equilibrium “steady-state” values of carbon stor- not. As forests mature or carbon builds up in soils age. The tier 2 model uses a coeff cient to adjust the over decades, estimates of their carbon stocks stay amount of carbon stored in a land cover type for its static in the tier 1 model, while they move closer to age, relative to the maximum possible storage of the maximum possible storage capacity of each land carbon in that land cover class. The tier 2 model also cover type in the tier 2 model. Managers interested accounts for the previous land use and cover type in using our models to track carbon sequestration and the amount of time that has passed since the where land use and cover types have changed in transition to the current type occurred. the recent past, or may change in the future, should use the tier 2 model if possible. In the Willamette 15.2.1.1 Predicting carbon storage and application, conservation planners or landowners sequestration levels using the tier 1 model to report the contribution of Predicting the total amount of carbon stored or their holdings can be conf dent that they are likely sequestered by a parcel can be important in the presenting conservative estimates. design of payment for ecosystem service programs or in the process landowners or corporations go 15.2.1.2 Predicting carbon sequestration patterns through to determine whether they should partici- in space pate in voluntary carbon markets. Carbon storage Information describing the spatial con f guration of or sequestration levels may also be useful when carbon storage or sequestration is useful when inves- companies or organizations are trying to assess the tors, governments, corporations or conservation value of their current land assets. For instance, The groups are trying to identify the best places for action, Nature Conservancy or Exxon-Mobil may want to or the best course of action. Policy makers designing know how much carbon sequestration benef t is new subsidy or fee programs, or conservation groups being provided by lands currently under their choosing new sites for action, may want to know management. where carbon will be sequestered or lost as a result of We used the application of both tier 1 and tier 2 a proposed management program. We asked how models in the Willamette Basin, Oregon (USA) to well the tier 1 and tier 2 models agree on these kinds ask whether these differences in model complexity of spatial questions by taking the difference between lead to important differences in the prediction of the carbon storage maps produced by the two mod- absolute levels of carbon storage and sequestration els for the Willamette Basin (USA) ( Figure 15.2 ). We on the landscape for conditions in 1990 and three found differences in the prediction of carbon storage future land-use scenarios. All maps used to represent for initial conditions in the 1990 landscape. These dif- land use and land cover future scenarios were ferences, however, show no clear pattern and seem to developed by a multi-stakeholder alliance (Hulse be fairly randomly distributed across the landscape. et al . 2002 ; US EPA 2002; Baker et al. 2004 ) that By 2050, the age class and landscape transition con- scripted three different management pathways: (1) siderations included in the tier 2 model lead to much conservation, where greater emphasis is placed on higher estimates of carbon storage in forested areas ecosystem protection and restoration, (2) develop- than those given by the tier 1 model (especially for ment, where market forces are given greater free the conservation scenario in 2050; Figure 15.2 ). Tier 2 rein across all components of the landscape, and shows lower estimates for carbon storage in con- (3) planned trend, where current policies are imple- verted areas. mented as written and recent trends continue. The type and age of land use and land cover classes In all cases, there was close agreement between included in the scenario being examined affected the simple and complex models ( Figures 15.1 a, b, c). how much the models disagreed. This can be seen As predictions progressed farther into the future, most clearly by comparing the estimates of carbon the tier 2 model predicted higher carbon storage sequestration (or change in carbon storage) over time 270 HOW MUCH INFORMATION DO MANAGERS NEED?

(a) 1000 1000

800 800

600 600

400 400

200 200

Tier 2 2000 Carbon Stock 0 Tier 2 2010 Carbon Stock 0 0 200 400 600 800 1,000 0 200 400 600 800 1,000 Tier 1 2000 Carbon Stock Tier 1 2010 Carbon Stock 1000 1000

800 800

600 600

400 400

200 200

Tier 2 2020 Carbon Stock 0 Tier 2 2030 Carbon Stock 0 0 200 400 600 800 1,000 0 200 400 600 800 1,000 Tier 1 2020 Carbon Stock Tier 1 2030 Carbon Stock 1000 1000

800 800

600 600

400 400

200 200

Tier 2 2040 Carbon Stock 0 Tier 2 2050 Carbon Stock 0 0 200 400 600 800 1,000 0 200 400 600 800 1,000 Tier 1 2040 Carbon Stock Tier 1 2050 Carbon Stock

Figure 15.1 The relationship between the tier 1 and tier 2 estimates of carbon stock (MgC ha-1) in the Willamette Basin (USA) for the conservation scenario (a), development scenario (b). and plan trend scenario (c) across decades from 1990 to 2050. The solid line shows a 1:1 relationship between tier 1 and tier 2 results. Data points are values for individual parcels in the Willamette Basin. The models agree relatively well over short time projections, but the tier 2 model consistently predicts higher levels of carbon storage than the tier 1 model at mid- to high carbon storage values as time progresses.

between 1990 and 2050 under each scenario for the Managers interpreting spatial maps of carbon two tiers ( Figure 15.3 ). The models show the greatest storage estimates should be aware that the tier 2 disagreement in prediction of change under the model will depict a more heterogeneous landscape, development scenario where some forested areas are with bigger differences in carbon stocks among allowed to grow to maturity, yielding higher tier 2 land cover classes and thus, bigger apparent differ- estimates, and where other forested areas and some ences in the opportunities for investment. With the agricultural lands are converted to commercial tim- tier 2 model, old forests will store much more car- ber harvest and housing development, yielding bon than young agricultural lands, whereas the dif- lower tier 2 estimates of change. ference will not be so stark in the tier 1 model. In TESTING AGREEMENT BETWEEN SIMPLE AND COMPLEX ECOSYSTEM SERVICE MODELS 271

(b) 1000 1000

800 800

600 600

400 400

200 200 Tier 2 2010 Carbon Stock Tier 2 2000 Carbon Stock 0 0 0 200 400 600 800 1,000 0 200 400 600 800 1,000 Tier 1 2000 Carbon Stock Tier 1 2010 Carbon Stock 1000 1000

800 800

600 600

400 400

200 200

Tier 2 2020 Carbon Stock 0 Tier 2 2030 Carbon Stock 0 0 200 400 600 800 1,000 0 200 400 600 800 1,000 Tier 1 2020 Carbon Stock Tier 1 2030 Carbon Stock 1000 1000

800 800

600 600

400 400

200 200 Tier 2 2040 Carbon Stock 0 Tier 2 2050 Carbon Stock 0 0 200 400 600 800 1,000 0 200 400 600 800 1,000 Tier 1 2040 Carbon Stock Tier 1 2050 Carbon Stock

Figure 15.1 continued

other words, managers looking for the best places agement questions are focused specif cally on to invest, or the places to avoid, will have an easier f nding the places with the highest levels of car- (or clearer) decisions when the tier 2 model is used. bon storage or sequestration. These “priority As with the absolute predictions, this discrepancy areas” are the places where the greatest return on in model predictions gets more severe the farther investment can be expected either in terms of stor- one looks into the future (although the gap will age or sequestration, if acquisition prices and stop growing at some point in the future when all accessibility are consistent across the landscape. land cover classes reach their maximum storage We asked if the simple and complex models iden- capacity). tify the same “priority areas” in the Willamette Rather than looking at the full spectrum of car- Basin. For carbon storage, we def ned priority bon values found on the landscape, many man- areas as the best 25% of the landscape for storing (c) 1000 1000

800 800

600 600

400 400

200 200 Tier 2 2000 Carbon Stock Tier 2 2010 Carbon Stock 0 0 0 200 400 600 800 1,000 0 200 400 600 800 1,000 Tier 1 2000 Carbon Stock Tier 1 2010 Carbon Stock

1000 1000

800 800

600 600

400 400

200 200 Tier 2 2030 Carbon Stock Tier 2 2020 Carbon Stock 0 0 0 200 400 600 800 1,000 0 200 400 600 800 1,000 Tier 1 2020 Carbon Stock Tier 1 2030 Carbon Stock

1000 1000

800 800

600 600

400 400

200 200 Tier 2 2040 Carbon Stock Tier 2 2050 Carbon Stock 0 0 0 200 400 600 800 1,000 0 200 400 600 800 1,000 Tier 1 2040 Carbon Stock Tier 1 2050 Carbon Stock

Figure 15.1 continued

Current Development Plan Conservation 1990 2050 2050 2050

Difference in Carbon Stock Mg C per ha

–90050 – –20000 –19999– –1 0–1 2–60000 60001–80000 80001–136000

Figure 15.2 Difference between tier 1 and tier 2 estimates of carbon stock distributions across space under current and future conditions in the Willamette Basin (USA). Tier 2 estimates are higher than tier 1 estimates in dark areas and lower than tier 1 estimates in light areas. TESTING AGREEMENT BETWEEN SIMPLE AND COMPLEX ECOSYSTEM SERVICE MODELS 273

Development Plan Conservation

Difference in Carbon Sequestration Mg C per ha per year

–94984– –1 0 1–2500 2501–8500 8501–40000 40001–60000 60001–129281

Figure 15.3 The difference between tier 1 and tier 2 estimates of carbon sequestration rates across space between 1990 and 2050 under three possible scenarios for the Willamette Basin (USA). Tier 2 estimates were higher than tier 1 estimates in dark areas and lower than tier 1 estimates in light areas.

carbon. We then used the Kappa statistic to calcu- 15.2.2 Crop pollination late the degree of overlap between the priority The tier 1 and tier 2 pollination models both produce areas identif ed by tier 1 and tier 2 models. The an index of pollinator abundance in native habitat Kappa statistic ranges from -1 to 1, with -1 indicat- areas and on farm f elds. This index is a relative met- ing no overlap of high service areas in space, 0 ric that gives a sense of how many pollinators are indicating overlap purely due to chance, and 1 likely to be present on the landscape, but it is not an indicating perfect overlap. We found that the two estimate of actual pollination rates. The main differ- carbon models identif ed nearly identical loca- ence between the simple and complex pollination tions as high priority carbon storage areas under models is how accurately they represent the compo- both current (1990) and future conditions with sition of the native pollinator community and the Kappa values of 0.9 or higher ( Table 15.2 ). Given characteristics of each species present on the land- these results, decision-makers can be conf dent scape. The tier 1 model assumes there is only one that the selection of priority areas for carbon stor- generic species of pollinator on the landscape while age will not be signif cantly affected by the level the tier 2 model considers each species separately of model complexity used. and incorporates species specif c characteristics for foraging distance, nest site choice, and f oral resource Table 15.2 Spatial overlap of high carbon storage specif city (see Chapter 10 for more details). areas identif ed by tier 1 and tier 2 models

Carbon Stock 75 th percentile (tons C parcel-1 ) 15.2.2.1 Predicting crop pollination levels Scenario Tier 1 Tier 2 Kappa In agricultural areas, many conservation groups recommend the retention of islands of native habitat 1990 Base Scenario 232,138 244,450 0.94 within the production landscape to serve as reser- 2050 Conservation Trend 242,935 309,201 0.90 voirs for native pollinators. Farmers considering this 2050 Development Trend 219,713 275,167 0.90 approach might want to weigh the economic value of 2050 Plan Trend 217,968 278,128 0.93 the additional yields they receive from native pollina- H i g h s t o r a g e a r e a s w e r e i d e n t i f ed as those having per parcel carbon stocks greater than or equal to the 75 th percentile value. The Kappa statistic identi f es the degree tion against the income they lose from taking an area of spatial overlap between tiers with a value of 1 indicating perfect overlap. out of crop production. Being able to predict the level 274 HOW MUCH INFORMATION DO MANAGERS NEED? of pollination (and thus additional crop yield) that end of the spectrum, the tier 1 model will predict a can be expected from restoring or protecting a parcel lower level of pollination from a land use and cover is essential to this and other types of decisions. class that the generic species f nds incompatible. Using California’s major agricultural region, the The tier 1 model is likely most appropriate where Central Valley, as a case example, we asked whether native pollinator communities are not diverse, or the differences between simple and complex pollina- where pollinator preferences for habitat types are tion models are important for predicting pollination similar among all pollinator species. When these levels across a landscape. We used watermelon as the conditions do not hold, it will be more appropriate focal agricultural crop for this study. We assessed the to use the tier 2 model. The models differ the most agreement of simple (tier 1) and complex (tier 2) pol- when the landscape of interest contains land use lination models by generating predictions for 40 000 and cover classes that vary dramatically in their randomly selected parcels (entire landscape consists suitability for different pollinators. of 600 000 parcels) ( Figure 15.4 ). The predictions of relative abundance of pollinators on watermelon 15.2.2.2 Predicting crop pollination patterns in space crops in each parcel were very highly correlated ( R2 = Pollination is a service driven by relatively small 0.99). The tier 1 pollination model predicts lower lev- scale processes (on the order of meters to kilom- els of pollination than the tier 2 model at the low end eters), so the question of where services are proof the pollination spectrum and higher relative polli- vided is especially important for management nation levels at the high end of the spectrum. The dif- decisions with implications for pollination services. ference in predictions of pollination provision can be For investments in pollinator habitat to be fruitful, explained by the difference in pollinator community habitat patches need to be within pollinator forag- composition represented by the two models. The tier ing range distance of crops that require insect 2 model depicts several different species of pollina- pollination. Of course, different crops sell at differ- tors with potentially different preferences for each ent prices, so providing viable pollinator habitat habitat type, while the tier 1 model depicts one generic near the most lucrative crops will also be preferred pollinator representing average characteristics of the from an investment perspective. pollinator community. A land use and cover class that We analyzed how well model predictions of the is very compatible for the generic species in the tier 1 distribution of pollination services agreed by sub- model will generate a high score in tier 1. At the other tracting tier 1 estimates of the relative pollinator

0.8

0.6

0.4

0.2 Tier 2 Pollination Score

0 0 0.2 0.4 0.6 0.8 Tier 1 Pollination Score

Figure 15.4 Relationship between tier 1 and tier 2 predictions of a pollinator abundance index in California’s Central Valley for watermelon crops. The black line shows a 1:1 relationship, or perfect agreement between the models. TESTING AGREEMENT BETWEEN SIMPLE AND COMPLEX ECOSYSTEM SERVICE MODELS 275

Difference in Pollination –0.3– –0.2 –0.2– –0.1 –0.1– –0.01 –0.01–0.01 0.01–0.03 0.03–0.1 0.1–0.4

Figure 15.5 Difference in spatial patterns of the pollinator abundance index predicted by tier 1 and tier 2 models in California’s Central Valley for watermelon crops. The tier 2 model predicted higher pollinator levels than the tier 1 model in dark areas and lower levels in light areas. abundance index from tier 2 estimates across space tat as very good for nesting and feeding, whereas the ( Figure 15.5 ). The tier 2 model predicts lower pollina- tier 2 model includes some species that do not view tion provision in a large area in the western part of the this habitat as perfectly compatible, yielding an over- Central Valley. This area is dominated by high quality all lower pollination score. Similarly, the green areas, pollinator habitat. The generic bee species repre- located along thin riparian buffers, are seen as incom- sented in the tier 1 model sees this high quality habi- patible habitat by the single insect species in the tier 1

100 90 80 70 60 50 40 30 % Model Agreement 20 10 0 100 90 80 70 60 50 Rank of Pollinator Abundance Index Estimate (%)

Figure 15.6 Summary of spatial agreement between tier 1 and tier 2 pollinator abundance index models applied in California’s Central Valley for watermelon crops. The x axis shows the percentile cutoff used to def ne “high priority areas” and the y axis shows how often the models chose the same parcels within that percentile. If one was interested in f nding the parcels on the landscape that have the highest 10% of abundance estimates (the 90th percentile), the models do not agree well (only 15% of those parcels overlap in space). If one was interested in locating the highest 25% of abundance estimates on the landscape (75th percentile), the models agree very well (92% overlap of parcels in space). 276 HOW MUCH INFORMATION DO MANAGERS NEED? model, but they are seen as viable habitat for some 2005 ). For some services, such as carbon sequestra- other species represented in the tier 2 model. The tion, there is a rapidly growing body of data on same recommendations given in the previous section which to validate models. Measures of above apply here; the tier 1 model will be most appropriate ground biomass are readily available for many sys- for landscapes with few pollinator species, pollina- tems. Empirical results for carbon storage in soil, tion communities with similar habitat preferences, or however, are subject to considerable heterogeneity landscapes with relatively homogenous, high quality making model prediction and validity testing for pollinator habitat. this component of the carbon model diff cult. How As with other services, targeting investments rapidly carbon is sequestered with land use or land towards parts of the landscape with the most pollina- cover change is also subject to uncertainty. We have tion potential is a common approach. Here, rather found validation of ecosystem service models to than using the Kappa statistic to look at agreement be especially challenging because model outputs among priority sites def ned by one cutoff value, we are often new metrics that incorporate information look at how well predictions agree across space at about social and biological systems that are not every percentile from the 50th to the 100th (Figure commonly measured today. For instance, our water 15.6 ). If we choose the same threshold as we did for pollution regulation model combines information carbon, and ask the two models to identify the best on hydrological processes that determine ecosys- quarter of the landscape for pollination service provi- tem service supply with information from water sion (75th percentile), we see again that the models treatment plants that determine service demand agree very well (92% overlap in space) (Figure 15.6 ). and value. We can validate individual parts of the However, if we had a more limited budget and needed model based on readily available information (see to focus on the best 10% of the landscape (90th percen- Chapter 4 ), but it is diff cult to validate the f nal tile), the models pick quite different parts of the land- model outputs. Improvements in this area will only scape, agreeing only 15% of the time. So while the come as we amass data sets on metrics that relate to selection of priority areas focusing on the best 25% of integrated socio-ecological systems. For other serv- the landscape for pollinator habitat or below can be ices, the simplicity of the tier 1 models leads to esti- done equally well with either pollination model, pri- mation of a proxy for the service of interest. For ority setting that chooses to set more selective stand- example, the crop pollination model returns an ards will be affected by model complexity. index of pollinator abundance based on habitat availability. We can relate this proxy to empirical, f eld-based observations of pollinator abundance or 15.3 Future directions and open visitation rates to ask if the model does well in pre- questions dicting the rank order of pollination levels or spatial It is encouraging that the simple and complex models distribution, but we cannot verify the absolute level of carbon storage and crop pollination agree well in of pollination provided. terms of rank order predictions of ecosystem service Even when data become available, identifying level and in patterns of ecosystem service distribution which model is most appropriate for representing across space. The tests used in this chapter could be system dynamics can be complex ( Pascual et al. applied to tier 1 and tier 2 model pairs for all of our 1997 ). Pascual et al. (1997 ) suggest that instead of ecosystem service models (see Chapters 4 –6 , 8 , 9 , 11, trying to f nd a single model that does well in deal- 13 ). Model results should also be compared in land- ing with all components of a system, a more fruitful scapes with different biophysical and socio-economic approach may be to focus on f nding management characteristics to ensure that patterns of similarity choices that are robust to the model chosen for use. seen in our sample landscapes are robust. We have attempted to take that approach here for a Although understanding how simple and com- subset of decisions, suggesting that identifying the plex models relate to each other is important, con- best 25% of a landscape for service provision is gen- fronting models with empirical data is essential for erally robust to model choice for both carbon validating model results ( Van Nes and Scheffer sequestration and pollination. Decisions regarding FUTURE DIRECTIONS AND OPEN QUESTIONS 277 the placement of investments by conservation Grand, J., Cummings, M. P., Rebelo, T. G., et al . (2007). groups, f nancial institutions or others in the private Biased data reduce eff ciency and effectiveness of con- sector, along with decisions about where to direct servation reserve networks. Ecology Letters , 10 , 364–74. mitigation and offsetting or payments for services Grantham, H. S., Moilanen, A., Wilson, K. A., et al . (2008). could all be made with tier 1 or tier 2 models with Diminishing return on investment for biodiverstiy data in conservation planning. Conservation Letters , 1 , 190–8. equal conf dence. This approach can easily be Hulse, D., Gregory, S., and Baker, J. (2002). Willamette applied to the many other models discussed in this River Basin planning atlas: trajectories of environmental book. and ecological change. Oregon State University Press, The biggest challenge in terms of model selection Corvallis. and testing entails questions of “how much” as Ludwig, D., and Walters, C. J. (1985). Are age-structured opposed to “where are the best places.” There need models appropriate for catch-effort data? Canadian to be national ecosystem monitoring programs Journal of Fisheries and Aquatic Sciences , 42 , 1066. that report spatially explicit data on water quality, Nelson, E., Polasky, S., Lewis, D. J. et al . ( 2 0 0 8 ) . E f f ciency carbon sequestration and other vital ecosystem of incentives to jointly increase carbon sequestration outputs if we are to understand which models to and species conservation on a landscape. Proceedings of the National Academy of Sciences of the USA 105 use and how much conf dence to have in model , , 9471–6. results. Pascual, M. A., Kareiva, P., and Hilborn, R. (1997). The Inf uence of Model Structure on Conclusions about the References Viability and Harvesting of Serengeti Wildebeest. Conservation Biology , 11 , 966–76. Army Corps of Engineers (ACOE). (2004). Chicago District Pierce, S. M., Cowling, R. M., Knight, A. T., et al . (2005). 2004 mitigation requirements. Chicago District of the U.S. Systematic conservation assessment products for lan- Army Corps of Engineers, Chicago. duse planning: interpretation for implementation. Baker, J. P., Hulse, D. W., Gregory, S. V., et al . (2004). Biological Conservation , 125 , 441–8. Alternative futures for the Willamette River Basin, Reyers, B., Nel, J., Egoh, B., et al . (2005). A conservation Oregon. Ecological Applications , 14 , 313–24. assessment of South Africa’s grassland biome: Integrating Cowling, R. M., Knight, A. T., Faith, D. P., et al . (2004). terrestrial, river, ecosystem services and f ne scale priorities. Nature conservation requires more than a passion for SCIR, Cape Town. species. Conservation Biology , 18 , 1674–7. Reyers, B., Rouget, M., Jonas, Z., et al. (2007). Developing Driver, A., Maxe, K., Rouget, M., et al . (2005). National spa- products for conservation decision-making: Lessons tial biodiversity assessment 2004: Priorities for biodiversity from a spatial biodiversity assessment for South Africa. conservation in South Africa. South African National Diversity and Distributions , 13 , 608–19. Biodiversity Institute, Pretoria. Stephens, P. A., Frey-Roos, F., Arnold, W., et al . (2002). Escobedo, F. J., Wagner, J. E., Nowak, D. J., et al . (2008). Model complexity and population predictions, The Analyzing the cost effectiveness of Santiago, Chile’s alpine marmot as a case study. Journal of Animal Ecology , policy of using urban forests to improve air quality. 71 , 343–61. Journal of Environmental Management , 86 , 148–57. United States Environmental Protection Agency (US EPA) Gan, T. Y., Dlamini, E. M., and Biftu, G. F. (1997). Effects of (2002). Willamette Basin alternative futures analysis. model complexity and structure, data quality, and objec- Environmental assessment approach that facilitates consen- tive functionson hydrologic modeling. Journal of sus building. U S E P A O f f ce of Research and Development, Hydrology , 192 , 81–103. Washington, DC. Gaston, K. J., and Rodrigues, A. S. I. (2003). Reserve selec- Van Nes, E. H., and Scheffer, M. (2005). A strategy to tion in regions with poor biological data. Conservation improve the contribution of complex simulation models Biology, 17 , 188–95. to ecological theory. Ecological Modelling , 185 , 153–64. CHAPTER 16 Poverty and the distribution of ecosystem services

Heather Tallis, Stefano Pagiola, Wei Zhang, Sabina Shaikh, Erik Nelson, Charlotte Stanton, and Priya Shyamsundar

16.1 Introduction of clean water, food from bushmeat and native plants, medicinal plants, and protection from “nat- The planet’s stocks of natural assets continue to ural disasters,” such as storms and f oods diminish while widespread poverty persists ( Millennium Ecosystem Assessment 2005 ). For throughout the developing world (Chen and example, conventional household economic sur- Ravallion 2008). Recognizing the link between veys fail to include directly the contributions of humans and nature has never been more salient for natural assets to rural household welfare (Cavendish policy-makers and researchers concerned with con- 2000), making it diff cult to assess how they will be servation and poverty alleviation (Sanderson and affected by public or private programs that change Redford 2003; Millennium Ecosystem Assessment the status of these resources. Similarly, ecosystem 2005 ). Information detailing the specif c links services that make up a country’s aggregated natu- between the poor and the environment is building ral capital are largely absent in national accounting. (e.g. Albla-Betrand 1993; Scherr et al. 2003; Delang Progress in economic valuation of ecosystem serv- 2006) and efforts are underway to determine how ices is needed to support the development of a international agreements concerned with poverty standardized methodology for the inclusion of eco- alleviation can, and should, incorporate conserva- system values into Standard National Accounts tion as a means to their ends. For example, Roe and (Mäler et al. 2008) and other policy decisions. The Elliott (2004) note that United Nations (UN) fundamental problem with incorporating ecosys- Millennium Development Goals (MDGs), adopted tem services into the balance sheets used in deci- by the General Assembly of the UN, have direct ties sion-making is the lack of tools that easily track the to environmental condition and stability. Of the status of or changes in these services, and their dis- eight goals, only one is explicitly environmental, tributional effects on human well-being. but the success of f ve others will rely on healthy In this chapter, we examine how the emerging ecosystems (Roe and Elliott 2004). f eld of modeling and mapping ecosystem services Incorporation of these advances in local and can address this gap. We begin by discussing the regional decision-making has proven challenging. need for a detailed understanding of the linkages Creating policies that account for interactions and between poverty and ecosystem services, and pro- trade-offs among environmental, economic and vide a brief review of the literature on these links social values is diff cult today because many of the ( Section 16.2 ). We then examine the opportunities connections between humans and the environment for and diff culties with an integrated approach to are not formally recognized by political and eco- mapping poverty with ecosystem services (Section nomic systems. Instead, decisions made today 16.3 ). Finally, we demonstrate how several of the based on costs and benef ts to society leave out mapping and valuation models described in earlier many of the public goods and services provided to chapters (Chapter 8 ) and developed elsewhere the poor by the environment, such as the provision can be applied in this context using case studies of

278 ECOSYSTEM SERVICES AND THE POOR 279 the Amazon Basin and Guatemala’s highlands (Pattanayak and Sills 2001), and the insurance pro- ( Section 16.4 ). vided by natural resources can make it feasible for households to recoup after natural disasters 16.2 Ecosystem services and the poor (McSweeney 2005). Large changes in access to or availability of these services are likely to have sig- 16.2.1 Dependence of the poor on ecosystem nif cant effects on the poor (Ferraro 2002; World services Bank 2008 ), making it essential that policy-makers recognize the possible costs that choices related Earth’s ecosystems provide myriad goods and serv- to resource sectors impose on the poor through ices that are essential to the well-being of all people, changes in services ( World Bank 2008 ). but natural capital contributes disproportionately to the welfare of the poor (see Section 16.3.2 for a discussion of “poor”), in some cases signif cantly, 16.2.2 The poor as agents or victims of because often the poor have limited capacity to pur- environmental degradation chase goods and services from elsewhere ( Table 16.1 ) (World Bank 2008 ). For instance, Vedeld et al. (2004) Depending on the context, the poor may be agents found that about 22% of rural household income of degradation, its victims, or both. When the poor can be attributed to the harvest of goods from for- are agents of environmental degradation, they can ests, contributing almost twice as much to the cause declines in ecosystem service provision for incomes of the poor as to the non-poor ( Table 16.1 ). themselves and for others. For example, the use of This f nding cannot be generalized to all rural forests for fuel wood and other non-timber forest households, but it is relevant to those living on products by the poor has been the main cause of the fringes of forests, those that are largely depend- forest degradation in India (Baland et al. 2006) and ent on natural resources for subsistence, or those Tanzania (Luoga et al. 2000; Ndangalasi et al. 2007). who are engaged in agricultural activities that rely Similarly, degradation in Peru’s Pacaya-Samiria heavily on natural capital and ecosystem processes National Reserve was largely the result of the sub- ( Box 16.1 ). sistence use of resources by households living In addition to providing income, natural resources around the reserve (Takasaki et al. 2004). Degradation can serve as safety nets for the poor during times of can affect the functions of local ecosystems, increase stress ( Justice et al. 2001; World Bank 2008 ). The ecological fragility, and increase the vulnerability of poor have been shown to respond to known agri- the poor to natural shocks (Shyamsundar 2001). cultural risks and sudden agricultural shocks by Despite its apparent irrationality, one of the rea- increasing their dependence on natural resources sons that households degrade ecosystem services

Table 16.1 Environmental income as percentage of total income in resource-poor and resource-rich areas

Study Resource-rich areas Resource-poor/ Average low access areas

Poor Rich Poor Rich Poor Rich

Jodha (1986) 9–26 1–4 Cavendish (2000) 44 30 Vedeld et al. (2004)a 32 17 Narain et al. (2005) 41 23 18 18 Chettri-Khattri (forthcoming) b 20 14 2 1

a Data reported are from multiple earlier studies. b Nontimber forest product (NTFP) income only. In most cases, “poor” refers to poorest 20% and “rich” to the richest 20% of households. Adapted with permission from World Bank ( 2008 ). 280 POVERTY AND THE DISTRIBUTION OF ECOSYSTEM SERVICES

Box 16.1 Can the natural capital of agroecosystems provide a pathway out of poverty?

C. Peter Timmer in economic development and poverty reduction, we will not be able to understand the role of natural capital in Historically, the major pathway out of poverty has been the poverty reduction directly. structural transformation of economies, where the share of The major dilemma, for economists and policy-makers, is in agriculture in employment and value added to the national placing a monetary value on the output from agriculture. For economy declines as the share of urban industry and many decades rich countries have sought mechanisms to modern services rises. Labor productivity is higher in urban place a higher value on their agricultural sectors than market activities, and migration from rural to urban jobs raises prices would indicate, and thus, implicitly, value the underlying wages at both ends. In a broad sense, this transformation natural resources committed to agricultural production more has been a transition from dependence on biological highly. At least three rationales for supporting agriculture in processes of production— especially in agriculture—to rich countries at taxpayer and consumer expense are physical processes of production—primarily in manufactur- increasingly accepted by mainstream policy analysts as ing processes for metals, chemicals, and automobiles. ref ecting appropriate public action in the face of market Ultimately, the main source of economic growth—and failures. These are: support for the multiple functions that poverty reduction—has been in knowledge-intensive agriculture performs, beyond the commodity production that processes such as f nance, information technology, and is offered for sale (“multi-functionality”); support for “local” communications. In short, reducing poverty has meant food systems that might offer reduced carbon footprints for reducing reliance on natural capital. most food consumers and possibly even fresher and healthier This evolution away from apparent dependence on food; and support for bio-fuel production as a mechanism to natural capital as the source of economic growth and break dependence on imported fossil fuels and slow livelihoods for the poor obviously missed a key point: we all emissions of greenhouse gases. have to eat. It was easy to miss the point: in rich countries with highly productive agriculture, there are more lawyers than farmers. The structural transformation has as its Multi-functionality and the non-market endpoint “a world without agriculture,” or at least a world where the farm sector behaves economically like all other contributions of agriculture sectors in the economy. Bucolic landscapes, green buffers to urban density, But the importance of our dependence on agriculture as preservation and development of rural societies, domestic the most eff cient way for human society to capture solar food security, and f ood alleviation through proper land energy in a form that we can consume has returned with a management all have economic value even if there is no vengeance, in the form of high food prices. There is market price for their “production.” These non-commodity vigorous debate over the causes of the price resurgence, outputs, although essential to economic, environmental but most analysts feel that the link between energy prices and social well-being, are unpaid by-products of commod- and food prices that has been established by bio-fuel ity production. If farmers are paid only the market price for programs in the Unites States (ethanol from corn) and their commodities, the by-products will not be produced in Europe (bio-diesel from vegetable oils) now means that optimal amounts, and may be lost altogether if farmers are high fuel prices mean high food prices. “Renewable forced out of business because of international competitive energy” largely means capturing it from the sun, and pressures. photosynthesis remains the most eff cient way of doing Efforts to value in economic terms the f ow of multiple that over large expanses of land. services from natural ecosystems, including agriculture, Natural capital means far more than agriculture, of need far more analytical research and empirical testing. course, and most researchers in the f eld spend most of With better valuation will come better designed initiatives their energy understanding the value to local economic to conserve natural resources and better mechanisms to productivity of biodiversity from natural ecological systems. pay the provider of these services, including farmers. From But this focus is too narrow. Unless we understand the an economic perspective, simply paying farmers to do more broader context in which natural capital has value, of what they do anyway cannot be an eff cient use of f scal especially in the link between capturing solar energy as an or natural resources. Agriculture performs multiple agricultural activity and the subsequent role of agriculture functions, but f nding ways for the market to value, and pay ECOSYSTEM SERVICES AND THE POOR 281

for, these functions will be essential to sustainable Bio-fuels and the potential to reverse production. the structural transformation Bio-fuels are not exactly new. Although coal, the f rst fossil fuel, was known in China in pre-historic times, and was Local food systems traded in England as early as the 13th century, it was not used widely for industrial purposes until the 17th century. Buying food that is produced “locally” is the current Until then, bio-fuels were virtually the only source of energy agenda for two related causes: the anti-globalization for human economic activities, and for many poor people movement and the sustainability movement. The anti- they remain so today. But the widespread use of fossil fuels globalization movement has its roots in a clear sense of since the Industrial Revolution has provided a huge subsidy lost control over something as deeply felt as where the to these economic activities—because coal and later food on our tables comes from. Modern supply chains seem petroleum were so cheap—a subsidy which seems to be impervious to consumer desires to control what they eat. nearing an end. Are bio-fuels the answer to growing The sustainability movement has its roots in the broader scarcity of fossil fuels? environmental movement that now links to climate change Not surprisingly, the answer depends on the role of as the key challenge to quality of life in rich and poor agriculture in individual countries, the pattern of commod- countries alike. Can transporting food thousands of miles, ity production and the distribution of rural assets, especially often on jet freighters, possibly be a sustainable way of land. It is certainly possible to see circumstances where eating? Will buying and consuming foods produced locally small farmers respond to higher grain prices by increasing make any difference to either of these agendas? output and reaping higher incomes. These incomes might Economic eff ciency has a hard time entering these be spent in the local, rural non-farm economy, stimulating debates. Both the anti-globalization and sustainability investments and raising wages for non-farm workers. In movements specif cally reject market prices as the basis for such environments, higher grain prices could stimulate an evaluating decisions about what consumers should upward spiral of prosperity. consume, because these prices have too many subsidies An alternative scenario seems more likely however, and distortions to ref ect real opportunity costs in terms of partly because the role of small farmers has been under natural resources used. There is some merit to these so much pressure in the past several decades. If only arguments. The question is, should the “local food large farmers are able to reap the benefits of higher movement” receive more policy support? grain prices, and their profits do not stimulate a dynamic Consumers, especially wealthy consumers, like to know rural economy, a downward spiral can start for the poor. where their food comes from and buy from producers who High food prices cut their food intake, children are sent are neighbors. The rapid growth of farmers’ markets, of to work instead of school and an intergenerational organic food, and of “local food” sections in supermarkets poverty trap develops. If the poor are numerous enough, is testimony to this basic desire. The trend bears watching, the entire economy is threatened, and the structural because it is the ultimate form of agricultural protection. transformation comes to a halt. The share of agriculture Expanded trade has been the basis of much economic in both employment and GDP starts to rise, and this growth, and restricting it could have serious and unfore- reversal condemns future generations to lower living seen consequences. standards.

that they rely on is that the impact of slow and incre- tion, programs and policies designed to incentivize mental reductions in resource availability on welfare management activities that enhance ecosystem serv- is small due to substitution effects. Households adapt ice provision would help advance goals of both pov- to destruction of one resource over time by obtaining erty alleviation and conservation. their resources from alternate areas or switching to The poor often are victims of changes in ecosystem alternate resources. As long as the opportunity cost services caused by the activities of other sectors or of spending time to alter resource-use patterns is low, people in other locations (Box 16.2 ). In these cases, the the welfare impact of degradation is likely to be poor may benef t indirectly from regulations (such as small. In cases where the poor are agents of degrada- protection, see Andam et al. 2010) or incentive 282 POVERTY AND THE DISTRIBUTION OF ECOSYSTEM SERVICES

Box 16.2 Poverty and ecosystem service mapping at work in Kenya

Norbert Henninger and Florence Landsberg and combines that with poverty rates in 222 administrative areas (Figure 16.B.1; Plate 12). Most of the poorer A new atlas of Kenya, designed to improve understanding communities are located in the drier plains downstream of of the relationships between poverty and the environment, the foothills of the Aberdare Range and Mount Kenya. The was released in 2007 ( World Resources Institute 2007 ). The quantity and quality of the surface water supply for these atlas and its 96 different maps include signif cant policy poorer communities is highly dependent upon the use of and economic development analyses that will be useful to land and water resources by the upstream communities. If policy-makers worldwide. This collection of maps is a step upstream users withdraw large quantities of water, little is forward from the landmark f ndings of the 2005 left for families downstream. If upstream users contaminate Millennium Ecosystem Assessment—that 15 of the world’s the water supply, families downstream bear the conse- 24 ecosystem services are degraded. It will help enable quences. Communities and decision-makers need to be other countries to develop their own similar maps. aware of these relationships to make better management Professor Wangari Maathai, 2004 Nobel Peace Laureate, and policy decisions. For example, upstream investments in said of the Atlas, “As a result of this type of work, we will improved watershed management to reduce water never be able to claim that we did not know. Planting trees pollution and water shortages could yield two benef ts: has been a way to break the cycle of diminishing resources improved ecosystem health and benef ts to poor down- for the women of the Green Belt Movement. I see the ideas stream communities. However, any mechanism to pay for and maps in this Atlas to be much like a small seedling. If these changes in the supply of ecosystem services cannot nurtured, if further developed and grown, and if used by rely on the downstream communities because of their lack both government and civil society, this seedling carries the of resources. promise of breaking the cycle of unenlightened decision- Similarly, other maps in the Atlas show how and where making that is not accountable to the people most affected people derive benef ts from the land and how that relates by economic or environmental changes; that does not to the spatial pattern of human well-being. The Atlas is consider the impact on our children and grandchildren.” designed to inspire improved analysis of poverty-environ- As an example, one map from Nature’s Benef ts in ment relationships and informed decision-making. Kenya outlines the upper watersheds of the Tana River

N POVERTY RATE (percent of population below poverty line) M M E R U T. K A M E R U N.P. E N Y C E NT R A L > 65 M E R U T H A R A K A S O U T H 55–65 N Y E R I E M B U 45–55 KI MURANGA RINYAGA 35–45 M W I N G I ABERDARE RANGE <= 35 MARAGUA M B E E R E Tana River OTHER FEATURES Upper Tana boundary Masingo K I A M A B U Reservoir T H I K A District boundaries Major national parks and reserves (over 5,000 ha) WATER BODIES AND RIVERS M A C H A K O S KITUI Permanent rivers NAIROBI Water bodies

Figure 16.B.1 Map of the Tana River headwaters in Kenya, and the distribution of poor communities. (See Plate 12.)

programs targeted at government or wealthier actors away from urban poor to areas with signif cantly to control ecosystem service degradation in poorer lower population densities (Ruhl and Salzman 2006; areas. For example, through the Clean Water Act, the BenDor et al . 2008). The equity impacts of this regula- US Government has unintentionally redistributed tion are rarely addressed in the Clean Water Act deci- wetland services (e.g., f sh for food, f ood mitigation) sion-making process (BenDor et al. 2008). An MAPPING POVERTY AND ECOSYSTEM SERVICES 283 alternative regulatory approach could intentionally More detailed poverty maps are typically created direct wetland mitigation activities towards areas by using comprehensive data from small-sample with poor populations, thereby improving the poor’s household budget surveys to obtain a predictive access to wetlands’ many services. relationship for poverty rates that is then applied to data from a national census at the highest available 16.3 Mapping poverty and ecosystem level of disaggregation (Poggi et al. 1998; Elbers et al. services 2002). This approach has been used to generate poverty maps for several countries (Hentschel et al. The relationship between the poor and natural 2000; Minot 2000; Müller et al. 2006; Bedi et al. 2007; resources is mediated by factors at various scales, Nelson and Chomitz 2007). However, even this such as labor and credit markets, property rights f ner scale analysis still produces maps based on and other institutions, and information about administrative divisions (census tracts) and not best practices (Bluffstone 1995; Duraiappah 1998; ecosystem boundaries. Nelson and Chomitz (2007) Wunder 2001; Adhikari 2005). In some cases, a lack dealt with this limitation by converting a poverty of markets contributes to degradation of natural map of Guatemala to a watershed map. Even with systems; in other cases growth in markets can lead poverty maps aligned to ecosystem boundaries, the to ecosystem declines. Weak governance institu- assumption of uniform distribution within an eco- tions, ill-def ned property rights or lax enforcement, system can seriously distort analysis. Within a high discount rates, and population growth will all watershed, for example, it may matter whether the likely continue to contribute to degradation of local poor are concentrated in the steeper upper slopes or natural capital. the f atter riparian zones ( Box 16.2 ). Given the complexities of connections between the poor and the environment, it is not easy to map pov- 16.3.2 Poverty indicators erty and ecosystem services together in a way that is robust and practically useful. Ideally, poverty and In addition to challenges with mapping poverty at ecosystem service mapping would be done such that an appropriate resolution for analysis, we also face (1) the resolution of both poverty and ecosystem the challenge of def ning poverty in different set- service information is suff cient to represent patterns tings. Poverty historically has been def ned in in each accurately, (2) the poverty indicators chosen strictly economic terms, with income as the com- are directly tied to the component of human well- mon indicator. Some analysts now argue that con- being of interest and well matched with the ecosys- sumption is a better measure of poverty, as it is tem service(s) of concern, and (3) the institutions that more closely related to well-being and ref ects control the provision of services are represented. capacity to meet basic needs through income and Here, we discuss what can be done today as f rst access to credit. It also avoids the problem of income steps toward this ideal and what challenges remain. f ows being erratic at certain times of the year, especially in poor agrarian economies where f uctuations can cause reporting errors. 16.3.1 Data resolution All money-based indicators have the limitation Robust poverty analyses require uniform and high- that they cannot ref ect individuals’ feelings of well- quality data that are often unavailable, especially in being and access to basic services. In recent years, a developing countries. In most cases, indicators such broader understanding has developed in which as poverty rates are only available for relatively poverty encompasses not only deprivation of mate- large administrative units. These data are often of rially-based well-being, but also a broader depriva- little use for detailed analysis because administration of opportunities ( World Bank 2001 ; UNEP 2004 ). tive boundaries seldom match those of the ecosys- The Millennium Ecosystem Assessment (MA) rec- tems of interest, and because neither ecosystem ognized f ve linked components of poverty: the nec- services nor the poor are likely to be distributed essary material for a good life, health, good social uniformly within these boundaries. relations, security, and freedom of choice ( Millennium 284 POVERTY AND THE DISTRIBUTION OF ECOSYSTEM SERVICES

Ecosystem Assessment 2005 ). Consider just one of When choosing the appropriate poverty these, security. A household’s ability to address risks indicator(s) to map, we should also consider the and threats can change dramatically even as income ecosystem service(s) of interest and how the poor or consumption remain stable. Factoring in the effect relate to those services. Pairings between ecosystem of vulnerability, analysts estimate that monetary- services and poverty indicators can be either direct based indicators can understate poverty and ine- or indirect. Pairings are “direct” if a change in the quality by around 25% ( World Bank 2001 ). In ecosystem service directly inf uences the poverty response, efforts have been made to develop non- indicator of choice (Table 16.2 ). If there is not a monetary poverty indicators related to health, nutri- causal link between the service and the indicator, tion, or education, as well as composite indices of the pairing is “indirect” ( Table 16.2 ). wealth (Wodon and Gacitúa-Marió 2001). In cases where the poor are agents of ecosystem Poverty measures used in mapping are typically service change, indirect pairings can be useful in def ned relative to a poverty line which is a cut-off mapping and modeling exercises used to design separating the poor from the non-poor. For instance, new programs. Consider a carbon offset program the headcount index is a measure of poverty inci- where a private sector buyer from a developed dence that computes share of the population below country wants to make payments to landholders the poverty line. An important distinction must be in the tropics to plant trees in order to offset the made between poverty rate, which is the propor- buyer’s carbon emissions. Mapping exercises tion of people in an area that are poor, and poverty could combine information on carbon sequestra- density, which is the number of people in an area tion potential and any indirect indicator of poverty that are poor. Many previous efforts to map poverty to identify areas where sequestration projects have found that areas with high poverty rates are could meet economic and conservation goals. An often areas with low population density, and thus a indirect indicator is appropriate here because pov- small absolute number of poor people. Poverty erty in this location is not directly related to the rates may be most relevant if an analysis aims to ecosystem service in question. The only require- locate segments of a population that are worst off, ment for the desired welfare transfer to the poor is but poverty density may be most relevant if the that people who are poor have ownership rights or analysis aims to f nd regions with the greatest control over the deforested or degraded lands number of poor. where carbon sequestration potential is high.

Table 16.2 Pairing ecosystem services with poverty indicators

Ecosystem service Poverty indicator

Water Child hunger Infant mortality HDI Income UBN Literacy poverty index

Water purif cation D I D b I I D c I Provision of food I D D I I I I Provision of medicinal plants I D D I I I I Timber production I I I I I I I Carbon sequestration I I I I I I I Crop pollination I D a I I I I I Erosion control I I I I I I I

a If local food crops need insect pollination. b If diarrhea from waterborne disease is signif cant cause of infant mortality. c If one of the unsatisf ed basic needs is clean water. Pairs where the poverty indicator could be directly in f uenced by a change in the ecosystem service are “direct” (D). Pairs where there is not a causal linkage between the service and the indicator are “indirect” (I). HDI = Human Development Index. UBN = Unsatisf ed Basic Needs. CASE STUDIES 285

When the poor are victims of environmental degra- non-timber forest product harvest in the Amazon dation, or benef ciaries of services, improvements in (Porro et al. 2008; see f nal model in Chapter 8; well-being occur through improvements in service results presented here are output from an earlier delivery, not through payments or incentives to the version of the NTFP model). The model estimates poor. In these cases, directly pairing ecosystem service the relative level of current NTFP harvest as a func- provision with poverty indicators tied to the service(s) tion of the association between harvested species of interest is appropriate. For example, consider a and habitat types, current NTFP stock (assuming change in the wetland mitigation example given that more pristine forests were harvested less in the above where the government requires developers to past and have a higher stock of products for harvest target offsets to benef t the poor. The most appropriate today), travel time from population centers (³ 1000 poverty indicators to use are those associated with people) along roads and waterways, and ease of wetland benef ts: hunger where wetlands provide f sh product harvest and current legal protection from for consumption, access to clean drinking water where harvest (assuming enforcement) (Porro et al. 2008; wetlands provide water purif cation, or f ood vulner- Peralvo et al. 2008). Forest regions with greater ability where wetlands provide storm surge protec- stocks of an NTFP, that were easier to reach, and tion. Using an income-based indicator, such as the were currently open to all households were assumed percent of the population below the poverty line, to to be harvested the most; the region with the high- recommend the allocation of improved wetland serv- est likely harvest has an index score of 1.0 for that ices would be inappropriate if household income is NTFP. We focused on the provision of wood, fruits not sensitive to changes in wetland services. and nuts, and medicinal plants sold in the market and wood, f ber, hunted bushmeat, fruits and nuts, and medicinal plants used for subsistence 16.4 Case studies ( Figure 16.1 ). The following two case studies represent some of Next, we used both direct and indirect pairings the analyses we can do today. They highlight the of these NTFP harvest projections with poverty types of decisions that can be informed by currently indicators to determine if current NTFP harvest is available data and methods, while identifying aligned spatially with areas of poverty. We also remaining challenges. investigated the impact of projected road expansion and deforestation over the next 20 years on the provision of NTFPs (deforestation scenario from 16.4.1 Deforestation in the Amazon Basin the Instituto de Pesquisa Ambiental da Amazonia, The Amazon Basin is one of the world’s most threat- road development scenario from the Initiative for ened and most diverse ecosystems, in terms of both Integration of Regional Infrastructure in the South biological and cultural diversity (~380 ethnic groups; (Porro et al. 2008)). These analyses are the f rst step Porro et al. 2008). Many of the people who reside in understanding whether the poor are at a greater here, poor and non-poor alike, rely directly on risk of losing their livelihood and well-being as a ecosystem services for their subsistence and liveli- result of new roads and deforestation. hoods (e.g., Clement 1993; Barham et al. 1 9 9 9 ; We represented current poverty with percentage Pattanayak and Sills 2001). Alarming rates of forest of underweight (UW) children (Figure 16.2a ; loss in the basin (Malhi et al. 2008) cause great con- Plate 11a) and the percentage of the population with cern for biodiversity loss, but we still have little unsatisf ed basic needs (UBN) ( Figure 16.2c; sense of what this loss means for society beyond Plate 11c). The resolution of UW children data var- species extinction and climate change (e.g., Laurance ied by country and was generally very coarse (cen- 1998; Ferraz et al. 2003). In this case study, we address sus blocks range in area from 13 202 km2 to 3 778 current and future provision of non-timber forest 690 km 2). Basic needs refer to any human need products (NTFP) and implications for the poor. where lack of satisfaction is considered to be an First, we used a mapping and valuation model indicator of deprivation or poor living conditions (see Nelson et al. 2009) to predict current levels of (Abaleron 1995). Basic needs, and the appropriate 286 POVERTY AND THE DISTRIBUTION OF ECOSYSTEM SERVICES

(a) (c) (e) (g)

0 0 0 0 0.153 0.090 0.110 0.392 0.161 0.364 0.400 0.596 0.164 0.588 0.702 0.599 0.999 0.999 0.999 0.999 (b) (d) (f) (h)

0 0 0 0 0.427 0.152 0.184 0.220 0.494 0.361 0.409 0.526 0.502 0.439 0.416 0.557 0.999 0.999 0.999 0.999

Figure 16.1 Distribution of forest product harvest index in the Amazon Basin in 2000. The relative harvest index for bushmeat for subsistence (a), f ber for subsistence (b), fruits and nuts for subsistence (c) or market (d), medicinal plants for subsistence (e) or market (f), and wood for subsistence (g) or market (h) varied across the Basin. All units are a relative harvest index in which the parcel with the highest likely harvest received a score of 1.0.

measures used to def ne their satisfaction, vary dra- We mapped two direct pairings of ecosystem matically across geographies. Therefore, a stand- services with a poverty indicator: the percentage ardized method for measuring UBN exists and of underweight (UW) children with the provision involves def ning the basic needs of the population of food (fruits and nuts) for subsistence ( Figure of interest, the appropriate measures for those 16.2b ; Plate 11b) and underweight children with needs, and the thresholds below which each need is the provision of bushmeat for subsistence. All considered unmet and people in such conditions other pairings were indirect, linking UBN to har- can be considered poor (Abaleron 1995). vest of marketed or subsistence NTFPs. In these In this case, the basic needs considered were: cases we assumed that households near areas of access to housing (type of material used for house greater NTFP harvest would be in a better position f ooring, walls and roof and the number of people to satisfy their basic needs by directly consuming per room), access to sanitation (type of water sup- or selling harvested NTFPs than households not ply source in the house and type and accessibility of located near these bases of consumption and bathrooms), access to education (presence or income supplementation. absence of at least one school-aged child not enrolled In 2000, we estimated that forest product harvest in school) and economic capacity (a calculation for both subsistence and market sale was relatively based on the age and number of household mem- higher in places where people had high UBN or UW bers, their highest level of education and their con- children ( Figure 16.3 ). Superf cially, this f nding may dition) (Schuschny and Gallopin 2004). Calculations suggest that forest products do not improve well- were made using census data (at the scale of munic- being. However, even though people in these regions ipalities) from 1993 for Peru, Ecuador, Brazil and are poor, forest products do make up a critical part Colombia and from 2001 for Bolivia. The size of of local incomes, supporting the health and nutri- municipalities varies dramatically across the region tional well-being of many Amazonian forest-related and many important patterns in UBN are likely households (Shanley et al. 2002). Deforestation leads missed by the often large regions included in a sin- to the loss of many of the most prominent and most gle census block. Threshold levels for each need prof table fruit and medicinal species that are not were set according to Feres and Mancero (2001). found in secondary forests (Shanley et al. 2 0 0 2 ) , t h u s CASE STUDIES 287

(a) (c)

0.8 0.0 7.7 19.3 17.5 95.3

(b) (d)

Figure 16.2 Poverty indicators and representative forest product harvest distributions in the Amazon Basin. The incidence of underweight children is highest in northern Peru and eastern Ecuador (a) while unsatisf ed basic needs are highest in Bolivia (c). High poverty areas def ned as those above the 75th percentile for underweight children (outlined in dark black) are shown in a direct pairing, overlaid with the harvest index of fruits and nuts for subsistence (b). High poverty areas def ned as those above the 75th percentile for unsatisf ed basic needs are shown in an indirect pairing, overlaid with the harvest index of wood for market sale (d). Units for underweight children are percentage of the population under the age of 5 that is underweight. Units for unsatisf ed basic needs are the percentage of the population with unsatisf ed basic needs. The legends and units for (b) and (d) are the same as in Figure 16.1 (c) and (h). (See Plate 11.)

continued forest loss is likely to contribute to declin- analysis suggests that proposed expansion of roads ing household health and nutrition. and development in the Basin will not likely result Finally, we found that over 20 years, the poor and in greater losses to the poor through decreases in non-poor alike will lose access to nearly all prod- forest products. This is probably because the region ucts analyzed ( Figure 16.3a , c). However, the great- is so large and the future time window so relatively est losses in harvest are likely to occur in areas short that the overall change in the Basin is rela- inhabited by the non-poor, or people with lower tively small. Most change also happens near already rates of underweight children and UBN developed areas that tend to be far from indigenous (Figure 16.3b , d). That is to say, this preliminary regions where rural poverty is generally higher 288 POVERTY AND THE DISTRIBUTION OF ECOSYSTEM SERVICES

(a) (b)

0.35 12 0.30 10 0.25 8 0.20 6 0.15 4 0.10 0.05 2 0.00 0 Relative Abundance Index Fruits, Nuts Meat Fruits, Nuts Meat (c) Decline in Ecosystem Service (%) 0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 Relative Abundance Index 0.00 Med Fruits, Wood Fiber Meat Wood Med Fruits, Nuts Nuts Subsistence Market (d) 10

Decline in Ecosystem Service (%) 0 Med Fruits, Wood Fiber Meat Wood Med Fruits, Nuts Nuts Subsistence Market

Figure 16.3 Distribution of forest product harvest index in 2000 and loss by 2020 between the poor and non-poor in the Amazon Basin. The harvest index for fruits, nuts, and bushmeat collected for subsistence use was higher in areas with high rates of underweight children (a, black bars) than in areas with low rates (a, gray bars). The harvest index of products for subsistence or market sale were also relatively more abundant in areas with high rates of unsatisf ed basic needs (c, black bars) than in areas with low rates (b, gray bars). Both the poor and the non-poor are projected to lose non-timber forest product provision in the next 20 yrs, but the non-poor will lose more, on average (b, d, gray bars) than the poor (b, d, black bars). High rates are def ned as the upper 25th percentile. Meat = bushmeat, Med = medicinal plants.

(Porro et al. 2008). However, those currently enjoy- by these indicators. Further, small losses to the poor ing low poverty rates are likely to experience the may do greater harm than larger losses to the non- greatest losses and may be pushed across an impor- poor. Information on the importance of specif c for- tant threshold as they lose access to forest products, est products to different groups, and an assessment resulting in a growing population of poor as def ned of additional ecosystem services and revenue CASE STUDIES 289 streams is needed to explore the impact predicted dicted to see substantially greater declines in the losses will have. provision of forest products than areas that are bet- If we look not at the entire basin but at each counter off ( Figure 16.4 ). Losses in the provision of try separately, we see a somewhat different story. In medicinal plants harvested for market sale were Brazil and Peru, regions with high UBN are pre- also higher in underprivileged areas of Colombia

(a) (b)

0.009 0.010 0.008 0.009 0.007 0.008 0.006 0.007 0.005 0.006 0.005 0.004 0.004 0.003 0.003 0.002 0.002

0.001 Loss in Wood for Market 0.001 Loss in Medicine for Market 0.000 0.000 Brazil Colombia Peru Ecuador Bolivia BrazilPeru Ecuador Bolivia Colombia (c) (d) 0.009 0.006 0.008 0.007 0.005

0.006 0.004 0.005 0.004 0.003 0.003 0.002 0.002

Loss in Food for Market 0.001 0.001 0.000

Loss in Hunting for Subsistence 0.000 BrazilPeru Colombia Ecuador Bolivia Brazil Peru EcuadorColombia Bolivia (e) 0.014 0.012 0.010 0.008 0.006 0.004 0.002 0.000 Loss in Medicine for Subsistence BrazilPeru Colombia Ecuador Bolivia

Figure 16.4 Predicted losses of forest product harvest after 20 years of deforestation and road expansion in individual countries in the Amazon Basin. Patterns were different at the country scale. Areas with high rates of unsatisf ed basic needs (black bars) showed greater declines in the harvest index for medicinal plants for market sale in Brazil, Peru and Colombia (a). This same pattern of disproportional loss in areas of high unsatisf ed basic needs (black bars) held for wood (b) and food for market sale (c) as well as hunted bushmeat (d) and medicinal plants for subsistence use (e) in Brail and Peru. High rates of underweight children or unsatisf ed basic needs are those above the 75th percentile value for the country. Units in all graphs are the decline in relative abundance index. Error bars show standard error. 290 POVERTY AND THE DISTRIBUTION OF ECOSYSTEM SERVICES

( Figure 16.4a ). We need additional information are thought to improve provision (Wunder 2001; about resource access, demand and changes in other Pagiola and Platais 2007). Although the approach ecosystem services and other factors associated was conceptualized as a mechanism to improve the with well-being to make sound conclusions about eff ciency of natural resource management, most how these changes are likely to affect the poor, but land users in upper watersheds are thought to be these trends suggest that the rural poor in certain poor (Heath and Binswanger 1996; CGIAR 1997 ), Amazonian regions may be most vulnerable to and because most ecosystem services are thought to declines in natural capital in these countries over come from such areas (Nelson and Chomitz 2007), the next 20 years. In addition, these analyses only many have assumed that most potential PES recipi- pertain to the poor living within the Amazon basin, ents are poor. Others indicate that the linkages and do not include poor in urban settings in each between potential PES recipients and poverty are country. Our analyses highlight the utility of simple more complex and show mixed results (Grieg-Gran mapping in identifying areas where the coupling of et al. 2005; Pagiola et al. 2005; Ravnborg et al. 2007; institutional information with ecosystem service Nelson and Chomitz 2007; Pagiola et al. 2008). information should be pursued. Further, while PES can benef t poor landowners, other poor populations may be negatively affected by the changes in land use through higher prices or 16.4.2 Potential for payments for lost employment (Zilberman et al. 2008). environmental services in highland Guatemala To examine whether PES approaches bene f t the Recent years have seen considerable interest in the poor in practice, Pagiola et al . (2007) analyzed the development of programs of Payments for Environ- spatial distribution of poverty in areas that are mental Services (PES). The PES approach aims to important to the provision of water services in address the classic problem of environmental exter- highland Guatemala (Figure 16.5 ). With about 56% nalities by establishing a mechanism through which of its population under the poverty line, Guatemala service users can compensate land users that pro- has one of the highest poverty rates in Central vide the desired service, or that adopt land uses that America (World Bank 2004 ). Guatemalan poverty

(a) (b)

N N

0.04–0.20 0.005–0.01 0.20–0.80 0.01–0.08 0.80–1.50 0.08–0.20 1.50–3.45 0.20–0.68 02040 60 120 02040 60 120 Kilometers Kilometers (c) (d)

0.02–0.2 N 0.2–1 1–2.5 Missing value 0–1.5 2.5–6.5 1.5–4.5 6.50–55.4 Guatemala city metro 4.5–20 02040 60 100 20–817 02040 60 100 Kilometers Kilometers

Figure 16.5 Water supply areas for principal surface water users in highland Guatemala. Major uses highlighted are hydroelectric power generation in generating capacity per hectare of upstream area (kW ha-1 ) (a), large-scale irrigation as irrigated area per hectare of upstream area (ha ha -1 ) (b), domestic water supply as households served per hectare upstream area (households ha-1 ) (c) and coffee mills as production quantity per hectare of upstream area (quintal ha-1 ) (d). CASE STUDIES 291

(a)

Water supply areas Missing value 0.001–0.25 0.25–0.50 0.50–0.75 0.75–0.98 02040 80 120

Kilometers

(b)

Water supply areas Missing value 0.0001–0.2 0.2–0.5 0.5–2.5 2.5–9.6

02040 80 120

Kilometers

Figure 16.6 Spatial distribution of poverty in Guatemala. Patterns are shown for both the poverty rate (number of poor) (a) and poverty density (number of poor ha-1 ) (b). The water supply areas for major water use points are outlined in black. The poor are def ned as those under Guatemala’s off cial poverty line, which is estimated using data from the 1994 census and consumption data from a household survey 1998–9. The household-unit imputations are aggregated to small statistical areas to estimate the percentage of households living below the poverty line (Nelson and Chomitz 2007). 292 POVERTY AND THE DISTRIBUTION OF ECOSYSTEM SERVICES is predominantly rural and extreme poverty is per area basis) for water service provision are those almost exclusively rural : over 81% of the poor live serving mid-size users (who often have much in rural areas. Pagiola et al. (2007) asked whether smaller water supply areas). these poor could potentially benef t from payments The poverty rates in the water supply areas var- for the provision of water services by comparing ied substantially ( Figure 16.6 , Figure 16.7 ). While water provision areas to the distribution of poverty some water supply areas had high poverty rates, in highland Guatemala (Nelson and Chomitz 2007) others had low poverty rates. The water supply (Figure 16.6 ). They mapped the areas that provide areas for hydroelectric power production had rela- water services (“water supply areas”) by identify- tively high poverty rates of 67% on average, but ing the specif c location of the intakes used by with a very high variance ( Figure 16.7a ). Poverty major users to obtain their water and then delineat- rates were lowest in the water supply areas that ing the portions of the watershed that contribute provide Guatemala City’s domestic water supplies water to those intakes. For example, Figure 16.5 (Figure 16.6 ). The poverty density in water supply shows a sample map, highlighting the water sup- areas also varied substantially (Figure 16.6b ). The ply areas serving hydroelectric power plants. The average poverty density within water supply areas relative importance of each water supply area was was 103 poor km-2 . This is slightly more than the estimated by constructing an index based on the average poverty density in the highland areas of the measures of the magnitude of the service they pro- country of 83 poor km-2 , but the difference was not vide (in this example, installed generating capac- signif cant (Pagiola et al. 2007). ity) and the size of the water supply area. This Across the entire highlands region, there was no analysis showed that the most valuable areas (on a correlation between the importance of a water sup-

(a) (b) 100 100

80 80

60 60

40 40

Poverty rate (%) Poverty 20 rate (%) Poverty 20

0 0 051015 20 021 3 4 5 67 Importance for HEP generation (kW/ha) Importance for domestic water supply (hhs/ha) (c) (d) 100 100

80 80

60 60

40 40

Poverty rate (%) Poverty 20 Poverty rate (%) Poverty 20

0 0 0.00 0.20 0.40 0.60 0.80 051015 20 Importance for Irrigation (Irrig ha/ha) Importance for coffee production (quintals/ha)

Figure 16.7 Relationship between poverty rate and importance of water supply areas. Patterns are shown for water used to generate hydropower (a), for domestic consumption (b), for general irrigation (c), and for coffee production (d). Importance is def ned by allocating the amount of supply delivered to a use point equally across the supplying watershed. INCLUDING INSTITUTIONS: THE WAY FORWARD 293 ply area and the poverty rate of people living within currently provides and benef ts from ecosystem it ( Figure 16.7 ). However, there were at least some services, and who will gain or lose from future man- areas with high water service provision and a high agement changes. incidence of poverty. This suggests that payments targeted to these areas would have the potential to meet the joint goals of ecosystem service provision References and poverty alleviation. However, these analyses do not include information about the nature of insti- Abaleron, C. A. (1995). Marginal urban space and unsatis- f ed basic needs: the case of San Carlos de Bariloche, tutions in the region governing control of the water Argentina. Environment and Urbanization, 7, 97–116. supply or the relationship between the poor and Adhikari, B. (2005). Poverty, property rights and collective water-regulating land-use activities. The poor in action: understanding the distributive aspects of com- important water supply watersheds may have no mon property resource management. Environment and control over land-use practices that alter the provi- Development Economics, 10, 7–31. sion of water-related services, and as such, pay- Albla-Betrand, J. M. (1993). The political economy of large ments made to the poor may not result in ecosystem natural disasters. Clarendon Press, Oxford. service returns. If the poor control water resources Andam, K. S., Ferraro, P. J., Sims, K. R. E., et al . (2010). and PES mechanisms were developed in all the Protected areas reduced poverty in Costa Rica and water supply areas, 1.76 million people, or 73% of Thailand. Proceedings of the National Academy of Sciences the poor in highland Guatemala could potentially of the USA , 107, 9996–10001. Baland, J., Bardhan, P., Das, S., et al. (2006). Managing the be reached. This f gure engenders enough promise environmental consequences of growth: Forest degradation in that taking the next step to identify water resource the Indian mid-Himalayas . National Council of Applied institutions in highland Guatemala would be well Economic Research, India, New Delhi. worth the investment. Barham, B. L., Coomes, O. T. and Takasaki, Y. (1999). Rainforest livelihoods: income generation, household 16.5 Including institutions: the way wealth and forest use. Unasylva, 50, 34–42. forward Bedi, T., Coudouel, A. and Simler, K. (2007). More than a pretty picture: using poverty maps to design better policies As the case studies demonstrate, overlaying maps and interventions . World Bank, Washington, DC. of poverty indicators and ecosystem services can be BenDor, T., Brozovic, N., and Pallathucheril, V. G. (2008). an informative f rst step in understanding the com- The social impacts of wetland mitigation policies in the United States. Journal of Planning Literature , 22 , plex relationships between the poor and the envi- 341–57. ronment, but it is not suff cient. Current poverty Bluffstone, R. (1995). The effects of labor markets on defor- mapping seldom makes explicit connections estation in developing countries under open access: an between the poor and the resources they rely on. example from rural Nepal. Journal of Environmental For example, poverty maps rarely distinguish Economics and Management, 29, 42–63. between land owners and the landless, simply Cavendish, W. (2000). Empirical regularities in the poverty- because the necessary information is not available. environment relationship of rural households: Evidence Land users that are renters have less ability to dic- from Zimbabwe. World Development, 28, 1979–2003. tate how resources are managed. Similarly, even if C G I A R . ( 1 9 9 7 ) . Report of the study on CBIAR research priori- land users have title to their land, their tenure may ties for marginal lands . Consultative Group on International be insecure or not supported by local institutions. Agricultural Research, Technical Advisory Committee Secretariat, Food and Agriculture Organization of the Again, this will inhibit their ability to change United Nations, Rome. practices and alter the level of ecosystem services Chen, S., and Ravallion, M. (2008). The developing world is provided by the landscape. Information about insti- poorer than we thought, but no less successful in the f ght tutions such as property rights, ownership, or man- against poverty . World Bank, Washington, DC. agement responsibility, and their stability must be Clement, C. (1993). Native Amazonian fruits and nuts: incorporated into poverty mapping exercises to Composition, production and potential use for sustainable maximize their utility for understanding who development. UNESCO, New York. 294 POVERTY AND THE DISTRIBUTION OF ECOSYSTEM SERVICES

Delang, C. O. (2006). The role of wild food plants in pov- ments for sustainable development. Proceedings of the er4ty alleviation and biodiversity conservation in tropi- National Academy of Sciences of the USA, 105, 9 5 0 1 – 6 . cal countries. Progress in Development Studies, 6, 275–86. Malhi, Y., Roberts, J. T., Betts, R. A., et al . (2008). Climate Duraiappah, A. K. (1998). Poverty environment degrada- change, deforestation, and the fate of the Amazon. tion: A review and analyses of the nexus. World Science, 319, 169–72. Development, 26, 2169–79. Millennium Ecosystem Assessment. (2005). Ecosystems and Elbers, C., Lanjouw, J. O., and Lanjouw, P. F. (2002). Micro- human well-being: synthesis . Island Press, Washington, DC. level estimation of welfare . World Bank, Washington, DC. Minot, N. (2000). Generating disaggregated poverty maps: Feres, J. C. and Mancero, X. (2001). El metodo de las necesi- an application in Vietnam. World Development, 28, dades basicas insatisfechas (NBI) y sus aplicaciones en 319–31. America Latina. Division de Estadistica y Proyecciones Müller, D., Epprecht, M., and Sunderlin, W. D. (2006). Economicas, CEPAL-ECLAC, Bogota. Where are the poor and where are the trees? Targeting of pov- Ferraro, P. J. (2002). The local costs of establishing protected erty reduction and forest conservation in Vietnam . CIFOR, areas in low-income nations: Ranomafana National Park, Bogor. Madagascar. Ecological Economics, 43 , 2 6 1 – 7 5 . Narain, U., Gupta, S., and Van ‘t Veld, K. (2005). Poverty Ferraz, G., Russel, G. J., Stouffer, P. C., et al . (2003). Rates of and the environment: exploring the relationship between species loss from Amazonian forest fragments. Proceedings household incomes, private assets, and natural assets. of the National Academy of Sciences of the USA, 100, 14069–73. Poverty Reduction and Environmental Management Grieg-Gran, M., Porras, I., and Wunder, S. (2005). How can (PREM) Working Paper 05/09. Institute for Environmental market mechanisms for forest environmental services Studies, Vrije University, Amsterdam. help the poor? Preliminary lessons from Latin America. Ndangalasi, H. J., Bitariho, R., and Dovie, D. B. K. (2007). World Development, 33, 1511–27. Harvesting of non-timber forest products and implica- Heath, J., and Binswanger, H. (1996). Natural resource tions for conservation in two montate forests of East degradation effects of poverty and population growth Africa. Biological Conservation, 134 , 242–50. are largely policy-induced: the case of Colombia. Nelson, A., and Chomitz, K. (2007). The forest-hydrology- Environment and Development Economics, 1, 65–84. poverty nexus in Central America: An heuristic analysis. Hentschel, J., Lanjouw, J. O., Lanjouw, P. F., et al . (2000). Environment, Development and Sustainability, 9, 3 6 9 – 8 5 . Combining census and survey data to trace the spatial Nelson, E. N., Mendoza, G. M., Regetz, J. et al. (2009). dimensions of poverty: A case study of Ecuador. World Modeling multiple ecosystem services, biodiversity Bank Economic Review, 14, 147–65. conservation, commodity production and tradeoffs at Jodha, N. S. (1986). Common property resources and the landscape scales. Frontiers in Ecology and the Environment , rural poor in dry regions of India. Economic and Political 7 , 4–11. Weekly , 21 , 1169-81. Pagiola, S., Arcenas, A. and Platais, G. (2005). Can pay- Justice, C., Wilkie, D., Zhang, Q., et al . (2001). Central ments for environmental services help reduce poverty? African forests, carbon and climate change. Climate An exploration of the issues and the evidence to date Research, 17, 229–46. from Latin America. World Development, 33, 237–53. Laurance, W. F. (1998). A crisis in the making: responses of Pagiola, S., and Platais, G. (2007). Payments for environmen- Amazonian forests to land use and climate change. tal services: from theory to Practice. World Bank, Trends in Ecology and Evolution , 13 , 411–15. Washington, DC. Luoga, E. M., Witkowski, E. T. F., and Balkwill, K. (2000). Pagiola, S., Zhang, W. and Colom, A. (2007). Assessing the Economics of charcoal production in miombo wood- potential for payments for watershed services to reduce poverty lands of eastern Tanzania: some hidden costs associated in highland Guatemala . World Bank, Washington, DC. with commercialization of the resources. Ecological Pagiola, S., Rios, A. R., and Arcenas, A. (2008). Can the Economics, 35 , 243–57. poor participate in payments for environmental serv- McSweeney, K. (2005). Natural insurance, forest access ices? Lessons from the Silvopastoral Project in and compounded misfortune: forest resources in small- Nicaragua. Environment and Development Economics, 13, holder coping strategies before and after Hurricane 299–325. Mitch, north eastern Honduras. World Development, 33, Pattanayak, S. K., and Sills, E. O. (2001). Do tropical forests 1453–71. provide natural insurance? The microeconomics of non- Mäler, K., Aniyar, S., and Jansson, A. (2008). Accounting for timber forest product collection in the Brazilian Amazon. ecosystem services as a way to understand the require- Land Economics, 77, 595–612. INCLUDING INSTITUTIONS: THE WAY FORWARD 295

Peralvo, M., Benitez, S., Nelson, E., et al . (2008). Mapping Shyamsundar, P. (2001). Poverty-environment indicators . spatial patterns of supply and demand of ecosystem services World Bank, Washington, DC. in the Amazon Basin. The Nature Conservancy, Quito. Takasaki, Y., Barham, B. L., and Coomes, O. T. (2004). Risk Poggi, J., Lanjouw, J. O., Hentschel, J., et al . (1998). coping strategies in tropical forests: Flood, illnesses and Combining census and survey data to study spatial dimen- resource extraction. Environment and Development sions of poverty . World Bank, Washington, DC. Economics, 9, 203–24. Porro, R., Borner, J., Jarvis, A., et al. (2008). Challenges to U N E P . ( 2 0 0 4 ) . Human well-being, poverty and ecosystem managing ecosystems sustainably for poverty alleviation: services: exploring the links. Premier Printing, securing well-being in the Andes/Amazon. Amazon Winnipeg. Initiative Consortium, Belem. Vedeld, P., Angelsen, A., Bojö, J., et al . (2004). Forest envi- Ravnborg, H. M., Damsgaard, M. G., and Raben, K. (2007). ronmental incomes and the rural poor. Forest Policy and Payment for ecosystem services-issues and pro-poor opportu- Economics , 9 , 869–79. nities for development assistance . Danish Institute for Wodon, Q., and Gacitúa-Marió, E. (2001). Measurement and International Studies, Copenhangen. meaning: combiing quantative and qualitative methods for Roe, D., and Elliott, J. (2004). Poverty reduction and biodiver- the analysis of poverty and social exclusion in Latin America . sity conservation: rebuilding the bridges. Oryx, 38, 137–9. World Bank, Washington, DC. Ruhl, J. B., and Salzman, J. (2006). The effects of wetland World Bank. (2001). World Development Report 2000/2001: mitigation banking on people. National Wetlands attacking poverty. Oxford University Press, Oxford. Newsletter, 28, 9–14. World Bank. (2004). Poverty in Guatemala . World Bank, Sanderson, S. E., and Redford, K. H. (2003). Contested Washington, DC. relationships between biodiversity conservation and World Bank. (2008). P overty and the environment: under- poverty alleviation. Oryx, 37, 389–90. standing linkages at the household level. World Bank, Scherr, S. J., White, A., and Kaimowitz, D. (2003). A new Washington, DC. agenda for forest conservation and poverty reduction . Forest World Resources Institute. (2007). Nature’s benef ts in Kenya: Trends, Washington, DC. an atlas of ecosystems and human well-being. World Schuschny, A. R., and Gallopin, G. C. (2004). La distribu- Resources Institute, Nairobi. cion espacial de la pobreza en relacion a los sistemas ambien- Wunder, S. (2001). Poverty alleviation and tropical forests: tales en America Latina . Division de Desarrollo Sostenible What scope for synergies? World Development, 29, y Asentamientos Humanos, United Nations, Santiago. 1817–33. Shanley, P., Luz, L., and Swingland, I. R. (2002). The faint Zilberman, D., Lipper, L., and McCarthy, N. (2008). When promise of a distant market: a survey of Belem’s trade in could payments for environmental services benef t the non-timber forest products. Biodiversity and Conservation, poor? Environment and Development Economics , 13 , 11, 615–36. 255–78.

CHAPTER 17 Ecosystem service assessments for marine conservation

Anne D. Guerry, Mark L. Plummer, Mary H. Ruckelshaus, and Chris J. Harvey

17.1 Introduction examination of trade-offs in ecosystem services provided under alternative management scenarios Humans always have benef ted from marine eco- (National Research Council 2004). systems—either obviously in the form of food Using an ecosystem services framework also has resources, or more subtly in the form of cultural and the potential to draw a larger and more diverse recreational opportunities. For example, 80–85 mil- population of people to marine and other conserva- lion tons of f sh were landed in marine capture f sh- tion efforts, beyond those who value the environ- eries worldwide in 2006, and f sh account for ment purely for its direct uses. For example, many approximately 15% of the annual animal protein residents are drawn to the Puget Sound region, USA consumption by humans ( FAO Fisheries Department because of the sound’s physical beauty and con- 2009 ). A growing recognition of the degradation of comitant aesthetic benef ts to their well-being. global marine ecosystems has led to numerous calls Indeed, existence values have been found to be for a shift toward more holistic, ecosystem-based among the “most important” benef ts provided by management of marine systems ( Pew Oceans the Puget Sound system (Iceland et al. 2008 ). If such Commission 2003 ; US Commission on Ocean Policy cultural values can be included in tallies of the con- 2004 ; Council on Environmental Quality 2009 ). sequences of ecosystem protection, conservation Ecosystem-based management is a coordinated efforts are likely to engage a greater fraction of the effort to manage the diverse human impacts that population. Helping people to see the many ways affect an ecosystem to ensure the sustainability of their well-being is affected by marine and coastal the ecosystem services it provides (Rosenberg and environments is key to the success of conservation. Mcleod 2005). Two key aspects of ecosystem-based In principle, marine ecosystem services are not management are relevant here. First, ecosystem- fundamentally different from their terrestrial coun- based management fundamentally recognizes the terparts. In practice, however, the valuation and inseparability of human and ecological systems. mapping of ecosystem services in marine environ- Human well-being is derived from ecosystems ments is not as well developed as it is for terrestrial through ecosystem services and, in turn, human ecosystems. As described in Chapters 4 –13 , there behavior affects natural systems. Second, ecosys- have been some early successes applying ecosystem tem-based management is inherently multifaceted, service mapping and modeling tools in diverse ter- encompassing suites of services, rather than the tra- restrial and freshwater settings. These approaches ditional approach of sector-by-sector management. and models all start with basic land cover and land- Importantly, the framework of ecosystem services use data layers. The same approach can work for can provide performance metrics for different marine environments—marine systems have patchy management strategies that attempt to balance habitats that provide f ows of ecosystem services, multiple objectives by allowing for the explicit and management actions can alter those habitats

296 ECOSYSTEM SERVICES PROVIDED BY MARINE ENVIRONMENTS 297 and f ows. Several challenges must be addressed, restrial and marine policies, which have been his- however. Maps of habitat type and habitat use are torically disconnected. much harder to come by in marine systems than Fortunately, there are advanced aspects of marine they are on land. We cannot readily “see” many science that will provide a good foundation for eco- parts of the marine ecosystem and its habitat types system service analyses. In particular, although using satellite imagery or other remote sensing basic mapping data are less ref ned in marine envi- technology. Moreover, marine habitats and the proc- ronments, marine science has a rich ecosystem- esses that maintain them are more transient and based modeling tradition to draw on for quantifying three-dimensional than their terrestrial counter- ecosystem services. For example, modeling for parts, and associations between particular species f sheries management (e.g., Christensen and and habitats are harder to document. Walters 2004 ; Pauly et al. 2000 ; Fulton et al. 2 0 0 4 a , b ) Another challenge stems from the ways in and water use impacts in the Everglades and which humans interact with marine environ- Florida Bay (e.g., US Geological Survey 1997 ) pro- ments. While f shery harvest, one of the most vide sophisticated system and food web models important marine ecosystem services, is straight- that can be extended to evaluate a more compre- forward to measure, its ecological effects and hensive suite of human activities and ecosystem potential impacts on other ecosystem services are services. harder to discern. In addition, many of our actions that affect the marine environment take place on 17.2 Ecosystem services provided land. For example, coastal development; land-use by marine environments practices that produce nutrient, sediment, and pathogen inputs to freshwater; and increases in Global oceans provide a wealth of ecosystem bene f ts impervious surfaces can severely degrade near- that span all four major categories of services identi- shore marine systems (Carpenter et al. 1998 ; Mallin f ed by the Millennium Ecosystem Assessment et al. 2000 ; Diaz and Rosenberg 2008 ). Incorporating ( Millennium Ecosystem Assessment 2005 ): provi- an ecosystem service perspective into marine sioning, regulating, cultural, and supporting services management, then, facilitates integration of ter- ( Table 17.1 ). Marine ecosystems provide goods and

Table 17.1 Ecosystem services provided by oceans and coasts††

Subcategory Examples

Provisioning services Food Capture Fisheries Tuna, mahi-mahi, crab, scallops Aquaculture Salmon, oysters, shrimp, seaweed Wild foods Mussels, clams, seaweed Fiber Mangrove wood (construction, boat-building), seagrass f ber Biomass fuel Mangrove wood (charcoal), biofuel from algae Water Shipping, tidal turbines Genetic resources Individual salmon stocks, marine diversity for bioprospecting Biochemicals, natural medicines, and Medicines Anti-viral and anti-cancer drugs from sponges pharmaceuticals Food additives Seaweed harvest for carrageenans Regulating services Air quality regulation Sea salt and spray help cleanse the atmosphere of air pollution*

Climate regulation Major role in global CO2 cycle Water regulation Natural stormwater management by coastal wetlands and f oodplains Erosion regulation Nearshore vegetation stabilizes shorelines

(continues) 298 ECOSYSTEM SERVICE ASSESSMENTS FOR MARINE CONSERVATION

Table 17.1 continued

Subcategory Examples

Water purif cation and waste treatment Uptake of nutrients from sewage wastewater, detoxif cation of PAH’s by marine microbes, sequestration of heavy metals Disease regulation Natural processes may keep harmful algal blooms and waterborne pathogens in check Pest regulation Grazing f sh help keep algae from overgrowing coral reefs Pollination/assistance of external fertilization Innumerable marine species require seawater to deliver sperm to egg Natural hazard regulation Coastal and estuarine wetlands and coral reefs protect coastlines from storms Cultural services Ethical values Novn-use Spiritual fulf llment derived from estuaries, coastlines, and marine waters Existence values Non-use Belief that all species are worth protecting, no matter their direct value to humans Recreation and ecotourism Non-consumptive use SCUBA diving, beachcombing, whale watching, boating, snorkeling Consumptive use Fishing, clamming Supporting services Nutrient cycling Major role in carbon, nitrogen, oxygen, phosphorus, and sulfur cycles Soil formation Many salt marsh surfaces vertically accrete; eelgrass slows water and traps sediment Primary production ~40% global NPP ** Water cycling 96.5% of earth’s water is in oceans ***

* Rosenfeld et al . ( 2002 ). ** Schlesinger ( 1991 ). ** Melillo et al . ( 1993 ). *** Gleick ( 1996 ). ††The taxonomy of services is adapted from the Millennium Ecosystem Assessment ( 2005 ).

services from both biotic (e.g., depend on food webs) obvious provisioning services include timber and and abiotic (e.g., depend on the presence of seawa- f ber from mangroves and seagrass beds, and bio- ter) aspects of natural capital. Assessment reports chemicals for cosmetics and food additives. The within ( Agardy et al. 2005 ) and based on the potential also exists for developing novel natural Millennium Ecosystem Assessment (UNEP 2006) and products from marine species with medical applica- other synthesis documents (Peterson and Lubchenco tions (Carté 1996). In addition, the ocean may 1997 ; Costanza 2000 ; Patterson and Glavovic 2008; become an important energy source: biofuels from Wilson and Liu 2008 ) provide useful overviews of algae and power generation from wave and tidal these services, as do descriptions of the particular energy have potential for more widespread use. services provided by f sh populations ( Holmlund And f nally, the world’s oceans provide the high- and Hammer 1999 ), coral reef ecosystems ( Moberg ways for the global shipping trade. and Folke 1999 ), and mangroves ( Ronnback 1999). Marine systems also are responsible for a wide Provisioning services include the most high-pro- range of regulating services. Most prominent of f le marine ecosystem services such as food from these is natural hazard regulation. As was vividly capture f sheries, aquaculture, and wild foods. On highlighted by the human losses wrought by the average, each person alive in 2006 ate 16.7 kg of f sh 2004 Asian tsunami and 2005 hurricanes on the Gulf that year (18% of that total came from marine aqua- Coast of the USA, coastal and estuarine wetlands culture; the proportion from capture f sheries is dif- have value for their ability to reduce storm surge f cult to calculate given non-food uses of wild f sh) elevations and wave heights ( Danielsen et al. 2005 ; ( FAO Fisheries Department 2009 ). Some of the less Travis 2005 ; Box 17.1 ). Other regulating services ECOSYSTEM SERVICES PROVIDED BY MARINE ENVIRONMENTS 299

Box 17.1 Nonlinear wave attenuation and the economic value of mangrove land-use choices

Edward B. Barbier an entire mangrove ecosystem to shrimp aquaculture? For example, deciding how much of a mangrove forest extending Although most ecologists have concluded that ecosystem size 1000 m seaward along a 10-km coastline to convert to and functional relationships are non-linear, the lack of data or shrimp aquaculture may depend critically on whether all the mapping of these relationships has often precluded estimating mangroves in the 10 km2 ecosystem are equally bene f cial in how the value of an ecosystem service varies across an terms of coastal storm protection (Barbier et al. 2 0 0 8) . ecological landscape. However, recent collaborations between Suppose that it is assumed initially that the annual per ha ecologists, hydrologists and economists have demonstrated values for the various ecosystem benef ts are “uniform,” and this effect for the wave attenuation function of mangroves, thus vary linearly, across the entire 10 km2 m a n g r o v e which in turn impacts on the land-use choices for conserving landscape. Following this assumption, a mangrove area of or developing mangrove forests. 10 km 2 would have an annual storm protection value of 1000 Barbier ( 2007 ) conducted a comparison of land-use values times the $1 879 ha–1 “ p o i n t e s t i m a t e,” which yields an between various mangrove ecosystem benef ts and annual total benef t estimate of nearly $1.9 million. Barbier conversion of the mangrove to shrimp ponds in Thailand. He et al. ( 2008 ) show how this assumption translates into a found that all three ecosystem services - coastal protection, comparison of the net present value (10% discount rate and wood product collection and habitat support for off-shore 20-year horizon) of shrimp farming to the three mangrove f sheries—have a combined value ranging from $10 158 to services - coastal protection, wood product collection and $12 392 ha –1 in net present value terms over the 1996 to habitat support for off-shore f sheries—as a function of 2004 period of analysis, and that the highest value of the mangrove area (km2 ) for the example of a 10 km2 c o a s t a l mangrove by far is its storm protection service, which yields landscape. Figure 17.A.1 shows the comparison of benef ts. an annual benef t of $1 879 ha–1 annually, or a net present The f gure also aggregates all four values to test whether value of $8 966 to $10 821. an “integrated” land-use option involving some conversion But what if these per hectare values for mangroves were and some preservation yields the highest total value. When used to inform a land-use decision weighing conversion of all values are linear, as shown in the f gure, the outcome is

$20, 000, 000 $18, 000, 000 $16, 000, 000 $14, 000, 000 $12, 000, 000 $10, 000, 000 $8, 000, 000

Net present value Net present $6, 000, 000 $4, 000, 000 $2, 000, 000 $0 0 12345678910 Mangrove area (sq km)

Shrimp Farming (Commercial Coastal protection Returns) Wood products All values Habitat-fishery linkage

Figure 17.A.1 Linear ecosystem service returns from mangroves. continues 300 ECOSYSTEM SERVICE ASSESSMENTS FOR MARINE CONSERVATION

Box 17.1 continued a typical “all or none” scenario; either the aggregate values 1 9 9 7 ; B a r b i e r et al. 2008 ). In other words, wave attenuation will favor complete conversion or they will favor preserving is greatest for the f rst 100 m of mangroves but declines as the entire habitat. Because the ecosystem service values are more mangroves are added to the seaward edge. large and increase linearly with mangrove area the B a r b i e r e t a l ( 2 0 0 8 ) e m p l o y t h e n o n - l i n e a r w a v e preservation option is preferred. The aggregate value of the attenuation function for mangroves based on the f eld mangrove system is at its highest ($18.98 million) when it study by Mazda et al. ( 1 9 9 7 ) t o r e v i s e t h e e s t i m a t e o f is completely preserved, and any conversion to shrimp storm protection service value for the Thailand case study. farming would lead to less aggregate value compared to The result is depicted in Figure 17.A.2 . full preservation, thus any land-use strategy that considers The storm protection service of mangroves still dominates all the values of the ecosystem would favor mangrove all values, but small losses in mangroves will not cause the preservation and no shrimp farm conversion economic benef ts of storm buffering by mangroves to fall However, not all mangroves along a coastline are equally precipitously. The consequence is that the aggregate value effective in storm protection. It follows that the storm across all uses of the mangroves, shrimp farming and protection value is unlikely to be uniform across all ecosystem values, is at its highest ($17.5 million) when up to mangroves. The reason is that the storm protection “service” 2 km2 of mangroves are allowed to be converted to shrimp provided by mangroves depends on their critical ecological aquaculture and the remainder of the ecosystem is preserved. function in terms of “attenuation” of storm waves. That is, Taking into account the “nonlinear” relationship the ecological damages arising from tropical storms come between an ecological function and the value of the mostly from the large wave surges associated with these ecosystem service it provides can therefore have a storms. Ecological and hydrological f eld studies suggest that signif cant impact on a land-use decision at the landscape mangroves are unlikely to stop storm waves that are greater scale. Other ecosystem services, including those for than 6 m ( Forbes and Broadhead 2007 ; Wolanski 2007 ; mangroves, are likely to have similar effects. For example, a Alongi 2008; Cochard et al. 2 0 0 8 ) . O n t h e o t h e r h a n d , study of the nursery habitat function of mangroves in the where mangroves are effective as “natural barriers,” against Gulf of California, Mexico reveals that the function’s storms that generate waves less than 6 m in height, the inf uence on the productivity of off-shore f sheries does not wave height of a storm decreases quadratically for each 100 scale-up in direct proportion to the area of the mangrove m that a mangrove forest extends out to sea ( Mazda et al. forests in the nearby lagoons ( Aburto-Oropeza et al. 2008 ).

$18, 000, 000 $16, 000, 000 $14, 000, 000

$12, 000, 000 $10, 000, 000

$8, 000, 000 $6, 000, 000 Net present value Net present $4, 000, 000 $2, 000, 000

$0 012345678910 Mangrove area (sq km)

Shrimp Farming (Commercial Coastal protection Returns) Wood products All values Habitat-fishery linkage

Figure 17.A.2 Nonlinear ecosystem service returns from mangroves. ECOSYSTEM SERVICES PROVIDED BY MARINE ENVIRONMENTS 301 provided by marine systems include the transfor- of tourism, and is one of the world’s most pro- mation, detoxif cation, and sequestration of wastes f table industries (United Nations Environment ( Peterson and Lubchenco 1997). Programme 2006 ). Rich cultural services are provided by marine Finally, the oceans provide essential supporting systems. Human coastal communities—both native services that underpin many of the world’s ecolog- and non-native—often def ne their identities in ical functions. The oceans are the center of the glo- relation to the sea. In the U.S., people love to live bal water cycle; they hold 96.5% of the earth’s water near the ocean; one study predicts average increases ( Gleick 1996 ) and are a primary driver of the atmos- of 3,600 people a day moving to coastal counties phere’s temperature, moisture content, and stabil- through 2015 (Culliton 1998 ). Globally, coastal ity (Colling 2001 ). Oceans are also key players in tourism is a key component of many economies the global cycles of carbon, nitrogen, oxygen, phos- (Box 17.2 ). It is one of the fastest growing sectors phorus, sulfur, and other key elements (Peterson

Box 17.2 Valuation of coral reefs in the Caribbean

Emily Cooper and Lauretta Burke Reef-related tourism and f sheries In the Caribbean, nearly 70% of coral reefs are threatened Tourism is Tobago’s largest economic sector, contributing by human activities—including over-f shing, dredging, 46% of GDP and employing 60% of the workforce (WTTC sewage discharge, increased runoff from agricultural 2005). WRI conducted a f nancial analysis of reef-related activities, and coastal development ( Burke and Maidens tourism, including net revenues from all reef-related 2004 ). Degradation of reefs not only results in a activities, accommodation, and other spending on reef- tremendous loss of biodiversity but also leads to a decline related days. In addition, the study drew on a local-use in the services they provide to coastal communities, survey to estimate recreational use of the reefs and coralline resulting in lost revenue from declining tourism and f shing, beaches by local residents each year. In total, coral increased poverty and malnutrition, and increased coastal reef-associated tourism and recreation contributes an erosion. estimated US$43.5 million to the national economy per year. Many of these damaging activities occur because an Revenues from reef-associated f sheries tend to be individual or group seizes an immediate benef t, without dwarfed by tourism in the Caribbean, but f shing is an knowing or caring about the long-term consequences. important cultural tradition, safety net, and livelihood for Quantifying the value of ecosystem services provided by many people. Coastal f shing communities are often among reefs can help to facilitate more sensible, far-sighted the most vulnerable groups to degradation of the decision-making by drawing attention to the economic ecosystem, as they may have fewer income alternatives. A benef ts associated with reefs, and by demonstrating the f nancial analysis of reef-related f sheries in Tobago found true costs of poor coastal management. In 2005 the World that annual economic benef ts are between US$0.8–1.3 Resources Institute (WRI) launched a project to assess the million (0.7–1.1 million). economic contribution of three reef-related ecosystem services to countries in the Caribbean: reef-related f sheries, tourism, and shoreline protection. These three services were The role of coral reefs in protecting chosen because they are (a) relatively easy to measure using published information, (b) easily understood by the shoreline politicians and decision-makers, and (c) especially As part of this valuation effort, WRI developed an important to local economies. National-level studies have innovative method for evaluating the shoreline protection been completed for St. Lucia, Tobago, and Belize. In Tobago, services provided by coral reefs. By integrating data on one of two pilot sites, the project estimated the value of coastal characteristics, storm events, and coral reef location these three services at US$62–78 million per year ( Burke and type into a Geographic Information System, we are et al. 2008 ). able to evaluate the role of coral reefs in maintaining the

continues 302 ECOSYSTEM SERVICE ASSESSMENTS FOR MARINE CONSERVATION

Box 17.2 continued stability of a country’s shoreline. In Tobago (Fig. 17.B.1), the benef ts that are often undervalued or unnoticed. Even relative reef contribution is zero in areas not protected by a ballpark values help to support an economic case for coral reef, and ranges from 27% where the shoreline has including these types of ecosystem services in decision- relatively good protection due to other factors, to 42% making processes. Going forward, policy-makers in many where the shoreline would be most vulnerable without the Caribbean countries may f nd it worthwhile to invest in reef. The relative share of protection provided by coral reefs economic valuation to support decision-making, including is particularly high behind the Buccoo Reef in southwest conducting cost-benef t analyses and assessing the effects Tobago, as well as along several portions of the windward of coral reef degradation on the value of these services over coast. After assessing the relative protection provided by time. coral reefs, we integrate property values for vulnerable Working with local partners, WRI has tied the economic areas to arrive at an estimate of US$18–33 million in f ndings to some clear opportunities for improved coastal “potentially avoided damages” per year. management in Tobago. For instance, the Buccoo Reef Marine Park (BRMP) in the southwest of the country is a Policy relevance cornerstone of the tourism industry—60% of international visitors take trips into the park—and provides signif cant This type of valuation produces a picture of the current coastal protection to a heavily developed and low-lying estimated value of these three services. The method has the section of the island. Applying the same methods as at the advantage of being simple, replicable, and transparent, and national level but looking over a 25-year period, we it is a useful exercise for drawing attention to reef-related estimate that tourism associated with BRMP contributes

Shoreline Protection by Coral Reefs—Relative Reef Contribution

Realtive Reef Contribution none 27% 29% 30–33% 33–36% 36–42% Landmass_ima.shp Reefs

Coral Reef Data from Millennium Coral Reef Mapping Project and R, Laydoo. Updated at WRI based on reef observations from Trinidad Institute of Marine Affaris (IMA).

Prepared at WRI, September 2007

Figure 17.B.1 Shoreline protection by coral reefs . MAPPING AND MODELING THE FLOW OF MARINE ECOSYSTEM SERVICES: A CASE STUDY OF PUGET SOUND 303

between US$128 and 156 million in net present value over enforcement, re-routing a sewer drain out of the lagoon, 25 years. The reef offers protection valued at between improving watershed management and installing US$140 and 250 million over the same period (using a 3% sediment traps, or building a sewage treatment plant for discount rate). The park is meant to be a no-take zone, so the area. A cost-effectiveness assessment of these f shing is not assessed. interventions was outside the scope of this study, but There is little enforcement of park regulations, and would be fairly straightforward. Local conservation the reefs suffer from over-fishing as well as sediment groups hope the valuation findings will draw attention and nutrient runoff. Steps to preserve the reefs could to a threatened and valuable resource, and point to the include (from lowest cost to highest): increasing need for change.

and Lubchenco 1997 ) and are responsible for use) and benef ts (recreational and other cultural approximately 40% of global net primary produc- benef ts) of improved water quality in the tivity (Schlesinger 1991 ; Melillo et al. 1 9 9 3 ) . T h e Stockholm Archipelago. Recent examinations of oceans are home to vast reservoirs of genetic and shoreline stabilization and trade-offs with aquac- ecological diversity, arguably the most fundamen- ulture are illustrative of a general growing interest tal of supporting services as it is directly linked to in services from coastal environments (Box 17.1 , the rate of evolution and therefore the ability to B a r b i e r et al. 2008 ). adapt to a changing climate (Pergams and Kareiva, in press). 17.3 Mapping and modeling the f ow The valuation of marine ecosystem services lags of marine ecosystem services: behind efforts aimed at terrestrial systems, a case study of Puget Sound although coastal wetlands ( Batie and Wilson 1978 ; L y n n e et al. 1981 ; Farber 1988 ; Bell 1989), coral reefs Ecosystem services are a useful currency for cost- ( Spurgeon 1992 ; Moberg and Folke 1999; Cesar benef t analyses or assessments of the trade-offs 2 0 0 0 ; B r a n d e r et al. 2007 ), and mangroves ( Bennett among alternative strategies for achieving multi- and Reynolds 1993 ; Gilbert and Janssen 1998 ; ple objectives in marine systems. This is especially Ronnback 1999 ; Ruitenbeek 1994; Barbier true when those ecosystem objectives explicitly 2000 ; Sathirathai and Barbier 2001; Barbier 2003 , include human well-being in addition to tradi- 2 0 0 7 ; B a r b i e r et al. 2008 ) are notable exceptions. tional conservation goals, which is the situation in Marine and coastal ecosystem services have been Washington’s Puget Sound region. In this section, included in a few comprehensive valuation exer- we present a small step forward in applying the cises. Costanza et al. ( 1 9 9 7 ) u s e d a ( m o s t l y ) b e n e - concept of quantifying dynamic f ows of ecosys- f ts-transfer approach, applying estimates of tem services to the management of Puget Sound. ecosystem service values for specif c terrestrial The Puget Sound ecosystem in Washington State and marine habitats to extrapolate the global value is home to 3.8 million people encompassed in a of ecosystem services. Without careful matching of 42 000-km2 basin, including temperate-latitude sites to ensure that the benef ts can and should be lands and rivers from the crests of the Cascade and transferred, however, this approach can be mis- Olympic mountains through a deep, fjord-type leading (Plummer 2009 ). One of the most interest- estuary to the Pacif c Ocean. The region’s marine ing discussions of ecosystem service valuation in environment produces basic provisioning services the marine environment entails four case studies such as commercial and tribal subsistence f sheries that demonstrate how valuing a suite of ecosystem for salmon (Oncorhynchus spp.) and other species, services has the potential to inform decision-mak- as well as clam, oyster, crab, and other shellf sh ing in the Swedish coastal zone (Soderqvist et al . harvests. It provides regulating services as global 2005 ). One of these case studies explores the costs as the carbon cycle, and as local as waste treatment (increased water treatment and reduced fertilizer through the breakdown of PAHs (polycyclic 304 ECOSYSTEM SERVICE ASSESSMENTS FOR MARINE CONSERVATION

aromatic hydrocarbons) and PCBs (polychlorin- a list of 24 services translated from the Millennium ated biphenyls) by eelgrass (Huesemann et al. Ecosystem Assessment into locally relevant termi- 2009 ). It offers numerous cultural services through nology. Across 12 different sectors, there was broad bird and whale watching, recreational f shing, agreement that water quantity and water regula- water recreation, educational opportunities, and tion, recreation and ecotourism, and ethical and simply the human value placed on the existence of existence values were of the utmost importance; the region’s biodiversity. Puget Sound also pro- capture f sheries, aquaculture, water purif cation vides a rich cultural heritage for native Indian and waste treatment also ranked highly (Iceland tribes. And underlying all of these are basic sup- et al. 2 0 0 8 ) . T r a d e - o f f s a r e l i k e l y t o o c c u r a m o n g porting services such as primary production and services even in this short list of valued ecosystem the provision of habitat for the Pacif c Northwest benef ts. Representing outcomes of management icons salmon and orcas ( Orcinus orca ). choices in terms of multiple benef ts, in currencies Using Puget Sound as a case study is motivated related to human well being, has promise for by the region’s move toward an ecosystem-level engaging a broader spectrum of the public in chart- management approach. In 2007, the Washington ing a path forward. State Legislature mandated formation of a new An early focus of the Partnership’s effort is near- State agency guided by a public-private council— shore habitats. This builds on the work of a number the Puget Sound Partnership (Partnership)— of previous planning efforts in Puget Sound, which whose charge is to recover the ecosystem by 2020. identif ed specif c actions aimed at either protecting The Partnership’s governance structure and man- existing nearshore habitats or restoring degraded date for ecosystem recovery presents an opportu- areas to provide improved function for species, nity to apply principles from ecological theory habitat maintenance, or human access (Shared and the science of ecosystem services to help pri- Strategy 2007; Puget Sound Nearshore Ecosystem oritize management actions for Puget Sound. The Restoration Project 2008 ; Alliance for Puget Sound Partnership recognizes that ecosystem recovery Shorelines 2008 ). These nearshore recovery schemes will require changes in the way local, state, in Puget Sound have broadly similar objectives in Federal and tribal governments act and—just as their common desire to increase the amount of func- importantly—changes in choices human residents tioning nearshore habitat. make about how they commute to work, where In the remainder of this chapter, we focus on the they buy their food, homes, and so forth (Puget suite of ecosystem services that nearshore habitats Sound Partnership 2006 ). To meet these chal- produce and support in Puget Sound, and how lenges, the Partnership has adopted a system- those services could change in response to a set of wide approach to restoring the ecosystem, and possible management actions. To illustrate this they have explicitly def ned their multiple objec- approach, we quantify the outcomes of nearshore tives in terms of what ecosystem services people protection or restoration for three distinct kinds of in the region care the most about (Puget Sound services that f ow from an important foundation Partnership 2008 ). species—eelgrass ( Zostera marina): (1) carbon Identifying these public values is an essential sequestration for climate regulation (a regulating step toward making an ecosystem services frame- service with global reach), (2) marine commercial work of practical use. In Puget Sound, a diverse harvest (a provisioning service), and (3) non-con- group of stakeholders including those from f sher- sumptive values (recreation and existence values) ies and aquaculture, tourism, ports and shipping, associated with species that belong to the Puget cities, counties, tribal governments, environmental Sound food web. We examine how changes in eco- interests, agriculture, forestry, homebuilding, and system services are created by changes in eelgrass business sectors were interviewed to identify those itself (carbon sequestration) and how changes in services they believe to be “most important.” The eelgrass produce changes in services provided interviewers f rst educated the participants about through higher levels of the food web (harvest, rec- the concept of ecosystem services and offered them reation, and existence values). MAPPING AND MODELING THE FLOW OF MARINE ECOSYSTEM SERVICES: A CASE STUDY OF PUGET SOUND 305

assume for the sake of illustration that the Partnership 17.3.1 Eelgrass can identify and implement policies capable of pro- E e l g r a s s (Zostera marina) is a widely distributed, tecting eelgrass and halting this decline. In addition, clonal seagrass that forms large, often monospecif c we examine other policies aimed at restoring eelgrass stands in shallow temperate estuaries worldwide. in areas where it used to occur. To assess the potential Much of the vegetative biomass of eelgrass is below for restoration, we built a spatially explicit habitat the surface of sediments and the above-ground suitability model for eelgrass in Puget Sound, with biomass tends to be highly seasonal. Ecosystem the aim of identifying locations in which eelgrass has services attributed to seagrass beds include the the potential to grow but where its current status is sequestration of carbon, the provision of habitat for unknown ( Figure 17.1 ; Plate 13, Davies et al. , i n p r e p - f sh and invertebrates, and the control of erosion aration). Our model suggests that an additional through sediment stabilization ( Williams and Heck 36 877 ha of Puget Sound’s benthic area has the 2001 ). Threats to eelgrass in Puget Sound are similar potential to be occupied by eelgrass. If even half of to those facing this habitat type elsewhere, including this were to be occupied in the future due to restora- mechanical damage (such as through dredging and tion projects, it would represent nearly a doubling of anchoring), eutrophication, some aquaculture prac- the current area of eelgrass and would yield area tices, siltation, coastal construction, invasions by similar to that estimated to be present historically non-native species, alterations to coastal food webs, (41 239 ha compared to 47 328 ha summarized from and climate change ( Williams and Heck 2001 ; Duarte Thom and Hallum 1990). 2 0 0 2 ; B a n d o 2 0 0 6 ) . E e l g r a s s b e d s c u r r e n t l y o c c u r along approximately 37% of the coast of Puget 17.3.2 How do changes in eelgrass habitats Sound, where they provide habitat for mobile organ- affect carbon storage and sequestration? isms such as crabs and small f shes and feeding habitat for larger consumers such as seabirds, salmon, Estimates of carbon storage and sequestration in and marine mammals ( National Marine Fisheries marine systems are rare. For example, a marine ana- Service 2007 ). They also provide spawning and rear- log does not exist that is similar to the look-up tables ing habitat for Pacif c herring ( Clupea pallasi ) , a k e y for carbon storage and sequestration values for species in the regional food web (Penttila 2007 ; various land use/land cover categories available in N a t i o n a l M a r i n e F i s h e r i e s S e r v i c e 2 0 0 7 ) . terrestrial systems (Intergovernmental Panel on To evaluate policy scenarios for the nearshore, we Climate Change 2006). Our approach to estimating ask how management actions are likely to affect eel- the ecosystem service value of eelgrass carbon seques- grass, and then how multiple ecosystem services tration is similar to that of Chapter 7 (this book) and is provided by eelgrass are likely to change as a result. summarized in Table 17.2a . We f rst estimate the We estimate ecosystem services using simple amount of carbon stored in eelgrass biomass and approaches for the sake of illustration, and do not soils. We then estimate the rate of carbon sequestra- include ecological nuances such as spatial or tem- tion for eelgrass in Puget Sound, noting that carbon poral variation in their production and delivery. also f ows through eelgrass to be consumed by her- Also, it is important to note that since we are exam- bivores, decomposed within the system, and exported ining changes in f ows of services that are likely to from the system. These two estimates yield ecosys- result from changes in a foundation species, we tem service values for the scenarios examined. In the focus on bottom-up effects; future scenario work end, our approach illustrates how changes in eelgrass will include the examination of top-down effects habitats result in changes in the amounts and values such as changes in harvest of key f sh species. of carbon storage and sequestration. We know from previous work that the total area of eelgrass in Puget Sound has declined from historical 17.3.2.1 Carbon storage levels ( Thom and Hallum 1990 ). Although the rea- Our estimates of carbon storage in Puget Sound eel- sons for this decline are not well understood (Thom grass beds and sediments range from 1–6.3 TgC and Albright 1990 ; Thom and Hallum 1990 ), we ( Table 17.2a ). The amount of C in the sediment pool 306 ECOSYSTEM SERVICE ASSESSMENTS FOR MARINE CONSERVATION

49°0’N N Legend Land

Suitable Areas for Eelgrass

02040 Kilometers

48°30’N

48°0’N

47°30’N Mapped Eelgrass-EFH Potential Eelgrass Habitat

Study Area

47°0’N 124°30’W 124°0’W 123°30’W 123°0’W 122°30’W

Figure 17.1 A map of Puget Sound showing areas our model predicts suitable for eelgrass beds (green). Inset maps show higher detail; orange represents currently mapped eelgrass from the NOAA Essential Fish Habitat data (TerraLogic GIS Inc. 2004). (See Plate 13.) Table 17.2a Summary of methods used for estimating how changes in ecosystem structure and function give rise to changes in services provided for carbon storage and sequestration, commercial f sheries, and food web support

Ecosystem service Estimation approach Parameter values Key assumptions

1 Carbon storage by eelgrass for climate C = (A )*(C B + CS ) A = 21 140 ha 40% of eelgrass biomass is C; C in soil is attributed to eelgrass 2 regulation A is area (ha) of eelgrass CB = 0.03–0.4 MgC/ha stabilizing and trapping sediments and preventing decomposition. 3 C B is carbon stored in biomass per ha CS = 50–300 MgC/ha

C S is soil organic carbon content per ha Carbon sequestration by eelgrass for ∆ C = NPP * S * A NPP = 300–600 gC/m2 /yr 4 Carbon exported to other systems (including the deep sea) has the climate regulation NPP is net primary productivity of eelgrass S = 5–15%5 potential to be sequestered, but because its ultimate fate is S is proportion of NPP stored in sediments A = 21 140 ha1 unknown it is not considered further here. A is area (ha) of eelgrass Commercial f sheries for food: current Observed landed biomass by species for commercial Pounds per species/year landed f sheries in different marine sub-basins in Puget Sound6 Food web support mediated through EwE food web model7 Sigmoid mediation function7 Eelgrass affects herring biomass through egg survival; food web eelgrass-herring interaction Mediation function def nes changes in herring egg responses to changes in eelgrass are mediated through herring vulnerability due to changes in eelgrass biomass Wildlife viewing and existence mediated EwE food web model7 Sigmoid mediation function7 Eelgrass affects herring biomass through egg survival; food web through eelgrass-herring interaction Mediation function def nes changes in herring egg responses to changes in eelgrass are mediated through herring vulnerability due to changes in eelgrass biomass EwE food web model

1. Gaeckle et al . ( 2007 ). 2. In one subtidal Puget Sound meadow Nelson and Waaland ( 1997 ) estimated annual above- and below-ground eelgrass biomass to average 256.3 gdw/m2 (seasonal range: 72.2 gdw/m2 in January to 445.0 gdw/m2 in July). These are similar to Webber et al . ( 1987 ) from another Puget Sound location. Yang et al . (unpublished data) surveyed 17 sites around the sound in the spring and found above- and below-ground biomass to range 17–217 gdw/ m2 . Because we are interested in estimating the C in relatively steady-state pools, we used winter biomass and chose a range 8–100 gdw/m2 . 3. Eelgrass sediments in Rhode Island have been characterized as having up to 300 MgC/ha (Payne 2007 ). Jesperson and Osher (2007 ) found soils to a depth of one meter in an estuary in Maine to average 136 MgC/ha with a range for different (generally unvegetated) habitats of 67–177 MgC/ha. We used a range of 50–300 MgC/ha for the sediment C estimates. To put this in context, the global average for wetland soils is 720 MgC/ha (US Department of Energy 1999 ) and Pacif c Northwest old-growth forest soils are estimated to hold 30–400 MgC/ha ( Homann et al . 2004 ). For comparison, Pacif c Northwest forests have been estimated to store 180 MgC/ha above-ground ( Lippke et al . 2003 ). 4. Globally, Mann ( 1982 ) estimated the NPP of coastal systems to range 300–1000 gC/m2 /yr. Duarte and Cebrian ( 1996 ) estimated seagrass NPP to be 548gC/m2 /yr; Mateo et al . ( 2006 ) estimated it to be 817 gC/m2 /yr. Estimates for Z. marina in Europe and Asia range 620–2 600 gC/m2 /yr (summarized by Stevenson 1988 ). Estimates of NPP of Z. marina in Alaska and Oregon are 1000–1500 and 316–450 gC/m2 /yr, respectively ( McRoy 1974 ; Kentula and Mcintire 1986 ). An estimate of NPP for above-ground Z. marina and epiphytes at one location in Puget Sound is 344 gC/m2 /yr ( Thom 1990 ). Thom (unpublished data) used an estimate of 600 gC/m2 /yr in Puget Sound for above- and below-ground biomass. 5. A carbon budget for generalized seagrass systems estimated that 15.9% of NPP is stored in sediments (Duarte and Cebrian 1996 ); because most studies have been conducted on a tropical genus that forms large mats of organic material, we assumed that 15 was an upper bound for Z. marina . 6. PacFIN (Pacif c Fisheries Information Network), unpublished data. 7. See the text for description of the EwE model. We varied the shape of the mediation function from nearly linear to steeply sigmoid; as model outputs were qualitatively similar across all steepness terms, we discuss the results for an intermediate function. Table 17.2b Summary of methods for estimating how changes in ecosystem services result in changes in their value for carbon storage and sequestration, commercial f sheries, and food web support

Ecosystem service Estimation approach Parameter values Key assumptions

Carbon storage and sequestration by Eelgrass protection: Reduction in expected damage from climate change T = 50 Eelgrass is mature; sediment C losses when eelgrass biomass is

eelgrass for climate regulation through carbon storage and sequestration. Estimate difference in C B = 0.21 lost span 3 periods: An initial period before eelgrass is lost 1 2 3 C stocks with and without eelgrass protection. , , C S = 175 in which sequestration continues; a second period in which

∆C B = 0 eelgrass sediment carbon is lost (at a constant rate); and a

∆C S = 0.525 third period in which sediment carbon remains stable at its

C min = 43.75 minimum level.

t1 = 10

T = Years of carbon t 2 = 20 sequestration p = $25

CB = Initial carbon stock for r = 3% biomass (MgC/ha) c = 3%

CS = Initial carbon stock for sediments (MgC/ha)

∆CB = Annual carbon sequestration for biomass (MgC/ha/yr)

∆CS = Annual carbon sequestration for sediments (MgC/ha/yr)

C min = Minimum carbon stock for sediments if eelgrass is lost(MgC/ha)

t1 = Time when eelgrass is lost and sediment carbon begins release

t 2 = Time when sediment carbon ends release p = Social value ($/MgC) r = Social discount rate c = Carbon discount rate Carbon storage and sequestration by Eelgrass restoration: Reduction in expected damage from climate change T = 50 C stored and sequestered in area to be restored is 0.

eelgrass for climate regulation through carbon storage and sequestration. Estimate difference in C B = 0 4 5 C stocks with and without eelgrass restoration. , C S = 0

∆C B = 0.0432

∆C S = 0.525

T B = 5 π = 0.5 T = Years of carbon sequestration p = $25 r = 3%

CB = Carbon stock for biomass (MgC/ha) c = 3%

CS = Carbon stock for sediments (MgC/ha)

∆CB = Annual carbon sequestration for biomass (MgC/ha/yr)

∆CS = Annual carbon sequestration for sediments (MgC/ha/yr)

T B = Years for restored eelgrass to reach “maturity” π = Probability of successful restoration p = Social value ($/MgC) Ecosystem service Estimation approach Parameter values Key assumptions

r = Social discount rate c = Carbon discount rate Commercial f sheries for food: current Net revenues by species for commercial f sheries in different marine Pounds and dollars sub-basins per species/year landed in each marine sub-basin6 Commercial f sheries for food: mediated Use food web model to examine changes in commercial harvest due to Harvest rates for all species do not change; non-trophic through eelgrass-herring interaction changes in eelgrass biomass 7 relationships between eelgrass and other species (e.g., Chinook salmon and Dungeness crab) are not examined. Wildlife viewing and existence mediated Use food web model to examine changes in biomass of species groups Non-commercial value is related to population size (lbs are through eelgrass-herring interaction due to changes in eelgrass biomass 7 used as a proxy for value); non-trophic relationships between eelgrass and other species (e.g., Chinook salmon) are not examined.

1. With protection, the amount of carbon in year t is CtCCtC()=++∆ . with B S S 2. Without protection, the amount of carbon in year t follows a step function:

CtCCtCttwo/1()=++ B S∆ S, <

⎡⎤()tt− 1 =+[]CtCCtCCSSSS11min12∆−[],.+ ∆− ttt≤ < ⎣⎦⎢⎥()tt21− = CttT, ≤≤ min 2 3. The economic value of eelgrass protection is derived by f rst considering the difference between the carbon stock in each year with and without protection:

CtCtwith()− w/1 o ()=++ ( CCtCCCtCtt B S∆− S ) ( B++ S∆ S ), <

⎛⎞⎡⎤()tt− 1 =++()[][]CCtCBS∆− S⎜⎟ CtCCtCC S+ 11min12∆− S S+ ∆− S , ttt≤ < ⎝⎠⎣⎦⎢⎥()tt21− =++(),CCtCC∆− ttT ≤≤ BS S min 2 Note that for the f rst period, before the eelgrass biomass has been lost, there is no difference between the two stock levels, so that the value of protection (in those years) is zero. We then attach an economic value to the difference in each year: T pC(()()) t− C t VProtection()= with w/ o . ∑ (1rc )tt (1 ) t=1 ++ 4. With restoration, the amount of carbon in year t is

Ct()= ∆ CBS+ ∆ C , t< T B = ∆≤≤CTtT, SB 5. Without successful restoration, the amount of carbon in any year is zero, and so the expected value of eelgrass restoration is

⎛⎞TB pC∆∆T pC VRestoration()=+p BS ⎝⎠⎜⎟∑∑(1++rc )tt (1 ) (1 ++ rc ) tt (1 ) tt==11 6. PacFIN (Pacif c Fisheries Information Network), unpublished data. 7. See the text for a description of the EwE model. 310 ECOSYSTEM SERVICE ASSESSMENTS FOR MARINE CONSERVATION dwarfs that in the biomass pool (such that the bio- 17.3.2.3 Valuing ecosystem service value changes mass pool is truly negligible). In comparison, total for carbon US forest carbon stocks are estimated to be in the The ecosystem service value of carbon storage and range of 40 000–50 000 TgC. Pacif c Northwest sequestration is based on the reduction in the (western OR and WA) forests are estimated to con- expected damage from climate change. Increasing tain approximately 351 TgC ( US Environmental levels of carbon dioxide and other greenhouse gases Protection Agency 2007 ). are linked to harmful changes in temperature and other aspects of climate. Controlling carbon dioxide by sequestering carbon therefore mitigates those 17.3.2.2 Carbon sequestration harmful effects, which counts as an economic ben- Seagrasses stabilize sediments, slow water motion, ef t. This value is enjoyed by society at large, and so and cause the deposition of organic matter from the it is referred to as the social value of carbon storage water column (Gacia and Duarte 1999 ; Gacia et al. and sequestration. 1999 ). Below the sediment surface, anoxia and light- For eelgrass, this ecosystem service value can be limitation inhibit microbial processing and photo- generated either by protecting current eelgrass or degradation ( Jesperson and Osher 2007 ), allowing investing in eelgrass restoration, as described above. for the build-up of C. Soil carbon generally has long Protection provides value if existing eelgrass areas residence times—particularly when submerged— are threatened and the projected amount without and is therefore considered “sequestered carbon” protection decreases over time; restoration provides (Wang and Hseih 2002 ). Despite ideal conditions for value when eelgrass would otherwise decline or production and preservation of organic matter, the remain stable in the future. In either case, the eco- C-sequestration capacity of the soils of coastal economic value is based on the difference in carbon systems has been under-studied ( Chmura et al. stocks over time for two scenarios, one with the 2003 ; Thom et al. 2003 ; Jesperson and Osher 2007 ). appropriate action (protection or restoration) and Our initial estimate of a sequestration rate in one without that action. Thus, it is important to Puget Sound is 3 171–19 026 Mg C/yr, or 11 627–69 762 understand what form of carbon sequestration and

Mg CO 2 /yr ( Table 17.2a ). This represents 0.02–0.1% storage (if any) would either replace eelgrass (for of the emissions of Washington State, 0.06–0.36% of the case of protection) or be replaced by eelgrass (in the emissions of King County (Seattle’s home), and the case of restoration). Calculating this economic up to 72% of the annual emissions of all transit value is relatively straightforward ( Chapter 7 ), but busses in King County (King County 2007 ). For settling on the values for some of the parameters is comparison, the carbon contained in all US forests fraught with controversy (Nordhaus 2007 ; Weitzman offset approximately 10% of total US CO 2 emissions 2007 ). For the purposes of this chapter, we pick in 2005 ( Woodbury et al. 2007 ). values merely to illustrate how the calculation and If a restoration policy is being pursued, particular resulting value depends on the scenarios described habitat types would be changing to eelgrass from a above ( Table 17.2b ). previous habitat type, and the original state would B a s e d o n t h e m e t h o d s o u t l i n e d i n T a b l e 1 7 . 2 b , have had its own C-storage/sequestration values. the social value of eelgrass protection is much Similarly, if eelgrass habitat is being lost, it is being higher than that of restoration: $1 496 to 4 585 versus replaced with another habitat type, and a similar $104 ha–1 . The range in values for protection re f ect comparison could be made. This makes marginal different assumptions about the loss of carbon from changes impossible to calculate without going sediments, with the high value representing an through the same exercise for all habitat types. assumed total loss of stored carbon when eelgrass Among possible habitat types, however, we expect is left unprotected and the low value resulting from eelgrass to have the greatest capacity for carbon a ten-year lag before carbon is released, a ten-year storage and sequestration, compared to other non- carbon release period, and a 25% minimum carbon vegetated intertidal and shallow subtidal habitats stock that remains in the sediments without eel- such as rocky reefs, cobble, mud-, or sand-f ats. grass. The large disparity is due in large part to the MAPPING AND MODELING THE FLOW OF MARINE ECOSYSTEM SERVICES: A CASE STUDY OF PUGET SOUND 311

(a) San Juan Islands

Whidbey Island

Strait of Juan de Fuca

Harvest (1998–2007, lbs) Farmed and Wild

11,000,000 Hood Canal North Central Puget Sound Crab South Central Puget Sound Groundfish Salmon Shellfish South Puget Sound Other species

(b) San Juan Islands

Whidbey Island

Strait of Juan de Fuca

North Central Puget Sound Harvest (1998–2007, $) Farmed and Wild

$13,000,000 Hood Canal

Crab South Central Puget Sound Groundfish Salmon Shellfish Other species

South Puget Sound

Figure 17.2 A map of the Puget Sound Partnership’s action areas showing the distribution of (a) landings (in UK£) and (b) revenue (in US$) of farmed and wild seafood from 1998 to 2007. (See Plate 14.) 312 ECOSYSTEM SERVICE ASSESSMENTS FOR MARINE CONSERVATION assumed loss of stored carbon when eelgrass is left eelgrass are likely to affect changes in the food web- unprotected. If leaving eelgrass unprotected does based ecosystem service f ows, we focus on the link not produce a signif cant loss of stored carbon in between eelgrass, a foundation species, and one the sediments, the value of protection and its other species with wide-ranging food web interac- advantage over restoration is diminished tions, Pacif c herring. accordingly. 17.3.3.1 How might changes in eelgrass habitats affect marine harvest and non-consumptive values? 17.3.3 Marine harvest and non-consumptive In order to begin to understand how changes in values nearshore environments are likely to affect changes Puget Sound’s living marine resources, though in the f ows of harvest and other services, we exam- depleted relative to historic times, remain a bounti- ined the habitat associations of the top 25 species ful source of provisioning and other ecosystem harvested in the sound (including: geoducks services. Commercial f sheries harvest over 35 spe- ( Panopea abrupta ), salmon, Dungeness crab, and cies of f nf sh and shellf sh, and generate more than oysters), and categorized their dependence on near- $50 million in annual revenue (Pacif c States Marine shore habitats. Only 5 of these species (spiny dog- Fisheries Commission, Pacif c Coast Fisheries f sh, (Squalus acanthias), and 4 salmon—steelhead; Information Network (PacFIN), unpublished data). sockeye (Oncorhynchus nerka), coho (Oncorhynchus Recreational harvest concentrates on Pacif c salmon kisutch), and pink ( Oncorhynchus gorbuscha)) did not and steelhead ( Oncorhynchus mykiss ) but also rely on nearshore habitats for at least one part of includes shellf sh such as Dungeness crab ( Cancer their life cycle. Thus, harvest levels of most species magister ) and butter clams (Saxidomus giganteus ) in the top 25 are likely to be sensitive to changes in (Washington State Department of Fish and Wildlife nearshore habitats, but further modeling is neces- (WDFW), unpublished data). Puget Sound Indian sary to understand how. tribes enjoy a rich tradition of ceremonial harvest. Pacif c herring are a key food web species that Aquaculture uses the ecological functioning of interacts with eelgrass—they aggregate in the near- Puget Sound to produce more than $30 million in shore prior to reproducing then spawn in shallow annual revenue for shellf sh and almost $20 million water, usually on submerged vegetation (eelgrass for Atlantic salmon ( Salmo salar ) (WDFW, unpub- or algae). Submerged vegetation provides spawn- lished data). Non-consumptive activities, such as ing substrate, food resources, cover, and nursery recreational whale and bird watching, also provide habitat (Thayer and Phillips 1977 ; Dean et al. 2000 ; an important f ow of services, while just the exist- Penttila 2007 ). Survivorship of eggs is higher with ence of some species produces value for Puget lower spawn density ( Galkina 1971; Taylor 1971 ). Sound residents and visitors. To survive, planktonic larvae must have suff cient In order to provide a perspective with which to supplies of microplankton; blooms of which are gauge the effects of changes in the ecosystem, we believed to be earlier, more dense and more consist- tallied the commercial harvest coming from Puget ent in sheltered bays ( Penttila 2007 ). The survival of Sound in biological (lbs.) and monetary ($) units larval herring (determined particularly by food (PacFIN; Figures 17.2a , b; Plate 14). This snapshot of availability and predation) is thought to have a sig- the seafood provisioning service allows us to exam- nif cant impact on the future abundance of the year- ine where particular types of harvest are highest. class (Alderdice and Hourston 1985 ). Juveniles For example, shellf sh, which are particularly lucra- spend several months inshore before moving into tive, are predominately produced in southern Puget deeper waters (Penttila 2007 ). Herring are impor- Sound. tant prey to seabirds, crabs, salmon, marine mam- The mobility of f sh and their use of multiple hab- mals, and numerous other groups (Haegele 1993a , itat types necessitates a food web-based modeling b ; Penttila 2007 ). Given the strong connection approach rather than a habitat-based one. To under- between herring and nearshore habitat, we focus stand how changes in nearshore environments and here on the consequences of how changes in MAPPING AND MODELING THE FLOW OF MARINE ECOSYSTEM SERVICES: A CASE STUDY OF PUGET SOUND 313

nearshore habitat give rise to changes in herring, ous mechanisms by which changes in eelgrass could and how such effects can propagate through the lead to changes in herring populations. Here we food web. present only indirect effects that act through herring ( Figure 17.3b ; Plate 15) and focus on species 17.3.3.2 Puget Sound food web model and/or groups whose relationship with eelgrass is Biomass dynamics of eelgrass and herring take either trophic (i.e., they consume it) or is mediated place in the context of a broader community of through direct or indirect trophic interactions with interacting species, and resulting feedbacks within herring. the food web are diff cult to anticipate without the We simulated a 50% decrease, a 50% increase, and benef t of models. We used the Ecopath with Ecosim a doubling of eelgrass biomass, and linked this to (EwE; Christensen and Walters 2004 ) software to herring egg vulnerability. Depending on eelgrass construct a food web model for the central basin of biomass, herring eggs became either more or less Puget Sound ( Figure 17.3a ; Plate 15; Harvey et al. vulnerable to predation by several groups (ducks 2010). EwE models trophically and reproductively and brants, gulls, ratf sh, Dungeness crabs, small link biomass pools using a mass-balance modeling nearshore f shes, and small crustacean omnivores). approach that satisf es two master equations Increases in eelgrass biomass yielded increases in describing production (as a function of catch, pre- herring and in turn increases in harbor seals (Phoca dation, migration, and biomass) and consumption vitulina , whose primary prey is herring), ducks and (as a function of production, respiration, and unas- brants (consumers of eelgrass), and greenlings (con- similated food). An initial mass-balanced snapshot sumers of small crustaceans who feed on herring of the ecosystem can then be used to explore eggs). Increases in eelgrass yielded decreases in dynamic simulations by expressing biomass f ux gulls and terns, skates, gadoids, lingcod (Ophiodon rates among pools through time. The model of elongatus), and numerous f atf sh. Most of these Puget Sound’s Central Basin (Harvey et al. 2010) has decreases result from competition with increased functional groups ranging from primary producers herring populations. Skates, however, likely decline to marine mammals and seabirds, as well as sev- with increases in eelgrass and herring because their eral f sheries. The model results we present below primary predators are harbor seals, a species that are preliminary outcomes that illustrate the com- increases with eelgrass and herring. Results for plex and often unforeseen nature of community decreases in eelgrass biomass generally mirrored responses to perturbations (Christensen and those of increases. Walters 2004 ). Manipulating eelgrass production in EwE has 17.3.3.3 Valuing commercial harvest and negligible effects on the food web through con- non-consumptive services from food web changes sumption—a result that ref ects our current under- The species in the Puget Sound food web model pro- standing of eelgrass as a relatively unimportant vide consumptive services that include commercial direct food source (Mumford 2007 ). In contrast, harvest, and non-consumptive services such as non-trophic effects of eelgrass—such as habitat pro- whale and bird watching and existence value visioning—are known to be very important (Thayer (because of data limitations on recreational f shing and Phillips 1977 ; Orth et al. 1984 ; Hosack et al. 2006 ; values, we do not consider recreational harvest in Mumford 2007 ). Such effects have important posi- this section). Of the functional groups in the Puget tive effects on other species and can be ref ected in Sound food web model, 15 are harvested commer- EwE through density-dependent mediation func- cially (Table 17.3 ). By weight, the most important tions (Ainsworth et al. 2008 ). Mediation functions f sheries are salmon (sockeye; chum, Oncorhynchus quantitatively link the vulnerability of a group (in keta; coho; and Chinook, Oncorhynchus tshawytscha , this case, herring eggs) to the biomass of a mediat- both wild and hatchery stocks), geoduck, Pacif c ing group (in this case, eelgrass): in other words, the herring, and Dungeness crab, in that order. By value, less eelgrass is present, the more vulnerable herring the same f sheries dominate but the geoduck’s high eggs are to their predators. This is but one of numer- price per pound makes it the most economically (a)

5.5 Transient orcas

Resident orcas 5.0

Porpoises Sea lions 4.5 Harbor seals Adult lingcod Diving birds Six gill shark Dogfish Juv. lingcod Chinook salmon Coho salmon 4.0 Large rockfish Skates

Octopus Pisc. flatfish Small rockfish Demersal fish Cancer crab Chum salmon Gulls Squid 3.5 Large gadoids Pacific hake Adult herring Sockeye salmon Smelt Pink salmon Surf perches Ratfish Juv. herring Sand lance Lg. jellies Other flatfish Seastars Pred. snails Ducks + brants 3.0 Benthic shrimp

Trophic Level Lg. zooplankton Sm. jellies Sm. crustaceans 2.5 Barnacles Clams Euphausiids Filter feeders Copepods 2.0 Sea cucumbers Mussels Soft infauna Sm. zooplankton Deposit feeders Urchins Geoducks Sm. grazers 1.5

1.0 Overstory kelp Detritus Eelgrass Understory kelp Benthic algae Phytoplankton

(b)

5.5 Transient orcas

Resident orcas 5.0

Porpoises Sea lions 4.5 Adult lingcod Harbor seals Dogfish Juv. lingcod Six gill shark Chinook salmon Large rockfish Diving birds 4.0 Skates Adult herring

Pacific hake Chum salmon Pisc. flatfish Demersal fish Cancer crab Gulls

3.5 Squid Trophic Level Juv. herring Lg. jellies Ratfish 3.0 Ducks + brants

2.5 Sm. crustaceans

Eelgrass

Figure 17.3 (a) The structure of the EwE food web model of the Central Basin of Puget Sound (without f sheries) and (b) a subset of the EwE food web model focusing on eelgrass and herring. (a) Box size is proportional to standing stock biomass; line thickness is proportional to the f ow of energy/material from the prey to the predator. Red colors represent detritus and the portion of the food web it supports, blues are benthic primary producers and those they support, and greens are phytoplankton and phytoplankton-supported groups. Consumers’ colors are a mix proportional to the amount of production that ultimately stems from those sources. In (b) dashed arrows indicate groups whose predation on herring eggs is mediated by the biomass of eelgrass. Colors are as those in (a). (See Plate 15.) MAPPING AND MODELING THE FLOW OF MARINE ECOSYSTEM SERVICES: A CASE STUDY OF PUGET SOUND 315 valuable commercial species ($5.4 million, average line in terms of how nearshore conditions deter- annual revenue, 2005–7; PacFIN data). This value is mine Puget Sound ecosystem service values. Two split about evenly between the northern and south- primary roadblocks prevent such an aggregation at ern parts of the Central Basin. Salmon and herring this point: (1) we lack complete data on these val- harvest, valued annually at $3.8 million and $159.3 ues—commercial species are relatively easy to thousand respectively, however, occur predomi- value, non-consumptive value species are not; and nantly in the southern part of the Central Basin. (2) we have not modeled the non-trophic relation- Fifteen functional groups in the food web, including ships between eelgrass and a number of important orcas, seals, ducks, sea stars, and the three groups of species (e.g., Chinook salmon, chum salmon, and wild salmon, arguably have non-consumptive eco- Dungeness crab). nomic values based on outdoor recreation or simply To address these two issues, we assume that the for their existence ( Table 17.3 ). non-consumptive value of a species is related to How are these values affected by the changes in population numbers, using pounds as a proxy nearshore conditions we have modeled? In biologi- metric for these values; and we limit our discus- cal terms, the food web model results show that the sion below to species whose primary interaction abundance of some species increases while it with eelgrass is through direct consumption of decreases for others. An ecosystem service value eelgrass or is mediated through their interactions framework provides us with a way of evaluating with herring (Table 17.3 ). Among these species is these trade-offs. Ideally, expressing values in a com- Pacif c hake (or whiting), Merluccius productus , mon metric (dollars) enables one to make a grand which is currently considered a “species of con- aggregation of all the changes, producing a bottom cern” by NOAA Fisheries. This status indicates some concern about the viability of the species but Table 17.3 Groups with signif cant commercial and non-consumptive insuff cient information is available to make a for- values in the Puget Sound food web mal ruling ( National Marine Fisheries Service

Commercial harvest value Non-consumptive 2004 ). We consider this species separately, then, value (group) because further decreases in its status might trigger additional legal protections. Other species that Geoducks Transient orcas (cetacean)* currently have legal protection under the ESA Sockeye and chum salmon Resident orcas (cetacean)* (wild & hatchery) (i.e., Chinook and summer chum salmon, steel- Chinook and coho salmon Porpoises (cetacean)* head, orca) are not examined here because they (wild & hatchery) neither consume eelgrass nor primarily interact Dungeness crab Gray whales (cetacean)* with eelgrass indirectly through their interactions Clams (various spp.) Harbor seals (pinniped)* with herring. Pacif c herring* Sea lions (pinniped)* For the limited set of commercial species identi- Shrimp (various spp.) Gulls (bird)* f ed in this way, the herring f shery is the most Sea cucumbers Piscivorous diving birds (bird)* important (in pounds harvested and revenue). Dogf sh* Murrelets (bird)* Assuming the harvest rates for all species do not Burrowing shrimp Ducks and brants (bird)* change, total herring harvest responds positively to Wild pink salmon Seastars (invertebrate) changes in eelgrass, approximately doubling as eel- Surf smelt* Wild Chinook and coho salmon Sea urchins Wild pink salmon salmon grass ranges from 50 to 200% of its baseline level. In Wild sockeye and chum salmon contrast, spiny dogf sh harvest increases by about Pacif c hake (species of concern) * 24% over that range, and surf smelt harvest decreases by about 23%; however, harvest yields for * Species whose primary interaction with eelgrass is mediated through their interactions with herring (or who directly consume eelgrass). both of these species are considerably less than that The commercial list includes all species groups with more than $1000 annual of herring. Over the modeled range of eelgrass harvest in 2003–7. The non-consumptive group is subjectively chosen to repre- changes and using the average prices for each f sh- sent species humans care about. The “group” for species of non-consumptive values indicates assignments to taxonomic groups for analysis of responses to ery (2005–7; PacFIN), the total harvest revenue for eelgrass and herring perturbations in the EwE model. this limited set of commercial species would 316 ECOSYSTEM SERVICE ASSESSMENTS FOR MARINE CONSERVATION increase by 82% or $942 000 as eelgrass increases to nearshore habitat conditions. Our point here, how- 200% of its baseline. ever, is not so much to “accurately” depict the ecol- For non-consumptive value species, the aggre- ogy of central Puget Sound, as to illustrate some gate weight of this group is negatively related to important issues for using ecosystem service val- increases in eelgrass biomass, decreasing by 15% ues. Commercial f sheries harvests are one of the across the range of eelgrass levels. Expressing their most straightforward and easily measured ecosys- total value as a simple summation of pounds, how- tem service values, and so producing a credible ever, implicitly assumes that these species have an “bottom line” for this ecosystem service is possible equal per-lb economic value. Although data are not as long as there are credible food web models. The available to provide any guidance to differentiate modeling also allows us to understand that trade- these values, dividing the group into subgroups offs among individual f sheries are still possible, def ned by taxonomy and legal status reveals poten- and so improvements in ecological conditions may tially important differences (Figure 17.4 ). Pinnipeds not be universally supported. The same can be seen have a strong positive relation with eelgrass bio- in the trade-offs among non-consumptive value mass, birds and cetaceans have a very weak posi- species. tive relation, and Pacif c hake exhibits a negative relation. 17.3.4 Suites of ecosystem services in space These results are heavily qualif ed, of course, by the absence in our modeling to date of ecological Overlaying carbon storage and sequestration serv- relations between eelgrass and species other than ices, and marine harvest and non-consumptive val- herring. It is not yet possible to assess the overall ues is another way to consider spatial variation in direction of the change in values captured in the ecosystem services. Herring spawn in twenty to Puget Sound food web in response to changes in twenty-one locations around Puget Sound and

1.40

1.30

1.20

1.10

1.00

0.90 Relative Biomass Relative

0.80

0.70

0.60 0.0 0.5 1.0 1.5 2.0 Eelgrass Index

Bird Invertebrate Cetacean Pinniped Pacific hake (ESA Species of Concern)

Figure 17.4 EwE model results for taxonomic groupings of non-consumptive value species (see Table 17.3 for group membership). The eelgrass index is eelgrass biomass/initial eelgrass biomass so values to the right of 1 represent increases in eelgrass biomass from the original baseline and values to the left represent decreases. FUTURE DIRECTIONS 317 observations during years of relatively high abun- tions are worth the lesser carbon services. An eco- dance suggest that they may expand their spawn- system service framework can answer this question ing activities adjacent to currently used meadows, easily if a common metric (e.g., dollars) is used to rather than colonizing new beds (Penttila 2007 ). measure the value of both sets of services. Even Therefore, eelgrass restoration, if undertaken, will absent that information, the framework can illus- likely produce more value in areas adjacent to doc- trate where trade-offs may exist among services. In umented herring spawning sites where the benef ts this way, policy-makers gain an understanding of of increased carbon storage and sequestration are the nature and extent of such trade-offs and can bet- most likely to be complemented by the benef ts of ter set priorities in accordance with public values. increased herring spawn, increased herring populations, and associated benef ts derived from the food web (e.g., Figure 17.5 ). 17.4 Future directions This example illustrates the need to consider ecosystem services en suite, rather than one-by-one. It The analyses presented here represent an initial might be, for example, that the most productive step in developing an ecosystem services frame- areas for eelgrass restoration in terms of carbon work to support ecosystem-based management in services are not adjacent to existing herring spawn- Puget Sound. In this f nal section, we sketch out ing locations. The question is then whether the additional steps that can move such a framework additional services derived from herring popula- closer to fruition.

Legend Herring Spawning Areas (WDFW) Mapped Eelgrass- EFH Suitable Areas for Eelgrass Land Kitsap Peninsula N

0123

Kilometers

Figure 17.5 A portion of the Kitsap peninsula in Central Puget Sound, showing current eelgrass beds (hatched), areas used by herring for spawning (stippling), and areas predicted to be suitable for eelgrass restoration (dark gray ). 318 ECOSYSTEM SERVICE ASSESSMENTS FOR MARINE CONSERVATION

A more complete evaluation of protection or res- using lessons learned from working in the Puget toration strategies will incorporate spatial variation Sound region to develop models for multiple marine in carbon storage and sequestration or food web ecosystem services. The Marine Initiative of the functions and assess specif c locations in terms of Natural Capital Project is developing a marine their current or potential production of these bene- InVEST tool that, like its terrestrial counterpart, will f ts. This information can provide useful guidance be an ecosystem services scenario assessment tool for for recovery of the Puget Sound nearshore by pro- application in ecosystem-based management proc- viding a map with sites categorized according to esses with diverse stakeholders and across multiple their likely ecosystem service benef ts under protec- scales. Building on the success of InVEST on land, we tion (for currently intact sites) or restoration (for will connect existing models through the land-sea sites that are currently degraded but with high interface to new and existing marine models. intrinsic potential) strategies. Ultimately, quantifying, mapping, and valuing Such an evaluation also needs to expand the marine ecosystem services has the potential to fun- modeling of marine ecosystem services beyond the damentally change the ways in which decisions current set covered. Incorporating more links about marine and coastal environments are made. between nearshore habitat conditions and the Making explicit the connections between human marine food web will allow us to investigate other activities in one sector and their effects on a broad potential trade-offs among the provisioning serv- range of other sectors forces decision-makers and ices of commercial f sheries for salmon and other the human communities they represent to think f nf sh, clam, oyster, crab, and other shellf sh har- about whole ecosystems and to manage them vests, as well as the numerous cultural services that accordingly. By making clear the life-sustaining include bird and whale watching and recreational services oceans and coasts provide, appropriately f shing. Waste treatment through the breakdown of valuing marine natural capital can help human PAHs (polycyclic aromatic hydrocarbons) and PCBs communities make better choices about how we use (polychlorinated biphenyls) by eelgrass should be these treasured environments. added to the list of marine ecosystem services included in the analysis. It also is possible in Puget Sound to extend the References ecosystem services approach to include upland activities and their associated ecosystem services— Aburto-Oropeza, O., Ezcurra, E., Danemann, G., et al . (2008). Mangroves in the Gulf of California increase basin-wide maps exist of current provisioning of f shery yields. Proceedings of the National Academy of water yields, water retention for f oods, water puri- Sciences, 105 , 10456–9. f cation potential, carbon storage, and commercial Agardy, T., Alder, J., Dayton, P., et al. (2005). Coastal sys- values of working landscapes in watersheds tems: assessment report, Chapter 19. Millennium ( Aukema et al. 2009 ; Rogers and Cooke 2009 ). The Ecosystem Assessment. Partnership is interested in understanding how Ainsworth, C., Varkey, D., and Pitcher, T. (2008). Ecosystem those watershed-based ecosystem benef ts affect simulations supporting ecosystem-based f sheries man- nearshore services provided to inform how and agement in the Coral Triangle, Indonesia. Ecological where to encourage different land-use practices Modeling, 214 , 361–74. around the region. Alderdice, D. F., and Hourston, A. S. (1985). Factors inf u- Clearly, developing a framework for assessing encing development and survival of Pacif c Herring (Clupea harengus pallasi) eggs and larvae to beginning of marine ecosystem services useful to policy-makers is exogenous feeding. Canadian Journal of Fisheries and an ambitious undertaking. Marine systems lack the Aquatic Sciences, 42 , 56–68. commonly available spatial data that inform assess- Alliance for Puget Sound Shorelines. (2008). http://www. ments of terrestrial services. As a result, building an shorelinealliance.org/ . assessment toolkit for marine environments may Alongi, D. M. (2008). Mangrove forests: Resilience, protec- always be reliant on a richer set of local data and tion from tsunamis, and responses to global climate models developed for a particular location. We are change. Estuarine, Coastal and Shelf Science, 76 , 1–13. FUTURE DIRECTIONS 319

Aukema, J., Vigerstol, K., and Foster, J. (2009). Application Cochard, S. L., Ranamukhaarachchi, G. P., Shivakotib, of InVEST models in Puget Sound , March 20, 2009. O. V., et al. (2008). The 2004 tsunami in Aceh and Unpublished report available from authors upon Southern Thailand: A review on coastal ecosystems, request. wave hazards and vulnerability. Perspectives in Plant Bando, K. J. (2006). The roles of competition and distur- Ecology, Evolution and Systematics, 10 , 3–40. bance in a marine invasion. Biological Invasions, 8 , Colling, A. (2001). Ocean circulation. Butterworth 755–63. Heineman/Open University, Milton Keynes. Barbier, E. B. (2000). Valuing the environment as input: Costanza, R. (2000). The ecological, economic and social Applications to mangrove-f shery linkages. Ecological importance of the oceans. In: C. R. C. Sheppard, Ed., Economics, 35 , 47–61. Seas at the millenium: an environmental evaluation, vol. 3: Barbier, E. B. (2003). Habitat-f shery linkages and man- global issues and processes. Pergamon Press, New York. grove loss in Thailand. Contemporary Economic Policy , Costanza, R., dArge, R., deGroot, R., et al. (1997). The value 21 , 59–77. of the world’s ecosystem services and natural capital. Barbier, E. B. (2007). Valuing ecosystem services as pro- Nature , 387 , 253–60. ductive inputs. Economic Policy , 22 , 177–229. Council on Environmental Quality. (2009). Interim Barbier, E. B., Koch, E. W., Silliman, B. R., et al . (2008). Report of the Interagency Ocean Policy Task Force. Coastal Ecosystem-Based Management with Nonlinear September 10, 2009. available at: http://www.white- Ecological Functions and Values. Science, 319 , 321–3. house.gov/administration/eop/ceq/initiatives/ Batie, S., and Wilson, J. (1978). Economic values attributa- oceans/interimrep . ble to Virginia’s coastal wetlands as inputs in oyster Culliton, T. (1998). Population: distribution, density, and production. Southern Journal of Agricultural Economics, growth . National Oceanic and Atmospheric Admini- 10 , 111–18. stration, Silver Spring, MD. Bell, F. (1989). Application of wetland valuation theory to Danielsen, F., Sorensen, M., Olwig, M., et al . (2005). The Florida f sheries . Florida Sea Grant Program, Florida State Asian tsunami: a protective role for coastal vegetation. University, Tallahassee. Science , 370 , 643. Bennett, E. L., and Reynolds, C. J. (1993). The value of a Davies, J., Guerry, A., Ruckelshaus, M. (In preparation). mangrove area in Sarawak. Biodiversity and Conservation, Mapping the potential distrubution of shallow-subti- 2 , 359–75. dal eelgrass in the greater Puget Sound region of Brander, L. M., Van Beukering, P., and Cesar, H. S. J. (2007). Washington state. The recreational value of coral reefs: a meta-analysis. Dean, T., Haldorson, L., Laur, D., et al. (2000). The distribu- Ecological Economics , 63 , 209–18. tion of nearshore f shes in kelp and eelgrass communi- Burke, L., and Maidens, J. (2004). Reefs at risk in the ties in Prince William Sound, Alaska: Associations with Carribean. World Resources Institute, Washington, DC. vegetation and physical habitat characteristics. Available online at: http://pdf.wri.org/reefs_carib- Environmental Biology of Fishes , 57 , 271–87. bean_front.pdf Diaz, R., and Rosenberg, R. (2008). Spreading dead zones Burke, L., Greenhalgh, S., Prager, D., et al. (2008). Coastal and consequences for marine ecosystems. Science , 321 , capital—economic valuation of coral reefs in Tobago and St. 926–9. Lucia. World Resources Institute, Washington, DC. Duarte, C. M. (2002). The future of seagrass meadows. Online at: http://www.wri.org/project/coral-reefs . Environmental Conservation, 29 , 192–206. Carpenter, S., Caraco, N., Correll, D., et al . (1998). Nonpoint Duarte, C., and Cebrian, J. (1996). The fate of marine auto- pollution of surface waters with phosphorus and nitro- trophic production. Limnology and Oceanography, 41 , gen. Ecological Applications, 8 , 559–68. 1758–66. Carte, B. K. (1996). Biomedical potential of marine natural FAO Fisheries Department. (2009). The state of world f sher- products. Bioscience , 46 , 271–86. ies and aquaculture—2008 (SOFIA). FAO, Rome. Cesar, H. S. J., Ed. (2000). Collected essays on the economics of Farber, S. (1988). The value of coastal wetlands for recrea- coral reefs . CORDIO, Kalmar Univeristy, Sweden. tion—an application of travel cost and contingent valua- Chmura, G., Anisfeld, S., Cahoon, D., et al . (2003). Global tion methodologies. Journal of Environmental Management, carbon sequestration in tidal, saline wetlands soils. 26 , 2 9 9 – 3 1 2 . Global Biogeochemical Cycles , 17 , 1111. Forbes, K., and Broadhead, J. (2007). The role of coastal forests in Christensen, V., and Walters, C. (2004). Ecopath with the mitigation of tsunami impacts. RAP Publication 2007/1, Ecosim: methods, capabilities and limitations. Ecological Food and Agricultural Organization of the United Nations, Modeling , 172 , 109–39. Regional Off ce for Asia and the Pacif c, Bangkok. 320 ECOSYSTEM SERVICE ASSESSMENTS FOR MARINE CONSERVATION

Fulton, E. A., Smith, A. D. M., and Johnson, C. R. (2004a). Hosack, G. R., Dumbauld, B. R., Ruesink, J .L., et al . (2006). Biogeochemical marine ecosystem models I: IGBEM - a Habitat associations of estuarine species: Comparisons model of marine bay ecosystems. Ecological Modeling , of intertidal mudf at, seagrass ( Zostera marina), and oys- 174 , 267–307. ter ( Crassostrea gigas) habitats. Estuaries and Coasts, 29 , Fulton, E. A., Parslow, J. S., Smith, A. D. M., et al . (2004b). 1150–60. Biogeochemical marine ecosystem models II: the effect Huesemann, M., Hausmann, T., Fortman, T., Thom, R. of physiological detail on model performance. Ecological and Cullinan, V. (2009). In-situ phytoremediation of Modeling, 173 , 371–406. PAH and PCB contaminated marine sediments with Gacia, E., Granata, T. C., and Duarte, C. M. (1999). An eelgrass ( Zostera marina ). Ecological Engineering , 35 , approach to measurement of particle f ux and sediment 1395–1404. retention within seagrass (Posidonia oceanica ) meadows. Iceland, C., Hanson, C., and Lewis, C. (2008). Identifying Aquatic Botany , 65 , 255–68. important ecosystem goods and services in Puget Sound . Gacia, E., and Duarte, C. (1999). Sediment retention by a World Resources Institute, Washington, DC. Mediterranean Posidonia oceanica meadow: the balance Intergovernmental Panel on Climate Change—National between deposition and resuspension. Estuarine and Greenhouse Gas Inventories Programme. (2006). IPCC Coastal Shelf Science, 52 , 505–14. Guidelines for National Greenhouse Gas Inventories , vol. 4: Gaeckle, J., Dowty, P., Reeves, B., et al. (2007). Puget Sound agriculture, forestry and other land use. Available at: Submerged Vegetation Monitoring Project, 2005 Monitoring http://www.ipcc-nggip.iges.or.jp/public/2006gl/vol4. Report. Puget Sound Assessment and Monitoring html Program, Washington State Department of Natural Jesperson, J., and Osher, L.. (2007). Carbon storage in the Resources, Nearshore Habitat Program, Aquatic soils of a mesotidal Gulf of Maine estuary. Soil Science Resources Division, Olympia. Society of America Journal, 71 , 372–9. Galkina, L. A. (1971). Survival of spawn of the Pacif c her- Kentula, M. E., and Mcintire, C. D. (1986). The autecology ring (Clupea harengus pallasii val. ) related to the abun- and production dynamics of eelgrass (Zostera marina L ) dance of the spawning stock. Rapports et Proces Verbaux in Netarts Bay, Oregon. Estuaries, 9 , 188–99. des Reunions , 160 , 30–3. King County. (2007). King County climate plan. Available Gilbert, A. J., and Janssen, R. (1998). Use of environmental at: http://www.metrokc.gov/exec/news/2007/pdf/ functions to communicate the values of a mangrove climateplan.pdf ecosystem under different management regimes. Lippke, B., Garcia, J. P., and Manriquez, C. (2003). Executive Ecological Economics , 25 , 323–46. summary: the impacts of forests and forest management on Gleick, P. (1996). Water resources. In: S. Schneider, Ed., carbon storage. Rural Technology Initiative, College of Encyclopedia of climate and weather. Oxford University Forest Resources, University of Washington. Press, New York. Lynne, G. D., Conroy, P., and Prochaska, F. J. (1981). Haegele, C. W. (1993a). Seabird predation of Pacif c her- Economic valuation of marsh areas for marine produc- ring, Clupea pallasi, spawn in British Columbia. Canadian tion processes. Journal of Environmental Economics and Field Naturalist, 107 , 73–82. Management , 8 , 175–86. Haegele, C. W. (1993b). Epibenthic invertebrate preda- McRoy, C. (1974). Seagrass productivity: carbon uptake tion of Pacific herring, Clupea pallasi, spawn in experiments in eelgrass, Zostera marina. Aquaculture , 4 , British Columbia. Canadian Field Naturalist , 107 , 131–7. 83–91. Mallin, M. A., Williams, K. E., Esham, E. C., et al . (2000). Harvey, C. J., Bartz, K.K., Davies, J., et al. (2010). A mass- Effect of human development on bacteriological water balanced model for evaluating food web structure and quality in coastal watersheds. Ecological Applications, 10 , community-scale indicators in the central basin of 1047–56. Pudget Sound. U.S. Dept. Commer., NOAA Tech. Mann, K. (1982). Ecology of coastal waters. University of Memo. NMFS-NWFSC-106, 180. California Press, Berkeley. Holmlund, C. M., and Hammer, M. (1999). Ecosystem Mateo, M. A., Cebrian, J., Dunton, K., et al . (2006). Carbon services generated by f sh populations. Ecological f ux in seagrass ecosystems. In: A. W. D. Larkum, R. J. Economics , 29 , 253–68. Orth, and C. M. Duarte, Eds., Seagrasses: biology, ecology, Homann, P. S., Remillard, S. M., Harmon, M. E., et al . and conservation. Springer, Dordrecht. (2004). Carbon storage in coarse and f ne fractions of Mazda, Y., Magi, M., Kogo, M., et al. (1997). Mangroves as Pacif c Northwest old-growth forest soils. Soil Science a coastal protection from waves in the Tong King Delta, Society of America Journal , 68 , 2023–30. Vietnam. Mangroves and Salt Marshes 1 , 127–35. FUTURE DIRECTIONS 321

Melillo, J. M., Mcguire, A. D., Kicklighter, D. W., et al . Pergams, O., and Kareiva, P. (In press). Support serv- (1993). Global climate-change and terrestrial net pri- ices: A focus on genetic diversity. In: A. Kinzig, Ed., mary production. Nature , 363 , 234–40. Ecosystem services. Princeton University Press, Millennium Ecosystem Assessment. (2005). Ecosystems and Princeton, NJ. human well-being: wetlands and water synthesis . World Peterson, C., and Lubchenco, J. (1997). Marine ecosystem Resources Institute, Washington, DC. services. In: G. Daily, Ed., Nature’s services . Island Press, Moberg, F., and Folke, C. (1999). Ecological goods and Washington, DC. services of coral reef ecosystems. Ecological Economics , Pew Oceans Commission. (2003). America’s living oceans: 29 , 215–33. charting a course for sea change. A report to the nation . Pew Mumford, T. F. J. (2007). Kelp and eelgrass in Puget Sound . Oceans Commission, Arlington, VA. Seattle District, US Army Corps of Engineers, Seattle, WA. Plummer, M.L. (2009). Assessing benef t transfer for the National Marine Fisheries Service. (2007). Sound science: valuation of ecosystem services. Frontiers in Ecology and synthesizing ecological and socio-economic information about the Environment , 7 , 38–45. the Puget Sound ecosystem. US Department of Commerce, Puget Sound Nearshore Ecosystem Restoration Project. National Oceanic and Atmospheric Administration, (2008). Puget Sound Future Scenarios . Draft report pre- Northwest Fisheries Science Center, Seattle, WA. pared by University of Washington Urban Ecology National Marine Fisheries Service. (2004). Endangered Research Lab. Available at: http://www.pugetsound- and threatened species; establishment of species of con- nearshore.org/index.htm cern list, addition of species to species of concern list, Puget Sound Partnership. (2006). Interim Report to the description of factors for identifying species of concern, Governor , Puget Sound Partnership. and revision of candidate species list under the Puget Sound Partnership. (2008). Action Agenda . Puget Endangered Species Act. Federal Register, 73 , 19975–9. Sound Partnership. December, 2008. Available at: National Research Council, Committee on Assessing and http://www.psp.wa.gov/aa_action_agenda.php Valuing the Services of Aquatic and Related Terrestrial Rogers, L. W., and Cooke, A. G. (2009). The 2007 Washington Ecosystems. (2004). Valuing ecosystem services: toward bet- State Forestland Database: Final Report. Prepared for the ter environmental decision-making . National Academies USDA Forest Service. Available at: http:/ /www. Press, Washington, DC. ruraltech.org/projects/wrl/f db/ Nelson, T., and Waaland, J. (1997). Seasonality of eelgrass, Ronnback, P. (1999). The ecological basis for economic epiphyte, and grazer biomass and productivity in subti- value of seafood production supported by mangrove dal eelgrass meadows subjected to moderate tidal ecosystems. Ecological Economics , 29 , 235–52. amplitude. Aquatic Botany , 56 , 51–74. Rosenberg, A., and McLeod, K. (2005). Implementing eco- Nordhaus, W. (2007). A review of the Stern Review on the system-based approaches to management for the con- economics of climate change. Journal of Economic servation of ecosystem services. Marine Ecology Progress Literature, 45 , 685–702. Series , 300 , 270–74. Orth, R. J., Heck, K. L., and Vanmontfrans, J. (1984). Faunal Rosenfeld, D., Lahav, R., Khain, A., et al. (2002). The role of communities in seagrass beds—a review of the inf u- sea spray in cleansing air pollution over ocean via cloud ence of plant structure and prey characteristics on pred- processes. Science , 297 , 1667–70. ator prey relationships. Estuaries , 7 , 339–50. Ruitenbeek, H. J. (1994). Modeling economy ecology link- Patterson, M., and Glavovic, B. Eds. (2008). Ecological eco- ages in mangroves—economic evidence for promoting nomics of the oceans and coasts. Edward Elgar, conservation in Bintuni Bay, Indonesia. Ecological Cheltenham. Economics , 10 , 233–47. Pauly, D., Christensen, V., and Walters, C. (2000). Ecopath, Sathirathai, S., and Barbier, E. B. (2001). Valuing mangrove Ecosim, and Ecospace as tools for evaluating ecosystem conservation in southern Thailand. Contemporary impact of f sheries. ICES Journal of Marine Science , 57 , Economic Policy , 19 , 109–22. 697–706. Schlesinger, W. (1991). Biogeochemistry: an analysis of global Payne, M. K. (2007). Landscape-level assessment of sub- change . Academic Press, San Diego. aqueous soils and water quality in shallow embayments in Shared Strategy for Puget Sound. (2007). Puget Sound Southern New England. Masters thesis, Department of salmon recovery plan . Plan adopted by the National Natural Resource Science, University of Rhode Island, Marine Fisheries Service. January 19, 2007. Seattle, WA. Kingston, RI. Soderqvist, T., Eggert, H., Olsson, B., et al . (2005). Economic Penttila, D. (2007). Marine forage f shes in Puget Sound . Seattle valuation for sustainable development in the Swedish District, US Army Corps of Engineers, Seattle, WA. coastal zone. Ambio , 34 , 169–75. 322 ECOSYSTEM SERVICE ASSESSMENTS FOR MARINE CONSERVATION

Spurgeon, J. P. G. (1992). The economic valuation of coral US Environmental Protection Agency. (2007). Inventory of reefs. Marine Pollution Bulletin , 24 , 529–36. U.S. greenhouse gas emissions and sinks: 1990–2005, Annex Stevenson, J. C. (1988). Comparative ecology of submersed 3.12, Methodology for estimating net carbon stock changes in grass beds in freshwater, estuarine, and marine environ- forest lands remaining forest lands . US Environmental ments. Limnology and Oceanography , 33 , 867–93. Protection Agency, Washington, DC. Taylor, F. H. C. (1971). Variation in hatching success in US Geological Survey. (1997). ATLSS: across-trophic-level pacif c herring ( Clupea pallasii ) eggs with water depth, system simulation: an approach to analysis of South Florida tempertature, salinity and egg mass thickness. Rapports ecosystems. Biological Resrources Division USGS, et Proces Verbaux des Reunions , 160 , 34–41. Miami, FL. TerraLogic GIS Inc. (2004). Public seagrass compilation for US Commission on Ocean Policy. (2004). An ocean blue- west coast Essential Fish Habitat (EFH) Environmental print for the 21st century. US Commission on Ocean Impact Statement . Pacif c States Marine Fisheries Policy, Washington, DC. Commission, Portland, OR. Wang, Y., and Hseih, Y. (2002). Uncertainties and novel Thayer, G., and Phillips, S. (1977). Importance of prospects in the study of the soil carbon dynamics. Eelgrass Beds in Puget Sound. Marine Fisheries Chemosphere , 49 , 703–24. Review , 39 , 1 8 – 2 2 . Webber, H. H., Mumford, T. J., and Eby, J. (1987). Remote Thom, R. M., Borde, A. B., Williams, G. D., et al . (2003). sensing inventory of the seagrass meadow of the Padilla Bay Climate change and seagrasses: climate-linked dynam- National Estuarine Research Reserve: areal extent and esti- ics, carbon limitation and carbon sequestration. Gulf of mation of biomass. NOAA technical report series OCRM/ Mexico Science , 21 , 134. MEMD, Reprint series 6. Padilla Bay National Estuarine Thom, R. M. (1990). Spatial and temporal patterns in plant Research Reserve, Seattle. standing stock and primary production in a temperate Weitzman, M. (2007). A review of the Stern Review on the seagrass system. Botanica Marina , 33 , 497–510. economics of climate change. Journal of Economic Thom, R. M., and Albright, R. G. (1990). Dynamics of ben- Literature , 45 , 703–24. thic vegetation standing-stock, irradiance, and water Williams, S., and Heck, K. L. (2001). Seagrass communi- properties in Central Puget Sound. Marine Biology , 104 , ties. In: M. Bertness, S. Gaines, and M. Hay, Eds., Marine 129–41. community ecology. Sinauer, Sunderland. Thom, R. M., and Hallum, L. (1990). Long-term changes in Wilson, M. A., and Liu, S. (2008). Evaluating the non- the areal extent of tidal marshes, eelgrass meadows and kelp market value of ecosystem goods and services pro- forests of Puget Sound . FRI-UW-9008. vided by coastal and nearshore marine system. In: M. Travis, J. (2005). Hurricane Katrina: scientists’ fears come G. Patterson and B. C. Glavovic, Eds., The ecological eco- true as hurricane f oods New Orleans. Science , 309 , nomics of the oceans and coasts . Edward Elgar, 1656–9. Cheltenham. United Nations Environment Programme. (2006). Wolanski, E. (2007). Estuarine ecohydrology . Elsevier, Marine and coastal ecosystems and human well-being: a Amsterdam. synthesis report based on the f ndings of the Millennium Woodbury, P. B., Smith, J. E., and Heath, L. S. (2007). Ecosystem Assessment. UN Environment Programme, Carbon sequestration in the US forest sector from Nairobi. 1990 to 2010. Forest Ecology and Management , 241 , US Department of Energy. (1999). Carbon sequestration: the 14–27. state of the science: carbon sequestration research and devel- World Travel and Tourism Council (WTTC). (2005). opment . US Department of Energy, Off ce of Fossil Trinidad and Tobago: the impact of travel and tourism on jobs Energy, Washington, DC. and the economy. CHAPTER 18 Modeling the impacts of climate change on ecosystem services

Joshua J. Lawler, Erik Nelson, Marc Conte, Sarah L. Shafer, Driss Ennaanay, and Guillermo Mendoza

18.1 Introduction on water availability. Climate change is projected to increase surface-water availability at far northern The Earth’s climate has already changed signif - latitudes, across much of Asia, and parts of eastern cantly in response to increased greenhouse-gas con- Africa and southeastern South America ( Milly et al. centrations from human activities, and future 2005 ). In contrast, water availability is projected to climatic changes are projected to be even more dra- decrease in the Middle East, mid-latitude western matic, with global average temperatures expected North America, southern Africa, and southern to rise between 1.1 and 6.4˚C by 2100, depending on Europe ( Milly et al. 2005 ). However, not all studies future emissions from human activities ( Alley et al. agree on the geographic distribution of potential 2007 ). With the rise in temperature will also come future hydrologic changes, particularly when changes in precipitation patterns, with some areas human population growth is taken into account becoming drier, and others wetter. Climate change ( Vörösmarty et al. 2000 ; Arnell 2004 ). Furthermore, of this magnitude is certain to have major impacts the effects of changes in temperature and precipita- on basic ecosystem processes such as primary pro- tion on water availability are not necessarily simple. duction, hydrological cycles, and nutrient cycles For example, across Africa, decreases in precipita- ( Cramer et al. 2001 ; Stewart et al. 2005 ; Betts et al. tion are projected to have very different effects on 2007 ). In this chapter, we discuss how models that runoff in areas with different annual rainfall regimes link land cover and land use to ecosystem services (de Wit and Stankiewicz 2006 ). A 10% decrease in can also be used to assess the impact of climate precipitation was projected to result in a 20% change on ecosystem services. Many studies have decrease in runoff in Ouagadougou, Burkina Faso, addressed how specif c ecosystem services, particu- whereas the same decrease in precipitation was larly water availability and agricultural production projected to produce a 77% decrease in runoff in or food security, are expected to respond to climate Okavango, Botswana ( de Wit and Stankiewicz change. Here, we go one step beyond and demon- 2006 ). strate how climate impacts can be integrated by Changes in water availability as well as changes examining several ecosystem services within the in temperature and atmospheric CO 2 concentra- same framework. tions will affect water demand, agricultural production, and, in some cases, food security (Alcamo and 18.2 Previous analyses of climate- Henrichs 2002 ; Arnell 2004; Battisti and Naylor driven changes in ecosystem services 2009 ). For example, there are clear linkages between recent changes in temperature and crop production Much of the world’s population relies on surface (Box 18.1)—increases in temperature have been water for drinking and irrigation. Given the clear linked to overall decreases in corn and soybean links between climate and hydrology, projected yields in the USA ( Lobell and Asner 2003 ). As one changes in climate will likely have signif cant effects would expect, however, the direction of this trend

323 324 MODELING THE IMPACTS OF CLIMATE CHANGE ON ECOSYSTEM SERVICES differs by geographic region—US corn and soybean services are based on average annual inputs of cli- yields in cooler areas have responded positively to matic conditions and provide annual average out- warming whereas yields in warmer areas have puts of service provision. Therefore, in regions responded negatively. Globally, crop yields have where climate change is expected to affect annual been projected to increase with a doubling of atmos- precipitation and temperature patterns, the tier 1 pheric CO2 concentrations ( Rosenzweig and Parry water-related service models provide informative 1994 ). Again, however, some regions will likely but temporally-coarse projections. However, in experience gains in crop yields while others will many cases, climate change is expected to shift sea- experience signif cant losses. sonal precipitation patterns, but leave annual total Individual wild species will also respond to cli- precipitation relatively unchanged. For these cases, mate change, resulting in changes in the ecosystem tier 1 models will be inappropriate and the ramif - services they provide. The most frequently docu- cations of climate change on water services can only mented effects of climate change on wild species be estimated with water-related service models that include shifts in distributions and shifts in phenol- make projections based on monthly or daily inputs ogy ( Walther et al. 2002 ; Parmesan and Yohe 2003 ; (tier 2 models). Root et al. 2003 ; Parmesan 2006 ). Most directly, the In most cases, the agricultural production models abundance or location of species that are themselves described in Chapter 9 can also be directly applied directly harvested for food may be altered by cli- to climate simulations. If the tier 1 agriculture pro- mate change. For example, increases in temperature duction model employs yield functions that include will likely reduce the overall range of many cool- average temperature or total water budget during and coldwater f sh, including salmonids. Further- the growing season as an input, then the effect of more, reductions in snowpack have the potential to changes in these variables on agriculture produc- alter f ow regimes potentially affecting spawning tion are easily simulated. The tier 2 agriculture pro- habitat and egg survival of salmonids. Battin et al. duction model discussed in Chapter 9 uses climate (2007 ) explored the potential combined effects of inputs at a daily or monthly resolution. In most climate change and stream restoration on Chinook cases, tier 2 models will more accurately capture salmon in a watershed in western Washington in changes in agricultural production because growth the western United States. Projected changes in the in most agricultural plants is a process that operates modeled salmon population for 2050 ranged from a on the scale of hours to months, not seasons to decrease of 40% to an increase of 19% depending on years. the climate-change scenario that was used as well Climate will also indirectly alter ecosystem serv- as the extent of habitat restoration that was assumed ices by altering vegetation type, land cover, and to take place. It is likely that studies of other eco- even land management ( Bachelet et al. 2001 ; nomically important wild species will also reveal Ramankutty et al. 2002 ). Climate is also known to changes in their productivity as a result of climate inf uence the abundance and distribution of plants change. and animals and in turn biodiversity. Several different approaches can be taken to modeling climate- 18.3 Using ecosystem-service models to driven shifts in land cover, land management, and evaluate the impact of climate change individual species. For example, dynamic global on natural and human systems vegetation models can be used to project changes in the distribution of plant functional types which Water supply is dependent on climate, geology, can be translated into maps of potential vegetation topography, land cover and land management change (e.g., Sitch et al. 2003 ). Other approaches ( Chapter 4 , this volume). By combining land-use take both climate-driven changes in potential vege- and land-cover data with climate projections it is tation and climate-change inf uenced land-use possible to examine likely changes in intercon- decisions into account (e.g., Bouwman et al. 2006 ). nected ecosystem services (see Chapters 5 through There are likewise different approaches to mode- 13 ). The simplest models (tier 1) for water-related ling climate-driven changes in individual species USING ECOSYSTEM-SERVICE MODELS TO EVALUATE THE IMPACT OF CLIMATE CHANGE 325

Box 18.1 An estimate of the effects of climate change on global agricultural ecosystem services

David Lobell tons. If we use $300 per ton as an average price for cereals, this amounts to $6 of annual losses for each To estimate climate impacts on agricultural services, we decade of warming. must consider two questions. First, how much is food These impacts relate only to warming. For a more worth? A simple measure is the world market price for an complete accounting, one would also want to evaluate incremental ton of food (this ignores ethical aspects such changes in atmospheric carbon dioxide, ozone, and as the value of a human life). As of August 2008, the prices patterns of precipitation. For example, future increases in per ton of wheat, rice, and maize—the three most widely the frequency and intensity of droughts or heavy rains grown crops—were roughly $330, $730, and $200, could hamper agriculture in many regions. In general, the respectively. Second, we must consider how climate largest net impacts are likely to be in systems with C4 changes are affecting production of these crops, a task crops (maize, sorghum, millet, sugarcane) that do not made diff cult because farmers are constantly adjusting to benef t greatly from higher CO . These crops are especially their environment. Numerous studies have shown that 2 important to millions of poor in Africa, as well as global incremental warming is clearly benef cial to production in livestock production and the rapidly growing biofuels some regions (e.g. Siberia) and harmful in others (e.g. industry. India), even after accounting for farmer adjustments. Thus, the answer to the question at hand clearly depends on the region and scale of interest. Focusing on the global scale, one approach to 10 estimating how climate changes are affecting agricultural production is to compare year-to-year changes in crop 5 yields with year-to-year changes in average temperatures over the areas and seasons where the crop is grown 0 (Figure 18.A.1 ). This approach ignores the many important aspects of climate other than average temperature, but –5 trends for most of these other aspects are much less Change (%) Yield pronounced than for temperature. As an average, global –10 production of several key crops exhibit a linear decline with warming of roughly 5–8% per degree, a number that –1.0 –0.5 0 0.5 agrees well with many site-level studies. According to the GS Temperature Change (C) IPCC, global warming in the past 25 years has proceeded at approximately 0.18 ˚C per decade. Combining these two Figure 18.A.1 The relationship between year-to-year changes in numbers we see that warming costs roughly 1% of global globally averaged maize yields and average growing season (GS) production per decade. Global cereal production is roughly temperatures over maize growing regions, 1961–2002 (Data from 2 billion tons per year, so that 1% represents 20 million Lobell and Field ( 2007 )).

distributions or abundances—both correlative and Conversely studies of the impact of climate on bio- mechanistic models with a range of complexities diversity have neglected concordant impacts on have been used (e.g., Pearson et al. 2002 ; Carroll ecosystem services. Here we examine the Willamette 2007 ). The outputs of these models can also be taken Basin in Oregon, USA, and consider climate impacts as inputs to ecosystem service models. on irrigation water demand among select crop pro- A primary concern of conservation is the spatial duction systems, carbon storage in forests, and bio- patterning of biodiversity and critical ecosystem diversity. As described below, we used a limited set services. Previous studies that assessed climate of climate simulations, vegetation model outputs, impacts on ecosystem services did not also include and projected species distributions as inputs to the impacts on biodiversity (e.g., Alcamo et al. 2005 ). ecosystem service models. These limited inputs do 326 MODELING THE IMPACTS OF CLIMATE CHANGE ON ECOSYSTEM SERVICES not take many of the important uncertainties in 165 km future climate-change projections or system responses into account. Furthermore, these input data are of relatively coarse spatial and/or tempo- Oregon ral resolution, further limiting the conclusions that Portland can be drawn from the results. Although the accuracy and generality of our model projections are limited, the examples we develop are illustrative of a general but useful approach to combining projec- Salem tions of climate impacts with spatially explicit maps Valley FloorAlbany Ecoregion of ecosystem services to determine how those serv- 270 km ices change under different climate scenarios. Corvallis

Coast Mountain Range Ecoregion

18.4 Climate impacts on ecosystem services in the Willamette Basin of Eugene Oregon

We focus on climate impacts on three services and the Cascades Mountain Range Ecoregion biodiversity conservation provided by the Willamette Basin of Oregon in the western United States. The Willamette Basin is a 2.93 million hectare watershed Figure 18.1 Map of the Willamette Basin, Oregon, USA. that includes the urban areas of Portland, Salem, Albany, Corvallis, and Eugene-Springf eld as well as signif cant areas of agricultural land on its central addition, the UKMO-HadCM3 model tends to pro- f oor and forests in the surrounding mountains. The duce climate-change projections that lie just above lowlands once supported expansive areas of oak the middle of the range of climate sensitivity def ned savannah and grasslands but little of these habitats by the other AOGCMs at the global scale. remain today. Most of the uplands are covered by Building the projected future climate dataset conifer forest ( Figure 18.1 ). We have chosen this val- involved downscaling recent historic climate data ley as the site for our case studies because it has a and then applying anomalies from the future cli- diversity of land uses and it allows us to take advan- mate simulation to this historic climate dataset. tage of the extensive data sets that have been created Monthly climate data for 1901–2002 were created for this area ( Schumaker et al. 2 0 0 4 ; P o l a s k y et al. 2005 , from the University of East Anglia Climatic 2 0 0 8 ; N e l s o n et al. 2009 ). To draw our conclusions, we Research Unit (CRU) CL 2.0 (New et al. 2 0 0 2 ) a n d rely on a single, middle of the twenty-f rst century, TS 2.1 (Mitchell and Jones 2005 ) climate data climate simulation. Projected spatial patterns of future applied to a 30-second grid (~0.6 km 2 f o r t h e temperature and precipitation were derived from the Willamette Valley) of the study area using a locally- UKMO-HadCM3 coupled atmosphere–ocean gen- weighted, lapse-rate-adjusted interpolation method eral circulation model (AOGCM; Gordon et al. 2000) developed by P. J. Bartlein (University of Oregon, driven by the Intergovernmental Panel on Climate personal communication). For future climate data, Change (IPCC) Special Report on Emissions Scenarios we used monthly climate data simulated for (SRES) mid-high A2 emission scenario (Nakicenovic 2000–99 by the UKMO-HadCM3 coupled atmos- et al. 2000 ). Recent data show that anthropogenic phere–ocean general circulation model (AOGCM; greenhouse-gas emissions are already exceeding the Gordon et al. 2000) and obtained from the World higher SRES A1f emission scenario. Therefore, our Climate Research Programme’s (WCRP’s) Coupled use of the A2 scenario should yield relatively con- Model Intercomparison Project phase 3 (CMIP3) servative projections for changes to biodiversity and multi-model dataset. Anomalies were calculated as ecosystem services under a no-action scenario. In differences (for temperature data) and ratios (for CLIMATE IMPACTS ON ECOSYSTEM SERVICES IN THE WILLAMETTE BASIN OF OREGON 327 precipitation and percent cloud cover data) between climatic conditions. For the future time period, we each 2000–99 simulated monthly value and 1961–90 averaged the projected monthly precipitation totals 30-year mean monthly values calculated from the and monthly average temperatures from 2041 to UKMO-HadCM3 simulated monthly data for the 2070 and used these 30-year monthly means to rep- twentieth century. The anomalies were bi-linearly resent the average climatic conditions for the mid- interpolated to the 30-second grid of the study area dle of the century. In this example, we used observed and applied to CRU-derived 1961–90 30-year mean climate in 2000 to represent current (baseline) cli- data to create the future climate data used in this mate. Ideally, we would have used an averaged study. 30-year period (as we did with the future projec- Importantly, we were not simply interested in cli- tions) to represent baseline conditions. However, mate impacts, but climate impacts assuming differ- our choice of a baseline period was limited by the ent scenarios of land use and land cover. Projections availability of specif c crop-yield information in the of human land use were taken from Hulse et al. Basin. ( 2002 ) and represent alternative maps of agriculture, In addition to climate projections, forecasting forest, and residential and industrial development future irrigation demand requires an estimate of assuming three different development scenarios. where vineyards and berry f elds would be located The three scenarios include a “development” trajec- by the middle of the 21st century. We use the pro- tory that portrays an increase in development pres- jected “plan trend” land-use scenario to identify sure, a “plan trend” trajectory that continues counties that will likely be involved in agricultural development at current rates and in current pat- production in 2050 (Hulse et al. 2002 ). The allocation terns, and a “conservation” trajectory that limits of vineyards and berry f elds in a county in 2050 development in ways that promote the conservation was based on current allocations of these crop types of natural systems in the Basin. All three land-use in the county. Of all the counties in the Basin, only scenarios assume that human population in the Columbia County does not currently produce any Basin will increase from 2.0 million in 1990 to 3.9 of our focal agricultural products; therefore, we million people in 2050. excluded it from our irrigation analyses. By assuming the proportion of crop lands within a county dedicated to berry or grape production in 18.4.1 Water availability for agriculture in the 2050 would be the same as the proportion in 2000, Willamette Basin we have likely overestimated the area dedicated to At the heart of the Willamette Basin is a fertile river these crops in some regions and underestimated it valley that produces a wide variety of agricultural in others. It is likely that some farmers will take goods, ranging from staple grains to wine grapes. various measures to adapt to changes in climate. Although the region receives signif cant amounts of Farmers will inevitably cease the cultivation of precipitation, summers are dry and some of the crops that demand too much irrigation or that fail to crops grown in the valley must be irrigated to meet thrive under new climatic conditions. In some cases, production goals. Here, we explore the potential new crops will be planted and, in more extreme cir- impacts of future climate change on the provision of cumstances, agriculture will be abandoned entirely. freshwater relative to the demand for irrigation Although it would be more realistic to try to capture water for berries and wine grapes, two of the val- the effects of potential adaptation strategies in our ley’s most valuable crops. We determine, all else analyses, doing so would introduce additional being equal, how much irrigation water would be uncertainties. Early contributions of economic stud- necessary to achieve current crop yields under a ies that addressed climate-change impacts on agri- hypothesized future climate. cultural production attempted to allow for the To estimate climate-driven changes in irrigation mitigating impacts of adaptation (Mendelsohn et al. demand, we compared the volume of water needed 1994 ). More recent efforts, however, have aban- for agricultural production under current condi- doned the focus on adaptation to avoid the poten- tions to the volume required under projected future tial biases associated with the functional form 328 MODELING THE IMPACTS OF CLIMATE CHANGE ON ECOSYSTEM SERVICES assumptions necessary to allow for adaptation Most of the Basin, including the areas dedicated to ( Deschenes and Greenstone, 2007 ; Schlenker et al. wine grape and berry production, is projected to 2005 ). In the interest of simplicity, in our case stud- experience decreases in water yield ( Figure 18.2 ; ies, we did not model the potential effects of adap- Plate 16). The largest decreases are along the crest of tation strategies. the Cascade Range, but decreases are also forecast We estimated crop-specif c water requirements for much of the basin f oor. The largest increases are based on local climatic conditions. These require- projected for the foothills of the Cascade Range and ments are outputs of CropWat, a tool developed by the northwestern corner of the basin. The decreases the Land and Water Development Division of FAO in water yields, in conjunction with increasing tem- to aid in the design and management of irrigation peratures, led to increased irrigation demands in schemes ( FAO 1992 ). The inputs to this tool include berry and grape f elds across all counties expected to climate data, crop coeff cients (e.g., root depth and grow these products in 2050 ( Table 18.1; recall that depletion fraction) and growth stages, and soil data the spatial pattern of cropland in 2050 is based on for calculating plant water use. The growth-stage 2050 projections but the fraction of these lands in the specif c crop coeff cients for the different agricul- f ve focus crops is based on observed 2000 patterns). tural products were taken from AgriMet, a satellite- All counties were projected to require at least a 50% based network of automated agricultural weather increase in irrigation water to maintain current berry stations operated and maintained by the US Bureau and grape yields, with some counties expected to of Reclamation. Although the AgriMet data pro- see as much as an 85% increase in irrigation needs. vided the crop coeff cients as a percentage of These analyses make it clear that increased irriga- growth, the length of the growing season in CropWat tion demand for berries and grapes is one of is measured in days. The necessary crop coeff cient the potential impacts of climate change for the values were interpolated assuming linear plant Willamette Basin. Of course, our analyses were growth. The soil data for the counties in the region done at relatively coarse temporal and spatial scales, were derived from the SSURGO Soil Data Viewer whereas climatic extremes at f ne temporal scales, (USDA 2008 ). Using the climate, land-use, and soils such as frosts or water shortages, will signif cantly data, we ran CropWat for the f ve crops we focused impact agricultural yields. Changes in the nature or on, raspberries, blackberries, strawberries, blueber- frequency of climatic extremes, which are also asso- ries, and wine grapes. ciated with climate change ( Alley et al. 2007 ) and We translated the crop-speci f c water require- specif c weather events could signif cantly alter irri- ments from CropWat into the total volume of irriga- gation demands. Mismatches between precipitation tion water required within each county for year 2000 events and water demand at a f ner temporal reso- production of each of the f ve focus crops by multi- lution may increase irrigation demand above the plying each crop’s water depth by the total area levels projected by our models. Finally, as discussed within each county in 2000 dedicated to the growth above, some farmers will likely change cultivation of that specif c crop (crop area data are from the practices and/or crops in response to changes in cli- Oregon Agricultural Information Network (OAIN)). mate, thus, altering irrigation requirements. We used the tier 1 water-availability model described in Chapter 4 to project average annual 18.4.2 Carbon storage in the Basin’s upland water depth across the landscape for current and forests predicted mid-century climatic conditions. The tier 1 water model requires temperature, precipita- The uplands of the Willamette Basin are blanketed in tion, and land-cover inputs. Again, land cover was forests, mostly conifer. The forests have contributed assumed to change based on the 2050 plan-trend to the region’s timber economy for over a century scenario. Unlike the two other case studies below, and include remnants of old-growth forests that pro- our water-availability projections do not include vide important habitat for a variety of species. Here, the effects of potential climate-driven shifts in we project the potential impacts of climate change on vegetation. carbon storage in the Basin’s upland forests. We used CLIMATE IMPACTS ON ECOSYSTEM SERVICES IN THE WILLAMETTE BASIN OF OREGON 329

Difference (mm)

High: 413

Low: –60

Berry/grape-producing pixels

Figure 18.2 Change in annual average water availability (yield) between present day and mid-century for the Willamette Basin. (See Plate 16.)

Table 18.1 Projected mid-century climate-induced increases in water irrigation volumes required to maintain current yields for f ve different crops in the Willamette Basin

Benton Clackamas Lane Linn Marion Multnomah Polk Washington Yamhill (%) (%) (%) (%) (%) (%) (%) (%) (%)

Raspberries 53.87 58.03 71.82 63.82 49.80 54.40 46.54 48.37 44.76% Blackberries 53.87 58.03 71.82 63.82 49.80 54.40 46.54 48.37 44.76% Strawberries 62.14 67.35 78.68 71.23 57.40 64.37 56.44 57.33 54.67% Blueberries 53.31 57.25 70.88 63.04 49.33 53.72 31.81 47.85 44.35% Wine grapes 67.60 n/a 97.55 n/a 61.38 n/a 56.21 62.70 55.55% Total 58.61 58.82 84.89 65.00 51.92 55.07 50.99 52.30 52.16%

the tier 2 carbon-storage model documented in for 2050 (the development, plan trend, and conserva- C h a p t e r 7 a n d i n N e l s o n et al. ( 2009 ) to estimate the tion scenarios described above). We integrated the carbon stored in both above and below ground bio- climate-driven vegetation projection and the landmass and soil on a given landscape. use change scenarios in two different ways. First, we To investigate the potential effect of climate-in- assumed that climate-driven changes in vegetation duced vegetation changes on carbon storage in the would be the dominant driver of land cover in 2050. Basin’s upland forests, we combined potential natu- This alternative assumes that land owners and man- ral vegetation change data simulated for the Basin agers will allow their land to convert to the forest for 2050 with the three projections of land-use change types as projected by the potential natural vegetation 330 MODELING THE IMPACTS OF CLIMATE CHANGE ON ECOSYSTEM SERVICES change maps. This approach leads to more closed narios (three land-use scenarios and two methods mixed and hardwood forests in the Basin’s uplands. for combining land-use and climate-driven vegeta- We refer to this as the “unmitigated management” tion projections) relative to the mid-century land- scenario. For the second approach, we assumed that use scenario maps that assume no climate-driven forest management actions would be the dominant change in vegetation. driver of forest land cover. In this scenario, timber Before running the carbon-storage model, each of companies, individual land owners, and natural the six climate-change-affected and three baseline resource management agencies would manage for- 2050 land-use maps, projected at a 30-m grid resolu- ests in such a way that the closed conifer forests that tion, were summarized on a grid composed of provide much of the economically valuable timber in 500-ha hexagons. We did this to reduce the number the Basin’s uplands today would be maintained of spatial units that had to be run through the car- regardless of changes in climate. In this scenario, for- bon-storage model. These hexagons merely served est management activities are allowed to mitigate the as spatial units over which the land-cover data were potential effects of climate change on closed canopy summarized. For example, a 500-ha hexagon could conifer forest, however, climate change is still have 200 ha in 0- to 20-year closed conifer forest, 100 assumed to be the driving factor for changes in all ha in 21- and 40-year closed conifer forest, and 300 other land-cover types. We refer to this second sce- ha in mixed conifer forest. nario as the “mitigated management” scenario (see In general, climate change is projected to have a the chapter’s online appendix for details on how negative effect on carbon storage for the Willamette these mitigated scenario maps were created). Basin ( Table 18.3 ; see chapter’s online appendix for The unmitigated and mitigated management calculation details). This trend is consistent across scenarios represent two extreme cases—the actual all six land-cover and management scenarios when contribution of land management to mitigating cli- compared to their relevant 2050 baseline maps mate-driven land-cover change will likely lie some- ( Figure 18.3 ). The 2050 landscapes that consider where between these two extremes. Table 18.2 projected climate-driven changes in vegetation are summarizes the projected vegetation transitions for estimated to store less carbon because the vegeta- the six climate-affected mid-century land-cover sce- tion simulations project a shift from carbon rich,

Table 18.2 Modeled changes in mid-century vegetation cover in the Willamette Basin (in hectares) as a result of hypothesized climatic changes relative to projected mid-century land-cover maps that do not consider climate change

2050 LULC Scenario

Plan trend Development Conservation

Unmitigated Mitigated Unmitigated Mitigated Unmitigated Mitigated

Gain in closed mixed forest 147 420 97 458 147 394 96 029 147 314 124 293 Gain in closed hardwood forest 34 027 25 155 34 068 23 674 34 191 27 776 Loss in closed mixed forest 20 596 14 611 20 569 14 583 20 706 14 717 Loss in closed hardwood forest 1 783 1 783 1 777 1 777 1 784 1 784 Loss in closed conifer forest 159 067 106 218 159 116 103 343 159 014 135 569 Loss in semi-closed hardwood forest 0.18 0.18 0 0 0 0

The projected climate-driven vegetation maps were superimposed on three different potential mid-century land-use scenarios that do not consider climate change (the plan trend, development, and conservation scenarios). The projected vegetation maps were merged with each mid-century land-use scenario map twice. In the f rst merge, known as the unmitigated management approach, the vegetation map always determined land cover when there was a disagreement between the land-use scenario and vegetation maps. In the second merge, the mitigated management approach, closed conifer forest 0–60 years of age was managed for timber and continued to be closed conifer forest despite any projected climate-driven transitions to other forest types. Likewise, 50% of all closed mixed forest was assumed to remain in mixed forest despite any projected climate change due to management actions. Thus, the mitigated mid-century land-cover maps assume that land managers will be able to maintain a desired forest type despite climate change—an assumption that may not be valid in many cases. The unmitigated approach makes no such assumption. All changes are relative to the relevant baseline mid-century land-use scenario map. CLIMATE IMPACTS ON ECOSYSTEM SERVICES IN THE WILLAMETTE BASIN OF OREGON 331

Change in Mg per 500 Ha Hexagon –90,000 and less

–70,000 to –80,000

–50,000 to –60,000 –30,000 to –40,000

–10,000 to –20,000 0 1 to 10,000 Development 2050; Conservation 2050; Mitigated Unmitigated

Figure 18.3 Difference in stored biomass and soil carbon in the Willamette Basin on two mid-century land-cover maps that incorporate projected climate-induced vegetation shifts relative to projected mid-century land-cover maps that do not consider climate change. The climate-change-affected development scenario map assumes mitigated land-cover change. The climate-change-affected conservation scenario map assumes unmitigated land-cover change (see description in legend for Table 18.2 for a brief description of the mitigated and unmitigated scenarios). When compared to their baselines, the development scenario map with mitigated change represents the lowest landscape-level loss in carbon storage of all the solutions presented in Table 18.3 (the 5th percentile solution of 5.6% (57.4 Tg)), whereas the conservation scenario with unmitigated change represents the greatest landscape-level loss in carbon storage of all the solutions presented in Table 18.3 (the 95th percentile solution of 7.0% (78.2 Tg)). Each spatial unit on the map is a 500-hectare hexagon.

older closed conifer forests, to relatively carbon the corresponding “mitigated” management sce- poor, younger closed mixed forests and closed narios. The mitigated scenarios prevent larger losses hardwood forests ( Smith et al. 2006 ) ( Table 18.3 ). of carbon because they restrict climate-driven Despite the slightly larger negative effect of cli- changes in much of the established closed conifer mate change on carbon storage under the conserva- and mixed conifer-hardwood forests from being tion land-use scenario (Table 18.3 ), more carbon manifested. was stored under this scenario than under either Our results highlight potential negative and posi- the plan-trend or the development scenario. Given tive feedback loops in the climate system. As atmos- climate change, the conservation land-use scenario pheric greenhouse-gas concentrations rise, the resulted in the median storage of 1040.5 Tg of car- climate will change, driving shifts in vegetation. In bon compared to 928.0 Tg stored under the plan- some areas, vegetation may respond by sequestering trend scenario and 944.5 Tg stored under the more carbon and thus will help to mitigate emis- development scenario. We found that forest man- sions. In other areas, such as the Willamette Basin, agement strategies may also be able to mitigate vegetation responses to warming may result in some of the effects of climate change with respect to reduced capacity for carbon storage and sequestra- carbon storage. The “unmitigated” management tion, thus amplifying the anthropogenic inf uence on scenarios resulted in less carbon storage than did climate. The models described in this book, when 332 MODELING THE IMPACTS OF CLIMATE CHANGE ON ECOSYSTEM SERVICES

Table 18.3 Percentage loss (and absolute loss in Tg) in biomass and soil carbon storage in the Willamette Basin due to hypothesized climatic changes relative to projected mid-century land-cover maps that do not consider climate change

Management approach 2050 LULC scenario

Plan trend Development Conservation

Median Median Median [5th–95th Percentile] [5th–95th Percentile] [5th–95th Percentile]

Unmitigated 6.7% (66.2) 6.6% (66.6) 7.0% (77.8) [6.6 (65.7)–6.7% [6.5 (66.3)–6.6% [6.9 (77.5)–7.0% (66.4)] (67.0)] (78.2)] Mitigated 6.0% (59.9) 5.7% (57.5) 6.7% (75.6) [6.0 (59.6)–6.1% [5.6 (57.4)–5.7% [6.7 (75.0)–6.8% (60.3)] (58.0)] (75.8)]

See the legend for Table 18.2 for a brief description of the various mid-century land-use scenario maps considered in this table. There is variability in results because we randomly simulated forest-age structure (this information is not provided by the mid-century land-cover maps) before calculating carbon-storage values. We repeated this process 1000 times. See the chapter’s online appendix for simulation details. All losses are relative to the relevant baseline mid- century land-use use scenario map.

linked with ecosystem or vegetation models, provide assess biodiversity value in diverse, human-domi- a useful tool for investigating where these tradeoffs nated landscapes. may occur and where management may be useful if To model climate-driven changes in species dis- increased carbon storage and sequestration is a goal. tributions in the Willamette Basin, we used projected range shifts developed for a hemispheric assessment of species’ geographic responses to cli- 18.4.3 Terrestrial vertebrate diversity in the mate change ( Lawler et al. 2009 ). These range shifts Basin are based on the same climate-change simulation The Willamette Basin hosts a rich biota that inhabits used in the agriculture production and carbon-stor- diverse environments including oak savannah, age case studies above. Although the previous two mixed hardwood and conifer forests, remnant ripar- case studies used the climate and vegetation projec- ian gallery forests, and young and mature temper- tions at their native resolutions, (e.g., the 30-second ate conifer rainforest. Here, we explore some of the grid), species range-shift projections were gener- potential effects of climate change on terrestrial ver- ated at a 50-km resolution. These projections were tebrate diversity in the Willamette Basin. Specif cally, made for the average simulated climatic conditions we investigate the effects of hypothesized clim- for a period 2041–70. Our analyses use a set of bird, driven shifts in species distributions and vegetation mammal, and amphibian species for which we have change (i.e., habitat availability) on two measures of data on current and projected future geographic biodiversity status described in detail in Chapter 13 , ranges as well as general habitat associations. This marginal biodiversity value (MBV) and countryside set includes 187 species (136 bird, 47 mammal, and species area relationship (SAR). In short, MBV is a 4 amphibian species). The modeled range shifts parcel-level measure of biodiversity based on both resulted in a potential loss of 8 of these species from the number of species with geographic range in a the Basin by the middle of the century. At the same parcel and the proportion of the parcel that likely time, the climate model projects potential range serves as habitat for each of the species. SAR is a expansions that could result in 17 new species mov- landscape-level estimate of biodiversity value based ing into the Basin by the middle of the century, on multi-habitat species area relationships (Pereira including 3 mammals and 14 birds (see Table 18.4 and Daily 2006 ). The measure is referred to as a for a complete list of species with potential ranges “countryside” metric because it was designed to that were projected to move in or out of the Basin). CLIMATE IMPACTS ON ECOSYSTEM SERVICES IN THE WILLAMETTE BASIN OF OREGON 333

As in the carbon-storage case study described study, land cover on all maps was summarized by above, we imposed projected climate-driven 500-ha hexagon. Finally, unlike the carbon-storage changes in vegetation (i.e., habitat availability) on case, we included the whole Basin in the biodiver- the three 2050 land-use projections (Hulse et al. sity analysis, not just its upland forest areas. 2002 ) assuming two different land management Using these various scenarios, we generated sev- responses (the unmitigated management and eral alternative landscapes across which we evalu- mitigated management schemes described in the ated vertebrate diversity. First, we combined the carbon-storage example), thereby generating six 2000 land-cover map with the maps of current spe- 2050 maps affected to some degree or another by cies distributions. This provided a map of current climate change. We also considered the three 2050 vertebrate biodiversity. Second, we combined each land-use projections that assume no climate-change of the nine 2050 land-cover maps, including the effects (the baseline 2050 land covers). We also use a three baseline maps, with maps of projected species year 2000 land-cover map of the Basin for additional distributions. This provided projections of how cli- comparison purposes (this year 2000 map does not mate change and land-use change would likely assume any climate-driven change in habitat avail- affect vertebrate diversity values when both range ability). Finally, as we did in the carbon-storage case shifts and climate-driven changes in habitat are taken into account and when range shifts alone are Table 18.4 Species with geographic ranges projected to shift out of or taken into account. Finally, we combined each of the into the Willamette Basin by mid-century nine 2050 land-cover maps, including the three baseline maps, with maps of current species distri- Emigrants Immigrants butions (i.e., the 2050 land cover assuming no Gulo gulo (wolverine) Myodes gapperi (Southern red-backed climate-driven shifts in species ranges). This last vole) combination provided an estimate of how both Anas acuta (Northern Pintail) Onychomys leucogaster (Northern land-use and climate-driven changes in habitat grasshopper mouse) would affect vertebrate diversity when range shifts Histrionicus histrionicus Ovis canadensis (Bighorn sheep) were not taken into account. (Harlequin Duck) Bucephala islandica (Barrow’s Aeronautes saxatalis (White-throated For all projected vertebrate diversity maps, MBV Goldeneye) swift) values were calculated for each 500-ha hexagon and Picoides dorsalis (Three-toed Amphispiza bilineata (Black-throated the SAR score was calculated for the entire land- woodpecker) sparrow) scape. As mentioned earlier, MBV measures the pro- Vireo olivaceus (Red-Eyed Carpodacus cassinii (Cassin’s Finch) portion of the modeled biodiversity supported by a Vireo) hexagon (or, conversely, the proportion of the land- Ammodramus savannarum Tringa semipalmata (Willet) scape’s total biodiversity that would be lost if the (Grasshopper Sparrow) hexagon were suddenly unable to support any spe- Xanthocephalus xanthocephalus Coccyzus americanus (Yellow-billed cies; see this chapter’s online appendix for species- (Yellow-Headed Blackbird) Cuckoo) habitat compatibility scores for all modeled species). Falco columbarius (Merlin) One of the limitations of the MBV index is that it Phalaenoptilus nuttallii (Common Poorwill) cannot be used to indicate the overall status of ter- Polioptila caerulea (Blue-gray restrial biodiversity on the landscape (the summa- Gnatcatcher) tion of MBV values across all hexagons always Recurvirostra americana (American equals one and does not allow for direct compari- Avocet) son of collective biodiversity status across scenar- Sphyrapicus nuchalis (Red-naped ios). Instead, an MBV map indicates the relative Sapsucker) provision of biodiversity across a landscape at some Tyrannus tyrannus (Eastern Kingbird) point in time. In contrast, the SAR scores allow for Anas crecca (Green-winged Teal) comparison across different landscapes, including Aythya collaris (Ring-necked Duck) the same landscape at different points in time. The Melanerpes lewis (Lewis’ Woodpecker) SAR score for a species indicates the proportion of 334 MODELING THE IMPACTS OF CLIMATE CHANGE ON ECOSYSTEM SERVICES its range space that contains compatible habitat and middle of the three 2050 maps are more sub- raised to a power that is between 0 and 1 (the z stantial than the differences between the middle score). A low z score is indicative of a species that and rightmost maps. only needs to f nd compatible habitat in a small por- Climate change had a negative effect on overall tion of its range space to be relatively secure; biodiversity value as assessed by the SAR score increases beyond this threshold improve the condi- (Table 18.5 ). Under the climate-change scenario, tion for that species, but not dramatically. In con- SAR scores decreased from the year 2000 to 2050 by trast, the status of a species with a high z score is 5.18 to 7.68% depending on the land-use scenario, more affected by any changes in habitat in its range the z score, and whether climate-driven range shifts space. The landscape’s SAR is the (weighted) aver- only or climate-driven range shifts and climate- age score for all species on the landscape. We calcu- driven vegetation changes were considered. In con- lated landscape SAR values in this case study trast, in the absence of climate change, SAR scores assuming equal species weights and that each spe- decreased between 0.23 and 1.19% under the devel- cies had the same z score. Because we do not have opment scenario and increased between 0.12 and information on the species’ z scores, we calculated 0.48% under the conservation scenario. The pattern each landscape’s SAR score four times, exploring a of land-cover change under the conservation sce- range of different z scores. nario mitigated some of the deleterious effects of In general, simulated climate-driven changes in climate-driven changes in species ranges on SAR species ranges and habitat availability resulted in scores. However, this mitigating effect was rela- notable changes in the spatial distribution of hexa- tively small. For example, with a z score of 0.64, gon MBVs from 2000 to 2050 under both the devel- climate-driven changes in species ranges resulted in opment and conservation land-use scenarios a decrease in the SAR score of 6.18% under the con- ( Figure 18.4 ; we do not include the 2050 plan-trend servation scenario, which was only slightly lower maps in Figure 18.4 because these maps tend to lie than the 7.25% decrease under the development between the extremes provided by the conservation scenario. When both climate-driven range shifts and development scenarios). Climate impacts on and climate-driven changes in vegetation were con- MBV values appeared to be greatest along the crest sidered, the conservation scenario did little to miti- of the Cascade Range and across portions of the gate the effects of climate change, particularly at basin f oor. The distribution of MBVs on the maps higher z scores. Although these results indicate that that included climate-change effects, range shifts general conservation strategies—such as those only, or both climate-driven range and habitat avail- embodied in our conservation scenario—may be ability changes under “unmitigated” management able to mitigate some of the potential effects of cli- (the last two 2050 maps for both scenarios in mate change on biodiversity, they ultimately sug- Figure 18.4 ), are signif cantly different than the dis- gest that conservation planning will need to directly tributions on the 2050 baseline maps (the f rst 2050 address climate change with specif c adaptation map for both scenarios in Figure 18.4 ) in several strategies for offsetting climate-driven habitat respects. First, climate change is projected to result changes and range shifts. in higher MBV values at higher elevations in the Caution is needed when interpreting our analy- Cascade Range (the darker areas on the far right ses. First, it is not clear what a 5–8% decrease in side of the maps) and some northern portions of the landscape-level SAR scores means to the actual con- Coast Range. Second, climate change is projected to servation status of the modeled species. Do these result in lower MBV values in the northern foothills apparently modest decreases in SAR scores suggest of the Cascade Range and across portions of the that a species or two might go locally extinct? If so, Basin f oor. These changes are a result of projected the loss of even a single species from a region may shifts in species distributions to higher elevations. have cascading ecological effects—a nuance not In general, climate-driven changes in species ranges captured in our analyses. Second, although the veg- had a greater effect on the change in the spatial dis- etation models capture some of the basic changes in tribution of MBV values than did climate-driven habitat availability, they likely fail to adequately changes in habitat—the differences in the leftmost describe changes in the habitats of many species DISCUSSION AND CONCLUSIONS 335

No Climate Change Climate Change Climate Change Causes Related Effects Causes Shift Shift in Range on Landscape in Ranges Only and Vegetation Development 2050 Conservation 2050

2000

No MBV Increasing MBV

2050

Figure 18.4 Marginal biodiversity values (MBVs) for the Willamette Basin for present day and mid-century. The six mid-century maps represent two different land-use scenarios (a development scenario and a conservation scenario) combined with three different levels of climate impact (none, climate- driven range shifts, and climate-driven range shifts and changes in habitat, i.e., vegetation). The f rst of the three mid-century maps assumes no climate change (species’ range space does not change between present day and mid-century and no climate-driven vegetation changes occur). The second of the three mid-century maps assumes climate-change-induced species’ ranges but no climate-driven vegetation changes occur. The last of the three mid-century maps assumes climate-driven shifts in species’ ranges and climate-driven vegetation changes occur (in this case, we assume that vegetation transitions are not mitigated—see description in legend for Table 18.2 for a brief description of the mitigated and unmitigated scenarios). Given the coarse resolution of the range-shift projections, the effects of range changes are revealed by changes in the blocky patterns in the maps, whereas changes in habitat are much more f nely resolved. White areas on the maps do not include any usable habitat. that have critical relationships with specif c plant in the Willamette Basin and where the climate species. In general, the biodiversity models pre- impacts are likely to be the greatest. sented in this book do not account for many of the factors that have been shown to affect species status 18.5 Discussion and conclusions on the landscape. Finally, the range-shift projections used in our analyses are based solely on climatic Projected future climatic changes are likely to have conditions. Shifts in species ranges will further be signif cant effects on the functioning of ecological modif ed by changes in the distribution of preda- systems. These effects will in turn alter the degree to tors, prey, competitors, and forage species. These which natural and managed systems can provide limitations have likely led us to underestimate the many of the ecosystem services on which we cur- potential impacts of climate change on biodiversity rently depend. Here, we use several ecosystem in the Basin. Nonetheless, the models presented service evaluation models and a biodiversity model here give us a f rst approximation of how climate to assess the potential impacts of climate change on change might affect terrestrial vertebrate diversity irrigation demand, carbon storage, and biodiver- 336 MODELING THE IMPACTS OF CLIMATE CHANGE ON ECOSYSTEM SERVICES

Table 18.5 Landscape-level SAR scores in the Willamette Basin for the collection of modeled terrestrial vertebrates in the present day and mid-century with and without hypothesized climatic changes

0.11 0.25 0.64 1

2000 a 0.855 0.737 0.533 0.421 2050 Development LULC Scenario No climate-change effects a 0.853 0.734 0.529 0.416 [ −0.23%] [−0.41%] [−0.75%] [−1.19%] Climate change causes shifts in range only 0.790 0.680 0.494 0.393 (% change from 2000) b [ −7.57%] [−7.68%] [−7.25%] [−6.56%] Climate change causes shifts in range and vegetation 0.790 0.681 0.495 0.394 (% change from 2000)b [ −7.57%] [−7.67%] [−7.17%] [−6.38%]

Conservation LULC Scenario No climate-change effects a 0.856 0.739 0.535 0.423 [0.12%] [0.27%] [0.38%] [0.48%] Climate change causes shifts in range only 0.795 0.685 0.500 0.399 (% change from 2000) b [ −7.02%] [−7.00%] [−6.18%] [−5.18%] Climate change causes shifts in range and vegetation 0.792 0.682 0.493 0.390 (% change from 2000)b [ −7.39%] [−7.46%] [−7.48%] [−7.38%]

a Species with geographic ranges overlapping the Willamette Basin in the present day are included. b All species with geographic ranges overlapping the Basin in the present day and/or in the middle of the century are included. The species with range in the present day only contribute a SAR score of 0 to the mid-century SAR score. sity. The three case studies provided here are meant models for assessing potential climate-change to serve as examples of how the models described impacts, rather than as providing quantitative esti- in this book can be applied to assess climate impacts. mates of how climate change is likely to affect the Our assessments were driven by a single future cli- Willamette Basin. mate simulation based on the SRES mid-high A2 To provide quantitative projections capable of emissions scenario. Thus, our results should be seen being used by decision makers for long-term as relatively conservative, and as a limited estimate planning purposes, the analyses presented here of projected future climate-change impacts on eco- should be modif ed in three important ways. First, it system services. Our results are also likely to be is important to quantify the extent to which combin- conservative for two additional reasons. First, they ing the projections of multiple models can result in a do not consider the potentially multiplicative effects compounding of model uncertainty. For example, of climate change on a given sector or service there are uncertainties in the AOGCM projections, through the inf uence of multiple climate drivers additional uncertainties in the dynamic global veg- and interactions between services and sectors. And etation model and climate-envelope model projec- second, only mean temperature and precipitation tions, and additional uncertainties inherent in the changes were used to estimate impacts. Analyses ecosystem-service models themselves. All of these show that many climate-change impacts, including uncertainties need to be quantif ed to capture the ecosystem services, are likely to be just as (if not full range of potential climate-change impacts on a more) sensitive to changes in climate thresholds given ecosystem service. A second limitation is that and extremes as they are to shifts in monthly, sea- we used one future climate simulation from a single sonal, or annual means. For all of these reasons, AOGCM in our analyses and a single emissions sce- these results should be viewed as a demonstration nario. Given the wide variability in projections from of the potential application of the ecosystem-service different AOGCMs and the broad range of potential DISCUSSION AND CONCLUSIONS 337 future emissions, it will be necessary to investigate comments and suggestions. We are also grateful to the effects of alternative climate-change projections Peter Kareiva for his insightful editing and to Evan to begin to understand the range of possible climate Girvetz for useful discussions. S. Shafer was sup- impacts on ecosystem services. And lastly, capturing ported by the US Geological Survey’s Earth Surface the potential impacts of climate change will require Dynamics Program. evaluating the effects of realistic adaptation strategies. Our analysis of irrigation demand did not take into account that farmers will likely shift crops or References modify irrigation methods in response to climate Alcamo, J., and Henrichs, T. (2002). Critical regions: A change. Nor did we account for the multitude of model-based estimation of world water resources sensi- ways in which forest managers might respond to tive to global changes. Aquatic Sciences—Research across climate-driven changes in tree-species composition. Boundaries, 64 , 352–62. Despite these limitations, the case studies investi- Alcamo, J., Vuuren, D. v., Ringler, C., et al . (2005). Changes gated here demonstrate the importance of making in nature’s balance sheet: model-based estimates of even preliminary estimates of potential climate future worldwide ecosystem services. Ecology and impacts on ecosystem services. For example, one Society , 10 , Art. 19. can use the projected climate-induced change in Alley, R., Berntsen, T., Bindoff, N. L., et al . (2007). Climate each ecosystem service as a metric of “impact.” The change 2007: the physical science basis, summary for policy- land-use scenario that yields the lowest impact score makers . Working Group I contribution to the Intergovernmental Panel on Climate Change, Fourth is the land-use scenario that provides the greatest Assessment Report, Geneva. potential for adaptation to climate change. In our Arnell, N. W. (2004). Climate change and global water assessments, this was generally the conservation resources: SRES emissions and socio-economic scenar- scenario. Although, it is important to note that care ios. Global Environmental Change , 14 , 31–52. must be taken in assessing the results of such sce- Bachelet, D., Neilson, R. P., Lenihan, J. M., et al . ( 2 0 0 1 ) . nario comparisons. For example, although we found Climate change effects on vegetation distribution and a greater climate-driven loss in carbon storage under carbon budget in the United States. Ecosystems , 4 , the conservation scenario, this differential resulted 164–85. from the fact that the conservation scenario gener- Battin, J., Wiley, M. W., Ruckelshaus, M. H., et al . (2007). ated more carbon storage to begin with and thus Projected impacts of climate change on salmon habitat there was more to lose by the middle of the century. restoration. Proceedings of the National Academy of Sciences of the USA , 104 , 6720–5. More generally, given that climate change is expected Battisti, D. S., and Naylor, R. L. (2009). Historical warnings to have profound effects on both human and natural of future food insecurity with unprecedented seasonal systems, it will be critical to consider estimates of heat. Science, 323 , 240–4. potential climate impacts when forecasting ecosys- Betts, R. A., Boucher, O., Collins, M., et al . (2007). tem services. As evidenced by the results of our sim- Projected increase in continental runoff due to plant ple case studies, climate change will alter many of responses to increasing carbon dioxide. Nature , 448 , the ecosystem services we rely on today. Both miti- 1037. gating and adapting to climate impacts will require Bouwman, A. F., Kram, T., and Goldewijk, K. K. (2006). an understanding of how ecosystems and ecosystem Integrated modelling of global environmental change. An services will respond to climate change. Models overview of IMAGE 2.4 . Netherlands Environmental designed to assess ecosystem services will likely Assessment Agency (MPN), Bilthoven. Carroll, C. (2007). Interacting effects of climate change, land- prove indispensable in our struggle to address cli- scape conversion, and harvest on carnivore populations mate change in the coming decades. at the range margin: marten and lynx in the Northern Appalachians. Conservation Biology, 21 , 1092–104. Acknowledgments Cramer, W., Bondeau, A., Woodward, F. I., et al . (2001). Global response of terrestrial ecosystem structure and function to We thank Michelle Marvier, Katherine Hayhoe, CO 2 and climate change: results from six dynamic global Nathan Schumaker, and David Turner for useful vegetation models. Global Change Biology , 7 , 357–73. 338 MODELING THE IMPACTS OF CLIMATE CHANGE ON ECOSYSTEM SERVICES

Deschenes, O., and Greenstone, M. (2007). The economic Pearson, R. G., Dawson, T. P., Berry, P. M., et al . ( 2 0 0 2 ) . impacts of climate change: evidence from agricultural SPECIES: A Spatial Evaluation of Climate Impact on output and random f uctuations in weather. American the Envelope of Species. Ecological Modelling, 154 , Economic Review , 97 , 354–85. 289–300. de Wit, M., and Stankiewicz, J. (2006). Changes in surface Pereira, H. M., and Daily, G. C. (2006). Modeling biodiver- water supply across Africa with predicted climate sity dynamics in countryside landscapes. Ecology , 87 , change. Science , 311 , 1917–21. 1877–85. FAO. (1992). CROPWAT, a computer program for irrigation Polasky, S., Nelson, E., Lonsdorf, E., et al . (2005). Conserving planning and management by M. Smith. FAO Irrigation species in a working landscape: land use with biological and Drainage Paper 26, Rome. and economic objectives. Ecological Applications , 15 , Gordon, C., Cooper, C., Senior, C. A., et al. (2000). The sim- 1387–401. ulation of SST, sea ice extents and ocean heat transports Polasky, S., Nelson, E., Camm, J., et al . (2008). Where to put in a version of the Hadley Centre coupled model with- things? Spatial land management to sustain biodiver- out f ux adjustments. Climate Dynamics , 16 , 147–68. sity and economic returns. Biological Conservation , 141 , Hulse, D., Gregory, S., and Baker, J. (2002). Willamette River 1505–24. Basin planning atlas: trajectories of environmental and eco- Ramankutty, N., Foley, J. A., Norman, J., et al. (2002). The logical change . Oregon State University, Corvallis. global distribution of cultivable lands: current patterns Lawler, J. J., Shafer, S. L., White, D., et al. (2009). Projected and sensitivity to possible climate change. Global Ecology climate-induced faunal change in the western hemi- and Biogeography , 11 , 377–92. sphere. Ecology , 90 , 588–97. Root, T. L., Price, J. T., Hall, K. R., et al. (2003). Fingerprints Lobell, D. B., and Asner, G. P. (2003). Climate and manage- of global warming on wild animals and plants. Nature , ment contributions to recent trends in U.S. agricultural 421 , 57–60. yields. Science , 299 , 1032. Rosenzweig, C., and Parry, M. L. (1994). Potential impact Lobell, D. B., and Field, C. B. (2007). Global scale climate– of climate change on world food supply. Nature , 367 , crop yield relationships and the impacts of recent warm- 133–8. ing. Environmental Research Letters, 2 , 014002. Schlenker, W., Hanemann, W. M., and Fisher, A. C. (2005). Mendelsohn, R., Nordhaus, W. D., and Shaw, D. (1994). Will U.S. agriculture really benef t from global warm- The impact of global warming on agriculture: A Ricardian ing? Accounting for irrigation in the hedonic approach. analysis. American Economic Review, 84 , 7 5 3 – 7 1 . American Economic Review, 95 , 395–406. Milly, P. C. D., Dunne, K. A., and Vecchia, A. V. (2005). Schumaker, N. H., Ernst, T., White, D., et al . (2004). Global pattern of trends in streamf ow and water avail- Projecting wildlife responses to alternative future land- ability in a changing climate. Nature , 438 , 347–50. scapes in Oregon’s Willamette Basin. Ecological Mitchell, T. D., and Jones, P. D. (2005). An improved Applications , 14 , 381–400. method of constructing a database of monthly climate Sitch, S., Smith, B., Prentice, I. C., et al. (2003). Evaluation observations and associated high-resolution grids. of ecosystem dynamics, plant geography and terrestrial International Journal of Climatology , 25 , 693–712. carbon cycling in the LPJ dynamic global vegetation Nakicenovic, N., Alcamo, J., Davis, G., et al. (2000). Special model. Global Change Biology , 9 , 161–85. report on emissions scenarios: a special report of Working Smith, J. E., Heath, L. S., and Skog, K. E. (2006). Methods for Group III of the Intergovernmental Panel on Climate Change. calculating forest ecosystem and harvested carbon with stand- Cambridge University Press, Cambridge. ard estimates for forest types of the United States. Gen Tech Rep Nelson, E., Mendoza, G., Regetz, J., et al . (2009). Modeling NE-343. US Department of Agriculture, Forest Service, multiple ecosystem services, biodiversity conservation, Northeastern Research Station, Newtown Square. commodity production, and tradeoffs at landscape Stewart, I. T., Cayan, D. R., and Dettinger, M. D. (2005). scales. F rontiers in Ecology and the Environment, 7 , 4–11. Changes toward earlier streamf ow timing across west- New, M., Lister, D., Hulme, M., et al. (2002). A high-resolu- ern North America. Journal of Climate , 18 , 1136–55. tion data set of surface climate over global land areas. USDA. (2008). Soil Survey Geographic (SSURGO) Database Climate Research , 21 , 1–25. for Willamette Valley, Oregon. Parmesan, C., and Yohe, G. (2003). A globally coherent f n- Vörösmarty, C. J., Green, P., Salisbury, J., et al. (2000). Global gerprint of climate change impacts across natural sys- water resources: vulnerability from climate change and tems. Nature , 421 , 37–42. population growth. Science , 289 , 284–8. Parmesan, C. (2006). Ecological and evolutionary responses Walther, G.-R., Post, E., Convey, P., et al. (2002). Ecological to recent climate change. Annual Review of Ecology and responses to recent climate change. Nature , 416 , Systematics , 37 , 637–69. 389–95.

CHAPTER 19 Incorporating ecosystem services in decisions

Emily McKenzie, Frances Irwin, Janet Ranganathan, Craig Hanson, Carolyn Kousky, Karen Bennett, Susan Ruffo, Marc Conte, James Salzman, and Jouni Paavola

19.1 Introduction 19.1.1 Role of ecosystem service information in decisions The world now faces unprecedented and interconnected challenges – reducing poverty, addressing cli- Scientif c models move us from abstract, conceptual mate change, and halting widespread environmental arguments about the importance of ecosystem serv- degradation. Scientif c information on the links ices to specif c quantif cation of the level, value and between humans and nature can help solve these spatial distribution of ecosystem service benef ts. problems ( Lubchenco 1998 ; Bingham et al . 1995 ; This is of great relevance to the real world because Pielke 2007; MA 2005a). However, to seize this oppor- such information can be applied to a range of poli- tunity, scientif c knowledge must be translated into cies and decisions, as in the examples from Oregon action (NRC 2004; Daily et al . 2 0 0 9 ; C a s h et al . 2003 ). and Hawaii in Chapter 14 (see also Turner et al . Advances in scientif c modeling such as those 2000 ; Naidoo and Ricketts 2006; Nelson et al . 2009 ). described in this book are necessary but not suff - A modeling framework that captures impacts on cient. Here we provide an explicit examination of the multiple ecosystem services over alternative sce- channels through which the science of mapping and narios enables stakeholders to weigh tradeoffs can valuing ecosystem services can improve decisions. serve as a basis for negotiation (Ghazoul 2007 ). Opportunities to use information on ecosystem Without such information, decision-makers tend to services occur in all sectors of the economy and at use intuitive or heuristic approaches that ignore all levels of decision-making: a mayor aims to ecosystem service values and distributional issues increase f ood protection for a city’s citizens; a busi- ( Kiker et al . 2005 ). As noted by Ascher and Healy ness requires a reliable supply of water for its man- ( 1990 ), “information shapes many aspects of how ufacturing process; an international development resource issues are viewed and addressed: the focus agency attempts to reduce poverty by increasing of attention, the way problems come to be def ned, small farm production. Although not always termed and the ways that success or failure . . . is attributed “ecosystem services,” in each example, considera- to a project or policy.” tion of the benef ts we gain from the environment is There are several characteristics of ecosystem essential for making wise decisions. Information on service models that are relevant to common policy ecosystem services can tell us how and which serv- contexts (Table 19.1 ). For example, the development ices are relevant to our goals, whether important of stakeholder-driven scenarios can ensure that eco- services are at risk, where services are provided, system service valuations are aligned with the prob- who is affected, and the trade-offs of different lems of interest to decision-makers, revealing choices; all key pieces of information for the design sources of conf ict and building consensus (Henrichs and implementation of a broad set of policy et al . 2008 ). This does not happen automatically, mechanisms. however, and typically requires active, iterative and

339 340 INCORPORATING ECOSYSTEM SERVICES IN DECISIONS

Table 19.1 Characteristics of ecosystem service models with relevance for decision-making

Model characteristics Relevance to decisions

Integrated framework, including multiple ecosystem services • Provides consistent standard for evaluating projects and policies • Encourages consideration of all ecosystem services including those that have not been emphasized in past policies • Enables analysis of trade-offs between different ecosystem services, stakeholders, and geographic areas • Informs and encourages coordinated, multi-sectoral, ecosystem-based management Packaged as a “tool” • Translates complex scientif c results into policy relevant outputs, e.g., trade-off curves, eff ciency frontiers (see Chapters 3 and 14) • Facilitates capacity building in ecosystem service modeling and valuation • Increases uptake due to lower costs and reduced diff culty of doing ecosystem service studies Different levels of model complexity • Tailors modeling to the level of certainty suff cient for different policy contexts (see Chapter 15 ) Produces information on biophysical changes • Enables use of biophysical ecosystem service units when they suff ce for specif c decision contexts • Can feed into decision-support tools such as multi-criteria analysis Produces information on economic values • Emphasizes connections between environmental sustainability and economic development • Provides common monetary metric, facilitating comparison of policy alternatives • Can feed into cost–benef t analysis Scalable • Enables selection of spatial scale most relevant to decision context and level of governance Scenario based • Provides structured way to consider implications of possible futures and policy alternatives • Tailors analysis to priority policy questions • Explores uncertain aspects of the future • Challenges assumptions about how systems operate Stakeholder driven • Encourages uptake and use of results • Enables local knowledge and policy priorities to shape the analysis Spatially explicit • Visual appeal for communication and advocacy • Determines where to target investments, policies, and payments • Determines locations where ecosystem services and biodiversity overlap • High-resolution maps enable policy responses to be targeted and context specif c Produces information on opportunity costs • Determines the lower bound for payments for ecosystem services, and other incentive schemes Possible to disaggregate to individual services • Enables users to focus on individual services targeted by specif c policies, such as payments for water yield Enables analysis of distribution of ecosystem service costs • Helps identify who bears benef ts and costs of different policies, thereby identifying and benef ts possible locations for payments for ecosystem services • Helps clarify how environmental policies affect social goals and priorities, such as poverty reduction inclusive communication between scientif c experts between ecosystems and human well-being and stakeholders ( Lubchenco 1998). ( Leopold 1949 ; Daly 1968 ; Wilson 1998 ; MA 2005c). 19.2 Putting ecosystem services When government ministries initiate programs or on the agenda companies launch new ventures, they rarely consider how ecosystem services will affect their suc- People continue to degrade our environment in part cess, or the costs of replacing degraded services because we simply do not appreciate the links ( Ranganathan et al . 2008 ). PUTTING ECOSYSTEM SERVICES ON THE AGENDA 341

Information on ecosystem services and their val- 19.2.1 Building better development policies ues can also inf uence decisions over major devel- The Millennium Development Goals (MDGs) pro- opment projects. Ecosystem services that lack vide an example where the importance of ecosystem market prices are often not considered in project services to achieve poverty alleviation was not ini- evaluations, enabling other interests to determine tially (but is now increasingly) recognized. In 2001, decisions ( Balmford et al . 2002 ). Economic valuation leaders adopted eight goals aimed at cutting global of non-marketed ecosystem services makes their poverty by 50% by 2015. Although the MDGs origi- values clear in monetary terms, enabling compari- nally included an environmental sustainability goal, son of all costs and benef ts of proposed projects. it was narrowly def ned, without consideration of This can help create political consensus around the role of ecosystem services in achieving other decisions that sustain ecosystem services. For exam- goals ( WRI 2005 ). Goals to increase incomes and ple, in Borneo a rapid assessment of the economic reduce hunger did not account for how water, fertile value of ecosystem services provided by standing soil and pollinators support agriculture. When coun- forests inf uenced conservation policy decisions tries made plans to achieve the MDGs, the links over proposed oil palm plantations. The study between development and nature were not assessed the ecosystem service benef ts of carbon addressed. Later analysis on the ecosystem service storage, the avoidance of health costs from forest benef ts of conservation found that signif cant envi- f res, and the benef ts of forest-agriculture mosaics. ronmental investment is required to achieve poverty The information appears to have played a role in reduction (Pearce 2005 ). Focusing targets on ecosys- the policy decision, with the government represent- tem services is now increasingly recognized as criti- ative declaring the oil palm development would cal to achieve all eight MDGs ( WRI 2005 ). not go forward because Borneo “is a resource of life Some countries now explicitly consider ecosys- for Kalimantan” ( Naidoo et al . 2009 ). tem services in national poverty reduction strate- It is possible to integrate ecosystem services into gies. Tanzania’s 2005 National Strategy for Growth existing decision-support and environmental and Reduction of Poverty exemplif es this shift. It assessment tools (Le Quesne and McNally 2004 ). If includes 15 environmental targets to protect and mandated and enforced, these tools can ensure sys- enhance ecosystem services ( Assey et al . 2007 ). tematic, transparent evaluation of ecosystem service Distributional and social information is particularly impacts. Examples include strategic environmental important for policy formulation in developing assessments (SEA) and regulatory impact assess- countries such as Tanzania; investments that ments for evaluating policies and legislation, typi- increase the supply of services and improve the cally at the level of an economic sector or region, welfare of certain groups of people may reduce and cost–benef t analyses (CBA) and environmental services that support the livelihoods and well-being impact assessments for individual projects. of others living in poverty (see Chapter 16 ). In Although historically restricted to human health Kenya, these types of trade-offs were explored by impacts of pollution, many CBAs and SEAs now overlaying poverty maps—showing where poor consider a wider range of ecosystem services (DAC people live and aspects of their well-being—with Network on Environment and Development maps of ecosystems and their services. This identi- Co-operation (ENVIRONET) 2008). For example, a f ed where poverty indicators overlap with service recent SEA for a district planning process in Rwanda supply and demand areas ( Snel 2004 ; WRI 2007 ). linked ecosystem service degradation to food, For example, in the upper Tana region around water, and fuel scarcity ( UNDP 2007 ). Mt. Kenya, a large number of poor communities rely directly and exclusively on ecosystems to pro- 19.2.2 Building better business strategies vide and f lter drinking water. This type of information can be used to make sure that polices that Businesses can use information on ecosystem serv- impact ecosystems do not exacerbate the poverty of ices to shape their investments and strategies. Just vulnerable communities. as governments can f nd that ecosystem services are 342 INCORPORATING ECOSYSTEM SERVICES IN DECISIONS linked to the wellbeing of citizens, businesses can pollute ecosystems with value to communities, discover that ecosystem services contribute to prof- leading to lawsuits and f nes, and challenging their its. Although businesses can adversely impact eco- license to operate. Conversely, businesses that sus- systems through consumption of natural resources, tainably manage land or water resources may pollution, and land conversion, businesses also increase eff ciency, differentiate their brand, reduce depend on ecosystems. Agribusiness, for example, costs, and even generate new sources of revenue depends on pollination, and control of pests and through markets for ecosystem services. Although erosion (see Chapter 10 ). Approximately 75% of the these emerging markets face challenges in prac- world’s 100 top agricultural crops rely on natural tice, they offer rewards to early business entrants pollinators (Klein et al . 2007 ). Property developers (e.g., markets for carbon offsets or certif ed tim- benef t from the coastal protection provided by ber—see Section 19.3 ). coral reefs, coastal forests and coastal wetlands Most companies, however, fail to connect ecosys- ( Turner et al . 1998 ; Cesar 2000 ). The tourism indus- tems with their business bottom line ( Hanson et al . try benef ts from these ecosystems’ aesthetic beauty 2008 ). Many are not aware of the extent to which ( De Groot 1994 ). they impact or depend on ecosystem services, nor Because of these impacts and dependencies, the the ramif cations of those impacts. Likewise, busi- degradation of ecosystem services presents signif - ness tools such as environmental management sys- cant risk to—and opportunities for—corporate tems and environmental impact assessments are performance (see Table 19.2 and Box 19.1). For often not attuned to the risks and opportunities example, industries relying on steady supplies of arising from use and degradation of ecosystem clean freshwater face operational risks when services. Rather, they are suited to “traditional” upstream deforestation increases sedimentation of issues of pollution and natural resource consump- rivers, disrupting business operations and increas- tion. As a result, companies may be unprepared for ing costs (see Chapter 6 ). Companies may face per- ecosystem service risks or miss new sources of rev- mit restrictions or damage brand image if they enue associated with ecosystem change.

Table 19.2 Business risks and opportunities arising from dependencies and impacts on ecosystem services

Type Risk Opportunity

Operational • Increased scarcity or cost of inputs • Increased eff ciency • Reduced output or productivity • Low impact industrial processes • Disruption to business operations Regulatory and legal • Extraction moratoria • Formal license to expand operations • Lower quotas • New products to meet new regulations • Fines • Opportunity to shape government policy • User fees • Permit or license suspension • Permit denial • Lawsuits Reputational • Damage to brand or image • Improved or differentiated brand • Challenge to social “license to operate” Market and product • Changes in customer preferences • New products or services (public sector, private sector) • Markets for certif ed products • Markets for ecosystem services • New revenue streams from company-owned or managed ecosystems Financing • Higher cost of capital • Increased investment by progressive lenders and socially responsible • More rigorous lending requirements investment funds PUTTING ECOSYSTEM SERVICES ON THE AGENDA 343

Box 19.1 An assessment of ecosystem services helps a paper and packaging business respond to emerging risks

Craig Hanson grazing. Selective controlled grazing is, however, widely practiced. One approach for businesses to identify connections between ecosystem services and their bottom line is to A trends analysis of these six ecosystem services uncovered conduct a Corporate Ecosystem Services Review (ESR) a number of emerging risks and opportunities facing ( Hanson et al . 2008). The ESR is a structured methodology Mondi. Freshwater in the three plantation watersheds is that helps managers proactively develop strategies to becoming increasingly scarce due to the proliferation of manage business risks and opportunities arising from their invasive alien plant species, increasing demand for water company’s dependence and impact on ecosystems. from nearby farmers, and climate change. This scarcity Conducting an ESR involves identifying priority ecosystem threatens to increase the cost of water, reduce the services for a company (a facility, product or supply chain), availability of wood f ber, and expose the company to analyzing trends in these services, and identifying business reputational and regulatory risk. Meanwhile, new risks and opportunities. With this information, the company opportunities are arising as ecotourism grows in the region can develop response strategies. and new markets emerge for biomass fuel. Mondi is a leading international paper and packaging Through the ESR process, Mondi identif ed several business. As of 2006, the company was Europe’s largest strategies for managing these risks and opportunities: producer of off ce paper, with operations in 35 countries. Much of the company’s pulp comes from its plantations in • Improve water-use efﬁ ciency. To improve water South Africa. Mondi conducted an ESR to understand what eff ciency, the company is now clearing invasive species business risks—and opportunities—might arise as changes more aggressively. It will also better match tree species to in ecosystem services affect its plantations. site conditions and more frequently burn grasslands. These During its ESR, Mondi assessed the dependence and strategies complement the company’s past efforts to impact of these plantations on 24 different ecosystem remove plantations from wetlands, thereby restoring services. The analysis identif ed six services as having the natural hydrological systems. most impact on Mondi’s corporate performance: • Use invasive species as biomass fuel. Mondi can tap into the growing market for biomass fuel by using the • Freshwater. Pine and eucalypt plantations signif cantly invasive species cleared from its plantations as feedstock depend upon and affect the quantity of freshwater in their for power and heat generation. Potential users of the watersheds. feedstock are Mondi’s own paper mills and a new • Water regulation. Plantations rely on and impact the biomass pellet manufacturer located close to one surrounding ecosystems that regulate the timing of water plantation. f ows. • Promote coppiced small-scale tree farms (woodlots) • Biomass fuel. As a byproduct, plantations generate for biomass fuel. Using the company’s forestry biomass that can be used for energy by the company’s mills expertise, Mondi can help local landowners and and local villages. villages establish woodlots on degraded land for • Global climate regulation. Plantations can sequester growing biomass fuel on coppiced rotations. Mondi carbon dioxide thereby mitigating climate change (albeit could provide seedlings and offer extension services. dependent on the stage of the carbon cycle—see Mondi could also purchase the wood to use in its mills C h a p t e r 7 ) . or sell to nearby wood pellet manufacturers. These • Recreation and ecotourism. One of the plantations is woodlots would provide additional income for villagers located next to the isiMangaliso Wetland Park—a World and thereby strengthen Mondi’s reputation and Heritage Site—and contains wetlands and grasslands that stakeholder relationships. could potentially provide new opportunities for recreation • Engage policy-makers to improve freshwater policies. and ecotourism. Mondi can support policies that encourage water-use • Livestock. The plantations preclude surrounding eff ciency in South Africa and, leveraging its expertise in villagers from using the land for large-scale livestock water management, provide input into policy design. 344 INCORPORATING ECOSYSTEM SERVICES IN DECISIONS

Information on the value of ecosystem services services are private goods (e.g., agricultural pro- can play a role in raising business awareness of the duction and timber), and as such, are easily traded links between ecosystems and prof ts. Spatially in markets that regulate their provision. Other explicit information on where ecosystem services important services (e.g., pollination, climate regula- are supplied can help businesses determine where tion, and cultural values) are “public goods,” which to invest in ecosystem protection and restoration or means private individuals cannot reap prof ts from where to avoid activity that degrades ecosystems. providing them. Governments can develop regulations or incentives to alter the provision of these services. Here, we discuss a range of policy instru- 19.2.3 Building public awareness ments for affecting the provision of both public and Because people often fail to connect their wellbeing private goods. to ecosystem conditions (Irwin and Ranganathan 2007 ), there is a need to build public awareness 19.3.1 Regulation about this linkage before we can expect the public to hold decision-makers accountable (Pielke 2007 ). Government regulations have only recently begun Education and awareness programs may be espe- to consider the full range of services provided by cially effective if they are directed at landowners or ecosystems, investing in them to meet human needs managers who lack information. For instance, if a eff ciently and effectively ( Ruhl et al . 2007 ). Several farmer does not realize that an increase in habitat kinds of regulations have recently been altered to for native pollinators will raise yields, simply pro- include ecosystem services: licensing and permit- viding the type of information provided by the ting, zoning and land-use planning, and environ- models outlined in Chapter 10 may induce land- mental standards. owners to set aside habitat. An example comes from Currently, most governments grant licenses and a US Federal “Conservation Buffers” program, permits for extractive activities (e.g. mining, oil and where training and advice is given to farmers on gas) and infrastructure development (e.g. roads, designing buffer zones to f lter pollution along developments) based on projected impacts to biodi- streams and wind barriers to reduce soil erosion versity and a subset of environmental conditions (NRCS 2008). Some consider technical training to be linked to human health. However, standard assess- more successful at altering land-use practices than ments usually fail to account for the full set of social direct payments (Daily and Ellison 2002 ; Salzman impacts associated with proposed activities. 2005 ). Governments are slowly moving toward permitting When “externalities” exist—impacts that affect and licensing procedures that more fully account people other than the decision-maker—information for environmental and social damages. Models such alone will not suff ce. For example, a farmer may as those discussed in this book can be informative learn that applying fertilizer causes water quality for these new, expanded impact assessments. For problems downstream but continue to do so because example, the Colombian Ministry of Environment, it leads to higher crop yields and the resulting water Housing and Territorial Development is expanding quality problem affects others. In these cases, edu- the impacts considered in its licensing process for cation and awareness programs need to target the activities related to oil and gas, mining, and infra- broader public to create demand for reform. structure to avoid, minimize, mitigate and compensate for biodiversity and ecosystem service damages. 19.3 Instruments for sustaining The Ministry plans to use information on where and enhancing ecosystem services services are supplied and realized to identify risks from industrial activities, and to shape permit con- When ecosystem service management becomes an ditions and offset compensation programs. institutional priority, several policy instruments can Regulations of common pool resources, such as be used to inf uence the way people interact with protected areas in marine, coastal and inland the environment (MA 2005d). Some ecosystem water ecosystems, are widespread. However, they INSTRUMENTS FOR SUSTAINING AND ENHANCING ECOSYSTEM SERVICES 345 are usually designed to maintain a single ecosys- 19.3.2 Market-based approaches tem service, such as f sheries or freshwater. To support regulation, governments may construct Regulatory limits can be designed to supply a market-based instruments that create incentives for broader range of services such as water purif ca- people to account for ecosystem services. The creation, waste treatment and recreation. For example, tion of property rights, typically enforced and regu- in the United Kingdom, the Department of lated by governments, is common to many of these Environment, Food and Rural Affairs (DEFRA) approaches, facilitating trade in activities that estimated the economic value of a range of ecosys- increase or maintain services. In theory, market- tem services affected by marine protected areas to based approaches are cost effective as services can inform the design of new regulations—Marine be provided at the lowest possible cost (Montgomery Conservation Zones—that are part of a national 1972 ). This is particularly the case when producers Marine Bill (McVittie and Moran 2008 ; Moran et al. have f exibility—they can provide a service in dif- 2007 ). The impact assessment supporting the fering ways and at differing costs ( Revesz and Marine Bill legislation demonstrated that the ben- Stavins 2004 ). In practice, eff ciency advantages can ef ts of Marine Conservation Zones outweigh the be outweighed by the costs of operating these— costs to industry. If the Marine Bill is passed, often complex—mechanisms. Here we discuss sev- information on the value and location of ecosys- eral forms of market-based approaches that have tem services and biodiversity is intended to inform been used to inf uence ecosystem service provision selection of protected areas that provide multiple and detail how quantitative analyses can improve benef ts matching policy priorities. A similar shift or facilitate the development of these policy is happening in the United States, where a new instruments. ocean policy will likely require an ecosystem- based approach to marine zoning that explicitly considers a wide range of ecosystem services 19.3.2.1 Cap and trade programs ( Lubchenco and Sutley 2010). Under “cap and trade” programs, the government Governments can also use information on the sets a limit on an environmental externality, such as provision or value of ecosystem services to ban or pollution. Permits to emit are issued to f rms, who set standards for activities that degrade the envi- can then trade them. Cap and trade programs have ronment beyond a socially def ned limit. In the reduced emissions at lower costs than regulations, a Republic of the Marshall Islands, information on notable example being the market for SO2 permits the value of coastal ecosystem services—protection in the United States. Carbon markets are now from erosion, along with tourism, recreation and emerging as a mechanism to achieve greenhouse f sheries benef ts—was used to advocate banning gas emissions reductions. In this context, the scope reef blasting and near-shore dredging (McKenzie of cap and trade programs may expand to encom- et al. 2006). In addition, liability rules can hold those pass ecosystem services. For example, payments for degrading ecosystem services responsible for the Reduced Emissions from Deforestation and Forest damage they cause, and enforce compensation for Degradation (REDD) allow benef ciaries of the cli- negligent actions, such as pollution (Thompson mate regulation service provided by forests to com- 2008b ). In these cases, information on the economic pensate those who conserve them (UNFCCC 2008 ). value of degraded services can lead to more accu- In this case, maps of carbon storage and projected rate estimates of environmental damages, and help estimates of future carbon change can be used to build consensus about compensation among con- identify areas where carbon investments in natural f icting stakeholders ( Van Beukering et al . 2007 ). For or restored systems would be most prof table. example, after the Exxon Valdez oil spill in Alaska in 1989, a study estimated the lower bound eco- 19.3.2.2 Voluntary markets nomic value of damages at $2.8 billion, based on the In addition to regulated markets, voluntary markets stated willingness to pay to prevent another similar are emerging, such as the Chicago Climate Exchange spill ( Carson et al . 1992 ). for carbon. Given that most ecosystem services are 346 INCORPORATING ECOSYSTEM SERVICES IN DECISIONS public goods or plagued by externalities, voluntary exchange for a guarantee that a specif ed amount of markets are only one part of the solution. In volun- an ecosystem service, or a particular land use or tary carbon markets, individuals or companies buy practice believed to provide that service, will be certif ed reductions in carbon emissions for moral delivered ( Wunder 2005 ). A famous example of a reasons or to improve their reputation. As men- PES program involves payments by Nestle (the tioned above, certif cation schemes exist to meet owner of the Vittel mineral water company) to demand for “charismatic” carbon activities that farmers in north eastern France to safeguard the deliver additional benef ts (CCBA 2008b). For exam- water supply. Upstream farmers had replaced natu- ple, the Climate, Community and Biodiversity ral grasslands—which f ltered and cleaned the Alliance recently awarded the Juma Reserve project water—with corn and cattle, resulting in nitrate in Brazil a Gold rating, verifying additional benef ts contamination of the aquifer. After calculating the through income generation and the promotion of higher costs of water purif cation plants, Nestle local business and biodiversity (CCBA 2008a). decided to use a PES approach. The company signed Integrated models of multiple ecosystem services long-term contracts with farmers to manage their can estimate how activities to reduce greenhouse animal waste, graze cattle sustainably, and reforest gas emissions will change a larger suite of ecosys- water f ltration areas ( Perrot-Maitre 2006 ). tem services. Such estimates might be used to grant The majority of PES schemes are government- certif cation to projects whose projected impacts funded. For example, in the well-known Catskills benef t multiple services. Follow up monitoring will example, the city of New York decided to protect be essential in determining whether these projected the Delaware and Catskills watersheds, which pro- benef ts are realized, but modeling can help mini- vide 90% of its drinking water (Heal 2000 , NRC mize negative impacts of early carbon investments. 2004) . Regulations were important for driving the The Voluntary Carbon Standard (VCS—a widely scheme; the US Safe Drinking Water Act required accepted certif cation scheme for carbon offsets) is that water must be f ltered unless watersheds are adding the need for a different kind of analysis. suff ciently protected. The costs of a new f ltration Under VCS, projects must demonstrate that their plant vastly outweighed the costs of restoring and emissions reductions are additional to what would protecting the watersheds ( Postel and Thompson have happened anyway. To pass, a project must be 2005 ). To protect the watersheds, a set of policy neither legally mandated nor common practice, and measures were introduced alongside standard pay- face unique barriers to implementation that carbon ments, including land purchase and easements. f nance could help overcome (VCS 2008). Modeling Within the f rst f ve years of the policy initiative, the the consequences of different scenarios can help city purchased approximately 14 000 ha of land and estimate additional ecosystem service benef ts, by another 1000 ha of conservation easements, thereby comparing a “baseline” scenario (what is assumed doubling the buffer zones around important reser- to happen without the project), with a “project” sce- voirs. The scheme had to be carefully negotiated nario. The difference in service delivery helps iden- with diverse landowners; in this case there were 477 tify the truly additional quantity, or that eligible for property owners ( Postel and Thompson 2005 ). compensation. When framed by baseline scenarios, Those providing water quality services beyond the the models described in the rest of this book can regulated minimum are rewarded with a payment make this type of calculation, estimating both addi- package, including additional income, but also in- tional carbon sequestration benef ts (see Chapter 7 ) kind benef ts, such as farm management consulta- and the additional benef ts of other services. tions ( Appleton 2002 ). Beyond the New York City watershed case, most 19.3.2.3 Payments for ecosystem services (PES) existing water-related PES programs have been Payments for ecosystem services (PES) represent an established without information on where invest- increasingly popular method for creating incentives ments will provide the greatest ecosystem service to sustain services. PES involve contracts in which returns, and payments are made based on one party agrees to compensate another party in implementation of activities, not on actual changes INSTRUMENTS FOR SUSTAINING AND ENHANCING ECOSYSTEM SERVICES 347 in ecosystem service levels. This approach can lead ers for managing the land sustainably and provid- to ineff cient programs that fail to produce desired ing ecosystem services, such as regulating nutrient ecosystem service returns because the marginal runoff and soil erosion, and maintaining the spirit- changes in ecosystem services associated with a ual and symbolic values of agricultural landscapes given management change are not constant ( Jack ( Baylis et al . 2008 , Ilbery and Bowler 1998 ). et al. 2008 ). For example, a kilometer of fencing Although growing in scale, agri-environment installed along a riverbank where cattle graze will programs nevertheless remain a small proportion usually reduce erosion much more than a kilometer of agricultural expenditure (OECD 2003 ). Reducing of fencing installed around croplands in the upper agricultural subsidies has proved particularly diff - reaches of the same watershed. Although policy cult in Europe and the United States due to per- creation in such contexts is diff cult, it can be dra- ceived impacts on farmers’ livelihoods and lifestyles matically simplif ed with the use of models that (Myers 2001 ). As subsidies are often maintained by reveal how management changes in different rent-seeking by special interest groups who lose out regions affect desired ecosystem services. More from subsidy reform, reform policies are more pop- complex, but eff cient payment designs (e.g. differ- ular if done gradually and combined with public ential taxes, trading zones) can be established with awareness campaigns (Pearce and Finck von such information. In addition, models such as those Finckenstein 1999). Such campaigns can draw on described in this book can help locate areas for new valuation studies that quantify the effects of subsi- PES schemes by identifying where there are sources dies on ecosystem services to build political sup- of service supply and benef ciaries who may be port, and create transparency and accountability willing to pay for service delivery. This is particu- about who wins and who loses from policy change. larly powerful when combined with information on land tenure, to determine who can inf uence service 19.3.2.5 Offsets delivery and should receive payments. For further Offset schemes attempt to ensure that degradation discussion on additional conditions under which of ecosystem services in one location is compen- PES can be most effective, see Jack et al . ( 2008 ). sated through activities elsewhere. For example, wetland mitigation banking is a consolidated offset 19.3.2.4 Government f scal incentives scheme in the United States to encourage compli- Governments can create further f nancial incentives ance with the Clean Water Act. This regulation to supply ecosystem services using subsidies, sub- requests that developers avoid dredging wetlands sidy reform and taxes. Subsidies differ from PES connected to other water bodies and offset any una- schemes in that payments are not conditional on voidable damages through creation of wetlands service delivery, or some proxy for it. Subsidies that elsewhere. Landowners that restore, establish, account for the full economic value of ecosystem enhance, and (in exceptional circumstances) pre- services can help to maintain or increase their pro- serve wetlands gain permits that can be sold to vision. Many existing subsidies, however, actually developers f lling wetlands. However, mitigation encourage activities that degrade services, by focus- banking has been controversial because of the ing on a narrowly def ned outcome and ignoring scheme’s potential to redistribute wetland ecosys- ecosystem service losses (Myers 2001 ). Signif cant tem services between human populations and the cases exist from agricultural subsidies. Farmers in mixed quality of offsets relative to the wetlands Europe and the United States have long received destroyed ( Ruhl et al . 2007 ). Offsets work best when payments linked to crop productivity ( OECD 1999 ). there is a well-def ned, scalable unit that can be Such payments promote land conversion, leading exchanged, whether it is an acre of wetland or a to erosion of topsoil, and increasing use of fertiliz- pound of CO2 , and when exchanges are fungi- ers and pesticides that affect water quality. Recent ble—an acre of wetland in one location produces reforms have attempted to replace agricultural sub- the same services as an acre elsewhere (Salzman sidies with payment programs (often called and Ruhl 2000 ). Since an acre of wetland seldom “agri-environment” schemes), which reward farm- provides the same service in different watershed 348 INCORPORATING ECOSYSTEM SERVICES IN DECISIONS contexts, standard, reproducible, and relatively to certain institutions—governments are the only easy ways of measuring ecosystem services are entity with authority to levy taxes, for example— essential for establishing trades that actually avoid and some will be constrained by the nature of the the loss of ecosystem services. Combining biophysi- service (its “production function”). Others require cal models with information on benef ciaries can strong systems of governance and institutions that determine how offsets affect the distribution of may not exist in developing countries, such as clear services among stakeholder groups, allowing agen- property rights and competitive markets (Pearce cies to preempt equity issues associated with devel- 2005 ). To choose the best approach, information is opments and their offsets. required on a range of issues, many of which are summarized in Table 19.3 . For further details within 19.4 Choosing the right instrument the conservation context, see McKenney et al. ( 2008 ) who describe a set of enabling conditions affecting Decision-makers need to select the policy mecha- the success of ecosystem service projects. nism most appropriate to their local context if they A summary of the main conditions affecting suc- are to succeed in delivering ecosystem services. cessful use of each policy mechanism is given in How to choose? Some approaches are appropriate Table 19.4 .

Table 19.3 Information needed to select a policy instrument to provide ecosystem services

The service • How do interventions alter the level of service being produced? • Can these changes be measured? • What (if any) other policies and interventions are compatible with providing the service? • Is a particular spatial pattern of land uses required? • Can the service be broken down into discrete, fungible units? • At what scale is the service produced? • Are other services co-produced with the desired one? • Is there uncertainty regarding the link between actions and service delivery? • Are there unpredictable shocks that could alter service levels? Producers • If other land uses are compatible with service provision, are behaviors that increase or decrease service levels observable and enforceable? • Are there any relevant information asymmetries between producers, the decision-maker, and benef ciaries? Beneﬁ ciaries • Who benef ts from the ecosystem service? Where are they located? • Will this change over time? • Can others be prevented from benef ting from the service (in the language of economics, is the service a public good)? Costs and funding • Who is bearing the costs? Where are they located? • What are the costs of various approaches for providing the service? (This should include direct costs, indirect costs such as reductions in the supply of other ecosystem services, and transaction costs.) • How much are people willing to pay? How much funding is available (and over what time frame) and where is it coming from? Goals • What are the goals of the policy? Goals could be improving livelihoods, biodiversity conservation, or simply cost-effectiveness. Institutional context • What institutions are needed for an approach and are they present (for example, contracts enforced by a court of law)? • Are there clear property rights? • What other laws are in existence that might affect implementation of the approach? Do these conf ict with or enhance any of the approaches to providing the service? Views of stakeholders • Who wins and who loses with each approach? • Which stakeholder groups could prevent implementation? • Whose support is needed and what are their views? Performance over time • How will each approach respond to changing conditions over time, such as price changes, changes in technology, changes in funding levels, or changes in drivers of service degradation? BUILDING STRONGER ORGANIZATIONS 349

Table 19.4 Enabling conditions for policy instruments

Cap and trade Public and private Taxes, subsidies and Regulations Information provision schemes, voluntary payments liability rules markets, offsets

Ecosystem service— • Well-def ned, • Alternative land • Changes in service • No f exibility in • Alternative land uses biophysical fungible units uses compatible levels can be how to supply compatible with attributes with supplying attributed to service supplying service service individual behavior or specif c actions (liability) • Specif c spatial arrangements of land uses required • Ecosystem service at or near tipping point

Producers of • Heterogeneity in • Small number of • Small number of • Producers already • Producers unsure how ecosystem service supply costs among large producers producers (liability) subject to other to increase service providers forms of regulation levels • Flexibility among • Land-use changes • Action clearly • Large point sources • Producers unable to providers on how to would not happen correlated with capture demand for increase services without payment supply or degradation service due to of service that is information asymmetry amenable to a f scal incentive (tax or subsidy) • Numerous point • Opportunity costs of sources land-use change not prohibitively high

Beneﬁ ciaries • Willing and able • Bene f ciaries can • No collective action • A certain level of • Benef ciaries would pay buyers organize to make problems restricting service must be for service if they knew suff cient payments use of liability rules guaranteed it was being provided

Costs and beneﬁ ts • Cost of trading • Transaction costs not • Revenue needed • Minimal transaction • Costs of providing is low prohibitive (taxes) costs to regulator information of suff cient • Funding available to • Insuff cient funding • Landowners face quality not prohibitive sustain program for payments similar costs for over time (liability/taxes) achieving service level

Quantif ed, mapped information on ecosystem priate to local contexts. But the tables also highlight services and their values can answer a number of that biophysical and economic models will usually these questions, particularly those relating to serv- need to be complemented by additional informa- ices and their biophysical attributes—the f rst rows tion, looking at the institutional context, property in Tables 19.3 and 19.4 . Ecosystem service maps can rights, and other social and economic conditions. identify where services are supplied. Distributional and land tenure information can identify who bears 19.5 Building stronger organizations the costs and who benef ts from service delivery. This underlines the fact that ecosystem service Effective actions to provide ecosystem services modeling can help select policy instruments appro- require strong and legitimate organizations and 350 INCORPORATING ECOSYSTEM SERVICES IN DECISIONS processes for making decisions ( Pearce and Second, organizations must establish mecha- Seccombe-Hett 2000 ). The Millennium Ecosystem nisms to distribute benef ts and costs of ecosystem Sub-global Assessments found that “effective” service change equitably, both at the local level and responses to ecosystem degradation tended to between local, regional and global levels (see involve collaboration across governance levels. C h a p t e r 1 6) . E c o s y s t e m d e g r a d a t i o n i s a s i g n i f cant Organizations must be equipped to implement factor—sometimes the principal one—exacerbating coordinated, cross-sectoral policies that consider poverty (MA 2005b) because the rural poor depend multiple services and trade-offs. There are three key on ecosystems for their livelihoods (Paavola 2008 ). prerequisites that need to be met for organizations Yet local communities have often been left out of to become effective implementers of ecosystem environmental management decisions, typically service policy. dominated by national or international interests. First, ecosystem services must move from the One positive example of a step in the right direc- periphery to the center of environmental and natu- tion entails Namibia, which in 1996, adopted the ral resource management organizations’ agendas. Nature Conservation Act giving rural communities Currently, natural resource ministries typically rights to form conservancies in communal lands to focus on provisioning services, such as f sh, timber build tourism and wildlife industries. By 2007, 50 and water, and neglect regulating services. conservancies benef ted over 230 000 rural people Fortuantely, there are indications that governments and covered 14% of the country’s area. Improved are giving regulating services—and ecosystem serv- management has increased wildlife populations ices in general—a more central role. For example, and reduced overgrazing (WRI 2008 ). In this case, the US Forest Service has recently reformulated its decentralizing authority to representative local mission to focus on conserving ecosystem services. institutions and establishing clear property rights In the United Kingdom, DEFRA has adopted an strengthened environmental management. Equity action plan to take a more holistic approach to poli- is not only about the distribution of costs and ben- cy-making to maintain ecosystem services and ef ts: participation is also important (Paavola 2007 ). ensure their values are ref ected in decisions If boundaries of participation are set locally, a for- (DEFRA 2007). est assessment might focus on timber and non-tim-

Box 19.2 Cultural evolution as an enabling condition for the use of ecosystem services in decisions

as it became clear that the norm of “putting aside a Paul R. Ehrlich, Lee D. Ross, and Gretchen C. Daily nature reserve to save a charismatic species” was failing The idea of “ecosystem services” originated through cultural to stem the loss of biodiversity. The “ecosystem service evolution, and this idea holds the secret to their continuation. norm” now spreading makes plain the dependence of Most human evolution does not involve changes in our DNA, human wellbeing on investment in natural capital (e.g., but rather alterations in the gigantic library of non-genetic D a i l y a n d M a t s o n 2 0 0 8; G o l d m a n et al . 2 0 0 8 ; K a r e i v a a n d information—the culture—of Homo sapiens . Culture Marvier 2007 ; MA, 2003). Policy efforts to reward comprises such evolving entities as languages, beliefs, scientif c investments in natural capital must tap into, and reinforce, theories, and systems of justice and oppression, in addition to this emerging norm. the information embodied in books, songs, computer disks, Numerous failed civilizations attest to the diff culty of and artifacts ranging from potsherds to jumbo jets. The directing cultural evolution toward environmentally decisions we make and the policies we devise to manage our sustainable practices ( Diamond 2005 ; Tainter 1988 ). On ecosystem services are shaped by our culture. Improving those Easter Island, early inhabitants surely knew that chopping decisions and policies thus depends on cultural evolution. down the entire palm forest would not be a good thing. The idea of ecosystem services entered the modern They depended heavily on porpoise meat, hunted from mainstream of cultural evolution about three decades ago, palm wood canoes. Nonetheless, they cut down every last BUILDING STRONGER ORGANIZATIONS 351

tree. They then wiped out coastal food resources, resorted the basis of decency norms that no child should go in desperation to eating rats, and f nally turned on each hungry in school. other. Possession of “scientif c” understanding—the The ecosystem service outcomes that are promised, importance and declining availability of palms would have whenever feasible, should reinforce existing norms and been evident to all—did not prevent disaster. Other islands show results in a satisfying time frame. In cases where we followed equally grim paths. But on others, truly sustainable know a lot scientif cally, as with carbon sequestration, we economies emerged. What accounted for the difference? It might do well to promise outcomes that are specif c, appears that size matters. Tikopia, a model of success, is attainable within a generation (or two), and congruent with only about 1.7 mi2 . Kirch ( 1997 ) proposes that where shared values (for example preserving forests). Where we everyone knew everyone else, ecological limits to human know less or have less local/regional control, as with activities were more likely to be accepted, and major provision of water quantity, it may prove detrimental to “policy changes” (like giving up pork) and new institutions make specif c promises. Our scientif c understanding is sure (regulating fertility) adopted. Conversely, the Easter-scale to evolve greatly; policy instruments, and the institutions (64 m 2 ) islands were prone to dividing into “them” and that shape them, need to evolve with this. “us” in a race to the bottom. On the question of how much to pay, relying on f nancial Two island lessons are particularly relevant to today. incentives or disincentives can be tricky. Payments for First, like the islanders, we know enough scientif cally to ecosystem services may actually reduce their supply if they recognize trouble and start moving in the right direction. remove cultural norm violation as a factor inf uencing Second, human beings evolved as a small-group animal. behavior. Yet carefully tuned payments are important for Our future prospects depend on whether, in a population of establishing new norms, especially where suppliers and 6 billion-plus, we can design and implement cooperative benef ciaries of services are widely separated. solutions. On the question of how long payments should be made, Social psychology is central to understanding cultural evidence from social psychology argues for shorter-term evolution, and has important, often counter intuitive (one generation, say) payments to achieve greater support. lessons for fostering change. Consider a popular policy Once a program is adopted and becomes the status quo, option: payments for ecosystem service (PES). Three key people are unlikely to violate group norms by not “doing issues arise: what should people pay for; how much can their share.” They are unlikely to exert political pressure to and should they pay; and what time frame for payment “opt out” (even if they had not initially favored “opting is best? in”). On the question of what people should pay for, social Underlying these tactical questions are deeper ones. psychologists stress that overselling, as a way of What combination of services should be targeted in policy, achieving greater public “buy-in” for the measures being given that there are trade-offs among services? Who advocated, should be avoided as it may be both decides? The answers may emerge quite differently in counterproductive in the long run and unnecessary. different places. Yet, our culture is evolving rapidly in the School lunch programs in the United States, for example, way people think about the environment, and we have a were promoted as a means to achieve a range of social window of opportunity to foster this change productively goals (kids with full bellies, it was claimed, could learn around the world. This will require cooperative efforts to better, leave school with better job prospects, and be less develop and deploy innovative policy approaches likely to turn to crime). American taxpayers did not see worldwide. A Millennium Assessment of Human Behavior these effects, and resistance to the program increased. may be the best shot at achieving the “small group” The irony is that Americans would probably have communication necessary to do this ( Ehrlich and Kennedy supported lunch programs for disadvantaged students on 2005 ). With luck, it might keep us from eating each other.

ber forest products. If set more broadly, water processes for framing ecosystem service modeling regulation and f ltration, and climate regulation assessments. may receive greater emphasis (Reid et al . 2 0 0 6 ) . I f Third, there is a need to coordinate across sectors, local voices are excluded, important ecosystem levels and timeframes. Organizations, such as a services may continue to be degraded. This empha- national forestry agency or a city council, follow sizes the need for participatory stakeholder political boundaries and jurisdictions, rather than 352 INCORPORATING ECOSYSTEM SERVICES IN DECISIONS the ecosystem service boundaries. Yet effective laws, and wrestle with powerful vested interests. management requires partnerships and networks As a f rst step, there is much to learn from the exist- across ecosystem boundaries. “Bridging organiza- ing understanding of best practice. Objective evalu- tions” can lower the costs of collaboration and con- ations are required in two areas: f rst, to evaluate f ict resolution ( Folke 2005 ). An increasingly popular how ecosystem service interventions perform over approach, co-management, involves government time and in different contexts, and thereby learn agencies working with local stakeholders and how to be most effective (Berkes et al . 2003 ; Ostrom organizations. A national agency’s authority to 2005 ; Carpenter et al. 2009 ); second, to assess con- adopt regulations, provide scientif c data, and tinually the information, and level of certainty of obtain funding and political support complements that information, required for different interven- local organizations’ on the ground understanding tions, so that scientif c models and tools can be of the natural resource and management capability ref ned in light of user needs. ( Irwin and Ranganathan 2007 ; WRI 2005). For exam- Our constant challenge is to ensure that science ple, locally managed marine areas in Fiji and else- and practice are effectively integrated, by working where in the Pacif c use traditional practices to set across disciplines and political boundaries, and aside portions of f shing grounds as restricted areas, sharing ideas and experiences. Ecosystem service allowing f sheries to recover. Local communities science needs to be grounded in sound theory but, lead in designating restricted areas, monitoring, to be most effective, it must always keep the f nal and enforcement. The national government used its application—the “practice”—f rmly in mind. regulatory authority to introduce a 12-mile limit on f shing by foreign trawlers ( WRI 2008 ). References 19.6 Future directions Appleton, A. F. (2002). How New York City used an ecosystem services strategy carried out through an urban- Our scientif c understanding of ecosystem services rural partnership to preserve the pristine quality of its is growing increasingly sophisticated—both in the drinking water and save billions of dollars. A paper for natural and social sciences. However, to improve Forest Trends. Tokyo. outcomes on the ground, information must be sali- Ascher, W. and Healy, R. (1990). The policy process and ent to the world’s problems, and channeled into complexity in natural resource policy-making. Natural resource policy-making in developing countries: Environment, behavior change. We have seen that information on economic growth, and income distribution. Duke University ecosystem services is already making a difference. Press, Durham. Governments, businesses and the public increas- Assey, P., Bass, S., Cheche, B., et al. (2007). Environment at ingly realize the signif cance of ecosystem services the heart of Tanzania’s development: lessons from for achieving their ends. A variety of policy instru- Tanzania’s National Strategy for Growth and Reduction ments and business practices are being imple- of Poverty—MKUKUTA National Resource Issues Series mented more effectively to maintain important 6. International Institute for Environment and services. And organizations are evolving to manage Development, London. ecosystem services effectively and equitably, using Balmford, A., Bruner, A., Cooper, P., et al. (2002). Economic approaches such as adaptive co-management. reasons for conserving wild nature. Science, 297, Models with the characteristics described in this 950–3. Baylis, K., Peplow, S., Rausser, G., et al . (2008). Agri- book can be powerful forces for change. They can environmental policies in the EU and United States: A alter perceptions about the importance of the envi- comparison. Ecological Economics, 65, 753–64. ronment for people. They can also answer many of Berkes, F., Colding, J., and Folke, C. (2003). Navigating the questions needed to select, design and imple- social-ecological systems: Building resilience for complexity ment different policy instruments. and change. Cambridge University Press, Cambridge. While progress is evident, it is diff cult to change Bingham, G., Bishop, R., Brody, M., et al. (1995). Issues in the way people perceive and value ecosystems; one ecosystem valuation: Improving information for deci- may have to f ght f erce political battles, build new sion making. Ecological Economics, 14, 73–90. FUTURE DIRECTIONS 353

Carpenter, S. R., Mooney, H. A., Agard, J., et al . (2009). Goldman, R. L., Tallis, H., Kareiva, P., et al . (2008). Field Science for managing ecosystem services: Beyond the evidence that ecosystem service projects support biodi- Millennium Ecosystem Assessment. Proceedings of versity and diversify options. Pr oceedings of National National Academy of Sciences, 106, 1305–12. Academy of Sciences, 105, 9445–8. Carson, R. T., Mitchell, R. C., Hanemann, W. M., et al . Hanson, C., Ranganathan, J., Iceland, C., et al. (2008). The (1992). A contingent valuation study of lost passive use Corporate Ecosystem Services Review: guidelines for identi- values resulting from the Exxon Valdez oil spill. fying business risks and opportunties arising from ecosystem Attorney General of the State of Alaska. change . World Resources Institute, Washington, DC. Climate Community and Biodiversity Alliance (CCBA). Heal, G. (2000). Nature and the market-place: capturing the (2008a). The Juma Sustainable Development Reserve Project: value of ecosystem services, Island Press, Washington, reducing greenhouse gas emissions from deforestation in the DC. States of Amazonas, Brazil. CCB—Validation Report. TUV Henrichs, T., Zurek, M., Eickhout, B., et al. (2008). Scenario SUD Industrie Service GmbH. development and analysis for forward-looking ecosys- Climate Community and Biodiversity Alliance (CCBA). tem assessments. Ecosystems and human well-being: A (2008b). Climate, Community and Biodiversity Project manual for assessment practitioners. Draft for external Design Standards , second edi. CCBA, Arlington VA. review. Cesar, H. S. J. (2000). Collected essays on the economics of coral Ilbery, B. and Bowler, I. (1998). From agricultural produc- reefs , CORDIO, Kalmar University, Sweden. tivism to post-productivism. In: B. Ilbery, Ed., The geog- DAC Network on Environment and Development raphy of rural change. Prentice Hall, London. Co-operation (ENVIRONET). (2008). Strategic environ- Irwin, F., and Ranganathan, J. (2007). Restoring nature’s mental assessment and ecosystem services. OECD. capital , World Resources Institute, Washington, DC. Daily, G. C., and Ellison, K. (2002). The new economy of Jack, B., Kousky, C. and Sims, K. E. (2008). Designing pay- nature: The quest to make conservation prof table. Island ments for ecosystem services: Lessons from previous Press, Washington, DC. experience with incentive-based mechanisms. Daily, G. C., and Matson, P. A. (2008). Ecosystem services: Proceedings of National Academy of Sciences, 105, 9465–70. from theory to implementation. Proceedings of National Kareiva, P., and Marvier, M. (2007). Conservation for the Academy of Sciences, 105, 2455–6. people. Scienti f c American , 297 , 50–7. Daily, G. C., Polasky, S., Goldstein, J., et al . (2009). Kiker, G. A., Bridges, T. S., Varghese, A., et al . (2005). Ecosystem services in decision-making: Time to deliver. Application of multicriteria decision analysis in envi- Frontiers in Ecology and the Environment, 7, 21–8. ronmental decision making. Integrated Environmental Daly, H. E. (1968). On economics as a life science. Journal of Assessment and Management, 1, 95–108. Political Economy, 76, 392–406. Kirch, P. V. (1997). Microcosmic histories: Island perspec- De Groot, R. (1994). Environmental functions and the eco- tives on “global” change. American Anthropologist, nomic value of natural systems. In: A. M. Jansson, 99 , 3–42. M. Hammer, C. Folke, and R. Costanza, Eds., Investing Klein, A.-M., Vaissiere, B. E., Cane, J. H., et al . (2007). in natural capital: the ecological economics approach to sus- Importance of pollinators in changing landscapes for tainability . Island Press, Washington, DC. world crops. Proceedings of the Royal Society Biological DEFRA. (2007). Securing a healthy natural environment: An Sciences, 274, 303–13. Action Plan for embedding an ecosystems approach. Le Quesne, T. and McNally, R. (2004). The green buck— Department for Environment Food and Rural Affairs, using economic tools to deliver conservation goals: a WWF London. f eld guide , WWF-UK. Diamond, J. (2005). Collapse: how societies choose to fail or Leopold, A. (1949). A Sand County Almanac, and sketches succeed . Viking, New York here and there . Oxford University Press, Oxford. Ehrlich, P. R., and Kennedy, D. (2005). Millennium assess- Lubchenco, J., and Sutley, N. (2010). Proposed ocean pol- ment of human behavior. Science, 309 , 562–3. icy for ocean, coast and great lakes stewardship. Science , Folke, C. (2005). Adaptive governance of social-ecological 3 28, 1485–6. systems. Annual Review of Environment and Resources, 30, Lubchenco, J. (1998). Entering the century of the envi- 441–73. ronment: A new social contract for science. Science, Ghazoul, J. (2007). Recognising the complexities of ecosys- 279, 491–7. tem management and the ecosystem service concept. Millennium Ecosystem Assessment (MA). (2003). GAIA—Ecological Perspectives for Science and Society, 16, Ecosystems and human well-being: a framework for assess- 215–21. ment . Island Press, Washington, DC. 354 INCORPORATING ECOSYSTEM SERVICES IN DECISIONS

Millennium Ecosystem Assessment (MA). (2005b). Ostrom, E. (2005). Understanding institutional diversity. Ecosystems and human well-being: synthesis. Island Press, Princeton University Press, Princeton. Washington, DC. Paavola, J. (2007). Institutions and environmental govern- Millennium Ecosystem Assessment (MA). (2005c). Living ance: a reconceptualisation. Ecological Economics, 63, beyond our means: natural assets and human well-being: 93–103. Statement from the Board. World Resources Institute, Paavola, J. (2008). Livelihoods, vulnerability and adapta- Washington, DC. tion to climate change: lessons from Morogoro, Tanzania. Millennium Ecosystem Assessment (MA). (2005d). Environmental Science & Policy, 11 , 642–54. Ecosystems and human well-being: policy responses . Island Pearce, D. W. (2005). Investing in environmental wealth for Press, Washington DC. poverty reduction: environment for the MDGs . United McKenney, B., Morris, B. and McKenzie, E. (2008). Nations Development Programme, United Nations Framework for assessing the viability of an ecosystem service Environment Programme, International Institute for approach to conservation: the top 10 screening criteria. Environment and Development, IUCN, World Resources Nature Conservancy, Natural Capital Project, and Institute, New York. WWF-US. Pearce, D. W., and Finck Von Finckenstein, D. (1999). McVittie, A., and Moran, D. (2008). Determining monetary Advancing subsidy reform: towards a viable policy values for use and non-use goods and services: Marine package. Finance for sustainable development: Testing new biodiversity—primary valuation. Department for policy approaches. United Nations Division for Sustainable Environment, Food and Rural Affairs, London. Development. Montgomery, D. W. (1972). Markets in licenses and eff - Pearce, D. W., and Seccombe-Hett, T. (2000). Economic cient pollution control programs. Journal of Economic valuation and environmental decision-making in Theory, 5, 395–418. Europe. Environmental Science & Technology, 34, Moran, D., Hussain, S., and Fofana, A. (2007). Marine Bill 1419–25. marine nature conservation proposals: Valuing the ben- Perrot-Maitre, D. (2006). The Vittel payments for ecosys- ef ts. Department for Environment, Food and Rural tem services: A perfect “PES” case?, International Affairs, London. Institute for Environment and Development. Myers, N. (2001). Perverse subsidies: How tax dollars can Pielke, R. A., Jr. (2007). The honest broker: making sense of undercut the environment and the economy . Island Press, science in policy and politics. Cambridge University Press, Washington, DC. Cambridge. Naidoo, R., Malcolm, T., and Tomasek, A. (2009). Economic Postel, S. L., and Thompson, B. (2005). Watershed protec- benef ts of standing forests in highland areas of Borneo: tion: capturing the benef ts of nature’s water supply quantif cation and policy impacts. Conservation Letters, services. Natural Resources Forum, 29, 104–5. 2 , 35–44. Ranganathan, J., Raudsepp-Hearne, C., Lucas, N., et al. Naidoo, R., and Ricketts, T. H. (2006). Mapping the eco- (2008). Ecosystem services: a guide for decision makers . nomic costs and benef ts of conservation. PLoS Biology, World Resources Institute, Washington, DC. 4, 2153–64. Reid, W., Berkes, F., Wilbanks, T. J., et al. (2006). Bridging Nelson, E., Mensoza, G. M., Regetz, J., et al . (2009). scales and knowledge systems: Concepts and applications in Modeling multiple ecosystem services, biodiversity ecosystem assessment . World Resources Institute, conservation, commodity production and tradeoffs at Washington, DC. landscape scales. Frontiers in Ecology and the Environment, Revesz, R. L., and Stavins, R. (2004). Environmental law 7, 4–11. and policy. In: A. M. Polinsky and S. Shavell, Eds., The NRC. (2004). Valuing ecosystem services: toward better envi- handbook of law and economics. Elsevier Science, ronmental decision-making. National Academies Press, Amsterdam. Washington, DC. Ruhl, J. B., Kraft, S. E., and Lant, S. L. (2007). The law NRCS. (2008) Buffer strips: common sense conservation. and policy of ecosystem services . Island Press, United States Department of Agriculture, Natural Washington, DC. Resources Conservation Science. Salzman, J. (2005). Creating markets for ecosystem serv- OECD (1999). Agricultural policies in OECD countries: ices: Notes from the f eld. NYU Law Review, 80, Monitoring and evaluation. Organisation for Economic 870–961. Co-operation and Development, Paris. Salzman, J., and Ruhl, J. B. (2000). Currencies and the com- O E C D ( 2 0 0 3 ) . The greening of the WTO Green Box . Organisation modif cation of environmental law. Stanford Law Review, for Economic Co-operation and Development, Paris. 53, 607–94. FUTURE DIRECTIONS 355

Snel, M. (2004). Poverty-conservation mapping applica- economics toolkit. Joint Nature Conservation Committee, tions. IUCN World Conservation Congress. Cooperazione Peterborough, UK. Italiana, IUCN, UNEP and GRID Arendal. Voluntary Carbon Standard (VCS). (2008). Voluntary Tainter, J. (1988). The collapse of complex societies . Cambridge Carbon Standard 2007.1: Specif cation for the project- University Press, Cambridge. level quantif cation, monitoring and reporting as well as Thompson, D. (2008b) Union Pacif c to pay $102 million validation and verif cation of greenhouse gas emission for forest f re. Associated Press. reductions or removals. Turner, R. K., Lorenzoni, I., Beaumont, N., et al . (1998). Wilson, E. O. (1998). Consilience: the unity of knowledge. Coastal management for sustainable development: ana- Kopf, New York. lysing environmental and socio-economic changes on W R I ( 2 0 0 5 ) . The wealth of the poor—Managing ecosystems to the UK coast. The Geographical Journal, 164, 269(1). f ght poverty . World Resources Institute, Washington, DC. Turner, R. K., Van den Bergh, J. C. J. M., Soderqvist, T., et al . WRI (2007). Nature’s benef ts in Kenya: An atlas of ecosystems (2000). Ecological-economic analysis of wetlands: and human well-being . World Resources Institute, Scientif c integration for management and policy. Washington, DC; Department of Resource Surveys and Ecological Economics, 35, 7–23. Remote Sensing, Ministry of Environment and Natural UNDP (2007). Pilot integrated ecosystem assessment of Resources, Kenya; Central Bureau of Statistics, Ministry Bugesera. United Nations Environment Programme, of Planning and National Development, Kenya; Repubulika y’u Rwanda and United Nations International Livestock Research Institute; Nairobi. Environment Programme. WRI (2008). Roots of resilience—growing the wealth of the UNFCCC (2008). United Nations Climate Change poor. World Resources Institute, Washington, DC. Conference in Bali. Wunder, S. (2005). Payments for environmental services: Van Beukering, P., Brander, L., Thompson, E., et al . (2007). Some nuts and bolts. Occasional Paper 42. Center for Valuing the environment in small islands: An environmental International Forestry Research, Jakarta. This page intentionally left blank Index

accounting for ES, incentives 363 lodge-controlled lands, map 199 marginal biodiversity value accounting tools 4 Trans-Amazon Highway 199 (MBV) 233 , 235 Adapting Mosaic scenario, MA 10–12 anthropocentric value 16–17 , 20–7 range shifts 332–5 adaptive management of benef t–cost analysis 16–17 species–area relationships landscapes 11 , 71 aquaculture, mangrove lands 28 , 34 (SAR) 236–7 , 241 African Americans 221–2 Asia, South, agriculture and terrestrial 229–46 agricultural ES biodiversity 230 see also marine conservation ; estimate of effects of climate atmosphere–ocean general circulation terrestrial biodiversity change 325 model (AOGCM) 326–7 biodiversity model monetary value 280 attributes, environmental, nature- based landscape level 235 native animal pollinators 157 , 183 tourism 189–92 , 200 , 203 linking with models for cultural non-market contributions 280 avoided costs 24 , 96, 98–100 services 218 nutrient uptake 151 , 161–2 dredging costs 42 , 90 , 98–100 biogas trading, Nepal 131 provisioning services 153–5 sediment retention modeling 40 , Biological Intactness Index (BII) value 159 42 , 90 , 98–100 system 229 supply and use of water treatment costs 99 bioprospecting 250 production 152–3 black people, and conservation Tier 1 modeling 154–8 , 324 bats, and pest control 39–40 organizations 221–2 Tier 2 modeling 158–62 , 324 bees see pollination services boating 200 limitations and next steps 163–4 behavior change 352 Bolivia, Noel Kempff Project value of regulatory and supporting benef ciaries of ES 43 (NKCAP) 115 ES inputs 157–9 and poverty 278–93 Borden Ranch case, wetlands in USA 27 yield functions 154–5 , 324 poverty alleviation Borneo, oil palm development 341 agriculture 150–67 agroecosystems as Budyko model, dryness index 54–5 and biodiversity, South Asia 230 pathway 280–1 bushmeat harvest 7–8 , 250–1 , 254 Hawai’i 254–5 and human development 7–8 , 341 business strategies, high food prices 280 trade-offs 341 improvement 341–2 LULC scenarios 151–2 benef t transfer studies mapping impact on ecological estimate of total economic value vs calibration processes 161–2 value of individual services 36 T ier 2 model 64–5 , 90 temperature change, and crop nature-based tourism 200 , 203 verif cation, validation (models) 90 production 323 value of ES 36 water balance calibration water use, total crop benef t–cost analysis 16–17 , 30 , 341 constant 66 transpiration 161–2 see also cost–benef t analyses California, Central Valley, pollination water-related services 327–8 (CBA) services 273–6 AgriMet 328 bio-fuels 281 camping, visit rates 193 , 195 Alaska, subsistence activities 218–20 and high food prices 280 cap and trade programs 345 Amazon Basin case study , biocentric value, willingness to carbon offsets 114 , 347 deforestation 285–90 pay 17–18 tree biomass estimation 119–20 Amazonian Peru biodiversity carbon sequestration/storage 111–25 , cattle ranchers 198 and agriculture, South Asia 230 261 , 268–73 forest protection economics and bees 168 aggregate market value 254 tourism 196–9 Endangered Species Act (US), changes in ES value 308 lodge revenues 198 inconsistency of approach 19 maximization 252

357 358 INDEX carbon sequestration/storage (cont.) coral reef valuation 301–3 linking with models for other ES offset markets 114–15 design based on valuation of and biodiversity 218 REDD 114–16 , 119 , 345 coastal ecosystems 28–30 methods predicting levels 269 economic value of degraded eight dimensions 208 predicting patterns in space 269–72 services 345 for ES valuation 207–9 Puget Sound 310–12 mangrove lands, effect of shrimp integrating cultural services and soil model 123 aquaculture 28 , 34 non-use values 213–15 Tier 1 supply model 112–21 storm surge alleviation 298 non-use values 25 , 207–26 Tier 1 valuation model 118–21 Colombia, government regulation of T ier 1 methods 216–20 Tier 2 supply model 121–4 ES 344–5 Tier 2 methods 220–2 Tier 2 valuation model 122 community forestry, degraded state see also timber forests 131 dams, f ood risk management 80–1 carbon stock 268–73 conf ict resolution 352 decision-making using ES 3–6 , 339–55 carbon storage congestion, nature-based tourism 203 better business strategies 341 harvested wood products conservation areas, location 44 better development policies 341 (HWPs) 113 conservation, forests, compensation building public awareness 344 marine habitats 305–10 for conservation 345 business risks and per hectare 112 conservation, marine 296–318 opportunities 342 (table) predicting levels 269 conservation organizations, and characteristics of relevant ES carbon storage pools 112 people of color 221–2 models 340 (table) Caribbean, coral reef valuation 301–3 conservation planning (SCP) 229 , 254 ecological production function census data, correlations 218 conservation site selection approach 35–7 China opportunity costs, Global framework 5 ecosystem function conservation Agro-Ecological Zone (GAEZ) impacts of climate change 323–38 areas 3 datasets 152 incomplete balance sheets ecosystem service payments spatially explicit models 250 , 252 intrinsic rights 18 , 19 , 27 (1998–2010) 3 consumer surplus, nature-based methods and tools for non-market f ood damage and forestry tourism 197 , 200 valuation 35 programs 107–8 consumptive/non-consumptive uses multi-criteria decision-making Forest Ecological Benef t of ES 16 , 22–4 (MCDM) 225 , 340 Compensation Fund contingent valuation 21 , 23 , 30 , 41 policy mechanisms 339 , 345 , 348 (FEBCF) 107 coral reef valuation 301–3 role of ES information 339–40 Natural Forest Protection Program corridor restoration deforestation (NFPP) 107 benef ts relative to costs 250 , 256 Amazon Basin case study 285–90 Sloping Lands Conversion Nepal 130–1 loss of livelihood and well Program (SLCP) 107 cost–benef t analyses (CBA) 16 , 30 , being 285–90 CITES, and US Endangered Species 260 , 341 non-timber forest production Act 19 anthropocentric value 16–17 , (NTFP) 279 , 285–90 Clean Water Act (US) 27 20–7 degraded state forests 131 Borden Ranch case 27 corridor restoration 250 GHG emissions 111 intrinsic rights 19 , 27 ecosystem service models 260 degraded services 131 redistribution of wetland services 282 crop pollination services 15 , 24 , 40 , liability rules 345 , 349 climate change 168–87 , 273–6 deliberative monetary valuation adaptation strategies 327–8 crop production, and climate (DMV) 224 and crop production 323 change 323 demand for ES 39–40 climate simulation (mid-21C), CropWat 328 development Willamette Basin cultural evolution, and ES 350 policies, decision-making using (Oregon) 326–37 cultural keystone species 207 ES 341 climate-driven changes in ES, cultural services 206–26 pressure, climate simulation modeling 323–38 def ned 206 (mid-21C) 326 previous analyses 323–4 distinguishing values from discount rates 30 , 60 , 82 , 98 , 104 , 121 , climate-envelope models 336 services 209–12 123 co-management 352 environmental/social processes, distribution of ES, and poverty 278–93 coastal protection and service production 212–13 dredging costs, avoided 42 , 90 , 98–100 benef t–cost analysis 30 interdependency, double-counting, dryness index, Budyko model 54–5 coastal zone services and and trade-offs 214–15 dynamic global vegetation benef ts 30 limitations and next steps 222–3 models 324 , 336 INDEX 359

Easter Island, degradation 350–1 f sheries, changes in ES value 308 habitat-quality and habitat rarity ecological processes (supply), vs f ood control 23 maps 229 ecosystem services (supply f ood risk harvest-right holders and demand) 38–40 calculating f ood damage 81 exclusive harvest volume and ecological production function integrated f ood risk value 144–6 changes in output of marketed management 80–1 limitations and next steps 146–7 goods and services 36–7 LULC Tier 1 and Tier 2 protection of property rights 136 modeling approach 35–7 modeling 73–88 supply, use and value of economic growth see also water-related services provisioning service 141–2 and sustainable environment 10 food, local food systems 281 harvested wood products (HWPs) see also development Food Quality Protection Act, carbon storage 113 economic valuation swampbuster clause 27 steady-state harvest volume and methods 5 food web support, changes in ES value 133 nutrient retention 90 value 308 volumes, costs and product ecosystem production functions, forests 129–50 prices 132–8 map/assess natural capital 6 compensation for conservation 345 Hawai’i, policies and payments for ES 4 ecosystem services (ES) degraded state forests 131 Hawai’i, Kamehameha Schools 254–5 changes in value, carbon storage/ provisioning service contrasting directions with sequestration, commercial supply and value 141–2 agricultural lands 254–5 f sheries, and food web Tier 1 132 cultural values 257 support 308 Tier 2 141–6 educational services 256 free services (6 Fs) 8 see also deforestation ; timber land use/land cover maps 258 models see modeling of ES production residential housing multiple ecosystem services 34–50 framing effects, willingness-to-pay development 254–5 norms 350 (WTP) 224 scenarios 254–6 , 258 predicting ES patterns in space 268 free services (6 Fs) 8 social value of carbon valuation methods 21 (table) fuels, bio-fuels 280–1 reduction 254 see also value of ES fuelwood, and provision of subsidies spatially explicit scenarios 254–5 Ecosystem Services Review (ESR) 343 for biogas 131 Strategic Plan 2000–2015 257 ecotourism see nature-based tourism future directions of ES practice 9 sugarcane ethanol scheme 254–5 education and awareness timber production 254–5 programs 244 Galapagos Islands win-win solutions 249 , 254 eelgrass contest between extraction and herring, Puget Sound (Washington carbon storage 305–10 conservation 31 State) case study 316–17 Puget Sound (Washington State) ecotourism 28–31 history, study of renewable case study 305–18 general theory of change 12 resources 4 eff ciency frontier 257–60 GHG emissions Hortonian f ows/runoff 75–6 , 101 Willamette Basin (Oregon) 257 , 259 deforestation 111 household harvest volume 143 Endangered Species Act (US) 19 see also climate change human development ethics, intrinsic rights 18 , 19 , 27 GHG f ux 161 human well-being 279–83 evapotranspiration coeff cient 66 Global Agro-Ecological Zone (GAEZ) poverty 7–8 , 278–95 , 341 existence value (passive use datasets 151–3 hunting 16 , 191 , 194 , 217 value) 16 , 24–5 , 31 , 250 opportunity costs of subsistence 219 , 289 evaluation 224 conservation 152 hydro energy demand, monsoonal mapping and quantifying 220–1 , global vegetation models 324 , 336 f oods 80 223–4 governance, incentives for people to Hydrologic Engineering Centers Exxon Valdez oil spill, economic account for ES 363 River Analysis System value of degraded government (HEC–RAS) 74 services 345 f scal incentives to supply ES 347 Hydrologic Response Units and property rights, enforcement/ (HRUs) 63 , 74 , 101 Federal Emergency Management regulation 345 Hydrological Simulation Agency (FEMA) 81 Guatemala Program–Fortran (HSPF) 53 , 74 feedback 9 payments for ES (PES) 290–3 hydropower 53 , 60–2 f sh/f shing water services case study 290–3 correlation with wildlife impact assessments 341 viewing 200 habitat, loss of habitat area, and loss impacts of climate change on ES, and seed-dispersal services 182 of ecosystem value 26 modeling 323–38 360 INDEX

incentives, for people to account for keystone species 26 , 207 market-based instruments 345–8 ES 363 Kuhn–Tucker model 201–2 cap and trade programs 345 information and ES decisions 264–77 governmentf scal incentives 347 future directions and open land management, unmitigated/ payments for ecosystem services questions 276–7 mitigated scenarios 330 (PES) 346 used to inform management by land use and land cover (LULC) selection of policy mechanism 348 diverse decision-makers 267 patterns 43–4 voluntary markets 345–6 information from ES, impact on agriculture 151–2 Marshall Islands, value of coastal decision-making 323–38 cultural services 220 ecosystem services 345 infrastructure variables, visitation f ood risk measurement of ES model 194–6 Tier 1 modeling 74–84 anthropocentric value 16–17 , 20–7 insects, control by bats 39 Tier 1 valuation 78 ecological production function institutions, new 351 Tier 2 modeling 84–6 approach 36 intangibility, cultural keystone Tier 2 valuation 84 valuation methods 21 (table) species 207 land-use change 216–18 willingness to pay 17 , 20, 220–1 intrinsic rights 18 , 19 , 27 alternative land-use plans 252–3 mentha oil, Nepal 131 vs value of ES 18 landscape level biodiversity Millennium Development Goals InVEST models 32 , 37–47 model 235 (MDGs) 278 , 341 applying as part of a stakeholder least squares regression model, annual Millennium Ecosystem Assessment process 43 (table) visits to Oregon State Parks 201 (MA) classif cation of ES modeled 38 liability rules 345 , 349 xv 3–5 , 20 , 29 , 34–5 , 206 , 278 , 304 , (table) degraded services 345 , 349 350–1 future directions and open local food systems 281 ES provided by oceans and questions 47–9 lodge revenues, Amazonian Peru 198 coasts 297–8 land management decisions 257 loss of habitat area, and loss of f ve linked components of land use and land cover (LULC) ecosystem value 26 poverty 282–4 patterns 43–4 LULC see land use and land cover Technogarden and Adapting Mosaic marine 318 (LULC) patterns scenarios 10–12 multiple ES 37–47 mitigation banking, wetlands Natural Capital Project 318 mahogany timber production, and scheme 347 nutrient retention 96 pollination services 183 mitigation and offsetting, InVEST 45 as part of stakeholder process 43 mainstreaming natural capital into modeling of ES 5 , 249–62 scenario-driven modeling 43–4 decisions 3–14 agricultural ES 154–64 , 324 applications 44–6 management analysis of location, type, and three-step structure 40 (table) adaptive management of intensity of demand for tiered system 46–7 landscapes 11 , 71 services 58 valuation 59 land, unmitigated and mitigated atmosphere–ocean general vs SWAT, spatial patterns in water management scenarios 330 circulation model yield 68 mangrove lands (AOGCM) 326–7 water supply 53–72 nonlinear wave attenuation 299–300 benef t–cost analysis 260 irrigation 41–2 , 53 shrimp aquaculture 28–9 , 34 biodiversity model 218 , 235 CropWat 328 marginal biodiversity value landscape level 235 demand, climate-driven (MBV) 233 , 235 linking with models for cultural changes 327–8 relative RMBV 235 services 218 surface water supply value, marginal value, vs total value 25 calibration, verif cation, validation example 157–8 marine conservation 296–318 (models) 64–5 , 90 see also water-related services Caribbean, coral reef carbon sequestration 123 island lessons 350–1 valuation 301–3 climate-driven changes 323–38 isochrones 77 mangrove land-use choices, climate-envelope 336 nonlinear wave complexity, sensitivity of ES jay, seed dispersal 39 attenuation 299–300 decisions 264–77 oceans and coasts, ES conservation site selection 250 , 252 Kendall tau ranking statistics 67 provided 297–8 cost–benef t analyses (CBA) 260 (table) Puget Sound case study 303–18 cultural services, linking with Kenya, poverty Marine Conservation Zones 345 biodiversity 218 ES mapping 282 market prices, marginal and total decision-making, characteristics of trade-offs 341 values 21 relevant ES models 340 (table) INDEX 361 dryness index, Budyko model 54–5 Mondi, Ecosystem Services Review time costs of visiting 189 , 203 dynamic global vegetation (ESR) 343 travel costs 196 , 202 models 324 , 336 monetary value of ES visitation data 191 , 194–203 ecological production agricultural ES 280 NatureServe 230 function 35–7 deliberative monetary valuation negotiation 339 , 346 ecosystem service models 249–62 (DMV) 224 Nepal f ood risk 73–88 market/non-market valuation corridor restoration 130–1 global vegetation models 324 , 336 methods 260 Terai Arc Landscape (TAL) 130 impacts of climate change on monsoonal f oods, hydro energy Nestle, payments for ecosystem ES 323–38 demand 80 services (PES) 346 Kuhn–Tucker model 201–2 Monte Carlo simulation analysis 162 , New York City watershed case 346 management 260 242 New Zealand, logging restriction 136 methods/tools, future directions 9 Muir, John 221 Noel Kempff Mercado Climate nature-based tourism 190–202 multi-criteria decision-making Action Project (NKCAP), nutrient retention models 90–104 (MCDM) 225 Bolivia 115 parks multiple ecosystem services 34–50 non-timber forest products infrastructure variables 194–6 ecological production function (NTFPs) 129–50 least squares regression model 201 approach 35–7 A m a z o n B a s i n case study 279 , 285–90 visitation model 194–203 InVEST tool 37–47 non-use values 25 , 206–26 pollination services 174–7 , 275 methods and tools for non-market norms Precipitation Runoff Modeling valuation 35 ecosystem service norm 350 System (PRMS) 62–3 , 84 , social norms 351 89–90 , 102 , 105 Native Americans 221–2 nutrient export 102 , 103–5 Puget Sound (Washington State) natural capital nutrient retention case study mapping/assessing 6 constraints and limitations 104–8 food web model 313–16 value to rural poor 7–8 economic valuation 90 mapping and modeling f ows of Natural Capital Project (NatCap) 37 , Tier 1 model 90–9 ES 303–17 257 Tier 2 model 99–104 prediction sites, eelgrass 306 see also InVEST models random utility model 202 nature-based tourism 30 , 188–205 Oahu see Hawai’i, Kamehameha scenario-driven modeling 43–4 , attributes Schools 340 , 346 , 364 spatial correlations 189 , 200 , 203 oak, seed dispersal 39 sediment retention modeling 42 , t o u r i s m l i n k a g e s 1 8 9 – 9 0 , 2 0 0 , 2 0 3 oceans and coasts, providing 90 , 98–100 benef t transfer 200 , 203 ES 297–8 soil erosion modeling 94–6 choice-based sampling 203 offset schemes 347 spatially explicit models 42–3 , 250 , consumer surplus 197 , 200 oil palm development, Borneo 341 252 environmental attributes 189–90 , opportunity costs of storm peak mitigation 73–88 200 , 203 conservation 152–3 Tanzania, Eastern Arcs Mountain for F3 state-of-the-art (Tier 3) option value 25 , 31 Watershed 136–43 model 201–2 Oregon see Willamette Basin (Oregon) terrestrial biodiversity 229–36 forest protection in Amazonian organizations testing agreement, simple vs Peru 197–8 building 349–52 complex ES models 268–73 Kuhn–Tucker model 201–2 and conf ict resolution 352 UKMO-HadCM3 coupled limitations and next steps 202–3 model 326 linkages to environmental Paraguay, Mbaracyau Forest vegetation, dynamic global attributes 190 Biosphere Reserve 250–5 vegetation models 324 , 336 random utility model 202 particulate matter air pollution 267 water treatment 110–13 social and environmental payments for ecosystem services water use modeling 53–92 processes 188–90 (PES) 5 , 48 , 346 , 351 Willamette Basin (Oregon), social interactions 188 Guatemala 290–3 nutrient and sediment substitute sites and activities 189–94 payments by Nestle 346 retention 89–106 Tier 1 tourism supply and use people of color, and conservation see also InVEST models ; land use model 190–3 , 197–8 organizations 221–2 and land cover (LULC) Tier 2 tourism supply and use Peru patterns ; scenario-driven model 193–7 cattle ranchers 198 modeling ; sediment retention Tier 3 (state-of-the-art) forest protection economics and modeling ; headings above model 201–2 tourism 196–9 362 INDEX

Peru (cont.) dependence on ES 279–83 def ned 188 lodge revenues 198 including institutions 293 see also nature-based tourism lodge-controlled lands, map 199 Kenya, ES mapping and recreational services 24 Madre de Dios 198 poverty 282 Reduced Emissions from Tambopata National Reserve mapping poverty and ES 283–5 Deforestation and Forest (TNR) 198–9 potential for payments for ES, in Degradation (REDD) 71 , pest control 39–40 highland Guatemala case 114–15 philosophical foundations, value of study 290–3 carbon investments in natural or ES 16–20 UBN (unsatisf ed basic restored systems 345 planning processes 341 , 344 needs) 285–9 regulation of ES poaching, anti-poaching units 131 poverty indicators 283–5 direct vs indirect use values 23–4 policy evaluations 340 , 341 environmental income, % of total government policies 344–5 policy mechanisms 339 , 345 , 348 income in resource-poor/ reservoir sedimentation, avoided 40 trajectory, alignment 46 resource-rich areas 279 revealed expenditure methods 22–3 pollination services 15 , 24 , 40 , pairing with ES 284 river channel hydraulic software, 168–87, 273–6 Precipitation Runoff Modeling HEC–RAS 81 animal-mediated 157 , 182–3 System (PRMS) 62–3 , 84 , roadmap 6–7 bees 89–90 , 102 , 105 runoff, direct 75–6 , 86 diversity 168 principle-based values 208 , 213 Curve Number 75–6 , 86–7 , 101 foraging 40 , 171–82 production functions see ecological beyond agriculture 182 production function Safe Drinking Water Act 346 California 177–8 project and policy evaluations 340 , 341 s a t u r a t i o n o v e r l a n d f ow 56 , 75–6 , 85–6 Central Valley, California 273–6 property rights, enforced/regulated scenario-driven modeling 43 , 340 , Costa Rica 178 by governments 345 346 , 364 crop vulnerability across provisioning services 21–3 , 129–50 , 2030 landscape scenarios, Sierra Europe 170 , 183 153–5 , 159 Nevada Conservancy 237 economic value of pollinators agriculture 153–5 , 159 2050 land-use change 253 169–70, 342 direct vs indirect use values 21–3 Adapting Mosaic , MA 10–12 mahogany timber production 183 forests 141–2 agriculture (LULC) 151–2 monetary value of global harvest-right holders 141–2 alternative (Willamette Basin) 252 services 169–70 Puget Sound Partnership, action applications 44 predicting crop pollination areas (map) 311 climate simulation (mid-21C) 326 levels 273 Puget Sound (Washington State) case conservation 254 predicting crop pollination patterns study 303–18 Hawai’i, Kamehameha in space 274 carbon sequestration 310–12 Schools 254–6 , 258 relative crop yields 172 , 175 , 177 , 182 carbon storage habitats 305–10 InVEST models 43–6 and seed dispersal, f sh/ changes in habitats 305–12 land management, unmitigated/ f shing 182 eelgrass 305–18 mitigated management 330 sensitivity analysis 177 , 181 model prediction sites (map) 306 land-use change 216–18 , 253 Tier 1 modeling 174–6 , 275 food web model 313–16 stakeholder-driven 254 , 339 Tier 1 valuation 175 future directions 317–18 Technogarden , MA 10–12 Tier 1 vs Tier 2 172 , 274 mapping and modeling f ows of SCS–curve number (CN) spatial agreement 275 ES 303–17 approach 75–6 , 86–7 , 101 Tier 2 modeling 176–7 , 275 changes in ES structure and sediment export 58 , 94 Tier 2 valuation 177 function 307–9 sediment retention modeling 58 , pollutants marine harvest and 89–106 particulate matter air pollution 267 non-consumptive values avoided dredging costs 42 , 90 , regulation, social value 96–7 312–16 98–100 water treatment 91 , 101 , 103 , 105 , 108 spatial variation, herring spawn avoided water treatment costs 99 poverty alleviation locations 316–17 export coeff cients 91–2 agroecosystems as pathway 280–1 sheetwash erosion 90 , 94 , 96 , 105 and human development 7–8 , 341 random utility model 202 social value 90 , 96–7 solar energy 280 range shifts, biodiversity 332–5 Tier 1 biophysical models 90–6 p o v e r t y a n d d i s t r i b u t i o n o f E S 2 7 8 – 9 5 ranking, Kendall tau ranking testing 105–8 agents/victims of environmental statistics 67 (table) Tier 1 economic valuation 96–9 degradation 279–83 rattan, Nepal 131 Tier 2 biophysical models 99–102 data resolution 283 recreation Tier 2 economic valuation 102–4 INDEX 363 seed dispersal 39–40 Willamette Basin (Oregon) 218 , time costs, visiting nature-based sensitivity analyses 65–6 , 236–42 254 , 255 tourism 189 , 203 and analysis limitations 242 strategic environmental assessments tourism see nature-based tourism pollination services 181 (SEA) 341 trade-offs sensitivity of ES decisions, to model streamf ow recession, and water converting to a win–win 254 complexity 264–77 retention 70 possibility of markets in carbon shrimp aquaculture, mangrove subsidy programs 347 credits 254–5 lands 28–9 , 34 InVEST 45–6 travel costs 22 , 23 , 30 Sierra Nevada Conservancy subsistence activities nature-based tourism 196 , 202 (California) 236–45 Alaska 218–20 Yosemite 23 2030 landscape scenarios 237 hunting 219 , 289 tree biomass and carbon content 119 site selection for conservation 250–2 ‘swampbuster clause’ 27 tree volume, estimation 119 Sloping Lands Conversion Program SWAT 40 , 53 , 60 , 66–70 , 76 , 86 , 101 , (SLCP), China 107 106 , 108 , 164 UKMO-HadCM3 coupled model social norms 351 systematic conservation planning AOGCM 326 social processes (SCP) 229 Universal Soil Loss Equation cultural services 212–13 (USLE) 40 , 90 nature-based tourism 188–90 Tanzania use of ES social value poverty reduction strategies 341 def ned 39–40 carbon reduction 254 surface water supply value, use values 206 , 210 ecosystem services 90 , 96–7 example 157–8 sediment retention modeling 90 , 96–7 Tanzania, Eastern Arcs Mountain validation of ES models 64 , 177–8 , soil erosion modeling 94–6 Watershed 117–18 206 Universal Soil Loss Equation peak f ow mitigation map 80 Tier 2 model 90 (USLE) 40 , 90 storm volume upstream of city of value of ES 15–33 , 206 Soil and Water Assessment Tool Ifakara, Tanzania 79 aesthetic, artistic, educational, (SWAT) 40 , 53 , 60 , 66–70 , 76 , Tier 1 intermediate access model spiritual and/or scientif c 86 , 101 , 106 , 108 , 164 example 136–41 values 206 soil water storage 73 , 90–2 Tier 2 model (harvest volume/ anthropocentric value 16–17 , 20–7 solar energy, poverty alleviation 280 effort observable) 141–3 area-based benef t estimates 36 spatial correlations Technogarden , MA scenario 10–12 benef t transfer studies 36 census data 218 temperature change, and crop biocentric value 17–18 environmental attributes and production 323 carbon sequestration 254 , 308 tourism 189 , 200 , 203 terrestrial biodiversity 229–46 case studies 27–31 spatially explicit models, limitations and next steps Galapagos Islands 28–31 conservation site selection 250 , 242–3 vegetation and coastal 252 Tier 1 and 2 examples with protection 28–30 species value 26 sensitivity analysis 236–42 wetlands in USA 27–8 keystone species 26 , 207 Tier 1 habitat-quality and rarity coastal protection 345 species–area relationships model 229–33 consumptive/non-consumptive (SAR) 236–7 , 241 Tier 2 models 233–6 uses of ES 16 , 22–4 stakeholder-driven scenarios 254 , 339 theory of change 12 cultural services 209–12 state-of-the-art tourism model 201–2 theory to implementation 5–6 direct, non-consumptive use storm depth 74–5 , 78 , 83–4 , 86 Tikopia 351 values 22–3 storm peak timber production 129–50 direct vs indirect use values 23–4 direct runoff 75–6 , 86 harvest volumes, costs, and economic value of pollinators return periods 73 , 75 , 78 , 82–6 product prices 132–8 169–70 , 342 storm peak mitigation 73–88 intermediate services 41 (table) existence/passive use value 16 , economic valuation of the provisioning supply, use and 24–5 , 31 , 220–1 , 223–4 , 250 landscape 83 value 129–50 f sheries 308 limitations and next steps 86 Tier 1 132–41 food web support 308 LULC, Tier 1 and Tier 2 Tier 2 141–6 forests 141–2 modeling 73–88 steady-state harvest volume and harvested wood products SCS–curve number (CN) value 133 (HWPs) 133 approach 75–6 , 86–7 , 101 sustainable 251 timber production 129–50 Tier 1 and Tier 2 modeling 73–88 see also harvested wood products harvest-right holders 144–6 Tier 2 supply and use model 84–6 (HWPs) and loss of habitat area 26 364 INDEX value of ES (cont.) Water Evaluation and Planning wildlife viewing, correlation with marginal biodiversity value System (WEAP) 63 f shing 200 (MBV) 233 , 235 water partition 71 Willamette Basin (Oregon) marginal vs total value 25–6 water retention, and streamf ow alternative future marine (oceans and coasts) recession 70 scenarios 252–3 297–8 water runoff, direct 75–6 , 86 avoided dredging costs 42 , 90 , methodological approaches for ES Curve Number 75–6 , 86–7 , 101 98–100 valuation 207–10 water saturation overland f ow 56 , avoided water treatment costs monetary value of ES 224 , 260 , 280 75–6 , 85–6 99 non-use values 24–5 , 206–26 water treatment berry and grape production option value 25 , 31 avoided costs 99 327 philosophical foundations 16–20 m o d e l i n g f ows relevant to biodiversity pollination services 169–70 , 342 nutrient and sediment conservation, opportunity pollutant regulation 96–7 retention 110–13 costs 152–3 principle-based values 208 , 213 pollutant loading 91 , 101 , 103 , 105 , range shifts 332–5 provisioning services 21–3 , 129–50 , 108 vertebrates 332–7 153–5 , 159 water use modeling 53–92 boating 200 regulating services 23–4 hydropower 53 , 60–2 camping visit rates 193 , 195 revealed expenditure irrigation 41–2 , 53 carbon sequestration and methods 22–3 a n d m a r g i n a l w i l l i n g n e s s t o stock 268–73 social values 90 , 96–7 , 254 pay 22 carbon storage in upland species value 26 , 207 Tier 1 water supply model 54–9 forests 328–32 travel cost method of estimating 23 hydrologic improvements 71 climate impacts on ES 326 use values 206 , 210 limitations 62 adaptation strategies 327–8 virtue-based values 208 , 213 sensitivity analyses and climate simulation (mid-21C) vs intrinsic rights 18 , 19 , 27 testing 65–6 326–37 water-related services 58 , 61 , 106 valuation 59–62 carbon storage 328–32 cultural values 218 Tier 2 water supply model 62–70 s c e n a r i o s : d e v e l o p m e n t and cultural values 218 valuation 65 pressure; plan trend; surface water supply water-related services 40 conservation 326 value 157–8 availability for agriculture 327–8 terrestrial vertebrate wetlands in USA 27–8 Clean Water Act 19 , 27 , 282 diversity 332–7 see also monetary value of ES intrinsic rights 19 , 27 day visit rates 195 value of species climate-driven changes in eff ciency frontier 257 , 259 assigning 26 ES 323–4 environmental attributes (EA) 185, keystone species 26 , 207 and cultural values 218 189–90 , 192 vegetation effects of changes in temperature f shing, correlation with wildlife and coastal protection 28–30 and precipitation on water viewing 200 dynamic global vegetation availability 323 irrigation demand 327 models 324 , 336 US Safe Drinking Water Act 346 land management 330 and land cover valuation, storm water balance calibration unmitigated and mitigated peak mitigation 73–88 constant 66 management scenarios 330 retention of nutrients and water f ows 218 m o d e l i n g f ows relevant to sediment 89–109 ] water retention, and streamf ow nutrient and sediment verif cation, Tier 2 model 90 recession 70 retention 89–106 vertebrate diversity, range water supply as ES 53–92 , 106 nature-based tourism 191–5 , shifts 332–5 water yield 65–9 200–1 virtue-based values 208 , 213 calculation 63 policies to provide carbon visitation model 194–203 Tier 1 model 54–9 sequestration 37 , 45 camping visit rates 193 , 195 vs run of river hydropower policy trajectory 46 and congestion 200 , 203 production 65 scenarios infrastructure variables 194–6 see also f ood risk ; irrigation alternative 252 ordinary least squares regression wetlands in USA climate simulation (mid-21C) model 201 Borden Ranch case 27 326 Voluntary Carbon Standard mitigation banking 347 conservation 254 (VCS) 346 swampbuster clause 27 land-use change for 2050 voluntary markets 345–6 value of ES 27–8 253 INDEX 365 state parks and landscape features, water availability for marginal and total values 21 , 22 characteristics, and agriculture 327–8 World Climate Research Programme attributes 192 water treatment, pollutant (WCRP), CMIP3 326–7 storm peak mitigation 218 , loading 91 , 101 , 103 , 105 , 254 , 255 108 yield functions 154–5 Tier 1 and 2 use value 197 willingness-to-pay (WTP) 17 , 20 agricultural ES 154–5 , 324 Tier 2 visitation model 194 , conservation programs 220–1 Yosemite, travel cost method of 200–2 framing effects 224 valuation 23 This page intentionally left blank (a) (b)

North Fork North Fork

Value of Land for Hydropower Production ($) Change in Water Yield (mm) 21305 105 Detroit Detroit

1482 0 Dam Watersheds Dams Dam Watersheds Green Peter Dams Green Peter

Fall Creek Fall Creek Lookout Point Lookout Point

01020Kilometers 01020Kilometers

0 Detroit

–1106 Dam Watersheds Dams Green Peter

Fall Creek Lookout Point

01020Kilometers

Plate 1 Hypothetical example application of tier 1 model of water provisioning for hydropower generation in the Willamette river watershed. The example evaluates f ve sub-catchments of hydropower stations at North Fork (41MW), Detroit (115MW), Green Peter (92MW), Fall Creek (6.4MW), and Lookout Point (138MW). (a) The net present value of landscape water provision services for hydropower; (b) changes in water yield as a result of hypothetical deforestation of all land below 1000 m above sea level; and (c) the changes in landscape value for hydropower under the deforestation scenario. (See Figure 4.1.)

Normalized SWAT and tier 1 Phosphorous export 3.5

3 tier 1 = 0.621 × SWAT 2 2.5 R = 0.385

1.5

0.5

–0.5

Normalized tier 1 Phosphorous export Correlation Groups –1 1 2 –1.5 3 N –2 –1 0 1234 4 Normalized SWAT Phosphorous export 5 6 7 Group 1 Group 2 Group 3 Group 4 Group 5 Group 6 Group 7

Plate 2 Aggregated sub-catchment phosphorous export comparison between our model and SWAT (graph) and agreement of spatial phosphorous export patterns predicted by the two models (map) in the Williamette Valley, Oregon (USA). The graph on the left depicts a correlation between the normalized tier 1 model outputs and the normalized SWAT outputs. The groupings in the graph depict sub-catchments whose tier 1 outputs lie within a threshold distance of SWAT outputs given the correlation between normalized tier 1 and SWAT outputs. Note that Groups 4 and 6 represent sub-catchments in which tier 1 outputs are unexpectedly low and Groups 5 and 7 represent sub-catchments in which tier 1 outputs are unexpectedly high. The map illustrates the sub-catchment groups identif ed in the graph. (See Figure 6.8.) K e n y a T a n z a n i a

Dar Es Salaam Lower bound

Carbon stock (Mg / ha)

0–88 150–313 783–1,122 Upper bound 88–150 313–783

Plate 3 Tier 1 carbon storage estimates for 1995 in Tanzania’s Eastern Arcs Mountains and their watersheds. The polygons formed with the green lines represent Eastern Arc Mountain blocks, which rise from the surrounding woodlands and savannas. These blocks were once largely forested, but now consist of a mixture of agriculture, forest, and woodlands. Blue lines are major rivers. Black squares represent major cities. Timber plantations cover approximately 0.3% of the study landscape. Spatially explicit land cover and other landscape data are from the Valuing the Arc project (2008; Mwakalila 2009). See the chapter’s SOM for details on data used in the maps. (See Figure 7.2.)

(a) 22 km Plate 4 The value of carbon sequestered in soil across two alternative LULC scenarios. (a) The year 2000 landscape. (b) The per-hectare monetary value of carbon sequestration in soil from 2000 to 2050 for each LULC scenario. The top row of maps gives mean results across all model simulations. The bottom rows of maps give the results from one particular run of the model. The black outlines on the parcels indicate parcels that experience LULC change in some portion of its 20 Km area at some point between 2000 and 2050. The Carbon Row crops Oak Sequestration Scenario map ref ects a program of Alfalfa Maple/basswood Grass, pasture, Aspen/White Birch afforestation, restoring prairie pothole, and converting row and range (inc. CRP) Lowland Shrub deciduous crops to pasture and perennial grassland. In the CRP Loss Wetlands/Marsh Water Scenario any parcel that was primarily in CRP in 2000 was Roads/urban/ barren converted to row crops or a hayf eld by 2050. (See Figure 7.4.) (b) Carbon Sequestration Scenario CRP Loss Scenario Mean Values

Present value per hectare < 0 0 1–441 442–602 603–928

Sample Draw > 928 (a) Land Cover (b) Nest Suitability

Ditchside edge High: 1 Farm parcel edge Residential edge Low: 0 Roadside edge Agriculture Organic Ag Pasture Riparian Scrub Forest Unclassified Water

High: 1 High: 1

Low: 0 Low: 0

(e) Farm Abundance (f) Service Value Map

High: 0.25 High: 1

Low: 0 Low: 0

W E Kilometers 04.5 9 18 27 36 S

Plate 5 Example results of pollination model for watermelon in Yolo County, California. The model uses (a) land cover data as input and derives maps of (b) nesting habitat and (c) f oral resources. From this, it generates (d) a pollinator supply map that describes an index of pollinator abundance on the landscape. Based on the supply map, the model generates (e) a pollinator abundance map on farm parcels (i.e., “farm abundance”). After using a simple yield function to translate farm abundance into relative yield the model distributes yield or economic value back onto the surrounding landscape to generate (f) the value map. All steps are the same for tier 1 and tier 2 models; results here are tier 2, based on data supplied in supplemental online appendix. (See Figure 10.2.)

Current Conservation Growth Plate 6 Maps of parcel habitat quality scores when the “Roads” and “Urban” parameter Landscape LULC Scenario LULC Scenario combinations are used in the Sierra Nevada illustrative example. We ran the tier 1 model on a grid map with a cellular resolution of 400 m × 400 m (16-ha grid cells). In these maps we present the mean habitat quality score ( Q ) of all grid cells within 500 hectare hexagons, our parcels in this illustrative example. There are 23 042 500-ha hexagons in the Sierra Nevada Conservancy. In both future LULC scenarios the majority of residential development is centered on Sacramento, and generally along the western foothills. In the Growth scenario, montane hardwood is the land cover type that loses the most area to development (158 268 ha). In the Conservation scenario, annual grassland is the land cover type that loses the most area (17 798 ha). See the chapter’s SOM for all tier 1 model details.

Roads parameter combination (See Figure 13.2.) Urban parameter combination

Parcel Q scores

>0– 0.91– 0.93– 0.95– 0.97– 0.99– <0.9 0.92 0.94 0.96 0.98 1.00 MBV RMBV ratios (a) (b) Current Conservation Growth Landscape LULC Scenario LULC Scenario

2–19 0–1.74 20–47 1.75–3.99

Minimum C 48–272 4.00–6.96 273–325 6.97–13.49 326–333 13.50–15.09 334–456 15.10–16.54 457–1,045 16.55–18.50 (c) (d)

27.60

Maximum C 0 2.21

(e) (f)

Federally Threatened (FT) Herpetofauna MBV 0 1 RMBV ratios 0 1

0 1–229 Plate 7 The spatial distribution of MBV and 0 230–291 25 292–412 413–577 RMBV ratio scores for federally threatened 578–766 767–975 Bolivia herpetofauna (FT) using minimum and N Brazil

20 0 20 40 60 Kilometers maximum species-habitat suitability scores in WE Paraguay the Sierra Nevada illustrative example. S Argentina The Growth scenario creates a much greater loss in FT subgroup species effective habitat area in Plate 8 Net present values in US$ha-1 for selected ecosystem services in the the foothills of the Sierra Nevada than the Mbaracyau Forest Biosphere Reserve, Paraguay. Conservation scenario does. See the chapter’s (a) Sum of all f ve services; (b) sustainable bushmeat harvest; (c) sustainable timber SOM for all tier 2 model details. (See Figure 13.4.) harvest; (d) bioprospecting; (e) existence value; and (f) carbon storage. (See Figure 14.2.) Source: Naidoo and Ricketts (2006).

LULC Scenarios Land Use/Land Cover Classes Undefined Native Shrubland/Sparse Ohia Uluhe Shrubland Closed Ohia Forest Open Koa-Ohia Forest Open Ohia Forest Water Island of O¢ ahu Wetland Vegetation Sugarcane Ethanol Agriculture Alien Shrubs and Grasses Alien Trees and Shrubs High Intensiyt Developed Low Intensiyt Developed Alien Grassland Alien Shrubland Christmas Berry Shrubland Koa Haole Shrubland Alien Forest Baseline Diversified Agriculture Closed Kukui Forest LULC Map & Forestry Kiawe Forest and Shrubland Kiawe-Koa Haole Forest and Shrubland Uncharacterized Forest Uncharacterized shrubland Very Sparse Vegetation to Unvegetated Koa Reforestation Field Buffer Sugarcane (irrigated) Sugarcane (unirrigated)

Residential Subdivision

Plate 9 Land use/land cover maps on the north shore of O’ahu. The area shown here includes all of Kamehameha Schools’ north shore land holdings, as well as small adjacent parcels that make for a continuous region. The baseline map is from the Hawai’i Gap Analysis Program’s land cover layer for O`ahu (Hawai’i Gap Analysis Program 2006). (See Figure 14.6.) 257 H G F E

Agriculture Rural-Residential 246 Managed Forestry Conserved C UGB B I

235 Expected Number of Species A

224 0 5 10 15 20 25 30 Billions of Dollars

Plate 10 Eff ciency frontier showing maximum feasible combinations of economic returns and biodiversity scores. Land-use patterns associated with specif c points along the eff ciency frontier (points A–H) and the current landscape (point I). (See Figure 14.8.) Source: Polasky et al. (2008).

(a) (c)

0.8 0.0 7.7 19.3 17.5 95.3

(b) (d)

Plate 11 Poverty indicators and representative forest product harvest distributions in the Amazon Basin. The incidence of underweight children is highest in northern Peru and eastern Ecuador (a) while unsatisf ed basic needs are highest in Bolivia (c). High poverty areas def ned as those above the 75th percentile for underweight children (outlined in dark black) are shown in a direct pairing, overlaid with the harvest index of fruits and nuts for subsistence (b). High poverty areas def ned as those above the 75th percentile for unsatisf ed basic needs are shown in an indirect pairing, overlaid with the harvest index of wood for market sale (d). Units for underweight children are percentage of the population under the age of 5 that is underweight. Units for unsatisf ed basic needs are the percentage of the population with unsatisf ed basic needs. The legends and units for (b) and (d) are the same as those in Figure 16.1c and h. (See Figure 16.2.) N POVERTY RATE (percent of population below poverty line) M M E R U T. K E Y A M E R U N N C E NT R A L .P. > 65 M E R U T H A R A K A S O U T H 55–65 N Y E R I E M B U 45–55 KI MURANGA RINYAGA 35–45 M W I N G I <= 35 ABERDARE RANGE MARAGUA M B E E R E Tana River OTHER FEATURES Upper Tana boundary K I A M A B U Masingo Reservoir T H I K A District boundaries Major national parks and reserves (over 5,000 ha) WATER BODIES AND RIVERS M A C H A K O S KITUI Permanent rivers NAIROBI Water bodies

Plate 12 Map of the Tana River headwaters in Kenya, and the distribution of poor communities. (See Figure 16.B.1.)

49°0’N N Legend Land

Suitable Areas for Eelgrass

02040 Kilometers

48°30’N

48°0’N

47°30’N Mapped Eelgrass-EFH Potential Eelgrass Habitat

Study Area

47°0’N 124°30’W 124°0’W 123°30’W 123°0’W 122°30’W

Plate 13 A map of Puget Sound showing areas our model predicts suitable for eelgrass beds (green). Inset maps show higher detail; orange represents currently mapped eelgrass from the NOAA Essential Fish Habitat data (TerraLogic GIS Inc. 2004). (See Figure 17.1.) (a)

Whidbey Island

Strait of Juan de Fuca

Harvest (1998–2007, Ibs) Farmed and Wild

11,000,000 Hood Canal North Central Puget Sound

Crab South Central Puget Sound Groundfish Salmon Shellfish South Puget Sound Other species

(b)

Whidbey Island

Strait of Juan de Fuca

North Central Puget Sound Harvest (1998–2007, $) Farmed and Wild $13,000,000 Hood Canal

Crab South Central Puget Sound Groundfish Salmon Shellfish Other species

South Puget Sound

Plate 14 A map of the Puget Sound Partnership’s action areas showing the distribution of (a) landings (in UK£) and (b) revenue (in US$) of farmed and wild seafood from 1998 to 2007. (See Figure 17.2.) (a)

5.5 Transient orcas

Resident orcas 5.0

Porpoises Sea lions 4.5 Harbor seals Adult lingcod Diving birds Six gill shark Dogfish Juv. lingcod Chinook salmon Coho salmon 4.0 Large rockfish Skates

1.0 Detritus Eelgrass Understory kelp Overstory kelp Benthic algae Phytoplankton

(b)

5.5 Transient orcas

Resident orcas 5.0

Porpoises Sea lions 4.5 Adult lingcod Harbor seals Dogfish Juv. lingcod Six gill shark Chinook salmon Large rockfish Diving birds 4.0 Skates Adult herring

Pacific hake Chum salmon Pisc. flatfish Demersal fish Cancer crab Gulls

3.5 Squid Trophic Level Juv. herring Lg. jellies Ratfish 3.0 Ducks + brants

2.5 Sm. crustaceans

Eelgrass

Plate 15 (a) The structure of the EwE food web model of the Central Basin of Puget Sound (without f sheries) and (b) a subset of the EwE food web model focusing on eelgrass and herring. (a) Box size is proportional to standing stock biomass; line thickness is proportional to the f ow of energy/material from the prey to the predator. Red colors represent detritus and the portion of the food web it supports, blues are benthic primary producers and those they support, and greens are phytoplankton and phytoplankton- supported groups. Consumers’ colors are a mix proportional to the amount of production that ultimately stems from those sources. In (b) dashed arrows indicate groups whose predation on herring eggs is mediated by the biomass of eelgrass. Colors are as those in (a). (See Figure 17.3.)

Plate 16 Change in annual average water availability (yield) between present day and mid-century for the Willamette Basin. (See Figure 18.2.)

Difference (mm)

High:413

Low:–60

Berry/grape-producing pixels SUPPLEMENTARY TABLES OF LONSDORF ET AL: CROP POLLINATION SERVICES (CHAPTER 10 IN KAREIVA ET AL: NATURAL CAPITAL (2011) © OXFORD UNIVERSITY PRESS.)

Supplementary Table 10.S1 Floral resource and nesting suitability values for land-use land cover in Costa Rica.

Land cover Nesting (Nij )

F N 1 N 2 Forest 1 1 1 Coffee 0.5 0.2 0.1 Cane 0 0 0 Pasture/grass 0.2 0.2 0.1 Scrub 0.3 0.3 0.2 Bare 0.1 0 0.1 Built-up 0.3 0.2 0.1

Supplementary Table 10.S2 Species foraging distances and nesting suitability values for Costa Rica

Pollinator species/guild Foraging distance N s1 N s2

(α s ; meters) Apis mellifera 663 1 0 Huge Black 2002* 214 0 1 Melipona fasciata 578 0 1 Nannotrigona mellaria 70 0 1 Partamona cupira/Trigona fussipennis/Trigona corvina 87 0 1 Plebia jatiformis 28 0 1 Plebia frontalis 34 0 1 Trigona (Tetragona) clavipes 55 0 1 Trigona (tetragonisca) angustula 22 0 1 Trigona dorsalis 60 0 1 Trigona fulviventris 77 0 1 Trigonisca sp. 663 0 1

* Unidentif ed species

1 SUPPLEMENTARY TABLES OF LONSDORF ET AL: CROP POLLINATION SERVICES (CHAPTER 10 IN KAREIVA ET AL: NATURAL CAPITAL (2011) © OXFORD UNIVERSITY PRESS.)

Supplementary Table 10.S3 Nesting suitability and f oral resource values for land cover in California.

Land cover Nesting type ( N ij ) Floral resources (F kj ) Ground Wood Stem Cavity Spring Summer Fall Water 0 0 0 0 0 0 0 Conventional agriculture 0.25 0 0 0.05 0.05 0.15 0 Organic agriculture 0.5 0.5 0.5 0.5 0.75 0.5 0.25 Pasture 1 0 0.25 1 0.75 0.5 0.1 Scrub forest 1 1 0.5 1 0.75 0.25 0.25 Riparian 1 1 1 1 1 0.25 0 Roadside edge* 1 0.1 0 0.5 0.75 0.75 0 Residential-suburban edge* 0.1 0.5 0.5 0.5 0.5 0.5 0.5 Ditchside edge* 1 0.25 1 1 0.5 0.5 0 Farm parcel edge* 1 0 0 0.5 0 0.25 0

*Only included in tier 2 analysis.

Supplementary Table 10.S4 Species dispersal and species nesting suitability estimates for nest types in California.

Pollinators Foraging Nesting type (N si ) Flight Season (F sk ) distance Ground Wood Stem Cavity Spring Summer Fall (αs ; meters) Agapostemon texanus 195 1 0 0 0 0.5 0.5 0 Anthophora urbana 1273 1 0 0 0 0 1 0 Apis mellifera 1091 0 0 0 1 0.4 0.4 0.2 Ashmeadiella aridula 36 0 1 1 0 0.5 0.5 0 Bombus californicus 1715 0 0 0 1 0.5 0.5 0 Bombus vosnesenskii 1035 0 0 0 1 0.4 0.4 0.2 Ceratina nanula 30 0 0 1 0 0.5 0.5 0 Diadasia enavata 724 1 0 0 0 0 0.67 0.33 Dialictus sp. 14 1 0 0 0 0.5 0.5 0 Evylaeus sp. 57 1 0 0 0 0.4 0.4 0.2 Halictus farinosus 436 1 0 0 0 0.4 0.4 0.2 Halictus ligatus 96 1 0 0 0 0.4 0.4 0.2 Halictus tripartitus 53 1 0 0 0 0.4 0.4 0.2 Hylaeus 20 0 1 0 0 0.5 0.5 0 Lasioglossum 159 1 0 0 0 0.5 0.5 0 Megachile 541 0 1 1 0 0.5 0.5 0 Melissodes 448 1 0 0 0 0 0.67 0.33 Osmia regulina 176 0 1 0 0 0.5 0.5 0 Peponapis pruinosa 1049 1 0 0 0 0 0.67 0.33