sign in sign up
<<
Home , Aquatic ecosystem, Ecological extinction, Ecology, Ecosystem, Lake ecosystem, Marine food web

Aquatic Webs This page intentionally left blank Aquatic Food Webs An Approach

EDITED BY Andrea Belgrano National Center for Genome (NCGR), Santa Fe, NM, USA Ursula M. Scharler University of Maryland Center for , Chesapeake Biological Laboratory (CBL), Solomons, MD, USA and Smithsonian Environmental Research Center, Edgewater, MD, USA Jennifer Dunne Pacific Ecoinformatics and Computational Lab, Berkeley, CA USA; Santa Fe Institute (SFI) Santa FE, NM, USA; Rocky Biological Laboratory, Crested Butte, CO USA

AND Robert E. Ulanowicz University of Maryland Center for Environmental Science, Chesapeake Biological Laboratory (CBL), Solomons, MD, 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 Dare es Salaam Hong Kong Karachi Kuala Lumpur Madrid Melbourne Nairobi New Delhi Shanghai Taipei Toronto With offices in Austria Brazil Chile Czech Republic France Greece Guatemala Hungary Italy Japan South Korea Poland Portugal Singapore Switzerland Thailand 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, 2005 The moral rights of the authors have been asserted right Oxford University Press (maker) First published 2005 All rights reserved. No part of this publication may be reproduced, stored in a retrieval , 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 . Enquiries concerning 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 this same condition on any acquirer British Library Cataloging in Publication Data (Data available) Library of Congress Cataloging-in-Publication Data Aquatic food webs : an ecosystem approach / edited by Andrea Belgrano ... [et al.]. p. cm. Includes bibliographical references and index. ISBN 0-19-856482-1 (alk. paper) – ISBN 0-19-856483-X (alk. paper) 1. Aquatic ecology. 2. Food chains (Ecology) I. Belgrano, Andrea. QH541.5.W3A68225 2004 577.6016–dc22 2004027135 ISBN 0 19 856482 1 (Hbk) 9780198564829 ISBN 0 19 856483 X (Pbk) 9780198564836 10987654321 Typeset by Newgen Imaging (P) Ltd., Chennai, India Printed in Great Britain on acid-free paper by Antony Rowe, Chippenham FOREWORD CURRENT AND FUTURE PERSPECTIVES ON FOOD WEBS

Michel Loreau

Food webs have been approached from two basic on food webs that is complementary to the energetic perspectives in ecology. First is the energetic view approach (DeAngelis 1992). Material cycles are articulated by Lindeman (1942), and developed by among the most common of the positive during the following decades. loops discussed by Ulanowicz in his concluding In this view, food webs are networks of pathways remarks, and may explain key properties of eco- for the flow of in , from its systems (Loreau 1998). The stoichiometry of ecolo- capture by in the process of photo- gical interactions may further strongly constrain synthesis to its ultimate dissipation by hetero- food-web (Sterner and Elser 2002; Elser trophic . I would venture to say that the and Hessen’s chapter). There has also been con- analysis advocated by siderable interest in the relationship between bio- Ulanowicz and colleagues in this book is heir to diversity and ecosystem functioning during the last this tradition. A different approach, rooted in decade (Loreau et al. 2002). Merging the theories that ecology, was initiated by May (1973) bear upon food webs and the maintenance of and pursued by Pimm (1982) and others. This diversity is urgently needed today, and may provide approach focuses on the dynamical constraints new insights into food-webs structure and ecosys- that arise from species interactions, and empha- tem functioning (Hillebrand and Shurin’s chapter). sises the fact that too much interaction (whether in The structure and functioning of ecological sys- the form of a larger number of species, a greater tems is determined not only by local constraints connectance among these species, or a higher and interactions, but also by larger-scale processes. mean interaction strength) destabilises food webs The importance of regional and historical influ- and ecological systems. The predictions resulting ences has been increasingly recognised in com- from this theory regarding the diversity and con- munity ecology (Ricklefs and Schluter 1993). The nectance of ecological systems led to a wave of extent to which they shape food webs, however, comparative topological studies on the structure of has been relatively little explored. The recent food webs. Thus, the two traditions converge in development of theory (Leibold the search for patterns in food-web structure et al. 2004) provides a framework to start exam- despite different starting points. This book results ining spatial constraints on the structure and from the confluence of these two perspectives, functioning of webs (Melian et al.‘s which are discussed in a number of chapters. chapter). At even larger scales, food webs are Patterns, however, are generally insufficient to the result of evolutionary processes which deter- infer processes. Thus, the search for explanations of mine their current properties. Complex food webs these patterns in terms of processes is still very much may readily evolve based on simple ecological alive, and in this search the energetic and dynamical interactions (McKane 2004). The of food- perspectives are not the only possible ones. Bio- web and ecosystem properties is a fascinating geochemical cycles provide a functional perspective topic for future research.

v vi FOREWORD

This book provides a good synthesis of recent Loreau, M. 1998. Ecosystem development explained by research into aquatic food webs. I hope this within and between material cycles. Pro- synthesis will stimulate the development of new ceedings of the Royal of London, Series B 265: 33–38. approaches that link and ecosystems. Loreau, M., S. Naeem, and P. Inchausti. Eds. 2002. Bio- diversity and ecosystem functioning: synthesis and per- spectives. Oxford University Press, Oxford. References May, R. M. 1973. Stability and in model ecosys- tems. Princeton University Press, Princeton. DeAngelis, D. L. 1992. Dynamics of cycling and McKane, A. J. 2004. Evolving complex food webs. The food webs. Chapman & Hall, London. European Physical Journal B 38: 287–295. Leibold, M. A., M. Holyoak, N. Mouquet, P. Amarasekare, Pimm, S. L. 1982. Food webs. Chapman & Hall, London. J. M. Chase, M. F. Hoopes, R. D. Holt, J. B. Shurin, R. Law, Ricklefs, R. E., and D. Schluter. Eds 1993. D. Tilman, M. Loreau, and A. Gonzalez. 2004. The in ecological communities: historical and geographical per- metacommunity concept: a framework for multi-scale spectives. University of Chicago Press, Chicago. community ecology. Ecology Letters 7: 601–613. Sterner, R. W., and J. J. Elser. 2002. : Lindeman, R. L. 1942. The trophic-dynamic aspect of the of elements from to the . ecology. Ecology 23: 399–418. Princeton University Press, Princeton. Contents

Foreword v Michel Loreau

Contributors ix

Introduction 1 Andrea Belgrano

PART I Structure and 5

1 Biosimplicity via stoichiometry: the evolution of food-web structure and processes 7 James J. Elser and Dag O. Hessen

2 Spatial structure and dynamics in a marine 19 Carlos J. Melia´n, Jordi Bascompte, and Pedro Jordano

3 Role of network analysis in comparative ecosystem ecology of 25 Robert R. Christian, Daniel Baird, Joseph Luczkovich, Jeffrey C. Johnson, Ursula M. Scharler, and Robert E. Ulanowicz

4 Food webs in —seasonal dynamics and the impact of variability 41 Dietmar Straile

5 Pattern and process in food webs: evidence from running waters 51 Guy Woodward, Ross Thompson, Colin R. Townsend, and Alan G. Hildrew

PART II Examining food-web theories 67

6 Some random thoughts on the statistical analysis of food-web data 69 Andrew R. Solow

7 Analysis of size and complexity of randomly constructed food webs by information theoretic metrics 73 James T. Morris, Robert R. Christian, and Robert E. Ulanowicz

8 Size-based analyses of aquatic food webs 86 Simon Jennings

vii viii CONTENTS

9 Food-web theory in marine ecosystems 98 Jason S. Link, William T. Stockhausen, and Elizabeth T. Methratta

PART III Stability and diversity in food webs 115

10 Modeling food-web dynamics: complexity–stability implications 117 Jennifer A. Dunne, Ulrich Brose, Richard J. Williams, and Neo D. Martinez

11 Is maintained by food-web complexity?—the adaptive food-web hypothesis 130 Michio Kondoh

12 Climate forcing, food web structure, and community dynamics in pelagic marine ecosystems 143 L. Ciannelli, D. Ø. Hjermann, P. Lehodey, G. Ottersen, J. T. Duffy-Anderson, and N. C. Stenseth

13 Food-web theory provides guidelines for 170 Enric Sala and George Sugihara

14 Biodiversity and aquatic food webs 184 Helmut Hillebrand and Jonathan B. Shurin

PART IV Concluding remarks 199

15 Ecological network analysis: an escape from the machine 201 Robert E. Ulanowicz

Afterword 208 Mathew A. Leibold

References 211

Index 255 Contributors

Daniel Baird, Department, University of Port Simon Jennings, Centre for Environment, Elizabeth, Port Elizabeth, South . and Science, Lowestoft Laboratory NR33 Jordi Bascompte, Integrative Ecology Group, Estacio´n 0HT, UK. Email: [email protected] Biolo´gica de Don˜ana, CSIC, Apdo. 1056, E-41080, Jeffrey C. Johnson, Institute of Coastal and Marine Sevilla, Spain. Email: [email protected] Resources, East Carolina University, Greenville, NC Andrea Belgrano, National Center for Genome Resources 27858, USA. (NCGR), 2935 Rodeo Park Drive East, Santa Fe, NM Pedro Jordano, Integrative Ecology Group, Estacio´n 87505, USA. Email: [email protected] Biolo´gica de Don˜ana, CSIC, Apdo. 1056, E-41080, Ulrich Brose, Technical University of Darmstadt, Sevilla, Spain. Department of Biology, Schnittspahnstr. 3, 64287 Michio Kondoh, Center for , Darmstadt, Germany. Institute of Ecology, Rijksstraatweg 6, Nieuwersluis, Robert R. Christian, Biology Department, East Carolina P.O. Box 1299, 3600 BG Maarssen, The Netherlands. University, Greenville, NC 27858, USA. Email: Email: [email protected] [email protected] P. Lehodey, Oceanic Fisheries Programme, Secretariat Lorenzo Ciannelli, Centre for Ecological and of the Pacific Community, BP D5, 98848 Noumea Evolutionary Synthesis (CEES), Department of Biology cedex, New Caledonia. University of Oslo, Post Office Box 1066, Blindern, Mathew Leibold, Section of Integrative Biology, The N-0316 Oslo, Norway. Email: lorenzo.ciannelli@ University of Texas at Austin, 1 University Station, bio.uio.no. C0930 Austin, TX 78712, USA. Email: mleibold@ Janet T. Duffy-Anderson, Alaska mail.utexas.edu Center, NOAA, 7600 Sand Point Way NE, 98115 Jason S. Link, National Marine Fisheries Service, Seattle, WA, USA. Northeast Fisheries Science Center, 166 Water St., Jennifer A. Dunne, Pacific Ecoinformatics and Computa- Woods Hole, MA 02543, USA. Email: jlink@ tional Ecology Lab, P.O. Box 10106, Berkeley, CA 94709 whsunl.wh.whoi.edu USA; Santa Fe Institute, 1399 Hyde Park Road, Santa Fe, Michel Loreau, Laboratoire d’Ecologie, UMR 7625 NM 87501 USA; Rocky Mountain Biological Laboratory, Ecole Normale Superieure 46, rue d’ Ulm F-75230, P.O. Box 519, Crested Butte, CO 81224 USA Email: Paris Cedex 05, France. Email: [email protected] [email protected]. Joseph Luczkovich, Biology Department, East Carolina James J. Elser, School of Sciences, Arizona State University, Greenville, NC 27858, USA. University,Tempe,AZ85287,USA.Email:j.elser@ Neo D. Martinez, Pacific Ecoinformatics and asu.edu Computational Ecology Lab, P.O. Box 10106, Dag O. Hessen, Department of Biology, University of Berkeley, CA 94709 Rocky Mountain Biological Oslo, P.O. Box 1050, Blindern, N-0316 Oslo, Norway. Laboratory, P.O. Box 519, Crested Butte, CO 81224 Alan G. Hildrew, School of Biological Sciences, Queen USA. Mary, University of London, Mile End Road, London, Carlos J. Melia´n, Integrative Ecology Group, Estacio´n E1 4NS, UK. Email: [email protected] Biolo´gica de Don˜ana, CSIC, Apdo. 1056, E-41080, Helmut Hillebrand, Institute for , University of Sevilla, Spain. Cologne, Gyrhofstrasse 15 D-50931 Ko¨ln, Germany. Elizabeth T. Methratta, National Marine Fisheries Email: [email protected] Service, Northeast Fisheries Science Center, D.Ø. Hjermann, Centre for Ecological and Evolutionary 166 Water St., Woods Hole, MA 02543, USA. Synthesis (CEES), Department of Biology University of James T. Morris, Department of Biological Sciences, Oslo, Post Office Box 1066 Blindern, N-0316 Oslo, University of South Carolina, Columbia, SC 29208, USA. Norway. Email: [email protected]

ix x CONTRIBUTORS

Geir Ottersen, Institute of Marine Research, P.O. Box William T. Stockhausen, National Marine Fisheries 1870 Nordnes, 5817 Bergen, NORWAY Current Service, Northeast Fisheries Science Center, Address: Centre for Ecological and Evolutionary 166 Water St., Woods Hole, MA 02543, USA. Synthesis, Department of Biology, P.O Box 1050 Andrew R. Solow, Woods Hole Oceanographic Blindern, N-0316 Oslo, NORWAY. Institution, Woods Hole, MA 02543, USA. Email: Enric Sala, Center for Marine Biodiversity and [email protected] Conservation, Scripps Institution of , Geroge Sugihara, Center for Marine Biodiversity and La Jolla, CA 92093-0202, USA. Email: Conservation, Scripps Institution of Oceanography, [email protected] La Jolla, CA 92093-0202, USA. Ursula M. Scharler, University of Maryland, Ross Thompson, Biodiversity Research Centre, Center for Environmental Science, Chesapeake University of British Columbia, Vancouver, Canada. Biological Laboratory (CBL), Solomons, MD 20688, Colin R. Townsend, Department of Zoology, USA; Smithsonian Environmental Research University of Otago, New Zealand. Center, Edgewater, MD, USA. Email: scharler@ Robert E. Ulanowicz, University of Maryland, Center cbl.umces.edu for Environmental Science, Chesapeake Biological Jonathan B. Shurin, Department of Zoology, Laboratory (CBL), Solomons, MD 20688, USA. Email: University of British Columbia, 6270 University Blvd. [email protected] Vancouver, BC V6T 1Z4, Canada. Richard J. Williams, Pacific Ecoinformatics and Nils Chr. Stenseth, Centre for Ecological and Computational Ecology Lab, P.O. Box 10106, Berkeley, Evolutionary Synthesis (CEES), Department of Biology CA 94709 USA; Rocky Mountain Biological Laboratory, University of Oslo, Post Office Box 1066 Blindern, N- P.O. Box 519, Crested Butte, CO 81224 USA; San Fran- 0316 Oslo, Norway. cisco State University, Computer Science Department, Dietmar Straile, Dietmar StraileLimnological 1600 Holloway Avenue, , CA 94132 USA. Institute, University of Konstanz, 78457 Konstanz, Guy Woodward, Department of Zoology, Ecology Germany. Email: dietmar.straile@ and Science, University College, Cork, uni-konstanz.de Ireland. INTRODUCTION Aquatic food-webs’ ecology: old and new challenges

Andrea Belgrano

Looking up ‘‘aquatic food web’’ on Google provides that linked latitudinal gradients in aquatic species a dizzying array of eclectic sites and information diversity, food-web complexity, and community (and disinformation!) to choose from. However, stability. even within this morass it is clear that aquatic Following that early MacArthur hypothesis, we food-web research has expanded greatly over the find it timely to also ask, How complex are aquatic last couple of decades, and includes a wide array food webs? of studies from both theoretical and empirical The first book on theoretical food-web ecology perspectives. This book attempts to bring together was written by May (1973), followed by Cohen and synthesize some of the most recent perspec- (1978). Since then, Pimm (1982) and Polis and tives on aquatic food-web research, with a parti- Winemiller (1996) have revisited some of the ideas cular emphasis on integrating that knowledge proposed by May and Cohen and discussed them within an ecosystem framework. in different contexts, and trophic flow models have It is interesting to look back at the pioneering been proposed and used widely for aquatic and work of Sir Alister Hardy in the early 1920s at particularly marine ecosystems (e.g. Wulff et al. Lowestoft Fisheries Laboratory. Hardy studied the 1989; Christensen and Pauly 1993). However, feeding relationship of the North herring with recent advances in ecosystem network analysis planktonic assemblages by looking at the species (e.g. Ulanowicz 1996, 1997; Ulanowicz and Abarca- distribution patterns in an attempt to provide Arenas 1997) and the network structure of food better insights for the of the webs (e.g. Williams and Martinez 2000; Dunne North Sea fisheries. If we take a look in his food- et al. 2002a,b; Williams et al. 2002) in relation to web scheme (Figure 1), it is interesting to note that ecosystem dynamics, function, and stability clearly he considered species diversity in both phyto- set the path for a new, complementary research and , and also specified agenda in food-web analysis. These and many body-size data for the different in the other studies suggest that a new synthesis of food web. Thus, it appears that already almost available information is necessary. This new 100 years ago the concept of constructing and synthesis is giving rise to novel that drawing links among diverse species at multiple generalizes across and scales, for example, trophic levels in a network-like fashion was in the the discovery of universal scaling relations in food- mind of many aquatic researchers. web structure (Garlaschelli et al. 2003), and is also In following decades, researchers began to underpinning new approaches and priorities for consider links between food-web complexity and whole-ecosystem conservation and management, ecological community stability. The classic, and still particularly in marine systems. contentious MacArthur hypothesis that ‘‘Stability Aquatic food-web research is also moving beyond increases as the number of link increase’’ (1955) an exclusive focus on taxa from gave rise to studies such as that by Paine (1966) to fish. A new look at the role that marine microbes

1 2 INTRODUCTION

Figure 1 The food web of herring Clupea harengus Hardy (1924). From Parables of Sea & Sky—The life, work and art of Sir Alister Hardy F. R. S. Courtesy of SAHFOS—The CPR Survey, Plymouth, UK.

Figure 2 The : impressionist version. A -eye view of the ’s euphotic layer. is an organic continuum, a gel of tangled polymers with embedded strings, sheets, and bundles of fibrils and , including living organisms, as ‘‘hotspots.’’ Bacteria (red) acting on marine (black) or (green) can control sedimentation and primary ; diverse microniches (hotspots) can support high bacterial diversity. (Azam, F. 1998. Microbial control of oceanic flux: the plot thickens. Science 280: 694–696.) (See Plate1) INTRODUCTION 3 play in the global ocean (Azam and Worden 2004) including humans. Aquatic food webs may pro- suggests that oceanic ecosystems can be character- vide a particularly useful empirical framework for ized as a complex dynamic molecular network. developing and testing an of The role of microbial food webs (Figure 2—see ecology that will take into account the complex also, Plate 1—Azam 1998) needs to be considered network of interactions among biotic and abiotic to understand the nonlinearities underlying the components of ecosystems. relationship between the pelagic and benthic domains. Acknowledgments Emerging challenges in aquatic food-web research include integrating genomic, biogeochemical, This work was funded in part or in full by the US environmental, and economic data in a modeling Dept of Energy’s Genomes to Life program (www. effort that will elucidate the mechanisms govern- doegenomestolife.org) under the project ‘‘Carbon ing the ecosystem dynamics across temporal and Sequestration in Synechococcus sp.: From Mole- spatial scales at different levels of organization cular Machines to Hierarchical Modeling’’ (www. and across the whole variety of species diversity, genomes-to-life.org). This page intentionally left blank PART I Structure and function

Many use food webs to portray ecolo- of trophic interactions. Along those gical communities as complex adaptive systems. lines, several of the chapters extend their scope However, as with other types of apparently com- beyond traditional food-web ‘‘snapshot’’ analyses plex systems, underlying mechanisms regulate to take into account and time when assessing food-web function and can give rise to observed changes in food-web structure and species structure and dynamics. These mechanisms can composition. sometimes be summarized by relatively simple By comparing food webs from different envir- rules that generate the ecosystem properties that onments and by encompassing organisms from we observe. bacteria to vertebrates, we start to see some com- This section of the book presents and discusses mon, general constraints that act to shape and responses of food webs to trophic interactions, change food-web structure and function. These transfer efficiency, length of food chains, changes include biological stoichiometry, body-size, and in community composition, the relative importance the distribution of interaction strengths. Insights of versus detrital pathways, climate from ecological network analysis also provide new change, and the effects of natural and anthro- tools for thinking about dynamical and energetic pogenic disturbances. In addition, research is properties of food webs, tools which complement beginning to incorporate spatial and temporal a wide array of more long-standing approaches. This page intentionally left blank CHAPTER 1 Biosimplicity via stoichiometry: the evolution of food-web structure and processes

James J. Elser and Dag O. Hessen

Introduction on purine would reveal yet more details, finally yielding individual molecules In these days of the vaunted genome and the and each individual chemical reaction pathway. rest of the proliferating ‘‘gnomes’’ (transcriptome, The fascinating but intimidating journey just proteome, metabolome, etc.) and the unveiling completed should be familiar to food-web eco- of astonishing complex pleiotropy and / logists, for whom Figure 1.1(a) (Lavigne 1996) has genome interactions, it may seem headstrong to achieved near-iconic status as a symbol, sure to propose that there is something more complex stimulate uneasy laughter in the audience, of the than the genome currently under study in modern daunting complexity confronted by food-web biology. Nevertheless, we propose that the ecologists. If we were to follow the metabolic ‘‘entangled bank’’ of food webs, the trophic example and zoom in on the northwest Atlantic connections among interacting organisms in food web, we would, of course, encounter more ecosystems, is indeed as complex and bewildering and more detail. The node ‘‘,’’ for example, as the emerging genome and its products. The might resolve itself into larval, juvenile, and complexity increases further when considering mature cod, each connected, by its feeding, in the myriad pathways of matter and energy that different ways with other parts of the web. Further the species interactions build upon. Consider, for inspection might then reveal the individual cod example, a simplified map of central cellular themselves, each with a distinct genome and a metabolism (Figure 1.1). Here we can see, in basic unique physiological and behavioral repertoire. outline, the key pathways by which energy and How, then, can we deal with this layered com- key resources are metabolized in maintenance and plexity in food webs? And how could any con- growth of the . Note the complexity of necting thread of simplicity and unifying principles the diagram both in terms of the numbers of nodes be spotted in this overwhelming complexity? It is and the numbers and types of connections among our view that, just as the individual molecules in different components. As shown by the shading, metabolism are the critical level of resolution for different parts of the overall metabolism can be the molecular confronting the genome classified into different functional roles, in this case and its products, the level of the individual into 11 categories. Note also that we used the word organism should be of central importance for the ‘‘simplified.’’ That is, if we were to zoom in on the food-web ecologist. This is because, just as particu- nucleotide synthesis area of the diagram, more lar individual molecules (not classes of molecules) details would emerge, with more nodes (chemical are the actual participants in metabolic networks, categories) and pathways appearing (you can do it is individual organisms (not species, popula- this yourself on the Internet at www.genome.ad. tions, functional groups) that do the actual eating jp/kegg/). Yet more magnification, for example,

7 8 AQUATIC FOOD WEBS

(and thus make the trophic connections) in food (a) webs. Thus, organism-focused reasoning based on sound physiological principles is likely to be of great assistance in unraveling food webs. Further- more, the constituents of food webs are not fixed entities; rather, they are products and agents of continuous evolutionary change. And since evolu- tion operates primarily at the level of individual reproductive success, it seems that evolutionary thinking should play a central role in under- standing how and why food webs are shaped the way they are. Molecular/ coming to grips with the daunting complexity of the genome and its products (Figure 1.1) have a powerful ally in the fact that each node of a metabolic network is the product (and a reactant) in an enzymatically driven chemical reaction. Thus, all parts of the network must obey strict rules of mass balance and stoichiometric combination in the formation and destruction of the constituent parts. Indeed, these simplifying principles form the basis of various emerging theories through which cell biologists hope to make progress in understanding func- (b) tional interconnections among and products in metabolism (e.g. metabolic , Dykhuizen et al. 1987; Wildermuth 2000; stoichiometric , Hobbie et al. 2003; metabolic flux balancing, Varma and Palsson 1994). But why should such powerful tools be left to molecular biologists? Luckily, food-web ecologists can also take advantage of the considerable traction afforded by the firm laws of because, just as every node in a biochemical network is a chemical entity, so is every node in a food web. That is, each individual organism forming a connection point in Figure 1.1(b) is an aggregation of biochemicals and chemical elements and is sustained by the net outcome of the coupled biochemical pathways shown in Figure 1.1(a). Thus, the interactions among food-web components in terms of consump- Figure 1.1 Two entangled banks demonstrating the intimidating tion (and the imposed by nontrophic task that lies before biology. Ultimately, food webs relations of excretion and nutrient ) (a) (Lavigne 1996) are the outcome of dynamic interactions are also constrained by the firm boundaries of among various organisms that acquire resources from the abiotic mass balance and stoichiometric combination. environment and each other in order to drive their metabolism (b) (www.genome.ad.jp/kegg/) and leave offspring. These principles and their applications are known as ‘‘ecological stoichiometry’’ (Sterner and BIOSIMPLICITY VIA STOICHIOMETRY 9

Elser 2002), while their more recent extension to the partnership of an intimate comedy; and if we would catch realms of , behavior, physio- the spirit of the piece, our attention must not all be logy, and cellular/ are known as absorbed in the characters alone, but most also be ‘‘biological stoichiometry’’ (Elser et al. 2000b). The extended to the scene, of which they are born, on which they play their part, and with which, in a little while, approach of ecological stoichiometry simplifies the they merge again. bewildering ecological complexity in Figure 1.1(b) by focusing on key ecological players in food webs and characterizes them in terms of their relative Stoichiometric imbalance, ‘‘excess’’ carbon (C), (N), and (P) carbon, and the functioning of demands. In biological stoichiometry, metabolic food webs complexity in Figure 1.1(a) is simplified by focus- ing on major biochemical pools (e.g. rRNA, total Conventional food-web diagrams show binary protein, RUBISCO) that determine overall orga- feeding links among species. Flowchart analyses of nismal demands for C, N, and P and attempts food webs go beyond a binary depiction of feeding to connect those biochemical demands to major relations in using dry-weight, energy (Joules), or evolutionary forces operating on each organism’s carbon as a common currency to express the life history or metabolic strategy. magnitude of particular connections. The advant- In this chapter we will review some basic prin- age of using C-based flow charts is quite obvious ciples and highlight some of the most recent since, not only does C account for some 50% of dry findings from the realm of ecological stoichiometry mass in most species and taxa, it also allows for in food webs, to illustrate how a multivariate inclusion on import and export of inorganic carbon perspective on energy and chemical elements in the same scheme. However, with the realization improves our understanding of trophic relations. that the supply and availability of P is a key More details of these (and other) are avail- determinant of the binding, flux, and fate of C in able in Sterner and Elser (2002); in this chapter we freshwater food webs, in many cases more informa- seek to highlight some findings that have emerged tion may be gained from P-based flow charts. Such since publication of that work. We will then dis- a view will also provide a better representation of cuss recent movements to integrate stoichiometric the of elements, and thus differentiate study of food webs with the fact that food-web between ‘‘old’’ and ‘‘new’’ production. In a P-lim- components are evolving entities and that major ited system, any extra of P will allow for evolutionary impose functional trade- binding of more than 100 of C in offs on organisms that may have profound impli- . Due to different elemental ratios in dif- cations for the structure and dynamics of food ferent food-web compartments, conventional webs. Our overarching view is that stoichiometric flowchart diagrams will normally turn out quite theory can help in integrating food-web ecology different in terms of C or P (Figure 1.2). Neither is and evolution into a more comprehensive frame- more ‘‘true’’ or correct than the other; instead, each work capable of making a priori predictions about provides complementary information on pools and major food-web features from a relatively simple key processes. Since P is a conservative element set of fundamental assumptions. In advocating this that is lost from aquatic systems only by sedimenta- view we hope to continue to add substance to the tion or outflow, it will normally be frequently vision of food webs offered nearly 100 years ago recycled and may thus bind C in stoichiometric by Alfred Lotka (1925): proportions a number of over a . In addition to pictures of who is eating whom in For the drama of life is like a puppet show in which stage, scenery, actors, and all are made of the same stuff. food webs, we need to understand the outcome of The players indeed ‘have their exits and their entrances’, those feeding interactions for the . but the exit is by way of a translation into the substance ‘‘Trophic efficiency’’ is key aspect of food webs of the stage; and each entrance is a transformation scene. that captures important aspects of this outcome So stage and players are bound together in the close (here we use ‘‘efficiency’’ to refer to the fraction of 10 AQUATIC FOOD WEBS

explanations may be invoked to explain such 100 mg C m–3 100 mg C m–3 per day patterns but, as argued by Cebrian et al. (1998), surely the high transfer efficiency of C in plank- Z tonic systems may be attributed to both high cell A quotas of N and P relative to C (high-quality food B for reasons given below) coupled with decreased importance of low-quality structural matter like and cellulose that are poorly assimilated. F A striking feature of the cross-ecosystem compar- ison compiled by Cebrian et al. (1998) is the close correlation paths among autotroph turnover rate, D specific nutrient (N, P) content, and the trophic efficiency. These associations make perfect sense SE D SE D from a stoichiometric point of view: while con- Egestion/exudation Grazing sumers are not perfectly homeostatic (cf. DeMott Death 2003), they have a far closer regulation of elemental ratios in somatic than autotrophs (Andersen 1 mg P m–3 1mg P m–3 per day and Hessen 1991; Hessen and Lyche 1991; Sterner and Hessen 1994) and their input and output of elements must obey simple mass balance prin- A Z ciples. As a general rule, limiting elements are expected to be utilized for growth and transferred in food chains with high efficiency, while non- F limiting elements, by definition present in excess, B must be disposed of and may be recycled (Hessen 1992; Sterner and Elser 2002). Thus, when feeding on low C : N or low C : P food, a considerable share D of N and P may be recycled (Elser and Urabe Release by grazers Grazing 1999), while C-use efficiency is high (Sterner and Death Elser 2002). However, typically autotrophs have higher C : element ratios than consumers (Elser Figure 1.2 Pools and fluxes of carbon (a) and phosphorus (b) in a et al. 2000a). Thus, when consumers feed on diets pelagic food web of a eutrophic (data from Vadstein et al. that are high in C : N or C : P, nutrient elements are 1989). A: algae, B: bacteria, D: and other kinds of nonliving reclaimed with higher efficiency by the dissolved and particulate matter, F: heterotrophic flagellates, and Z: metazoan zooplankton. Boxes denotes , arrows denote (Elser and Urabe 1999), while much of the C is fluxes. Note the entirely different size of pools and fluxes for C and P. unassimilated and must be egested, excreted, or respired (DeMott et al. 1998; Darchambeau et al. 2003). Since performance is strongly energy or C produced by a certain level that is impaired in these high C : nutrient systems, more C transferred to higher trophic levels). Efficient must enter detrital pathways, as is clearly shown trophic systems typically have steep slopes from by Cebrian’s studies. autotrophs at the base of the food web to top These factors point to fundamental differences predators, and subsequently should also support a between ecosystems not only with regard to the higher number of trophic levels than less efficient transfer and sequestration of carbon, but also systems. Typically, planktonic systems are among with regard to community composition and eco- those with high trophic efficiency, with on system function in more general terms. While the extreme low end (Hairston and Hairston 1993; realizing that pelagic food webs are among the Cebrian and Duarte 1995; Cebrian 1999). Several most ‘‘efficient’’ ecosystems in the , there is BIOSIMPLICITY VIA STOICHIOMETRY 11 certainly a huge scatter in trophic efficiency also (Faerovig et al. 2002; Urabe et al. among pelagic systems, that is, considerable 2002a). At a given (low) level of P, high light yields variation appears when phytoplankton biomass or more algal biomass than low light treatments, production is regressed against zooplankton but with lower food quality (higher C : P). The net (Hessen et al. 2003). Such scatter may be caused by outcome will be slow herbivore growth rates at time-lag effects, external forcing, algal species’ high light, with a higher asymptotic biomass composition, and associated as well of adults. This is because high growth rate and as by top-down effects, but it is also clear that high reproduction require a diet that balances the alterations in trophic transfer efficiency due to grazer’s demands in terms of energy, elements, stoichiometric constraints could be a strong con- and , while a standing stock of tributor. Thus, understanding food-web dynamics (nearly) nonreproducing adults can be sustained requires understanding the and impacts of on a low-quality diet since their basic metabolic nutrient limitation of . requirements mostly rely on C (energy). This implies a shift from a high to low biomass: pro- Stoichiometry, nutrient limitation, duction ratio. Eventually, the nutrient constraint in and dynamics in food webs low quality (high autotroph C : P) systems may be overcome by feedbacks from grazers. Such intra- Since nutrient limitation of autotroph production and interspecific facilitation (Sommer 1992; Urabe only occurs, by definition, when nonnutrient et al. 2002a) may induce a shift in population resources such as light are sufficient, a particularly dynamics under a scenario of increasing grazing intriguing outcome of stoichiometric analysis in since an increasing amount of P will be available freshwaters, and one that is rather counter- per unit of autotroph biomass due to the combined intuitive, is that high inputs in the effect of grazing and recycling (cf. Sterner 1986). form of photosynthetically active may Thus, understanding the biological role of lim- reduce secondary (herbivore) production (Urabe iting in both autotrophs and consumers and Sterner 1996; Sterner et al. 1997; Hessen et al. provides a basis for better prediction of how 2002; Urabe et al. 2002b). The rationale is as of should respond follows: when photosynthetic rates are high due to changing environmental conditions that alter to high light intensity but P availability is low nutrient supply, light intensity, or other environ- (a common situation in freshwaters), C is accumu- mental conditions. However, surprisingly little lated in biomass out of proportion with P. Thus, attention in mainstream textbooks on population C : P in the phytoplankton increases, meaning dynamics has been given to food quality aspects potential reduced C-use efficiency (P-limitation) (e.g. Turchin 2003). According to the stoichiometric in P-demanding grazers such as . The growth rate hypothesis (described in more detail outcomes of such effects have been shown by later) and supported by an increasing body of Urabe et al. (2002b), who applied deep shading experimental data (Elser et al. 2003), taxa with that reduced light intensities nearly 10-fold to field high body P-contents commonly have high growth enclosures at the Experimental Lakes Area, where rates and can thus rapidly exploit available C : P ratios are generally high (Hassett et al. resources but are probably especially susceptible 1997) and Daphnia have been shown to be P-limited for stoichiometric food quality effects. What are (Elser et al. 2001). The outcome was a nearly five- the dynamic consequences of this under different fold increase in zooplankton biomass in unenriched conditions of nutrient limitation in the food web? enclosures after the five-week . For reasons given above, the a priori assumption However, the negative effects of high algal would be that predator–prey interactions should C : P ratio on zooplankton can be a transient situa- be more dynamic when the system sets off with tion and high energy (light) input may eventually high quality (low C : P) autotroph biomass. Low sustain a high biomass of slow-growing - autotroph C : P will stimulate fast growth of plankton, as demonstrated by long-term the consumer and relatively high recycling of P for 12 AQUATIC FOOD WEBS

autotroph reuse; the system should therefore (a) Qmin = 0.010, Qz = 0.018 rapidly reach an equilibrium where food abund- ance is limiting to the grazer. On the other hand, a system with high C : P in the autotrophs should have slow grazer response and low recycling of P, 1400 yielding sluggish and perhaps erratic dynamics as 1200 the system operates under the simultaneous effects 1000 of changing food , quality, and nutrient 800 Bz recycling. Indeed, recent models (Andersen 1997; 600 Hessen and Bjerkeng 1997; Loladze et al. 2000; 400 Muller et al. 2001) taking grazer P-limitation 200 and recycling into account clearly demonstrate 0 1000 0.025 this kind of dynamic dependency on 2000 0.020 and consumer stoichiometry. As demonstrated by 3000 0.015 4000 Figure 1.3 (Hessen and Bjerkeng 1997), the ampli- Ba 0.010 5000 Qa tudes and periods of autotroph–grazer limit cycles 6000 0.005 depends both on food quantity and quality. When P : C in the autotroph becomes low, this constrains grazer performance and a high food biomass of low quality may accumulate before the grazer slowly builds up. With assumptions of a more (b) Qmin = 0.003, Qz = 0.018 efficient elemental regulation in the autotroph (i.e. lower minimum P : C), limit cycles or amplitudes will be smaller, but the periods will increase. One intriguing feature of stoichiometric model- 1400 ing is the potential of the grazer, like a 1200 P-demanding Daphnia, under a scenario of high 1000 food biomass but low food quality (Andersen 1997; 800 B Hessen and Bjerkeng 1997). External enrichment z 600 of P to the system will also invoke strong shifts in 400 due to stoichiometric mechan- 200 0 0.025 isms (Andersen 1997; Muller et al. 2001). The 1000 relevance for these theoretical exercises for natural 2000 0.020 Bz 0.015 systems remains to be tested, however. Clearly 3000 isocline 4000 0.010 Ba the assumption of two compartment dynamics 5000 0.005 Qa represent an oversimplification, since a consider- 6000 able share of recycled P and organic C will enter Figure 1.3 Three-dimensional limit cycles for two scenarios the bacteria or detritus pool, thus dampening the with Daphnia grazing on algae with different flexibility in their dynamics predicted from the simplified model P : C ratio (Qa). The solid line gives the trajectory, while projections of the three-dimensional trajectory are given on the Ba–Bz assumptions. 1 plane and the Qa–Ba plane. Ba: algal biomass (mgCl ), Bz: Thus, one central outcome of stoichiometric 1 grazer (Daphnia) biomass (mgCl ), Qa: algal P : C (mgP:mg C). theory in consumer–resource systems is deviation In the upper panel, the lower bound of Qa (Qmin) is set to 0.010(a), from the classical straight Lotka–Volterra isoclines while Qmin in the lower panel is 0.003(b). P : C in the grazer (Qz) (Andersen 1997; Murdoch et al. 2003). From is in both cases fixed at 0.018. By increasing Qz (higher P : C ratio, these analyses, it appears that a combination of lower C : P ratio) slightly in the lower scenario, the grazer will go extinct, and the system will stabilize at a high algal biomass inter- and intraspecific facilitation during periodic near Qmin. nutrient element limitation by consumers results in a deviation from straightforward negative density BIOSIMPLICITY VIA STOICHIOMETRY 13 dependence in consumer and a much and developmental stages according to their mode richer array of population dynamics appears. of feeding (, , herbivores, In this way, stoichiometry can provide a logical , filtrators, raptors, , etc). explanation for Allee effects (positive density- We suggest that it might also be useful to adopt dependence) and hump-shaped curves for density- a subtler categorization based on dietary, stoichio- dependent responses. metric requirements. In fact, in many cases the Much of the preceding discussion has had more specific dietary requirements of a taxon may herbivores and other primary consumers (e.g. be the ultimate cause for an organism being a detritivores) in mind. What about the role of or a and, as we discuss stoichiometry higher in food webs? Since meta- below, is probably also an important factor con- zoans do not vary too much in their biochemical tributing to widespread omnivory among taxa. makeup, predators are less likely to face food Hence one could speak about the ‘‘stoichiometric quality constraints compared with herbivores and niche’’ of a particular species, in the sense that especially detritivores (Sterner and Hessen 1994). species (or stages) with high P (or N) requirements in general have high P requirements due to would succeed in situations that supply nutrient- investment in bone (Sterner and Elser 2002); this rich food compared with species with lower could be seen as another reason, in addition to nutrient requirements. For example, for freshwater their large body size, why P-rich Daphnia should food webs it has been suggested that when be preferred prey relative to P-poor . planktonic algae are deprived of P and develop A more important issue is, however, how the high C : P ratios, P-demanding species like Daphnia predicted dynamics due to stoichiometric mechan- acutely suffer from ‘‘P-starvation’’ and, probably isms might be associated with the potential prey due to decreased C-use efficiency, become com- susceptibility to predators. A reasonable assump- petitively inferior to less P-demanding members of tion would be that grazers in low food quality the plankton community like Bosmina (DeMott and systems would be more at risk for predatory Gulati 1999; Schulz and Sterner 1999). Thus, the mortality simply because, all else being equal, stoichiometric niche space available to Bosmina slow growth would render the population more may extend to higher of food C : P than susceptible to the impacts of any given rate of in Daphnia. But what about the other end of the mortality loss. However, the effects might not C : P continuum? Interestingly, in a brand-new quite be so straightforward. For example, fast- stoichiometric wrinkle, recent evidence shows that growing individuals generally also require high extremely low C : P may cause decreased growth in rates of food intake; in turn, more active feeding Daphnia (Plath and Boersma 2001) and the cater- might increase risk (Lima and Dill pillar Manduca sexta (Perkins et al. 2003). While the 1990). Furthermore, there may be some inherent mechanistic bases of these responses remain and unappreciated physiological–developmental obscure, they appear to represent the other side of impacts associated with rapid growth such that the stoichiometric niche in the P-. overall mortality is elevated in fast-growing indi- Another aspect of stoichiometric effects on bio- viduals, over and above potentially accentuated diversity and food-web structure relates to the predation risk (Munch and Conover 2003). number of trophic levels and the degree of omnivory. Hastings and Conrad (1979) argued that Stoichiometry, omnivory, and the the evolutionary stable length of food chains evolution of food-web structure would be three, and that the main determinant of the number of trophic levels is the quality of prim- The fact that different species or taxa have differ- ary production (and therefore, to at least some ent stoichiometric or dietary requirements has degree, its C : nutrient ratio) and not its quantity, as important bearings on the dietary preferences that is often implied in discussions of food-web length. weave food webs together. Ecologists have com- Omnivory may be seen as a compromise between monly generated a coarse classification of species exploiting large quantities of low quality resources 14 AQUATIC FOOD WEBS at low metabolic cost, or utilizing lower quantities that they often cannot even build their bodies of high-quality food at high metabolic costs. From from available food? Why do some species seem a stoichiometric point of view, omnivory may be to be more sensitive to the effects of stoichio- seen as a way of avoiding nutrient deficiency while metric food quality? In this section we follow at the same time having access to a large the advice of Holt (1995) by describing some of energy. This ‘‘best of two ’’ strategy is recent findings that illuminate some of these evolu- clearly expressed as life-cycle omnivory, like in tionary questions in the hopes that perhaps in the crayfish where fast-growing juveniles are carni- future we will encounter Darwin as well as Lin- vorous, while adults chiefly feed on detritus or deman in the reference sections of food-web (cf. Hessen and Skurdal 1987). papers. The fact that organisms can be potentially Beyond expanding the diet to include more limited not only by access to energy (carbon) but nutrient-rich prey items and thus inducing also by nutrients has obvious implications for omnivory as discussed earlier, another obvious coexistence of potential competitors. While this evolutionary response to stoichiometrically unbal- principle has been well explored for autotrophs anced food would be for a consumer to evolve a (e.g. Tilman 1982), the same principle may be lower body requirement for an element that is invoked for with different require- chronically deficient in its diet. Several recent ments for key elements (cf. Loladze et al. 2004). studies emphasizing terrestrial biota have pro- In fact, this will not only hold for interspecific vided evidence for just such a response. Fagan competition, but also for intraspecific competition, et al. (2002) examined the relative nitrogen content since most species undergo ontogenetic shifts in (%N of dry mass, N : C ratio) of folivorous nutrient requirements. Indeed, it now seems species and documented a significant phylogenetic that this coexistence principle can be extended signal in which the recently derived insect group to explain the evolution and maintenance of (the ‘‘Panorpida,’’ which includes Diptera and omnivory (Diehl 2003), since utilization of diff- Lepidoptera) have significantly lower body N erent food resources in species with different content than the more ancestral groups Coleoptera nutrient contents promotes and stabilizes feeding and Hemiptera which were themselves lower than diversification. the still older Lower Neoptera. Their analysis eliminated differences due to body size, gut con- Biological stoichiometry: the tents, or feeding mode as possible explanations for convergence of ecological the pattern. They noted that the divergence of and evolutionary time major insect groups appears to have coincided with major increases in atmospheric CO2 con- It should be clear from the preceding material that centrations (and thus high plant C : N ratio) stoichiometric imbalance between food items and and hypothesized that clades of that consumers has major effects on the dynamics emerged during these periods of ‘‘nitrogen crisis’’ and structure of key points in the food web and in plant biomass were those that had an efficient especially at the autotroph–herbivore interface. N economy. Signs of evolutionary response to Indeed, the effects of stoichiometric imbalance on stoichiometric imbalance in insects is also seen at a herbivores are often extreme and suggest that finer scale in studies by Jaenike and Markow there should be strong selective to alle- (2002) and Markow et al. (1999), who examined the viate these impacts. The fact that such impacts body C : N : P stoichiometry of different species of nevertheless remain implies that there may be Drosophila in relation to the C : N : P stoichiometry fundamental trade-offs and constraints on evolu- of each species’ primary resource. Host tionary response connected to organismal stoichio- involved different species of rotting cactus, fruit, metry. So, why is it that consumer organisms, mesquite exudates, and mushrooms and presented such as herbivorous zooplankton or insects, a considerable range in nutrient content. They maintain body nutrient contents that are so high showed a significant correlation between host BIOSIMPLICITY VIA STOICHIOMETRY 15 nutrient content and the nutrient content of the various comparisons involving physiological, associated fly species: taxa specializing on ontogenetic, and cross-species comparisons, nutrient-rich mushrooms had the highest nutrient P-content increases with growth rate (Main et al. contents (in terms of %N and %P) while those 1997; Carrillo et al. 2001; Elser et al. 2003; Makino specializing on the poorest quality resource and Cotner 2003), RNA content also increases with (mesquite flux) had the lowest body nutrient growth rate (Sutcliffe 1970; Vrede et al. 2002; Elser contents. Signs of evolutionary response to stoichio- et al. 2003; Makino and Cotner 2003; Makino et al. metric resource limitation can be found even in 2003), and, importantly, increased allocation to the structure of themselves. RNA quantitatively accounts for the increased Using genomic data for the Escherichia P content of rapidly growing organisms (Elser et al. coli and the Saccharomyces cerevisiae, 2003; Figure 1.4(a)). The mechanistic connections Baudoin-Cornu et al. (2001) showed that protein among RNA, P, and growth appear to manifest involved in C, N, or S metabolism were quite directly in how nutrient supply affects herb- significantly biased in favor of amino acids having ivore dynamics, as illustrated in a recent study of a low content of C, N, and S, respectively. That is, plant–herbivore interaction in the Sonoran enzymes involved in uptake and processing of N (Figure 1.4(b)). In this study (Schade et al. 2003), were disproportionately constituted of amino acids interannual variation in rainfall led to increased having relatively few N atoms, a response that P supply under wet conditions, which in turn makes adaptive sense. lowered foliar C : P in velvet mesquite . These studies suggest that heterotrophic con- Consequently, mesquite-feeding had higher sumers can indeed respond in evolutionary time to RNA and correspondingly higher P contents, reduce the degree of stoichiometric imbalance consistent with a P-stimulated increase in growth between their biomass requirements and their rate, along with higher population densities on often-poor diets. And yet the overall nutrient mesquite branches with low foliar C : P. In sum, content (e.g. %N and %P) of herbivorous these findings of tight and ecologically significant (zooplankton, insects) remains at least 10-times associations of growth, RNA, and P contents and often 100- to 1,000-times higher than is found suggest that major aspects of an organism’s in autotroph biomass (Elser et al. 2000a). Some ecology and life history have an important fundamental benefit of body nutrient content must stoichiometric component. Given the strong effects exist. The nature of the benefits of higher body of stoichiometric imbalance in trophic relations nutrient content is perhaps becoming clearer now (Figure 1.3), it seems then that evolutionary based on work emerging from tests of the ‘‘growth adjustments bearing on growth rate will impact rate hypothesis’’ (GRH hereafter), which proposes on the types of organisms and species that come that variation in the C : P and N : P ratios of many to dominate key positions in food webs under organisms is associated with differences in growth different conditions. rate because rapid growth requires increased Thus, better understanding of the genetic basis allocation to P-rich RNA (Hessen and Lyche 1991; and evolutionary dynamics of the coupling among Sterner 1995; Elser et al. 1996; Elser et al. 2000b). growth, RNA, and P may help in understanding In this argument, a key life history parameter, how food webs self-organize. Elser et al. (2000b) growth or development rate, is seen to inherently suggested that the genetic basis of variation in require increased investment in P-intensive bio- growth and RNA (and thus C : P and N : P ratios) chemicals, implying a trade-off in that fast- lies in the genes encoding for RNA, the rDNA. growing organisms are likely to find themselves In particular, they reviewed evidence suggest- constrained by an inability to acquire sufficient P ing that rDNA copy number and length and con- from the environment or diet. Work with zoo- tent of the rDNA intergenic spacer (IGS, where plankton and other organisms such as insects and promoter–enhancer sequences are found) were key microbes now makes it clear that the fundamental variables underlying growth rate variation because core of the GRH is correct (Figure 1.4(a)): in of their effects on transcriptional capacity and ebvr iooe fet ouaindnmc.I td fadsr csse,itrnulvraini analldt neana difference interannual ( to trees led mesquite rainfall velvet in variation to interannual panel) ecosystem, left desert (upper a availability of study P a soil In in dynamics. population affects ribosomes herbivore -otn fdvrehtrtohcognss h aaicue nti ltivledt from data involve plot this in included data The organisms. heterotrophic diverse of P-content 1.4 Figure 16 ihfs-rwn ueie eedmntdb niiul arigol h ogIS iue rmGrkoae l (2002). al. et lines Gorokhova and from low long while Figures both IGS. carrying mixed long individuals the the of by only dominated proportion were carrying relative juveniles individuals the slow-growing by in with generations dominated lines differed five fecundity were for lines high juveniles selection Selected IGS: Artificial fast-growing left). long (c) with (bottom the (2003). just contents al. or P et IGS and Schade short from RNA Figures parthenogenetic juvenile higher. in in was panel) and supply left (top P fecundity soil weight-specific and on rainfall when right) (bottom densities nrsos,msut-edn evl ( weevils mesquite-feeding response, In rsaenzolntnincluding zooplankton n hne u opyilgclPlmtto.I ahcs,RArc adtu -ih aawr rmognsswt eaieyfs rwhrate. growth an fast and relatively 0.97 with of organisms slope from a were has data combined P-rich) data thus the (and of RNA-rich all case, to each fit In relationship P-limitation. A physiological to due changes and (a)

% Total-P WEBS FOOD AQUATIC 0 1 2 3 012 rmtegnm hog eaoimt h odwb a aito nalcto oPrc iooa N xlisvraini the in variation explains RNA ribosomal P-rich to allocation in Variation (a) web. food the to metabolism through genome the From (c) % RNA-P Daphnia 0.2 0.4 0.6 0.8 1.2 1.4 10 12 0 2 4 6 8 0 1 iii setosa Sibinia niiulrltosisivleitrpcfi oprsn,cagsdrn noeei development, ontogenetic during changes comparisons, interspecific involve relationships Individual . y %RNA 0.97 = Hi R g %P 2 o Hi Low h 0.78 = X a ihrRAadPcnet bto et n hsahee ihrpopulation higher achieved thus and left) (bottom contents P and RNA higher had ) 0.62 + Fecundity %P ahi pulicaria Daphnia rspsvelutina Prosopis 3 (b) 100 R 0.1 0.2 0.3 0.4 0.5 20 40 60 80 2 0 0

f07.Fgr rmEsre l 20) b rnmsino olPinto P soil of Transmission (b) (2003). al. et Elser from Figure 0.78. of Sibinia %RNA or %P 0.5 1.5 2.5 0 1 2 3 Growth rate

May 2000 Soil P rdcdcreae hne njvnl rwhrt tpright) (top rate growth juvenile in changes correlated produced ,laigt eraei oirC:Prto(pe right). (upper ratio P : C foliar in decrease a to leading ), %RNA g May 2000 h %P Rainfall % Apr 2001 2001 Apr Low .cl,Doohl melanogaster Drosophila coli, E. %P

No. of Sibinia per g Mesquite leaf C : P 100 200 300 400 500 600 700 800 900 0 1 2 3 0 May 2000 0 800 400 efC:P : Leaf C Apr 2001 n various and , 1200 s BIOSIMPLICITY VIA STOICHIOMETRY 17 thus rates of RNA production. Thus, rDNA If the results of Gorokhova et al. are confirmed variation may also be linked to differences in and shown to hold for other biota that comprise stoichiometric requirements of biota. Emerging food webs, then we would suggest that there is data suggest that indeed it is the case that varia- reason to question the traditional distinction tions in the rDNA can have important stoichio- between ecological and evolutionary time. That is, metric ramifications (Gorokhova et al. 2002; ecologists are used to considering species shifts Weider et al. 2004). (e.g. during seasonal succession in the plankton) in Of these data, most striking is the study of terms of the sorting out of various ecological Gorokhova et al. (2002), who performed an artificial transactions such as competition and predation selection experiment on the progeny of a single among taxa that have fixed genetic structure on the female of Daphnia pulex, selecting on weight- timescale of the study. The results of Gorokhova specific fecundity (WSF) of the animals. The et al. suggest that, even with a clonal organism, treatments responded remarkably quickly to the genetic recombinations with important ecophysio- selection regimen—animals selected for low logical impacts can arise on time scales that would fecundity were significantly lower than random easily be encompassed, for example, by a single controls or high fecundity selected animals within growing season in a lake. This same point of rapid three rounds of selection. After five rounds of evolution in food webs is also demonstrated by selection, animals were assayed for juvenile studies of rapid evolution of digestibility-growth growth rate along with RNA and P contents. The trade-offs in rotifer-algae (Fussmann analyses showed that lineages showing low WSF et al. 2003, Yoshida et al. 2003). Not all evolutionary had experienced a correlated inverse response of change requires the slow propagation of small juvenile growth rate; that is, these females with point through the vast protein library low WSF produced offsprings that grew faster, that comprises the genome. Indeed it is becoming offering an opportunity to test the growth rate increasingly obvious that the genome can reorgan- hypothesis. As predicted by the GRH, these fast- ize quickly via structural mutations that affect growing juveniles had elevated RNA and P regulatory regions and gene copy number and via contents compared to the slower-growing counter- gene silencing/unsilencing mechanisms such as parts in the control and high-selected lines (Figure those resulting from insertion/deletion of trans- 1.4(c)). To determine if these differences had a posable elements. It is interesting to note that genetic basis, animals were screened for variation transposable elements have a significant role in the IGS of their rDNA. Consistent with the idea in silencing copies of the rDNA in Drosophila above that high RNA (fast growth) (Eickbush et al. 1997) and have been identified should be associated with long IGS, a dispropor- in the rDNA of various other taxa, including tionate number of the (fast-growing, high RNA, Daphnia (Sullender and Crease 2001). To the extent high P) animals in the low fecundity treatment that these reorganizations impact traits that are carried only long IGS variants while (slow- ecophysiologically relevant (such as those that growing, low RNA, low P) animals in the high affect RNA production and thus organismal WSF and control lines disproportionately carried C : N : P stoichiometry), the shifting genomes of both long and short IGS variants. Recall that this interacting species will need to be incorporated experiment involved selection on the offspring of a (somehow!) into food-web ecology. single parthenogenetic female and all offspring throughout were also produced by parthenogen- esis. Nevertheless, it appears that functionally Conclusions significant genetic variation can arise in a few generations even within a ‘‘clonal’’ organism, as is Lotka’s ecological play remains to a large degree a becoming increasingly recognized (Lushai and mysterious entertainment. There is a long road Loxdale 2002; Loxdale and Lushai 2003; Lushai ahead for food-web ecologists before we can expect et al. 2003). to really understand how food webs self-organize 18 AQUATIC FOOD WEBS and are sustained in their complex interplay Children’’ begins with what sounds like a familiar with the abiotic environment. So, in some problem: ways this chapter is a plea for a form of a ‘‘bio- The rising hills, the slopes, simplicity’’ research program in approaching of statistics food-web study. That is, Darwin unlocked the lie before us. key to the origin of the biosphere’s bewildering the steep climb biodiversity with his breathtakingly simple of everything, going up, algorithm of variation and selection. Perhaps up, as we all go down. the complexity we see in food webs is more apparent than real and will also come to be seen to The poem ends with the advice: be the product of a simple set of rules ramifying stay together through time and space. Bearing in mind the learn the flowers warning ‘‘if your only tool is a hammer then every go light problem looks like a nail,’’ for the moment we So, building on this counsel, perhaps we should propose that the minimal biosimplicity ‘‘rulebox’’ try to stay together with our colleagues who are should include the Darwinian paradigm along confronting the genome, adapting their tools, data, with the laws of , mass con- and ideas to better understand our own questions, servation, and stoichiometric combination. What and perhaps offering some insights of our own. kind of food-web can be built using We do need to learn the flowers, their names, and such tools given the raw materials circulating in the names of those who eat them, and of those who the biosphere? eat those too. But, especially these days when the Answering such a question will take some time juggernaut threatens to bury us all in a and a large quantity of cleverness and a certain blizzard of data and detail, perhaps we should try amount of courage. We can take some solace in the to go light, approach the problems with a simpli- fact that we will be traveling this long road toward fied but powerful toolbox that includes stoichio- understanding of biological complexity with metric principles, and see what major parts of the our colleagues who study the genome and its puzzle will yield themselves to our best efforts. immediate metabolic products. While they might not themselves always realize that high- Acknowledgments throughput sequencers and microarray readers do not necessarily result in high throughput of the The authors are grateful for the support of the conceptual insights that might make sense of it all, Center for Advanced Study of the Norwegian a time will come (or has it already?) when all Academy of Letters and Sciences that made biologists, from the geneticists to the ecologists, this work possible. JJE also acknowledges NSF will lean on each other for insight and necessary grant DEB-9977047. We thank M. Boersma and data. For guidance at this point we turn to the K. Wiltshire for their helpful comments on an early American poet Gary Snyder whose poem ‘‘For the draft. CHAPTER 2 Spatial structure and dynamics in a

Carlos J. Melia´ n, Jordi Bascompte, and Pedro Jordano

Introduction studies in food-web structure and dynamics have dealt with either large (but see Hori and Noda 2001) The role of space in population and community or small (but see Caldarelli et al. 1998) number of dynamics has been recently emphasized (e.g. species, but make no explicit reference to space Hanski and Gilpin 1997; Tilman and Kareiva 1997; (Caswell and Cohen 1993; Holt 1996, 1997). Only a Bascompte and Sole´ 1998). Several models for the few studies have explored the role of space on a small coexistence of interacting species in heterogeneous subset of trophic interacting species (Holt 1997; environments have been formulated. These Melia´n and Bascompte 2002). include the energy and material transfer across The present study is an attempt to link structure ecosystem boundaries and its implication for and dynamics in a spatially structured large succession and diversity (Margalef 1963; Polis et al. marine food web. We use data on the diet of 5526 1997), the geographic mosaic of specimens belonging to 208 fish species (Randall (Thompson 1994), the regional coexistence of 1967) in a community in five different competitors via a competition–colonization trade- habitats (Opitz 1996; Bascompte et al., submitted). off (Tilman 1994), the random assembly of com- First, we analyze structure by addressing how munities via limitation (Hubbell 2001), simple trophic modules (i.e. tri-trophic food chains and (Wilson 1992). As a general (FCs) and chains with omnivory (OMN) with the conclusion of these approaches, succession, dis- same set of species are shared among the five persal, local interactions, and habitats. Second, we extend a previous meta- have appeared strongly linked to the persistence community model (Mouquet and Loreau 2002) by of diversity. However, the underlying pattern of incorporating the dynamics of trophic modules ecological interactions in a spatially structured in a set of connected communities. Specifically, ecosystem and its implications for the persistence the following questions are addressed: of biodiversity remains elusive by the lack of synthetic data (Loreau et al. 2003). 1. How are simple trophic modules composed Introducing space and multiple species in a by the same set of species represented among single framework is a complicated task. As Caswell habitats? and Cohen (1993) argued, it is difficult to analyze 2. How does the interplay between dispersal and patch-occupancy models with a large number of food-web structure affect species dynamics at both species because the number of possible patch states local and regional scales? increases exponentially with . Therefore, most spatial studies have dealt with a Data collection: peculiarities few number of species (Hanski 1983), predator– and limitations prey systems (Kareiva 1987), or n-competing species (Caswell and Cohen 1993; Tilman 1994; Mouquet The Caribbean fish community here studied covers and Loreau 2003). On the other hand, the bulk of the geographic area of Puerto Rico–Virgin .

19 20 AQUATIC FOOD WEBS

Data were obtained in an area over more than number of chains (with identical species at all 1000 km2 covering the US Virgin Islands of trophic levels) shared by each pair of communities. St Thomas, St John, and St Croix (200 km2), the To assess whether this shared number is higher or British Virgin Islands (343 km2), and Puerto Rico lower than expected by chance we develop a null (554 km2). The fish species analyzed and asso- model. This algorithm randomizes the empirical ciated data were obtained mainly from the study data at each community, yet strictly preserves the by Randall (1967), synthesized by Opitz (1996). ingoing and outgoing links for each species. In this Spatially explicit presence/absence community algorithm, a pair of directed links A–B and C–D matrices were created by considering the presence are randomly selected. They are rewired in such a of each species in a specific only when that way that A becomes connected to D, and C to B, particular species was recorded or provided that none of these links already existed breeding in that area (Opitz 1996; Froese and Pauly in the network, in which case the rewiring stops, 2003). Community matrices include both the and a new pair of links is selected. trophic links and the spatial distribution of 208 fish We randomized each food web habitat 200 taxa identified to the species level. Randall’s list of times. For each pair of habitats we compare each species was completed by Opitz (1996), successive pair of replicates and count the shared which included more with affinities to number of simple tri-trophic FCs and chains with reefs of the Puerto Rico–Virgin islands area, based OMN containing exactly the same set of species. on accounts in Fischer (1978). Note that our trophic Then we estimated the probability that a pair-wise modules are composed only by fishes, and that all comparison of a random replicate has a shared fish taxa is identified to the species level, which number of such modules equal or higher than the implies that results presented here are not affected observed value. Recent algorithm analysis suggest by trophic aggregation. that this null model represents a conservative The final spatially explicit community test for presence–absence matrices (Miklo´s and includes 3,138 interactions, representing five Podani 2004). food webs in five habitat types. Specifically, the We calculated the number of tri-trophic FCs, and habitat types here studied are /estuaries OMN chains common to all pairs of communities, (m hereafter; 40 species and 94 interactions), coral and compared this number with that predicted by reefs (c hereafter; 170 species and 1,569 interactions), our null model (Figure 2.1(b) and (d)). The coral beds/algal mats (a hereafter; 98 species reef habitat shares with all other habitats a number and 651 interactions), sand (s hereafter; 89 species of FCs and OMN larger than expected by chance and 750 interactions), and offshore reefs (r hereafter; (P < 0.0001 in all pair-wise comparisons except 22 species and 74 interactions). To a single habitat for the mangrove comparison, where P < 0.002 85 species are restricted while 46, 63, 12, and 2 and P < 0.01 for FCs and OMN, respectively). species occupy 2, 3, 4, and 5 habitats, respectively. Similarly, seagrass beds/algal mats and sand (a/s Global connectivity values (C) within each habitat contrasts) share a significant number of FCs are similar to previously reported values for food and OMN (P < 0.0001). Globally, from the 10 webs (Dunne et al. 2002). Specifically, Cm ¼ 0.06, possible intercommunity comparisons, five share a

Cc ¼ 0.054, Ca ¼ 0.07, Cs ¼ 0.095, and Cr ¼ 0.15. number of modules higher than expected by chance (Figure 2.1(a) and (c) where arrows are Food-web structure and null model thick when the pair-wise comparison is statistically significant, and thin otherwise). This suggests that We consider tri-trophic FCs (Figure 2.1(a)) and FCs habitats sharing a significant proportion of trophic with OMN (Figure 2.1(c)). We count the number modules are mainly composed by a regional pool and species composition of such trophic modules of individuals. within the food web at each community. We then The average fraction of shared FCs and OMN make pair-wise comparisons among communities between habitat pairs is 38%24.5% and 41%25%, (n ¼ 10 pair-wise comparisons) and count the respectively, which still more than 50% of SPATIAL STRUCTURE AND DYNAMICS 21

(b) 1,500 (a)

1,000 c m a Frequency (FC) Frequency 500

r s

0 c/s c/a a/s c/r m/c m/a m/s s/r a/r m/r Habitat pairs

(d) 1,500 (c)

1,000 c m a

Frequency (OMN) Frequency 500

r s

0 c/s c/a a/s c/r m/c m/a m/s s/r a/r m/r Habitat pairs

Figure 2.1 The food-web modules studied here are (a) tri-trophic FCs, and (c) OMN chains. Circles represent the five different habitat types. For each habitat pair, the link connecting the two habitats is thick if the number of shared trophic modules is significant, and thin otherwise; (b) and (d) represent the frequency of shared tri-trophic FCs and OMN chains, respectively in all pair-wise community comparisons. Black and white histograms represent the observed and the average expected value, respectively. Habitat types are mangrove/estuaries (m), coral reefs (c), seagrass beds/algal mats (a), sand (s), and offshore reefs (r). As noted, coral reefs (c), share with the rest of the habitats a number of FCs and OMN larger than expected by chance, which suggest a high degree of connectance promoted by dispersal. different species composition trophic modules trophic modules across space in larger structures, between habitats. However, it is interesting to note which suggest a cohesive spatial structure (Melia´n that 15 species (specifically, herbivorous species and Bascompte 2004). from Blenniidae and Scaridae families, and top species from Carcharhinidae and Sphyrnidae famil- Dynamic metacommunity model ies) are embedded in more than 75% of trophic modules, which suggests that a small number of In order to assess the local and regional dynamics species are playing an important role in connecting of the structure studied, we extend a previous through dispersal local community dynamics. metacommunity model (Mouquet and Loreau Note that these highly connected species link 2002, 2003) by incorporating trophic modules 22 AQUATIC FOOD WEBS

(tri-trophic FCs and FCs with OMN) in a set of where the sum stands for all the other commun- interacting communities. The model follows the ities l. Rik represents the amount of resources formalism of previous models available to species i in community k (Levins 1969) applied to the scale of the individual XS : (Hastings 1980; Tilman 1994). At the local scale Rik ¼ aijkPjk,(24Þ (within communities), we consider a collection j¼1 of identical discrete sites given that no site is ever where aijk is the predation rate of species i on occupied by more than one individual. The regio- species j in community k, and the sum is for all nal dynamics is modeled as in mainland– prey species. Similarly, Cik represents the amount models with (Gotelli 1991), but with of consumption exerted on species i by all its an explicit origin of immigration that is a function predators in community k, and can be written as of emigration from other communities in the meta- follows: community (Mouquet and Loreau 2003). Therefore, XS : the model includes three hierarchical levels Cik ¼ ajikPjk,(25Þ (individual, community, and metacommunity). j¼1

The model reads as follows: where ajik is the predation rate of species j on dP species i in community k, and the sum is for all ik ¼ yI V þ (1 dÞc P V m P dt ik k ik ik k ik ik predator species. : : þ RikPik CikPik (2 1Þ We have numerically simulated a metacommu-

At the local scale, Pik is the proportion of sites nity consisting of six species in six communities. In occupied by species i in community k. Each com- each community, either two simple tri-trophic FCs, munity consists of S species that indirectly com- or two OMN chains are assembled with the six pete within each for a limited species. The two trophic modules within each proportion of vacant sites, Vk, defined as: community are linked only by indirect competition XS between species within the same trophic level. We : Vk ¼ 1 Pjk,(22Þ assumed a species was locally extinct when its j¼1 proportion of occupied sites was lower than 0.01. where Pjk represents the proportion of sites occu- Mortality rates (mik) are constant and equal for all pied by species j within the same trophic level in species. Dispersal success (y) was set to 1. community k. The metacommunity is constituted We considered potential reproductive rates to fit by N communities. d is the fraction of individuals the constraint of strict regional similarity, SRS dispersing to other habitats, and dispersal success, (Mouquet and Loreau 2003). That is, species within y, is the probability that a migrant will find a new each trophic level have the same regional basic community, cik is the local reproductive rate of reproductive rates, but these change locally among species i in community k, and mik is the mortality communities. Under SRS, each species within each rate of species i in community k. trophic level is the best competitor in one com- For each species in the community, we munity. Similarly, we introduce the constraint of considered an explicit immigration function Iik. strict regional trophic similarity (SRTS). That is, Emigrants were combined in a regional pool of each consumer has the same set of local energy dispersers that was equally redistributed to all requirements but distributed differently among other communities, except that no individual communities. Additionally, we assumed a direct returned to the community it came from (Mouquet relationship between the resource’s local repro- and Loreau 2003). After immigration, individuals ductive rate and the intensity it is predated with were associated to the parameters corresponding (Jennings and Mackinson 2003). to the community they immigrated to. Iik reads as: Under the SRS and SRTS scenarios, regional XN species abundance and intercommunity variance d I ¼ c P ,(2:3Þ are equal for each of the two species within the ik N 1 il il l6¼k same trophic level. Regional abundance in OMN is SPATIAL STRUCTURE AND DYNAMICS 23 higher, equal, and lower for top, intermediate, and (a) 0.4 basal species, respectively. Local abundances differ significantly between the two modules explored. 0.3 Specifically, when there is no dispersal (d ¼ 0) there is local exclusion by the competitively 0.2 superior species (Mouquet and Loreau 2002). This occurs for the basal and top species in the simple 0.1 trophic chain. The variance in the abundance of the basal and top species between local communities is 0 thus higher without dispersal for tri-trophic FCs 0 0.2 0.4 0.6 0.8 1 (Figure 2.2(a)). d However, the situation is completely different (b) 0.4 for OMN. Now, intercommunity variance is very low for both the basal and top species in the 0.3 absence of dispersal, and dramatically increases with d in the case of the top species. When the 0.2 communities are extremely interconnected, the top species disappears from the two communities < 0.1 (Pik 0.01), and is extremely abundant in the remaining communities. For intermediate species, varianceIntercommunity variance Intercommunity 0 increasing dispersal frequency decreases the 0 0.2 0.4 0.6 0.8 1 intercommunity variance, except when d ranges d between 0 and 0.1 in FCs (Figures 2.2(a) and (b)). Figure 2.2 Intercommunity variance in local species abundance Finally, we can see in Figure 2.2(b) (as compared for the basal (continuous line), intermediate (broken line), and top with Figure 2.2(a)) that intercommunity variance (dotted) species as a function of the proportion of dispersal between for high d-values is higher in a metacommunity communities (d ). (a) Represents tri-trophic FCs and (b) OMN with OMN. Thus, the interplay between dispersal chains. Parameter values are mik ¼ 0.2, cik for basal species is 3, among spatially structured communities and food- 2.8, 2.6, 2.4, 2.2, and 2 from the first to the sixth community, respectively. For intermediate species cik is 1.5, 1.4, 1.3, 1.2, 1.1, and web structure greatly affects local species abund- 1, respectively from the first to the sixth community. Top species ances. The results presented here were obtained reproductive values are 0.8, 0.75, 0.7, 0.65, 0.6, and 0.55, with a single set of species parameters. Under the respectively. Predation rates of intermediate and top species j on SRS and SRTS scenarios, results are qualitatively species i in community k are 0.6, 0.5, 0.4, 0.3, 0.2, and 0.1, robust to deviations from these parameter values. respectively. The initial proportion of sites occupied by species i in community k,(Pik) is set to 0.05. As noted, in closed metacommunities, tri-trophic FCs show an extreme variation in local < Summary and discussion abundances for both the basal and top species (Pik 0.01) in two and three communities, respectively. On the other hand, OMN shows It is well known that local communities can be the highest intercommunity variance for high dispersal rates (d ¼ 1). structured by both local and regional interactions The top species becomes unstable, and goes extinct in two local communities (P < 0.01). (Ricklefs 1987). However, it still remains unknown ik what trophic structures are shared by a set of interacting communities and its dynamical implications for the persistence of biodiversity. immigration (Karlson and Cornell 2002). It has The present study is an attempt to link local and been recently shown that in the regional food-web structure and dynamics in a Caribbean strongly influence the local community spatially structured marine food web. structure of fish on neighboring coral reefs Communities in five habitats of the Caribbean (Mumby et al. 2004). Additionally, empirical have shown significantly similar trophic structures studies have shown that dispersal among habitats which suggest that these communities are open to and local species interactions are key factors for 24 AQUATIC FOOD WEBS metacommunity structure (Shurin 2001; Cottenie In summary, the similarity in the trophic mod- et al. 2003; Kneitel and Miller 2003; Cottenie ules reported here suggests a strong link among and De Meester 2004), and the persistence of local the spatially structured communities. The level of and regional diversity (Mouquet and Loreau 2003). connectivity among these local communities and However, it still remains unclear how the interplay the type of trophic modules alter local abundance between dispersal and more complex trophic of species and promote local changes in diversity. structures affects species persistence in local com- It still remains unexplored how the results here munities ( et al. 2002; Kneitel and Miller 2003). presented change by the introduction of a larger In the present work, closed communities (d ¼ 0) number of interacting modules in a set of spatially with tri-trophic FCs showed an extreme variation structured communities. Our result predicts a in local abundances for both the basal and top relative stability in the composition of basal species (Figure 2.2(a)). On the other hand, OMN species, and a dramatic influence in the abundance shows the highest intercommunity variance for of top species depending on the connectivity high dispersal rates (d ¼ 1). The top species (i.e. dispersal) among distinct habitats. becomes unstable, and goes extinct in two local communities (Figure 2.2(b)). Recent empirical stu- Acknowledgments dies have shown that increasing dispersal fre- quency in intermediate species decreases the We thank the editors of this book for inviting us variance among local communities (Kneitel and to contribute this chapter. We thank Miguel A. Miller 2003), a pattern consistent with theoretical Fortuna and Mayte Valenciano for their useful results presented here (see dotted line in Figure comments on a previous draft. Funding was 2.2(a) and (b)). Further data synthesis and theore- provided by the Spanish Ministry of Science tical work is needed here to integrate the func- and Technology (Grants REN2003-04774 to JB and tional links between habitats and the local REN2003-00273 to PJ, and Ph.D. Fellowship dynamics of species embedded in food webs. FP2000-6137 to CJM). CHAPTER 3 Role of network analysis in comparative ecosystem ecology of estuaries

Robert R. Christian, Daniel Baird, Joseph Luczkovich, Jeffrey C. Johnson, Ursula M. Scharler, and Robert E. Ulanowicz

Introduction improve our inferences on trophic structure and dynamics. Assessments of trophic structure through eco- Estuaries are excellent ecosystems to test the logical network analysis (ENA) have been done in veracity of the inferences of ENA for three reasons. a wide variety of estuarine and coastal environ- First, more NAs have been conducted on estuaries ments. For example, some have used it to compare than on any other kind of ecosystem. Second, trophic structures within ecosystems focusing estuarine environments are often stressed by on temporal conditions (Baird and Ulanowicz natural and anthropogenic forcing functions. This 1989; Baird et al. 1998) and among ecosystems affords opportunities for evaluating controls on focusing on spatial conditions (e.g. Baird and trophic structure. Third, sampling of estuaries has Ulanowicz 1993; Christensen 1995). These compar- often been extensive, such that reasonable food isons have used carbon or energy as the currency webs can be constructed under different condi- with which to trace the interactions of the food tions of stress. Finally, other modeling approaches webs, although other key elements such as have been used in numerous estuarine ecosystems. nitrogen and phosphorus have also been used in Results of these alternate modeling approaches can ENA (Baird et al. 1995; Ulanowicz and Baird 1999; be compared to those of ENA to test the coherence Christian and Thomas 2003). One of the primary of inferences across perspectives of ecosystem features of ENA is that the interactions are structure and function. These conditions set the weighted. That is, they represent rates of flow of stage for an evaluation of the status of ENA as a energy or matter and not simply their existence. tool for comparative ecosystem ecology. Other kinds of comparisons have been attempted Comparative ecosystem ecology makes valuable less frequently. Effects of currency used to track contributions to both basic ecology and its applica- trophic dynamics has received little attention tion to environmental management. Given the (Christian et al. 1996; Ulanowicz and Baird 1999), critical position of estuaries as conduits for mater- and comparisons of ENA with other modeling ials to the and often as sites of intense approaches are quite rare (Kremer 1989; Lin et al. human activities in close proximity to important 2001). There is a need to expand the applications natural resources, ENA has been used frequently of network analysis (NA) to address specific for the assessment of the effects of environmental questions in food-web ecology, and to use it conditions within estuaries related to management. more frequently to explain and resolve specific Early in the use of ENA in ecology, Finn and management issues. The NA approach must be Leschine (1980) examined the link between fertiliza- combined with other existing methods of identi- tion of saltmarsh grasses and shellfish production. fying ecosystem performance to validate and

25 26 AQUATIC FOOD WEBS

Baird and Ulanowicz (1989) expanded the detail Fath and Patten 1999), provides some of the same accessible in food webs and the consequences of analyses found in NETWRK4 but includes others this increased detail in their seminal paper of based on the theoretical considerations of how seasonal changes within the . systems interact with their environment. Lastly, In 1992, Ulanowicz and Tuttle determined through social NA is beginning to be applied to ecological ENA and field data that the overharvesting of systems. A software package so used is UCINET may have had significant effects on a (www.analytictech.com/ucinet.htm; Johnson et al. variety of aspects of the food web in Chesapeake 2001; Borgatti et al. 2002). Although several Bay. Baird and Heymans (1996) studied the methods and software packages exist for evaluat- reduction of freshwater inflow into an ing weighted food webs, none has been developed in South Africa and noted changes in food-web and validated to an extent to give a good under- structure and trophic dynamics. More recently, standing of the full implications of the variety of Brando et al. (2004) and Baird et al. (2004) evalu- results. ated effects of and its symptoms on We have organized this chapter to address the Orbetello , Italy, and Neuse Estuary, use of ENA associated with estuarine food webs in USA, respectively. All of these studies involved the context of comparative ecosystem ecology. comparisons of conditions linked to human Comparisons within and among estuaries are first impacts. considered. ENA provides numerous output The first comprehensive review of the meth- variables, but we focus largely on five ecosystem- odologies and use of ENA, an associated software level variables that index ecosystem activity NETWRK4, and application in marine ecology and organization. We address the ability of was published in 1989 (Wulff et al. 1989). Other recognizing ecosystem-level change and patterns approaches to ENA have been developed and of change through the use of these indices. applied to food webs. The software programs Then we compare several estuarine food webs to and ECOSIM have been used through- budgets of biogeochemical cycling to assess the out the world to address various aspects of aquatic correspondence of these two facets of ecosystems. resources management (see www.ecopath.org/for Again we use these same indices and relate them summary of activities; Christensen and Pauly to indices from the biogeochemical budgeting 1993). In parallel with NETWRK4, ECOPATH was approach of –Ocean Interaction in developed by Christensen and Pauly (1992, 1995) the Coastal Zone (LOICZ) program. How do the and Christensen et al. (2000), based on the original two modeling approaches compare in assessing work of Polovina (1984). The dynamic simulation ecosystems? Finally, comparisons of food-web module, ECOSIM, was developed to facilitate diagrams are problematic if the food webs are the simulation of fishing effects on ecosystems at all complex. Recently, visualization tools (Walters et al. 1997). NETWRK4 and ECOPATH from biochemistry and social networks have include, to various extents, similar analytical tech- been used to portray food webs. We explore this niques, such as input–output analysis, Lindeman new approach in the context of intrasystem trophic analysis, a analysis, comparisons. and the calculation of information-theoretical indices to characterize organization and develop- Estuarine food-web comparisons ment. However, some analyses are unique to each. There are several differences in the input meth- We highlight how food webs are perceived to odology between the NETWRK4 and ECOPATH change or remain stable across a variety of condi- software, which lead to differences in their tions. First, we compare systems temporally from outputs. Heymans and Baird (2000) assessed intra and interseasonal to longer-term changes. these differences in a case study of the northern Within a relatively unimpacted ecosystem, food Benguela system. Environs analysis, webs may tend to be relatively stable with developed by Patten and colleagues (reviewed by differences among times related to altered, ROLE OF NETWORK ANALYSIS 27 -related metabolism and differential of each level represents the ingestion of the growth, migrations and ontogenetic changes in next level as a percentage of the ingestion of populations (Baird and Ulanowicz 1989). Human the focal level. The geometric mean of individual impacts may alter these drivers of change and add level efficiencies is the system’s TE. Ulanowicz new ones. Multiple food-web networks for an has characterized the degree of organization and ecosystem tend to be constructed under common maturity of an ecosystem through a group of sets of rules, facilitating temporal comparisons. information-based indices (Ulanowicz 1986). Then we compare food webs among ecosystems /developmental capacity (A/C) is a where major differences may exist in the very ratio of how organized, or mature, systems are, nature of the food webs. Interpreting such differ- where ecosystems with higher values reflect ences is more difficult than intrasystem com- relatively higher levels of organization. Thus, these parisons and must be viewed with more caution. five indices can be used to describe both extensive We have used studies of intersystem comparisons and intensive aspects of food webs. While our where effort was made by the authors to minimize focus is on these indices, we incorporate others as differences in rulemaking and network structure. appropriate to interpret comparisons. Should networks be constructed under different constraints, such as inconsistent rules for aggre- Temporal comparisons gation, the interpretation of differences in the NA results is difficult and should be viewed with more There are surprisingly few estuarine ecosystems caution. for which food-web networks have been examined Ecological network analysis provides a myriad during different times. Most networks represent of output variables and indices. Each has its annual mean food webs. We provide a brief review own sensitivity to differences in network of some for which we have direct experience structure. Generally, indices of population (i.e. at and can readily assess the focal ecosystem-level compartment-level) and cycling structure are more indices. These are ecosystems for which NETWRK4 sensitive than ecosystem-level indices in terms of was applied rather than ECOPATH, because of responsiveness to flow structure and magnitude of some differences in model construction and flows (Baird et al. 1998; Christian et al. 2004). Also, analysis (e.g. general use of gross primary pro- because currency and timescale may differ among duction in NETWRK4 and net primary production networks, direct comparisons using different in ECOPATH). The shortest timescale examined flow currencies are difficult. We focus on five has been for a winter’s Halodule wrightii ecosystem ecosystem-level output variables of ENA, four of in Florida, USA (Baird et al. 1998) where two which are ratios. These are described in greater sequential months were sampled and networks detail elsewhere (Kay et al. 1989; Christian and analyzed. Seasonal differences between food webs Ulanowicz 2001; Baird et al. 2004). The first adds were a central part of the Baird and Ulanowicz all flows within a network, total system through- (1989) analysis of the food web in Chesapeake put (TST), and reflects the size, through activity, Bay. Almunia et al. (1999) analyzed seasonal of the food web. Combinations of flows may be differences in Maspalomas Lagoon, , interpreted as occurring in cycles, and the per- following the cycle of domination by benthic centage of TST involved in cycling is called the versus pelagic primary producers. Florida Bay, Finn Cycling Index (FCI; Finn 1976). The turnover which constitutes the most detailed quantified rate of biomass of the entire ecosystem can be network to date, has been analyzed for seasonal calculated as the sum of compartment production differences (Ulanowicz et al. 1999). Finally, inter- values divided by the sum of biomass (P/B). decadal changes, associated with hydrological Networks can be collapsed, mathematically into a modifications, were assessed for the Kromme , or Lindeman Spine, with the proces- Estuary, South Africa (Baird and Heymans 1996). sing of energy or matter by each trophic level iden- Table 3.1 shows the five indices for each temporal tified (Ulanowicz 1995). The trophic efficiency (TE) condition for these ecosystems. 28 AQUATIC FOOD WEBS

Table 3.1 Temporal changes in ecosystem-level attributes for different estuarine ecosystems

Time period TST (mg C m2 per day) FCI (%) P/B (day1) TE (%) A/C (%)

St Marks, intraseasonal January 1994 1,900 16 0.037 4.9 36 February 1994 2,300 20 0.041 3.3 32 Neuse, intraseasonal Early summer 1997 18,200 14 0.15 5.0 47 Late summer 1997 17,700 16 0.30 4.7 47 Early summer 1998 18,600 16 0.24 3.3 47 Late summer 1998 20,700 16 0.33 4.9 46 Chesapeake, interseasonal 1,300,000 24 n.a. 9.6 45 Summer 1,700,000 23 n.a. 8.1 44 Fall 800,000 22 n.a. 10.9 48 Winter 600,000 23 n.a. 8.6 49 Maspalomas Lagoon, interseasonal Benthic-producer-dominated system 13,600 18 n.a. 11.4 40 Transitional 12,300 23 n.a. 12.8 38 Pelagic-producer-dominated system 51,500 42 n.a. 8.7 45 Florida Bay, interseasonal Wet 3,460 26 n.a. n.a. 38 Dry 2,330 n.a. n.a. n.a. 38 Kromme, interdecadal 1981–84 42,830 12 0.012 4.5 48 1992–94 45,784 10 0.011 2.8 46

Note: Flow currency of networks is carbon; n.a. means not available.

Ecological network analysis was applied to a and a 12% increase in P/B. However, organization winter’s H. wrightii ecosystem, St Marks National of the food web (A/C) decreased, and dissipation , Florida, USA (Baird et al. 1998; of energy increased lowering the TE. Although Christian and Luczkovich 1999; Luczkovich et al. statistical analysis of these changes was not done, it 2003). Unlike most applications of ENA, the field would appear that the indices do reflect perceived sampling design was specific for network con- effects of increased metabolism. struction. From these data and from literature The food web of the Neuse River Estuary, NC, values, the authors constructed and analyzed one was assessed during summer conditions over two of the most complex, highly articulated, time-and years (Baird et al. 2004; Christian et al. 2004). The site-specific food-web networks to date. Two Neuse River Estuary is a highly eutrophic estuary sequential months within the winter of 1994 were with high primary production and long residence sampled with the increase of 5C times of water. Temperature was not considered to from January to February. Metabolic rates, calcu- differ as dramatically from early to late summer, lated for the different and migrations but two major differences distinguished early and of fish and waterfowl affected numerous attributes late summer food webs. First was the immigration of the food webs (Baird et al. 1998). The changes in and growth of animals to the estuary during the focal indices are shown in Table 3.1. Activity summer, which greatly increased the biomass of estimated by the three indices was higher during several compartments. Second, the warmer period with >20% more TST, and FCI, commonly occurs during summer, stressing both ROLE OF NETWORK ANALYSIS 29 nekton and . Hypoxia was more dramatic increased during the pelagic phase (Table 3.1). in 1997 (Baird et al. 2004). Benthic biomass The proportional increase in TST could be inter- decreased during both summers, but the decrease preted as eutrophication, but the system has no big was far more dramatic during the year of more sources of material input from outside the system. severe hypoxia. Changes in the ecosystem-level Almunia et al. (1999) explained the increase in A/C indices were mostly either small or failed to show as a shift in resources from one subsystem the same pattern for both years (Table 3.1). A/C (benthic) to another (pelagic). The FCI was lowest changed little over summers or across years. TST, during the benthic-dominated stage and highest FCI, and TE had different from early to late during the pelagic-dominated stage, and matter summer for the two years. Only P/B showed was cycled mainly over short fast loops. The relatively large increases from early to late pelagic-dominated stage was interpreted as being summer. Thus, inferences regarding both activity in an immature state, but this interpretation is and organization across the summer are not counter to the highest A/C during the pelagic readily discerned. We have interpreted the results stage. The average TE dropped from the benthic- to indicate that the severe hypoxia of 1997 reduced dominated stage to the pelagic-dominated stage, the overall activity (TST) by reducing benthos and and the ratio of detritivory to herbivory increased their ability to serve as a food resource for nekton. accordingly. Highest values of detritivory coin- But these ecosystem-level indices do not demon- cided with lowest values of TE. strate a stress response as effectively as others Florida Bay showed remarkably little change considered by Baird et al. (2004). in whole-system indices between wet and dry The food web in Chesapeake Bay was analyzed (Ulanowicz et al. 1999; www.cbl.umces. for four seasons (Baird and Ulanowicz 1989). Many edu/bonda/FBay701.html). Although system-level of the changes linked to temperature noted for the indices during the wet season were about 37% within-season changes of the food web in St Marks greater than the same indices during the dry hold here (Table 3.1). TST and P/B are highest in season, it became apparent that this difference was summer and lowest in winter, although the other almost exclusively caused by the change in system measure of activity, FCI, does not follow this activity (measured as TST), which was used to pattern. However, FCI is a percentage of TST. scale the system-level indices to the size of the The actual amount of cycled flow (TST FCI) does system. The fractions of A/C and the distribution follow the temperature-linked pattern. TE failed to of the different components of the overhead were show a pattern of increased dissipation with almost identical during both seasons. Ulanowicz higher temperatures, although it was lowest et al. (1999) concluded that the Florida Bay eco- during summer. Organization, as indexed by A/C, system structure is remarkably stable between the showed the greatest organization in winter and two seasons. (FCI was high during the wet season least in summer. Hence, in both of the aforemen- (>26%) but could not be calculated for the dry tioned examples, times of higher temperature and season since the computer capacity was exceeded therefore, higher rates of activity and dissipation by the amount of cycles (>10 billion).) of energy were linked to transient conditions of Lastly, we consider a larger timescale of a decreased organization. These findings are corrob- decade for the Kromme Estuary, South Africa. orated for spring—summer comparisons of these Freshwater discharge to this estuary was greatly food webs are discussed later in the chapter. reduced by 1983 due to water diversion and Maspalomas Lagoon, Gran Canaria, shows, over damming projects, greatly lessening nutrient the year, three successive stages of predominance additions, salinity gradients, and pulsing (i.e. flood- of primary producers (Almunia et al. 1999). The ing; Baird and Heymans 1996). Can ecosystem- system moves from a benthic-producer-dominated level indices identify resultant changes to the via an intermediate stage to a pelagic- web? Although there was a slight increase in TST, producer-dominated system. The analysis of the trend was for a decrease in all other measures system-level indices revealed that TST and A/C (Table 3.1). However, all of these were decreases of 30 AQUATIC FOOD WEBS less than 20%, with the exception of TE. This and processes among systems of different geo- general, albeit slight, decline has been attributed to graphic locations, ranging from studies on estuaries the stress of the reduced flow regime (Baird and in relatively close proximity (Monaco and Heymans 1996). The TE decreased during the Ulanowicz, 1997; Scharler and Baird, in press) decade to less than half the original amount. to those of estuarine/marine systems spanning Thus, much less of the primary production was three (Baird et al. 1991). Perhaps inferred to pass to higher, commercially important, the most extensive comparison has been done trophic levels. Further, TE of the Kromme under a by Christensen (1995) on ecosystems using reduced flow regime was among the smallest for ECOPATH to evaluate indices of maturity. These ecosystems reviewed here. comparisons are limited, as discussed previously, In summary, most temporal comparisons because of differences in rules for constructing and considered were intra-annual, either within or analyzing networks. We review here some of the among seasons. Seasons did not have comparable estuarine and coastal marine comparisons that meaning among ecosystems. The Chesapeake have taken into account these issues, beginning Bay networks were based on solar seasons, but with our focal indices. Maspalomas Lagoon and Florida Bay networks The geographically close Kromme, Swartkops, were not. All indices demonstrated intra-annual and Sundays Estuaries, differ in the amount of change, although the least was associated with A/C. freshwater they receive, and consequently in the This is to be expected as both A and C are logar- amount of nutrients and their habitat structure ithmically based indices. Summer or warmer (Scharler and Baird 2003). Input–output analysis seasons tend to have higher activity (TST and FCI highlighted the differences in the dependencies or FCI TST), as expected. In some cases this was (or extended diets) of exploited fish and inverteb- linked to lower organization, but this was not rate bait species. were found to play an consistent across systems. We only include one important role in the Sundays Estuary (high interannual, actually interdecadal, comparison, but freshwater and nutrient input) as a food source to differences within a year for several systems were exploited fish and bait species, as great as those between decades for the Kromme whereas detritus and detritus producers were of Estuary. Interannual differences in these indices comparatively greater importance in the Kromme for other coastal ecosystems have been calculated (low nutrients) and Swartkops (pristine freshwater but with different currencies and software (Brando inflow, high nutrients) Estuaries (Scharler and et al. 2004; others). Elmgren (1989) has successfully Baird, in press). used trophic relationships and production esti- When comparing some indicators of system mates to assess how eutrophication of the Baltic performance such as TST, FCI, A/C, and TE of Sea over decades of enhanced nutrient loading has the Kromme, Swartkops, and Sundays Estuaries, modified production at higher trophic levels. Even it revealed an interplay between the various though the sample size remains small, it appears degrees of physical and chemical forcings. The that intra-annual changes in food-web structure Kromme Estuary is severely freshwater starved and trophic dynamics can equal or exceed those and so lacks a frequent renewal of the nutrient across years and across different management pool. Freshets have largely disappeared as a phy- regimes. Obviously, more examples and more sical . The Sundays system features thorough exploration of different indices are nee- increased freshwater input due to an interbasin ded to establish the sensitivities of ecosystem-level transfer, and the Swartkops Estuary has a relat- indices to in ecosystem condition. ively pristine state of the amount of freshwater inflow but some degree of anthropogenic Interecosystem comparisons (Scharler and Baird, in press). NA results showed that the Swartkops was more impacted Ecological network analysis has been used in inter- due to a low TST and a high average residence system comparisons to investigate the structure time (ART, as total system biomass divided by ROLE OF NETWORK ANALYSIS 31 total outputs) of material and least efficient to pass estuarine and marine upwelling systems. They on material to higher trophic levels. The Kromme used a slightly different approach, in that only the was more self-reliant (higher FCI) than the Sundays residual flow matrices (i.e the straight through (lowest FCI). The Sundays was also the most active flows) were used for this calculation, since the featuring a comparatively high TST and low ART. cycled flows were believed to inflate the inputs However, a comparatively high ascendency in the to the various end compartments. In this study, Sundays Estuary was not only a result of the high the most productive systems in terms of producing TST, which could have implicated the con- planktivorous fish from a unit of primary pro- sequences of eutrophication (Ulanowicz 1995a), duction were the upwelling systems (Benguela and but it also featured the highest AMI (the informa- Peruvian) and the Swartkops Estuary, compared to tion-based component of ascendency)(Scharler and the Baltic, Ems, and Chesapeake (Baird et al. 1991). Baird, in press). The Kromme Estuary had the In terms of carnivorous fish, the Benguela upwel- comparatively lowest A/C, and lowest AMI, ling system was the most efficient, followed by the and Baird and Heymans (1996) showed that since Peruvian and Baltic (Baird et al. 1991). the severe freshwater inflow restrictions, a decline Trophic efficiencies have also been used to make of the internal organization and maturity was assumptions about the productivity of a system. apparent. In perhaps the first intersystem comparison using Intercomparisons of estuaries and coastal aquatic NA, Ulanowicz (1984) considered the efficiencies ecosystems have often focused on other issues with which primary production reached the top in addition to the focal indices of this chapter. predators in two gut ecosystems in Crystal One important issue has been the secondary River, Florida. Monaco and Ulanowicz (1997) production of ecosystems, which is of special identified that fish and macroinvertebrate catches interest in terms of commercially exploited species. in the Chesapeake Bay were higher compared to As Monaco and Ulanowicz (1997) stated, there can the Narragansett and Delaware Bay, despite be differences in the efficiency of the transforma- its lower system biomass, because the transfer tion of energy or carbon from primary production efficiencies between trophic levels were higher. to the commercial species of interest. By relating Similarly, transfer efficiencies calculated from the output of planktivorous and carnivorous fish, material flow networks were used to estimate the and that of suspension feeders to primary produc- primary production required to sustain global tion, it became apparent that in Narragansett Bay fisheries (Pauly and Christensen 1995). Based on twice as many planktivorous fish and 4.6–7.4 as a mean energy transfer efficiency between trophic many carnivorous fish were produced per unit levels of 48 ecosystems of 10%, the primary primary production than in Delaware or Chesapeake production required to sustain reported catches Bay, respectively. The latter, on the other hand, and was adjusted to 8% from a previous produced 1.3 and 3.5 as much suspension feeding estimate of 2.2%. biomass than Narragansett and Delaware Bay In the context of the direct and indirect diet of from one unit of phytoplankton production exploited and other species, it can be of interest (Monaco and Ulanowicz 1997). This analysis was to investigate the role of benthic and pelagic performed on the diet matrix to quantify a con- compartments. The importance of benthic pro- tribution from a compartment (in this case the cesses in the indirect diet of various age groups primary producers) in the network to any other, of harvestable fish was determined with input– over all direct and indirect feeding pathways, and output analysis by Monaco and Ulanowicz (1997). is described as part of an input–output analysis in The indirect diet is the quantified total consump- Szyrmer and Ulanowicz (1987). tion by species j that has passed through species i This approach of tracing the fate of a unit along its way to j (Kay et al. 1989). They found of primary production through the system was that benthic processes in the Chesapeake Bay also applied by Baird et al. (1991) who calculated was highly important to particular populations of the fish yield per unit of primary production in juvenile and adult . Indirect material 32 AQUATIC FOOD WEBS transfer effects revealed that the Chesapeake Bay pointed out that the system P/B ratio is not relied more heavily on its benthic compartments necessarily a reflection of the maturity of the sys- compared to the Narragansett and Delaware Bays tem, but due to NA results reinterpreted maturity and that disturbances to benthic compartments in the context of physical forcing (e.g. the upwel- may have a comparatively greater impact on the ling systems (Peruvian, Benguela) are considered system (Monaco and Ulanowicz 1997). The pattern to be mature under their relatively extreme phy- changes somewhat with season, as discussed in sical forcings, although they have a higher system the section, ‘‘visualization of network dyna´mics’’ P/B ratio than the estuarine systems (Chesapeake, of this chapter. Ems, Baltic, Swartkops)). The shallow Kromme, Swartkops, and Sundays Comparison of whole-system indices between Estuaries were found to rely more on their benthic the Chesapeake and Baltic ecosystems provided biota in terms of compartmental throughput and managers with a surprise (Wulff and Ulanowicz the total contribution coefficients in terms of 1989). The conventional wisdom was that the compartmental input (Scharler and Baird, in Baltic, being more oligohaline than the Chesapeake, press). In terms of carbon requirements, the would be less resilient to stress. The - Kromme and Swartkops Estuaries depended two- izational status of the Baltic, as reflected in the third on the benthic components and one-third on relative ascendency (A/C) was greater (55.6%) the pelagic components, whereas the Sundays than that of the Chesapeake (49.5%) by a significant Estuary depended to just over half on its benthic amount. The relative redundancy (R/C) of the components. The Sundays Estuary was always Chesapeake (28.1%) was correspondingly greater perceived to be ‘‘pelagic driven,’’ probably due to than that of the Baltic (22.0%), indicating that the high phyto and zooplankton standing stocks, the Chesapeake might be more stressed than the which are a result of the regular freshwater and Baltic. The FCI in the Chesapeake was higher nutrient input. By considering not only direct (30%) than in the Baltic (23%). As greater cycling is effects, but also all indirect effects between the indicative of more mature ecosystems (Odum compartments, the regular freshwater input sup- 1969), this result seemed at first to be a counter- pressed somewhat the dependence on benthic indication that the Chesapeake was more stressed, compartments, but has not switched the system to but Ulanowicz (1984) had earlier remarked that a predominantly pelagic dependence (Scharler and a high FCI could actually be a sign of stress, Baird, in press). especially if most of the cycling occurs over short Indicators of stress, as derived from ENA, have cycles near the base of the trophic ladder. This been discussed in several comparative studies. was also the case in this comparison, as a decom- Baird et al. (1991) proposed a distinction between position of cycled flow according to cycle length physical stress and chemical stress. The former has revealed that indeed most of the cycling in the in general been influencing ecosystems, such as Chesapeake occurred over very short cycles (one upwelling systems, for a time long enough so that or two components in length), whereas recycle the systems themselves could evolve under the over loops that were three or four units long was influence of this type of physical forcing. Fresh- significantly greater in the Baltic. The overall water inflow into estuaries similarly determines picture indicated that managerial wisdom had the frequency of physical disturbance, due to been mistaken in this comparison, as the saltier frequent flooding in pristine systems and restric- Chesapeake was definitely more disrupted than tions thereof in impounded systems. On the other the Baltic. hand, chemical influences are in general more recent through anthropogenic pollution, and the Intermodel and technique comparisons systems are in the process of changing from one response type (unpolluted) to another (polluted) Another modeling protocol was developed that adjusts to the chemical type of forcing (Baird under the auspices of the International Geosphere– et al. 1991). With this perspective, Baird et al. (1991) Biosphere Program (IGBP), an outcome of the 1992 ROLE OF NETWORK ANALYSIS 33

Rio Summit and established in 1993. The inorganic P), one can derive whether (1) an estuary aims of the IGBP are ‘‘to describe and understand is a sink or a source of C, N, and P, that is D the physical, chemical and biological processes Y ¼ fluxout fluxin, where Y ¼ C, N, or P; (2) that regulate the earth system, the environment the system’s metabolism is predominantly auto- provided for life, the changes occurring in the trophic or heterotrophic, that is, (p r) ¼ D system, and the influence of human actions.’’ DIP(C : P)part, where (p r) is In this context, the Land Ocean Interactions in the minus respiration; and (3) nitrogen fixation (nfix) Coastal Zone (LOICZ) core project of the IGBP was or denitrification (denit) predominates in the sys- D D established. LOICZ focuses specifically on the tem, where (nfix denit) ¼ DIN DIP(N : P)part functioning of coastal zone ecosystems and their (Gordon et al. 1996). A summary of attributes for role in the fluxes of materials among land, sea, and this modeling approach is shown in Table 3.2. atmosphere; the capacity of the coastal ecosystems This section explores the possibility of linkages to transform and store particulate and dissolved between the two different methodologies of ENA matter; and the effects of changes in external and LOICZ biogeochemical budgeting protocol. forcing conditions on the structure and functioning The rationale for this hypothesis is: of coastal ecosystems (Holligan and de Boois 1993; Pernetta and Milliman 1995). The magnitude and frequency of N and P loadings and The LOICZ biogeochemical budgeting proce- the transformation of these elements within the system, ultimately affect the system’s function. Since we postulate dure was subsequently developed that essentially that system function is reflected in network analysis consists of three parts: budgets for water and salt outputs, we infer that there should exist correspondence movement through coastal systems, calculation of in the biogeochemical processing, as indexed by the rates of material delivery (or inputs) to and LOICZ approach, and trophic dynamics, as indexed by removal from the system, and calculations of rate network analysis outputs. of change of material mass within the system (particularly C, N, and P). Water and salt are To do this, we used ENA and LOICZ variables considered to behave conservatively, as opposed and output results from six estuarine or brackish to the nonconservative behavior of C, N, and P. ecosystems based on input data with a high level Assuming a constant stoichiometric relationship of confidence (Table 3.3). We first performed (e.g. the Redfield ratio) among the nonconservative Spearman’s and Kendall’s correlation analyses nutrient budgets, deviations of the fluxes from the between the ENA and LOICZ output results of a expected C : N : P composition ratios can thus be number of system indices of the six ecosystems. assigned to other processes in a quantitative From the correlation matrices we selected those fashion. Using the flux of P (particularly dissolved variables which showed correlation values of 80%

Table 3.2 System properties and variables derived from NA and the LOICZ biogeochemical budgeting protocol used in

LOICZ variables Description of variable/system property

Nutrient loading From land to ocean, two macronutrients and their possible origins Dissolved inorganic nitrogen (DIN) (mol m2 per year) Products of biogeochemical reactions Dissolved inorganic phosphorus (DIP) (mol m2 per year) Materials responding to human production, that is, domestic (animal, human) and industrial waste, and sewage, , atmospheric fallout from vehicular and industrial emissions 2 DIN (mol m per year) Fluxout Fluxin 2 DIP (mol m per year) Fluxout Fluxin Net ecosystem metabolism (NEM) (mol m2 per year) Assumed that the nonconservative flux of DIP is an approximation of net metabolism: (p r) ¼DIP(C : P) NFIXDNIT (mol m2 per year) Assumed that the nonconservative flux of DIN approximates N fixation minus denitrification: (nfix denit) ¼ DIN DIP(N : P) 34 AQUATIC FOOD WEBS

Table 3.3 Ecosystem-level attributes used for comparison of NA results of estuarine food webs with biogeochemical budgeting models

System Morphology ENA

Volume (m3) Area (m2) A/C (%) TST (mgC m2 per day) FCI (%) P/B (day1) TE (%)

Kromme 9.00Eþ06 3.00Eþ06 33.7 13,641 26 0.73 6.2 Swartkops 1.20Eþ07 4.00Eþ06 28 11,809 44 3.65 4 Sundays 1.40Eþ07 3.00Eþ06 43 16,385 20 10.95 2.6 1.74Eþ13 3.70Eþ11 55.6 2,577 23 29.2 16.2 Cheasapeake 3.63Eþ10 5.90Eþ09 49.5 11,224 23 51.1 9 Neuse 1.60Eþ09 4.60Eþ08 46.8 11,222 15.4 94.9 4.5

LOICZ models (mol m2 per year)

DIP DIN (nfixdenit) (pr) NEM DIN loading DIP loading GPPa

Kromme 6.80E03 1.59E01 4.93E01 7.27E01 3.01E02 7.30E04 2.15Eþ02 Swartkops 6.25E02 1.01Eþ01 9.13Eþ00 6.65Eþ00 1.40Eþ00 6.85E02 1.10Eþ02 Sundays 4.70E03 2.26E01 3.02E01 5.03E01 1.43Eþ00 1.18E02 8.53Eþ03 Baltic Sea 9.41E03 1.52E01 1.17E01 5.12E01 1.33E01 1.25E03 2.50Eþ03 Cheasapeake 1.42E02 5.95E01 2.78E01 1.51Eþ00 4.82E01 1.00E02 6.27Eþ03 Neuse 1.69E04 1.64E03 4.02Eþ01 6.55Eþ00 1.77E01 1.56E02 5.04Eþ03

a GPP is in C and was also calculated through stochiometry for N and P. These values are not included here. and higher, on which we subsequently performed Table 3.4 Unrotated factor loadings of the selected system factor analysis. The system properties and the variables listed in Table 3.1 values of the ecosystem properties on which the Variables/property Principal component factor analysis was based are given in Table 3.3. The output from factor analysis yielded eigen 1234 values of six principal components, of which the Network variables first four principal components account for 98.4% A/C 0.79 0.31 0.13 0.51 of the variance between the system properties. FCI(%) 0.94 0.04 0.10 0.19 The factor loadings for each of the LOICZ and NA TST 0.08 0.81 0.25 0.49 variables are given in Table 3.4, and taking þ0.7 P/B(day1) 0.55 0.02 0.81 0.03 and 0.7 as the cutoff values, certain variables are Trophic Efficiency (%) 0.23 0.85 0.07 0.46 correlated with one another and can be interpreted GPP-C 0.70 0.61 0.05 0.35 as varying together on these principal factors. GPP-N 0.69 0.65 0.04 0.30 The first three principal components explain 87.7% GPP-P 0.70 0.65 0.03 0.29 of the variance and none of the factor loadings of LOICZ variables the fourth principal component exceeded the cut- DIN loading 0.47 0.76 0.20 0.37 off value, so this factor was not considered further. DIP loading 0.87 0.34 0.32 0.17 A number of inferences can be made: DIP 0.89 0.03 0.26 0.37 DIP 0.94 0.17 0.19 0.20 1. The first principal component explains 46% of the (nfixdenit) 0.55 0.10 0.75 0.34 variance and which includes three LOICZ and three (pr) NEM 0.44 0.19 0.88 0.02 ENA variables. Table 3.4 (under the first principal Note: Four principal components are extracted (columns 1–4). component) shows that of the LOICZ-derived variables DDIN and DDIP correlate negatively with The FCI correlates negatively with the A/C DIP loading, which means that the magnitude of and carbon GPP (gross primary productivity) of DIP loading will somehow affect the flux of DIN the ENA-derived properties, and one can thus and DIP between the estuary and the coastal sea. expect lower FCI values in systems with high A/C. ROLE OF NETWORK ANALYSIS 35

This inverse relationship has in fact been reported 3. The third principal component, which accounts in the literature (cf. Baird et al. 1991; Baird 1998). for 17% of the variance (Table 3.4) shows posit- From the linkage between the ENA and the LOICZ ive correlations between two LOICZ variables modeling procedure, we can infer from these ((nfix-denit), net ecosystem metabolism (p r)), results that there appears to be a positive correla- and one ENA system-level property, the P/B. tion among DIN and DIP flux, GPP, and ascen- We can construe from these relationships that the dency. Systems acting as nutrient sinks may thus P/B is influenced by the magnitude and nature of well be positively associated with GPP and ascen- one or both of the two LOICZ-derived properties. dency, and such systems are thus more productive (higher GPP) and organized (higher A/C). The data The underlying associations are summarized in given in Table 3.3 show to some degree that the a scatter plot of the ecosystem positions relative to Baltic Sea, the Chesapeake Bay and the Neuse the first two principal components (Figure 3.1) and River Estuary have high A/C associated with their a cluster (Figure 3.2), which essentially reflects performance as nutrient sinks. the results from factor analysis presented above. 2. Of the variance, 25% is explained by the second The three systems in the middle of Figure 3.1 principal component, which had high factors occupy a relatively ‘‘neutral ’’ in the con- scores for one LOICZ-and two ENA-derived text of their responses to the variability of the NA variables (Table 3.4). The results would indicate and LOICZ parameters given on the x- and y-axes, some positive correlation between DIN loading and appears to relate to the analyses of Smith et al. and TST, but both are negatively associated with (2003) that a large proportion of the estuaries for the TE index (Table 3.3). which biogeochemical results are available cluster

2 ihFILow FCI Low GPP High FCI

Kromme 1 Swartkops

Baltic 0 Neuse

Chesapeake

–1 High GPP

Sundays –2 –3 –2 –1 012 Low High ∆DIP ∆DIN High DIN Low DIP Loading

Figure 3.1 System position within the plane of the first two principal components. 36 AQUATIC FOOD WEBS

Cluster tree In addition, the FCI from NA and DIP loading TST from LOICZ vary together as well, which suggests DIN loading that overall cycling is linked to phosphorous DIP loading in some way. This result is similar in many ways to FCI the factor analysis above, especially the variables NEM P/B that score highly on principal factors 1 and 3 nfix (Table 3.4). GPP-C The fundamental differences between the net GPP-N flux methodology of LOICZ modeling and the GPP-P gross flows of material inherent in food-web net- ∆DIP works must be kept in mind, but the correlation ∆DIN A/C between the methodologies is encouraging in our Trophic efficiency search for better understanding of ecosystems function. We should thus emphasize the possible 0.0 0.5 1.0 1.5 Distances linkages and the complimentary results derived from these methodologies. ENA results have rarely Figure 3.2 Cluster tree of ENA and LOICZ variables. been related to other approaches. Comparisons, such as this, are essential to broaden our under- around neutral values of (p r)(or NEM) and standing of how ecosystems function and are (nfix denit). These three systems, namely the structured in a holistic way. Baltic Sea, the Chesapeake Bay, and the Neuse River Estuaries are large in terms of aerial size and Visualization of network dynamics volume compared to the three smaller systems (namely the Swartkops, Kromme, and Sundays The of dynamic, complex food webs has Estuaries), which are scattered at the extreme been problematic in past, due to the multiple ranges of the variables. Table 3.3 shows that species and linkages that must be rendered. This the larger systems are bigger in volume and size display limitation has prevented the visualiza- by 3–5 orders of magnitude, but that the DIN and tion of changes that occur at the level of the whole DIP loadings on a per meter square basis of all six food web. Seasonal changes, changes over longer systems fall within in the same range. Other periods of time, impacts due to fishing or , noticeable differences are the shorter residence and pollution impacts can all affect food-web times of material, the small volume and low rate of structure, but unless this can be quantified and inflows compared with the three big- visualized, it is difficult for most to appreciate. ger systems. Although the scales of the axes in Most current approaches involve either simplify- Figure 3.1 are nondimensional, the positions of the ing the food web by aggregating species into various systems reflect the relative order of the and by displaying ‘‘wiring dia- four variables plotted on the x- and y-axis, grams’’ of the underlying structure. We use net- respectively, and corresponds largely with the work statistical modeling software to analyze empirical outputs from ENA and the LOICZ bud- the similarities in the food webs and display the geting protocol. Finally, a cluster tree (Figure 3.2), results using three-dimensional network modeling which shows the similarity of the variables using and visualization software. an average clustering of the Pearson correlation r We used the visualization technique described (as a distance measure ¼ 1 r), groups the NA and in Johnson et al. (2001, 2003) and Luczkovich et al. LOICZ variables in a hierarchical manner. Using a (2003) to display a series of food webs of the distance of <0.2 as a cutoff, the P/B ratio from Chesapeake Bay, originally described by Baird NA is closely grouped with the [nfix denit] of and Ulanowicz (1989). This technique involves LOICZ, which suggests that overall production arranging the nodes (species or carbon storage and nitrogen balance is linked in these estuaries. compartments) of the food-web network in a ROLE OF NETWORK ANALYSIS 37 three-dimensional space according to their sim- Table 3.5 The compartments in the four seasonal models of ilarity in feeding and predator relationships, as Chesapeake Bay and their numbers measured by a model called regular equivalence. Compartment Spring Summer Fall Winter Trophic In the regular equivalence model, two nodes in name level close position in the three-dimensional graph have linkages to predator and prey nodes that them- 1 Phytoplankton x x x x 1.00 2 Bacteria in x x x x 2.00 selves occupy the same trophic role, but not suspended POC necessarily to the exact same other nodes. Thus, 3 Bacteria in x x x x 2.00 here we visualize the change in trophic role of the sediment POC compartments in Chesapeake Bay as they change 4 Benthic x x x x 1.00 from spring to summer. 5 Free bacteria x x x x 2.00 In the example we display here, Baird and 6 Heterotrophic x x x x 3.00 Ulanowicz (1989) modeled the carbon flow in a microflagellates 36-compartment food web of the Chesapeake Bay. 7 x x x x 2.75 The model was adjusted seasonally to reflect the 8 Zooplankton x x x x 2.16 measured changes in carbon flow among the com- 9 Ctenophores x x x x 2.08 partments. This model was originally constructed 10 Sea Nettle x 3.44 11 Other suspension x x x x 2.09 using the program NETWRK4. We obtained feeders the input data from the NETWRK4 model from 12 Mya arenaria x x x x 2.09 the original study and converted them to text 13 Oysters x x x x 2.08 data using a conversion utility from Scientific 14 Other x x x x 3.00 Committee on Oceanographic Research (SCOR) polychaetes format (Ulanowicz, personal communication). 15 Nereis sp. x x x x 3.00 The carbon flow data in a square matrix for each 16 Macoma spp. x x x x 3.00 season was imported into UCINET (Borgatti et al. 17 Meiofauna x x x x 2.67 2002) to compute the regular equivalence coeffi- 18 Crustacean x x x x 3.00 cients for each compartment (or node). Due to deposit feeders migrations and seasonal fluctuations in abundance, 19 Blue x x x x 3.51 20 Fish larvae x 3.16 the model had 33 compartments in spring, 36 in 21 Alewife and x x x x 3.16 summer, 32 in fall, and 28 in winter. They are blue herring listed in Table 3.5 along with their identifica- 22 Bay anchovy x x x x 2.84 tion codes and seasonal presence and absences. 23 Menhaden x x x x 2.77 The algorithm for computing regular equivalence 24 Shad x x 3.16 (REGE), initially places all nodes into the same 25 Croaker x x x 4.00 class and then iteratively groups those that 26 Hogchoker x x x x 3.91 have similar type of connections to predators and 27 Spot x x x 4.00 prey. Finally, a coefficient ranging from 0 to 1.00 28 White perch x x x x 3.98 is assigned to each node, which reflects their sim- 29 Catfish x x x x 4.00 ilarity in food-web role. These coefficients have 30 Bluefish x x x 4.59 31 Weakfish x x x 3.84 been found to have a relationship with trophic 32 Summer flounder x x x 3.99 level, as well as differentiate the benthos and 33 Striped bass x x x 3.87 plankton based food webs (Johnson et al. 2001; 34 Dissolved x x x x 1.00 Luczkovich et al. 2003). After the REGE coefficients organic carbon were computed, the matrices for each season were 35 Suspended POC x x x x 1.00 concantentated so that a 144 36 rectangular 36 Sediment POC x x x x 1.00 matrix of the coefficients was created. The com- Source: From Baird and Ulanowicz (1989). bined four-season REGE coefficient matrix was analyzed using a stacked correspondence analysis 38 AQUATIC FOOD WEBS

(Johnson et al. 2003), which makes a singular ing with the node number) and summer (black value of the rows and column data nodes with labels beginning with ‘‘SU’’ and ending in a multivariate space. We used the row scores with the node number) food web of the Chesapeake (the 36 compartments in each of the 4 seasons) to network shows groupings of nodes that have plot all 144 points in the same multivariate space. similar predator and prey relationships, so that The network and correspondence analysis coordinate they form two side groups at the base of the data were exported from UCINET to a coordinate web, and a linear chain of nodes stretching file so that the food web could be viewed in Pajek upwards (Figure 3.3). The arrows show the shift (Batagelj and Mrvar 2002). (Note: we have also in coordinate position from spring to summer used real time interactive molecular modeling (we omit the arrows showing carbon flow here software Mage for this purpose; see Richardson and for clarity). The vertical axis in this view (note: Richardson (1992)). Pajek was used to create the normally axis 1 is plotted along the horizontal, but printed versions of this visualization. we rotated it here to have high trophic levels at the The three dimensional display of the spring top) is correspondence analysis axis 1, which is (gray nodes with labels beginning ‘‘SP’’ and end- significantly correlated (r ¼ 0.72) with the trophic

(a) (Trophic Position) (Trophic Correspondence Axis Correspondence 1

Correspondence Axis 2 (Pelagic to Benthos)

(b) Figure 3.3 (a) The food-web network of the Chesapeake Bay in spring (gray ‘‘SP’’ node labels) and summer (black, ‘‘SU’’ node labels), displayed using Pajek. The arrows show the shift in coordinate position from spring to summer. The stacked correspondence analysis row scores were used to plot the positions in three-dimensional space. (b) Another view showing the shift along the first and third axes, which represent trophic position as before and

(Trophic Position) (Trophic degree of connectedness to the network.

Correspondence Axis Correspondence 1 The two compartments that were absent from the summer network: sea nettles (10) and other suspension feeders (11) are shown as moving into the center in the summer and becoming Correspondence Axis 3 connected to the network. ROLE OF NETWORK ANALYSIS 39 levels (Table 3.5) that were calculated based on in the summer. This can also be seen in the case annualized carbon flows for each compartment by of bay anchovy (22), which takes in more zoo- Baird and Ulanowicz (1989). In Figure 3.3 (a), there plankton (8), and spot (27) which increase is a group at the right side of the base of the web, consumption of ‘‘other polychaetes’’ (14) in the which is composed of compartments that are summer (Baird and Ulanowicz 1989). In all of these associated with detritus or the benthos, includ- cases, an increase in consumption of species with ing bacteria in the sediment particular organic lower trophic positions is driving this change in material (POC; 3), benthic diatoms (4), Nereis (15), the visualization. other polychaetes (14), crustacean deposit feeders Another interpretation of the coordinate move- (18), and sediment POC (36) (note: the number ments is that the species which derive energy from in the parenthesis is the serial number of each the in the summer are moving toward compartment in Table 5). On the left side of the the center on axis 2. For example, free bacteria (5) base of this web, there are plankton-associated and zooplankton (8) move toward the center of the groups, including phytoplankton (1), free bacteria diagram from spring to summer as they increase (5), heterotrophic microflagellates (6), zooplankton their consumption of (34) (8), and dissolved organic carbon (34), and sus- and ciliates (7), respectively, while crustacean pended POC (35). In the center at the base of the deposit feeders (18) move toward the center since web, we find the bacteria in the suspended POC (2), they consume less sediment POC (36) in the sum- which is midway between the benthic group and mer. Thus, the degree to which the whole ecosystem the plankton group, due to the fact that these shifts from benthic to pelagic primary production bacteria are important as food of consumers in can be easily visualized. This also can be visualized both groups. Stretching in a near-linear chain dynamically across multiple seasons. We do not above the base are various consumers that are show the other seasons here, but interactively, one higher trophic levels. Compartments with low can turn off and on a similar display for each season effective trophic levels (TL) include oysters and show that the nodes in the fall and winter move (13; TL ¼ 2.08), soft-shelled clams Mya arenaria back towards the springtime positions. This can also (12; TL ¼ 2.09), bay anchovies (22; TL ¼ 2.77), other be done over multiple years, if the data were suspension feeders (11; TL ¼ 2.09), and menhaden available, or in varying salinities, temperatures, and (23; TL ¼ 2.77). Higher on the correspondence under different management schemes. analysis vertical axis are compartments catfish (29), white perch (28), spot (27), hogchoker (26), Conclusions alewife and blueback herrings (21), summer flounder (32), striped bass (33), bluefish (30), shad Estuarine and coastal ecosystems have been loca- (24), and larval fish (20). tions where numerous studies have incorporated One way to interpret these visualizations is that ENA to assess food-web structure and trophic those compartments that move the most in the dynamics. ENA also affords a valuable approach coordinate space show the greatest seasonal to comparative ecosystem ecology. Numerous change in trophic roles. Species that move ecosystem-level indices are calculated and com- downward along axis 1 are consuming more of plement indices at lower level of . the primary production or consuming more prey Comparisons of five ecosystem-level indices of at low trophic levels. This is also shown for the food webs over various temporal and spatial scales whole system as higher TST and P/B ratios in appeared to correspond with our understanding of the summer (Table 3.1), but our visualization levels of development and stress within several shows the contribution of individual compart- estuarine systems. Intra-annual variations in these ments to the system-wide changes. Some good indices within an ecosystem were equal to or examples are other suspension feeders (11), which exceeded that for the limited number of cases of move downward on the trophic position axis, interannual comparisons. Interecosystem compar- because they feed more on the phytoplankton isons are more difficult because of differences in 40 AQUATIC FOOD WEBS rules for network construction used for different multiple conditions. Thus, we evaluated how ENA ecosystems, but patterns in calculated indices were can be used in comparative ecosystem ecology and consistent with expectations. Finally, two relatively offer two new approaches to the discipline. new approaches to understanding estuarine eco- systems, namely models of biogeochemical budgets and visualization tools were compared to Acknowledgments the focal indices. The biogeochemical modeling complemented the ecosystem-level network indi- We would like to thank several for ces, providing an extended assessment of the the support of this work. We thank the limited number of ecosystems evaluated. Visual- Foundation and the Biology Department of East ization of food webs is problematic when those Carolina University for support of Dan Baird as food webs are complex. This problem is exacer- Rivers Chair of International Studies during fall bated when one wants to compare food-web 2003. Funding for Christian came from NSF under structures. We demonstrated a relatively new grants DEB-0080381 and DEB-0309190. Ursula approach to visualizing food webs that enhances Scharler is funded by the NSF under Project one’s ability to identify distinctions between No. DEB 9981328. CHAPTER 4 Food webs in lakes—seasonal dynamics and the impact of climate variability

Dietmar Straile

Introduction indeed play a key role for an understanding of As a result of increased green-house gases, global the effects of on lakes. Although surface temperatures increased strongly during the food-web related impacts of climate change will twentieth century and will most likely increase probably be among the most difficult impacts to between 1.4–5.8C within the twenty-first century predict, they might have also some of the most (Houghton et al. 2001). Early signs of climate- severe consequences, for example, leading to related gradual changes in lake ecosystems have regime shifts (Sanford 1999; Scheffer et al. 2001a,b). been reported (Schindler et al. 1990; Magnuson et al. Studies on lakes provide some of the most 2000; Straile 2002; Livingstone 2003; Straile et al. complete investigations on the structure, dynamics, 2003) and further changes are expected to come. and energetics of food webs encompassing organ- Besides the increase in temperature, the increase in isms from bacteria to vertebrates. The life cycle of greenhouse gases per se (e.g. Urabe et al. 2003; organisms in temperate lakes is adapted to a highly Beardall and Raven 2004) as well as other asso- seasonal environment. Food-web interactions in ciated phenomena, for example, increase in UV-B these lakes depend on the seasonal overlap of the radiation (e.g. Schindler et al. 1996; Williamson occurrence of potential prey, competitor, or predator et al. 2002), and their interactions will affect species. This seasonal overlap, that is, the match– organisms in lake ecosystems (Williamson et al. mismatch of food-web interactions depends strongly 2002; Beardall and Raven 2004). This chapter can on the seasonal dynamics of the physical environ- obviously not cover in detail these issues in global ment of lakes such as temperature, light availability, change research; instead the emphasis will be on and mixing intensity. Consequently, climate vari- possible food-web effects of temperature increase ability influences food-web interactions and hence and variability. For recent reviews on various the structure, dynamics, and energetics of lake effects of climate forcing on lakes see (Carpenter food webs. On the other side, a thorough under- et al. 1992; Schindler 1997; Gerten and Adrian standing of climate effects on lakes will require 2002a; Straile et al. 2003). information on lake food webs and—as will be Studying spatial shifts in shown below—on the seasonal dynamics of lake and changes in —although they are food webs. Hence, this chapter is separated into important and do undoubtedly occur also in lake three sections: The section titled ‘‘The trophic struc- food webs (Straile et al. 2003; Briand et al. 2004)— ture of aquatic food webs’’ examines the trophic is not sufficient to predict the effects of climate structure of lake food webs, especially of lake food change on ecosystems as they do not consider webs as observed during the summer season. The indirect effects and feedbacks within the food web section ‘‘Seasonal sucession’’ analyses the seasonal (Schmitz et al. 2003). Food-web interactions may development of lake food webs, and hence provides

41 42 AQUATIC FOOD WEBS the more dynamical viewpoint necessary for exam- experiments (Carpenter and Kitchell 1993). Biomass ining the effects of climate on the food webs of lakes relationships in a suite of 11 Swedish lakes (Persson (section ‘‘Climate variability and plankton food et al. 1992) also largely corrobated the predictions webs’’). I will illustrate some aspects of lake food- of HSS and the food-chain theory in systems with web structure with examples from Lake Constance, three trophic levels (algae, zooplankton, plankti- which food web has been rather well studied vorous fish) and with four trophic levels (algae, during recent years (see Box 4.1). zooplankton, planktivorous fish, piscivorous fish). The success of HSS in aquatic systems is The trophic structure of aquatic evidence that the complexity of aquatic food webs food webs can indeed be aggregated into trophic levels (Hairston and Hairston 1993). Nevertheless, this The most successful models on food-web regula- assumption has hardly been tested explicitly with tion in pelagic lake food webs are based on the data from food webs. Recently, Williams and assumption that food webs can be aggregated into Martinez (2004) used highly resolved carbon flow trophic chains, that is, that trophic levels act models to provide a first test: based on mass- dynamically as populations. These assumption has balanced carbon flow models, they estimated the its origin in the famous ‘‘green world’’ hypothesis trophic position of web components by computing of Hairston, Smith, and Slobodkin (1960) (hereafter their food-chain length and their relative energetic referred to as HSS) although aquatic systems were trough chains of different length (Levine not discussed by HSS. According to HSS, the 1980). They suggested that most species can be relative importance of competition and predation assigned to trophic levels and that the degree of alternates between trophic levels of food webs omnivory appears to be limited. To provide such that carnivores compete, herbivores are con- another test of this idea, seasonally resolved carbon trolled by predation, which frees producers from flow models, which were established for the predation and results into competition at the pelagic food web of Lake Constance (Box 4.1) can producer trophic level. Similar ideas developed be used. To calculate trophic positions in the Lake more or less simultaneously in freshwater ecology. Constance food-web basal trophic positions of Already in 1958, Hrba´cˇek (1958) provided seminal phytoplankton and detritus/DOC were set to one. observations on trophic level dynamics in lakes: The latter implies that bacteria, which derive their he noted the relationship between fish predation, nutrition from the detritus/DOC pool, were the reduction of large-bodied herbivores, and the assigned to a trophic position of two. As a con- resulting decrease in transparency and increase sequence, two different flow chains can be recog- in phytoplankton density in lakes. An increasing nized: a grazing chain starting from phytoplankton number of subsequent studies used the aggrega- and a detritus chain starting from detritus, tion of food webs into trophic levels as a tool respectively bacteria (Figure 4.2(a)). However, as to analyze food-web interactions especially in (a) in Lake Constance, bacterial production is plankton food webs. The theory of trophic level considerably lower than primary production, and dynamics was further developed into the concepts (b) one group, that is, heterotrophic nanoflagellates of the (Carpenter et al. 1985) and (HNF) are the major consumers of bacterial pro- (Benndorf et al. 1984; Shapiro and duction, a strong impact of the detritus chain on Wright 1984). Food chain theory (Oksanen et al. the trophic position of consumers is only evident 1981) formally extended HSS ‘‘green world for HNF. All other groups rely energetically hypothesis’’ to more than three trophic levels directly (ciliates, rotifers, herbivorous ) based on the proposed relationship between or indirectly (carnivorous crustaceans, fish) on primary productivity and food-chain length. The the grazing chain (but see Gaedke et al. (2002) for trophic cascade has received strong support in the dependence of consumers on phosphorus). both mesocosm experiments (Brett and Goldman Trophic positions in the Lake Constance carbon 1996) as well as in whole lake manipulative flow models were not distributed homogenously FOOD WEBS IN LAKES 43

Box 4.1 The pelagic food web of Upper Lake Constance

Lake Constance is a large (500 km2) and deep weekly sampling. A special strength of this dataset is that (zmax ¼ 254 m) perialpine lake in central , which it encompasses both, the classical food chain as well as has been intensively studied throughout the twentieth the . For carbon flow models the century. The lake consists of the more shallow Lower pelagic food web was aggregated into eight different Lake Constance, and the deep Upper Lake Constance compartments (Figure 4.1(b)), of which five can be (Figure 4.1(a)). Due to its deep slope the latter has a truly assigned to the ‘‘classical food chain,’’ that is, pelagic zone, which seems to be energetically independent phytoplankton, rotifers, herbivorous crustaceans, from littoral subsidies. Like many other temperate lakes, carnivorous crustaceans, and fish, and three to the Lake Constance went through a period of severe microbial loop, that is, bacteria, heterotrophic eutrophication starting in the 1930s and culminating in the nanoflagellates, and ciliates (Figure 4.1(b)). In addition 1960s/1970s (Ba¨uerle and Gaedke 1998 and references flows between these eight compartments and the therein). Beginning with the 1980s total phosphorus detritus/DOC (dissolved organic carbon) pool were concentrations declined again. However, the response of considered (exsudation of phytoplankton, egestion and the plankton community to oligotrophication was delayed. excretion of consumers, DOC uptake by bacteria). To The pelagic food web of Upper Lake Constance during the analyze seasonal changes in carbon flows data were oligotrophication period has been analyzed within several subdivided into up to 10 seasonal time intervals per year years of intensive sampling (Ba¨uerle and Gaedke 1998). lasting between 14 and 102 days. For all seasonal time Different food-web approaches, that is, body-mass size intervals mass-balanced carbon and phosphorous flows distributions (Gaedke 1992, 1993; Gaedke and Straile were established and further processed with the 1994b), binary food webs (Gaedke 1995), and mass- techniques of network analyses (Ulanowicz 1986). balanced flow networks (Gaedke and Straile 1994a,b; The results shown here are based on 44 different Straile 1995; Gaedke et al. 1996; Straile 1998; Gaedke mass-balanced food-web diagrams for different seasonal et al. 2002) were applied to a dataset consisting of five, time intervals from the study years 1987 to 1991 respectively eight (Gaedke et al. 2002) years of almost (Straile 1995, 1998).

(a) (b) Sampling site Fish Germany Carnivorous crustaceans

Friedrichshafen Herbivorous Konstanz Rotifers crustaceans Ciliates

Romanshorn Switzerland HNF

0 5 10 km Bregenz Bacteria Phytoplankton Austria

Figure 4.1 (a) Map of Lake Constance, and (b) aggregation of the Lake Constance pelagic food web into eight trophic guilds. Cannibalistic food-web interactions were considered for ciliates, rotifers, and carnivorous crustaceans, but are not shown here. Also not shown are the flows between these compartments and the detritus/DOC pool.

(Figure 4.2(b)). Rather, the distribution seems to be trophic level will cause an overestimation of three-modal with peaks around trophic levels of trophic level omnivory, for example, herbivores do two, three, and below four. This is especially consume bacterivorous HNF (trophic level two to remarkable as the inclusion of the microbial three, Figure 4.2(a)). Interestingly the distribution food-web with bacteria considered as second of trophic positions is not symmetrical around the 44 AQUATIC FOOD WEBS

(a) 5 (b) Phytoplankton Rotifers Bacteria Herbivorous crustaceans HNF Carnivorous crustaceans 4 Cilliates Fish

3 Trophic position Trophic 2

1 –10 –5 0 5 10 15 20 25 30 35 0 10 20 Body size (log2(pg C)) %

Figure 4.2 (a) Relationship between trophic positions in the carbon flow models and mean body size of trophic guilds; (b) percentage distribution of trophic positions (trophic positions of phytoplankton, bacteria, and HNFs not included). different trophic levels, but rather skewed to the in a discussion about climate effects on plankton right, which might reflect the tendency of omni- food webs. Due to the small size and consequently vorous consumers to gain most of their energy on high intrinsic growth rates of the food-web com- the lowest trophic level on which they feed ponents, successional processes taking centuries in (Hairston and Hairston 1993). terrestrial systems will take place every year anew Regarding the relationship of trophic position in the plankton. Environmental variations experi- with body size, within both, detritus chain, and the enced by the plankton community in a lake during grazing chain trophic position increases with body one year are considered to be scale analogous of size. However, predominantly herbivorous groups gross climate change since the last glacial maximum do cover a large body size range between 210 and (Reynolds 1997). Different successional stages are 224 pg C (Figure 4.2(a)), that is, more than four usually not considered together when examining orders of magnitude. This data support a recently terrestrial food webs, rather than food webs of for assembled binary food web for Tuesday Lake example, grasslands, , and forests are studied which also provides information on body size of separately. This might suggests that there is in the species and in which a strong overall relationship pelagic zone of a specific lake not one food web, but between body size and trophic position (prey- rather a succession of different food webs. In fact, averaged trophic position calculated from the the classical paper on plankton succession in lakes, binary food web) was observed (Cohen et al. 2003). the PEG (Plankton Ecology Group) model (Sommer In many cases lake food webs hence provide text et al. 1986) distinguishes between 24 successional book examples for HSS and the trophic cascade. stages, in which the relative importance of abiotic However, most of these examples are taken from forcing through physical and chemical constraints the food webs of lakes during the summer situ- and of biotic interactions, that is, competition and ation. As climate variability will affect lakes predation, differ (Sommer 1989). Consequently, throughout the season, it is necessary to consider seasonal variability in driving forces will result into seasonal variability of lake food webs. seasonal differences in for example, species com- position, diversity, interactions strength between Seasonal succession food-web components, and finally food-web con- figurations. The small size of planktonic organisms is the cause Food-web structure including the number of why it is necessary to consider seasonal succession functionally relevant trophic levels changes during FOOD WEBS IN LAKES 45 the season. During winter, primary productivity of (a) 3.5 Herbivorous algae in many lakes will be strongly limited by crustaeans the availability of light, for example, due to deep 3 mixing and/or the formation of ice cover, both preventing algal blooms. During this time period, 2.5 growth of algae is constrained by abiotic condi- 2 tions and not by the activity of herbivores. Reynolds (1997) considers this successional state as (b) 3.5 an aquatic analog to bare land. Only with ice 3 thawing and/or the onset of stratification in deep lakes algal blooms can develop with subsequent 2.5 growth of herbivores. Within a rather short time of Carnivorous Trophic position Trophic several weeks different herbivore population 2 crustaeans increase in size and finally often control algae. (c) 4.5 Within this so-called clear-water phase (Lampert 1978) algal concentration is strongly reduced and 4 water transparencies rise again to values typical for winter. During this successional phase the food 3.5 web is functionally a two-trophic level system. Fish However, the ‘‘green world’’ returns due to—at 3 least partially—increased predation pressure on J FMAM JJA SOND herbivores due to fish, whose larvae surpass gape limitation and/or due to invertebrate predators Figure 4.3 Seasonal changes in trophic position of herbivorous zooplankton, carnivorous zooplankton, and fish. Trophic positions (Sommer 1989). In addition to predatory control of are shown on the time axis at the midpoints of the respective herbivores, the world in summer is also prickly time intervals. The dotted line represents a loess fit. and tastes bad (Murdoch 1966) as for example, large and spiny algal species develop which are difficult to ingest for herbivores. Hence the system herbivorous production to sustain a carnivorous has developed from a one level system in early feeding mode for a major invertebrate group spring where herbivores are not abundant enough (Straile 1995). As a consequence, ‘‘carnivorous to control phytoplankton development into a crustaceans’’ do rely largely on phytoplankton complex three–four level system in summer, which production, that is, they are in fact herbivorous collapses again toward winter. during winter and early spring. Only from late This development of trophic structure is also spring/early summer onwards, herbivore pro- evident for the trophic positions of the different duction is large enough to sustain energetically compartments of the Lake Constance flow model populations of carnivorous crustaceans. This is (Figure 4.3). The trophic position of herbivorous also the time when more specialized carnivorous crustaceans remains rather constant throughout crustaceans, the cladocerans Bythotrephes long- the year, at approximately two (Figure 4.3(a)). imanus and Leptodora kindtii develop from resting However, the trophic positions of carnivorous eggs and build up important populations in the crustaceans and of fishes do increase seasonally planktonic community. Hence, the food web (Figure 4.3(b) and (c)). This increase of both groups develops energetically from a three-trophic level is due to the increase in trophic position of system in winter/early spring to a four-trophic carnivorous crustaceans. During winter and early level system in summer. This is not due to the spring, carnivorous crustaceans consist of cyclo- addition of a new top predator, as fish are present poid copepods. Although considered as carni- during the whole year, but due to the combined vorous at least as more ontogenetically more effects of diet switches of taxa, that is, cyclopoid advanced stages, there is simply not enough copepods, and the addition of new intermediate 46 AQUATIC FOOD WEBS predators, that is, Bythotrephes and Leptodora. This provide evidence for shifts in trophic cascades does not imply that the pelagic food web during caused by intrinsically driven population dynamics summer can be considered dynamically as a simple of top predators including size-specific . four-trophic level system as among other things This is an important case study showing that spatial (Stich and Lampert 1981) and size refuges cannibalism and life-cycle omnivory can have (Straile and Ha¨lbich 2000) of prey species at dif- important food-web consequences culminating in a ferent trophic levels seem to blur cascading trophic trophic cascade, that is, less zooplankton and more interactions in Lake Constance. phytoplankton biomass during high summer when However, despite these seasonal differences in large-sized ‘‘gigantic cannibals’’ produced high food-web composition, structure, and dynamics, numbers of young-of-the-year fish (Persson et al. the season-specific pelagic food webs are clearly 2003). However, even for long-lived animals there interconnected temporally. This is either due to the is a need to adjust their life cycle to the seasonal species which manage to persist during all food- environment, that is, to seasonal changes in the web configurations within the year in the open food-web structure. Any analysis of climatic forcing water, or due to the existence of diapause and or of pelagic food-web hence will need to consider fast stages. The latter allow species—given changes in food-web structure occurring during sufficiently correct cues for the start and termina- seasonal succession and the of species tion of diapause—to exploit a specific season, to cope with this variability. respectively food-web configuration to ensure long-term persistence in a specific lake. These time- Climate variability and travelers (Hairston 1998) allow for the existence of plankton food webs a specific summer food web without the need for yearly new colonization events from other lakes. In Changes in climate may result into changes in addition, the latter is not really a possibility, as the timing and duration of successional stages (e.g. food webs in nearby lakes will be at a similar Straile 2002) as well as in modifications of food-web successional stage. This is also an interesting structure within distinct successional stages (e.g. contrast to terrestrial system where a landscape Weyhenmeyer et al. 1999). Climate warming has mosaic of different successional stages offers the been shown to increase the stratification period in possibility of a new colonization. Recent research lakes (Livingstone 2003) which may in turn inten- has shown that resting stages are important sify nutrient limitation and phytoplankton compe- components of the life cycle of nearly all plankton tition. Species need to adapt their life cycles, for taxa, for example, for phytoplankton (Hansson example, the timing of reproduction of long-lived 1996), ciliates (Mu¨ ller and Wu¨ nsch 1999), rotifers species, the timing of critical life history shifts, for (Hairston et al. 2000), and crustaceans (Hairston example, the timing of metamorphosis of and Ca´ceres 1996; Ca´ceres 1998; Jankowski and species, and the timing of diapause initiation and Straile 2004). Furthermore, bet hedging strategies termination to temporal shifts in food-web struc- even allow for multiyear dynamics and the persi- ture, respectively food-web shifts. Match or mis- stence of species even when reproduction fails match of specific life-history events with the within specific years (Ca´ceres 1998). successional state of the food web may crucially On the other hand, lake food webs in the various influence population dynamics of species. This has successional stages or even within several succes- been suggested for, for example, the interaction of sional cycles might be connected through the fish larvae and their zooplankton prey (Cushing presence of long-living organisms, which are for 1990; Platt et al. 2003) and the hatching of over- example, top predators. This can have important wintering eggs (Chen and Folt 1996). The latter dynamical consequences when for example, long- authors suggest that fall warming may result into lived piscivores suppress planktivorous fish over the maladaptive hatching of overwintering eggs of several years (Post et al. 1997; Sanderson et al. the copepod Epischura lacustris possibly resulting 1999; Persson et al. 2003). Persson et al. (2003) into the loss of the species from the lake. FOOD WEBS IN LAKES 47

As a consequence of the strong seasonal American lakes is associated with the NAO, res- dynamics within lake food webs, climate effects pectively ENSO (Anderson et al. 1996; Livingstone will be highly season-specific. Interestingly, and in 2000; Straile et al. 2003). Winter severity influences contrast to many food-web studies, which consider the mixing intensity in deep lakes, thereby influen- the summer situation, there is a wealth of studies cing nutrient distributions (Goldman et al. 1989; reporting ecological climatic effects of climate Nicholls 1998; Straile et al. 2003) and forcing in lakes during winter and early spring concentrations in the (Livingstone periods. Although climate forcing is also important 1997; Straile et al. 2003). In contrast, stratification in summer (see below), this might point to a special is usually strong during summer and interannual importance of winter meteorological forcing for meteorological variability might result into more temperate lake food webs. It is hence important to or less steep temperature gradients but unlikely in ask why this is the case: as often, there is not a mixing (but see George and Harris (1985) for more single explanation, but meteorological, physical, -exposed lakes). However, warmer summer and biological reasons contribute to this obser- temperatures may reduce mixing events in shallow vation: (1) there are strong changes and there is polymictic lakes. a high interannual variability in winter meteoro- 3. Species and life history and hence logical forcing, (2) there is a high susceptibility of plankton succession is highly sensitive to physical lake to meteorological forcing in the winter factors during the winter half year (Sommer et al. half year, (3) there is a high susceptibility of 1986). Ice cover duration (Adrian et al. 1999; species physiology and life history and hence of Weyhenmeyer et al. 1999), mixing intensity, and successional dynamics to changes in lake physics timing (Gaedke et al. 1998) will have a strong during the winter half year. impact on phytoplankton bloom formation. Tem- perature effects on growth rates of zooplankton 1. Strong changes and high variability in winter and fish seem to be especially important during meteorological forcing—the largest warming on spring. During the phytoplankton spring bloom our planet during the twentieth century has zooplankton growth is unlikely to be limited by occurred (1) during winter and (2) over Northern food concentration. Consequently, temperature Hemisphere land masses. Likewise, large-scale limits zooplankton growth rates (Gerten and climatic such as the North Atlantic Adrian 2000; Straile 2000; Straile and Adrian 2000). (NAO) and the El Nino Southern Similar arguments have been put forward regard- Oscillation (ENSO) do have their strongest tele- ing growth of whitefish larvae (Eckmann et al. connections to northern hemispheric 1988). As zooplankton is abundant in late spring— during winter (Hurrell 1995; Rodionov and Assel given that there is not a mismatch situation—fish 2003). In contrast, summer warming was not as growth depends on water temperatures. However, pronounced during the twentieth century, and higher winter temperatures may also have negative climate forcing during summer seems to be more impacts on consumers due to enhanced metabolic controlled by local and regional factors, but less requirements at low food abundance. This has been under the control of large-scale climate oscillations. suggested to have caused a decline of Daphnia 2. Lake physics will be especially sensitive to abundance in high NAO winters in Esthwaite climate variability during the winter half year. Water (George and Hewitt 1999). Additionally, During this time period important mixing events do competition with the calanoid copepod Eudiapto- occur and lakes may or may not be covered by ice. mus, which has lower food requirements and Physical conditions such as the presence/absence increased in abundance during winter in high and intensity of mixing and the presence/absence NAO years might have contributed to the decline and duration of ice cover may be highly sensitive— of Daphnia in Esthwaite Water. depending on lake morphology, latitude, and altitude—to changes in winter meteorology. For Another important reason for the importance example, ice cover duration in European and North of winter is that winter represents a population 48 AQUATIC FOOD WEBS bottleneck for many species. For example, in growth rates of daphnids during spring and an temperate lakes, the abundance of plankton is earlier onset of the clear-water phase has been strongly reduced during winter, and populations suggested to promote a in shallow rely on resting stages to overwinter (see above). lakes from a turbid phytoplankton dominated to Fitness benefits might be associated with the a clear macrophyte-dominated state. This might be ability to overwinter in the plankton as opposed to an example on how even small temporal changes in produce resting stages. For example, the dom- the food-web interaction between phytoplankton inance of the Planktothrix in Lake and daphnids may result into large-scale ecosystem Zurich in recent years has been attributed to a changes in lakes. However, the database used by series of mild winters resulting in less deep water Scheffer et al. (2001b) is confounded by manage- mixis (Anneville et al. 2004). ment effects and cannot be used as a strong support Also winter is often associated with reduced of their hypothesis (van Donk et al. 2003). consumption and starvation of juvenile fishes Differences in winter meteorological forcing can influencing fish life history (Conover 1992) and also change the nutrient availability for phyto- population dynamics (Post and Evans 1989). plankton due to changes in deepwater mixing or Furthermore severe winters associated with long- runoff with subsequent consequences for annual lasting ice cover and oxygen deficiency can cause primary productivity. For example, primary pro- winterkill of fish species with consequences for fish duction in Lake Tahoe and Castle Lake was related community composition (Tonn and Paszkowski 1986) to winter mixing depth and precipitation, respect- and lower trophic levels (Mittelbach et al. 1995). ively (Goldman et al. 1989). Likewise, decreased However, also summer food webs are affected summer transparency and increased epilimnetic by climate variability. This is due either to summer pH and O2 concentrations indicated that the phyto- meteorological forcing or indirect food-web medi- plankton summer bloom was more intense after ated effects of winter/early spring meteorological mild winters in Plußsee (Gu¨ ss et al. 2000). Food- forcing (Straile et al. 2003). For example, temper- web mediated effects of climate variability in ature effects on the growth rate of the cladoceran summer will also occur when fish numbers are Daphnia will change the timing of phyto- affected by winter severity due to starvation or plankton suppression, that is, the timing of the oxygen deficiency under ice (see above). clear-water phase in early summer (Straile 2000). Summer meteorological forcing can also have Higher temperatures associated with a positive significant effects on lake food webs. For example, phase of the NAO resulted into higher Daphnia in Lake , interannual differences in growth rates and earlier phytoplankton depression June stratification linked to the position of the Gulf in Lake Constance, in addition the duration of (George and Taylor 1995) are related to phytoplankton suppression increased (Straile the growth of edible algae and finally to Daphnia 2000). Similar shifts in the timing of the clear-water biomass (George and Harris 1985). Also, copepod phase were observed in a number of European species have been shown to increase in abundance lakes (Straile 2000, 2002; Anneville et al. 2002; as a result of warm summers possibly due to Straile et al. 2003). Overall warming trends since development of an additional generation (Gerten the 1970s resulted into an advancement of the and Adrian 2002b). clear-water phase of approximately two weeks in Effects of climate variability will also change the central European lakes (Straile 2002). This pheno- relative importance of species within successional logical trend of a predator–prey interaction is of stages. For example, phytoplankton species differ similar magnitude as phenological shifts in, for in respect to sedimentation and/or light limitation. example, plant, insect, and bird populations Climate-related changes in stratification hence will (Walther et al. 2002; Straile 2004). In shallow lakes, favour different species. For example, in Mu¨ ggelsee the spring clear-water phase is of crucial import- (Adrian et al. 1999) and Lake Erken (Weyhenmeyer ance for the growth and summer of et al. 1999), mild winters with less ice cover in macrophytes (Scheffer et al. 1993). Hence, higher response to high NAO years favored diatoms FOOD WEBS IN LAKES 49 species over other phytoplankton. Diatoms need 4 turbulent conditions to prevent them from sinking ))

out of the euphotic zone, a condition not met –2 under ice. In contrast, the dinoflagellate Peridinium 3 can already build up a large under (mg C m clear ice (Weyhenmeyer et al. 1999). Likewise, in 10 2 Rostherne , the intensity of summer mixing is crucial for which phytoplankton species will dominate: extremely stable stratification leading to 1

the dominance of Scenedesmus, with increasing Biomass (log Copepods mixing favoring Ceratium/Microcystis and finally Daphnia 0 Oscillatoria dominance (Reynolds and Bellinger 1992). The establishment of a more stable strat- 7891011 May temperature (°C) ification is of special concern since stratification may often favor buoyant species including toxic Figure 4.4 Response of Daphnia and copepod biomass in May to cyanobacteria over sedimentating species (Visser interannual differences in vernal warming as expressed as average et al. 1996). May water temperatures (0–20 m water depth). Both taxa show a < Also zooplankton growth rates will show differ- significant relationship with temperature (n ¼ 12, r ¼ 0.84, p 0.001 for daphnids, and n ¼ 12, r ¼ 0.57, p < 0.06 for copepods). ent responses to interannual temperature vari- ability. Based on a literature survey of laboratory data, Gillooly (2000) suggested that larger zoo- plankton species should respond more strongly to relative abundances of daphnids and copepods in increasing water temperatures than small ones. spring as an increase in food availability due to According to his survey, difference in generation eutrophication (Straile and Geller 1998). Eutrophi- time between the large cladoceran Eurycercus cation as well as temperature increase seems to lamellatus and the rotifer Notholca caudata is favor taxa with high intrinsic growth and develop- reduced from 88 days at 5C to 15 days at 20C mental rates. suggesting that large zooplankton should be corres- To summarize, due to the fast changes in food- pondingly favored with warming. A hypothesis web structure in pelagic systems, and the import- that remains to be tested with field data (Straile ance of food-web interactions, it is important to et al., in prep.). Also the life history of zooplankton adopt a food-web approach when studying climate species is of importance. rate of effects on lake ecosystems. Plankton systems seem parthenogenetically reproducing rotifers and cla- to be especially suitable to study indirect food-web docerans exceeds those of copepods which repro- mediated effects due to the small size of the duce sexually and do have a long and complex organisms concerned, their fast growth rates, and . As a consequence, daphnids should be the relative promptness in which indirect effects able to benefit more strongly from spring warming will occur. On the other hand, food-web variability than copepods. This can be demonstrated by due to meteorological forcing, that is, so-called examining interannual variability of daphnid and ‘‘natural experiments,’’ can also be considered as copepod biomass in relation to interannual vari- an important tool to analyze food-web regulation. ability to vernal warming. Both taxa are able to The highly seasonal signal of climate effects on build up higher biomasses in years with early lake food webs suggest that food-web studies vernal warming and consequently high May water in lakes need to consider seasonal variability. temperatures in Lake Constance (Figure 4.4). In addition, more attention should be paid to the However, the effect is much more pronounced for winter period. Many studies in the past seem to daphnid biomass, which increases several orders have neglected the winter due to the admittedly of magnitude. Interestingly, an increase in spring overall lower biological activity during winter. temperatures seems to have a similar effect on the Instead, the winter period may be a critical one for 50 AQUATIC FOOD WEBS lake functioning, and for climate effects on lake Acknowledgments functioning as it seems that the winter/early Data acquisition was mostly performed within the spring situation might set the stage for much what Special Collaborative Program (SFB) 248 ‘‘Cycling of will come during the period. When Matter in Lake Constance,’’ supported by Deutsche adopting a seasonal approach, more emphasis Forschungs-gemeinschaft. I thank all people contri- should also be paid to time and season as resources. buting to the establishment of the food-web dataset Resources, which will undoubtedly change with of Lake Constance. The manuscript benefited from climate change and to which organisms have to fit comments by K.-O. Rothhaupt. their life cycle. CHAPTER 5 Pattern and process in food webs: evidence from running waters

Guy Woodward, Ross Thompson, Colin R. Townsend, and Alan G. Hildrew

Introduction: connectance webs greater sampling effort and the improved resolution from of small-bodied organisms has revealed recurrent and complex patterns in stream food-webs includ- Early depictions of food webs were usually simple ing the ubiquity of omnivory, feeding loops and and consisted of few taxa and a limited number cannibalism, high linkage density, a decline in of links. These ‘‘connectance’’ webs seemed to connectance with increased species richness, and support models suggesting that the dynamical long food chains (Schmid-Araya et al. 2002a,b; stability of food webs was inversely related to their Woodward and Hildrew 2002a). complexity, and thus the prediction that simple The number of species and links described in webs should predominate in nature (May 1972, the food web of Broadstone Stream (southern 1973; Cohen 1978). Pioneering descriptions of food England), currently the most detailed of any lotic webs from running waters proved no exception system, increased by almost an order of magni- (Figure 5.1(a); see Hildrew 1992), though more tude from its earliest incarnation, as resolution was latterly streams have become particularly promin- improved among smaller species. The initial ent in the new generation of high quality, detailed description (Figure 5.1(b); Hildrew, Townsend, food-webs that challenge the received wisdom that and Hasham 1985) contained 24 species and 109 complexity is the exception rather than the rule (e.g. links, and described the trophic links of the Tavares-Cromar and Williams 1996; Benke and numerous but small-bodied predatory chir- Wallace 1997; Townsend et al. 1998; Williams and onomids for the first time in any food web. Ten Martinez 2000; Schmid-Araya et al. 2002a,b). years later, predatory links to the microcrustacea Many of the coarse groupings (e.g. ‘‘algae’’ or were resolved (Lancaster and Robertson 1995), ‘‘meiofauna’’) used previously to lump supposedly producing a web of intermediate completeness similar ‘‘trophic species’’ are now separated into (Woodward and Hildrew 2001; Figure 5.1(c)). This distinct taxa in these more detailed webs. Such was followed by the inclusion of the soft-bodied research shows that a large number of samples are meiofauna and algae (Figure 5.1(d); Schmid-Araya required to characterize complex food-webs, not et al. 2002a), and there are now 131 species and least because many predators have empty guts and over 841 links described (Woodward, et al. in press). species rank-abundance curves follow log-normal Although Broadstone Stream seems like a complex or geometric distributions (Tokeshi 1999; system, its acidity actually renders it relatively Woodward and Hildrew 2001). Yield-effort curves depauperate (e.g. fish and specialist grazers are for the detection of species and links suggest that absent, and there are few algae), and other streams small sample sizes are often insufficient to capture are certainly much more complex, potentially the true complexity of real food webs (e.g. yielding further surprises. For instance, a single Thompson and Townsend 1999). Most importantly, small, sandy stream in central Germany, the

51 52 AQUATIC FOOD WEBS

(a) (c)

(b)

(d)

Figure 5.1 Connectance food webs from the early and more recent stream literature: (a) Early stream food-web (redrawn from Cohen 1978) (b) Initial connectance web from Broadstone Stream (after Hildrew et al. 1985) (c) Intermediate resolution web from Broadstone Stream (after Woodward and Hildrew 2001) (d) Highly resolved Broadstone Stream food-web (after Schmid-Araya et al. 2002a).

Breitenbach, has over 1,000 species of 1996; Townsend et al. 1998; Schmid-Araya et al. (Allan 1995). Nonetheless, the sample size of 2002b; Thompson and Townsend 2003), and stream food-webs where individual feeding links encouraging progress has been made over the last have been resolved, while still very small, is 20 years in revealing the true complexity of trophic increasing (e.g. Tavares-Cromar and Williams networks in streams. PATTERN AND PROCESS IN FOOD WEBS 53

Small-bodied species are not the only ones prone that if a web is arranged into a matrix of con- to underestimation; many top predators, such as sumers and resources ranked in order of increas- semi-aquatic birds, are comparatively rare, mobile, ing body-size, feeding links are overrepresented, and often ignored by stream ecologists (Steinmetz relative to a random distribution, in the upper part et al. 2003). In addition, links between the main of the triangle above the leading diagonal (Cohen energy sources in stream food-webs, algae and et al. 1993a). The situation might be reversed in detritus of terrestrial or aquatic origin plus asso- host–parasite webs, where consumers are smaller ciated heterotrophic microbes, still remain largely than their resources, but little is known about the unresolved (Hildrew 1992). Probably the best- importance of parasites in stream food webs. characterized links within the microbe-detritus We might expect that the size difference between ‘‘sub-web’’ are those between leaf-litter and predators and prey ultimately becomes so great hyphomycete fungi (e.g. Gessner and Chauvet that the feeding links are broken; thus, piscivorous 1994). Unraveling the individual feeding links fish are unlikely to feed directly on rotifers, which between undoubtedly heterogeneous basal food therefore occupy a lower-size refugium (the ‘‘size- resources and their nonpredatory consumers disparity hypothesis’’ (Hildrew 1992; Schmid-Araya remains a major challenge for stream ecologists. et al. 2002a)). This suggests that stream food webs might be partially compartmentalized, such that, Food web patterns related to body size although upper triangularity could exist within ‘‘subwebs’’ (perhaps delimited by portions of the Stream food-webs often contain reciprocal preda- community size-spectrum, for example, meiofauna, tion links, in accord with the ‘‘niche’’ food-web macrofauna, megafauna), these subwebs might models of Warren (1996) and Williams and have relatively few connections between them. Martinez (2000) but not the earlier ‘‘random’’ or Size disparity between the top and bottom of food ‘‘cascade’’ models (e.g. Cohen et al. 1993a). These webs might thus also account for the decline in may involve body-size effects, including the web connectance with increasing species richness seasonal ontogenetic reversals that occur when (this last partly attributable to the inclusion of generations overlap, so that large individuals of the small-bodied meiofauna) apparent in a sample ‘‘small’’ species are able to eat small individuals of of stream food webs (Schmid-Araya et al. 2002b), ‘‘large’’ species (Woodward and Hildrew 2002a). though verifying such patterns is fraught with Moreover, both the trophic status and niche width methodological difficulty. of secondary consumers can be determined mainly Cohen et al. (2003) have demonstrated recently by body size rather than , with similar- the potential importance of body size as a struc- sized predators occupying almost the same posi- turing force in aquatic food-webs by examining tion within the food web, regardless of their relationships within a three-way data matrix: they species identity (Woodward and Hildrew 2002a). found strong trivariate correlations between the Other powerful influences of body size on food presence/absence of links, mean body-size, and the web structure and dynamics are the partitioning of mean abundance of each species. Small species food resources among predators, via indirect hori- were abundant, but low in the food web, with large zontal (i.e. potentially competitive) links (Cohen species being rarer and higher in the food web and et al. 1993a; Woodward and Hildrew 2002a) and also possessing a greater number of links. Although the expansion of prey size range with predator size these patterns have yet to be examined explicitly in (Allan 1982; Woodward and Hildrew 2002a). The a range of systems, they are true of the Broadstone latter occurs when small prey remain vulnerable to Stream food web (Woodward et al. in press). In growing predators that are increasingly able to addition, many of the component univariate or take larger prey. This ‘‘nested hierarchy’’ of feed- bivariate patterns they report have already been ing links confers an obvious size-related structure described in stream communities, such as inverse to food webs (Cohen et al. 1993a). A pattern is relationships between body-size and abundance produced, known as ‘‘upper triangularity’’, such (Schmid et al. 2000) and upper triangularity 54 AQUATIC FOOD WEBS

(Williams and Martinez 2000; Schmid-Araya et al. (Benke and Wallace 1997). While web structure 2002a). Cohen et al.’s (2003) findings may therefore and the strength of links clearly interact, and a apply to many food webs in running waters. We knowledge of both is necessary to understand return to body size effects in later sections. function, there is now a shift in emphasis to the construction of quantitative, and thus hopefully more realistic, food webs (e.g. Hall et al. 2000; From pattern to process Benke et al. 2001; Woodward et al. in press). Food-web ecologists working on streams have Community and ecosystem approaches to made recent progress by using semi-quantitative quantifying stream food webs measures of ‘‘interaction strength’’ or by detailed The analysis of connectance webs, which represent analysis of subsets of whole community webs (e.g. equally species and the links between them, focu- Benke and Wallace 1997; Woodward and Hildrew ses on patterns in structure (e.g. links per species; 2002a; Figure 5.2). However, the sample of such predator–prey ratios) rather than processes (e.g. webs is very small and interaction strength has energy flux, interaction strengths). Trivial inter- been expressed in a variety of ways, hindering actions are inevitably overemphasized and vice versa comparisons (indeed, Berlow et al. 2003 list no less

(a)

<0.1 0.1–1 1–10 10–50 >50 Ingestion rates (g m–2 yr–1)

(b)

Figure 5.2 Approaches to quantifying food webs: (a) the flux of and the trophic basis of production among consumers in a subset of a stream food-web (redrawn after Benke and Wallace 1997), (b) body size, trophic status, and feeding loops within the predator subweb of Broadstone Stream, with links expressed as per capita consumption of ‘‘prey’’ by ‘‘predators’’ (% of each ‘‘prey’’ population consumed per ‘‘predator’’ individual per 24 h). The area of the circles is proportional

to log10 mean body-mass (redrawn after Woodward and Hildrew 2002a). PATTERN AND PROCESS IN FOOD WEBS 55 than 11 separable uses of the term in the ecological species might account for much of a predator’s literature). Attempts to quantify webs have focused production, the population size of the prey may either on population/community dynamics (e.g. itself be unaffected. Conversely, a predator may Power 1990; Wootton et al. 1996) or on the flux of consume a large proportion of the production of energy or matter (the ‘‘ecosystem approach’’ e.g. one prey species, while that prey contributes little to Benke and Wallace 1997), with virtually none the overall production of the predator (a situation combining the two (but see Hall et al. 2000). The that can arise where predators are subsidized by former approach can itself be divided into ques- alternative food sources). This creates the potential tions about overall community dynamics (for for indirect effects, such as apparent competition, instance, how the distribution of interaction whereby prey compete for ‘‘enemy-free space’’ strength may affect community stability), and those rather than for more conventional resources (Holt seeking predictions about the dynamics of con- 1977). The potential for diffuse, apparent competi- stituent species populations (Berlow et al. 2003). tion can be gauged from prey-overlap graphs. The ecosystem approach has often been used to Those from Broadstone Stream (Woodward and view food webs from a mass–balance perspective in Hildrew 2001) show that virtually every prey spe- which, for example, the patterning and strength of cies shares at least two predator species with vir- flow pathways have been analyzed to assess indir- tually every other prey species, with many taxa ect effects, trophic and cycling properties, and the being preyed on by all of the dominant predators organizational status of the system (e.g. Christian (Figure 5.3). The strength of these indirect prey– et al. Chapter 3, this volume, review several such prey interactions is extremely difficult to measure studies from marine systems). but, if they prove generally prevalent, they could be Interaction strengths defined or measured in the a major driver of community dynamics. The blur- different ways are not necessarily related, so com- ring of the traditional boundaries between compe- parisons between quantified webs constructed tition and predation that occurs in real food webs using the different approaches should be made (e.g. and cannibalism), allows with caution. For instance, while a single prey for a variety of complex interactions to arise,

Sialis Plectrocnemia Tanypodinae

Figure 5.3 Prey overlap graphs derived from the

Sialis Broadstone Stream food-web. Links between prey join species that share pairs of the four predator taxa in the matrix. The arrows point in the direction of increasing predator body-mass. There is high overlap, especially when comparing pairs of predators of similar size (e.g. Sialis and Cordulegaster), suggesting a high degree of trophic redundancy in the food web and the potential for diffuse apparent competition among prey (constructed from

Plectrocnemia Cordulegaster the food-web data matrix in Woodward and Hildrew 2001). 56 AQUATIC FOOD WEBS including apparent competition, apparent mutual- will be donor-controlled. Supporting this, several ism, and even potentially neutral effects of pre- studies have reported the absence of trophic cas- dators on their prey, and a given predator can have cades in detritus-based stream food-webs, where mixed trophic impacts on the different members of omnivorous predators have had negative effects a web (Ulanowicz and Puccia 1990; Woodward and on two adjacent trophic levels (e.g. Hildrew 1992; Hildrew 2002c). Usio 2000; Rosemund et al. 2001; Woodward and Hildrew 2002a,b). This differs markedly from the situation in algal-based systems, in which cascades The population/community approach: the have been recorded, where consumers typically complexity-stability debate revisited exert strong top-down control on their resources Recent theoretical advances have allowed the (e.g. Power 1990; Townsend 2003), which may also development of more sophisticated mathematical be responding simultaneously to the bottom-up models that predict better the dynamics of food effects of nutrients or light (Hillebrand 2002). The webs (Polis 1998). In the case of the population/ presence of a large number of weak or donor- community approach, there is increasing support controlled interactions within a food web, such as for the argument that the complexity (i.e. many those in detritus-based streams, is thought to species and/or links) commonly observed in nature increase its stability (McCann 2000), and may serve can increase the stability of a food web (de Ruiter to buffer any potentially strong top-down effects et al. 1995; McCann et al. 1998; McCann 2000; of pairwise interactions between predators and Neutel et al. 2002). McCann et al. (1998) challenged primary consumers (Schmid-Araya et al. 2002a). conventional wisdom by demonstrating mathemati- The idea that a diffuse network of pathways within cally that complexity could enhance stability—if a diverse community can dampen strong, pairwise most links were weak. Intriguingly, some of the interactions, thereby increasing dynamic stability, earliest models also demonstrated this possibility is of course not new, having been proposed by (see May 1972) but this was largely ignored in favor Charles Elton almost 80 years ago (Elton 1927) and of models that assumed strong interactions and reiterated by MacArthur (1955). showed an inverse relationship between complexity These suggestions are supported by the results and stability. of a large-scale experiment in which leaf-litter Most dynamic food-web models constructed inputs to a detritus-based stream were manipu- prior to those of McCann et al. (1998) were based lated experimentally: the strength of links (in on the assumptions that interactions between terms of per unit biomass consumption of prey by species were strong (and often uniformally dis- predators) increased when leaf-litter was excluded, tributed) and that feeding relationships were linear suggesting that a large reservoir of detritus may (e.g. Type I rather than sigmoid Type III functional indeed weaken top-down effects and increase responses) (Polis 1998). An increasing body of stability (Hall et al. 2001). Subsidies from external evidence from real systems suggests that most energy sources are in general common in stream links in food webs are ‘‘weak’’ while fewer are food-webs, such as terrestrial invertebrates that fall ‘‘strong’’ (McCann 2000), and this seems true into the stream (e.g. Nakano et al. 1999), CPOM regardless of what measure of interaction strength and FPOM derived from terrestrial leaf-litter (e.g. is used (Berlow et al. 2003). In this context, Usio 2000; Usio and Townsend 2001; Woodward experiments in streams have often demonstrated and Hildrew 2002a) and marine-derived biomass that, although predators can have strong impacts from carcasses of anadromous salmon that die on a few prey species, they have much weaker, or after spawning (e.g. Wipfli and Caouette 1998). negligible, direct effects on many others (e.g. Refugia may also serve to weaken predator–prey Woodward and Hildrew 2002a). interactions and thus stabilize food webs (Closs The dominance of terrestrial detritus at the base 1996), and the physical complexity of streams may of many stream food-webs means that a large explain why trophic cascades appear to be rarer number of primary consumer–resource interactions than in simpler habitats, such as the pelagic zone PATTERN AND PROCESS IN FOOD WEBS 57 of lakes, where physical refugia are scarce and intractable. Consequently, the main strength of SIA strong cascading interactions can arise. Stabilizing, is that it provides a complementary, rather than density-dependent Type III functional responses alternative, approach to GCA, which is most can arise where prey populations are constrained commonly used to quantify ingestion (e.g. Hall by the availability of physical refugia, such as et al. 2001). interstitial between particles in the stream Despite the lack of methodological standardiza- bed. At low prey , the relative tion among studies, a similarly skewed distribu- availability of refugia is high and predators may be tion of ‘‘link strength’’ occurs in a wide range of unable to drive prey to extinction. As prey density terrestrial and aquatic food webs, irrespective of increases, however, per capita increa- which measures are used (e.g. de Ruiter et al. 1995; ses. There is now plenty of evidence from streams Benke and Wallace 1997; Wootton 1997; Emmerson of density-dependent mortality and the stabilizing and Raffaelli 2004; Woodward et al. in press). To effect of refugia (e.g. Hildrew and Townsend 1977; quantify the impact of a predator on a prey species Townsend 1989; Hildrew et al. 2004). (i.e. the interaction coefficient in a Lotka–Volterra

model, aij) and vice versa (i.e. aji), we need to know how much of a predator’s production is accounted Measuring interaction strength and food-web for by that prey species, and what proportion of dynamics in practice the prey’s production is ingested by the predator.

The term ‘‘interaction strength’’ is used loosely in Then aij can be estimated by dividing ingestion of a ecology (Berlow et al. 2003). However, by expres- prey species by the mean annual abundance of a sing carbon (or matter) flux in terms of g C (or dry predator or expressed as per unit biomass effects. mass) m2 per year and per capita consumption of Both halves of the community (Jacobian) matrix individuals m2 per year, common baselines can can therefore be calculated from secondary pro- be obtained for characterizing links, while also duction and ingestion data, as long as information approximating to the units used in most models is available on the assimilation and gross produc- (Cohen et al. 1993b). Detailed gut contents analysis tion efficiencies of the consumers. Measuring (GCA) can be used to assess both per capita con- secondary production and ingestion rates is sumption rates (e.g. Hildrew and Townsend 1982; laborious, as it requires extensive population sur- Speirs et al. 2000) and energy fluxes (e.g. Benke veying and exhaustive GCA. However, it does and Wallace 1997; Hall et al. 2000). Nevertheless, provide an approximate measure of ‘‘interaction large numbers of individuals need to be sampled strength,’’ in lieu of data obtained via the even because GCA provides only a snapshot of a pre- more logistically challenging, and indeed next to dator’s diet, which can vary in time, while most impossible, route of attempting to characterize guts are empty or contain few prey items, some of intergenerational population dynamics (as used in which are difficult to quantify (e.g. filamentous Lotka–Volterra type models) following pairwise algae, soft-bodied prey, biofilm). manipulations of the entire multispecies assem- Stable (SIA) uses the signature blage of a real food web. of stable isotopes of carbon and nitrogen to reveal, Stream ecologists are presently making concerted respectively, the overall basal resources of the web efforts to unite the currently disparate ecosystem and the trophic position (height) of a species and community approaches to food webs by com- within the web. This technique thus has the major bining detailed GCA with SIA. A promising avenue advantages that it is sensitive to what is assimi- for further progress is provided by the relatively lated by consumers, rather than what is simply new field of ecological stoichiometry (Chapter 5, ingested, and integrates feeding over a longer this Volume). Essentially, because any given species period. However, SIA mixing models cannot be must keep elemental ratios (e.g. C : N : P) in its body used to construct the complex, multispecies food- tissues within certain narrow limits, changes in the webs that can be described using GCA, partly ratios among food resources will determine which because the calculations soon become algebraically consumers will be most successful and this, in turn, 58 AQUATIC FOOD WEBS will determine species composition and population Spatial and temporal variation in dynamics and interactions in the web. food web pattern and process Stoichiometric techniques have been used most Variability in space extensively in the pelagic zone of lakes, where results suggest that an imbalance in C : N : P ratio Riverine food webs vary in both their structural can have profound consequences for plankton and dynamic properties over the three spatial population dynamics, even to the extent of creat- dimensions and with time (Woodward and ing alternative equilibria, whereby systems are Hildrew 2002c), and much of this variation is dominated either by algae or by rooted macro- patchy and scale-dependent (Townsend 1989). phytes (Elser and Urabe 1999). To date, very few Marked shifts can occur in both web topology and authors have employed stoichiometric approaches the strength of feeding interactions, even at the to the study of C : N : P ratios in lotic systems, microhabitat scale (Lancaster 1998; Stone and although many studies have examined either N : P Wallace 1996). or C : N ratios. For instance, the rate of N turnover The ecosystem size hypothesis (Cohen and in the tissues of primary consumers increases as Newman 1988) states that as the size of a habitat the C : N ratio declines, suggesting that the latter increases so too does species richness, habitat may act as a surrogate for biological activity availability, and habitat heterogeneity; these are all with regard to N flux in streams (Dodds et al. factors that may contribute to reduced omnivory 2000). A recent study demonstrated how global and greater dietary specialization, with consequent climate change could alter energy flux in detrital increases in food-chain lengths (Post et al. 2000; food chains in streams: raised CO2 reduced the Woodward and Hildrew 2002c). This hypothesis quality of the basal resources by increasing the has rarely been tested, but a study of 18 food webs, C : N ratio which, in turn, suppressed microbial collected using standardized methods, found a activity and the rate of decomposition (Tuchman weak positive relationship between ecosystem size et al. 2003). (estimated as wetted area of a 30-m reach) and Overall, because terrestrial leaf-litter is a poor mean food-chain length (Thompson and Townsend, quality food resource, being high in C and poor in press). in P and N, there is considerable scope for nutri- The question of spatial variation is also relevant ent recycling in streams to be strongly influenced at still larger scales. The by consumer-driven dynamics (Vanni 2002; (RCC), developed for North American river sys- Woodward and Hildrew 2002c). Large-scale tems with headwaters flowing through deciduous switches, along the river corridor, from a resource , postulated variations along a river in habitat base dominated by terrestrial detritus (very C conditions, resource quality and quantity, and in enriched) to one dominated by algae/aquatic ecological patterns (Vannote et al. 1980). Any macrophytes should, in theory, have strong effects longitudinal changes in the energy base and its upon nutrient uptake and storage by the con- associated should have important con- sumers within the web. A change in C : N : P ratio sequences for food-web structure. Despite distinct might have dramatic impacts on stream food-webs downstream variation in the nature of the basal as productivity increases (e.g. as P increases resources (changing from coarse to fine particulate during eutrophication). Indeed, Cross et al. (2003) matter), Rosi-Marshall and Wallace (2002) reported recently demonstrated marked differences in the remarkable consistency in overall food-web struc- C : N : P ratio of CPOM, driven by variations in ture, though not in energy flow along specific links microbial conditioning, across streams that spanned in their study stream. a gradient of nutrient enrichment. Intriguingly, they An important, though largely unexplored, also found P enrichment in the body-tissues of aspect of spatial scale in stream food webs relates consumers, suggesting a degree of deviation from to the enormous range in body size of the con- the usually assumed in stoichiometric stituent organisms and the consequent differences models (sensu Sterner 1995). in the spatial ‘‘grain’’ with which they perceive PATTERN AND PROCESS IN FOOD WEBS 59 their environment (Stead et al. 2003). Food webs cascade, with an increase in algal biomass when attempt to integrate information on groups ran- fish switched from terrestrial prey to stream ging from bacteria (for whom meaningful biolo- invertebrates (Nakano et al. 1999). Comparative gical scales may be measured in millimeters) to studies of streams receiving a variety of types of fish, , and birds (which may operate over litter, as compared with those receiving a single scales of meters to many kilometers). It has been type, also suggest that the heterogeneity of terres- suggested that food webs should be constrained by trial subsidies may be an important factor in the home range of the top predator in the system determining food-web characteristics (Thompson (Cousins 1996). For many streams this would and Townsend 2003). require inclusion of information appropriate to the There are also food-web linkages to the land- large scales over which predatory fish forage. The scape in the ‘‘reverse’’ direction. For instance, dispersal of adult insects along streams and aquatic insects accounted for around 60% of the between catchments is also likely to be important carbon assimilated by riparian spiders close to two in determining local food web structure in streams New Zealand streams (Collier et al. 2002) and (Woodward and Hildrew 2002c). aquatic prey supports populations of many ripar- In general, the definition of appropriate bound- ian birds (e.g. Ormerod and Tyler 1991). In the aries is a major challenge for food-web ecologists. Pacific northwest of nitrogen In streams, the limit is usually drawn at the derived from carcasses of migrating salmon is aquatic–terrestrial interface, but in reality this assimilated by riparian plants, incorporated into ‘‘’’ is leaky to materials and energy, and the terrestrial litter, and ultimately contributes to stream food-web is firmly embedded within that of stream productivity (Milner et al. 2000). the neighboring landscape (Ward et al. 1998). Expansion of streams and rivers across the While lateral links with the adjacent terrestrial aquatic/terrestrial ecotone during high flow events system have long been recognized as important in is also known to be important in determining studies of stream food-webs (e.g. Hynes 1975; Junk in-stream food webs. The lateral ‘‘flood pulse’’ is et al. 1989), attempts to quantify the energetic proposed to be a major force controlling pro- importance of terrestrial subsidies to streams are ductivity and biota in floodplains (Junk et al. 1989). more recent. Allen’s classic (Allen 1951) study of a Young and Huryn (1997) showed that dissolved stream identified ‘‘Allen’s paradox’’— organic carbon varies along the length of New the observation that in-stream invertebrate prey Zealand’s Taieri River, with the highest con- production was apparently insufficient to provide centrations where the river flows through flood- for the observed trout production. This paradox plain reaches. A similar role for floodplains in was resolved in part by realization of the import- maintaining and subsidizing river food-webs has ance of terrestrial invertebrates to trout diet been suggested for the tropics by Winemiller (1990, (Huryn 1996) and stream food-webs in general 1996). The consequences for food-web structure of (e.g. Nakano et al. 1999). The use of stable isotopes such energy subsidies from the land have been (e.g. Rounick and Hicks 1985) has also proved little explored and warrant future attention. As a useful in characterizing the relative importance of note of caution, however, Bunn et al. (2003) point biomass generated in the catchment and in the to accumulating evidence from stable isotopes that stream. terrestrial carbon provides little energy to aquatic Several large-scale manipulative experiments food webs in arid-zone flood plain rivers, despite have attempted to elucidate the role of terrestrial the sometimes enormous inputs. Rather, animals inputs by excluding allochthonous material. on Australian floodplain pools were labeled Exclusion of leaf-litter from a North Carolina mainly with C derived from aquatic microalgae, stream resulted in the loss of a trophic level from despite the latter’s apparent paucity. the stream food-web (Hall et al. 2000), while Most studies of stream food-webs have been similar experiments excluding terrestrial inverte- carried out over relatively limited spatial scales, brates from Japanese streams evoked a trophic and usually in an arbitrarily determined reach, 60 AQUATIC FOOD WEBS while experimental studies have commonly Seasonality in the life history of invertebrates resorted to the use of microcosms, mesocosms, or can affect food-web structure via ontogenetic in-stream channels. While such studies may shifts. In Broadstone Stream, pulses of invertebrate measure adequately the changes and processes recruitment in autumn and summer resulted in that operate on small scales, they cannot incorpo- marked changes in size spectra through time rate properly species that operate on large spatial (Woodward and Hildrew 2002a). For predators, scales (particularly fish). Experiments at a large which tend to be relatively long lived, this entailed scale, such as the whole catchment manipulations changes in diet as a series of prey moved through a of litter input at Coweeta, USA (Hall et al. 2000) temporal ‘‘window’’ of vulnerability, in which they and in Japan (Nakano et al. 1999), are particularly were detectable as prey but had not yet either notable exceptions. achieved an upper size refugium or had emerged as adults. Woodward and Hildrew (2002a) showed that the diet of predators changed with ontogeny, Variability in time with that of their prey, and with variations in the Food webs have been constructed over a wide range abundance of prey. In Broadstone this produces a of different timeframes: ranging from data gathered much simpler web structure in winter than sum- on a single day to collated webs that represent mer, and a much more complex summary web decades of cumulative research effort (Hall and than is ever realized at one time (Schmid-Araya Raffaelli 1993). Time is critical because temporal et al. 2002a). In contrast, the asynchrony of inver- variation in community structure and functioning tebrate life histories in New Zealand streams is inevitable on one or more timescales, particularly means that a common core of invertebrates is in dynamic systems such as streams. Overall, represented through the year across a range of size underlying temporal variation in stream food-webs classes (Thompson and Townsend 1999). means that summary food webs gathered over one Stream systems are characterized by a relatively or many years are likely to misrepresent true high degree of physical disturbance that can vary (instantaneous) food-web structure. This is because seasonally. Highly variable and/or unpredictable species may be grouped together that never actually discharge in small streams acts as a disturbance of occur at the same time, thus producing anomalously variable frequency and intensity (Poff and Allan low values of connectance in the summary web 1995) and can modify population, community, and (Thompson and Townsend 1999). ecosystem processes (Townsend 1989; Death and A number of studies of detritus-based streams Winterbourn 1995). The pattern of bed-disturbance have shown considerable seasonal variability in can also dictate species composition and richness, community structure (Winemiller 1990; Closs and and the distribution of species traits (Death Lake 1994; Tavares-Cromar and Williams 1996), and Winterbourn 1995; Townsend et al. 1997). due in part to seasonal patterns in the inputs Townsend et al. (1997) found that insect species of terrestrial leaf-litter. Thus, autumn pulses of that are small-bodied, mobile as adults, relatively leaf-litter may be important in determining the streamlined, and generalist in habitat preference synchrony of life-cycles evident in many Northern were more common in highly disturbed streams. Hemisphere streams, which in turn influences Those characters were grouped into those that food-web structure. However, seasonal variation conferred (the ability to survive spates) in stream food-webs is also evident in tem- and those that conferred resilience (the ability to perate regions with relatively muted cycles. For recover quickly after spates). instance, Thompson and Townsend (1999) repor- Although few studies have investigated the ted seasonal food-web patterns in New Zealand influence of disturbance on food-web structure, streams that had no autumnal pulse of leaf-litter, Townsend et al. (1998) found that, in 10 streams that modest seasonality in rainfall and a fauna of differed in bed disturbance regime, the number invertebrates with predominantly asynchronous of dietary links per species was negatively related life-cycles. to intensity of disturbance. Jenkins et al. (1992) PATTERN AND PROCESS IN FOOD WEBS 61 hypothesized that disturbance could also affect forests often shade the channel pro- food-chain length, because species higher in the foundly, reducing algal productivity, but may food chain are rarer and are more likely to be lost by increase the supply of organic matter. chance from the system during a disturbance event, Lindeman (1942) predicted more trophic levels but stream studies have not provided unequivocal (longer food chains) in productive systems, and data in support of this (Townsend et al. 1998). data from New Zealand grassland and forested The persistence of food-web structure has been streams (Townsend et al. 1998; Thompson and relatively poorly studied on timescales greater than a Townsend, unpublished) showed a weak, positive single year (e.g. Winemiller 1990; Closs and Lake relationship between productivity and food-chain 1994; Tavares-Cromar and Williams 1996 and above). length (as well as the total number of trophic links Bradley and Ormerod (2001) followed invertebrate and links per species). Grassland food-webs (‘‘high communities in eight Welsh streams over 14 years. productivity settings’’) had more algal species, dis- They found that the persistence of communities played greater internal connectance and had longer was related to the occurrence of climatic cycles, food chains, while forest food-webs (‘‘low pro- with mild, wet winters destabilizing communities. ductivity settings’’) tended to be based on detritus, A long-term study (>20 years) showed that Broad- with short food chains and low prey: predator stone Stream (Woodward et al. 2002) has maintained ratios (Figure 5.4), patterns that have also been a broadly consistent food-web structure over very described in detritus-based food-webs from Canada long periods, overlain with a gradual shift in (Tavares-Cromar and Williams 1996), response to a rise in pH since the 1970s. (Closs and Lake 1994), and the United States (Thompson and Townsend 2003). Thompson and Anthropogenic effects on stream Townsend (2003) also compared streams in exotic food-webs pine forests in southern New Zealand with those at two locations in eastern North America and The pervasive effects of human activities on run- found that food-web structure was essentially ning waters, from local to global scales, are not identical, providing support for the occurrence usually analyzed in terms of stream food-webs of ecological equivalence with taxonomically per se (see Malmquist et al. (in press) for a recent different species performing the same functional view of the prospects for the running water roles. environment in general). Here we address three specific issues in relation to stream food-webs: Invasion of stream food-webs catchment land-use change, species introductions- invasions, and larger-scale environmental impacts The deliberate translocation of fish for sport fish- such as air-borne and climate change. eries provides many examples of invasion (e.g. Townsend 1996). Others have occurred through accidental introductions (e.g. Schreiber et al. 2003), Land-use change or as a result of a relaxation of physicochemical Land-use effects are among the most profound constraints (e.g. Woodward and Hildrew 2001). influences on stream food-webs at both local and Not all invaders persist (e.g. Woodward and landscape spatial scales (Woodward and Hildrew Hildrew 2001), but those that do may occupy an 2002c). for agricultural development, unoccupied trophic niche but not exert strong and afforestation with exotic, plantation , effects on prey (e.g. Wikramanayake and Moyle can alter (e.g. Leeks 1992), water 1989), and thus have weak effects on food-web chemistry (e.g. Hildrew and Ormerod 1995) and the structure or processes. Others occupy a similar disturbance regime (e.g. Fahey and Jackson 1997) niche to an existing species and replace it (e.g. Dick over large scales. Conversion of forest to et al. 1993), but may feed at a different rate and reduces shading of the channel, increases the sup- thus alter ecosystem processes (e.g. Power 1990; ply of nutrients, and enhances algal productivity. McIntosh and Townsend 1995). In extreme cases, 62 AQUATIC FOOD WEBS

(a) and algal consumption is reduced as grazing invertebrates become more nocturnal (McIntosh and Townsend 1995). Trout do not appear to influence the structural properties of the food web (i.e. as assessed by the population/community approach), because in general they simply replace galaxiids (Townsend et al. 1998), but in terms of ecosystem processes they cause major changes in the allocation of bio- mass and the flux of materials. Galaxiids consume a moderate amount of invertebrate production, leaving sufficient grazing invertebrates to exert top-down pressure on algal biomass. After inva- sion, however, a trophic cascade is induced, Detritus whereby trout consume a larger fraction of inver- tebrate production, thus reducing grazing pressure (b) on algae, and resulting in an accumulation of algal biomass (McIntosh and Townsend 1995) and higher primary productivity (Huryn 1998). In Broadstone Stream, the population of the dragonfly Cordulegaster boltonii, previously very rare, irrupted probably as a result of reduced acidity (Woodward and Hildrew 2001), and food- web patterns were recorded before, during and after this invasion. The is large, voracious and, by virtue of mouthpart morphology, able to feed on a wide variety of prey items. The invasion by C. boltonii increased mean food-chain lengths, web complexity and omnivory but top-down Detritus Plant Material control of prey abundance was observed for only a few, particularly vulnerable, prey species. No Figure 5.4 Generalized food-web diagrams from (a) grassland species has been lost, probably reflecting the and (b) forested streams showing differences in food-web importance of abundant refugia in the hetero- architecture (Thompson and Townsend unpublished). geneous streambed, though a competing top pre- dator might be deleted in the longer term (Woodward and Hildrew 2001). It is of interest an invader may drive vulnerable competitors that both S. trutta and C. boltonii are large, poly- or prey to extinction or have widespread con- phagous top predators that are able to achieve sequences that resonate through the food web. ‘‘size refugia’’ from predation by all except larger Despite attempts to model the effects of inva- conspecifics (Huryn 1996; Woodward and Hildrew sions on food webs (e.g. Mithen and Lawton 1986), 2002a), characteristics that have been described for empirical studies are rare. The impact of brown certain other successful stream invaders (Power trout (Salmo trutta) has been investigated by com- 1990; Charlebois and Lamberti 1996). paring New Zealand streams with and without this invader. The equivalent were Air-borne pollutants and climate change galaxiid fishes, which are lost through predation and competition from trout (Townsend 2003). The In general, the study of the effects of pollutants presence of trout elicits changes in prey behavior, and toxins on aquatic communities has been limited PATTERN AND PROCESS IN FOOD WEBS 63 largely to ascertaining lethal doses for inverte- Global climate change potentially has an even brates and fish, although a limited number of more pervasive and larger scale effect on food-web stream studies have reported reduced species structure and function. Mean river temperature in diversity (e.g. Winner et al. 1980) and more general Europe probably increased by approximately 1C effects on ecosystem functioning (e.g. Molander during the twentieth Century (Webb 1996) and et al. 1990). Food-web studies are an appropriate similar trends are predicted elsewhere (Matthews unit of study for improving understanding of the and Zimmerman 1990; Hogg and Williams 1996; wider consequences of pollutants in the environ- Keleher and Rahel 1996). The effects of the poten- ment (see Clements and Newman (2002) for a tial consequent increase in productivity and wider discussion of community-level ). nutrient flux in streams are unknown. In addition, The best known in this context are the air-borne global warming may drive species with narrow acidifying pollutants that affect large parts of the thermal tolerances locally extinct (Matthews and globe. Zimmerman 1990). For example, salmonid habitat Acidification of streams affects the biota and may become limited to higher altitudes (Keleher influences the concentration and toxicity of and Rahel 1996), with an overall reduction in the aluminum and other metals (Hildrew and Ormerod area available to fish (Mohseni et al. 2003). Higher 1995; Friberg et al. 1998). Acid reduces stream temperature can also cause reduced invertebrate pH in areas with a susceptible , and density, earlier onset of maturity, and altered sex can further exacerbate the effects of ratios (Hogg and Williams 1996). Higher atmo- acidification (Hildrew and Ormerod 1995). Many spheric CO2 can reduce flux rates in detrital food fish species are intolerant of low pH and their chains via an increase in the C : N ratio (Tuchman absence from acid streams allows large, generalist et al. 2003). Global changes in the El Nino Southern predatory invertebrates to proliferate (e.g. Hildrew Oscillation (ENSO) and the North Atlantic Oscil- et al. 1984; Hildrew 1992), resulting in greater lation (NAO), and more generally in patterns of connectance and a predominance of diffuse links, rainfall and , also have the potential to alter which may enhance stability and reduce the profoundly food-web structure and functioning likelihood of trophic cascades (Schmid-Araya (Puckridge et al. 2000; Bradley and Ormerod 2001). et al. 2002a,b). For example, Closs and Lake (1994) showed Acidification has sometimes been associated increases in food web size, prey : predator ratio, with a reduction in the productivity of algae and food-chain length as a stream food web reas- (Ledger and Hildrew 2000a) and bacteria (Edling sembled after a drying event. and Tranvik 1996), with consequences for nutrient recycling (Mulholland et al. 1987). Sutcliffe and Biodiversity and ecosystem function in Hildrew (1989) describe an apparent switch from stream food-webs an algivore-dominated to a detritivore-dominated community associated with low pH, a pattern Recognition of the implicit links between food-web confirmed by Lancaster et al.’s (1996) analysis of structure and ecosystem processes has under- acid-stream food webs. Specialist invertebrate pinned recent attempts to position food-web eco- grazers, such as mayflies and , are usually logy within the context of the debate about missing from acid streams, which are generally biodiversity and ecosystem function (Paine 2002; less productive than circumneutral systems. Petchey et al. 2004), with significance for both However, Ledger and Hildrew (2000a,b; 2001) ecological theory and (i.e. dealing showed that the suite of acid-tolerant ‘‘detritivor- with the consequences of ). Many ous’’ species that remain grow well on algae and classic studies on biodiversity-ecosystem function are effective grazers, indicating that ecosystem have focused on horizontal (i.e. competitive) processes are maintained despite a loss of biodi- interactions among terrestrial primary producers versity, and that trophic links in stream food-webs (e.g. Lehman and Tilman 2000). More recently, can be somewhat flexible. however, aquatic ecologists have focused attention 64 AQUATIC FOOD WEBS on vertical (i.e. predatory) as well as horizontal species richness (e.g. from three species onwards) interactions (e.g. Downing and Leibold 2002; Paine in stream systems. 2002; Petchey et al. 2004). Because the extent of dietary overlap tends to In temperate forested streams, detrital food increase with resource availability (Nakano et al. chains predominate and decomposition processes 1999; Schmid and Schmid-Araya 2000; Woodward are the means not only by which energy and and Hildrew 2002a), the degree of redundancy in organic matter are incorporated into stream food- stream food-webs, and hence the dependence of webs but are also pathways for the uptake of ecosystem processes on biodiversity, is likely to toxins that can impair ecosystem function (Bird vary both spatially and temporally. Redundancy in and Schwartz 1996). Consequently, leaf-litter the food web may be further enhanced by ‘‘trophic breakdown has been the primary focus of such plasticity’’ (one source of dietary generalism), research in streams (although predation and filtra- whereby individual species alter their feeding tion rates have also been examined more recently; behavior in response to resource availability, see Jonsson and Malmqvist 2003). Most studies so that they are able to exploit alternative food have measured how decomposition is affected by sources when available. The example of grazing manipulation of either species richness of detriti- being taken over by detritivores in species-poor vores (macrofaunal shredder) or of microbial acidic streams may be a case in point (see above: . Shredders are often the dominant Ledger and Hildrew 2000a,b; 2001). Further, a processors, accounting for up to 75% of the mass 7-year litter exclusion experiment revealed that loss of coarse particles (Hieber and Gessner 2002), shredders and microbes responded by shifting although microbial breakdown seems to become their feeding from leaves to wood when leaf litter increasingly important with decreasing latitude was excluded (Eggert and Wallace 2003). (Irons et al. 1994). Many stream studies have Trophic plasticity is also evident across ‘‘guilds’’ shown positive effects of invertebrate species among the primary consumers: for instance, fine richness on decomposition (e.g. Jonsson et al. 2001; particulate and dissolved organic matter bound Jonsson et al. 2002), others have reported no up in eplilthic biofilms can provide a significant relationship or found that species identity, rather energy source to taxa that are traditionally viewed than richness per se, was the main driver (Jonsson as algal grazers, as revealed by the uptake of iso- and Malmqvist 2003). This idiosyncratic pattern of topically labeled biofilms in cave streams by snails results resembles that found in marine ecosystems (Simon et al. 2003). Moreover, extensive sharing of (Emmerson et al. 2001). resources can also occur across trophic levels in A number of factors will determine the nature of streams: many predators are extremely omnivor- the relationship between biodiversity and decom- ous and can consume significant quantities of position. First, there appears to be a high degree of algae, macrophytes, and detritus (Usio 2000; ‘‘trophic redundancy’’ within stream food-webs, as Schmid-Araya et al. 2002a; Woodward and suggested by the prevalence of generalist, as Hildrew 2002a). A high level of redundancy in opposed to specialist, feeding modes (Mihuc 1997; stream food-webs accords with the ‘‘insurance Woodward and Hildrew 2002a), resulting in com- hypothesis’’ of ecosystem function (e.g. Walker plex, reticulate food webs with a high proportion 1992; Naeem 1998), in that trophically equivalent of taxa with similar consumers and resources taxa may be lost without any obvious effects on (Benke and Wallace 1997; Williams and Martinez process rates, at least in the short term, because the 2000; Woodward and Hildrew 2001). Even in remaining taxa are able to compensate for changes quantified webs, intraguild variation in diets may in community composition. be slight, particularly among similar-sized con- It is worth noting that all studies of biodiversity sumers (Woodward and Hildrew 2002a). Trophic and ecosystem function in streams to date have equivalence among species might explain why involved manipulation of a single trophic level curves plotting the increase in a specified eco- (usually the primary consumers, rather than the system process often begin to saturate at quite low basal resources). However, ecosystem processes, PATTERN AND PROCESS IN FOOD WEBS 65 such as leaf breakdown, can be influenced by both gradients (e.g. Ha¨ma¨la¨inen and Huttunen 1996: top-down and bottom-up processes, which can act Wright et al. 2000). In a revealing experiment on simultaneously (e.g. Rosemond et al. 2001). It is the crucial role of species identity, a ‘‘fixed’’ dangerous to generalize when higher trophic sequence in the loss of species of shredders asso- levels have been ignored because there are many ciated with two perturbations, acidification and examples of indirect, predator-mediated impacts eutrophication, resulted in a different response in on links between the primary consumers and the the rate of leaf breakdown, as compared to the basal resources in streams (e.g. Power 1990; effects of ‘‘random’’ losses (Jonsson et al. 2002). Flecker and Townsend 1994; Huryn 1998; Nakano Indeed, species losses from systems are not ran- et al. 1999). Members of the primary consumer dom but are determined by differential vulner- in streams vary in their vulnerability to ability to environmental factors (e.g. pollutants, predation (e.g. Woodward and Hildrew 2002b,d) disturbance) or trophic effects (e.g. invasion/ and the identity of species included in experiments expansion of a predator) and life history char- may skew results, even to the extent of changing acteristics (slow growth rate, low fecundity). the sign of the relationship between biodiversity Studies exploring the consequences of the pre- and ecosystem function (Petchey et al. 2004). dicted loss of species, in the order of their known Petchey et al. (2004) modeled the consequences for vulnerability to particular changes, are the most ecosystem function of species loss in multitrophic likely to offer useful insight into the way that real systems and found that function was affected not systems function and can be managed. only by the number of species that were lost, but Species that are particularly prone to extinction also by the trophic status of the deleted species because of traits such as body size also may con- and whether omnivory was present in the web. tribute disproportionately to stream food-web Another potentially fundamental flaw inherent structure and dynamics, with attendant influences in many biodiversity-ecosystem function experi- on ecosystem function (Petchey et al. 2004). Fish ments is the equal weighting of the species can have particularly strong effects in stream represented (either as numbers or biomass), even food-webs, and may induce trophic cascades that though in real communities geometric or log- alter both decomposition and primary production normal species rank-abundance curves are the rule (e.g. Power 1990; Flecker 1996). However, fish are rather than the exception (Tokeshi 1999). The lack not the only important ‘‘keystone’’ taxa; the larger of realism associated with the equal weighting of members of each guild in a food web are fre- species in biodiversity-ecosystem function experi- quently the main drivers of species interactions ments may be analogous to the equal ranking and and ecosystem processes in streams. For instance, inadequate sampling of species and links in early leaf breakdown is often dominated by the larger connectance food webs, where unrealistic patterns macroinvertebrate shredders such as crayfish and were modeled repeatedly despite the inevitably amphipods (Usio and Townsend 2001; Dangles spurious results (e.g. the supposed rarity of et al. 2004). Similarly, strong top-down effects on omnivory) (Hall and Raffaelli 1993; Polis 1998). primary production are often associated with large A recent study (Dangles and Malmqvist 2004) has grazers, such as snails. Secondary production has highlighted the pitfalls of this over-simplistic also been regarded as driven mainly by larger approach, and revealed the importance of incor- individuals, particularly within taxa but also porating more realistic levels of dominance in between taxa (e.g. hydropsychid caddis among biodiversity experiments. filter-feeders (Benke and Wallace 1997); stoneflies Finally, the random assortment of species, typi- among invertebrate predators (Benke et al. 2001). cal of biodiversity-ecosystem function experi- Large invertebrate predators, such as dragonflies ments, is unrealistic because community assembly (Woodward and Hildrew 2002b) and omnivorous is often nonrandom, as demonstrated by the crayfish (Usio 2000), can also have relatively strong predictability of community (i.e. taxonomic) com- impacts on prey populations, whereas the often far position in streams in response to environmental more abundant small species, such as tanypod 66 AQUATIC FOOD WEBS midge larvae, generally appear to have much fish can have profound impacts on primary pro- weaker or negligible effects (Woodward et al. in duction), it seems that many are relatively weak press). interactors which, if removed from a system, have little discernible effect because their role is assumed Conclusion by another, trophically similar, species. The few biodiversity-ecosystem function experiments that Food webs are complex, and stream food-webs are have been carried out to date with stream assem- no exception with even relatively ‘‘simple’’ depau- blages have shown that, for a given process rate, perate systems, such as acid streams, containing saturation often occurs rapidly as species richness many hundreds of feeding links. However, most increases, lending support to the ‘‘insurance of these links may be energetically and dynamically hypothesis’’ of ecosystem function, which predicts trivial: increasingly, theoretical advances and that high species redundancy may serve to buffer improved empirical data suggest that most links the effects of disturbance. However, as progres- are weak, and that this may stabilize even very sively more species are lost this buffering capacity complex webs (McCann 2000; Neutel et al. 2002). will eventually be impaired, and large changes in The ubiquity of generalist feeding in streams creates ecosystem function can occur when certain species high levels of redundancy within webs, whereby (e.g. top predators; ) are lost. many species are functionally similar, and may also Arguably the most pressing challenge at present is give rise to ecological equivalence among webs, to be able to predict the consequences of species loss whereby different assemblages perform the same fromloticsystems:todothiswewillneedtounite functional role (e.g. at low pH and mayfly the community and ecosystem approaches to food- grazers are replaced by stoneflies, which can act as web ecology, to identify species traits associated facultative grazers and maintain herbivory). with vulnerability to extinction, and then to use this Although some individual species can have very information to perform multitrophic experiments powerful effects on community structure and eco- across a range of scales, in order to develop more system processes (e.g. trophic cascades induced by realistic, predictive mathematical models. PART II Examining food-web theories

Ideally, food-web theory results from the rigorous In aquatic systems, the interplay between models and creative interplay between qualitative or and data is particularly crucial given the tension quantitative models, including appropriate null or between the preservation and exploitation of random models, and observed data collected from many aquatic taxa. Theory can play an important field and lab, along gradients or from experiments. role in conservation, and food-web theory in The following chapters touch on some import- particular is helping to underpin new whole- aspects of food-web theory, including the sta- ecosystem approaches to marine resource manage- tistical analysis of community food webs for ment. For example, analysis of abundance–body testing for patterns of regularities, such as con- mass relationships in marine food webs is pro- stant connectance; testing parametric models as viding new ways to predict species abundance, the cascade model; and ecological network food-web structure, and trophic transfer efficiency analysis and the patterns of information theory in the presence and absence of fisheries exploita- based metrics associated with network size and tion, as well as characterizing changes in those function. properties related to environmental variability. This page intentionally left blank CHAPTER 6 Some random thoughts on the statistical analysis of food-web data

Andrew R. Solow

Introduction Joel Cohen and Fre´de´ric Briand of a freely dis- seminated database called ECOWeB containing For the purposes of this chapter, the term food web well over 100 published food webs. On the basis is taken to mean the binary predator–prey links of these observed food webs, Cohen and his among the species (or other elements) in a biolo- co-workers identified a number of regularities and gical community. The description and analysis of went on to propose a simple one-parameter model food-web data has a long history, dating back in of an S-species food web. This and related work the United States at least to the landmark paper of is discussed in the collected articles by Cohen Forbes (1887). One line of food-web research has et al. (1990). The model proposed by Cohen and been the search for general patterns or regularities Newman (1985) is called the cascade model and is in the structure of observed food webs and the discussed in further detail below. construction of parsimonious models that repro- Neo Martinez and others questioned the degree duce these regularities. It is not clear that treating of resolution of the food webs in the ECOWeB food-web structure in vacuo is either efficient or collection (Martinez 1992, 1993). Upon analyzing a sensible (Winemiller and Polis 1996). If structural smaller number of more highly resolved food regularities are present in food webs, then pre- webs, Martinez found that the regularities identi- sumably they are present for a reason. The most fied in the ECOWeB data did not survive obvious possibility is that certain structural fea- improved resolution and proposed an alternative tures have beneficial implications for the dynamics model—the constant connectance model. The of the community. The connection between food- enterprise did not end there. New and improved web structure and community dynamics has been food-web data have been published on a regular pursued for decades by Robert May (1973), Stuart basis (e.g. Link 2002; Schmid-Araya et al. 2002), Pimm (1982), and many others. This suggests that new regularities have been discovered (e.g. Solow it may be profitable to change the question from and Beet 1998; Camacho et al. 2002), and new ‘‘What regularities are present in observed food models have been proposed (e.g. Williams and webs?’’ to ‘‘What food-web structures have bene- Martinez 2000). ficial implications for community dynamics and The purpose of this chapter is to discuss some are these structures present in observed webs?’’ Be statistical issues that arise in the analysis of food- that as it may, this chapter will focus exclusively web data. There have been many informal statist- on the analysis of food-web data. ical analyses of food-web data. The relatively small To search for general patterns and to construct existing body of formal statistical work in this area models, it is necessary to have a reasonably large has focused on assessing the goodness of fit of the collection of observed food webs. A significant cascade model (Solow 1996; Neubert et al. 2000) or contribution in this area was the compilation by testing proposed regularities (e.g. Murtaugh 1994;

69 70 AQUATIC FOOD WEBS

Murtaugh and Kollath 1997; Murtaugh and The likelihood and statistical inference based upon Derryberry 1998). This chapter focuses on statistical it are discussed in the monograph by Azzalini issues in modeling food-web data. The current (1996). What follows is a very brief review of the ^ practice in this area is, first, to identify regularities basic theory. The maximum likelihood estimate y in the data and, second, to show via simulation of y is found by maximizing L(y) (or its logarithm) that a proposed model either does or does not over y. The likelihood can also serve as a basis exhibit the same behavior. Although it can be for testing the null hypothesis H0: y ¼ y0 that useful, from a statistical perspective, this approach the parameter y is equal to a specified value y0 is ad hoc. Not only are its statistical properties against the general alternative hypothesis H1: unknown, but regularities do not (usually) con- y 6¼ y0. Specifically, the likelihood ratio test statistic stitute a complete food-web model in the sense is given by: described below. The main goal here is to discuss a ^ L ¼ 2( log L(y) log L(y )): (6:3) more formal approach to modeling. 0 Provided n is large enough (and other regularity conditions hold), under, H , L has an approximate Likelihood methods for 0 w2 distribution with degrees of freedom given food-web models by the dimension dim(y)ofy, so that H0 can be

Consider a community consisting of S species s1, rejected at significance level a if the observed value ... L s2, ,sS. The term species is used loosely here. of exceeds the upper a-quantile of this distri- It is common in food-web analysis to refer to bution. A 1 a confidence for y is given by trophic species. A trophic species is defined as a the set of values of y0 for which H0: y ¼ y0 cannot set of one or more species with the same predators be rejected at significance level a. Finally, to test and prey. The food web for this community can the overall goodness of a fitted model (an exercise be represented by an S-by-S binary predation or inferior to testing against an alternative model), ^ adjacency matrix a ¼ [ajk] where: if the fitted model is correct, then 2 log L(y) has an approximate w2 squared distribution with dim(y) a ¼ 1, if s preys on s jk k j degrees of freedom. ¼ 0, otherwise: ð6:1Þ

This predation matrix is viewed as a realization of Some examples a matrix A. A random variable is As an illustration, consider the problem of fitting completely characterized by its distribution. Thus, the cascade model of Cohen and Newman (1985). a food-web model is complete if it completely Under the cascade model, there is an unknown specifies the distribution of A. As noted, unless a ranking r ¼ (r(s )r(s ) r(s )) of the S species in set of descriptive features can be shown to com- 1 2 s a food web and the elements of A are independent pletely characterize the distribution of A, it does with: not constitute a complete description. Let fy be the probability mass function of A under a particular prob(Ajk ¼ 1) ¼ 0, if r(sj) r(sk) food-web model, where y is the unknown (possi- ¼ g=s, otherwise, (6:4) bly vector-valued) parameter of the model. Unlike g g S a collection of regularities, fy(a) is a complete food- for an unknown constant with 0 . That is, ... under the cascade model, species can prey only on web model. Also, let a1, a2, , an be the observed predation matrices of a random sample of n food other species of lower rank and the probability that ... they do so is inversely proportional to the size of webs of sizes, S1, S2, , Sn, respectively. Statistical inference about the parameter y can be the food web. A consequence is that the expected based on the likelihood function: number of links increases linearly with web size S, 2 Yn and not with S as would be expected if predation L(y) ¼ fy(ai): (6:2) probability is independent of S. Note that, if the i¼1 species are relabeled so that, r(sj) ¼ j, then the STATISTICAL ANALYSIS OF FOOD-WEB DATA 71 matrix A will be upper triangular. If no such size S is a random variable with mean p(S) ¼ g/S relabeling is possible, then the food web contains a and variance: loop, which is forbidden under the cascade model. Var P(S) ¼ d p(S)(1 p(S)): (6:6) In terms of the previous notation, for a single food web, under the cascade model, the parameter y Under this model, which is due to Williams (1982), includes the ranking r and the constant g. The the expected number of links is the same as under likelihood is identically 0 if ar(sj) r(sk) ¼ 1 for any pair the cascade model, but the variance is inflated of species sj,sk with r(sj) r(sk). Any ranking with by the factor 1 þ (S 1)d. Using this model, this feature is inadmissible. Let y be the sum of the Solow (1996) showed that the smallest value of d elements of a. For any admissible ranking, the that explains the extra-binomial dispersion in the likelihood is given by the binomial probability: number of links is around 0.023. To put this result into context, under this model, the greatest vari- ! S ability in predation probability occurs when S ¼ 8, S y = y = 2 : : fy(a) ¼ 2 (g S) (1 g S) (6 5) in which case the expected predation probability is y around 0.5 with standard deviation around 0.08. Murtaugh and Kollath (1997) also found over- Importantly, this likelihood is the same for all dispersion in food-web data. admissible rankings. The overall likelihood for a Like the cascade model, almost all food-web collection of n independent webs—each with at models include a ranking of species. However, the least one admissible ranking—is just the product cascade model is special in the sense that the over n terms each with the binomial form (6.5). likelihood is the same for all admissible rankings. However, if even a single food web in this collec- In statistical terminology, under the cascade tion has no admissible ranking, then the overall model, for any admissible ranking, the sufficient likelihood is identically 0. statistic y is ancillary for the ranking. As a result, in Solow (1996) used likelihood methods to fit and fitting the cascade model, it is not necessary to test the cascade model to 110 ECOWeB food webs. estimate the true ranking of the species. This con- First, the cascade model was fit by maximum genial property is not shared by even the simplest likelihood. The estimate of g is 4.11, which is close alternatives to the cascade model. For example, to the value of 4 proposed by Cohen. However, the Neubert et al. (2000) tested the cascade model goodness of the fitted cascade model is easily against several alternative models in which pre- rejected. One possible explanation is that the sys- dation probability varied with the rank of the tematic part of the model—namely, a predation predator, the rank of the prey, or both. In this probability exactly inversely proportional to S—is application, rank was determined by body mass. incorrect. To explore this possibility, Solow (1996) Consider the first possibility, which Neubert et al. fit a more general model under which predation (2000) referred to as predator dominance—see, probability is inversely proportional to Sb. Cohen also, Cohen (1990). If the rank of each species is (1990) referred to this as the superlinear-link- known, then the problem amounts to testing scaling model. The maximum-likelihood estimate whether the proportion of potential prey that is of the parameter b is 0.87, which is close to the actually consumed is constant across predators. value of 1 under the cascade model. However, the This can be done through a standard w2 test for goodness of this fitted model is also easily rejected. homogeneity of probabilities in a 2-by-S contin- An alternative explanation of the lack of fit of gency table. Briefly, the two rows in the table the cascade model is that the predation probability represent ‘‘eaten’’ and ‘‘not eaten,’’ the S columns for food webs of size S is not fixed, but varies represent the species according to rank, and the from food web to food web, resulting in extra- entries in the table represent the number of binomial variation in the number of links. To potential prey that are and are not eaten by each explore this possibility, Solow (1996) assumed that species (so, for example, the sum of the entries in the predation probability P(S) for a food web of column j must be j 1). Of course, to construct this 72 AQUATIC FOOD WEBS table, it is necessary to know the rank of each will serve as the test statistic. Second, randomize species. Neubert et al. (2000) used the species size the location of the 1s in the upper triangular part as a for rank, which is certainly a reasonable of the observed predation matrix. This produces approach (Cohen et al. 1993). However, the pro- an equally likely realization of the cascade model blem is more difficult if the species ranks are not conditional on the number of links. Third, as known, for in that case, the number of potential before, find the maximum value of the w2 statistic prey that is not eaten by each species is unknown. over the linear extensions of the partial order Under the assumption that species can only prey imposed by the (randomized) feeding pattern. on species of lower rank, the observed food web Fourth, repeat the randomization procedure a large a determines a partial ordering of the species. number of times and estimate the significance level For example, suppose that, in a four-species food (or p-value) by the proportion of randomized food 2 web, s1 is observed to prey on s2 and s3 is observed webs with values of wmax that exceed the observed to prey on s4. If these are the only observed spe- value. cies, then there are six possible complete rankings: > > > in brief, but obvious notation, (1 2 3 4), Conclusion (1 > 3 > 2 > 4), (3 > 1 > 2 > 4), (3 > 1 > 4 > 2), (3 > 4 > 1 > 2), and (1 > 3 > 4 > 2). These are called To an outsider, the recent history of food-web the linear extensions or topological sortings of the research—at least, the kind of food-web research partial ordering (1 > 2, 3 > 4). An algorithm for discussed in this chapter—seems absolutely generating the linear extensions of a partial remarkable. The availability of the ECOWeB food ordering is given in Varol and Rotem (1981). This webs led to a burst of activity involving the iden- algorithm is relatively simple, but not computa- tification of structural regularities and the develop- tionally efficient. A more efficient, but less simple, ment of models. However, virtually none of this algorithm is given in Pruesse and Ruskey (1994). work—some of which is quite elegant—survived To give an idea of the statistical complications the realization that the ECOWeB webs were gross arising from an unknown ranking, return to the caricatures of nature. Ecologists are now left with a test for predator dominance of Neubert et al. relatively small number of what appear—but are (2000). The following approach—which is not not guaranteed—to be more realistic food webs. directly based on the likelihood—could be used It seems that some sorting out is now in order to when the species ranking is not known. First, help ensure that future research in this area lies identify the linear extension that gives the largest along the most profitable lines. An admittedly 2 2 value wmax of the w statistic in the corresponding small part of this sorting out ought to focus on 2-by-S contingency table. This is the ranking that proper ways to approach statistical inference in 2 most favors predator dominance. The value wmax food-web models. CHAPTER 7 Analysis of size and complexity of randomly constructed food webs by information theoretic metrics

James T. Morris, Robert R. Christian, and Robert E. Ulanowicz

Introduction predicated on the importance of autocatalysis (i.e. loops) during growth and Understanding the interrelationships between development. Autocatalysis has both extensive properties of ecosystems has been a central theme and intensive properties. The extensive metric of ecology for decades (Odum 1969; May 1973). (characteristic of size) is the total amount of flow Food webs represent an important feature of eco- within the system, that is total system throughput systems, and the linkages of food-web properties (TST). When the intensive metric (characteristic have been addressed specifically (Pimm 1982; of complexity) of flow diversity is scaled by TST, Paine 1988). Food webs consist of species net- the result is called the system’s developmental works, and the effect on ecosystem processes of capacity. Flow patterns become increasingly the imbedded species diversity has become a constrained as a system develops and matures, a major focal point for research (Tilman 1997; characteristic represented by average mutual McGrady-Steed et al. 2001; Chapin et al. 2000). In information (AMI). When AMI has been scaled by recent years the properties of food webs have been TST, the ensuing quantity is called the system placed into a network perspective, but a major ascendency. These metrics, and related ones, pro- limitation to addressing these properties is sim- vide ways to gauge and characterize systems and plification associated with data availability and changes within systems. For example, eutrophica- software limitations (Paine 1988; Cohen et al. tion appears to increase system ascendency via an 1993). One major simplification involves aggregat- increase in TST that more than compensates for of species into guilds or trophic species a drop in AMI (Ulanowicz 1986b). This quantita- to reduce the number of taxa to be considered tive definition distinguishes simple enrichment (Ulanowicz 1986a). Sizes of published food-web (which does not induce a drop in AMI) from networks are less than 200 taxa (Martinez 1991) eutrophication (which does). when binary feeding relationships (i.e. diet item or The objectives of this study were twofold. One not) are considered, and no more than 125 taxa objective was to assess the effects of different (Ulanowicz et al. 1998) when feeding relationships assumptions used to construct the hypothetical are quantified (i.e. as fraction of total diet). We webs. A second objective was to evaluate how have constructed hypothetical food webs with recent theoretical metrics of food-web organization weighted feeding relationships that overcome the and complexity (Ulanowicz 1986a, 1997) respond limitation of aggregation. Food webs ranged in to two key variables that vary with ecosystem size up to 2,426 taxa. development, namely web size and function. Ulanowicz has developed several metrics of Size was examined in terms of the number of taxa system growth and development based on informa- and amount of processing (i.e. TST). Function tion theory (Ulanowicz 1986, 1997). They are

73 74 AQUATIC FOOD WEBS was evaluated on the basis of the number and allochthonous import of organic matter. GPP was distribution of connections among taxa. We evalu- set at 1,000 kcal m2 per year. Allochthonous ated the interrelationships among size, con- import was 100 kcal m2 per year. Total input to nectivity, and network metrics of both hypothetical unstructured webs was fixed at 1,100 kcal m2 per and empirical webs. year. These standard conditions facilitated com- parison of networks. The food webs were donor-controlled in the Methods sense that the flow from one taxon to the next was Food-web construction proportional to the biomass of the donor taxon. The model structure was purely descriptive and We generated a large diversity of food webs for not meant to have any predictive capability of analysis by creating thousands of randomly con- temporal dynamics. Rather, every feasible food structed networks varying in size (number of taxa) web generated is a description of possible energy and distribution of connections. Random, donor- flows within a hypothetical network constructed controlled food webs were generated by populat- from realistic principles. We posit that real food ing a transfer matrix (A) with random coefficients webs lie within the state space of these randomly that were drawn from realistic probability sets, generated food webs, provided that the sample randomly partitioning exogenous inputs (f) among size of hypothetical webs is sufficiently large. primary producers (gross production), and solving Solving for the steady state (equation 7.2) from for the biomass vector (x) for the steady-state a randomly generated transfer matrix can result in condition: negative biomass numbers for any number of taxa. dx=dt ¼ 0 ¼ f þ Ax, ð7:1Þ Such webs were identified as being nonfeasible and were discarded. Hypothetical food webs were x ¼A1f, ð7:2Þ generated until a sufficiently large population of where f represents a vector of inputs (gross feasible (positive biomass for all taxa) webs were primary production, autochthonous inputs). obtained to make a convincing analysis of their Inputs to all structured webs (Figure 7.1) were properties and trends. Four kinds of networks divided into gross primary production (GPP) and are considered here (Figure 7.1) and the rules are described below. We constructed networks using different sets Trophic Transfer of rules concerning the nature of connections. relationships coefficients Every structured food web consisted of ‘‘taxa’’ that were each defined as belonging to a group Uniform (0–1) of primary producers, primary, secondary, and Structured tertiary consumers, or detritivores, plus an organic Lognormal matter or detritus compartment (Figure 7.2; plate 2). For each web that was generated, the number of taxa within each group was determined randomly. Uniform (0–1) By this method we generated feasible webs that Unstructured consisted of as few as 7 or as many as about Lognormal 2,200 taxa. For each structured web we generated three Figure 7.1 Terminology and classification of hypothetical food web types of connections among food-web members construction. Realistic trophic structure was imposed on structured (Figure 7.2). These included mandatory flows webs (Figure 7.2), while the taxa in unstructured webs were randomly from all taxa to organic matter or detritus, man- connected without regard to trophic identity. Structured webs had no constraints on the relative numbers of taxa within each trophic level. datory flows from lower to higher trophic Transfer coefficients were drawn either from lognormal or uniform levels, and nonobligatory flows among taxa, such distributions. as a flow from a secondary consumer to a primary INFORMATION THEORETIC METRICS 75

1,000 Primary Primary Secondary Tertiary producers consumers consumers consumers

Organic 100 Detritivores matter

Figure 7.2 This figure shows the basic architecture of structured food-webs constructed with realistic trophic relationships. The energy input to every randomly constructed food web was standardized at 1,000 kcal m2 yr1 of GPP and 100 kcal m2 yr1 of exogenous organic matter input. Solid arrows (!) denote mandatory flows of energy and their direction. For example, every primary consumer in the web is made to consume at least one primary producer. Dotted arrows ( ) denote flows that are possible, but not mandatory. The rules for making these connections are discussed in the text. consumer. Consumer taxa belonging to the same The structure we have described above is trophic group were not connected. With flows general, and definitions of trophic components and of the second and third kind, a random number flows are flexible. For example, loss from the generator was used to determine the presence of organic matter compartment can be defined as a flow between two specific nodes. Mandatory either a respiratory flow and/or an export from flows from lower to higher trophic levels defined the system depending on how the microbial com- the trophic positions of taxa. For mandatory flows munity is defined. If detritivores include microbes, from lower to higher trophic levels, such as flow then export from organic matter is best defined as from primary producers to herbivores, the prob- a loss from the system rather than a respiratory ability of flow from a lower to a higher-level taxon loss. At the level of the whole web, the entirety of was assumed to be inversely proportional to living components can be thought of as including the number of taxa within the lower trophic level. microbes that are parasitic or saprophytic on For example, with four primary producers in the multicellular organisms from neighboring trophic food web, there was a 25% probability that any levels. one of them flowed to any one of the herbivores. A second type of web architecture, termed Unit random numbers were drawn for each pair of unstructured (Figure 7.1), was purely random in donor and recipient taxa to make the connections, construction. There were no rules of trophic subject to the requirement that every higher-level structure of any kind to govern the presence or taxon had to feed on at least one lower level absence of flows, except that the distribution of taxon. Connections of the third kind, or non- GPP was restricted to a group of taxa representing obligatory flows as from a tertiary to a secondary between 5% and 75% of total taxa in the web as consumer, were made using a unit random num- determined by random number. Every taxon, ber drawn for each possible pair of taxa after first including ‘‘primary producers,’’ had an equal drawing a unit random number that set the overall probability of feeding on any other taxon as connectivity of the web. For each one of these determined by the overall connectivity of the web, possible connections, a connection or flow was which was determined by a uniformly distributed made if the value of a unit random number was random variable ranging from 0.2 to 0.7 In addit- less than the overall connectivity. This procedure ion, these unstructured networks had no detritus generated food webs that ranged from sparsely to component to accumulate and degrade the pro- densely connected. ducts of mortality. These networks were simply 76 AQUATIC FOOD WEBS random distributions of respirations and of Network analysis exogenous inputs (gross production) and feeding For every feasible food web generated, we com- among nodes. puted total system biomass, the number taxa within each trophic group, the net production, and Respiration rates total respiration within each trophic group, the number of connections, total system throughput, Weight-specific respiration rate coefficients were and various metrics derived from information randomly chosen from uniform distributions theory (Ulanowicz 1986, 1997). These included spanning ranges representing realistic rates taken average or AMI, flow diversity from the literature. The respiration of primary (SI), total system throughput (TST), full develop- producers consisted of two parts: dark respira- ment capacity (Cd ¼ TST SI), and ascendancy tion and . Photorespiration was (TST AMI). accounted for by subtracting a fraction equivalent Average mutual information is inversely to 25–50% of the rate of GPP. The rate of photo- proportional to the degree of in a respiration was randomly chosen for each primary system and is a measure of the predictability of producer. A weight-specific rate of dark respirat- flow. Adopting the convention that flows can ion ranging from 40% to 70% of plant weight be represented by a matrix T (row-column order) per year was also randomly chosen for each prim- ij in which flow moves from a species in column ary producer (Lambers 1985; Landsberg 1986). j to a species on row i, then AMI (bits) was Heterotrophs (consumers and detritivores) had computed as: specific respiration rates ranging from 19% to X X 3,400% per year of their biomass, which is a = = AMI ¼ pðTijÞ log2½fpðTijÞ pðTjÞg pðTiÞ, range that spans a wide spectrum of body size i j (Gordon et al. 1972). For each heterotoph, a unit ð7:3Þ random number was drawn to assign a respiration rate within the appropriate range. Based on where Tij is the flow from species j to species i, numerous sources (Morris and Lajtha 1986; p(Tij) is the joint probability given by: Schomberg and Steiner 1997; Asaeda and Nam pðTijÞ¼Tij=TST, ð7:4Þ 2002; Van Santvoort et al. 2002; Qualls and X X Richardson 2003; Salovius and Bonsdorff 2004), pðTiÞ¼ pðTijÞ; and pðTjÞ¼ pðTijÞ, detritus was given an annual decay rate ranging j i from 6.6% to 164%. In the case of the unstructured networks, the specific respiration rate assigned to and where TST is total system throughput: each taxon was like that of the heterotrophs. X X TST ¼ jTijjð7:5Þ i j Transfer coefficients The Shannon flow diversity (SI) is based on the Transfer coefficients representing energy flow individual joint probabilities of flows from each among taxa were either drawn from a population species j to each species i: of unit random numbers (0–1) or from a lognormal X X with m ¼2.3 and d ¼ 2, where the mean of the : : 2 SI ¼ ½pðTijÞ log2ðpðTijÞ ð7 6Þ distribution is given by exp(m þ d /2). The log- i j normal distribution allows for coefficients greater than one as would be the case for biomass turn- Empirical food webs over rates greater than once annually. The two alternative methods of generating transfer coeffi- The hypothetical data were compared against data cients were applied to the structured webs as well consisting of biomass and flow distributions of as to the unstructured webs. 31 empirically derived food webs from various INFORMATION THEORETIC METRICS 77 ecosystems. The empirical webs were analyzed of the means of generating transfer coefficients using the same suite of metrics as described above. (uniform or lognormal). The probability distribution However, the biomass and flows of the empirical was described by a sigmoid: Y ¼ 1/{1þexp(.02X–11)}, food webs were first scaled so as to transform where Y is the probability and X is the number the total GPP of the empirical webs to a constant of taxa. For webs with 500 taxa the probability of 1,000 units across all webs in order that they successfully generating a feasible random web was would be consistent with the GPP of hypothetical 0.5. For webs with 1,000 taxa the probability declined webs and to remove GPP as a variable. The pro- to 0.0001. cedure was to scale the input vector f and absolute For each web generated, we calculated the actual flows by the quotient 1000/Sf. number of connections as well as the maximum possible number of connections, which depends on size as well as the specific distribution of taxa Results among trophic groups (Figure 7.4). Among struc- tured webs, the total number of connections per Food-web dimensions web averaged 280,763 324,086, with a maximum Food webs that have been described empirically of 1,885,606 and a minimum of 23. The number of are necessarily aggregated to various degrees and connections per taxon averaged 263 220, with a typically overlook connections or flows that are maximum of 1050 and a minimum of 2. The set of minor. The average number of taxa in the set of hypothetical webs varied greatly in the total empirical food webs was 39 36 (1 SD). The number of connections and connections per taxon, largest empirical food web contained 125 taxa; irrespective of size (number of taxa) (Figure 7.4). the smallest contained 4. Empirical food webs Both sparsely and densely connected webs were averaged 350 563 (1 SD) total connections and generated, even among the largest sizes. The 5 5 connections per taxon (Table 7.1). The mini- average number of major (flows greater than 5% of mum and maximum numbers of connections per the total input flows) flows per taxon in the set of web were 4 and 1,969. Minimum and maximum hypothetical webs was similar to the empirical connections per taxon were 1 and 15.7. webs and averaged 2.1 1.2 per taxon. This may The accounting of flows or connections among be an artifact of a process that produces a dis- the taxa that comprise empirical food webs is tribution like that of the broken stick model likely to miss rare items in an organism’s diet, and (MacArthur 1957), but the distribution of flow we arbitrarily defined rarity as any flow that con- stems ultimately from the transfer coefficients stitutes 5% or less of an organism’s diet. Thus, in a donor-controlled fashion, and their distribu- major flows were greater than 5% of an organism’s tion does not resemble a broken stick distribution diet. Using this definition of a major flow, the (Figure 7.5). On the other hand, GPP imposes theoretical maximum number of major flows per a limit on flow in a manner similar to the resource taxon approaches 20, provided the flows into space that MacArthur (1957) argued should set a taxon are equal in magnitude. The theoretical a limit on the distribution of species. Ulanowicz minimum number of major flows is zero as, for (2002) used an information- theoretic homolog of example, when there are greater than 20 uniform the May–Wigner stability criterion to posit a flows. Among empirical food webs the number of maximal connection per taxon of about 3.01 (which major flows per taxon averaged 2.3 1, and ranged accords with these results). from 1 to 4.5. The values of individual fractional flows into Hypothetical webs had a mean size of 827 444 each taxon as a percentage of the total flow into taxa per web, with a maximum size of 2,426 and each taxon were computed for each food web a minimum of 7. The probability of generating a generated, excluding gross photosynthesis, by feasible (biomass of all taxa must be positive), type of web construction. For each food web gen- structured web declined precipitously at sizes erated, a frequency distribution of these fractional exceeding 500 taxa (Figure 7.3) and was independent flows was computed, and then the global average 78 AQUATIC FOOD WEBS

Table 7.1 Summary statistics from empirical (n ¼ 31) and hypothetical food-webs. Statistics were computed from 800 samples each of the lognormal and uniform variety of unstructured webs and 641 and 1006 samples of structured lognormal and uniform webs, respectively

Variable Web type Mean Standard deviation Minimum Maximum

Number of connections Empirical 5.1 4.8 1.0 15.8 per taxon Unstructured lognormal 349.7 242.6 1.3 1025.3 Unstructured uniform 358.8 247.2 2.0 1049.7 Structured lognormal 179.7 156.5 3.2 849.0 Structured uniform 172.6 138.6 4.2 829.6 Number of major connections Empirical 2.3 1.0 1.0 4.5 per taxon Unstructured lognormal 3.1 0.8 1.5 4.9 Unstructured uniform 0.8 0.9 0.0 5.1 Structured lognormal 2.9 0.5 1.0 4.0 Structured uniform 1.7 0.8 0.5 5.4 TST (kcal m2 yr1) Empirical 3,412 662 2,175 4,684 Unstructured lognormal 18,120 11,311 2,209 50,934 Unstructured uniform 14,051 7,992 2,299 36,381 Structured lognormal 4,117 801 2,382 11,250 Structured uniform 4,052 637 2,912 9,133 AMI Empirical 1.6 0.3 1.0 2.0 Unstructured lognormal 3.3 0.6 1.0 4.5 Unstructured uniform 1.6 0.4 0.9 2.5 Structured lognormal 2.7 0.4 1.2 4.0 Structured uniform 2.0 0.3 1.2 2.9 Flow diversity (bits) Empirical 4.0 1.0 1.8 5.7 Unstructured lognormal 13.4 3.2 1.3 17.5 Unstructured uniform 15.0 3.7 2.0 19.8 Structured lognormal 9.2 1.7 3.2 14.7 Structured uniform 10.5 1.7 4.2 16.5 Full capacity Empirical 13,905 5,001 3,935 25,104 Unstructured lognormal 273,323 205,619 2,892 889,544 Unstructured uniform 236,335 167,057 4,824 720,361 Structured lognormal 39,146 15,008 7,736 165,117 Structured uniform 43,379 13,731 13,059 143,630 Ascendency as a fraction Empirical 0.41 0.07 0.31 0.64 of full capacity Unstructured lognormal 0.26 0.04 0.19 0.76 Unstructured uniform 0.12 0.05 0.05 0.52 Structured lognormal 0.30 0.03 0.23 0.38 Structured uniform 0.20 0.04 0.11 0.29

frequency distribution of fractional flows was The distributions of transfer coefficients were computed by the type of web (Figure 7.5). also computed and averaged over all webs within Regardless of the type of web construction, the each class of web construction to determine if they overwhelming majority of flows were less than differed from the distributions expected on the 5% of the total flow into each taxon. Random basis of the random number generator. If the dis- webs without realistic structures had the highest tributions differed from the expected, then the web percentage of these small flows; 98% 6(1 SD) structures must be selecting the distribution. Webs of flows were less than 5% of the total. Among constructed of transfer coefficients drawn from webs with structure, 95% 9 of flows were of a the uniform, unit distribution had frequencies magnitude less than 5% of the total. exactly as expected (Figure 7.5). Webs constructed INFORMATION THEORETIC METRICS 79 of transfer coefficients drawn from lognormal constraints. TST among structured webs was distributions also did not differ from the expected. bounded by upper and lower asymptotes. These Note that transfer coefficient values greater than 1 asymptotes were sensitive to GPP (not shown), were grouped into one category for analysis, which in these examples was always set to which accounts for the upturn in frequency on the 1,000 kcal m2 per year. Mean TST among struc- tail of the distribution. Among webs constructed of tured webs was about 4,000 kcal m2 per year lognormally distributed transfer coefficients, 12.5% (Table 7.1) or four times GPP. That this mean is 0.3 of coefficient values were greater than one. very near to four times GPP tells us that the mean effective number of trophic levels ( ¼ TST/GPP) is about 4, which is consistent with real food-webs. Total system throughput Of course, the structured webs were designed Total system throughput, which is simply the with 4 trophic levels (Figure 7.2), but it does sum of all flows among the taxa, including respirations and exogenous flows (i.e. GPP, losses and allochthonous inputs), was bounded within 1.0 the structured webs. Among the unstructured webs, TST increased with web size without limit, 0.8 at least within the range of the food-web sizes Y = 1/{1 + exp(0.02x – 11)} generated (Figure 7.6; Plate 3). This increase in 0.6 TST, seemingly without limit and in defiance of the laws of physics, was a consequence of the high 0.4 degree of flow reciprocity that occurs among Probability unstructured webs constructed without realistic 0.2 trophic relationships. TST increased by approxi- mately 15 and 21 kcal m2 per year per taxon in 0.0 unstructured uniform and unstructured log- 0 100 200 300 400 500 600 700 800 normal webs, respectively. TST was limited among Network size (Total number of nodes) structured webs, as finite inputs dictate a finite Figure 7.3 Probability of generating a feasible, random network TST (Ulanowicz 1997), at least when food webs with realistic structure using transfer coefficients drawn from either have realistic structures and thermodynamic uniform ( ) or lognormal ( ) distributions.

3.0 3 106 1,400

2.5 3 106 Maximum possible1,200 Maximum possible 1,000 2.0 3 106 Actual800 Actual 1.5 3 106 600 1.0 3 106 400 3 Number of connections

3 Connections per taxon 500.0 10 200

0.0 0 0 500 1,000 1,500 2,000 0 500 1,000 1,500 2,000 Number of taxa Number of taxa

Figure 7.4 Density of connections of structured, hypothetical food webs as a function of the number of taxa. Shown are the maximum possible number of connections ( ), determined by the greatest number of feasible connections for a particular distribution of taxa, and the corresponding actual number of connections ( ) for approximately 6,800 networks (see also Plate 2). 80 AQUATIC FOOD WEBS support the choice of coefficients used to 2 connections per taxon, TST was either greater construct the hypothetical webs. The mean TST than 40,000 or less than 10,000 kcal m2 per year, (normalized to GPP ¼ 1,000 kcal m2 per year) of while at the highest connectivity, TST focused on a the empirical webs was 3,400 662 kcal m2 per single cluster less than 10,000 kcal m2 per year. year (Table 7.1). This suggests a strong and peculiar control by the The pattern of TST when plotted against the structure of the web over its function. TST of average number of major (>5% flow) connections structured and empirical webs was independent of per web was quite different between the two types the connectivity. of unstructured webs and between the unstruc- tured and structured webs (Figure 7.6). Among Flow diversity unstructured normal webs, TST was greatest at low connectivity and declined toward a lower Flow diversity (SI) among empirical and hypo- asymptote as connectivity increased. In contrast, thetical webs increased with size (taxa) (Figure 7.7; the TST of unstructured, lognormal webs was Plate 4). Unstructured, uniform webs had the bifurcated (Figure 7.6). For example, for webs with highest SI, averaging 15 3.7 bits, empirical webs

100 40 From random webs 35 95 From structured webs 30 25 From lognormal distributions 90 20 From uniform distributions 4 15 10 2 5 Frequency of occurence (%) of occurence Frequency Frequency of occurence (%) of occurence Frequency 0 0 0 20 40 60 80 100 0.0 0.2 0.4 0.6 0.8 >10 Percent of total input flow Transfer coefficient values

Figure 7.5 Average frequency distribution of flow strengths into taxa expressed as fractions of the total flow into each taxon (left). Average frequency distribution of transfer coefficients (right). Note that the transfer coefficient is equivalent to the fraction of the standing stock of a donor taxon that flows into a recipient taxon in a unit of time in steady state.

(a)20,000 (b) 60,000

50,000

) 15,000 –1

yr 40,000 –2 10,000 30,000

20,000

TST (kcal m 5,000 10,000

0 300 600 900 1,200 1,500 012345 Number of taxa Major connections per taxon

Figure 7.6 TST of different types of hypothetical food-webs as a function of (a) the number of taxa and (b) of the average number of major connections (flows >5% of inputs) per taxon. Hypothetical webs were either unstructured in design with uniformly or lognormally distributed transfer coefficients, or structured in design with uniformly or lognormally distributed transfer coefficients. Empirical webs are denoted by (see also Plate 3). INFORMATION THEORETIC METRICS 81

(a)20 (b) 20

15 15

10 10 SI (bits)

5 5

0 0 300 600 900 1,200 1,500 023451 6 Number of taxa Major connections per taxon

Figure 7.7 Flow diversity (SI) of different types of hypothetical food webs as a function (a) of the number of taxa and (b) of the average number of major connections (flows >5% of inputs) per taxon. Hypothetical webs were either unstructured in design with uniformly or lognormally distributed transfer coefficients, or structured in design with uniformly or lognormally distributed transfer coefficients. Empirical webs are denoted by (see also Plate 4). had the lowest (4 1 bits). The SIs of empirical and to flow diversity and total system throughput, we hypothetical webs were equivalent among webs of can place limits on Cd. For fully connected, comparable size, while structured webs had lower unstructured webs of uniform flow and m2 m 2 flow diversities for a given size than unstructured flow paths, Cdmax ¼TST log2[1/(m m)] ¼ 2 webs (Figure 7.7, Table 7.1). For a web with m taxa amlog2[1/(m m)], where a is a proportionality and m2 flow paths, the SI reaches an upper limit constant, equivalent to 15 and 21 kcal m2 per year 2 of log2(1/m ), provided the flows are uniform per taxon in unstructured uniform and unstructured in size. Alternatively, if the flows among taxa are lognormal webs, respectively. For an arbitrary range dominated by n major flows per taxon, then the of sizes starting with m ¼ 100, this gives 20 103 to 3 2 3 SI of structured webs reduces to log2(1/nm), 28 10 kcal bits m per year and 166 10 to which gives a limit for SI of about 12.1 for a large 232 103 kcal bits m2 per year for m ¼ 600, depend- web of size m ¼ 1500 and n ¼ 3. The latter calcula- ing on a, and these values are consistent with the tion is consistent with the results obtained Cd observed for the unstructured webs without from the structured webs (Figure 7.7). Note that realistic trophic structure (Figure 7.8; Plate 5). the mean number of major connections per taxon For webs with realistic structure or webs with for this web type was about 3 (Table 7.1). Hence, limited TST, a fully connected web also would 2 the imposition of a realistic structure seems to be have Cd ¼TST log2[1/(m m)], which for selecting for webs with a flow diversity that is the hypothetical webs here, were TST 4GPP, dominated by a few major flows among taxa, we estimate that Cdmax would be about whereas the unstructured webs are more fully 84 103 kcal bits m2 per year when m ¼ 1,500. This connected and have greater uniformity of flow is near the upper range of Cd observed among (Figure 7.5). the hypothetical webs with realistic structure (Figure 7.8). When a web is dominated by a few flows per taxon, a more realistic example, the Full development capacity Cdmax is approximated by TST log2(1/3m), 2 The total potential order or complexity (kcal bits m which gives Cdmax 4 GPP log2(1/3m)or per year) is termed full developmental capacity (Cd) 35 103 kcal bits m2 per year. This value is in the and is the product of TST and the flow diversity (SI) middle of the observed Cd for structured webs (Ulanowicz 1986; Ulanowicz and Norden 1990). (Figure 7.8) and close to the mean Cd of the Based on the above generalizations about the limits structured, lognormal webs (Table 7.1). 82 AQUATIC FOOD WEBS

(a) 160 3 103 (b) 160 3 103 140 3 103 140 3 103 120 3 103 120 3 103 100 3 103 100 3 103 80 3 103 80 3 103

Capacity 60 3 103 60 3 103 40 3 103 40 3 103 20 3 103 20 3 103 0 300 600 900 1,200 1,500 012345 Number of taxa Major connections per taxon

Figure 7.8 Full development capacity of different types of hypothetical food-webs as a function (a) of the number of taxa and (b) of the average number of major connections (flows >5% of total inputs) per taxon. Hypothetical webs were either unstructured in design with uniformly or lognormally distributed transfer coefficients, or structured in design with uniformly or lognormally distributed transfer coefficients. Empirical webs are denoted by (see also Plate 5).

(a) 5 (b) 5

4 4

3 3

AMI 2 2

1 1

0 0 300 600 900 1,200 1,500 0123456 Number of taxa Major connections per taxon

Figure 7.9 AMI content of different types of hypothetical food webs as (a) a function of the number of taxa and (b) of the average number of major connections (>5% of input flows) per taxon. Hypothetical webs were either unstructured in design with uniformly or lognormally distributed transfer coefficients, or structured in design with uniformly or lognormally distributed transfer coefficients. Empirical webs are denoted by (see also Plate 6).

Average mutual information number of major connections per taxon, but the unstructured uniform webs had the lowest The AMI of empirical and hypothetical webs average number of major connections per taxon were consistent among webs of comparable size (Figure 7.9; Plate 6, Table 7.1). Among structured (Figure 7.9(a)). Among the hypothetical webs, the webs, the lognormal variety tended to have highest AMI was found among large, unstruc- higher AMI than the uniform variety (Figure 7.9) tured lognormal webs, while the lowest AMI and, while the total connectivity was equiva- was found among unstructured, uniform webs. lent (Table 7.1), the lognormal webs had the Unstructured webs had higher total number of greater number of major connections per taxon connections than structured webs, and unstruc- (Figure 7.9). tured lognormal webs had the greatest average INFORMATION THEORETIC METRICS 83

For a food web of size m with nm uniform AMI is equivalent to SI when flows are uniform flows of size TST/nm, the joint probability of flow and highly predictable. SI and AMI were posi- will be p(Tij) ¼ (TST/nm)/TST ¼ 1/(nm); the row tively correlated in both empirical and hypothe- and column sums of the joint probabilities, p(Ti) tical webs (Figure 7.10; Plate 7). However, the and p(Tj) respectively, are each equivalent to flows in the empirical and hypothetical webs were n/nm. Making these substitutions into equation 7.3 neither uniform nor highly predictable, and the and solving gives the maximum theoretical AMI was less than the SI (Figure 7.10). For a given

AMI ¼ log2(m/n). Thus, AMI will tend to rise flow diversity, the range of possible AMIs with network size as our results demonstrate increased as flow diversity increased. Thus, as (Figure 7.9(a)). A large food web of m ¼ 1,500 taxa flow diversity increases, the potential AMI also would have a maximum AMI of about 10.5 bits increases. As expected, the hypothetical webs when there is n ¼ 1 uniform connection per taxon. constructed of uniformly distributed transfer AMI will decline as the connectedness of the food coefficients had lower AMI : SI than their log- web increases. For example, for a web of size m normal counterparts (Figure 7.10) due to greater taxa and n ¼ m connections of uniform flow per uniformity ( ¼ greater unpredictability of direc- taxon, the AMI reduces to AMI ¼ log2(m/m) ¼ 0. tion) of flow among the former web types. Thus, AMI rises with network size and declines Empirical webs clustered at the low end of the as the uniform number of connections (total and SI and AMI distributions, but were otherwise per taxon) rises (as the direction of flow becomes consistent with the results of the hypothetical less certain). networks. The relationship between AMI and SI is com- plex. For networks of n uniform flows per taxon, AMI ¼ log2(nm), and SI ¼log2(1/nm). SI may be Ascendency rewritten as SI ¼ [log2(nm) log2(1)] ¼ AMI. Thus, Interestingly, and importantly, structured webs with realistic trophic structure had greater As/Cd than unstructured webs, within each class of trans- 5 fer coefficient distribution (Table 7.1, Figure 7.11; Plate 8). Thus, the imposition of a realistic trophic structure, which constrained the flow dis- 4 tribution, raised the As/Cd. Average As/Cd of empirical webs was considerably greater than the 3 average As/Cd of the hypothetical webs (Table 7.1), but at the low end of the size range, empirical

AMI webs and hypothetical webs had similar As/Cd 2 (Figure 7.11). Thus, the small size of both empirical and hypothetical webs appears to constrain the 1 flow distribution and raise As/Cd. The As/Cd of hypothetical webs declined rapidly with increasing web size toward a relatively stable 0 0 5 10 15 20 average value (depending on web design), with SI (bits) considerable variability around the mean (Table 7.1, Figure 7.11). For a given size of web, the As/Cd Figure 7.10 AMI as a function of flow diversity (SI) in empirical ( ) of lognormal webs was greater than that of uni- and hypothetical webs. Hypothetical webs were either unstructured in design with uniformly or lognormally distributed transfer form webs (Figure 7.11), which is a consequence coefficients, or structured in design with uniformly or lognormally of the greater inequality of flow among the distributed transfer coefficients (see also Plate 7). lognormal webs. 84 AQUATIC FOOD WEBS

(a) 0.8 (b) 0.8

0.6 0.6

0.4 0.4 As/Cd

0.2 0.2

0.0 0.0 300 600 900 1,200 1,500 0123456 Number of taxa Major connections per taxon

Figure 7.11 Ascendency:capacity as a function of (a) web size and (b) of the average number of major connections (flows >5% of inputs) per taxon (right). Hypothetical webs were either unstructured in design with uniformly or lognormally distributed transfer coefficients, or structured in design with uniformly or lognormally distributed transfer coefficients. Empirical webs are denoted by (see also Plate 8).

Discussion had lower SI than unstructured uniform webs (Figure 7.7, Table 7.1). The rules of organization of the random webs had Average mutual information is a measure of the significant effects on the various information average constraint or probability of flow from one indices. These rules of organization defined both taxon to the next (Ulanowicz 1997). It is not easily the size of the networks (i.e. number of taxa and decomposed like flow diversity (SI) and has a TST) and function (i.e. number and distribution of behavior that is far more complex. AMI has a connections). Some of the responses of these minimum value of zero when a network is maxim- information metrics to the rules were quite pre- ally connected with uniform flows, or when there dictable, at least qualitatively, while others were is an equal probability of flow moving from one not. It was shown that that flow diversity (SI) taxon to any other taxon and the in the increased with network size (Figure 7.7), and with direction of flow is greatest. Conversely, AMI is uniformity of flow and connectedness. Average maximized when a web is minimally connected mutual information also increased with network and flows are uniform. In other words, there is no size (Figure 7.9), but the distribution of flows had uncertainty about direction of movement. greater impact on AMI than did network size. The Zorach and Ulanowicz (2003) provided an SI had an upper bound depending on the size of extension of AMI by relating it to the number of the network. For example, if flows are uniformly roles in a network, where a role is a unique flow distributed, then a fully connected web of size m pattern within a food web that is common to one with m2 m connections, excluding losses and or more species. The number of roles is a measure inputs, will have joint flow probabilities that are of the diversity of functions that are occurring determined by the species diversity m. The flows within a network. The empirical food webs ana- T are given by TST/(m2 m), and the joint i,j lyzed here had about 2–5 roles per food web, and probabilities (from equation 7.4) by 1/(m2 m). occupied a graphical area termed the window of From equation 7.6, the flow diversity would be vitality (Zorich and Ulanowicz 2003), similar to the equal to log [1/(m2 m)], which gives an SI 2 clustering we observed between AMI and the for a large web (m ¼ 1,500) of about 21 bits, and density of major connections (Figure 7.9(B)). This is this is consistent with the results for the unstruc- an idea similar to the network motifs or recurring tured webs with uniformly distributed transfer circuit elements described by Milo et al. (2004), coefficients (Figure 7.7). Unevenness in flow and seems to be a powerful way of characterizing reduces SI, and unstructured lognormal webs INFORMATION THEORETIC METRICS 85 and classifying webs. However, the application of We can make some generalizations about the these metrics to real webs suffers from the same behavior of As/Cd by making the simplifying problems of aggregation as the information theo- assumption that there are nm uniform flows in retic metrics. a web of size m. With this assumption, As/Cd

Interestingly, unstructured uniform and struc- reduces to log2(m/n)/log(1/nm). For one uni- tured uniform networks gave the lowest range of form connection per taxon (n ¼ 1), the quotient AMI (Table 7.1) and, therefore, the lowest number As/Cd ¼ 1 and is independent of size m. A net- of roles. Conversely, the lognormal networks gave work of n ¼ 1 is fully specialized. At the opposite the greatest range of AMI and highest number of extreme (n ¼ m), the quotient As/Cd ¼ 0, and the roles. Intuitively, a greater number of roles is network is maximally disorganized. Among hypo- consistent with the greater variation in flow thetical and empirical webs, the average As/Cd strength afforded by the lognormal distribution, were less than 0.5 (Table 7.1), which affirms a large and on a mechanistic level this is most likely a diversity of flows. Like the hypothetical webs, data consequence of the differences in the distributions on interaction strengths in natural food webs of transfer coefficients. However, the actual flow indicate that interaction strengths are character- distributions expressed as a percent of total inflow ized by many weak interactions and a few strong hardly differed among the different web types interactions, which is thought to increase stability (Figure 7.5). by dampening oscillations between consumers and Ascendency (As) is given by the product of AMI resources (McCann et al. 1998). times TST and is a measure of the absolute order in Do empirical food webs reflect real webs? This a system times the power generated, also descri- has been a major source of concern in ecology for bed as the performance of the system (Ulanowicz the past quarter century (Pimm 1982; Paine 1988; 1986, 1997). The quotient As/Cd is a measure of Martinez 1991; Cohen et al. 1993; others). We con- the degree of organization of the system. We had tend that networks of fully articulated food webs hypothesized that As/Cd would be higher for fall within the of our random networks. structured webs than for unstructured webs and The structured webs (both uniform and lognormal) this was confirmed by our hypothetical data. are a subset that should more closely contain organ- Recall that Cd (full capacity) is a measure of ized, real food webs. Of the metrics considered, the total potential order or complexity, and As the highly aggregated empirical food webs have is the realized complexity. All web types had a characteristics that fall within at least the smaller moderate range of As/Cd for webs of constant examples of the reference universe of structured size, and for large webs the As/Cd appeared to be webs. Reasonable similarities occur for total system insensitive to web size, which is consistent with the throughput (normalized for inputs), number of calculations made above. Thus, As/Cd was most major connections per taxon, SI, and Cd. However, sensitive to flow distribution and insensitive to AMI and As/Cd tend to be higher for empirical size, except among the smallest size classes where webs than their hypothetical counterparts of similar size apparently constrains the flow distributions size. This may be a matter of resolution of flows. and raises As/Cd. This raises the unsettling possi- Major flows (>5%) are dominant in empirical webs bility that the aggregation of species that is chara- because (1) the number of interaction possibilities is cteristic of the relatively small empirical webs low and (2) the ability to identify rare diet items is artificially raises As/Cd. This may be associated low (Cohen et al. 1993). Major flows in these webs with the general inability to define minor flows in may constitute all flow, whereas major flows in the highly aggregated, small food webs. Even larger hypothetical webs do not necessarily constitute all empirical webs lack resolution of flows of less than flows. These minor flows affect the AMI and thus a few percent, and no current empirical web has As and As/Cd. Thus, the true significance of these species-level resolution for smaller organisms metrics may not be realized within our current (Christian and Luczkovich 1999). means of characterizing food webs. CHAPTER 8 Size-based analyses of aquatic food webs

Simon Jennings

Introduction class (Kerr and Dickie 2001) although this is necessarily a simplification of a more complex Size-based analyses of aquatic food webs, where reality (Pimm 1991; Hall and Raffaelli 1993). body size rather than species identity is the prin- The systematic study of abundance–body-mass ciple descriptor of an individual’s role in the food relationships (size spectra) in aquatic communities web, provide insights into food-web structure and began when Sheldon and colleagues began to look function that complement and extend those from at the size distributions of phytoplankton in the species-based analyses. A strength of the focus on ocean using a modified Coulter Counter (Sheldon body size is that body size underpins predator– et al. 1972). They observed the remarkable reg- prey interactions and dictates how the biological ularity and surprisingly shallow gradient of the properties of individuals change with size. Thus relationship between abundance and body size. size-based food-web analyses provide an approach Subsequently, this approach was extended over for integrating community and ecosystem ecology the entire size spectrum from plankton to , with energetic and metabolic theory. and the links between size-based analyses of food- In aquatic ecosystems, the principle primary web structure and size-based analyses of the bio- producers are small unicellular algae, and these logical properties of organisms were soon appar- strongly support size-structured food chains ent (Kerr 1974). This work was a significant step in where most predators are larger than their prey understanding food-web processes, since previous (Sheldon et al. 1972). Individuals of most species studies of abundance–body-mass relationships begin life as larvae feeding at the base of food in both terrestrial and aquatic ecosystems had chains, but can end life as large terminal predators. focused primarily on taxonomically or trophically Lifetime weight increases of five orders of defined subsets of the entire food web. magnitude and more are typical of many species As more size spectra were compiled, it became (Cushing 1975), and a fast growing species may increasingly clear that their similar slopes begin life as a prey item for other species only to (Figure 8.1) reflected similar processes of biological become the main predator on the same group of organization. These processes appeared to be species within one year (Boyle and Bolettzky 1996). common to many aquatic ecosystems, even though The significance of size-based predation and the the physical and biological characteristics of the large scope for growth in aquatic animals means ecosystems, such as primary production, higher that body size is often a better indicator of trophic level production, mean temperature, seasonality, position than species identity. Thus there are com- and depth, were very different (Boudreau and pelling reasons to adopt size rather than species- Dickie 1992). Clearly, an understanding of the based analysis of aquatic food webs. In such processes that led to characteristic size spectra analyses, a small individual of a large species is would help to identify general laws that govern treated as functionally equivalent to a large indi- biological organization. The applied potential of vidual of a small species in the same body-mass

86 SIZE-BASED ANALYSIS OF AQUATIC FOOD WEBS 87

2 both subsets and the whole food web (e.g. Brown and Gillooly 2003). ) –2

1 The significance of body size k Cal m

10 Since body size determines both the biological properties of individuals and predator–prey 0 interactions, size-based food-web analysis has led to the integration of community and ecosystem

Biomass (log ecology with energetic and metabolic theory.

–1 –10 –8 –6 –4 –2 0 2 4 Body size and biological properties Body mass (log10 k Cal) Individual body mass in aquatic communities Figure 8.1 Relationships between abundance and body mass spans 20 orders of magnitude, from bacteria sup- (size spectra) for seven aquatic ecosystems. Slopes are porting the microbial loop to whales that filter comparable but differences in intercept reflect differences in several tonnes of each day. Remarkably, all primary production, and hence production at body mass, in each ecosystem. Redrawn with modifications from Boudreau and these organisms obey consistent scaling laws, Dickie (1992). which dictate how biological features change with size (Brown and West 2000). Species with smaller size-based food-web analysis was also quickly adult body size have higher metabolic rates (Peters recognized and, in addition to analyses of the 1983), higher intrinsic rates of increase (Denney impacts of human exploitation on food webs et al. 2002; Fenchel 1974), faster growth (Brey 1990, (Dickie 1976), Sheldon and Kerr (1972) boldly used 1999), greater annual reproductive output these methods to estimate the biomass of monsters (Gunderson and Dygert 1988; Charnov 1993), in Loch Ness! The methods developed were classic higher natural mortality, and shorter lifespan examples of the macroecological approach (Brown (Beverton and Holt 1959; Pauly 1980). Conversely, 1995; Gaston and Blackburn 2000). larger adult body size is correlated with lower While the study of abundance–body-mass rela- metabolic rate, lower intrinsic rates of increase, tionships has contributed to aquatic and terrestrial slower growth, lower annual reproductive output, ecology, studies in the two environments have lower natural mortality, and greater longevity taken rather different paths. In aquatic ecosystems, (Peters 1983; Charnov 1993). most research on abundance–body-mass relation- Scaling relationships between biological prop- has been closely tied to energetic analysis erties and body size can be used to parameterize of ecosystem function, with an emphasis on the abundance–body-mass relationships, such that size structure, biomass, and production of all they provide estimates of community metabolism, individuals in the food web (Dickie et al. 1987). production, turnover time, and other ecosystem Conversely, in terrestrial ecosystems, there has properties (Kerr 1974; Schwinghamer et al. 1986; been much more focus on abundance and body Boudreau and Dickie 1989; Jennings and Blanchard mass in taxonomically defined subsets of the 2004). whole food web, which either derive energy from the same source (e.g. plants) or do not (e.g. birds). Body size and trophic level In the latter case, the links to patterns of energy flow in the ecosystem are rarely defined (Gaston Aquatic food chains are strongly size-based and and Blackburn 2000). Only recently have there most predators are larger than their prey (Sheldon been clear attempts to integrate aquatic and et al. 1972). While species with a similar maximum terrestrial approaches, and to consider the links body size can evolve to feed at different trophic between abundance–body-mass relationships in levels, the whole food web is characterized by 88 AQUATIC FOOD WEBS a near-continuous rise in mean trophic level with most usefully made at an identifiable and compar- body size (Fry and Quinones 1994; Jennings et al. able stage in the life history (e.g. size at maturity), 2002a). Thus small species that feed at high trophic or, if detailed life history data are lacking, at a levels are never abundant relative to small species fixed proportion of maximum body size. The latter feeding at lower trophic levels. The trophic approach is consistent with the observation that continua across body-size classes show that key life history transitions occur at a relatively fixed (integer) trophic levels do not appropriately constant proportion of maximum body size (e.g. describe the structure of aquatic food webs (France Charnov 1993). et al. 1998), and I treat trophic level as a continuous The relative contributions of intra and inter- measure of trophic position that can refer to any specific differences in trophic level to the trophic point on the continuum. structure of a marine food web were described by In aquatic ecosystems, the commonly adopted Jennings et al. (2001, 2002a). They used nitrogen procedure of assigning a trophic level to a species stable isotope analysis to estimate trophic level is misleading without qualification. While phyto- (Box 8.1) and demonstrated that the interspecific plankton and some herbivores do remain at similar relationships between maximum body mass, trophic levels throughout their life history, the and trophic level at a fixed proportion of that trophic level of most aquatic species increases with body mass, were weak or nonsignificant (e.g. body size. So, comparisons among species are Figure 8.2(a)). The results were confirmed with

Box 8.1 Nitrogen stable isotope analysis

An impediment to the description of links between trophic level of individual or size class i can be body size and trophic structure was created by the estimated as unreliability or unsuitability of methods used to estimate 15 15 ðd Ni d NrefÞ trophic level. Conventional diet analysis did not always TLi ¼ þ TLb, help since species and individuals in the size spectrum k 15 15 switched diet frequently, digested prey at different rates, where d Ni is the mean d N of individual or size class and contained unidentifiable gut contents (Polunin and i and TLb is the trophic level assigned to the reference Pinnegar 2002). Diet analysis was also very labor material (e.g. 1 ¼ phytoplankton, 2 ¼ bivalves or intensive when applied to the range of species and size zooplankton that feed solely on phytoplankton) classes in the size spectrum and estimates of trophic Even when the d15N of a reference material is not level are needed for prey items, which necessitates study of known, estimates of fractionation can usefully be used to prey diet. An appealing alternative to diet analysis was assign relative trophic levels to size classes in the spectrum nitrogen stable isotope analysis. This provides estimates of and to estimate the predator–prey mass ratio (PPMR). trophic level, because the abundance of d15N in the While nitrogen stable isotope analysis does overcome tissues of consumers is typically enriched by 3.4‰ many of the disadvantages associated with diet analysis, relative to their prey. The abundance of d15N the method is also subject to a range of biases that must reflects the composition of assimilated diet and be considered in interpretation. There is considerable integrates differences in assimilated diet over time variation in fractionation around the reported mean of (Post 2002b). 3.4‰ and fractionation is influenced by many factors To estimate the mean trophic level of animals including rates of assimilation and excretion (Olive et al. in a size class, the tissue of all animals in the size 2003). While 3.4‰ is likely to be broadly robust when class can be homogenized and the homogenate prepared applied to analyses of multiple trophic pathways and many for analysis (Jennings et al. 2001). The analysis species (Post 2002b), as in the assessment of trophic level would give an abundance-weighted d15N for in the size spectrum, some variation is expected. In the the animals that were present. With knowledge absence of procedures to correct fractionation estimates, of d15N fractionation (k), and the d15N of reference it is appropriate to run sensitivity analyses to assess the 15 material at the base of the food chain (d Nref), potential effects of errors in the assumed fractionation. SIZE-BASED ANALYSIS OF AQUATIC FOOD WEBS 89

20 16

18 15

) 16 14 N (‰ 15 14 ) d 13 N (‰

12 15 d 12 10 4 6 8 10 12 14 16 log2 maximum total mass 11

16 10 –4–3–2–10123456 log2 body mass 14 ) Figure 8.3 The relationship between body mass and relative trophic 15 N (‰ level (expressed as d N) for the central North Sea fish and inver- 15 d 12 tebrate community (broken line) and relationships between body mass and relative trophic level for the 10 most abundant individual species in this community (solid lines). From Jennings et al. (2002a). 10 0 5 10 15 log2 body mass spectrum, but ‘‘within’’ the size window con- sidered, most of the animals contributing to the Figure 8.2 (a) Relationship between relative trophic level (expressed spectrum were included. 15 as d N) and maximum body mass of North Sea fishes and (b) the One potential weakness with size-based analyses relationship between trophic level and body mass for the whole fish community by body-mass class. From Jennings et al. (2001). is that the observed linear relationship between body size and trophic level will break down at higher size classes. Thus the very largest animals in aquatic ecosystems are often filter feeding a phylogenetically based comparative analysis sharks and whales, which typically ‘‘feed down (Harvey and Pagel 1991). Conversely, when all the food chain’’ on smaller and more productive animals in the communities were divided into size classes of prey. This is probably because body-mass classes without accounting for species the mobility of these large animals allows them to identity, trophic level rose almost continuously track waves of zooplankton production, providing with body mass (e.g. Figure 8.2(b)). a near continuous food supply, and because feed- Additional analyses of intra and interspecific ing down the food chain will give access to much relationships between body mass and trophic level higher levels of prey production. in fishes and invertebrates have shown that increases in trophic level across the size spectrum Predator–prey body-size ratios and were predominantly a consequence of intraspecific transfer efficiency increases in trophic level with body mass rather than larger species (species with greater maximum The structure of size-based food webs is deter- body mass) feeding at higher trophic levels. Thus mined by predator–prey interactions and the the increase in trophic level of individual species efficiency with which energy is passed from prey with body size makes an important contribution to to predators. Thus the predator-prey mass ratio the increase in the trophic level of the community (PPMR) and trophic transfer efficiency (TE) are with size (Figure 8.3). Analyses of this type have fundamental attributes of size-based food webs yet to be done for complete food webs rather and measurements of PPMR and TE are needed to than a defined body size ‘‘window’’ in the size support size-based analysis (Gaedke 1993). 90 AQUATIC FOOD WEBS

Diet analyses for species that dominate commun- 0.13 for energy transfer from phytoplankton ity biomass and production have provided estim- to zooplankton or benthic animals and 0.10 for ates of PPMR for some species and species groups zooplankton or benthic animals to fish. In fresh- in some ecosystems (Ursin 1973; et al. 1991) but water food webs, TE is typically 0.15 from primary they do not provide mean estimates of PPMR for all to secondary consumers (Blazka et al. 1980). animals in a size ‘‘window’’ in the spectrum. Mean estimates of PPMR can be made using size-based The dynamic size spectrum nitrogen stable isotope analysis, provided that individuals can be sampled in proportion to their For analytical purposes it is convenient to view abundance within size classes of all species. size spectra as stable; most field sampling pro- Estimating the abundance of all individuals in grams are designed to account for seasonal vari- food webs is a significant challenge. While meth- ability in biomass and production, and to estimate odologies for quantifying phytoplankton and annual means. Seasonal variations in the size zooplankton abundance are generally well tested spectrum are particularly pronounced at high and established, many fishes and larger inverte- latitudes and in small size classes with fast turn- brates are difficult to sample quantitatively. The over times (turnover time ¼ 1/P : B, and P : B scales 0.25 paucity of vulnerability and catchability estimates with body mass as P : B ¼ 2M where P:B is the needed to estimate their relative and absolute production to biomass ratio and M is body mass; abundance is a major impediment to progressing Ware 2000). In large size classes, turnover time is large-scale food-web studies. Moreover, many slow and there is greater temporal stability in different sampling techniques and high levels of biomass and production. replication are needed. Thus a recent attempt to The dynamics of the size spectrum, although not estimate the abundance and trophic level of all the focus of this chapter, have important ecological animals in part of a marine size spectrum required implications. For example, seasonal production three types of trawl nets plus acoustic, grab, and cycles lead to the propagation of waves of pro- corer surveys (Jennings et al. 2002b). duction up the size spectrum. These waves flatten If sampling difficulties can be overcome, indivi- and broaden as they reach larger individuals with duals can be divided into body-mass categories higher turnover times. Pope et al. (1994) produced 15 a of the propagating wave. To (typically log2) and biomass weighted mean d Nis determined (Box 8.1). Mean d15N can also be survive, animals needed to surf (track) the wave, determined mathematically from d15N versus size reproducing and growing at a rate that allowed relationships for individual species or species them to feed on the wave of abundant food while groups and estimates of their abundance. The mean avoiding a coevolving wave of larger predators. PPMR is calculated from the slope of the relation- 15 The slopes of size spectra between mean d N(y) and log2 body mass (x) (k/slope) (e.g. Figure 8.2(b), where PPMR ¼ 2 )andk is Explorations of the processes governing the slopes 15 the assumed mean fractionation of d N per trophic of size spectra have included detailed process- level, usually 3.4 ‰ (Jennings et al. 2002b). based models of predator–prey interactions and Consistent with Dickie’s (1976) analysis, the more simplistic models based on fundamental decline in abundance with body mass is a function ecological principles. All models are underpinned of the inefficient transfer of energy from prey to by the recognition that the scaling of metabolism predators. This inefficiency is most simply descri- with body size determines the energy require- bed by TE. TE describes the proportion of prey ments of animals in different size classes. production that is converted to predator produc- ¼ tion (TE Pn þ 1/Pn, where Pn þ 1 is predator pro- Predator–prey interactions in the size spectrum duction and Pn is prey production). Ware (2000) reviewed TE in marine food webs. It was typically Following observations of the relative constancy of higher at lower trophic levels, with a mean of slopes of abundance–body-size relationships in SIZE-BASED ANALYSIS OF AQUATIC FOOD WEBS 91 several ecosystems (Sheldon et al. 1972; Boudreau and and Dickie 1992), theoreticians began to explore Bn Fn Cnþ1 : : how these patterns might arise (Kerr 1974; Dickie ¼ (8 8Þ Bn1 Fnþ1 Cn 1976). Dickie’s (1976) theoretical development was Thus the ratio of the biomass at two successive particularly significant and described how the trophic positions is the ecological efficiency ratio of production and biomass at successive corrected by the ratio of the predation rates, and trophic levels, and hence the slope of the size will be independent of the mean body size of spectrum (in which body mass is an index of individuals at trophic levels. Further, since pre- trophic level), might be determined. dation rates will depend on metabolism, biomass Production (P) at trophic position (n) is a func- ratios will reflect the scaling of metabolism (R) tion of consumption (C) and the efficiency of use of with M, which we know to be R / M þ 0.75 (Peters food intake (K), 1983). Dickie’s (1976) analysis also implied that : Pn ¼ KnCn,(81Þ relative predation rates must increase rapidly as and consumption by a predator at trophic position trophic level decreases, and thus utilization of n þ 1is production at low trophic levels will be higher. Subsequent theoretical work focused on the : Cnþ1 ¼ Fnþ1Bn,(82Þ secondary structure of the size spectrum, with an where F is rate of predation by the predator and B emphasis on whether domes that corresponded is prey biomass. Thus with trophic groups were superimposed on the primary scaling that Dickie (1976) had described C B nþ1 ¼ F K n ,(8:3Þ (Thiebaux and Dickie 1992, 1993; Benoit and C nþ1 n P n n Rochet 2004). Empirical evidence for such domes where Cn þ 1/Cn is the ratio of food intake at has been found in some freshwater lakes (Sprules successive trophic positions; the ‘‘ecological effi- and Stockwell 1995), but they may not be a uni- ciency’’ of Slobodkin (1960). versal feature of aquatic size spectra in commun- In a stable and seasonally averaged size spec- ities when animals with a greater range of trum, the rates of production and total mortality morphologies and life histories are present. must balance, so Energy equivalence rule Pn ¼ Fnþ1 þ Un,(8:4Þ Bn While aquatic ecologists were modeling the detailed structure of size spectra (Dickie et al. where U is mortality due to factors other than 1987), terrestrial ecologists were compiling predation. empirical estimates of the slopes of abundance– Dickie (1976) showed that equations (8.3) and body-mass relationships for subsets of food webs (8.4) could be used to predict ratios of production (review: Gaston and Blackburn 2000). For com- and biomass at successive trophic positions, and munities that shared a common energy source, hence the slope of the size spectrum. Thus, such as plants using , numerical abund- Pn KnCn Fn : ance (N) typically scaled with body mass (M) ¼ ¼ Kn ,(85Þ 0.75 Pn1 Kn1Cn1 Fn þ Un1 as M . Since metabolic rate scaled with mass as M0.75 (Peters 1983), the rate of energy use in these and communities was expected to be independent of B P F þ U n ¼ n n n1 : (8:6Þ body size (Damuth 1981), a prediction now known Bn1 Pn1 Fnþ1 þ Un as the energetic equivalence hypothesis (Nee et al. Thus, 1991). Energetic equivalence, for example, has since been shown to predict the scaling of numerical Bn Bn abundance and body mass in phytoplankton and ¼ FnKn (8:7Þ Bn1 Pn plant communities (Belgrano et al. 2002; Li 2002). 92 AQUATIC FOOD WEBS

Extending energy equivalence to One feature of Brown and Gillooly’s (2003) ana- size-based food webs lysis was that each trophic level (phytoplankton, zooplankton, or fish) was assumed to extend over While available energy (E) scales as M0 when a range of body sizes, an approach that was individuals share a common energy source, size- inconsistent with evidence for a trophic continuum. based food webs are characterized by larger Jennings and Mackinson (2003) formalized and predators eating smaller prey. Since the transfer tested the Cyr (2000) and Brown and Gillooly of energy from prey to predators is inefficient, (2003) approach in a size-structured marine food the total energy available to individuals must web where trophic level was shown to increase scale

(a) (b)

PZF B~M10 B~M10.25 E E Figure 8.4 Relationships between biomass

or ~ 0 or E M ~ –0.25 B B E M abundance (B) and rate of energy use (E )as a function of body mass (a) within and log log (b) among trophic levels (here P: phytoplankton, Z: zooplankton, and F: fish). Predicted scalings across trophic levels assume TE ¼ 0.10 and PPMR ¼ 10,000 : 1. Redrawn with modifications log M log M from Brown and Gillooly (2003). SIZE-BASED ANALYSIS OF AQUATIC FOOD WEBS 93

0.1

0.0

–0.1

–0.2

–0.3 0.20 Scaling of B with M –0.4

–0.5 0.15 800 Transfer 800:1 600 600:1 efficiency 400400:1 200200:1 0.10 Figure 8.5 Sensitivity of the slope of the biomass Predator-prey abundance (B)–body-mass (M) relationship (size body mass ratio spectrum) to changes in PPMR and TE. combinations of parameters, slopes of the size First, in species and size composition. Second, spectrum for the whole food web are expected to in maximum food-chain length, because the focus range from 0 to 0.2 (Figure 8.5). on average chain length in idealized communities may obscure the presence of long but rare food Food-chain length chains that support individuals of intermediate body size. Maximum food-chain length is an important food- Jennings and Warr (2003) investigated empirical web property and affects community structure, relationships between maximum food-chain length, ecosystem processes, and contaminant concen- PPMR, primary production, and environmental trations in animal tissue (Post 2002a). Maximum stability in marine food webs with a food-chain length has been correlated with of community assembly. Their analyses provided resource availability, ecosystem size, environ- empirical evidence that smaller mean PPMR was mental stability, and colonization history. Some of characteristic of more stable environments (stability these correlations may result from environmental measured as annual temperature variability) and effects on predator–prey mass ratios. that food chains were longer when mean PPMR Size-based food-web analysis provides a method was small. Their results also demonstrated that the for understanding relationships between predator– heaviest fish predators at each site rarely fed at the prey mass ratios and food-chain length. In a highest trophic level and the longest food chains community of given size and species composition, supported fish predators with intermediate body changes in PPMR must influence food-chain length size. However, both maximum and maximum (Figure 8.6). Thus in the community with mean food-chain lengths were correlated. PPMR ¼ a (Figure 8.6), the trophic level of the The results suggested that environmental factors individual with the maximum body size (Mmax) favoring smaller mean PPMR allowed individuals TLa represents maximum food-chain length and is feeding at high trophic levels to persist. Jennings higher than the trophic level of the same sized and Warr (2003) proposed that relationships individual in the community with PPMR ¼ b (TLb). between environmental variables and food-chain Real communities, however, differ from the length were explained by the effects of environ- idealized communities in Figure 8.6 in two ways. ment on PPMR. If maximum food-chain length can 94 AQUATIC FOOD WEBS

Since abundance–body-mass scaling will partly depend on PPMR and PPMR is smaller in more TL a stable environments, steeper abundance–body- mass relationships may be found in more stable environments if there is not a compensatory rela- TLb tionship between PPMR and TE. These relation- a ships have yet to be investigated.

b Abundance–body-size relationships in subsets of the food web

log body mass Mmax While terrestrial macroecologists have often focused on abundance–body-mass relationships Figure 8.6 Relationships between PPMR and maximum for taxonomically or ecologically defined groups of food-chain length in communities of the same size and animals (Gaston and Blackburn 2000) such species composition. When PPMR is relatively small (a) the research has not been a consistent focus of aquatic maximum food-chain length is long (TLa). When PPMR is relatively large (b), the maximum food-chain length will science. Nevertheless, there are studies of groups fall (TLb). of individuals which share, or do not share, a common energy source. depend on PPMR, then feeding behavior, as Oceanic phytoplankton are an example of a mediated by the interactions between individuals group that share an energy source, since they all and their environment, will have a fundamental use sunlight to photosynthesize. Large-scale and effect on maximum food-chain length. The life long-term research has shown that the slope of history, morphology, and motility of a predator the abundance–body-mass relationship in phyto- places ultimate constraints on the prey it can eat, plankton communities is consistent with that but within these constraints, prey size selection predicted by the energetic equivalence hypothesis, will depend on the environment, competition, prey N / M þ 0.75 (Li 2002). Thus the sunlight energy availability, prey processing costs, prey population used by all phytoplankton cells in a size class is the stability, and prey production (Stephens and Krebs same as in any other size class (Li 2002). Interest- 1986; Cohen et al. 1993). ingly, the universality of energetic equivalence is It was not clear from these analyses whether such that it predicts the abundance–body-mass the significant relationship between the measure relationship when marine phytoplankton and of environmental variability and PPMR demon- terrestrial plants are included in the same analysis strated causality, as annual temperature variation (Belgrano et al. 2002). may be a surrogate for many other types of Benthic infaunal communities do not share the physical and biological variability. For example, same energy source, but they are effectively sus- at deeper sites with lower annual temperature tained by energy reaching the seafloor as phyto- variation, and wave effects are smaller, the plankton or detritus and are often treated as a water is clearer, and there are more complex functional unit. Unlike the complete food web, the seabed habitats, so predators may be able to track size structure of infaunal communities does not and feed on larger, mobile, and patchily dis- primarily reflect size-based predation, since many tributed prey more easily. Conversely, in turbid of the largest animals (e.g. bivalve molluscs, bur- shallow sites subject to high levels of physical rowing ) feed at lower trophic levels disturbance, tracking and catching large active than some of the smaller animals (e.g. predatory prey in the may be more difficult polychaete worms). and more fishes, of all body sizes, may feed on Benthic infaunal communities exhibit remark- smaller but more evenly distributed and more ably consistent abundance–body-mass relationships productive benthic animals on the seafloor. (Schwinghamer 1981; Schwinghamer et al. 1986; SIZE-BASED ANALYSIS OF AQUATIC FOOD WEBS 95

Duplisea 2000) and Dinmore and Jennings (2004) hoteliers. With hindsight, it is clear that Sheldon investigated whether these slopes could be and Kerr (1972) had actually underestimated predicted from the energy available to different monster abundance, since sufficiently monstrous size classes of animals. Energy availability (E)at animals may well have fed down the food chain on size was calculated from the scaling of E with M, smaller and more productive size classes of prey, based on estimated trophic level at size (from or fed opportunistically on scantily clad terrestrial nitrogen stable isotope analysis) and TE. In the energy inputs to the Loch! infaunal community they studied, trophic level Species-size–abundance data underpin popu- (d15N) decreased with increasing M and the lation, community, and ecosystem analyses and slope of the relationship between d15N and M was have been collected during monitoring and 1.07. Thus E / M0.13, B / M0.53, and N / M0.47 research programs in many aquatic environments when TE was 0.125 and fractionation was 3.4‰ over many years. Examples are the plankton, d15N per trophic level. The predicted scalings of B benthos, and fish surveys conducted to assess the and N with M were 0.48 and 0.54, respectively impacts of climate, pollution, and fisheries on and did not differ significantly from the predicted aquatic ecosystems. These data can be used for scalings. retrospective size-based food-web analysis, when This analysis suggested that theory most usefully dietary data for species-based food-web analysis applied to the prediction of abundance–body-mass are not available (Murawski and Idoine 1992). relationships in complete food webs may have a Assessments of the effects of fisheries exploitation role in predicting abundance–body-mass relation- on aquatic food webs have frequently been con- ships in subsets of the food web. The analysis of ducted using size-based approaches (Duplisea and the benthic infaunal community is probably robust Kerr 1995; Rice and Gislason 1996; Gislason and because it is a relatively defined functional unit, Rice 1998). but the approach would be less likely to work Since the slope of the unexploited size spectrum if applied to taxonomically rather than function- can be predicted from PPMR and TE, and since ally defined subsets. This is because taxonom- there is little evidence that either parameter is ically defined subsets typically receive energy affected by exploitation to the same extent as inputs from many sources that can rarely be well biomass (Ware 2000; Jennings and Warr 2003), the quantified. slope of the unexploited size spectrum provides a useful baseline or reference level when assessing the relative impact of human activities on the Practical applications of size-based . Jennings and Blanchard (2004), food-web analysis for example, used estimates of TE and empirical Size-based food-web analyses provide a method measurements of PPMR in a marine ecosystem to for assessing the large-scale direct and indirect predict the slope of the unexploited size spectrum effects of human activities on marine ecosystems and compared this with the slope as predicted (e.g. climate change, fishing). The potential from contemporary data (Figure 8.7). Slopes of applications of size-based food-web analysis to the size spectra were predicted using the appro- management problems were well recognized by ach previously described, and intercepts were the first scientists to work in this field, and a classic predicted from primary production (PP), where paper by Sheldon and Kerr (1972) used knowledge production at any higher trophic level (PTL) of the regularity of structure in the size spectrum declines as to predict the population density of monsters in P ¼ PP TETL1,(8:9Þ Loch Ness. They estimated that 10–20 monsters of TL approximately 1500 kg were present. This ensured where TE is the estimated transfer efficiency. that Loch Ness continued to attract wealthy By comparing the unexploited theoretical and monster hunters and sceptical tourists, with happy observed size spectra, Jennings and Blanchard economic consequences for local landlords and (2004) predicted that the current biomass of large 96 AQUATIC FOOD WEBS

Trophic level 4.0 4.2 4.4 4.6 4.8 5.0 5.2

1 TE = 0.150 )

10 TE = 0.125 0 TE = 0.100

Biomass (log Figure 8.7 Comparison between predicted –1 slopes of unexploited size spectra when TE ¼ 0.100, 0.125, or 0.150 and the slope of the size spectrum in the fished North Sea in 2001. Circles indicate biomass at body-mass for 2001. Trophic levels were –2 assigned to body-mass classes using nitrogen stable 12345 isotope analysis. From Jennings and Body mass (log10) Blanchard (2004).

fishes weighing 4–16 and 16–66 kg, respectively TE among ecosystems appears to be quite limited, was 97.4% and 99.2% lower than in the absence of and this places constraints on the slopes of fisheries exploitation. Given the scaling relation- size spectra, a better understanding of the links ships between body size and biological properties, between PPMR and TE would significantly the size spectrum could also be parameterized improve understanding of links between aquatic to estimate the effects of fishing on turnover food-web structure and function. time. The mean turnover time of the exploited Recent attempts to link size and species-based community was almost twice as fast as that of the analyses of food webs are an important step (Jen- unexploited community (falls from 3.5 to 1.9 years) nings et al. 2002a; Cohen et al. 2003), since Even in data poor systems, the relative con- understanding of the relationships between struc- stancy of TE and PPMR means that there are a ture and diversity in aquatic food webs is still limited number of solutions for the slope of the poor. Key to this process is the development of size spectrum (Figure 8.5). This makes it relatively analytical methods that account for the changing easy to make crude (order of magnitude) estimates trophic role of species with body size while of changes in the relative abundance of small and preserving the significance of species identity. large animals following exploitation. Are there consistent patterns of change in trophic position with body size, for example, and how do these patterns of change differ in rare or com- Conclusions mon species and relate to life histories and the environment? Size-based food-web analyses provide insights into Macroecological approaches in aquatic and ter- food-web structure and function that complement restrial ecology have followed rather different and extend those from species-based approaches. developmental paths, perhaps reflecting the One key strength of size-based analyses is that dominant characteristics of aquatic and terrestrial they provide a transparent approach for integrat- ecosystems and largely separate research funding ing community and ecosystem ecology with ener- mechanisms (Beddington and Basson 1994; Chase getic and metabolic theory. Thus they describe 2000). The renewed focus on the common bio- how the environment affects PPMR and how, in logical properties of individuals and their role in turn, predator–prey body-mass ratios and transfer controlling community and ecosystem structure efficiency might determine abundance–body-mass (Brown and West 2000; Brown and Gillooly 2003) relationships. Although variability in PPMR and should encourage wider examination and testing SIZE-BASED ANALYSIS OF AQUATIC FOOD WEBS 97 of macroecological principles and help to identify example) in an ecosystem context and provides a cross-system generalizations (Belgrano et al. 2002). basis for describing the wider impacts of human Despite the early recognition that size-based activities. food-web analysis could help to address environ- mental management issues (Dickie 1976; Pope et al. Acknowledgments 1988; Boudreau and Dickie 1989; Murawski and Idoine 1992), the application of size-based food- I am very grateful to Andy Belgrano for the invi- web analysis to management issues is still largely tation to write this chapter, to Steve Mackinson for overlooked. This is unfortunate, since the applica- comments on the manuscript and to the UK tion of a size-based approach helps to place the Department of Environment, Food and Rural management of populations (exploited fishes, for Affairs for funding (M0731). CHAPTER 9 Food-web theory in marine ecosystems

Jason S. Link, William T. Stockhausen, and Elizabeth T. Methratta

Introduction morphed into focusing on binary pairs of species rather than the entire food web. Food webs provide the basis for describing com- However, multispecies food webs were still munity and ecosystem structure. Food webs also being considered in those and following decades. provide the framework for integrating population Clarke (1946) developed a food-web for Georges dynamics, community structure, species inter- Bank that encapsulated the thinking that ecosys- actions, community stability, biodiversity, and tems were like machines, with gear-like mechan- ecosystem productivity. Food-web interactions isms that represent how energy flowed through a ultimately determine the fate and flux of every food web. The concurrent trophic-dynamic con- population in an ecosystem, particularly upper cept of Lindeman (1942) similarly developed from trophic levels of fiscal importance (May 1973; food-web observations in a north-temperate lake. Pimm 1982). Additionally, food webs often pro- Together, the seminal work of Lindeman and vide a context for the practical management of Clark provided the foundations for much of the living resources (Crowder et al. 1996; Winemiller subsequent research that explored the energetics and Polis 1996). As such, food webs have high and structure of food webs (e.g. Odum 1957, 1960). heuristic value for ecological theory and have been The importance of species interactions and their the subject of considerable interest in general essential role in regulating the flow of energy in an ecology. ecosystem was clearly established by the end of this era. Earliest food webs—first era

The development of ecological theory concerning Complexity is prominent—second era the patterns exhibited by food webs and the under- lying processes they reflect can loosely be divided In the middle of the twentieth century, the into four temporal eras. Some of the earliest food- emphasis on energy flow of food webs in ecology web studies were marine examples (Petersen 1918 continued (in many ways developing into the field quoted in Pomeroy 2001; Hardy 1924). The work by of ) but the prominent ‘‘food-web Lotka (1925) and Volterra (1926) to understand the theory’’ of that era was largely encapsulated by relative roles of predation and fishing in European Elton’s (1958) and Paine’s (1966) exploration of com- represented some of the first attempts to plexity and stability in ecological communities. characterize food-web interactions. Their work The hypothesis that complexity engendered stabi- explored the importance of species interactions lity in ecological communities enjoyed considerable after the cessation and restart of fisheries brought support and approached the status of a mathe- about by WWI. The emphasis on interactions matical theorem (MacArthur 1955; May 1973) prior became an integral part of ecology, and ultimately to the theoretical attacks of May (1972, 1973)

98 FOOD-WEB THEORY IN MARINE ECOSYSTEMS 99 and others. As the field of food-web ecology pro- structure, as it implies that connectance declines gressed, the study of large marine systems was hyperbolically with increased species richness. de-emphasized. The conclusion from this era was Cohen (1989) also presented a model of commun- that complexity was an important, if not the ity organization, the ‘‘cascade’’ model, that gave central, factor for food webs. ‘‘remarkable quantitative agreement’’ between his empirical laws and the model’s predictions based on a single parameter—the expected linkage Complexity and stability—third era density. Subsequent to this era, May (1972, 1973), built During this era and extending into the present upon results by Gardner and (ca. the past 25 years), in many respects analysis of Ashby (1970) and developed a criterion for com- topological food-web catalogs has dominated the munity stability related to connectance, whereby a search for pattern and process in food webs. Few system is stable if i(SC)1/2 < 1 (where i is mean of the food webs studied were marine ecosystems, interaction strength, S is the number of species, and those few examples were estuarine, coastal, and C is the connectance of a food web, principally or intertidal and not representative of the vast derived from S and the number of links between majority of the world’s oceans. It is interesting species or species interactions, L), based on ana- that marine food webs have been under- lysis of randomly constructed webs. Under the represented in these collections even though the assumption that mean interaction strength i is first food-web diagrams were constructed for independent of food-web size, May’s formula marine systems (Petersen 1918 quoted in Pomeroy predicts that connectance should decline hyper- 2001; Hardy 1924). bolically as the number of species increases, with intriguing consequences for stability. Networks and current synthesis—fourth era This prediction ushered in the next era of food- web theory and led to numerous comparisons of At the beginning of the fourth (and current) era in empirically based, topological food-web catalogs the development of food-web theory, the edifice of (Briand and Cohen 1984, 1987; Cohen and Briand empirical evidence supporting prior generalizations 1984; Cohen 1989; Schoener 1989; Sugihara et al. based on topological food-web catalogs was ser- 1989; Cohen et al. 1990; Pimm et al. 1991; Havens iously challenged on a number of grounds. Chief 1992). Additional topological web metrics (e.g. among these criticisms was that the scaling— Pimm 1982; Cohen et al. 1990; Bersier et al. 2002) spatial, temporal, or taxonomic—used to construct were introduced in these and associated studies the catalogued food webs, was too limited in scope to identify other food-web patterns. By the end of to adequately capture all the S and L of a given the 1980s, food-web theorists had developed a set food web (Polis 1991; Hall and Raffaelli 1991, 1993; of empirical relations based on these food-web Cohen et al. 1993; Martinez 1993; Goldwasser and catalogs, some of which included >100 webs. Roughgarden 1997; Solow and Beet 1998, Martinez Cohen (1989) summarized several of these et al. 1999). Other related concerns were that the empirical relations as five ‘‘laws’’: (1) excluding criteria for aggregation were inconsistent, ontolo- cannibalism, cycles are rare, (2) food chains are gical changes in diet were absent, cannibalism was short, (3) the proportions of top, intermediate, and ignored, and inconsistencies in sampling effort, basal species (%T, %I, %B) are independent of spatio-temporal resolution and spatio-temporal food-web scale (the ‘‘species scaling law’’), (4) the aggregation giving rise to sampling artifacts could proportions of link types (%T-I, %T-B, %I-I, %I-B) not be assessed (Paine 1988; Winemiller 1990; are independent of food-web scale (the ‘‘link Hall and Raffaelli 1991; Kenny and Loehle 1991; scaling law’’), and (5) linkage density (LD ¼ L/S)is Martinez 1991; Polis 1991). independent of food-web scale (the ‘‘link-species To be fair, in part these criticisms reflected that scaling law’’). The last of these laws supports few, if any, of the collected food webs had ori- May’s stability criterion as a constraint on trophic ginally been developed with the intent of testing 100 AQUATIC FOOD WEBS food-web theory (Warren 1994). Additionally, and structure needs to be further resolved, despite more recent and exhaustive food webs disagreed enhanced caveats to account for increased food with many of the previous theoretical predictions. web complexity, and (2) the vast majority of For example, there is much more omnivory, can- food web and network analyses neglect marine nibalism, and cycling than was previously expected ecosystems. (Polis and Strong 1996; McCann et al. 1998; Closs et al. 1999). The new webs also exhibited higher The marine environment species diversity and topological complexity than typical in older webs (e.g. Winemiller 1990; Hall Our premise is that all types of ecosystems are and Raffaelli 1991; Martinez 1991; Polis 1991; unique, but food-web theory has been developed Goldwasser and Roughgarden 1993; Polis and (certainly tested further) principally from fresh- Strong 1996; Reagan et al. 1996). The average water and terrestrial food webs, largely omitting number of links per species (i.e. linkage density, marine food webs. Additionally, not all marine

LD) and food-chain lengths were greater in newer food webs are equal, with differences in the food webs. It was also shown that many food-web expected number of species, dominant processes, properties were sensitive to the criteria used to physical forcing factors, stability, production, aggregate species, as well as the level of aggrega- structure, interaction strength, etc., observed tion, thus casting previous scale invariance laws among the various types of marine ecosystems (see (species, link, and link-species scaling laws) into the section below for a categorization of marine question (Winemiller 1990; Martinez 1991). ecosystems). Steele (1985) has noted some of the Martinez (1991, 1992) also challenged the obvious differences between marine and other eco- hyperbolic scaling relation for connectance (Pimm systems. Here we expand upon those considera- 1982; Cohen and Newman 1988) and hypothesized tions as they relate particularly to food webs. ‘‘constant connectance,’’ whereby directed con- nectance is scale invariant. The empirical and Scale theoretical framework outlined by Cohen (1989) to extend May’s (1972, 1973) work underwent rigor- As difficult as it is to construct and study fresh- ous critique in the past decade, severely ques- water and terrestrial food webs, it is much worse tioning the determinants of food-web stability. in marine ecosystems. Observations and data are During the past 10 years there has also been a much more costly to obtain in marine systems. The call to move beyond just ‘‘S and L’’ and topological typical sampling endeavor in marine ecosystems is food webs, examining explicitly the flow of energy akin to trying to estimate population parameters of and rates of species interactions in food webs small mammals in a field or forest via towing a (bioenergetic and biodemographic, respectively butterfly net from a low-flying aircraft. Suffice it to Winemiller and Polis 1996). In particular, the say that sampling marine ecosystems is difficult, importance of interaction strength relative to food- and as a direct result S and L are going to be sorely web stability is still being explored (Haydon 1994; under sampled. Food-web theories which rely on de Ruiter et al. 1995; Raffaelli and Hall 1996; those parameters are consequently difficult to test McCann et al. 1998; Closs et al. 1999). Much of the in marine food webs. This parallels the realization shift in effort for food-web studies is reflected in reached >10 years ago for freshwater and terres- the recent emphasis on network structure and trial food webs but is exacerbated by the challenge theory (Dunne et al. 2002a,b, 2004; Krause et al. of sampling in the marine environment. 2003). In some respects this shift in effort has pro- The spatio-temporal scales at which marine vided extensions or advances to prior ‘‘cascade,’’ ecosystems function are multiple and large (Steele ‘‘niche,’’ and scaling models (e.g. Williams and 1985; Mann and Lazier 1991). The scales over Martinez 2000; Dunne et al. 2004). Yet two points which biological interactions occur are probably remain despite current efforts; (1) much of the much larger for marine ecosystems compared to fundamental debate about complexity, stability, nonmarine systems. Often individuals of the same FOOD-WEB THEORY IN MARINE ECOSYSTEMS 101 population can be found dispersed over several The ontogenetic shifts in the size (either biomass thousand square kilometer. Many species have or length) of marine organisms routinely span 3 to 4 daily movements on the scale of hundreds of orders of magnitude, and in some instances span kilometer. Currents, fronts, and similar physical 5 to 6 orders of magnitude. This may be rivaled phenomena also occur at broad spatial scales and only by a few large tree and species in can transport organisms at greater rates and terrestrial ecosystems. Changes in body size over magnitudes than in other systems (Carr et al. the course of an organism’s life history typically 2003). Often species migrate or aggregate season- correspond to changes in diet and functional ally for significant biological events (i.e. spawning, trophic group for many marine species, a process foraging, etc.). Local and regional scale phenomena referred to as ‘‘metaphoetesis’’ (Cohen et al. 1993). are also amalgamated with broader oceanographic The result is that one species may actually function regimes, such as warming or a shift in the PDO or like three or four different species across its life NAO, with the result that species distributions, history. Thus, ontogenetic shifts in diet may have ranges, and productivity can shift. Cousins (1996) a significant influence on food-web topology and suggested that food webs be defined by ambits of energy flows. top predators or the dominant processes that affect The largest and most long lived species in the them. Applied to marine systems this principle ocean are fundamentally different than those in yields food webs with enormous spatial extents. other ecosystems (Steele 1991). These species The sum result of all these scaling considerations which correspond to the apex and top trophic is that there are likely to be a lot of species inter- levels in the ocean are also the most heavily actions (L) that go undetected. Additionally, it is exploited species (Pauly et al. 1998; Jackson et al. likely that L, and hence linkage density (LD) and 2001). The opposite is true in terrestrial ecosys- connectivity (C), are probably much higher than tems, where autotrophs tend to be the larger, typically reported for marine systems and food longer-lived organisms and the most widely har- webs in general, and may even be much higher vested for human use (Steele 1991). The resolution than predicted by food-web theory. of existing food webs reflects this disparity, parti- cularly for lower trophic levels. Basal trophic levels are highly unresolved in most marine food webs, Marine species with finer resolution often impractical for webs Differences in foraging behavior and the higher describing metazoan interactions. As a result, degree of ominivory of marine versus terrestrial producers are often represented oversimply, if at species have been hypothesized to explain all, in these food webs. the higher LD observed for marine food webs The taxonomic groups represented in the ocean (Cohen 1994). For instance, insects are often are numerous. The way in which species are dominant terrestrial herbivores and specialize on treated has also been found to affect the amount of one or a few tree species in an entire lifetime. This information derived from a food web, indicating strategy reduces the overall number of links for that the number of compartments (i.e. trophic that species and its food web. A single marine species; the largest set of organisms with identical copepod, by contrast, is likely to encounter mul- sets of predators and prey defines a group known tiple species of phytoplankton, microzooplankton, as a trophic species; see Cohen and Briand 1984; and micronekton over the course of its life history Cohen 1989) may influence both structural and and drive the LD much higher for marine food functional aspects of the food web (Schoener 1989; webs (Cohen 1994). Similarly, Link (2002) has Polis 1991; Hall and Raffaelli 1991, 1993; Martinez noted that most marine fish species are oppor- 1993; Cohen 1994; Solow and Beet 1998; Martinez tunistic generalists, with few trophic specialists. et al. 1999; Abarca-Arenas and Ulanowicz 2002).

Again the implication is that L, LD, and C will be Lumping and splitting of taxonomic and func- greater in these highly but loosely connected tional groups and the criteria by which those food webs. groups are linked have important ramifications 102 AQUATIC FOOD WEBS for food-web structure, metrics, and theory. In interactions are typically left out of metazoan food particular, reduced food-web resolution is asso- webs across ecosystems. Some evidence suggests ciated with lower S, L, LD,C, and mean chain that the aggregation of microbial groups in food lengths (Schoener 1989; Polis 1991; Hall and webs may reduce the informational value Raffaelli 1991, 1993; Martinez 1993; Cohen 1994; provided by food webs and that of Solow and Beet 1998; Martinez et al. 1999; Abarca- highly connected species at basal trophic levels Arenas and Ulanowicz 2002). Aggregation of may be even more destabilizing to an ecosystem marine species is common, but in some cases than the loss of top, highly connected species different life history stages are separated into (Abarca-Arenas and Ulanowicz 2002). Further different trophic species to better resolve a food studies examining the linkages between microbes web. (e.g. Christensen and Pauly 1992; Gomes and metazoans will provide a more comprehens- 1993). In some cases, the criteria used to assign a ive understanding of food-web structure and species to a trophic species are based on clear function. trophic distinctions across life history (Pimm and Uncommon species, or species with low abund- Rice 1987) whereas in other cases, trophic species ances, may play significant roles in ecosystem are more difficult to define. While taxonomy pro- functioning (e.g. Lyons and Schwartz 2001) but by vides one approach to the aggregation or splitting definition these species are hard to find. A recon- of species in a food web, the ‘‘trophic species’’ sideration of inconspicuous but potentially sig- concept (Cohen and Briand 1984; Cohen 1989) is nificant interactions such as parasitic, symbiotic, generally a widely accepted convention that has or host/ relationships may also be been shown to reduce methodological biases warranted for marine food webs. Keystone species (Cohen et al. 1990; Pimm et al. 1991; Martinez are defined by their ability to influence community 1994), but is not without its critics (Schoener 1989; structure in ways that are disproportionate to their Polis 1991). While constructing food webs, caution own abundance (Power and Mills 1995). In some should be taken when aggregating or splitting instances, sampling of shelf and deep-sea systems species, but it is likely that gross aggregation will has recovered numerous species represented by continue for most marine food webs. only a single individual (Etter and Mullineaux Additionally, there are marine taxa for which we 2001) but the overall contribution of these single- know little else other than that they exist. This does ton species to food-web dynamics is difficult to not even account for those species that we have assess. Species which are difficult to collect such as not even discovered or identified. For example, may also be functionally the discovery of the microbial loop highlights a significant. Due to their low nutritive value, gelat- potentially important but overlooked set of taxa. In inous zooplankton such as coelenterates, cteno- marine ecosystems, microbes play important roles phores, and are generally considered to be an in decomposition of detrital material and renewal unimportant food source for most economically of resources for primary producers (Schlesinger important fish species. However, gelatinous prey 1997). Furthermore, evidence suggests that mic- are thought to be consumed by over 100 fish spe- robes may be as important as zooplankton in cies and are the main food source for some pre- consuming organic carbon in some marine eco- dators including pelagic , moonfish, and systems (Cho and Azam 1988) and that the stromateoid fishes (Verity and Smetacek 1996). shunting of carbon into the microbial loop may They are also the dominant predators of many represent a in marine systems (Azam zooplankton and fish larvae (Purcell 1986). More- et al. 1983). The additional role of phagotrophic over, the majority of trophic cascades identified in is under-known but suspected to be pelagic marine ecosystems are initiated by gelat- an important link between microbes and the inous zooplankton (Verity and Smetacek 1996), zooplankton consumed by higher trophic-level indicating that these species are probably import- metazoans (Sherr et al. 1986). Despite their ant determinants of energy flow and ecosystem important functional roles, detailed microbial function. Overlooked or understudied organisms FOOD-WEB THEORY IN MARINE ECOSYSTEMS 103 are more likely the norm rather than the exception collection at those locales. However, there was in marine ecosystems. less of a latitudinal gradient than one may have expected or that has been observed in other Obiter dicta disciplines (e.g. limnology, fisheries science). There are three major points from the discussion We distinguished three types of food webs about the uniqueness of marine ecosystems. First, (Winemiller and Polis 1996) based upon their many of the S and L (and hence the food web information content: (1) topological or descriptive metrics and theory derived from them) are prob- webs, (2) flow or bioenergetic webs, and (3) inter- ably undetected, and thus underestimated, for action, functional, or biodemographic webs. marine ecosystems. Second, as marine food webs Topological webs are qualitative in nature; only become more exhaustive, S, L and those para- the presence/absence of interactions between meters derived from them are expected to be much groups is indicated. Bioenergetic and interaction higher than typically reported for most food webs food webs are quantitative; the relative strengths and probably much higher than generally sus- of trophic interactions between groups are indi- pected from food-web theory. Finally, by necessity cated. Bioenergetic webs quantify the transport (enforced via logistical constraints), there will con- through consumption of energy/matter among tinue to be some degree of aggregation in marine groups, while interaction webs depict the strength food webs for the foreseeable future. A more rig- of links between groups in terms of their influence orous look at how taxonomic or trophic resolution on the dynamics of community composition and in the construction of marine food webs affects structure. Using this distinction, we noted which topological metrics, network metrics, and energy type of web each of the food-web case studies flows is needed. were, but in some instances we also noted that the study could have fallen into more than one cate- gory. The vast majority of marine food webs were A compilation of marine food webs categorized as bioenergetic (84%), while seven were interaction food webs (10%) (Table 9.1). The We assembled case studies of marine food webs remaining food webs were topological (22%) and published in the past 25 years (Table 9.1). We periodically cooccurred with one of the other two limited our inclusion of examples to those food categories. Because most of the marine food webs webs that are more representative of larger ocean were highly aggregated, even the topological ecosystems (i.e. not an estuary, not a small webs, we did not attempt to calculate or compare embayment, not a rocky , etc.). The common food-web macrodescriptors for these list is by no means exhaustive, but the inclusive- marine food webs. ness and large number (107) of cases likely make However, we did note the number of species (or this list representative of the food-web work approximate number if not given, if species were executed in marine systems. aggregated, or if multiple webs were constructed). Given the caveats described in the section above, Metrics and categorization we recognize that it is not inconceivable for marine food webs to have an S on the order of several We categorized the type of food web into one of hundreds, with L on the order of several thou- seven ecosystems (bounded seas (6%), coastal sands or tens of thousands. Yet given the sampling (20%), (40%), (11%), constraints of working in marine ecosystems, the mid ocean gyre (4%), seamount (6%), or upwelling highest number of S observed was on the order (14%)) along a latitudinal gradient (equatorial of 100 (Table 9.1). Most marine food webs with S (11%), subtropical (31%), temperate (31%), or boreal of 40 or greater were, relatively speaking, well (27%)). Clearly the more difficult-to-reach ecosys- studied. Still, despite the caveats described above tems (sea mounts or mid ocean gyres) were the and the known limitations of food-web catalogs, the least studied, reflective of the high cost of data average S for the marine for food webs examined Table 9.1 Compilation of marine food webs

Food web Type of Latitude Source Topological Flow/ Interaction/ # Degree of spp. resolution Comprehensiveness system Energetic functional Spp. of spp. Apex Upper Middle Lower Basal coverage TLs TLs TLs TLs TLs

Baltic Sea Bounded seas Temperate Baird et al. x x — 15 Low Low Med Med Low Moderate 1991 Baltic Sea Bounded seas Temperate Harvey et al. — x — 15 Med High Low Low Low Moderate 2003 Baltic Sea Bounded seas Temperate Sandberg et al. — x — 10–15 n.a. Low Low Low Low Limited 2000 North Sea Bounded seas Temperate Christensen — x — 29 High High Med Med Low Reasonable 1995 North Sea Bounded seas Temperate Greenstreet et al. — x — 9 Low Low Low Low n.a. Limited 1997 North Sea Bounded seas Temperate Andersen and ? — x 20–25 Med Hi Med Low Low Moderate Ursin 1977 California Coastal Temperate Clarke et al. x x — 24 Hi Hi Med Low n.a. Reasonable 1967 Monterey Bay, Coastal Temperate Olivieri et al. — x — 16 Low Low Low Low Low Limited California 1993 North Central Coastal Temperate Ortiz and — x — 23 n.a. High High Med Low Moderate Chile Wolff 2002 Port Phillip Bay, Coastal Temperate Fulton and — x — 34 Med Med Med Low Med Moderate Australia Smith 2002 Suruga Bay, Japan Coastal Temperate Hogetsu 1979 — x — 16 n.a. High High High High Moderate Borneo Coastal Subtropical Pauly and — x — 13 Low Low Low Low Low Moderate Christensen 1993 Campeche Bank, Coastal Subtropical Vega-Cendejas et al. — x — 18 n.a. High High Med Low Moderate Mexico 1993 Central Gulf of Coastal Subtropical Arreguin-Sanchez and — x — 26 Low Med Med Low Low Moderate California Calderon-Aguilera 2002 Gulf of Thailand soft Coastal Subtropical Pauly and — x — 15 Low Low Low Med Low Limited bottom community Christensen 1993 Hong Kong territorial Coastal Subtropical Cheung et al. 2002 — x — 33 Med Med Med Low Low Reasonable water Malaysia Coastal Subtropical Pauly and — x — 13 Low Low Low Low Low Limited Christensen 1993 Phillipine Coastal Subtropical Pauly and — x — 17 High Med Low Low Low Moderate Christensen 1993 San Miguel Bay, Coastal Subtropical Bundy and Pauly —x—16n.a. Med Med Low Low Moderate Philippines 2000 (200 þ ) South China Sea, Coastal Subtropical Pauly and Christensen — x — 13 Low Low Low Low Low Limited coastal benthos 1993 South China Sea, Coastal Subtropical Pauly and Christensen — x — 15 Low Low Low Med Low Moderate shallow water 1993 South China Sea, Coastal Subtropical Pauly and Christensen — x — 13 Low Med Low Low Low Moderate Vietnam and 1993 China SW Gulf of Mexico Coastal Subtropical Arregun-Sanchez — x — 24 Med High Med Low Low Moderate et al. 1993 Tamiahua, Mexico Coastal Subtropical Abarca-Arenas and — x — 13 Med Med Low Low Low Moderate Valero Pacheco 1993 Northeast Greenland, Coastal Boreal Rysgaard et al. 1999 ? — — 15–20 n.a. n.a. n.a. Med Med Limited Northern British Coastal Boreal Beattie and — x — 53 Med High Med Low Low Reasonable Columbia Vasconcellos, 2002 Prince William Sound Coastal Boreal Okey and Pauly, 1999 — x — 50 Med Med Low Low Low Reasonable Bay of Biscay Continental Temperate Ainsworth et al. 2001 — x — 38 Med Med Low Low Low Moderate 1970 and 1998 Shelf Mid-Atlantic Bight US Continental Temperate Okey 2001 — x — 55 Med High High Low Low Reasonable Shelf SE Australia Continental Temperate Bulman et al. 2001 x — — 100 þ Med Med Med Med Low Reasonable Shelf SE Australia Continental Temperate Bulman et al. 2001 x — — 100 þ Med Med Med Med Low Reasonable Shelf Southeastern Continental Temperate Okey and Pugilese — x — 42 High High Med Med Low Reasonable United States Shelf 2001 US NW Atlantic Continental Temperate Link 2002 x — — 81 Med High High Med Low Reasonable Shelf Gulf of Thailand Continental Subtropical Christensen 1998 — x — 26 Med Med Low Low Low Reasonable Shelf Minnan-TaiwancHighentan, Continental Subtropical Qiyong et al. 1981 x — — 38 n.a. Med High High Low Moderate South China Sea Shelf South China Sea Continental Subtropical Silvestre et al. 1993 — x — 13 Low Low Low Low Low Limited Shelf Table 9.1 (Continued)

Food web Type of Latitude Source Topological Flow/ Interaction/ # Degree of spp. resolution Comprehensiveness system Energetic functional Spp. of spp. Apex Upper Middle Lower Basal coverage TLs TLs TLs TLs TLs

South China Sea, Continental Subtropical Pauly and — x — 13 Low Low Low Med Low Moderate Borneo Shelf Christensen 1993 South China Sea, Continental Subtropical Pauly and — x — 13 Low Low Med Low Low Limited deep shelf Shelf Christensen 1993 South China Sea, Continental Subtropical Pauly and — x — 13 Low Low Low Low Low Limited deeper shelf Shelf Christensen 1993 South China Sea, Continental Subtropical Pauly and — x — 15 Low Low Low Low Low Limited Gulf of Thailand Shelf Christensen 1993 South China Sea, Continental Subtropical Pauly and — x — 15 Low Low Low Med Low Moderate Gulf of Thailand Shelf Christensen 1993 South China Sea, Continental Subtropical Pauly and — x — 17 Med Med Med Med Low Moderate NW Phillipines Shelf Christensen 1993 South China Sea, Continental Subtropical Pauly and — x — 10 Low Low Low Low Low Limited pelagia Shelf Christensen 1993 South China Sea, Continental Subtropical Pauly and — x — 13 Low Low Low Med Low Moderate SW SCS Shelf Christensen 1993 Southwestern Gulf of Continental Subtropical Manickchand-Heileman x x — 19 Med High Med Low Low Moderate Mexico Shelf et al. 1998a,b US Gulf of Mexico Continental Subtropical Browder 1993 — x — 15 Med Med Low Low Low Moderate Shelf Venezuela Shelf Continental Subtropical Mendoza 1993 — x — 16 Med Med Low Low Low Moderate Shelf Yucatan, Mexico Continental Subtropical Arregun-Sanchez — x — 21 Med Med Med Low Low Moderate Shelf et al. 1993 Strait of Bali Continental equatorial Buchary et al. 2002 — x — 14 Med Med Low Low Low Moderate Shelf Antarctic Continental Boreal Constable et al. 2000, — — x 15–25 Med High Med Low n.a. Moderate Shelf Constable 2001 Antarctic Continental Boreal Ainley et al. 1991 x — — 20–25 n.a. High Med Med Low Moderate Shelf Arcitc Ocean Continental Boreal Tittlemier et al. 2002 ? x — 9–10 n.a. Low Med Low n.a. Moderate Shelf Barents Sea Continental Boreal Hop et al. 2002 ? x — 10–15 Med Med Med Low Low Limited Shelf Barents Sea Continental Boreal Bogstad et al. 1997, — — x 5–15 Med High Med Low n.a. Moderate Shelf Tjelmeland and Bogstad 1998 Barents Sea Continental Boreal Borga et al. 2001 ? x — 9–15 n.a. Med Med Med n.a. Moderate Shelf Eastern Bering Sea Continental Boreal Trites et al. 1999 — x — 25 Med Med Low Low Low Moderate Shelf Eastern Bering Sea Continental Boreal Aydin et al. 2002 — x — 38 Med Med Low Low Low Reasonable Shelf Hi Continental Boreal Hobson and Welch ? x — 40–45 Med Med High High Med Reasonable Shelf 1992 Lancaster Sound Continental Boreal Mohammed 2001 — x — 31 Med High Med Med Low Moderate 1980s Shelf Lancaster Sound, Continental Boreal Atwell et al. 1998, x ? — 27 Med Med Low High n.a. Reasonable Northwest Shelf Welch et al. 1992 Territores, Canada Newfoundland Continental Boreal Gomes 1993 x — x 26 Low Med Med Med Low Moderate Southern and Shelf Northeastern Grand Bank Newfoundland-Labrador Continental Boreal Bundy 2001, — x — 20–25 Med High High Med n.a. Moderate Shelf Bundy et al. 2000 Northern Polynya, Continental Boreal Fisk et al. 2001, ? x — 35–40 Med Med High High Med Reasonable Arctic Ocean Shelf Hobson et al. 1995 Norwegian and Continental Boreal Dommasnes et al. 2001 — x — 30 Med High Med Med Low Reasonable Barents Seas Shelf Prince Edward Island, Continental Boreal Kaehler et al. 2000 x ? — 38 n.a. Med High Med Low Reasonable Shelf Southern Ocean Continental Boreal Murphy 1995 — — x 4–20 Low High High Med n.a. Moderate Shelf Southern Ocean, Continental Boreal Reid and Croxall 2001 — x — 4–7 Low Med Med Low n.a. Limited Shelf West Antarctica Peninsula Continental Boreal Barrera-Oro 2002 x — — 25–35 Med Med Med Low Low Moderate Shelf West Greenland Continental Boreal Pederson and Zeller 2001, — x — 22 Med High Med Low Low Moderate Shelf Pederson 1994 Western Bering Sea Continental Boreal Aydin et al. 2002 — x — 36 Med Med Low Low Low Moderate Shelf Table 9.1 (Continued)

Food web Type of Latitude Source Topological Flow/ Interaction/ # Degree of spp. resolution Comprehensiveness system Energetic functional Spp. of spp. Apex Upper Middle Lower Basal coverage TLs TLs TLs TLs TLs

Bolinao, NW Philippines Coral reef Subtropical Alino¨ et al. 1993 — x — 25 Med Med Low Low Low Florida Keys Coral reef Subtropical Venier and Pauly 1997 — x — 20 Med Low Med Low Low Moderate Northern , Coral reef Subtropical Gribble 2001 — x — 24 Med Med Low Low Low Reasonable Australia South China Sea, Coral reef Subtropical Pauly and — x — 13 Med Med Low Med Low Limited coral reef Christensen 1993 South China Sea, reef flats Coral reef Subtropical Pauly and Christensen — x — 26 Low Low Med Med Low Moderate 1993 Caribbean Coral reef Equatorial Opitz 1993 — x — 11 Low Low Low Low Low Limited Caribbean Coral reef Equatorial Opitz 1993 — x — 20 Low Low Med Low Low Moderate Caribbean Coral reef Equatorial Opitz 1993 — x — 50 High Med Med Med Low Reasonable French Frigate Coral reef Equatorial Polovina 1984 — x — 12 Low Low Low Low Low Limited Moorea Island, French Coral reef Equatorial Arias-Gonzalez — x — 43–46 n.a. Med Med Med Med Reasonable Polynesia et al. 1997 Moorea Island, French Coral reef Equatorial Arias-Gonzalez — x — 43–46 n.a. Med Med Med Med Reasonable Polynesia et al. 1997 Takapoto Atoll lagoon Coral reef Equatorial Niquil et al. 1999 — x — 7–10 n.a. n.a. n.a. Low Low Limited Central North Pacific Mid ocean Equatorial Kitchell et al. 1999, — x — 27 High High Low Low Low Reasonable gyre 2002 Eastern Tropical Pacific Mid ocean Equatorial Olson and Watters, 2003; — x — 36 Med High Low Low Low Reasonable gyre Essington et al. 2002 Pacific Warm Pool Pelagic Mid ocean Equatorial Godinot and Allain 2003 — x — 20 High Med Low Low Low Moderate Ecosystem gyre Western and Central Pacific Mid ocean Equatorial Cox et al. 2002 — x — 25 High High Med Low Low Reasonable Ocean, warm pool pelagic gyre Azores Archipelago 1997 Seamount Temperate Guenette and Morato — x — 43 Med Med Low Low Low Moderate 2001 Faroe Islands (NE Atlantic) Seamount Boreal Zeller and Freire 2001 — x — 20 Med High Med Med Low Moderate Iceland Seamount Boreal Bjornsson 1998 — x x 3 n.a. Med Low n.a. n.a. Limited Icelandic Ecosystem 1950 Seamount Boreal Buchary 2001 — x — 24 Med High Med Med Low Reasonable Icelandic Ecosystem 1997 Seamount Boreal Mendy and Buchary 2001 — x — 25 Med High Med Med Low Reasonable New Zealand Southern Seamount Boreal Bradford-Grieve and — x — 18 Med Low Low Low Low Moderate Plateau Hanchet 2002; Bradford-Grieve et al. 2003 Benguela Current Upwelling Temperate Baird et al. 1991 x x — 16 Low Low Med Med Low Moderate Benguela Current Upwelling Temperate Yodzis 1998, 2000, x — x 29 Med High Med Low Low Reasonable et al. 1991 , Peru Upwelling Temperate Jarre et al. 1991 x x — 16 Med High Med Low Low Moderate Humboldt Current, Peru Upwelling Temperate Jarre-Teichmann and — x — 20 High High Low Low Low Moderate Pauly 1993 Morocco Atlantic Upwelling Temperate Stanford et al. 2001 — x — 38 Med Low Low Low Low Moderate Coast mid-1980s North Central Chile Upwelling Temperate Wolff 1994 — x — 17 Low Low Low Low Low Limited Northern Benguela Upwelling Temperate Heymans et al. 2004 — x — 17 n.a. Med Low Low Low Moderate Northern Benguela Upwelling Temperate Heymans and Baird — x — 24 Med Med Med Low Low Reasonable 2000 Northern Benguela Upwelling Temperate Shannon and — x — 24 Med Med Med Low Low Reasonable Jarre-Teichman 1999 Oregon/Wash. Upwelling Temperate Brodeur and Pearcy — — x 22 n.a. High High Low Low Limited Coast 1981 1992 Oregon/Wash. Upwelling Temperate Brodeur and Pearcy — — x 21 n.a. High High Low Low Limited Coast 1982 1992 Oregon/Wash. Upwelling Temperate Brodeur and Pearcy — — x 24 n.a. High High Low Low Limited Coast 1983 1992 Oregon/Wash. Upwelling Temperate Brodeur and Pearcy — — x 20 n.a. High High Low Low Limited Coast 1984 1992 Peruvian Current Upwelling Temperate Baird et al. 1991 x x — 15 Low Low Med Med Low Moderate Southern Benguela Upwelling Temperate Jarre-Teichman — x — 19 Med Med Med Low Low Moderate et al. 1998 110 AQUATIC FOOD WEBS was 20–25. This is a clear area for improvement in no means are we stating that even a marine food future studies. web with an S of 50–100 is exhaustive nor neces- We also categorized the degree of species res- sarily complete. olution across the various groups using five cat- egories loosely corresponding to a trophic level or Obiter dicta an associated group of trophic levels: apex, upper, There are a few general observations from this mid, lower, basal. We used a qualitative rank of compilation of marine food webs. First is that most low, medium, or high to characterize how well of the species emphasized, most of the species all the species in each group was represented and studied with any degree of resolution, and most of also how much aggregation of species occurred. the stated goals for assembling these food webs are With very few exceptions (<1%), most published related to marine fisheries. marine food webs did not address the basal trophic Major criticisms of aquatic and marine food levels with any degree of detail (Table 9.1). The webs are that (1) they are highly biased toward fish other trophic levels in 35–50% of marine food webs species, and (2) basal taxa are extremely aggre- had species with at least a medium or greater gated in these webs. These criticisms and obser- resolution. Surprisingly, >14% of marine food vations are in part artifacts of the expense and webs had apex trophic levels that were not con- challenges of sampling marine ecosystems, but sidered. When comparing these results to previous also reflect the national priorities experienced by web catalogs (Briand and Cohen 1984, 1987; Cohen many marine research scientists and organizations. and Briand 1984; Cohen 1989; Schoener 1989; A large proportion of researchers who have access Sugihara et al. 1989; Cohen et al. 1990; Pimm et al. to the necessary large-scale marine sampling 1991; Polis 1991; Havens 1992) this pattern con- equipment also have concurrent professional firms the continued trend of poor treatment for obligations to conduct stock assessments, deter- lower trophic levels and highlights the dearth of mine fisheries impacts on economically valuable information for many of the apex consumers in fish populations, and generally provide fishery marine food webs. management advice. Thus, the focus on data col- Finally, we noted the comprehensiveness of lection tends to be geared toward commercially species coverage relative to the overall number (or targeted species. Consequently most of our suspected number) of the species in the system. knowledge on marine species interactions is cen- We used a subjecitive rank, limited, moderate, or tered on these groups. As suggested by Raffaelli reasonable to provide an overall sense of exten- (2000), more collaborations among scientists from siveness for each of the food webs. By no means a broader array of marine and ecological dis- are we stating that any of the food webs with ciplines may ameliorate these problems. limited coverage are of poor quality, rather that The second observation is that other studies, they are not likely to fully capture S and L nor such as those exploring organochlorine or pesti- contribute directly to associated theories based cide biomagnification, isotope signatures, or upon those metrics. Many of the limited food webs focused studies on select marine organisms (e.g. (22%) were from organochlorine biomagnification sea birds, benthos, etc.) can provide useful and or isotope studies that encapsulated only portions ancillary information for marine food webs. of a food web. Most of the other limited food webs These studies need to be recognized as incomplete were from studies that emphasized a particular in a classical food-web sense, but should not taxonomic group that omitted or greatly aggre- be summarily ignored either. In particular, com- gated other species. We included these in our bining multiple and interdisciplinary studies from compilation, as they do provide some information, the same ecosystem may provide the basis for but their utility for testing some food-web theories constructing a more complete food web of that individually may be limited. That >75% of the system. The implication is that the literature marine food webs examined are of moderate or from other disciplines (e.g. environmental con- reasonable coverage is encouraging. However, by taminants, , , etc.) FOOD-WEB THEORY IN MARINE ECOSYSTEMS 111 may be an unmined source of valuable marine levels beyond a gross aggregation of basal species. food-web data. Because of this, the predator to prey ratio will The final observation from this compilation is probably continue to be consistent across most that, given the pros and cons of Ecopath/Ecosim food webs as well. Finally, it is likely that marine (E/E, Christensen and Pauly 1992; Walters et al. food webs will continue emphasizing commer- 1997; discussed in Hollowed et al. 2000 and cially valuable species due to the increased interest Whipple et al. 2000), this modeling framework has in collecting data on those organisms, similar to been a valuable tool for cataloging marine food other ecosystems. webs. With some exceptions, most marine food Estimating interaction strengths or functional webs have been published since the early 1990s, value in marine food webs still remains a major and the vast majority of these are bioenergetic flow challenge. We suspect that in simpler boreal models using E/E. However, unlike other network (relatively lower S and L) ecosystems that the analyses or topological food webs, many of the interaction strengths i will be higher than in more food-web macrodescriptors are lacking or not complex temperate ecosystems. Conversely, i in published with E/E models, yet they need to be highly complex but highly specialized food webs presented to further explore some of the basic (e.g. equatorial coral reefs) will continue to be tenets of food-web theory as applied to marine some of the highest interaction strengths measured systems. Additionally, most marine food webs, (sensu Emery 1978; Carr et al. 2002). Marine food including but not limited to those constructed webs are not unique, however, in that functional or under an E/E framework, exhibit a high degree interaction webs are the least represented type of of aggregation. We need to broaden our under- food webs. standing and better our resolution of marine food In many respects, the topologies of marine food webs in order to test food-web theory for ecosys- webs already confirm more recent and exhaustive tems that cover >70% of the planet’s surface. food-web studies (e.g. Winemiller 1990; Hall and Raffaelli 1991; Martinez 1991; Polis 1991; Food-web theory and marine Goldwasser and Roughgarden 1993; Reagan et al. ecosystems 1996). For example, if one examines the level of C in most marine ecosystems in relation to S, these Clearly, there is much work to be done in marine food webs are distinct outliers in the decreasing food-web research. More exhaustive and more hyperbolic curve of C versus S (e.g. Link 2002). In highly resolved food webs from marine ecosys- fact, many of the marine food webs do not fall on tems will broaden our understanding of food webs the curve at all. These observations have clear in general, let us compare marine food webs to implications for food-web stability and may ulti- their nonmarine counterparts, and will test how mately mean that marine food webs support the these webs support or contradict food-web theory. constant connectance hypothesis (Martinez 1992), Recognizing the current bounds on knowledge but need to extend it in terms of magnitude of S. about marine food webs, we are not precluded from making a few limited observations about Marine food-web disparities with theory marine food webs and how further exploration and inclusion of them may alter food-web theory. However, in some respects marine food webs are not likely to fit empirically based theory that was developed from land or freshwater. Using the Marine food web verification of theory example above, observations of S, C and lack of Marine food webs are likely similar to all other a hyperbolic decline in the relationship between food webs with the share of species primarily the two does not necessarily mean that marine clustered at intermediate trophic levels. Like food webs concur with the recent niche model aquatic food webs, they are also not likely to (Williams and Martinez 2000; Dunne et al. 2004). integrate lower trophic levels with upper trophic The reasoning is threefold. First, these and prior 112 AQUATIC FOOD WEBS theories were developed with S and L that were an Thus, it is premature to state whether food-web- order of magnitude or more lower than what theory will fully encapsulate all the observations actually occurs in marine food webs. Second, of marine food-web structure and functioning nonlinearities and the high probability of alternate (vis-a´-vis Dunne et al. 2004). The most immediate steady states are common in marine systems and useful test might be to compare large lake (Steele 1985). Steele (1985) and Link (2002) discuss systems to high latitude marine systems with how stability may neither be an observable careful controls for the number of species (e.g. in marine systems nor a useful con- focus on less speciose arcto-boreal marine sys- struct given the high degree of perturbation and tems in comparisons with temperate and boreal dynamics experienced by marine ecosystems. lake systems where there may be more parity Third, we strongly suspect that we have not in S). Such a comparison might allow useful yet incorporated or even collected all the salient contrasts while controlling for similarities and information to best elucidate the situation differences in life history patterns of the species (Raffaelli 2000). involved. For those few marine food webs that report topological metrics, these metrics consistently Conclusions stand apart from those of their terrestrial and freshwater counterparts. Marine food webs will The implications from studying marine food webs continue to have higher LD (Bengtsson 1994; Cohen and associated theory are widespread for living 1994; Link 2002), higher average chain lengths marine resource management. Food webs provide (Bengtsson 1994; Cohen 1994), higher maximum a context for the management of living resources chain lengths (Cohen 1994; Schoener 1989), and (Crowder et al. 1996; Winemiller and Polis 1996) higher C values (Bengtsson 1994; Link 2002). Changes in marine food webs can potentially alter Cannibalism, omnivory, and cycles will also con- all populations, including those that support or are tinue to be prominent features of most marine food economically valuable species. Changes in food- webs. Marine food webs will probably continue to web theory may also alter how we view marine challenge the scale invariance laws. Additionally, food webs, fundamentally altering the feasibility of network metrics from marine systems will also our expectations for particular management probably continue to confirm the amazing com- objectives. plexity found in food webs. In particular, indices If theories continue to explore the issue of food- of energy re-cycling within marine food webs will web stability and structure, it is likely that marine remain high (e.g. Manickchand-Heileman et al. food webs will expand this discussion, yet perhaps 1998b, Niquil et al. 1999; Abarca-Arenas and without providing conclusive evidence one way or Ulanowicz 2002). another. However, Link (2002) has argued that Most of the data presented for marine food webs assessing the stability of marine food webs may be do not fully cover the entire size range, distribu- a moot point given the ongoing harvesting pres- tion, and trophic functioning across the entire life sure component populations have experienced history of most marine species. Most of the data over the past several decades (Pauly et al. 1998; presented for marine food webs severely aggregate Jackson et al. 2001). Assuming that marine eco- under-known or unstudied organisms, if they do systems exhibit at least Lyapunov stability, then not omit them entirely. Most of the data presented two points stand out. One is that with the high for marine food webs represents only a small degree of interactions in most marine food webs, snapshot of the number of species and species the resilience is going to be very high for the entire interactions in an ecosystem. In fact, many of the system. That is, to return to a historical equili- marine food-web macrodescriptors (S, L, LD,C) brium will take a long time. Second, it is likely that and related network metrics which we had marine food webs may be perturbed beyond his- originally planned to examine and compare with torical equilibria and shifted to new stable states other systems were unfortunately unavailable. (Steele 1985; Link 2002). FOOD-WEB THEORY IN MARINE ECOSYSTEMS 113

Additionally, the sheer complexity of marine (NMFS 1999) approach is premised on a solid food webs makes them difficult to predict. How understanding of marine food webs. the populations of marine food webs will collec- The search for unifying concepts in ecology has tively fluctuate from one steady state to another far-reaching significance. Food webs are complex, remains a major, if not the key management chal- and we submit that marine food webs are probably lenge for global marine resource managers. Given some of the worst cases of food-web complexity. the complexity of marine food webs, the high We also submit that the unique biological and degree of omnivory, and the generalist nature of physical properties (as discussed above) displayed most fish, it is unclear if predicting unambiguous in marine ecosystems distinguish them from non- trade-offs in biomass allocation among species in marine food webs, perhaps even fundamentally marine ecosystems is entirely practical. very different than other types of food webs. However, some extant approaches are making Conversely, this also means that the potential to valuable contributions to better deal with marine make significant contributions to food-web theory food-web dynamics and to better manage living via the further examination of marine food webs marine resources. Many of the more recent net- may be quite high. work approaches and E/E (Christensen and Pauly 1992) explicitly explore trade-offs in energy flows Acknowledgments and biomass among groups of species in a food web. Other multispecies and multivariate models We thank all members of the Food-Web Dynamics (Hollowed et al. 2000; Whipple et al. 2000) have Program, past and present, for their dedicated also shown promise toward this end. Although the effort at collecting, auditing, and maintaining a requisite decision criteria remain to be fully food habits database of unprecedented scale and developed, the ability to predict the ‘‘climate if not scope, from which much of our local empirical the weather’’ is a promising intersection between observations are taken. We also thank the editors food-web theory and resource management. of this book for inviting us to contribute this Translating many of the food-web macro- chapter. Finally, we thank J. Dunne and M. Fogarty descriptors and network metrics into a decision who providing constructive comments on earlier criteria format remains a key area of research. versions of the manuscript which notably The entire ecosystem-based fisheries management improved its content. This page intentionally left blank PART III Stability and diversity in food webs

The following chapters link food-web stability and In addition to diversity, the structure of food species diversity in aquatic systems to underlying webs can also fundamentally influence community trophic dynamics across different habitats, spatio- dynamics and stability. For example, structure is temporal scales, and levels of organization. Previous shown to mediate the relationship between external studies of patterns of diversity have usually climate forcing and species dynamics in large, focused on single trophic levels. Food webs pro- pelagic, marine ecosystems. In general, a funda- vide a framework for integrating diverse types of mental insight emerging from these chapters is that diversity–stability studies, for example, studies it is necessary to combine analyses of the network that focus on guilds of aggregated species, studies structure of complex food webs with dynamical that look at diversity within trophic levels, approaches if we want to make progress on how to and studies that focus on the role of indirect implement the use of dynamical food-web models interactions. in a conservation and management context. This page intentionally left blank CHAPTER 10 Modeling food-web dynamics: complexity–stability implications

Jennifer A. Dunne, Ulrich Brose, Richard J. Williams, and Neo D. Martinez

Introduction beyond the one- or two-species population dynamics modeling paradigm, other research has Understanding the structure and dynamics of expanded the focus to include 3–8 species, explor- ecological networks is critical for understanding ing dynamics in slightly more complex systems. the persistence and stability of ecosystems. Deter- However, these interaction modules still present mining the interplay among network structure, a drastic simplification of the diversity and struc- network dynamics, and various aspects of stability ture of natural ecosystems. Other approaches have such as persistence, robustness, and resilience in focused on higher diversity ecological networks, complex ‘‘real-world’’ networks is one of the but have either left out dynamics altogether greatest current challenges in the natural and (e.g. network topology and carbon-budget studies) social sciences, and it represents an exciting and or ignored network structure in order to conduct dramatically expanding area of cross-disciplinary analytically tractable dynamical analyses. The inquiry (Strogatz 2001). Within ecology, food-web nature of modeling requires judicious simplifica- research represents a long tradition of both tions (Yodzis and Innes 1992), but such choices can empirical and theoretical network analysis (e.g. leave empirical ecologists suspicious and theore- Elton 1927; Lindeman 1942; MacArthur 1955; Paine tical ecologists perplexed, both wondering how to 1966; May 1973). All aspects of ecological network bridge the gap between simple mathematical research have increasing relevance (McCann 2000) models and complex natural systems. At worst, for a world facing biodiversity and habitat loss, theoreticians and empiricists end up ignoring , climate change, and other or deriding each other’s work, dismissing it as anthropogenic factors that are resulting in the irrelevant for being too abstract or too particular, drastic reorganization of many ecosystems, and in never the twain shall meet. some cases may lead to the collapse of ecosystems In this chapter, we focus on theoretical aspects of and the vital, underappreciated services they pro- the broader food-web research agenda, particu- vide human society (Daily 1997). larly the background and various approaches used Most theoretical studies of trophic dynamics for modeling food-web dynamics in abstract have focused narrowly on predator–prey or systems with more than two taxa. Much of this parasite–host interactions, and have thus ignored type of modeling has oriented itself around the network structure. In natural ecosystems such classic (May 1972) and enduring (McCann 2000) interaction dyads are embedded in diverse, com- complexity–stability debate, especially those plex food webs, where many additional taxa and aspects which relate to the theoretical and their direct and indirect effects can play important associated empirical food-web research into the roles for both the stability of focal species as well relationships between ecosystem complexity, often as the stability of the broader community. Moving characterized as number of links and/or number

117 118 AQUATIC FOOD WEBS of species in a community, and various aspects of We focus on one end of a spectrum of modeling ecosystem stability. ‘‘Stability’’ is a catchall term (Holling 1966), the use of abstract models to that has been variously defined to reflect aspects of elucidate general qualitative insights about eco- population and/or system equilibrium, persist- system structure, dynamics, and stability. We do ence, resilience, resistance, and robustness (see not discuss more explicitly applied approaches for McCann 2000 for some broad, but not exhaustive, simulating particular ecosystems, a strategy which stability definitions). In some cases stability refers has been used for many aquatic systems (e.g. to the outcome of internal dynamics while in mass-balance, carbon-budget models such as other cases it reflects the response of a population Ecopath/Ecosim: Christensen and Walters 2004). or system to a perturbation. The very notion of Such approaches are presented elsewhere in this what is a plausible definition of ecosystem stability book (Chapter 3 by Christian et al. and Chapter 7 has driven some of the fundamental shifts in by Morris et al.). These researchers have often modeling methodology that will be described in referred to their work as ‘‘network analysis.’’ the following sections. We do not address the more However, in our view, network analysis is a very empirically and experimentally based ‘‘diversity– broad term that encompasses all types of research stability’’ debate of recent years (e.g. Naeem et al. that treat a system as a network of nodes and 1994; Hector et al. 1999; Tilman et al. 2001), which ‘‘edges,’’ or links, regardless of how those nodes mostly focuses on the relationship between plant and links are defined or their relationships are species richness and primary productivity. analyzed (Strogatz 2001). In ecology, there has This chapter discusses how three approaches to often been a tension between research focused on dynamical modeling of ecological networks try to abstract models versus that focused on applied strike a balance between simplifying and embody- simulation models of particular ecosystems. We ing aspects of the complexity of natural systems to shall see that as abstract models seek to represent gain a better understanding of aspects of ecosystem greater ecological complexity, they take on some of stability. In particular, the degree to which the the characteristics of applied simulation modeling, different approaches incorporate diversity, network hopefully narrowing the gap between theoretical structure, nonlinear dynamics, and empirically and empirical work by carefully integrating char- measurable parameters will be recurring themes. We acteristics and goals of the two modeling extremes. do not address spatial heterogeneity (see Chapter 2 In this spirit, we will end by reviewing a recent by Melian et al.), metapopulation dynamics, age– application of a multispecies nonlinear dynamical class structure, environmental variability, or stoi- food-web model to the issue of how culling a top chiometry (see Chapter 1 by Elser and Hessen). predator is likely to affect the hake fishery yield in Inclusion of those important factors in dynamical the Benguela ecosystem (Yodzis 1998, 2000, 2001). food-web models may well alter many of our notions about the interplay between ecosystem Background complexity and stability. While we do not explicitly review studies focused on food-web structure or In the first half of the twentieth century, many topology (e.g. Cohen et al. 1990; Williams and ecologists believed that natural communities Martinez 2000; Dunne et al. 2002a; Garlaschelli et al. develop into stable systems through successional 2003), we do refer to them to the degree that they dynamics. Aspects of this belief developed into the have influenced dynamical approaches. A funda- notion that complex communities are more stable mental message of this chapter is that the insights than simple ones. Odum (1953), MacArthur (1955), from studies on the network structure of complex Elton (1958), and others cited an array of empirical food webs need to be merged with dynamical evidence supporting this hypothesis. Some popular approaches if we are to make serious headway examples included the vulnerability of agricultural regarding issues of the stability of complex net- monocultures to calamities in contrast to the works and the potential for using dynamical models apparent stability of diverse tropical , in a conservation and management context. and the higher frequency of invasions in simple MODELING FOOD-WEB DYNAMICS 119 island communities compared to more complex development of theories of deterministic chaos mainland communities. It was thought that a (e.g. May 1974, 1976). Analytically, May used local community comprised of species with multiple stability analyses of randomly assembled com- consumers would have fewer invasions and pest munity matrices to mathematically demonstrate outbreaks than communities of species with fewer that network stability decreases with complexity. consumers. This was stated in a general, theore- In particular, he found that more diverse systems, tical way by MacArthur (1955), who hypothesized compared to less diverse systems, will tend to that ‘‘a large number of paths through each species sharply transition from stable to unstable behavior is necessary to reduce the effects of as the number of species, the connectance of one species.’’ MacArthur concluded that ‘‘sta- (a measure of link richness—the probability that bility increases as the number of links increases’’ any two species will interact with each other), or and that stability is easier to achieve in more the average interaction strength increase beyond diverse assemblages of species, thus linking com- a critical value. We describe his analyses in more munity stability with both increased trophic links detail in the next section. May (1973) also pointed and increased numbers of species. Other types out that empirical evidence is not conclusive of theoretical considerations emerged to support with regard to a particular complexity–stability the positive complexity–stability relationship. For relationship (e.g. there are a number of highly example, Elton (1958) argued that simple predator– stable, productive, and simple natural ecosystems prey models reveal their lack of stability in the such as east-coast Spartina alterniflora ; oscillatory behavior they exhibit, although he complex continental communities have suffered failed to compare them to multispecies models the ravages of pests such as the Gypsy Moth, etc.). (May 1973). The notion that ‘‘complexity begets May’s analytical results and his conclusion that stability,’’ which already had great intuitive appeal in ‘‘general mathematical models of multispecies as well as the weight of history behind it, was thus communities, complexity tends to beget instabil- accorded a gloss of theoretical rigor in the mid- ity’’ (p. 74, 2001 edition) turned earlier ecological 1950s. By the late 1950s it took on the patina of ‘‘intuition’’ on its head and instigated a dramatic conventional wisdom and at times was elevated shift and refocusing of . His to the status of ecological theorem or ‘‘formal results left many empirical ecologists wondering proof’’ (Hutchinson 1959). While some subsequent how the astonishing diversity and complexity they empirical investigations raised questions about the observed in natural communities could persist, conclusiveness of the relationship (e.g. Hairston even though May himself insisted there was no et al. 1968), it was not until the early 1970s that the paradox. As May put it, ‘‘In short, there is no notion that complexity implies stability, as a the- comfortable theorem assuring that increasing oretical generality, was explicitly and rigorously diversity and complexity beget enhanced com- challenged by the analytical work of Robert May munity stability; rather, as a mathematical gen- (1972, 1973), a physicist by training and ecologist erality, the opposite is true. The task, therefore, is by inclination. to elucidate the devious strategies which make for May’s 1972 paper and his 1973 book Stability stability in enduring natural systems’’ (p. 174, 2001 and Complexity in Model Ecosystems (reprinted in edition). Most subsequent work related to food a 2nd edition in 1974 with an added preface, webs was devoted to finding network structures, afterthoughts, and bibliography; most recently species’ strategies, and dynamical characteristics reprinted in an 8th edition in 2001 as a Princeton that would allow complex communities to persist. Landmarks in Biology volume) were enormously In some cases, the research was undertaken to important on several fronts. Methodologically, corroborate May’s findings, and other research May introduced many ecologists to a much- sought to reanimate what many now felt to be expanded repertoire of mathematical tools beyond the ghost of positive complexity–stability past. Lotka–Volterra predator–prey models, and also set Whatever the motivation, important pieces of this the stage for major ecological contributions to the ongoing research agenda include examining the 120 AQUATIC FOOD WEBS role of omnivory, the number of trophic levels, calculated). For every population, positive real weak interactions between species, adaptive fora- parts of the eigenvalues indicate perturbation ging of consumers, and complex network structure growth, negative real parts indicate perturbation on various measures of ecosystem stability. decay, and zero real parts indicate neutral stability. However, substantially differing methodologies, Accordingly, if any of the eigenvalues has a positive stability definitions, and a variety of sometimes real part the system will be unstable, that is, at least apparently conflicting results have somewhat one of the species does not return to the equili- muddied the original question of whether com- brium. Stability may be measured either as a binary plexity begets stability. Instead, the question variable—return to equilibrium or not—or as a is becoming more usefully recast along the lines metric variable in form of the return time needed of May’s notion of ‘‘devious strategies’’—what by the population densities to settle back to the are the general characteristics of complex ecolog- equilibrium. ical networks that allow for, promote, and May (1972, 1973) used community matrices in result from the stability and persistence of natural which species were randomly linked with random ecosystems. interaction strength to show that the local stability decreases with complexity (measured as con- Local stability analyses of community nectance), diversity, and average interaction matrices strength among the species. The use of such random community matrices has attracted much In his seminal studies on food-web stability, May criticism. It was shown to be extremely unlikely (1972, 1973) measured local or neighborhood that any of these random communities even stability. In this analysis, it is assumed that the remotely resembles ecosystems with a minimum community rests at an equilibrium point where all of ecological realism such as containing at least one populations have constant values (Figure 10.1(a)). primary producer, a limited number of trophic The stability of this equilibrium is tested with levels, and no consumers eating resources that are small perturbations. If all species return to two or more trophic levels higher (Lawlor 1978). the equilibrium—monotonically or by damped The nonrandomness of ecosystem structure has oscillations—it is stable (Figure 10.1(a)). In contrast, been demonstrated in detail by more recent food- if the population densities evolve away from the web topology studies (e.g. Williams and Martinez equilibrium densities—monotonically or oscillatory— 2000; Dunne et al. 2002a, 2004; Morris et al. they are unstable (Figure 10.1(b)). Neutral stability Chapter 7, this volume). Accordingly, subsequent represents a third possibility, in which the pertur- work added more structural realism to those ran- bation neither grows nor decays as the population dom community matrices by including empirical densities either reach an alternative equilibrium or patterns of food-web structure and interaction oscillate with a constant amplitude in limit cycles strength distributions. (Figure 10.1(b)). Varying connectance in 10 species food webs, In a community of n species, this approach De Angelis (1975) tested May’s (1972) results within is based on an n n Jacobian community matrix a more realistic model environment and found of species interaction coefficients that describe three conditions for positive complexity–stability the impact of each species i on the growth of relationships: (1) the biomass assimilated by con- each species j at equilibrium population densities. sumers is a small fraction (<50%) of the amount In food webs, these coefficients may be positive of biomass removed from the resource population, (i.e. i is eaten by j), zero (no interaction), or negative (2) the consumers are subject to strong self- (i.e. i eats j). To describe interactions by single dampening (up to 20 times the maximum growth coefficients it is necessary to assume linear inter- rate of primary producers) which restricts their actions. The n eigenvalues of the community matrix population growth, and (3) a bias toward bottom- characterize its temporal behavior (see May 1973 up control of the interactions (i.e. resource biomass for a detailed description how these eigenvalues are has a stronger influence on the interaction strength MODELING FOOD-WEB DYNAMICS 121

(a) 0.29 Stable equilibrium Damped oscillations

0.27

0.25

0.23

0.21 Population density 0.19

0.17 0 200 400 600 800 1000 Time

(b) Limit cycles Amplified oscillation 0.33

0.28

0.23

Population density 0.18

0.13 0 200 400 600 800 1000 Time

(c) 0.45 Chaotic oscillations 0.4 0.35 0.3 0.25 0.2 0.15

Population density 0.1 0.05 0 600 700 800 900 1000 Time

Figure 10.1 Population dynamics in time series. values than consumer biomass), which also means such extreme self-dampening or low influence that consumers have very little influence on the of consumers on resources appear unlikely. dynamics of their resources. The realism of these Nevertheless, for a long time the results of conditions, however, is questionable as assimilat- De Angelis (1975) remained the only mathematical ion efficiencies for carnivores should generally model showing positive complexity–stability be higher (85%, Yodzis and Innes 1992) and relationships. 122 AQUATIC FOOD WEBS

In a more structural approach, Pimm and revealed significant compartmentalization. Girvan Lawton (1977, 1978) used community matrices of and Newman (2002) found ecologically meaningful four interacting species. They systematically varied compartments based on unweighted topology for food-chain length and introduced omnivory by a Chesapeake Bay food web (Baird and Ulanowicz letting one of the species consume resources of 1989), but had difficulty identifying compartments multiple trophic levels. They found that the stabil- in other more highly resolved food webs (Girvan ity of the community matrices decreases with the personal communication). Topological analysis of number of trophic levels. The results for omnivory, 16 diverse, well-resolved food webs showed that however, varied, as omnivory decreases stability their clustering coefficient (Watts and Strogatz in terms of the fraction of stable models that return 1998) is low in comparison to other types of net- to equilibrium but at the same time increases works (Dunne et al. 2002a). The low clustering stability of these stable models by decreasing their coefficients of food webs is in part attributable return time. Nevertheless, they concluded that the to their high connectance (links/species2) com- number of trophic levels in natural communities pared to other real-world networks (Dunne et al. is limited by the stability of population dynamics 2002a)—in effect, there is such a high concentration (Pimm and Lawton 1977), and they predicted that of links in most food webs that the possibility of omnivory should be a rare phenomenon in natural compartmentalization, at least from a purely food webs (Pimm and Lawton 1978). In particular topological perspective, is drowned out. In addi- the latter prediction was not found to be consistent tion to being based on unweighted links, which with empirical data as more detailed food webs may obscure compartmentalization, the clustering were compiled (e.g. Martinez 1991; Polis 1991). In coefficient misses compartments of non-omnivorous a further approach to analyze the relationship species based on a single basal species. In sum, between food-web structure and dynamics, Yodzis the issue of whether and how food webs are (1981) showed that empirical community matrices compartmentalized, and the impacts of compart- corresponding to 40 real food webs are more stable mentalization on stability, remains an open ques- than randomly assembled community matrices. tion that requires more research. This clearly indicated that natural food webs are Extending the approach of adding empirical structured in a way that stabilizes the internal realism to local stability analyses, de Ruiter et al. population dynamics, but the question remained (1995) parameterized the general community open as to which particular structural properties matrix model with empirical food-web structures were important. Interestingly, Moore and Hunt and interaction strengths among the species. In (1988) showed that in those 40 food webs species their empirical data, they found a pattern of strong are blocked into smaller subwebs of high con- top-down effects of consumers on their resources nectance with only a few links connecting the at lower trophic levels in food webs and strong subwebs. Based on analyses of May (1972), they bottom-up effects of resources on their consumers argued that such compartmentalization may at higher trophic levels. Adding empirical inter- increase network stability and thus be responsible action strength patterns to the community matrices for the comparatively high stability of empirical increased their local stability in comparison to food webs. matrices with random interaction strength values. Evidence for clustering or compartmentalization Most importantly, these results demonstrated in more detailed natural food webs has been that natural food-web structures as well as the suggestive, but scant. Krause et al. (2003), using distribution of interaction strengths within those food-web network data with unweighted links structures contribute to an increased local stability as well as with links weighted by interaction of the corresponding community matrices. Neutel frequency, carbon flow, or interaction strength, et al. (2002) explained this finding with results that identified significant compartments in three of five showed that weak interactions are concentrated in complex food webs. However, out of 17 versions long loops. In their analysis, a loop is a pathway of the 5 food webs examined, only 6 versions of interactions from a certain species through MODELING FOOD-WEB DYNAMICS 123 the web back to the same species, without visiting Nonlinear, nonequilibrium population other species more than once. They measured loop stability in food-web modules weight as the geometric mean of the interaction To overcome the limitations of the equilibrium strengths in the loop and showed that loop weight assumption and linear interactions included in decreases with loop length. Again, when applied community matrix analyses, an alternative to the community matrix, this empirically docu- approach to analyzing food-web stability emerged mented pattern of interaction strength distribut- in the early 1990s (e.g. Hastings and Powell 1991; increased its local stability in comparison to Yodzis and Innes 1992; McCann and Yodzis random interaction strength distributions. 1994a,b). Giving up those assumptions prohibited Thus, over the last 30 years, studies of the local the use of community matrices. Instead, these stability of community matrices have contributed to studies defined less analytically tractable, but more our understanding of the relationship between ecologically plausible models of nonlinear species’ complexity and stability. De Angelis (1975) interactions to study the population dynamics indicated some specific conditions for positive in species-poor food-web modules such as three- complexity–stability relationships, whereas none of species food chains. In effect, one aspect of the subsequent studies has inverted the negative ecological plausibility was traded for another: non- complexity–stability relationship demonstrated by linear, nonequilibrium dynamics were embraced, May (1972). Nevertheless, other studies con- but diversity was lost. Most of these nonlinear sistently suggested that local stability increases dynamical studies have used an allometric bio- when ecological realism is added to the community energetic consumer–resource model of species bio- matrix. In this sense, stability is gained by a limited mass evolution that developed by Yodzis and number of trophic levels (Pimm and Lawton Innes (1992). It defines for every species how much 1977), natural food-web structures (Yodzis 1981; biomass is lost due to metabolism or being con- Ulanowicz 2002), and interaction strength sumed and how much biomass is gained due to distributions (de Ruiter et al. 1995; Neutel et al. primary production or feeding on other species. In 2002). However, to gain mathematical tractability, this approach, primary producer growth is based the community matrix approach, with or without on logistic growth equations, and consumer the addition of more plausible ecological structure, growth is based on nonlinear functional responses makes several assumptions that may restrict the of consumption to resource densities (Skalski and generality of its results. First, community equili- Gilliam 2001). Another aspect of ecological brium is assumed, which excludes chaotic popu- plausibility is added through the basis of most of lation dynamics that may characterize many natural its parameters on species body size and metabolic systems (Figure 10.1(c), Hastings et al. 1993). categories (i.e. endotherm, ectotherm vertebrate, Second, only small perturbations are tested. Larger invertebrate, primary producer). Specifying this perturbations might yield alternative equilibria. bioenergetic model for every species in a food web Only addressing local or neighborhood stability yields a set of ordinary differential equations does not necessarily predict global stability, (ODEs). The numerical integration of the ODEs although they can coincide. Third, linear interaction results in time series of species’ biomass evolution coefficients are only a rough approximation of that may be used to characterize population species’ dynamics that will only hold for arbitrarily dynamics (e.g. stable equilibria, limit cycles, small variances in population densities. For any chaotic cycles, Figure 10.1). Furthermore, if species larger population variances, nonlinear species’ extinction is defined as falling below a threshold of interactions will be poorly approximated by linear biomass density (e.g. 1030) the persistence of the coefficients. This assumption of small population populations in the food web can be studied. This variances, however, is likely to hold once the approach led to the identification of persistent analyses are restricted to slightly perturbed popu- chaos in three-species food chains where the lations at equilibrium. However, that case is a rather population minima are bounded well away from small slice of a much broader ecological pie. 124 AQUATIC FOOD WEBS zero (Hastings and Powell 1991; McCann and migrate between different habitat compartments, Yodzis 1994a, Figure 10.1(c)). However, on a longer such as the pelagic and littoral food webs of lakes timescale, persistent chaos may transform into (Post et al. 2000). This suggests a potentially transient chaos due to sudden catastrophic popu- important role of spatial patterns on food-web lation crashes (McCann and Yodzis 1994b). stability that remains to be elucidated. Extending Together, these results demonstrate that species the size of the food-web modules, Fussmann and and community persistence is not restricted to Heber (2002) demonstrated that the frequency of equilibrium dynamics (Figure 10.1(a) and (b)), chaotic dynamics increases with the number with the potential for chaotic population dynamics of trophic levels, but decreases with other struc- (Figure 10.1(c)) that may or may not be persistent tural properties that cause higher food-web over short- and long-timescales. complexity. Interestingly, these results indicate Exploring another aspect of ecological complex- that—dependent on which process dominates in ity, McCann and Hastings (1997) investigated the a particular food web—population stability might impact of omnivory on dynamics by altering the increase or decrease with food-web complexity. network structure so that the top species of Overall, the numerical integration of nonlinear the food chain consumed both the herbivorous population dynamic models of food-web modules and basal species. They found that this type has added a great deal to our understanding of of omnivory stabilizes population dynamics by dynamical stability of multispecies systems. While either eliminating chaotic dynamics or bounding not as simple, abstract, and analytically tidy as their minima away from zero. In contrast to the community matrix analyses, these models appear to earlier community matrix work of Pimm and capture more faithfully some centrally important Lawton (1978), this suggests that omnivory is a aspects of natural ecological communities—aspects stabilizing feature of complex networks. It has to which appear very relevant to issues concerning be cautioned, however, that these results were complexity and stability. Most importantly, this obtained under the assumption that the modeling approach has clarified that non- has a higher preference for the intermediate species equilibrium population dynamics do not prevent than for the basal species, which rarely appears in species’ persistence (Hastings and Powell 1991; natural food webs (Williams and Martinez 2004a). McCann and Yodzis 1994a). Further results suggest Furthermore, the consequences of omnivory in how omnivory (McCann and Hastings 1997; Diehl a comparable simple experimental system are and Feissel 2001), weak interactions (McCann et al. highly dependent on nutrient enrichment, since 1998), and spatial processes (Post et al. 2000) may coexistence of both consumers is restricted stabilize food webs. Being restricted to species-poor to intermediate nutrient saturations (Diehl and food-web modules, however, none of these studies Feissel 2001). can adequately explore or test complexity–stability In an extended study of food-web modules, relationships in diverse networks that reflect more McCann et al. (1998) showed that weak inter- of the complexity of natural systems. While it has actions might generally dampen population oscil- been suggested that results from interaction module lations and thus stabilize dynamics in ecological studies can be generalized to more diverse, complex networks. They suggested that such weak inter- networks, that notion is based on two potentially actions among species might be the key for problematic assumptions: (1) population dynamics understanding the existence of complex com- respond similarly to independent variables in munities, if most interactions in such communities food-web modules and complex networks, and are weak. Post et al. (2000) demonstrated that a set (2) population stability coincides with community of two three-species food chains linked by the stability. It has been shown that species’ persistence same top consumer can be stabilized when the top does not coincide with population equilibria consumer can switch its preferences between (McCann and Yodzis 1994 a) and that the frequency the two intermediate species. Such preference of chaotic dynamics is influenced by food-web switching appears likely whenever consumers complexity (Fussmann and Heber 2002). MODELING FOOD-WEB DYNAMICS 125

Species persistence in complex limit cycle, and chaotic dynamics). Using such food webs food-web models, Williams and Martinez (2004c) found that persistence decreases linearly for com- More recently, the numerical integration approach munities of 15–50 species and connectance values has been extended to complex food-web models between 0.05 and 0.3, thus qualitatively replicating (e.g. Williams and Martinez 2004b,c). In this the results of May (1972) in a completely different approach, bioenergetic models are applied to model environment. Also, they found that food a diverse assemblage of species in complex trophic webs with random topology display very low networks. The structure of these complex networks persistence, while adding some network structure is defined by a set of simple topological models: in the form of the cascade model increases per- random, cascade, or niche model food webs sistence by an order of magnitude. Use of the niche (Williams and Martinez 2000). The niche model model structure improves persistence yet again has been found to successfully predict the general over the cascade model (Williams and Martinez structure of complex food webs, while the cascade 2004c). Given that this coincides with a decrease in model (Cohen et al. 1990) predicts some aspects empirical adequacy of the structural models, the and the random model predicts almost no aspects results strongly suggest that the complex network of food-web structure (Williams and Martinez structure of natural communities provides a higher 2000; Dunne et al. 2004). Dynamics based on the probability of species persistence. Interestingly, the Yodzis and Innes (1992) approach are run on each structure of the persistent food webs that remained of the three types of food-web structures, and the after species extinctions was more consistent with ensuing short- and long-term population dynamics empirical data than the original structural models, are investigated. The explicit integration of com- even in the case of the niche model (Williams and plex network structure with nonlinear, non- Martinez 2004c). These results suggest a tight equilibrium bioenergetic modeling allows for the relationship between food-web structure, dynam- exploration and testing of complexity–stability ics, and stability, since network structure influ- relationships in more plausibly complex food ences the population dynamics that in turn webs, while retaining certain aspects of modeling determine persistent food-web structures. In simplicity. The topology models, which are ana- addition to the important role of complex network lytically solvable (e.g. Camacho et al. 2002), use structure for persistence, Williams and Martinez two readily estimated parameters, species richness (2004b,c) showed that small variations in the (Brose et al. 2003a) and connectance (Martinez et al. functional response of consumption to resource 1999), and a few simple rules for assigning links. and/or predator density can have dramatic effects The dynamics model (Williams and Martinez on stabilizing the dynamics of particular species, 2004b,c), modified and generalized from Yodzis as well as overall species persistence. One import- and Innes (1992), while not analytically tractable, ant stabilizing variation in functional response is is defined by a narrow set of parameters that can the slight relaxation of consumption at low be measured relatively readily (e.g. allometric resource densities. Prior dynamical models have data), simple rules governing biomass gain and typically assumed Type II functional responses loss, and interaction dynamics that have been (Holling 1959) which lack this type of feeding studied extensively in empirical settings (e.g. relaxation, or have looked at stabilizing aspects of functional response). As described in the previous strong versions of relaxation-type responses (e.g. section, the bioenergetic dynamics model has the strong Type III) in model systems with only two additional benefit of having been studied exten- species (Murdoch and Oaten 1975; Hassell 1978; sively in a simple module setting. Yodzis and Innes 1992). The relaxation of feeding In these integrated structure–dynamics at low resource density is evocative of a variety of networks, global stability can be measured as well-documented ecological mechanisms (Skalski persistence, that is, the fraction of species with and Gilliam 2001) including predator interference, persistent dynamics (which includes equilibrium, , and refuge seeking. These types of 126 AQUATIC FOOD WEBS trophic and nontrophic behaviors allow rare or interesting patterns, such as periods of high low-biomass resource species to persist in both community diversity and complexity followed by natural and model ecosystems, increasing overall catastrophic extinction events leading to periods of community persistence. lower diversity and complexity. The community In another approach to integrating complex evolution approach has yet to be analyzed in structure and dynamics, and in contrast to most terms of complexity–stability relationships, and it prior dynamical studies, Kondoh (2003a,b, see also remains unclear what causes the extinction events Chapter 11 by Kondoh) allowed consumer pre- (Drossel et al. 2001). While the integration of evo- ferences for resources to adaptively vary in diverse lutionary and population dynamics in a food-web random and cascade model food webs. He found assembly model is intriguing, this formulation positive complexity–stability relationships in many is hampered by its extraordinarily large number of of these adaptive networks. However, the results parameters, rendering the interpretation of results are confounded due to inclusion of implausible as well as their relevance to natural systems diffi- linear nonsaturating species interactions and cult, if not impossible. internal growth terms that allow consumers to In summary, the recent extension of the persist without resources. A stability reanalysis of numerical integration approach to diverse model adaptive networks using the Williams–Martinez ecosystems with complex, ecologically plausible framework revealed that positive complexity– network structure offers an approach for stability relationships require an arbitrarily high exploring conditions that give rise to, and hin- fraction of adaptive consumers (>50%) with a der the stability of individual populations within high adaptation rate in simple cascade model food a complex ecosystem, as well as the stability of webs, and do not occur in the more empirically the whole system (Brose et al. 2003b; Williams well-corroborated niche model food webs (Brose and Martinez 2004b,c). Interestingly, many et al. 2003b). Together these results indicate that results from the integrated network structure positive complexity–stability relationships due to and dynamics models are consistent with prior adaptive foraging are unlikely to be found in local stability analyses: a generally negative natural systems (but see Kondoh 2003b). complexity–stability relationship and an increased Integrating events and assembly into stability of empirically consistent food- dynamical food-web models has shown that— web structures in comparison to random webs despite a high species-turnover—communities (Williams and Martinez 2004c). Integrating may assemble into persistent diverse networks adaptive or evolutionary processes in complex (Caldarelli et al. 1998; Drossel et al. 2001). How- community models may lead to further insights ever, the food-web network structures that evolve regarding how diverse, complex natural ecosys- are at best loosely comparable to empirical food tems may persist (Caldarelli et al. 1998; Drossel webs, and appear less complex in some fund- et al. 2001; Kondoh 2003a,b; Brose et al. 2003b). amental ways. For example, the connectances (the It remains a disadvantage in comparison to the number of potential feeding links that are actually local stability approach that resource competition realized, links/species2, as calculated using among basal species is generally not included in unweighted links) of eight example webs are very the nonequilibrium models, but it can be readily low (C ¼ 1–6%, mean of 3%, Drossel et al. 2001: added. In general, the utility of current and tables 1 and 2) and connectance decreases with future applications of these more complex species richness. In comparison, a set of 18 recent dynamical approaches will depend to a great empirical food webs have much higher con- degree on a careful balancing act between nectances (C ¼ 3–33%, mean of 14%, Dunne et al. implausible abstraction associated with fewer 2004: table 2) and display no significant relation- parameters and model clarity versus greater ship between species richness and connectance biological plausibility associated with more (see also Martinez 1992). Nevertheless, this evolut- parameters and the increased danger of model ionary food-web model gives rise to qualitatively obfuscation. MODELING FOOD-WEB DYNAMICS 127

Application to an many of those pathways may have negligible impacts on the hake–seal interaction, it cannot be Thus far, this chapter has hopefully helped to assumed a priori that scientists can ignore most of introduce readers, many of whom are interested in those pathways of influence when assessing the aquatic systems but perhaps have less of a back- interaction between the two species (Yodzis 1998). ground in mathematical models, to the world of Yodzis (1998) utilizes this more realistic food abstract models of population dynamics in simple web as the framework for studying multispecies to complex multispecies assemblages. The various dynamics determined by the nonlinear bioener- approaches to modeling have been used to explore getic consumer–resource model discussed pre- the conditions for and limitations upon stability viously (Yodzis and Innes 1992). Parameter values and persistence in idealized ecosystems. However, for this model were set for the Benguela ecosystem aspects of the models may also lend themselves to in one of three ways: using field data estimations more applied implementation, allowing assess- (e.g. population biomasses, average body mass), ment of stability of natural ecosystems in the face through allometric and energetic reasoning, and by of actual or anticipated perturbations. This issue allowing unknown parameters to vary randomly. is of particular, immediate concern in aquatic The use of some randomly varying parameters ecosystems that contain fisheries. While the use of means that probability distributions, rather than single-species models, which have dominated particular solutions, were calculated for the out- fisheries research and management, is generally come of interest: ‘‘What is the impact of reducing acknowledged as inadequate and is slowly giving fur seal biomass on the biomass of hake and other way to analysis of simple food-web modules commercial fishes?’’ In effect, Yodzis is simulating (Yodzis 2001), those modules may still be inade- an experimental press-perturbation—an experi- quate for discerning the likely outcome of pertur- ment that cannot be done in a scientific or ethical bations that occur in a diverse multispecies way on the actual ecosystem. This framework and context. Yodzis (1998, 2000) has examined this question requires two further simplifications: fundamentally important issue for both theory and Yodzis assumes local perturbations (e.g. small practice in the context of whether culling (remov- culls) and equilibrium dynamics (e.g. constancy in ing or killing) fur seals is likely to have a negative parameter values and dynamics). While we have or positive impact on the hake fishery in the already seen that there can be problems with these Benguela ecosystem. sorts of assumptions, they allow the use of the Within a fisheries context, one common view analytically tractable Jacobian matrix (used by May of Benguela food-web dynamics is that people 1972, 1973) to determine the outcome of small, harvest hake, fur seals eat hake, and therefore if long-term perturbations in a , people reduce the fur seal population there will be which is a reasonable approach to use for this more hake for people to harvest without a net particular question. Thus, the analysis combines decrease in the hake stock. However, this is elements of all three approaches to multispecies obviously an extraordinary oversimplification of dynamical modeling we have discussed, including a much more complex ecosystem. Instead, Yodzis local stability analyses of community matrices, (1998) considers a more realistic food web that nonlinear bioenergetic dynamics, and complex contains 29 taxa including phytoplankton, food-web structure. These modeling elements, in zooplankton, invertebrates, many fish taxa, and conjunction with empirical data on trophic inter- several other vertebrate taxa (modified from Field actions and various characteristics of taxa in the et al. 1991). While still a simplification of the Benguela ecosystem, form a plausible framework diversity and complexity of the system, this 29-taxa in which to simulate a cull of fur seals and its web better captures the variety of pathways that likely effects on commercial fishes in the context of can link components of the network. There are a complex ecosystem. over 28 million paths of potential influence from Using this modeling framework, Yodzis (1998) seals to hake in this food-web configuration. While found that if fur seals are culled, there is 128 AQUATIC FOOD WEBS a significant probability that two of three Conclusions commercial fishes (hake, anchovy, and We have reviewed three dynamical modeling mackerel) will have negative responses, and that approaches for studying complexity–stability across all of the commercial stocks it is more likely relationships in abstract food webs, and some that there will be a loss than a gain. Perhaps even of their central characteristics are compared in more importantly, Yodzis’ analyses (1998, 2000) Table 10.1. This type of modeling has its roots in underscore the dangers of relying on overly sim- linear, equilibrium dynamics research. Some plistic models to make predictions. One important studies continue to utilize equilibrium assump- issue is how sensitive the model is to inclusion of tions, which may be appropriate for certain kinds apparently unimportant trophic links—that is, can of questions, as discussed in the previous section. trophic links that reflect some relatively small However, since the early 1990s alternative percentage of a predator’s consumption be ignored approaches have taken advantage of increasing without affecting the outcome of the model? The computing power and advances in nonlinear answer is ‘‘no’’ for most trophic links. As the cutoff dynamics to explore population dynamics in level (of percentage of diet) for inclusion rises detail, using nonlinear and nonequilibrium above 10%, results start to vary drastically from assumptions which are more likely to reflect the outcomes seen on the complete web, thus showing types of dynamics that occur in natural systems. that ‘‘taxonomic completeness and resolution, Rather than being restricted to the binary stable/ detailed food-web structure, does make an enor- unstable outcomes that typify local stability ana- mous difference in predicting the outcomes of lyses, researchers can now explore a wider range generalized press-perturbations, such as culling of types of population and whole-system stability. fur seals in the Benguela ecosystem’’ (Yodzis 1998). This is especially the case with the extension of The interaction between hake and fur seals within nonlinear, nonequilibrium dynamical analysis to the Benguela food web is highly influenced by food webs with complex, plausible network many other species in the web (Yodzis 2000), and structure. For example, one new measure of- it is very likely that those types of influences are stability is the percentage of species in a model food found in many, if not most ecosystems. The pre- web that displays persistent dynamics (Martinez sence of these potentially strong diffuse effects in and Williams 2004c), which can be further broken complex food webs renders the use of small down into the percentages of species that display modules as a means of simplified community different types of persistent dynamics (e.g. analysis highly suspect for many ecosystems and stable equilibria, limit cycles, chaos) (Figure 10.1). many types of questions (Yodzis 2000).

Table 10.1 Summary of the stability analysis methodologies

Local stability Food-web module Complex food-web stability persistence

Interactions Linear Nonlinear Nonlinear Equilibrium assumption Yes No No Small perturbations Yes No No Chaotic dynamics allowed No Yes Yes Community size Small to diverse Small Diverse Population dynamics analyses Stable/unstable Yes Yes Community stability analyses Yes No Yes Community persistence Yes No Yes Competition among primary producers Yes No No Nutrient dynamics No No No MODELING FOOD-WEB DYNAMICS 129

Some of the promising new directions for this type and the more realistic the food-web structure, the of research include explorations into evolutionary greater the species persistence. From a purely and adaptive dynamics, nutrient dynamics, spatial topological perspective, increasing connectance heterogeneity, and nontrophic interactions. How- may buffer against the probability of cascading ever, adding new layers of ecological plausibility extinctions (Dunne et al. 2002b) partly because to models carries with it the risk of generating middle and high connectance webs have more a parameter space so large and uncontrolled that uniform link distributions (the distribution of interpretation of model outcomes is rendered an trophic links per species) than low connectance exercise in creative tale-spinning, and the links to webs, which tend to have power-law distributions empiricism that may have motivated the addition (Dunne et al. 2002a). Power-law networks appear may be rendered worthless. Any elaborations need especially vulnerable to loss of highly connected to be carefully justified, made as simple as pos- nodes (e.g. Albert et al 2000; Dunne et al. 2002b). sible, and explored and explained in detail. It should be clear that there is still an enormous Rather than being shackled just to questions of amount of work to be done to elucidate May’s whether species richness and connectance ‘‘beget’’ ‘‘devious strategies,’’ and in the face of widespread stability or instability, researchers are expanding biodiversity loss and ecosystem perturbation, there their search for the ‘‘devious strategies’’ that pro- is an increasingly compelling set of reasons to do mote different aspects of population and ecosys- that research. Food-web research has proven to be tem persistence and robustness. Regardless of the a useful vehicle for many theoretical and empirical methodology used, there are some emerging, questions in ecology, and some of the most heroic consistent results indicating that natural food webs work will be accomplished by people who are have nonrandom network structures and inter- willing to try to integrate the lessons from abstract action strength distributions that promote a high models with real-world applications (e.g. Yodzis degree of population stability and species persist- 1998). Beyond ecology, the analyses and results ence. As recently pointed out by McCann (2000), that emerge from the study of complex food webs model and empirical evidence (e.g. Berlow 1999) are likely to reverberate back outwards to influ- that many weak interactions can buffer a ence our understanding of other types of biotic few strong, unstable consumer–resource inter- and abiotic networks. actions, thus stabilizing the whole community, is consistent with MacArthur’s (1955) hypothesis Acknowledgments that having more feeding links acts to dampen large population swings. Nonrandom network Jennifer Dunne acknowledges the support from structure, regardless of interaction strength, also the following grants: NSF DEB/DBI-0074521, DBI- appears to buffer against population extinction, 0234980, and ITR-0326460. CHAPTER 11 Is biodiversity maintained by food-web complexity?—the adaptive food-web hypothesis

Michio Kondoh

properties of the food web. An example of such The complexity–stability debate a property is ‘‘population stability’’ (May 1973; Food-web complexity and population stability Goodman 1975; Pimm 1991). Theory suggests that there is statistical regularity in how population In nature, a number of species are interconnected stability changes with changes in the food-web by trophic links, resulting in a complex food-web structure (see Pimm 1991). Several components network (Warren 1989; Winemiller 1990; Martinez of population stability have been considered, 1991; Polis 1991; Goldwasser and Roughgarden including resilience (i.e. the rate at which popu- 1993). In a food web, a species affects, and is lation density returns to equilibrium after a dis- affected by, other species not only through direct turbance), feasibility (the system is at equilibrium resource–consumer interactions but also through with all existing populations), asymptotic local the indirect interactions mediated by other species stability (the system returns to equilibrium after (Schoener 1993; Menge 1994; Abrams et al. 1996). a slight disturbance), global asymptotic stability Indirect effects between species are quite sensitive (the system returns to equilibrium after any level to food-web topology and the positions of the of disturbance), persistence, and permanence (the focal species within the web (Yodzis 2000). Small system does not lose species). I briefly review these differences in food-web topology or interaction studies in the following section. strength can lead to large changes in indirect effects (Yodzis 2000). The high sensitivity of indirect effects to detailed Complexity–stability paradox food-web structure implies difficulty in relating population abundance and its variation to inter- May (1972) first tackled the issue of how popu- specific indirect effects and in finding a consistent lation stability is related to food-web structure by pattern in the strength, or even the sign, of the explicitly incorporating the effects of interspecific indirect effects in complex food webs. Does a par- indirect interactions on population dynamics ticular species within the web increase or decrease into a dynamic model. May (1972) performed when the level of a certain other species increases or linear stability analysis for a number of randomly decreases? What is the species-level consequence of generated community matrices and varied the the addition or removal of a link to/from the web? following parameters: number of species (species All those questions are unlikely to be answered for richness), the probability that a pair of species the difficulty in describing food-web structure with is connected (connectance), and the strength of a high degree of accuracy (Yodzis 2000). interspecific interactions relative to intraspecific Nevertheless, it may still be possible to predict negative interactions (interaction strength). The how food-web structure affects the grosser dynamic analysis showed that the local asymptotic stability

130 THE ADAPTIVE FOOD-WEB HYPOTHESIS 131 of the population, on average, decreased with architecture, with the expectation that a food web increasing food-web complexity. There was a with more realistic architecture should be stable threshold in food-web complexity above which the (DeAngelis 1975; Lawlor 1978; Yodzis 1981; system was unlikely to be stable. The condition for De Ruiter et al. 1995; McCann et al. 1998; Roxburgh stability is given by a(SC)1/2 > 1, where a is the and Wilson 2000; Neutel et al. 2002; Emmerson interaction strength, S is species richness (number and Raffaelli 2004). of species), and C is the connectance. The predic- The starting point of the latter idea is to recog- tion is that the relationship between food-web nize that for a given food-web complexity, complexity and population stability is negative. several possible food-web architectures exist, The negative complexity–stability relationship, some of which may be stable, while others are not. first predicted by May (1972), was supported by a An important implication of this heterogeneity is number of subsequent theoretical studies (Gilpin that the theoretical finding that a more complex 1975; Pimm and Lawton 1977, 1978; Keitt 1997; food web is less likely to be stable does not Schmitz and Booth 1997; Chen and Cohen 2001b). necessarily mean that all complex food webs are Increasing the species richness or connectance unstable. Indeed, studies that have examined how decreases feasibility (Gilpin 1975). Increasing the the incorporation of more detailed food-web length of trophic chains decreases resilience (Pimm structure alters the stability of the food web and Lawton 1977; but see Sterner et al. 1997). (DeAngelis 1975; Lawlor 1978; Yodzis 1981; Omnivory, that is, feeding at more than one De Ruiter et al. 1995; Roxburgh and Wilson 2000; trophic level, lowers local asymptotic stability Neutel et al. 2002; Emmerson and Raffaelli 2004) (Pimm and Lawton 1978). Increasing connectance have shown that a complex food web can be stable. or species richness lowers permanence and global In the following section, I focus on this approach, asymptotic stability (Pimm and Lawton 1977). which explicitly considers the effect on population Although the stability indices and complexity dynamics of realistic food-web architecture. indices used in these studies vary, they agree with each other in that a more complex food web is less Static food-web architecture and the resolution likely to persist, owing to its inherent instability. of the complexity–stability paradox The theoretical prediction that increasing com- plexity destabilizes populations contradicts the The negative complexity–stability relationship of intuition that large and complex food webs, which randomly generated food-web models does not in reality persist in nature, should be stable (Pimm necessarily predict a negative relationship in 1991). This is because such webs are unlikely to natural food webs. For one reason, within the be observed if complex food webs are unstable. whole parameter space of complexity (food- The counterintuitive prediction has catalyzed web topology, species-specific parameters, and extensive studies on this issue, resulting in a interaction-specific parameters), only a small frac- number of hypotheses to explain the maintenance tion is likely to be realized in nature (Lawlor 1978). mechanisms of complex food webs (DeAngelis The complexity–stability relationship, which is 1975; Lawlor 1978; Nunney 1980; Yodzis 1981; obtained from the whole parameter space, may not De Ruiter et al. 1995; McCann et al. 1998; Haydon represent the relationship within the ‘‘realistic’’ 2000; Roxburgh and Wilson 2000; Sole and parameter region if food-web models with Montoya 2001; Chen and Cohen 2001a; Vos et al. ‘‘realistic’’ parameter sets behave differently from 2001; Neutel et al. 2002; Emmerson and Raffaelli those with ‘‘unrealistic’’ parameter sets (DeAngelis 2004). Some studies have focused on how complex 1975; Lawlor 1978). food webs are assembled through the sequential Haydon (2000) explicitly showed that variation process of ‘‘trial (i.e. immigration or speciation) in stability within random food-web models and error (extinctions)’’ (Post and Pimm 1983; depends on food-web complexity. He analyzed a Law and Morton 1996; Wilmers et al. 2002). Other number of randomly generated community studies have focused on more detailed food-web matrices to examine how the maximum stability 132 AQUATIC FOOD WEBS changes with changing food-web complexity. The Vos et al. 2001). For example, if a predator has model analysis revealed that while more complex more potential prey species, it is likely that the food webs are, on average, less likely to be stable, interaction strength of each trophic link decreases the maximally stable subset of more complex food due to the limited ability of a predator in dealing webs is more stable than that of less complex food with multiple prey species. In the presence of webs (Haydon 2000). This result has two import- the negative correlation between connectance and ant implications. First, within food-web models interaction strength, increasing complexity may with varying complexity, the maximally stable stabilize population dynamics (May 1973; Lawlor food webs could be more complex. Second, the 1978; Nunney 1980; Vos et al. 2001). variance in stability is higher in more complex Analyses of food-web models with empirically food webs. These findings indicate the importance obtained topology and interaction strength (Yodzis of considering only a realistic subset of randomly 1981; De Ruiter et al. 1995; Roxburgh and Wilson generated food-web models in evaluating the 2000; Neutel et al. 2002; Emmerson and Raffaelli complexity–stability relationship, especially when 2004) often suggest that natural food webs the food web is highly complex. are constructed in a way that enhances stability. Studies that predict a negative relationship have Neutel et al. (2002) showed that pyramid-like been criticized, as the food-web models used in biomass distribution between trophic levels results these studies often assume an unrealistic food-web in an interaction strength that enhances food-web structure (Lawlor 1978). In May’s (1973) model, stability. Emmerson and Raffaelli (2004) pointed for example, interaction strength is randomly out that interaction strength derived from the assigned, and therefore, the models include bio- relative body sizes of prey and predator promotes logically implausible communities such as com- population stability. These studies suggest that munities without autotrophs or with extremely there may be natural processes ‘‘choosing’’ the long food chains. If the stability of food-web interaction strength that results in food-web models with unrealistic structures largely differs stability. The generality of this idea, however, is from the stability of food-web models with realistic still questionable, as these hypotheses have been structures, an analysis of randomly generated food tested for only a few select food webs. webs can lead to an incorrect prediction that is Although the predicted complexity–stability biased by the unrealistic models. Indeed, some relationship varies between studies, these studies studies that focused on the complexity–stability are similar in that they focus on the position of relationship within realistic parameter regions trophic links and interaction strength to explain have concluded that the exclusion of unrealistic the maintenance of complex food webs. The food-web models increases the stability of popu- message of these studies may be summarized as lations or even creates a positive complexity– follows: stability relationship (Lawlor 1978; Yodzis 1981). 1. A complex food web can be stable if realistic Some studies have found that when using food- interaction strengths and/or topology are con- web models with realistic structures, increasing sidered in constructing the food-web model. complexity does not destabilize populations 2. The instability of complex systems in previous (DeAngelis 1975; McCann et al. 1998). DeAngelis theoretical studies is due to averaging over many (1975) showed that if the effect of predators on food-web models with varying stability. prey is sufficiently larger than that of prey on predators (e.g. low conversion rate), increasing the complexity enhances the asymptotic stability of the Adaptation and complexity–stability equilibrium point. Other studies have proposed a relationship correlation between food-web complexity (species Flexible food-web Structure richness, connectance) and interaction strength, which influences the complexity–stability relation- Previously predicted complexity–stability relation- ship (May 1973; Lawlor 1978; Nunney 1980; ships were derived mainly from mathematical THE ADAPTIVE FOOD-WEB HYPOTHESIS 133 models that considered dynamic populations and evaluated (e.g. Pimm 1991). On the other hand, static linkage patterns (May 1972; DeAngelis 1975; if the timescale of structural changes is comparable Gilpin 1975; Pimm and Lawton 1977, 1978; Lawlor to that of population dynamics, it may influence 1978; Yodzis 1981; De Ruiter et al. 1995; Haydon the population dynamics (e.g. Holling 1959; 2000; Neutel et al. 2002). These models explicitly or Abrams 1982, 1984). implicitly assume that the interaction strength is a fixed property of a trophic link, that interaction Adaptation and food-web structure strength only changes for changing population abundances, or that a trophic link has a fixed A general through which food-web position within a web (Paine 1988). These simpli- flexibility arises is adaptation. Adaptation is a fied assumptions allow for a direct comparison of distinguishing feature of organisms and occurs at food-web models with varying structure and make several different biological levels. First, natural it possible to identify the effects of interaction selection drives population-level adaptation, or strength or food-web topology on population evolution (Stephens and Krebs 1987). A popu- stability. The theoretical contributions of this lation, which is described as a node in a food web, approach are unquestionable; however, at the same is in reality a heterogeneous unit comprising time, these assumptions have set limitations as to individuals with varying . The relative how interaction strength influences population abundances of each genotype change from stability and complexity–stability relationships. generation to generation according to their relative An important property of trophic links that reproductive contribution to the next generation. has been more or less overlooked in previous This results in phenotypic shifts at the population complexity–stability debates is flexibility. Earlier level. Second, adaptation occurs at the individual studies tended to assume static food-web struc- level. Organisms have the potential to learn tures. However, in reality, the position and through their experience and to modify their strength of a trophic link are not fixed, but behavior accordingly (Hughes 1990). Population- are spatiotemporally variable (e.g. Warren 1989; level characteristics, which are given by averaging Winemiller 1990). The linkage pattern changes across individuals within a population, can change over time as a trophic link may be activated or over time if each individual continuously adjusts inactivated. A trophic link that exists for a specific its behavior to the changing environment. If period during a life cycle (e.g. ontogenic niche the behavior or morphology that influences the shift) is disconnected during other periods. If such strength of a trophic interaction is dynamic, flexibility is a general property of a trophic link, then interaction strength should also be dynamic. then a food-web structure composed of a single set There are two types of that can of nodes and links represents only a certain time influence the strength of a trophic interaction, period and cannot serve to represent population that is, adaptive defenses by prey (Fryxell and dynamics over the long term. Lundberg 1997; Abrams 2000; Bolker et al. 2003) The effect of food-web flexibility on population and adaptive foraging by predators (Emlen 1966; stability depends on the timescale in which the MacArthur and Pianka 1966). Consider a predator food-web structure changes. If the food-web and a prey species connected by a trophic link. structure changes at a timescale that is sufficiently From the viewpoint of the prey, the trophic link is slower than the timescale of population dynamics, a path through which it loses its body mass, then the food-web structure does not change as opportunities for reproduction, or life. Therefore, population levels change. food-web structure can prey adapts to reduce the risk of predation through be approximated as static in evaluating the popu- a variety of behavioral or morphological changes lation dynamics. As Food-web flexibility makes the (antipredator defenses; Endler 1986; Lima and Dill parameters that characterize a food web change 1996), which include choosing a habitat with less over time, increasing flexibility only enlarges risk of predation, developing a morphological the parameter region where stability should be structure that prevents consumption, producing 134 AQUATIC FOOD WEBS toxins, and shifting their active time. Such anti- of a less-profitable resource lowers the net energy predator defenses weaken the strength of the gain per unit effort. Indeed, a number of examples trophic link. From the viewpoint of the predator, have been discussed in which organisms switch to on the other hand, a trophic link is a path through more valuable or abundant diets as the relative which energy and nutrients are gained. A predator abundance and/or quality of potential diets influences the strength of trophic interactions to changes (see Stephens and Krebs 1986). maximize its energetic gain from the potential Analyses of food-web models with relatively resources (Emlen 1966; MacArthur and Pianka simple structures have suggested that a foraging 1966). Some resources are extensively searched switch has a major effect on population dynamics for or captured to strengthen the trophic link, and community structure (Tansky 1978; Teramoto while others are discarded from the diet, thereby et al. 1979; Holt 1983; Sih 1984; MacNamara weakening the link. and Houston 1987; Wilson and Yoshimura 1994; Behavioral or morphological shifts potentially Abrams and Matsuda 1996; Holt and Polis influence population dynamics. A number of 1996; Sutherland 1996; McCann and Hastings 1997; theoretical studies have examined the effects of Kr˘ivan 2000; Post et al. 2000; Kondoh 2003a,b). For adaptation on simple systems such as prey– example, in a two prey–one predator system, the predator systems, one prey–two predator systems, adaptive diet choice of the predator not only and two prey–one predator systems (Tansky 1978; inhibits rapid growth of the prey but also allows Teramoto et al. 1979; Holt 1983; Abrams 1984, 1992, minor prey to rally without predation pressure. 2000; Sih 1984; Ives and Dobson 1987; MacNamara This prevents apparent competition (Holt 1977), and Houston 1987; Lima 1992; Matsuda et al. 1993, that is, a negative indirect effect between prey 1994, 1996; Wilson and Yoshimura 1994; Sutherland species sharing the same predator, leading to 1996; Abrams and Matsuda 1996; Holt and Polis species extinctions (Tansky 1978; Teramoto et al. 1997; Kr˘ivan 1998, 2000; Bolker et al. 2003). These 1979; Abrams and Matsuda 1996), and enhances studies have shown that the effects of adaptation the coexistence of intraguild prey and predators on population dynamics are diverse and, to a large (Holt and Polis 1997; McCann and Hastings 1997; extent, context dependent (see Abrams 2000; Kr˘ivan 2000) or competing prey species (Wilson Bolker et al. 2003). Inclusion of an adaptive diet and Yoshimura 1994; McCann and Hastings 1997). shift either stabilizes or destabilizes prey–predator More recent studies have examined how food- interactions and either increases or decreases the web flexibility arising from the adaptive foraging amplitude of population cycles. Consequences switch by the predator alters the dynamics of more depend on the parameters affected by the focal complex multi prey–multi predator systems trait (i.e. pleiotropic effects), relative speed of (Pelletier 2000; Kondoh 2003a,b). These studies adaptation to speed of population dynamics, the have revealed that the stability of food webs shape of the functional response, and the con- comprised of adaptive foragers (i.e. adaptive food straints under which a trait changes. webs) responds to increasing complexity in a way that is completely different from that of food webs without adaptive foragers. More specifically, an Foraging adaptations and complexity–stability adaptive foraging switch enhances the persistence relationships of complex food webs (Pelletier 2000; Kondoh Among the various foraging strategies for a pre- 2003a,b) and even generates a positive complexity– dator to maximize its energetic gains, one possible stability relationship (adaptive food-web hypo- strategy is to consume only diets of higher quality thesis; Kondoh 2003a,b). or quantity from a set of nutritionally substitu- In the following sections, I present dynamic table diets (a foraging switch or foraging shift; food-web models after Kondoh (2003a,b) to demon- Stephens and Krebs 1987). This is a natural strate the effects of an adaptive foraging switch on consequence if different strategies are required to population dynamics and the complexity–stability find or capture different diets, because utilization relationship. The model analysis examines how the THE ADAPTIVE FOOD-WEB HYPOTHESIS 135 relationship changes with changing levels of I assumed that fraction F of the nonbasal species foraging adaptation. Species richness and con- represents adaptive foragers, which allocate more nectance are used as indices of food-web com- effort to a prey species that offers higher gain per plexity. Population stability is measured by unit effort (i.e. product of density, Xi, and foraging population persistence, which is defined as the efficiency, fij). The dynamic equation for this probability that a randomly chosen species goes simple decisionmaking rule should meet the extinct within a given time period (Kondoh 2003a). following conditions: (1) more effort is allocated

to a prey species that is more abundant (larger Xj)

Model and results or has a higher predation efficiency (higher fji), (2) the total predation effort of a prey species Model description P ( k[sp:i’s resource aikÞ is kept constant over time, A food web is comprised of [pN] basal species and (3) the foraging effort is within the limited range of

[(1 p)N] nonbasal species, where N is the number 0 aij 1. For adaptive dynamics I use the equation of species and p is the fraction of basal species. For that is used in my previous studies (Appendix). simplicity, I consider a case where half of the Each adaptive consumer is assigned with a species are basal species (p ¼ 1/2). The basal spe- constant adaptation rate, Gi (Kondoh 2003a). The cies persist on their own, while the other species remaining fraction (1 F) of species in the web rely energetically on other species (either basal or does not have the ability to make an adaptive nonbasal species). Nonbasal species consume at foraging switch. These predators allocate their least one resource species, which is randomly foraging effort equally to all potential prey species chosen from the web. Basal species are never [aij ¼ constant ¼ 1/(number of potential consumer connected to each other by a trophic link. All other species of species i)]. pairs are connected with the connection prob- ability C. If a link connects a basal species with a Dynamic structure of an adaptive food web nonbasal species, the latter species always eats the former. The direction of a link between nonbasal The average number of potential prey per predator species is randomly determined (either prey or increases with increasing food-web complexity predator with the same probability of 0.5). Note (potential connectance and species richness). How that the expected connectance of a food web with the prey number changes with changing food-web species richness N and connection probability C complexity depends on the ability of the predator to is given by [(1 p) þ C{(N 3)/2 p(pN 2)}]/N. forage adaptively (Figure 11.1; see Kondoh 2003a,b). The behavior of the adaptive food web is given A nonadaptive forager allocates its foraging effort by combining population dynamics and adaptive to all potential prey species. Therefore, the expected dynamics (see Appendix for details). number of prey species utilized by a nonadaptive I use the Lotka–Volterra model for population predator (defined by a prey to which a certain > 13 dynamics (Appendix). The per-capita predation level of effort, aij 10 , is allocated) is given by rate of predator i on prey j is determined by [1 þ {C (N 2) (N þ Np 1)/2 (N 1)}]. This is an interaction-specific foraging efficiency (fij) and the increasing function of C and N, suggesting that the foraging effort of the predator (aij), aij ¼ fij aij. prey number per predator increases with increasing A predator specifies (nonbasal species) has a fixed food-web complexity. In contrast, an adaptive for- amount of ‘‘foragingP effort (finding effort or ager may consume only a fraction of the potential capturing effort)’’ ( j[sp:i’s resource aij ¼ constant ¼ 1Þ prey species, as prey species of low quality or and can increase the predation rate on a prey quantity are discarded from the actual diet. Indeed, species by allocating its foraging effort (aij) to that in food webs comprised of adaptive foragers, the prey species. An increase in foraging effort to one number of prey species per adaptive predator is prey species is associated with a decrease in the skewed to smaller numbers. total foraging effort allocated to all other potential A food-web-level consequence of an adaptive prey species. foraging switch is a reduction in the number of 136 AQUATIC FOOD WEBS

Adaptive food web Non adaptive food web Connectance 1.0 1.0 0.4 C = 0.1 F =0 0.8 0.8 0.3 0.4 0.25 0.6 0.6 0.5 0.7 0.2 N =6 0.4 0.4 0.75 0.2 1.0 0.2 0.1 1.0 0 0 0 0246810 0246810 0 0.1 0.2 0.3 0.4 1.0 1.0 0.4 0.8 0.8 0.3 0.6 0.6 0.2 0.4 0.4 N =8

Frequency 0.1 0.2 0.2 0 0 0

0246810 0246810 Snapshot connectance 0 0.1 0.2 0.3 0.4 1.0 1.0 0.4 0.8 0.8 0.3 0.6 0.6 0.2 0.4 0.4 N =10 0.2 0.2 0.1 0 0 0 0246810 0246810 0 0.1 0.2 0.3 0.4 Prey number Potential connectance

Figure 11.1 The effect of foraging adaptation on food-web structure. Number of prey per predator in a food web with varying food-web complexity (N ¼ 6, 8, 10 and C ¼ 0.0–1.0) in the presence or absence of adaptive forager. Snapshot connectance in a food web with varying species richness (N ¼ 6, 8, 10) and potential connectance (C ¼ 0.0–1.0) for varying fraction (F ¼ 0, 0.25, 0.5, 0.75, 1) of adaptive forager. p ¼ 0.5, G ¼ 0.25. The dotted line represents (potential connectance) ¼ (snapshot connectance).

active links and realized connectance, since an Long-term (potential) connectance adaptive forager tends to use only a small fraction of the potential prey species. We can define ‘‘snapshot connectance’’ as the proportion of active > 13 trophic links (aij 10 ). Snapshot connectance is given as a function of species richness, potential Connectance Snapshot connectance connectance, and a fraction of adaptive foragers (Figure 11.1). It increases with increasing complex- ity (potential connectance and species richness), Observation time (T) reaching an asymptote at high complexity levels. Figure 11.2 The ‘‘observed’’ connectance described as a cumulative In general, snapshot connectance decreases with function of observation time (T ). Snapshot connectance and potential an increase in the fraction of adaptive foragers. connectance are given as the connectance at T ¼ 0 and infinity, In an adaptive food web, a trophic link that is respectively. inactive at a given time point may be activated at another point, triggered by population fluctua- food-web topology makes connectance timescale tions or changes in parameters such as foraging dependent (Kondoh submitted). If the number of efficiency (Kondoh 2003a; Kondoh submitted). trophic links is evaluated in a short time period, The number of active links is therefore described then ‘‘observed’’ connectance should be identical to not as a timescale-independent constant but as a snapshot connectance. The number of active links cumulative function of observation time (Kondoh increases with an increased timescale of observa- submitted; Figure 11.2). This temporally varying tion, as a link is more likely to be activated during THE ADAPTIVE FOOD-WEB HYPOTHESIS 137 that time period, and the connectance evaluated to a resource decreases with increasing species from an infinitely long observation (referred to as richness or connection probability as the expected ‘‘long-term connectance’’ hereafter) should be iden- number of resource species increases. The model tical to potential connectance (Kondoh submitted). analysis shows that increasing species richness (N) or connection probability (C) always decreases Population dynamics in an adaptive food web population persistence (Figure 11.3). A population is more likely to go extinct in a web with higher The effects of foraging adaptations on the relation- C or N when parameter values change over time. ship between complexity (potential connectance In the presence of foraging adaptation, a com- and species richness) and population stability pletely different complexity–stability relationship are evaluated by comparing the relationships emerges (Figure 11.3). When the predator species for a varying fraction of adaptive foragers (F) and are all adaptive (F ¼ 1) and their adaptation rate is adaptation rate (G). I used population persistence sufficiently high (G ¼ 0.25), population persistence (Pp), the probability that a species randomly increases with increasing species richness or con- chosen from an N-species community does nection probability (Kondoh 2003a,b). The positive not become extinct within a given time period, as relationships become less clear with a decreasing an index of population stability. Population 1/N fraction of adaptive foragers (F) or decreasing persistence is given by Pp(C, N) ¼ Pc(C, N) , adaptation rate (G), confirming the role of adaptive where Pc is the probability that no species becomes foragers (Kondoh 2003a,b). extinct within the time period and is directly evaluated by measuring the proportion of food- web models where no species are lost within Adaptive food-web hypothesis a time period (t ¼ 105) in an assemblage of 10,000 Relationship between species richness randomly generated models. and population persistence The relationship in the absence of adaptive foraging is examined by setting F ¼ 0, G ¼ 0 The effect of changing species richness on popu- and aij ¼ 1/(number of potential prey species of lation stability depends on the fraction of adaptive predator i). The average foraging effort allocated foragers and their adaptation rate (Kondoh 2003a,b).

1.0 1.0 1.0 1.0

0.9 0.9 0.9 0.9

) 0.8 0.8 0.8 0.8 Pp

0.7 F =1 0.7 F = 0.75 0.7 F = 0.5 0.7 F = 0.25 G = 0.25 G = 0.25 G = 0.25 G = 0.25 0.6 0.6 0.6 0.6 0.05 0.15 0.25 0.35 0.05 0.15 0.25 0.35 0.05 0.15 0.25 0.35 0.05 0.15 0.25 0.35 1.0 1.0 1.0

0.9 0.8 0.8 N =2 0.6 0.6 4 Population persistence ( 0.8 0.4 0.4 6 0.7 F =1 0.2 F =1 0.2 F =0 8 G = 0.025 G = 0.0025 G =0 10 0.6 0.0 0.0 0.05 0.15 0.25 0.35 0.05 0.15 0.25 0.35 0.05 0.15 0.25 0.35 Potential connectance

Figure 11.3 The relationship between potential connectance and population persistence in food webs with varying fraction of adaptive foragers (F ¼ 0, 0.25, 0.5, 0.75, 1) and adaptation rates (G ¼ 0.25, 0.025, 0.0025). 138 AQUATIC FOOD WEBS

When there are only a few adaptive foragers or The relationship between potential connectance when their switching ability is low (i.e. low and population persistence depends on the fraction adaptation rates), population persistence tends to of adaptive foragers and their adaptation rate decrease with increasing species richness (Kondoh (Kondoh 2003a,b). In the absence of adaptation, 2003a,b). This implies that, for a given potential increasing the potential connectance tends to lower connectance, a species is more likely to become population persistence (Kondoh 2003a,b), implying extinct in a larger food web. In contrast, when the that a food web comprised of species with a wider fraction of adaptive foragers is sufficiently high resource range is more likely to lose species. In and the ability of foraging adaptation is high, contrast, in the presence of adaptation, the relation- the relationship is positive (Kondoh 2003a,b). ship is positive (Kondoh 2003a,b). This suggests A population is less likely to go extinct in a larger that a species is less likely to go extinct when the food web. Foraging adaptation inverts the negative species in the food web have, on average, a wider species richness–stability relationship. range of prey species. As potential connectance The positive relationship that emerges in the represents connectance that is observed by infinitely presence of adaptation (Kondoh 2003a,b) does not long observation, the result can be interpreted as a necessarily mean that a larger food web is less relationship between long-term connectance and likely to lose a species (i.e. higher community population persistence (Kondoh submitted). persistence of a larger food web). Even if popu- For a given fraction of adaptive foragers and lation persistence is independent of species rich- a given adaptation rate, snapshot connectance ness, a food web with more species is more always increases with increasing potential con- susceptible to species losses simply because it has nectance (Kondoh 2003a; Figure 11.1). Since the more species. The relationship between species relationship between potential connectance and richness and community persistence, therefore, snapshot connectance is positive, the sign of the depends on the relative intensities of the opposite relationship between snapshot connectance and effects of increasing species richness on commu- population persistence should be the same as the nity persistence. Community persistence, that is, relationship between potential connectance and the probability that a food web will not lose any population persistence. The relationship is positive species, will be higher in a community with more in the presence of adaptive foragers (Kondoh species only if the stabilizing effect that arises from submitted), while in the absence of adaptation, the foraging adaptation is stronger than the negative relationship is negative (Kondoh submitted). Note effect of having more species. Indeed, in the pre- that the positive relationship is stronger than the sent model, community persistence decreases with relationship between potential connectance and increasing species richness even when adaptive population persistence, as snapshot connectance is foragers are present, suggesting that the negative a saturating function of potential connectance effect is stronger than the positive effect. (Kondoh submitted). The applicability of the adaptive food-web hypothesis to a wider range of food-web models is Relationship between connectance and still an open question, because only a few studies population persistence have attempted to reexamine the hypothesis in In the presence of adaptive foragers, connectance other food-web models. In my own tests, a similar is a timescale-dependent variable, as the topology pattern has been obtained for a variety of food-web of an adaptive food web changes over time models with different assumptions. The models in (Kondoh submitted). This implies that the which the hypothesis holds include models using connectance–stability relationship is also timescale food-web topology of random (Kondoh 2003a,b), dependent and should, therefore, be separately cascade (Kondoh 2003a,b), and niche models defined for each timescale (Kondoh submitted). (unpublished data; but see Brose et al. 2003; where The connectance–stability relationship for each the adaptive food-web hypothesis does not hold as timescale is summarized as follows. another stability index is used), models using THE ADAPTIVE FOOD-WEB HYPOTHESIS 139 trophic interactions of Holling’s Type I and II Another mechanism that creates a positive (Kondoh 2003b; unpublished data) functional relationship in an adaptive food web is related to responses, and models with different fractions of predation pressure. Without adaptive foraging, species with a positive intrinsic growth rate prey species that share the same predator are (Kondoh 2003a,b). Thus, the hypothesis appears to unlikely to coexist as a result of predator-mediated be robust for a wide range of food-web models. negative indirect effects between prey species (apparent competition; Holt 1977). This implies that if increasing complexity increases the number Mechanisms creating the positive relationship of prey shared by the same predator, then it is It remains difficult to provide a formal explanation likely to increase the probability of species extinc- of how a foraging switch turns the previously tion. In contrast, in the presence of foraging negative complexity–stability relationship into a adaptation, increasing complexity may prevent positive relationship, as the present analysis is extinction of a minority species caused by preda- based on numerical calculations. However, it is tion pressure. The explanation for this is as follows. possible to speculate on possible mechanisms by Consider a prey species, whose abundance is low, making a gross approximation of how the extinc- and an adaptive predator. In a food web with N tion process is affected by food-web complexity in species and potential connectance C, the expected an adaptive food web. number of potential predator species for a prey Consider a population whose density becomes species is given by [(N B 1)C/2]. If the predator so low that further reductions in the population does not have an alternative resource, it will con- level will lead to extinction of the population. sume the prey irrespective of its abundance. This In this case, there are only two reasons for the implies that a prey population whose abundance is species to become extinct—a low growth rate low is still susceptible to predation pressure. Such owing to low resource consumption and a high trophic interactions destabilize prey–predator sys- mortality due to overconsumption by the predator. tems. If there is an alternative resource, a predator In the presence of adaptive food choice, increasing is likely to shift to another prey and, therefore, the complexity can enhance population persistence by focal prey species is unlikely to be consumed. either increasing resource availability or preventing Therefore, the probability that a population with overconsumption. low density is not consumed can be grossly An increase in food-web complexity may approximated by the probability that a predator enhance resource availability for an adaptive for- has an alternative prey species. The probability ager. Consider a predator species in a food web is given by [1 (1 C/2)(N 2)][(N 1)C/2], which with species richness N and connectance C. The increases with increasing food-web complexity expected number of potential prey species of the when the complexity is sufficiently high, suggest- predator is [1 þ (N 2)C/2], which increases with ing that food-web complexity potentially prevents either increasing C or increasing N. In the presence species extinction due to overconsumption. of adaptive diet choice, most trophic links are Further studies are required to understand the inactive and the abundances of a potential prey mechanism that creates the positive complexity– and a potential predator are less correlated with stability relationship in the adaptive food web, each other. If the predator is capable of switching as the present section only provides a few poss- to the most efficient resource, a larger number of ibilities. There may be other mechanisms. For potential prey will increase the probability that the example, complexity and foraging adaptation may consumer finds a more valuable or more abundant alter the dynamic trajectory of the system and resource. This increases the growth rate of the change the probability that the system will predator and thus decreases its extinction prob- approach its limits. In some food-web topologies, ability. The stabilizing effect of having more a diet switch can stabilize population dynamics, potential prey has been discussed in previous whereas other studies suggest that adaptive studies (see Petchey 2000). dynamics may destabilize population dynamics 140 AQUATIC FOOD WEBS and increase the extinction probability. It is thus Species richness necessary to extend the studies in which simple (food-web complexity) food webs are used, to more complex systems. + Maintenance Food-web complexity and an Enhancement

Although a complexity–stability relationship is a + community-level phenomenon, it is also possible Population persistence to regard this relationship as representative of an (Community service) interbiological level interaction between a com- Figure 11.4 The positive feedback between food-web complexity munity and a population. A food web is a bio- and population persistence. Species richness enhances population logical unit that determines the persistence of the persistence through the complexity–adaptation interactive effect, species within the web. The population persistence while increased population persistence maintains the species can therefore be viewed as an index that represents richness. the ability of an ecosystem to maintain a popu- lation (see Dunne et al. 2002 for a similar argu- food web, a species is less likely to be lost owing to ment). High species persistence indicates that a the high population persistence that arises from population receives a larger ‘‘service’’ from the the interactive effects of an adaptive foraging community. Applying this idea to the present switch and high food-web complexity. However, study, the adaptive food-web hypothesis suggests once a species is lost, the ecosystem service, which that in the presence of adaptive foraging, a more is measured in terms of population persistence, complex food web provides a larger community decreases, thereby enhancing further species service, supporting the species in the population. extinctions. This feedback lowers species richness The positive relationship between species and population persistence at the same time, richness and population persistence implies that resulting in cascading extinctions. A complex and community service is related to maintaining adaptive food web may be more fragile when it is species and species richness in an interesting way. under strong pressure, lowering species richness. On the one hand, as suggested by the positive complexity–stability relationship, population per- Conclusion sistence increases with increasing species richness (i.e. community service is greater in a larger food The adaptive food-web hypothesis suggests that web). On the other hand, species richness will adaptation, a distinguishing feature of organisms be higher when community service, which is at individual and population levels, potentially measured by population persistence, is higher. resolves the community-level paradox of how a These two interactions taken together imply posi- complex food web persists in nature. An adaptive tive feedback between species richness and popu- foraging switch provides a food web with flex- lation persistence (Figure 11.4; Kondoh 2003a). ibility, which enhances population persistence. Species richness is more likely to be maintained in Furthermore, this stabilizing effect is stronger in a a food web with more species, while high species more complex food web, resulting in a positive richness contributes to high population per- complexity–stability relationship. sistence. This implies that a community with high Changes in interaction strength caused by species richness is self-sustaining (Kondoh 2003a). adaptive responses to temporal population variab- This self- of a complex food web, ility are a realistic feature of trophic interactions, however, does not necessarily mean that a more and have been well studied in behavioral biology. complex food web will be less susceptible to loss of In this sense, the adaptive food-web hypothesis is species. This is because positive feedback may in line with previous studies that have asserted operate in reverse. In an adaptive and complex that a ‘‘realistic’’ food-web structure is the key to THE ADAPTIVE FOOD-WEB HYPOTHESIS 141 understanding the persistence of complex food systems in which a predator performs well. More webs (Yodzis 1981; De Ruiter et al. 1995; Roxburgh specifically, a trophic system will be less stable and Wilson 2000; Neutel et al. 2002; Emmerson with more timid predators, such as invading and Raffaelli 2004). The adaptive food-web hypo- predators, especially when the system is complex. thesis, however, differs from previous hypotheses The prediction that adaptive foraging alters the in three important ways. complexity–stability relationship suggests that the First, the adaptive food-web hypothesis predicts relationship might be affected by the relative body a positive relationship between food-web com- sizes, mobility, spatial distribution, and generation plexity and population stability, whereas most time of prey and predator, as these factors all previous studies have either shown that a complex potentially influence the predator’s capability of food web can be as stable as a less complex food diet choice. A larger predator is less likely to be web or have not dealt with the complexity– able to choose exclusively the small prey indi- stability relationship (Yodzis 1981; De Ruiter et al. vidual even if most profitable; a predator with 1995; Roxburgh and Wilson 2000; Neutel et al. high mobility is more likely to choose the most 2002; Emmerson and Raffaelli 2004, but see profit prey; even if the prey’s body size is much DeAngelis 1975; Nunney, 1980). The positive smaller than the predator’s body size, prey’s relationship predicted by the present hypothesis clustered distribution may allow for predator’s has the potential to drive a paradigm shift in the adaptive diet choice; generation time should issue of biodiversity maintenance, as it leads to a influence the relative timescale of population completely novel view that complex food webs are dynamics to adaptive dynamics. Considering these self-sustaining and supported by positive feedback effects through adaptation, it might be an inter- between complexity and population stability. esting approach in food-web study to look at how Second, the adaptive food-web hypothesis shows these factors affect population dynamics. Further, the importance of dynamic food-web structure it is notable that relative prey–predator body size, in biodiversity maintenance. The view that bio- mobility, and spatial distribution seem to be very diversity is maintained in a food web in which the much different between aquatic and terrestrial structure is changing continuously implies that systems. Predators tend to be larger than prey in the persistence of species should be considered aquatic system, while this pattern is less clear in separately from the persistence of the food web. The terrestrial systems; terrestrial systems are more persistence of species is supported by the flexible likely to form clustered distribution of less mobile and nonpersistent structure of the food web, organisms. This may imply that the habitat type whereas a species is less persistent in a nonadaptive may affect the complexity–stability relationship of food web with a persistent architecture. food webs. Third, the adaptive food-web hypothesis indi- The food-web models, with which the adaptive cates that evolutionary history or individual-level food-web hypothesis is derived, are very much experience plays a central role in determining simplified in many ways. A simplifying assump- population dynamics and thus the maintenance of tion is the low interspecific heterogeneity within a a community. The adaptive foraging switch of a web—there are only two types of behavior, non- predator in general requires information regarding adaptive foraging and adaptive foraging. While the relative quantity and/or quality of potential a nonadaptive forager does not switch diets at prey species so that it can evaluate which potential all, adaptive species are able to discriminate prey is energetically more valuable (Stephens and every potential resource species and their adaptive Krebs 1987). Lack of such information would result switches take place at the same rate. This is in incorrect responses by the predator to environ- however unlikely in real nature. Adaptive rates mental changes. Therefore, a trophic system in should be heterogeneous within a web. A con- which a predator performs poorly in terms of sumer with high capability of behavioral flexibility recognizing prey species or evaluating their rela- or short generation time would have a higher tive value is likely to behave differently from adaptive rate. An adaptive consumer may well 142 AQUATIC FOOD WEBS

discriminate a certain type of species, while being and eij is the conversion efficiency of predator i unable to discriminate another type of species. eating prey j. Ri and Fij are given by: How does such heterogeneity affect the adaptive X (r s X Þ (for basal speciesÞ food-web hypothesis? Inclusion of the heterogeneity R ¼ i i i i i m X (for nonbasal speciesÞ would be a next important step to generalize the i i : present hypothesis. ð11 2Þ and Acknowledgment Fij ¼ aijXiXj, (11:3Þ I thank Jennifer Dunne for her valuable comment on an earlier manuscript. This research is partly where ri is the intrinsic growth rate of the basal supported by Japan Society for the Promotion of species, si is the self-regulation term for the basal Science Research Fellowship for Young Scientists species, mi is the mortality rate of the non and the Grant for the Biodiversity Research of the basal species, and aij is foraging efficiency, which twenty-first Century COE (A14). is affected by the behavior of either prey j or predator i. The dynamics of foraging effort is described by: Appendix 0 1 X ... daij @ A The dynamics of biomass Xi of species i (1, , N)is ¼ G a f X a f X , dt i ij ij j ik ik k described by: k[sp:i ’s resource

XN XN (11:4Þ dXi ¼ Ri þ eijFij Fj, i dt j[sp:i ’s resources j[sp:i ’s consumers where Gi is the adaptation’ rate of consumer i. (11:1Þ In the analysis, fij is set to a random value between 0.0 and 1.0; aij(0) ¼ 1/[number of potential prey for where Ri is the intrinsic population growth rate, predator i]. Each numerical calculation starts with the second and third terms are the biomass gain the following initial condition: Xi ¼ 0.0–0.1, ri ¼ 0.1, and loss due to trophic interactions, respectively, si ¼ 0.1, e ¼ 0.15, mi ¼ 0.01. CHAPTER 12 Climate forcing, food web structure, and community dynamics in pelagic marine ecosystems

L. Ciannelli, D. Ø. Hjermann, P. Lehodey, G. Ottersen, J. T. Duffy-Anderson, and N. C. Stenseth

Introduction unavailable to the investigational tools of food web dynamics, specifically in situ experimental The study of food webs has historically focused on perturbations (Paine 1980; Raffaelli 2000; but their internal properties and structures (e.g. see Coale et al. 1996; Boyd et al. 2000). To date, diversity, number of trophic links, connectance) studies on population fluctuations and climate (Steele 1974; Pimm 1982; Cohen et al. 1990). forcing in marine ecosystems have been primar- A major advance of these investigations has been ily descriptive in nature, and there have been the recognition that structure and function, within few attempts to link the external forcing of cli- a food web, are related to the dynamic properties mate with the internal forcing of food web inter- of the system (Pimm 1982). Studies that have actions (e.g. Hunt et al. 2002; Hjermann et al. focused on community dynamics have done so 2004). From theoretical (May 1973) as well as with respect to internal forcing (e.g. competition, empirical studies in terrestrial ecology (Stenseth predation, interaction strength, and energy et al. 1997; Lima et al. 2002) we know that the transfer; May 1973), and have lead to important relative strength of ecological interactions among advances in community ecology, particularly in different species can mediate the effect of external the complex field of community stability (Hasting forcing. It follows that, different communities, or 1988). During the last two decades, there has been different stages of the same community, can increasing recognition that external forcing—either have diverging responses to a similar external anthropogenic (Parsons 1996; Jackson et al. 2001; perturbation. In a marine context, such pheno- Verity et al. 2002) or environmental (McGowan et al. menon was clearly perceived in the Gulf of 1998; Stenseth et al. 2002; Chavez et al. 2003)—can Alaska, where a relatively small increase (about profoundly impact entire communities, causing a 2C) in sea surface temperature (SST) during the rearrangement of their internal structure (Pauly mid-1970s co-occurred with a dramatic change of et al. 1998; Anderson and Piatt 1999; Steele and the species composition throughout the region Schumacher 2000) and a deviation from their (Anderson and Piatt 1999). However, in 1989 an original succession (Odum 1985; Schindler 1985). apparent shift of the Gulf of Alaska to pre-1970s This phenomenon has mostly been documented climatic conditions did not result in an analogous in marine ecosystems (e.g. Francis et al. 1998; return of the community to the pre-1970s state Parsons and Lear 2001; Choi et al. 2004). (Mueter and Norcross 2000; Benson and Trites The susceptibility of large marine ecosystems 2002). An even clearer example of uneven com- to change makes them ideal to study the effect munity responses following the rise and fall of an of external forcing on community dynamics. external perturbation is the lack of cod recovery However, their expansive nature makes them

143 144 AQUATIC FOOD WEBS from the Coast of Newfoundland and Labrador in We emphasize the mediating role of food web spite of the 1992 fishing moratorium (Parsons and structure (i.e. trophic interactions) between Lear 2001). external climate forcing and species dynamics. In this chapter, we review how marine This we do by summarizing studies from three pelagic communities respond to climate forcing. different and well-monitored marine pelagic

(a)

Barents Sea

Gulf of Alaska

Tropical Pacific

(b) Alaska current 10 °C Sub-Arctic gyre 40N Oyashio 20 °C California current Subtropical ° gyre Kuroshio 26 C 20N

NEC Equatorial countercurrent 0 SEC

–20S ° 29 C Subtropical Humboldt East-Australia gyre current current 26 °C –40S 20 °C

Average SST (°C) 5 10 15 20 25 30

Figure 12.1 (a) Location of the three pelagic ecosystems reviewed in the present study. Also shown are the detailed maps of each system. (b) Tropical Pacific (TP). (NEC ¼ North Equatorial Current, SEC ¼ South Equatorial Current.) See plate 9. (c) Gulf of Alaska (GOA). (ACC ¼ Alaska Coastal Current.) (d) Barents Sea (BS). CLIMATE FORCING, FOOD WEB STRUCTURE, COMMUNITY DYNAMICS 145

(c) –160°00' –158°00' –156°00' –154°00' –152°00' –150°00'

58°00' Spawning 58°00' area Bering Sea Shelikof Strait sland

Kodiak I eninsula ACC Stream Alaska P

° 56 00' Alaskan ° ° ° ° 56 00' Main 160 W 150 W 140 W juvenile 60°N habitat

° Gulf of Alaska 55 N

–160°00' –158°00' –156°00' –154°00' –152°00' –150°00'

(d) 80°N

76°N

72°N

Arctic water Atlantic water 100 0 100 200 km

Map presentation: Norwegian Polar Institute

20°E 30°E40°E50°E

Figure 12.1 (Continued) 146 AQUATIC FOOD WEBS ecosystems (Figure 12.1(a)): (1) the Tropical Pacific (Figure 12.1(b); see Plate 9). Under the influence of (TP); (2) the western Gulf of Alaska (GOA), and (3) the trade blowing from east to west, the the Barents Sea (BS). These communities are is transported along the same strongly impacted by climate (Anderson and Piatt direction (north and south equatorial currents: 1999; Hamre 2003; Lehodey et al. 1997, respec- NEC and SEC). During transport, surface water is tively), but have also fundamental differences in warmed up and creates a warm pool with a thick the way they respond to its forcing. In the GOA layer (about 100 m) of water above 29C on the and (particularly) in the BS systems, food web western side of the oceanic basin. The warm pool interactions play a major role in determining the plays a key role in the development of El Nin˜o fate of their communities, while in the TP trophic events (McPhaden and Picaut 1990). In the eastern forcing plays a minor role compared to the direct and central Pacific, this dynamic creates an equa- effects of climate. We suggest that such differences torial divergence with an upwelling of deep are to a large extent the result of dissimilar and relatively cold water (the ‘‘cold tongue’’) and food web structures among the three pelagic eco- a deepening from east to west. The systems. general east–west surface water transport is In this chapter we describe the physics, the counterbalanced by the north and south equatorial climate forcing, and the food web structure of countercurrents (NECC and SECC), the equatorial the investigated systems. We then examine their undercurrent (EUC) and the retroflexion currents community dynamics in relation with the food that constitute the western boundaries (Kuroshio web structures and climate forcing. The chapter and east Australia currents) of the northern and ends with generalizations on how to link trophic southern subtropical gyres. The TP presents a weak structure and dynamics in large, pelagic, marine seasonality, except in the far western region (South ecosystems. We emphasize climate and commu- China Sea and archipelagic waters throughout nity processes occurring in the pelagic compart- Malaysia, Indonesia, and the Philippines) that is ment at the temporal scales perceivable within a largely under the influence of the seasonally period less than a human generation (10–40 years). reversing monsoon winds. Conversely, there is We recognize that the present review is based on strong interannual variability linked to the El Nin˜o information and data that were not originally Southern Oscillation (ENSO). meant to be used in community studies, and for this reason it is unbalanced in the level of infor- Western Gulf of Alaska mation provided for each trophic assemblage. Typically, the information is available in greater The Gulf of Alaska (herein referred to as GOA) detail for species that are commercially important. includes a large portion of the sub-Arctic Pacific However, to our knowledge, this is the first domain, delimited to the north and east by the explicit attempt to link external (climate) and North American , and to the south and internal (trophic) forcing in the study of commu- west by the 50 latitude and 176 longitude, nity dynamics in large marine ecosystems (but see respectively (Figure 12.1(c)). In the present chapter Hunt et al. 2002; Hjermann et al. 2004), and should we focus on the shelf area west of 150 longitude— be most relevant to advance the knowledge of the most studied and commercially harvested structure and dynamics also in marine pelagic region of the entire GOA. The continental shelf food webs. of the GOA is narrow (10–150 km), and frequently interrupted by submerged valleys (e.g. the The and the physics ‘‘Skelikof Sea Valley’’ between Kodiak and the Semidi Islands) and archipelagos (e.g. Shumagin Tropical Pacific Islands). The offshore surface (<100 m) circulation The physical oceanography of the TP, roughly of the entire GOA is dominated by the sub-Arctic between 20N and 20S, is strongly dominated gyre, a counterclockwise circulation feature of by the zonal equatorial current systems the North Pacific. A pole-ward branch of the CLIMATE FORCING, FOOD WEB STRUCTURE, COMMUNITY DYNAMICS 147 sub-Arctic gyre, flowing along the shelf edge, maximum coverage was around 680,000 km2, while forms the Alaska current/Alaska stream. This in 1969 and 1970 it was as much as 1 million km2. current varies in width and speed along its This implies a change in ice coverage area of more course—from 300 km and 10–20 cm s1 east of 150 than 30% in only four years (Sakshaug et al. 1992). latitude to 100 km and up to 100 cm s1 in the GOA In any case, due to the inflow of warm water region (Reed and Schumacher 1987). The coastal masses from the south, the southwestern part of surface circulation pattern of the GOA is domi- the BS does not freeze even during the most severe nated by the Alaska coastal current (ACC) flowing winters. southwestward along the Alaska Peninsula. The ACC is formed by pressure gradients, in turn Climate forcing caused by freshwater discharge from the Cook Inlet area. The average speed of the ACC ranges Pacific inter-Decadal Oscillation 1 around 10–20 cm s , but its flow varies seasonally, The GOA and the TP systems are influenced by with peaks in the fall during the period of highest climate phenomena that dominate throughout the freshwater discharge (Reed and Schumaker 1987). Pacific Ocean. These are the Pacific inter-Decadal The ACC and its associated deep-water under- Oscillation (hereon referred to as Pacific Decadal currents, play an important biological role in the Oscillation, PDO; Mantua and Hare 2002), and the transport of eggs and larvae from spawning to ENSO (Stenseth et al. 2003). The PDO is defined as nursery areas of several dominant macronekton the leading principal component of the monthly species of the GOA (Kendall et al. 1996; Bailey and SST over the North Pacific region (Mantua et al. Picquelle 2002). Royer (1983) (cited in Reed and 1997). During a ‘‘warm’’ (positive) phase of the Shumacher 1987) suggested that the Norwegian PDO, SSTs are higher over the Canadian and coastal current is an analog of the ACC, having Alaskan coasts and northward winds are stronger, similar speed, seasonal variability, and biological while during a cool phase (negative) the pattern is role in the transport of cod larvae from the reversed (Figure 12.2; see also, Plate 10). The typi- spawning grounds to the juvenile nursery habitats. cal period of the PDO is over 20–30 years, hence the name. It is believed that in the last century there Barents Sea have been three phase changes of the PDO, one in 1925 (cold to warm), one in 1946 (warm to cold), The BS is an open arcto-boreal shelf-sea covering and another in 1976 (cold to warm; Mantua et al. 2 an area of about 1.4 million km (Figure 12.1(d)). 1997), with a possible recent change in 1999–2000 It is a shallow sea with an average depth of about (warm to cold) (McFarlane et al. 2000; Mantua and 230 m (Zenkevitch 1963). Three main current Hare 2002). The pattern of variability of the PDO systems flow into the Barents determining the main closely reflects that of the North Pacific (or Aleutian water masses: the Norwegian coastal current, Low) index (Trenberth and Hurrell 1994). The the Atlantic current, and the Arctic current system relationship is such that cooler than average SSTs (Loeng 1989). Although located from around 70 N occur during periods of lower than average sea to nearly 80 N, sea temperatures are substantially level pressure (SLP) over the central North Pacific, higher than in other regions at similar latitudes due and vice versa (Stenseth et al. 2003). It bears note to inflow of relatively warm Atlantic water masses that a recent study by Bond et al. (2004) indicates from the southwest. The activity and properties of that the climate of the North Pacific is not fully the inflowing Atlantic water also strongly influence explained by the PDO index and thus it has no the year-to-year variability in temperature south clear periodicity. of the oceanic Polar front (Loeng 1991; Ingvaldsen et al. 2003), as does regional heat exchange with the El Nin˜ o Southern Oscillation atmosphere (A˚ dlandsvik and Loeng 1991; Loeng et al. 1992). The ice coverage shows pronounced Fluctuations of the TP SST are related to the interannual fluctuations. During 1973–75 the annual occurrence of El Nin˜o, during which the equatorial 148 AQUATIC FOOD WEBS surface waters warm considerably from the Inter- period, of about 4–7 years. Although changes in national Date Line to the west coast of South TP SSTs may occur without a high amplitude America (Figure 12.2). Linked with El Nin˜o events change of the SO, El Nin˜o and the SO are linked so is an inverse variations in SLP at Darwin closely that the term ENSO is used to describe the (Australia) and Tahiti (South Pacific), known as the atmosphere–ocean interactions over the TP. Warm Southern Oscillation (SO). A simple index of the ENSO events are those in which both a negative SO is, therefore, often defined by the normalized SO extreme and an El Nin˜o occur together, while Tahiti minus Darwin SLP anomalies, and it has a the reverse conditions are termed La Nin˜as

(a) Pacific Decadal Oscillation Positive phase Negative phase Monthly values for the PDO index: 1900–June 2002 0.8 4 0.4 2 0.2 0 0.0 –2 –0.2 –4 –0.6 1900 1920 1940 1960 1980 2000

EI Niño Southern Oscillation Monthly values for the Niño 3.4 index: 1900–June 2002 EI Niño La Niña 4 0.8 0.4 2

0.2 0 0.0 –2 –0.2 –4 –0.6 1900 1920 1940 1960 1980 2000

(b) 58N, 170W 54N, 166W 54N, 162W 58N, 146W 56N, 138W 50N, 130W Location 44N, 126W 38N, 126W 32N, 120W 26N, 116W Jan. 1950 Jan. 1951 Jan. 1952 Jan. 1953 Jan. 1954 Jan. 1955 Jan. 1956 Jan. 1957 Jan. 1958 Jan. 1959 Jan. 1960 Jan. 1961 Jan. 1962 Jan. 1963 Jan. 1964 Jan. 1965 Jan. 1966 Jan. 1967 Jan. 1968 Jan. 1969 Jan. 1970 Jan. 1971 Jan. 1972 Jan. 1973 Jan. 1974 Jan. 1975 Jan. 1976 Jan. 1977 Jan. 1978 Jan. 1979 Jan. 1980 Jan. 1981 Jan. 1982 Jan. 1983 Jan. 1984 Jan. 1985 Jan. 1986 Jan. 1987 Jan. 1988 Jan. 1989 Jan. 1990 Jan. 1991 Jan. 1992 Jan. 1993 Jan. 1994 Jan. 1995 Jan. 1996 Jan. 1997 Jan. 1998 Jan. 1999 Jan. 2000 Jan. 2001 Jan. 2002 Jan. 2003 Date –2.50–2.00 –2.00–1.50 –1.50–1.00 –1.00–0.50 –0.50–0.00 0.00–0.50 0.50–1.00 1.00–1.50 1.50–2.00 2.00–2.50

Figure 12.2(a) SST anomalies during positive and negative phases of the PDO (upper panel) and ENSO (lower panel), and time series of the climate index. During a positive PDO phase, SST anomalies are negative in the North Central Pacific (blue area) and positive in the Alaska coastal waters (red area) and the prevailing surface currents (shown by the black arrows) are stronger in the pole-ward direction. During a positive phase of the ENSO, SST anomalies are positive in the eastern Tropical Pacific and the eastward component of the surface currents is noticeably reduced. (b) Time series of temperature anomalies in different locations of the North Pacific. The graph shows area of intense warming (yellow and red areas) associated with the ENSO propagating to the North Pacific, a phenomenon termed El Nin˜ o North condition (updated from Hollowed et al. 2001). See Plate 10. CLIMATE FORCING, FOOD WEB STRUCTURE, COMMUNITY DYNAMICS 149

(Philander 1990; Stenseth et al. 2003). Particularly (a) strong El Nin˜o events during the latter half of the twentieth century occurred in 1957–58, 1972–73, 1982–83, and 1997–98. Positive NAO index Typically, the SST pattern of the TP is under the influence of interannual SO-like periodicity (i.e. 4–7 years), while the extra-TP pattern is under the interdecadal influence of the PDO-like periodicity (Zhang et al. 1997). However, El Nin˜o/ La Nin˜a events can propagate northward and affect the North Pacific as well, including the GOA system, a Negative phenomenon known as Nin˜o North (Figure 12.2; NAO index Hollowed et al. 2001). During the latter half of twentieth century, there have been five warming events in the GOA associated with the El Nin˜o (b) NAO Index (Dec.–Mar.) 1864–2001 North: in 1957–58, 1963, 1982–83, 1993, and 1998. 6 The duration of each event was about five months, with about a year lag between a tropical El Nin˜o 4 and the Nin˜o North condition (Figure 12.2). The 2 likelihood of an El Nin˜o event to propagate to ) n the North Pacific is related to the position of the –S

n 0 Aleutian Low. Specifically, during a positive phase (L of the PDO, the increased flow of the Alaska –2 current facilitates the movement of water masses –4 from the transition to the sub-Arctic domain of the North Pacific, in turn increasing the likelihood of –6 186018701880189019001910192019301940195019601970198019902000 an El Nin˜o North event (Hollowed et al. 2001). Year It has also been reported that the likelihood of El Nin˜o (La Nin˜a) events in the TP is higher during Figure 12.3 The NAO. (a) During positive (high) phases of the NAO index the prevailing westerly winds are strengthened and moves a positive (negative) phase of the PDO (Lehodey northwards causing increased precipitation and temperatures over et al. 2003). and southeastern United State and dry anomalies in the Mediterranean region (red and blue indicate warm and cold anomalies, respectively, and yellow indicates dry conditions). Roughly North Atlantic Oscillation opposite conditions occur during the negative (low) index phase (graphs courtesy of Dr Martin Visbeck, www.ldeo.columbia.edu/ The BS is influenced by North Atlantic basin visbeck). (b) Temporal evolution of the NAO over the last 140 scale climate variability, in particular that repres- winters (index at www.cgd.ucar.edu/jhurrell/nao.html). High and ented by the North Atlantic Oscillation (NAO) low index winters are shown in red and blue, respectively (Hoerling (Figure 12.3; see also Plate 11). The NAO refers to et al. 2001). See Plate 11. a north–south alternation in atmospheric mass between the subtropical and subpolar North Atlantic. It involves out-of-phase behavior between in the direction and magnitude of the is the climatological low-pressure center near Ice- responsible for interannual and decadal fluctua- land and the high-pressure center near the tions in wintertime temperatures and the balance Azores, and a common index is defined as the of precipitation and evaporation over land on difference in winter SLP between these two loca- both sides of the (Rogers 1984; tions (Hurrell et al. 2003). A high (or positive) Hurrell 1995). The NAO has a broadband spec- NAO index is characterized by an intense Ice- trum with no significant dominant periodicities landic Low and a strong Azores High. Variability (unlike ENSO). More than 75% of the variance of 150 AQUATIC FOOD WEBS the NAO occurs at shorter than decadal time- species that are important consumers of other scales (D. B. Stephenson, web page at www. pelagic resources (e.g. micronekton), but are met.rdg.ac.uk/cag/NAO/index.html). A weak peak preyed upon, for the most part, by apex predators. in the power spectrum can, however, be detected Micronekton consist of small animals (2–20 cm) at around 8–10 years (Pozo-Vazquez et al. 2000; that can effectively swim. Typically, macronekton, Hurrell et al. 2003). Over recent decades and to a smaller extent, micronekton and apex the NAO winter index has exhibited an upward predators, include commercial fish species (, trend, corresponding to a greater pressure gra- cod, pollock, herring, and anchovies) and squids. dient between the subpolar and subtropical In the following, we summarize available informa- North Atlantic. This trend has been associated tion on food web structure, covering for the most with over half the winter surface warming part trophic interactions, and, where relevant in over the past 30 years (Gillett et al. (e.g. TP), also differences in spatial distribution 2003). among the organisms of the various trophic A positive NAO index will result in at least three assemblages. (connected) oceanic responses in the BS, reinforc- ing each other and causing both higher volume Tropical Pacific flux and higher temperature of the inflowing water (Ingvaldsen et al. 2003). The first response is The TP system has the most diverse species connected to the direct effect of the increasingly assemblage and most complex food web structure anomalous southerly winds during high NAO. among the three pelagic ecosystems included in Second, the increase in winter storms penetrating this chapter (Figure 12.4). Part of the complexity of the BS during positive NAO will give higher the TP food web is due to the existence of various Atlantic inflow to the BS. The third aspect is con- spatial compartments within the large pelagic nected to the branching of the Norwegian Atlantic ecosystem. The existence of these compartments Current (NAC) before entering the BS. Blindheim may ultimately control the relationships with et al. (2000) found that a high NAO index corres- (and accessibility to) top predators, and affect the to a narrowing of the NAC towards the community dynamics as well (Krause et al. 2003). Norwegian coast. This narrowing will result in In the vertical gradient, the community can be a reduced heat loss (Furevik 2001), and possibly divided into epipelagic (0–200 m), mesopelagic in a larger portion of the NAC going into the (200–500 m), and bathypelagic groups (<500 m), BS, although this has not been documented the last two groups being subdivided into migrant (Ingvaldsen et al. 2003). It should be noted that and non-migrant species. All these groups include the correlation between the NAO and inflow to organisms of the main taxa: fish, crustacean, and and temperature in the BS varies strongly . Of course, this is a simplified view of with time, being most pronounced in the early the system as it is difficult to establish clear vertical half of the twentieth century and over the most boundaries, which are influenced by local envir- recent decades (Dickson et al. 2000; Ottersen and onmental conditions, as well as by the life stage of Stenseth 2001). species. In addition to vertical zonation, there is a pro- Food web structure nounced east–west gradient of species composition and food web structure in the TP. Typically, there To facilitate the comparison of the three food webs, is a general decrease in biomass from the intense we have grouped the pelagic species of each system upwelling region in the eastern Pacific toward the in five trophic aggregations: primary producers, western warm pool (Vinogradov 1981). While zooplankton, micronekton, macronekton, and apex primary productivity in both the western warm predators. This grouping is primarily associated pool and the subtropical gyres is generally low, the with trophic role, rather than trophic level. equatorial upwelling zone is favorable to relatively Macronekton includes all large (>20 cm) pelagic high primary production and creates a large zonal CLIMATE FORCING, FOOD WEB STRUCTURE, COMMUNITY DYNAMICS 151

Tropical Pacific Primary producers Zooplankton Micronekton Macronekton Apex predators Seasea turtles Epipelagicepipelagic seabirdsSeabirds Meroplanktonmeroplankton marlinMarlin and Euphausidseuphausids Mesopelagicmesopelagic skipSkipjack jack yellowYellow fin Amphipodsamphipods migrant Phytoplanktonphytoplankton dolphinsDolphins Copepodscopepods Mesopelagicmesopelagic Youngyoung albaAlbacore core non-migrant and scombrids DOM Chaetognathchaetognath blueBlue shark Salpssalps Oth. sharks Picoplanktonpicoplankton Bathypelagicbathypelagic Piscivorouspiscivorous fish migrant bigBigeye eye Microzoo-microzoo- Detritusdetritus sworSwordfish dfish plankton Bathypelagicbathypelagic Largelarges squids quids Killer whales non-migrant Larvaceanslarvaceans

Gulf of Alaska Euphausiids Pollock Purpoises Eulachon Copepods Pacificcod Toothed Whales Capelin Arrowtooth flounder Chaetognath Baleen whales Phytoplankton Other Halibuth Pteropods Rockfish Juvenile Gadids Steller sealion Larvaceans Other Harbor seals Other Flathead sole

Barents Sea

Euphausiids Herring Cod Sea birds

Capelin Copepods Baleen whales Phytoplankton Polar cod Othergadids

Amphipods Juv gadids

Pinnipeds Other Shrimps Fatfish

Figure 12.4 Simplified representation of the food web for each studied system. Arrows point from the prey to the predator DOM: Dissolved Organic Matter 152 AQUATIC FOOD WEBS band, in the cold tongue area, of rich mesotrophic Fish, crustaceans (large euphausiids), and waters. However, primary productivity rates in cephalopods dominate the micronekton group, this area could be even higher as all nutrients with typical sizes in the range of 2–20 cm. These are not used by the phytoplankton. This ‘‘high- organisms, together with the gelatinous filter nutrient, low-’’ (HNLC) situation is feeders, are the main forage species of the top and due for a large part to iron limitation (Coale et al. apex predators (Figure 12.4). Many species of 1996; Behrenfeld and Kolber 1999). Another zooplankton and micronekton perform diel vertical important difference between the east and west TP migrations between layers of the water column appear in the composition of plankton (both phyto that are over 1,000 m apart. One important benefit and zooplankton). In regions where upwelling is of this evolutionary adaptation is likely a decrease intense (especially the eastern Pacific during La of the predation pressure in the upper layer during Nin˜a periods), diatoms dominate new and export daytime (e.g. Sekino and Yamamura 1999). The production, while in equatorial and oligotrophic main epipelagic planktivorous fish families are oceanic regions (warm pool, subtropical gyres) a Engraulidae (anchovies), Clupeidae (herrings, few pico- and nanoplankton groups (autotrophic sardines), Exocœtidae (flyingfish), and small bacteria of the microbial loop) dominate the phyto- Carangidae (scads), but an important component plankton community (Bidigare and Ondrusek 1996; also is represented by all juvenile stages of Landry and Kirchman 2002). large-size species (Bramidae, Coryphaenidae, Based on size, the zooplankton assemblage can Thunnidae). The oceanic anchovy (Enchrasicholinus be subdivided into micro- (20–200 mm), meso- punctifer) is a key species in the epipelagic food (0.2–2.0 mm), and macrozooplankton (2–20 mm). web of the warm pool as it grows very quickly and ciliates dominate the micro- (mature after 3–5 months) and can become very zooplankton group; however, nauplii of copepods abundant after episodic blooms of phytoplankton. are abundant in the eastern equatorial region, in Meso- and bathypelagic species include euphau- relation with more intense upwelling. Pico- and siids, deep shrimps of the Sergestidae, Peneidae, nanoplankton are consumed by microzooplankton, Caridae, and numerous fish families, Myctophidae, which remove most of the daily accumulation of Melamphaidae, Chauliodidae, Percichthyidae, biomass (Landry et al. 1995; Figure 12.4). Copepods and Stomiatidae. The micronekton consume a dominate the mesozooplankton group, as well large spectrum of prey species among which the as the entire zooplankton assemblage of the TP dominant groups are copepods, euphausiids, (Le Borgne and Rodier 1997; Roman et al. 2002). amphipods, and fish. More detailed analyses Gueredrat (1971) found that 13 species of copepods (Legand et al. 1972; Grandperrin 1975) showed that represented 80% of all copepod species in the prey composition can differ substantially between equatorial Pacific. However, meso- and macro- micronekton species, especially in relation to pre- zooplankton include a very large diversity of other dator–prey size relationships: smallest micro- organisms, such as amphipods, euphausiids, chaeto- nekton prey mainly upon copepods, medium size gnaths, and larval stages (meroplankton) of many of micronekton consumes more euphausiids, and species of molluscs, cnidaria, crustaceans, and fish. large micronekton are mostly piscivorous. Another group that has a key role in the functioning Tuna dominate the macronekton in the TP food of the pelagic food web but has been poorly studied web, although this group also includes large- is what we can name ‘‘gelatinous filter feeders.’’ size cephalopods, and sea turtles. Skipjack tuna In particular, this group includes appendicularians (Katsuwonus pelamis) is the most abundant and (a.k.a., larvaceans) and salps. Salps and larvaceans productive species of the TP and constitute the filter feed mainly on phytoplankton and detritus. fourth largest fisheries in the world (FAO 2002; Though by definition, the zooplankton described 1.9 million tons per year). Juveniles of other above is drifting in the currents, many species tropical tuna, particularly yellowfin tuna (Thunnus undertake diel vertical migrations, mainly stimulated albacares) and bigeye tuna (Thunnus obesus) are by the light intensity. frequently found together with skipjack in the CLIMATE FORCING, FOOD WEB STRUCTURE, COMMUNITY DYNAMICS 153 surface layer, especially around drifting logs (e.g. Ommastrephidae, Onychoteuthidae) than the that aggregate many epipelagic species. With these other billfishes. Blue sharks consume cephalopods well-known species, there are many other as a primary component of their diet and various scombrids (Auxis sp., Euthynnus spp., Sarda spp., locally abundant pelagic species (Strasburg 1958, Scomberomorus spp., Scomber spp.), and a large Tricas 1979). Whitetip and silky sharks are omni- variety of piscivorous fish (Gempylidae, vorous. They feed primarily on a variety of fish Carangidae, Coryphaenidae, Trichiuridae, Alepi- including small scombrids, cephalopods, and to a sauridae), and juveniles of apex predators (sharks, lesser extent, crustaceans. The whitetip sharks , swordfish, and sailfish). The largest bio- consume also a large amount of turtles (Compagno mass of the most productive species, skipjack and 1984) and occasionally stingrays and sea birds. yellowfin tuna, is in the warm waters of the Thresher, hammerhead, and mako sharks feed on Western and Central Pacific Ocean (WCPO), but various piscivorous fish including scombrids and warm currents of the Kuroshio and east Australia alepisaurids, and cephalopods. extend their distribution to 40N and 40S (roughly Marine mammals encountered in the TP system delineated by the 20C surface isotherm). Most permanent baleen whales, toothed whales, and macronekton species are typically predators of the . However, most of baleen whales are not epipelagic micronekton but many of them take permanent in the tropical pelagic food web as they advantage of the vertical migration of meso- and only migrate to tropical regions for breeding, while bathypelagic species that are more particularly they feed in polar waters during the summer. The vulnerable in the upper layer during sunset and diets of toothed whales (killer , whale, sunrise periods. and short-finned pilot whale) are mainly based on Apex predators of the TP food web include squids (the sperm whale often taking prey at adult large tuna (yellowfin T. albacares, bigeye considerable depths), and fish. Killer whales are Thunnus obesus albacore Thunnus alalunga), broad- also known to prey on large fish such as tuna and bill swordfish (Xiphias gladus), Indo-Pacific blue dolphinfish and sometime small cetaceans or ( mazara), (Makaira turtles. All the dolphins consume mesopelagic fish indica), (Tetrapturus audax), shortbill and squid. Spotted and common dolphins are spearfish (Tetrapturus angustirostris), Indo-Pacific known also to prey upon epipelagic fish like flying sailfish (Istiophorus platypterus), pelagic sharks, fish, mackerel, and schooling fish (e.g. sardines). seabirds, and marine mammals. The diets of apex Tropical seabirds feed near or above (flyingfish, predator species reflect both the faunal assemblage flying squids) the water surface, on a large variety of the component of the ecosystem that they of macroplankton and micronekton (mainly fish explore (i.e. epi, meso, bathypelagic) and their and squids), including vertically migrating species aptitude to capture prey at different periods of likely caught during sunset and sunrise periods. the day (i.e. daytime, nightime, twilight hours). They have developed a remarkably efficient All large tuna species have highly opportunistic foraging strategy associated with the presence of feeding behavior resulting in a very large spectrum subsurface predators (mostly tuna) that drive prey of prey from a few millimeters (e.g. euphausiids to the surface to prevent them from escaping to and amphipods) to several centimeters (shrimps, deep waters. Therefore, as stated by Ballance squids and fish, including their own juveniles) in and Pitman (1999), ‘‘subsurface predators would size. However, it seems that differences in vertical support the majority of tropical seabirds and behavior can be also identified through detailed would indirectly determine distribution and abund- analyses of the prey compositions: bigeye tuna ance patterns, and provide the basis for a complex accessing deeper micro- and macronekton species. community with intricate interactions and a Swordfish can also inhabit deep layers for longer predictable structure. This degree of dependence periods than most apex predators. This difference has not been found in non-tropical seabirds.’’ in vertical distributions is reflected in the diets, An example of such potential interaction is that swordfish consuming a larger proportion of squids a decrease in tuna abundance would not have 154 AQUATIC FOOD WEBS a positive effect on seabirds despite an expected generated by salinity gradients increase of forage biomass, but instead would (Sambrotto and Lorenzen 1987). have a negative effect as this forage becomes less Euphausiids, copepods, cnidarians, and chae- accessible. tognaths constitute the bulk of the zooplankton assemblage of the GOA food web. Given the paucity of feeding habits data at very low trophic Western Gulf of Alaska levels, we assume that the zooplankton species The trophic web of the GOA includes several feed mainly on phytoplankton (Figure 12.4). This is generalist (e.g. Pacific cod) and opportunistic (e.g. a common generalization in marine ecology sablefish) feeders (Figure 12.4). In addition, dif- (e.g. Mann 1993), which, nonetheless, underscores ferent species exhibit a high diet overlap, such as the complex trophic interactions within the juvenile pollock, and capelin, or the four dominant phytoplankton and zooplankton assemblage (e.g. macroneckton species (arrowtooth flouder, Pacific microbial loop). However, microbial loop organ- halibut, cod, and pollock; Yang and Nelson 2000). isms are particularly important in oligotrophic As with other systems included in our review, diet environments, such as the west TP system, and patterns can change during the species ontogeny. supposedly play a minor role in more productive In general, zooplanktivory decreases in importance marine ecosystems, such as the GOA, the BS sys- with size, while piscivory increases. Also, within tems, and the cold tongue area of the TP system. zooplanktivorous species/stages, euphausiids The most abundant and common micronekton replace copepods as the dominant prey of larger species in the GOA food web are capelin (Mallotus fish. In the GOA, as well as in the BS systems, the villosus Mu¨ ller), eulachon (Thalicthys pacificus), structure of the food web is also influenced by sandlance (Ammodytes hexapterus), juvenile gadids climate forcing, as shown by noticeable diet (including Pacific cod and walleye pollock), Pacific changes of many macronekton and apex predators sandfish (Trichodon trichodon), and pandalid over opposite climate and biological phases. shrimps (Pandalus spp.). Pacific herring (Clupea In contrast to tropical regions, primary produc- harengus pallasi) are also important, but their pre- tion in Arctic and sub-Arctic marine ecosystems sence is mainly limited to coastal waters and to the varies seasonally with most annual production northern and eastern part of the GOA. The range confined to a relatively short spring bloom. This of energy density of these micronekton species is seasonal pattern is mainly the result of water very broad (Anthony et al. 2000), as it is also the stratification and increased in the nutritional transfer to their predators. Eulachon upper water column during the spring. In the GOA have the highest energy density (7.5 kJ g1 of wet system, primary productivity varies considerably weight), followed by sand lance and herring also with locations. Parsons (1987) recognized four (6 kJ g1 of wet weight), capelin (5.3 kJ g1 of wet distinguished ecological regions: (1) the estuary weight), Pacific sandfish (5 kJ g1 of wet weight), and intertidal domain, 150 g C m2 per year; (2) the and by juvenile cod and pollock (4 kJ g1 of wet fjord domain, 200 g C m2 per year, (3) the shelf weight). Juvenile pollock feed predominantly on domain, 300 g C m2 per year, (4) the open ocean copepods (5–20%) and euphausiids (69–81%), the domain, 50 g C m2 per year. These values, parti- latter becoming more dominant in fish larger than cularly for the shelf area, are considerably higher 50–70 mm in standard length (Merati and Brodeur than those observed in the BS and similar to those 1996; Brodeur 1998). Other common, but less observed in the east TP. A number of forcing dominant prey include fish larvae, larvaceans, mechanisms can explain the high productivity in pteropods, crab larvae, and hyperid amphipods. the GOA system, including a seasonal weak Capelin diet is similar to that of juvenile pollock, upwelling (from May to September, Stabeno et al. feeding mostly on copepods (5–8%) and euphau- 2004), strong tidal currents with resulting high siids (72–90%) (Sturdevant 1999). Food habits of tidal mixing, high nutrient discharge from other micronekton species are poorly known in fresh water run off, and the presence of a strong this area; however, it is reasonable to assume that CLIMATE FORCING, FOOD WEB STRUCTURE, COMMUNITY DYNAMICS 155 most of their diet is also based on zooplankton include (26–44%, tanner crab and pagurids) species. and cephalopods ( 3–5%). Flathead sole The GOA shelf supports a rich assemblage of feed almost exclusively on shrimps (39%) and macronekton, which is the target of a large brittle stars (25%). Atka mackerel is zooplankti- industrial fishery. The majority are demersal vorous (64% copepods, 4% euphausiids), but also species (bottom oriented), such as arrowtooth feed on large jelly fish (Scyphozoa 19%). Sablefish flounder (Atherestes stomias), halibut (Hippoglossus can equally feed on a variety of prey, including stenolepis), walleye pollock (Theragra chalcogramma), pollock (11–27%), shrimps (5–11%), jellyfish Pacific cod (Gadus macrocephalus), and a variety of (9–14%), and fishery offal (5–27%). The rockfish of rockfishes (Sebastes spp.). Walleye pollock currently the GOA food web can be grouped among those constitute the second largest fisheries in the world that feed mainly on shrimps (rougheye, short- (FAO 2002, second only to the Peruvian anchoveta); spine), euphausiids (Pacific ocean perch, northern, however, the bulk of the landings comes from dusky), and squids (shortraker). the Bering Sea. In the GOA, arrowtooth flounder Apex predators of the GOA include seabirds, (A. stomias) presently dominate the macronekton , and cetaceans. Among seabirds the most assemblage, with a spawning biomass estimated at common are murres (a.k.a. common guillemot) over 1 million metric tones. (Uria aalge), black-legged kittiwakes Rissa tridactyla, The macronekton species of the GOA food a variety of cormorants (double-crested, red-faced, web can be grouped in piscivorous (arrowtooth and pelagic), horn and tufted puffins, storm- flounder, halibut), zooplanktivorous (pollock, atka petrels, murrelets, shearwaters, as well as three mackerel, some rockfish), -feeders (some species of albatross (laysan, black-footed, and rockfish, flathead sole), and generalist (sablefish short-tailed). Pinnipedia include Steller sea lions and cod) (Yang and Nelson 2000). Within these (SSLs, Eumetopias jubatus) and harbor seals (Phoca subcategories there have been noticeable diet vitulina). Cetacea include killer whales (Orcinus changes over time. For example, in recent years orca), Dall’s porpoise (Phocoenoides dalli), harbor (1996) adult walleye pollock diet was based porpoise (Phocoena phocoena), primarily on euphausiids (41–58%). Other import- (Megaptera novaeangliae), minke whale (Balenoptera ant prey include copepods (18%), juvenile pollock acutorostrata), and sperm whale (Physester macro- (10%), and shrimps (2%). However, in 1990 cephalus) (Angliss and Lodge 2002). Humpback, shrimps where more dominant in pollock diet minke, and sperm whales are transient species, (30%) while cannibalism was almost absent (1%) and are present in Alaska waters only during the (Yang and Nelson 2000). Adult Pacific cod has also feeding migration in summer. Apex predators undergone similar diet changes. Recently they feed have also undergone drastic diet changes during on benthic shrimps (20–24%), pollock (23%), crabs the last 30 years. For example, a new study (tanner crab bairdi, pagurids, 20–24%), (Sinclair and Zeppelin 2003) indicates that in and eelpouts (Zoarcidae). In contrast, during the recent times SSL fed on walleye pollock and Atka early 1980s they fed primarily on capelin, pandalid mackerel, followed by Pacific salmon and Pacific shrimps, and juvenile pollock (Yang 2004). cod. Other common prey items included arrow- Arrowtooth flounder feed predominantly on fish tooth flounder, Pacific herring, and sand lance. (52–80%), among which pollock is the most com- In contrast, prior to the 1970s, walleye pollock and mon (16–53%), followed by capelin (4–23%) and arrowtooth flounder were absent in SSL diet, while herring (1–6%). Shrimps are also well represented capelin was a dominant prey. The food habits in the diet of arrowtooth flounders (8–22%), of cetaceans are poorly known, but they can however their importance, together with that of reasonably be grouped in piscivorous (porpoises, capelin, has also decreased in recent years, while killer whales, and sperm whales, feeding mainly that of pollock has increased (Yang and Nelson on micronekton species) and zooplanktivorous 2000). Pacific halibut feed mainly on fish, parti- (minke and humpback whales). diets are cularly pollock (31–38%). Other common prey comprised of squid, euphausiids, capelin, sand 156 AQUATIC FOOD WEBS lance, and pollock. Piatt and Anderson (1996) Thysanoessa inermis and Thysanoessa longicaudata demonstrated a change in seabird diets since the are the dominant species in the western and last major reversal of the PDO, from one that central BS, while Thysanoessa raschii is more primarily comprised capelin in the late 1970s to common in the shallow eastern waters. Three another that contained little to no capelin in the hyperiid amphipod species are also common; late 1980s. Themisto abyssorum and Themisto libellula in the western and central BS, and Themisto compressa in the Atlantic waters of the southwestern BS. Barents Sea Close to the Polar Front very high abundance of The high latitude BS ecosystem is characterized by the largest of the Themisto species, T. libellula, have a relatively simple food web with few dominant been recorded (Dalpadado et al. 2002). species: for example, ! krill ! capelin ! cod, The dominant micronekton species of the BS or diatom ! copepod nauplii ! herring larvae ! food web are capelin, herring (C. harengus), and puffin (Figure 12.4). However, a more detailed polar cod (Boreogadus saida). The BS stock of inspection of the diet matrix reveals some level capelin is the largest in the world, with a biomass of complexity, mostly related with shift in diet that in some years reaches 6–8 million metric preferences among individuals of the same species tones. It is also the most abundant pelagic fish of but different age. The primary production in the the BS. Capelin plays a key role as an intermediary Barents Sea is, as an areal average for several of energy conversion from zooplankton produc- years, about 110 g C m2 per year. Phytoplankton tion to higher trophic levels, annually producing blooms that deplete the winter nutrients give rise more biomass than the weight of the standing locally to a ‘‘new’’ productivity of on average stock. It is the only fish stock capable of utilizing 40–50 g C m2 per year, 90 g C m2 per year in the the zooplankton production in the central and southern Atlantic part, and <40 g C m2 per year northern areas including the marginal ice zone. north of the oceanic polar front. In the northern C. finmarchicus is the main prey of juvenile capelin, part of the BS system, the primary production in but the importance of copepods decreases with the marginal ice zone (polar front) is important for increasing capelin length. Two species of krill, the local food web, although the southern part is T. inermis and T. raschii, and the amphipods more productive (Sakshaug 1997). T. abyssorum, T. compressa, and T. libellula dominate The zooplankton community is dominated by the diet of adult capelin (Panasenko 1984; Gjøsæter arcto-boreal species. The biomass changes inter- et al. 2002). The krill and amphipod distribution annually from about 50–600 mg m3, with a long- areas overlap with the feeding grounds of capelin, term mean of about 200 mg m3 (Nesterova 1990). especially in the winter to early summer. A number Copepods of the genus Calanus, particularly of investigations demonstrate that capelin can Calanus finmarchicus, play a uniquely important exert a strong top-down control on zooplankton role (Dalpadado et al. 2002). The biomass is the biomass (Dalpadado and Skjoldal 1996; Dalpadado highest for any zooplankton species, and the mean et al. 2001, 2002; Gjøsæter et al. 2002). Polar cod has abundance has been measured to about 50,000, a similar food spectrum as capelin, and it has been 15,000, and 3,000 thousand individuals per m2 suggested that there is considerable food com- in Atlantic, Polar Front, and Arctic waters, petition between these species when they overlap respectively (Melle and Skjoldal 1998). In the (Ushakov and Prozorkevich 2002). When abun- Arctic region of the BS, Calanus glacialis replaces dant, young herring are also important zoo- C. finmarchicus for abundance and dominance plankton consumers of the BS food web. Stomach (Melle and Skjoldal 1998; Dalpadado et al. 2002). samples show that calanoid copepods and appen- The Calanus species are predominantly herbivorous, dicularians makes up 87% of the herring diet by feeding especially on diatoms (Mauchline 1998). weight (Huse and Toresen 1996). Although data on Krill (euphausiids) is another important group capelin larvae and 0-group (half-year-old fish) of crustaceans in the southern parts of the BS. abundance suggests that the predation of herring CLIMATE FORCING, FOOD WEB STRUCTURE, COMMUNITY DYNAMICS 157 on capelin larvae may be a strong or dominant as 1,300 metric tons per day for the entire effect on capelin dynamics (Hamre 1994; Fossum population. It is suggested that this estimate 1996; Gjøsæter and Bogstad 1998), capelin larvae represents 63% of the total amount of food con- are only found in 3–6% of herring stomachs (Huse sumed by seabirds in the BS (Mehlum and and Toresen 2000). However, the latter authors Gabrielsen 1995). Consumption of fish by sea- comment that fish larvae are digested fast, and that birds is between 10% and 50% of the yearly predation on capelin larvae may occur in short, catch of fish in the fisheries. intensive ‘‘feeding frenzies,’’ which may have been In addition to seabirds, there are about 20 missed by the sampling. species of cetaceans, and seven species of pinni- The dominant macronekton species of the BS peds in the assemblage of the BS. are (Gadus morhua), saithe (Pollachius The majority of the cetaceans are present in the BS virens), haddock (Melanogrammus aeglefinus), on a seasonal basis only. Among these, the most Greenland halibut (Reinhardtius hippoglossoides), common are minke whale (B. acutorostrata), white long rough dab (Hippoglossoides platessoides), and whale (Delphinapterus leucas), white-beaked deepwater redfish (Sebastes mentella). The stock , and harbor porpoise (P. phocoena). Annual of Arcto-Norwegian (or northeast Arctic) cod is food consumption of minke whale has been currently the world’s largest cod stock. As estimated at approximately 1.8 tons, including macronekton predators, cod is dominating the about 140,000 tons of capelin, 600,000 tons of ecosystem. They are opportunistic generalists, the herring, 250,000 tons of cod, and 600,000 tons of spectrum of prey categories found in their diet krill (euphausiids) (Bogstad et al. 2000). Among being very broad. Mehl (1991) provided a list of pinnipeds, the most common is the harp seal 140 categories, however, a relatively small num- (Phoca groenlandica) whose abundance in the White ber of species or categories contributed more than Sea (a large inlet to the BS on the northwestern 1% by weight to the food. These are amphipods, coast of Russia) is 2.2 million. Other pinnipeds are shrimp (Pandalus borealis), herring, present in lower numbers. These include, ringed capelin, polar cod, haddock, redfishes, and juve- seal (Phoca hispida), harbor seal (P. vitulina), gray nile cod (cannibalism). The trophic interaction seal (Halichoerus grypus), and (Odobenus between cod and capelin is particularly strong in rosmarus). Yearly food consumption by harp seal the BS and, as shown later, occupies a central role in the BS is estimated at a maximum of 3.5 million in regulating community dynamics. However, cod tons, including 800,000 tons of capelin, 200,000– diet varies considerably during the life cycle, the 300,000 tons of herring, 100,000–200,000 tons of cod, proportion of fish in the diet increasing with age. about 500,000 tons of krill, 300,000 tons of amphi- The diet of haddock is somewhat similar to that of pods, and up to 600,000 to 800,000 tons of polar cod, but they eat less fish and more benthic cod and other fishes (Bogstad et al. 2000; Nilssen organisms. et al. 2000). In spite of the high capelin consumption Approximately, 13–16 million seabirds of more of harp seals, cod remains the primary consumer than 20 species breed along the coasts of the BS. of capelin in the BS, with an estimated annual The most plentiful fish-eaters among Alcidae removal (mean 1984–2000) of more than 1.2 million are Atlantic puffin (Fratercula arctica; 2 million tons (Dolgov 2002). breeding pairs) and guillemots, such as Bru¨ n- nich’s guillemot (Uria lomvia) and common Community dynamics guillemot (Uria aalge), 1.8 million and 130,000 breeding pairs, respectively. Bru¨ nnich’s guillemot In this section we summarize the community is the most important consumer of fish, of which changes of the reviewed systems in relation to polar cod (up to 95–100% on ) recent climate forcing. It is possible that some of and capelin (up to 70–80% on Spitsbergen) are the community changes that we describe are the dominant in the diets. Daily food consumption by result of human harvest, or a synergism between a Bru¨ nnich’s guillemot is 250–300 g, and as much human and environmental factors. At the current 158 AQUATIC FOOD WEBS state of knowledge it is impossible to quantify the relative contribution of climate and fishing.

However, as illustrated below, the synchrony of the Jan. 1960 Jan. 1964 Jan. 1968 Jan. 1972 Jan. 1976 Jan. 1980 Jan. 1984 Jan. 1988 Jan. 1992 Jan. 1996 Jan. 2000 Jan. 2004 biological changes among different components of La Niña dominant El Niño dominant the food web, and the large ecological scale of these changes point to the fact that climate must ? have occupied a central role. PDO negative PDO positive Tropical Pacific

Primary production may change drastically during Skipjack an El Nin˜o event. As the relax and the Yellowfin EPO warm pool extends to the central Pacific, the Yellowfin upwelling intensity decreases and the cold tongue WCPO retreats eastward or can vanish in the case of particularly strong events. The eastward move- ment of warm water is accompanied by the displacement of the atmospheric convection zone Bigeye WCPO allowing stronger wind stresses in the western Bigeye EPO region to increase the mixing and upwelling in the surface layer and then to enhance the primary production. Therefore, primary production fluc- South Pacific albacore tuates with ENSO in an out-of-phase pattern between the western warm pool and the central- eastern cold-tongue regions. Figure 12.5 PDO, ENSO, and recruitment time series of the main These changes in large-scale oceanic conditions tuna species in the Tropical Pacific (updated from Lehodey et al. 2003). WCPO Western and Central Pacific Ocean, EPO Eastern strongly influence the habitat of tuna. Within the ¼ ¼ Pacific Ocean. The y axis of the climate indices graph has been resource-poor warm waters of the western Pacific rotated. Note that during a positive PDO phase (before 1976), El Nin˜ o most of the tuna species are able to thrive, partly events become more frequent and tropical tunas recruitment because of the high productivity of the adjacent (skipjack, yellowfin, and bigeye) is favored. Conversely, during a oceanic convergence zone, where warm western negative PDO phase La Nin˜ a events are more frequent and the recruitment of subtropical tuna (albacore) is favored. See text Pacific water meets the colder, resource rich waters for more explanations regarding effect of climate on tuna of the cold tongue. The position of the convergence recruitment. zone shifts along the east–west gradient and back again in response to ENSO cycles (in some cases (i.e. albacore) showed the opposite pattern, with 4,000 km in 6 months), and has direct effect on the low recruitment following El Nin˜o events and high tuna habitat extension. In addition to the impacts recruitment following La Nin˜a events (Figure 12.5). on the displacement of the fish, ENSO appears to also affect the survival of larvae and subsequent Western Gulf of Alaska recruitment of tuna. The most recent estimates from statistical population models used for tuna After the mid-1970s regime shift, the GOA has stock assessment (MULTIFAN-CL, Hampton and witnessed a dramatic alteration in species com- Fournier 2001) pointed to a clear link between tuna position, essentially shifting from a community recruitment and ENSO-related fluctuations. The dominated by small forage fishes (other than results also indicated that not all tuna responded juvenile gadids) and shrimps, to another domin- in the same way to climatic cycles. Tropical species ated by large piscivorous fishes, including gadids (such as skipjack and yellowfin) increased during and flatfish (Anderson and Piatt 1999; Mueter El Nin˜o events. In contrast, subtropical species and Norcross 2000). In addition, several species of CLIMATE FORCING, FOOD WEB STRUCTURE, COMMUNITY DYNAMICS 159 seabirds and pinnipeds had impressive declines During the last 20–30 years, many micronekton in abundance. An analysis of the GOA community species of the GOA community have undergone a dynamics, for each trophic assemblage identified sharp decline in abundance, to almost extinction in the previous sections, follows. levels (Anderson and Piatt 1999; Anderson 2000). Evidence for decadal-scale variation in primary On average, shrimp and capelin biomass production, associated with the mode of variability decreased after the mid-1980s, while eulachon and of the PDO, is equivocal. Brodeur et al. (1999) sandfish are currently reaching historical high found only weak evidence for long-term changes levels (Figure 12.6). The biomass trend of juvenile in phytoplankton production in the northeast gadids (cod and pollock) is inferred by the Pacific Ocean, though Polovina et al. (1995) suggest recruitment estimates (Figure 12.7), based on fish that production has increased due to a shallowing stock assessment models (SAFE 2003). While cod of the mixed layer after the 1976–77 PDO reversal. recruitment has remained fairly constant, pollock A shoaling mixed layer depth exposes phyto- recruitment had a series of strong year-classes in plankton to higher solar irradiance, which in sub- the mid-1970s, followed by a continuous decline Arctic domains tends to be a in up to the 1999 year-class. However, the actual early spring and late fall. There is more conclusive strength of the 1999 year-class will only be avail- evidence for long-term variations in zooplankton able in coming years, as more data accumulates on abundance in the North Pacific Ocean. Brodeur the abundance of this cohort. and Ware (1992) and Brodeur et al. (1999) have In the macronekton guild, the most remarkable demonstrated that zooplankton standing stocks change in biomass has been that of arrowtooth were higher in the 1980s (after the 1976–77 phase flounder, with a sharp increase in both biomass change in the PDO), relative to the 1950s and and recruitment since the early 1970s (Figure 12.7). 1960s. Mackas et al. (2001) determined that inter- Arrowtooth flounder now surpass, in total annual biomass and composition anomalies for biomass, the once dominant gadid species, such as zooplankton collected off Vancouver Island, walleye pollock or Pacific cod. Halibut has also Canada, show striking interdecadal variations undergone a similar increase in biomass during (1985–99) that correspond to major climate indices. the available time series. The biomass of adult

Pandalid shrimps Eulachon 300 ) )

–1 250 –1 6 200 4 150 100 2 50 CPUE (kg km CPUE (kg km 0 0 1972 1976 1980 1984 1988 1992 1996 2000 1972 1976 1980 1984 1988 1992 1996 2000

Capelin Sandfish 4 ) )

–1 15 –1 3 10 2

5 1 CPUE (kg km CPUE (kg km 0 0 1972 1976 1980 1984 1988 1992 1996 2000 1972 1976 1980 1984 1988 1992 1996 2000

Figure 12.6 Abundance trend of non-gadid micronekton species of the GOA food web. Time series were obtained from survey data and are expressed as (CPUE kg km1) (Anderson and Piatt 1999; Anderson 2000). 160 AQUATIC FOOD WEBS

800,000 4e+6 200,000 600 Walleye Pacific cod pollock 500 600,000 3e+6 150,000 e-1 individuals)

400 g 400,000 2e+6 100,000 300

200,000 1e+6 50,000 200

Spawning biomass (metric tons) 0 0 Spawning biomass (metric tons) 0 100 1960 1970 1980 1990 2000 1975 1980 1985 1990 1995 2000 2005

1,200,000 7e+5 3.2e + 5 Arrowtooth 140,000 Flathead sole flounder 6e+5 1,000,000 120,000 2.8e + 5 5e+5 2.4e + 5

800,000 100,000 e-3 individuals) Recruits (millions of a g 4e+5 80,000 600,000 2.0e + 5 3e+5 60,000 400,000 1.6e + 5 2e+5 40,000 1.2e + 5 200,000 1e+5 20,000

Spawning biomass (metric tons) 0 0 Spawning biomass (metric tons) 0 8.0e + 4 1960 1970 1980 1990 2000 1975 1980 1985 1990 1995 2000 2005 Recruits (thousands of a Recruits (thousands of age-3 individuals) Recruits (thousands of age-2 individuals)

120 20 Spawning biomass (females only) left axis Pacific Halibut 18 100 Recruitment right axis 16 14 80 12 60 10 8 40 6 4 20 2

Spawning biomass (1000 metric tons) 0 0 1974 1979 1984 1989 1994 1999 2004 Recruits (millions of age-6 individuals)

Figure 12.7 Biomass time series for the dominant macronekton species of the GOA food web (SAFE 2003). Pacific halibut time series refers to the legal area 3A, which include most of the north GOA system (halibut data courtesy of Steve Hare, International Pacific Halibut Commission, Seattle WA, USA). The vertical dashed line indicates the 1976 regime shift. walleye pollock peaked in the early 1980s but has continuous decline, while the eastern stock stabil- since then declined (Figure 12.7), as a result of ized at low levels and continued to be treated as several successive poor year classes. Currently, ‘‘threatened’’. Harbor seals (P. vitulina) have also adult pollock (age 3 þ ) biomass is similar to that of declined in the GOA, compared to counts done the adult Pacific cod, whose biomass has generally in 1970s and 1980s. The exact extent of their increased since the 1960s, and is recently in slight decline is unavailable. However, in some regions, decline. particularly near Kodiak Island, it was estimated Among apex predators, the best-documented to be 89% from the 1970s to the 1980s (Angliss and studied case of species decline has been that of and Lodge 2002). Declines in seabird popu- the SSLs (Figure 12.8). In 1990, the National Marine lations have been observed in the GOA, though it Fisheries Service declared SSL a ‘‘threatened’’ should be noted that trends are highly species- species throughout the entire GOA region. and site-specific. For example, black-legged kitti- Later on, in 1997, the western stock (west of 144 wakes are declining precipitously on Middleton longitude) was declared ‘‘endangered’’ due to Island (offshore from Prince William Sound), but CLIMATE FORCING, FOOD WEB STRUCTURE, COMMUNITY DYNAMICS 161

Estimated Numbers of Adult and Juvenile SSLS 200 Russia (western stock) Alaska (western stock) Southeast Alaska (eastern stock) 150 British Columbia (western stock) Orogon and California (eastern stock)

100

50 Thousands of non-pups

0 1965 1970 1975 1980 1985 1990 1995 2000 Year

Figure 12.8 SSL population trends for various locations of the North Pacific. populations are increasing in nearby Prince zooplankton abundance, being able to graze down William Sound, in the central GOA. Likewise, the locally available zooplankton in a few days common murres are steadily increasing on Gull (Hassel et al. 1991). Capelin biomass has been Island, but are rapidly declining on nearby Duck fairly stable through the 1970s and early 1980s, but Island (both located in the vicinity of Cook Inlet). since then has had two major collapses, one in Cormorants on the other hand, appear to be 1985–90 and another in 1994–98 (Figure 12.9). The declining throughout the GOA (Dragoo et al. 2000). Norwegian spring-spawning herring has under- gone large fluctuations in abundance throughout the twentieth century (Toresen and Østvedt 2000; Barents Sea Figure 12.9). At the turn of the century the The bulk of primary production in the BS occurs in spawning stock biomass was around 2 million two types of areas: close to the ice edge and in the tons, increasing to more than 15 million tons open sea. The spring bloom along the ice edge can in 1945. From 1950, the biomass decreased occur as early as mid-April when the melt water steadily while the landings increased. In the late forms a stable, nutrient-rich top layer. As nutrients 1960s, the stock collapsed to a very low level are exhausted, but new areas of nutrient-rich water (<0.1 million tons), mainly because of over- is uncovered by the receding ice, the bloom exploitation (Dragesund et al. 1980). Strong regula- follows the ice edge in the spring and summer. tion of the fishery allowed the stock to recover In cold years, the increased area of ice leads to an very slowly during the 1970s, and more rapidly earlier and more intense ice edge bloom (Rey et al. after 1983 (due to the strong 1983 year-class). 1987; Olsen et al. 2002). Olsen et al. (2002) found Finally, the stock started to increase again around that this leads to a tendency for a higher annual 1990, reaching about 10 million tons in 1997. primary production in cold years. However, The cod stock declined from 3–4 million tons in most of it is ungrazed, due to mismatches with the 1950s to less than 1 million tons in the late primary copepods consumers, for example, capelin 1980s (Figure 12.10). changed the structure and cod. of the spawning stock greatly during the same The zooplankton shows large interannual period, by decreasing both the age at maturity variations in abundance, species composition, (Jørgensen 1990; Law 2000) and the mean age and timing of the development of each species. (G. Ottersen, personal communication) of the The capelin can have a significant impact on spawning stock from 10 to around 7 years. 162 AQUATIC FOOD WEBS

After this crisis in the cod fishery in the late 1980s, and saithe have decreased by around 50% since the the population has picked up; while the spawning 1960s and 1970s; the portion of large piscivorous stock was only 118,000 tons in 1987, it is now individuals has probably decreased even more. 505,000 tons. The other gadoid stocks of haddock The biomass of the long-lived deep-water redfish decreased from around 1 million tons in the (a) 10 1970s to 0.14 tons in 1986, and is still very small. The same can be said for Greenland halibut and 8 probably long rough dab. Abundance trends of BS pinnipeds are largely 6 unknown, though an upward trend has been noted for the abundance of harp seal, walrus, and com- 4 mon seal; ringed seal in the western part of the BS may be declining. Among seabirds, the common 2 guillemot has declined dramatically since the 1960s. Already in 1984, before the first collapse of the Biomass (million metric tons) 0 capelin stock, its abundance had declined by 75% 1975 1980 1985 1990 1995 2000 (compared to 1964) because of drowning in fishing gears and perhaps also the collapse in the herring (b) 20 stock around 1970. In 1984–85, the capelin stock collapsed, and in 1986 the large 1983 year-class of herring emigrated from the BS. As a consequence, 15 the largest colonies of common guillemot were further reduced by 85–90% in 1986–88. 10 Linking climate forcing, food web 5 structure and community dynamics

Biomass (million metric tons) In the TP, ENSO events are a central forcing 0 variable of the tropical tunas population dyna- 1950 1960 1970 1980 1990 2000 mics (Lehodey et al. 1997, 2003; Lehodey 2001). Figure 12.9 Time series of (a) capelin and (b) herring biomass The links between climate, habitat, and tuna in the BS. recruitment have been investigated in detail with a

Cod biomass and recruitment 2000 5 Cod total biomass Year versus cod recruit 4 1500 (vpa numbers age 3)

3 1000 2

500 1 e-3+ biomass (million metric tons) g A

Recruitment (millions of age-3 individuals) Recruitment 0 0 1950 1960 1970 1980 1990 2000

Figure 12.10 Cod biomass (age 3 and older) and recruitment (age 3) in the BS. CLIMATE FORCING, FOOD WEB STRUCTURE, COMMUNITY DYNAMICS 163 spatial population dynamic model (SEPODYM) number of juvenile skipjack between four and that describes the population responses of tuna to eight months of age. imagery indicated changes in both feeding and spawning habitats that this same area was the site of a major bloom (Lehodey et al. 2003). Results suggest that skipjack in phytoplankton some 4–8 months before and yellowfin recruitment increases in both the (Murtugudde et al. 1999). The catch in 1998 was an central and western Pacific during El Nin˜o events, all-time record; ironically, it led to a drop of 60% in as the result of four mechanisms: the extension of the price of skipjack, which were so abundant they warm water farther east (ideal spawning habitat is could not all be processed by the canneries. found in warm, 26–30C water), enhanced food While the main skipjack and yellowfin spawning for tuna larvae (due to higher primary production grounds in the western and central TP are associated in the west), lower predation of tuna larvae, and with the warm pool, those of albacore roughly retention of the larvae in these favorable areas as a extend through the central Pacific on each side of result of ocean currents. The situation is reversed the equatorial 5N–5S band. The out-of-phase during La Nin˜a events, when westward movement primary productivity between western (warm pool) of cold waters reduces recruitment in the central and central (equatorial upwelling) Pacific led the Pacific. When all the favorable conditions occur model to predict similar out-of-phase recruitment together, then high peaks of recruitment are fluctuations between species associated to one or observed. This was the case, for example, in the the other areas of the TP. In addition, the extension final phase of the powerful 1997–98 El Nin˜o event. of the warm waters in the central Pacific during El In the second half of 1998, the skipjack purse seine Nin˜o events that extends the skipjack spawning catch was concentrated in a small area in the grounds may conversely reduce those of the alba- equatorial central Pacific, and contained a high core (Figure 12.11). In summary, it appears that

60N 60 El Niño El Niño Oct.82–Mar.83 40N 40 1 5 20N 1 20 5 10 20 80 40 0 10 0 20 40 1 20S –20 5 1 40S –40 100E 140E 180 140W 100W 100E 140E 180 140W 100W

60N La Niña 60 La Niña Oct.88–Mar.89 40N 40 1 5

20N 1 20 1 20 40 10 80 0 1 0 40 1 20 10 20S –20 5

40S –40 100E 140E 180 140W 100W 100E 140E 180 140W 100W

Figure 12.11 Predicted spatial distribution of the larvae and juveniles of skipjack (left) and albacore (right) tuna in the TP during ENSO phases. 164 AQUATIC FOOD WEBS tuna recruitment and population fluctuations in occasional strong year-classes (e.g. 1984, 1988, and the TP would be controlled through physical, 1994), but never at the level of those observed bottom-up and ‘‘middle-down’’ (larvae predation before the 1980s. Bailey (2000) has shown that by epipelagic micronekton) rather than top-down during the years of the decline, pollock recruit- mechanisms, the intermediate ‘‘middle’’ component ment shifted from being controlled at the larval including the juvenile and young tuna. stage (1970s and initial part of the 1980s) to being The species dynamics of the GOA community controlled at the juvenile stage (late 1980s and appear to be strongly influenced by both climate 1990s). This change in recruitment control was due forcing and species interactions within the food to the gradual buildup of piscivorous macronekton web. For example, the gradual increase of mac- and consequent increase of juvenile pollock pre- ronekton during the mid-1980s resulted from a dation mortality. Ciannelli et al. (2004) have shown series of strong year-classes that followed the that macronekton predators, besides having a mid-1970s shift of the PDO index. High pollock direct effect on juvenile pollock survival (via pre- recruitment may have been the result of a series dation), can also indirectly affect their dynamics by of favorable conditions, including higher water amplifying the density-dependent mortality. temperature and lower spring wind stress (Bailey Micronekton species of the GOA community, and Macklin 1994) during the larval stages, as such as capelin and pandalid shrimps, might well as limited predation and density-dependent have suffered high predation mortality after the mortality during the juvenile stage (Ciannelli et al. predators buildup, with consequent decline in 2004). Immediately after the 1976 PDO regime abundance. To date, apart from the case of shift, both of these conditions were common in juvenile pollock described above, there is no the GOA area. Similarly, flatfish increase in direct evidence that predation was the primary recruitment was, in part, the result of a favorable cause of the decline of forage species in the GOA larval advection from the deep offshore spawning community. However, studies from other sub- to shallow inshore juvenile nursery grounds, Arctic ecosystems of the North Atlantic, point to conditions that appear commonly, particularly the fact that macronekton species set off strong during strong El Nin˜o events (Bailey and Pic- top-down control on their prey (Berenboim et al. quelle 2002). Intervention analysis applied to 2000; Worm and Myers 2003; Hjermann et al., many of the available GOA recruitment time 2004; but see also Orensanz et al. 1998). Also, in series indicates that large flatfish recruitment (i.e. the GOA community a top-down control of halibut and arrowtooth flounder) has significantly macronekton on micronekton is consistent with increased after the 1976 PDO shift, while pollock the changes of the groundfish diet observed in and cod recruitment has significantly increased the last 20 years (Yang and Nelson 2000; Yang during El Nin˜o North years (Hollowed et al. 2004). In addition to top-down forcing, during the 2001). As mentioned above, the frequency of El years following the PDO regime shift (i.e. late Nin˜o North events in the GOA region can 1970s and 1980s), capelin may also have been increase during positive PDO phases, thereby negatively influenced by strong competition with rendering more difficult the distinction between juvenile pollock. As indicated above, these two the effect of PDO or ENSO forcing in the North species have an almost complete diet and habitat Pacific community dynamics. overlap. In the GOA, capelin are at the south- The initial increase of macronekton biomass ernmost range of their worldwide distribution, after the late 1970s may have triggered a series of while, pollock are at the center of their range and food web interactions that directly and indirectly will be more likely to out-compete capelin during affected other species of the GOA community. For warm climate phases. example, with regard to pollock, the post 1985 Changes of the micronekton assemblage of the biomass decline was the result of mostly poor GOA food web had severe repercussions on apex recruitment from the mid-1980s to the late 1990s. predators, such as seabirds and pinnipeds. During this time frame, pollock had some Springer (1993) hypothesized that food-related CLIMATE FORCING, FOOD WEB STRUCTURE, COMMUNITY DYNAMICS 165 stresses have contributed to observed population (Ellertsen et al. 1989; Ottersen and Loeng 2000) declines of the seabirds. Among pinnipeds, the (Figure 12.12). The increased westerly winds over SSL decline in the western GOA is to these days the North Atlantic that are associated with a high one of the most interesting case studies in con- (positive) NAO phase, has, at least for the most servation biology (National Research Council recent decades, affected BS water temperature by 2003). Several hypotheses have been advanced to increasing (1) the volume flux of relative warm explain the decline and absence of recovery, water from the southwest; (2) cover; and including direct and indirect fishing effects (Alaska (3) air temperature. Increased BS water tem- Sea Grant 1993), climate change (Benson and Trites perature influences growth and survival of cod 2002), nutritional stress (Trites and Donnelly 2003), larvae both directly through increasing the develop- parasites and disease agents (Burek et al. 2003), ment rate and indirectly through regulating and, recently, top-down forcing (Springer et al. C. finmarchicus production. Variation in availability 2003). The majority of these hypotheses acknow- of C. finmarchicus nauplii is an important factor for ledge the importance of direct and indirect food formation of strong cod year-classes. In fact, the web interactions. For example, Springer et al. match–mismatch hypothesis of Cushing (1982, (2003) suggested that an increase of 1990) states that the growth and survival of cod predation was responsible for the Steller’s decline. larvae depends on both the timing and magnitude According to this concept, killer whales turned of C.finmarchicus production. In addition, an to SSLs after the baleen whales of the GOA dis- increase of inflow from the zooplankton-rich appeared due to an ongoing legacy of the post Norwegian Sea further increases availability of Second World War whale hunt. In addition, one food for the cod larvae (Ottersen and Stenseth might speculate that never fully 2001). High food availability for larval and juvenile recovered after the postwar decline due to a lack of fish results in higher growth rates and greater sustainable and highly nutritious fish prey. Top- survival through the vulnerable stages when year- down foraging by killer whales has played a major class strength is determined (Ottersen and Loeng role also on the decline of (Enhydra lutris) 2000). from western Alaska, with rather dramatic effects Cod and herring can potentially eat a large on the sea urchins (increase) and (decrease) amount of capelin (herring eats larvae, cod eats populations (Estes et al. 1998). In contrast to top- larger stages). Therefore, capelin can be expected down forcing, the nutritional stress hypothesis to experience high predation after a warm year (a.k.a., ‘‘junk-food’’ hypothesis) suggests that the with favorable conditions for cod and herring SSL decline was due to a shift in their diet toward recruitment. This was confirmed by Hjermann et al. prey with lower energy and nutritional value (e.g. (2004), who found that capelin cohorts that are pollock, cod) as a consequence of the reduced spawned two years after a warm year tend to be availability of the high-energy forage species weak. It appears that the predation-mediated effect (Trites and Donnelly 2003). of climate has been the main mechanism of the Probably, to a greater extent than in the GOA two collapses of the capelin stock in the last two food web, the pelagic community of the BS decades (1984–86 and 1992–94); a third collapse is dominated by a few very abundant species, appears to be occurring at the time of writing. resulting in strong interspecific interactions (Hamre These collapses (stock reduction of >97%) had a 1994). Specifically, the relationship between cod, huge impact on the BS community assemblage and capelin, and young herring have been viewed as food web structures. The most apparent impact particularly important for the ecosystem function- was a drastic reduction in population of some ing (Hamre 1994, 2000; Hjermann et al. in press). seabird species (such as the common guillemot The recruitment of herring and cod is strongly U. aalge; Krasnov and Barrett 1995; Anker-Nilssen associated with the temperature of the BS; speci- et al. 1997) and mass migration of a huge number fically, in cold years, recruitment is always low, of harp seal toward the Norwegian coast; while in warm years, it may be low or high incidentally about 100,000 seals subsequently 166 AQUATIC FOOD WEBS

5.0 C) ° 4.5

4.0

3.5

Temperature ( Temperature 3.0 1980 1982 1984 1986 1988 1990 1992 1994 1996 4,000

3,000

2,000

1,000 Age 1–2 herring Age 1–2

biomass (million tons) 0

3e+5

2e+5 (age 3) 1e+5 Cod recruitment Cod recruitment 0 9,000 7,500 6,000 Figure 12.12 From top to bottom, time 4,500 series of water temperature, herring (age 1–2) 3,000 biomass, cod recruitment (age 3), capelin and 1,500 amphipod (P. libellula and P. abyssorum) Capelin biomass (1000 million tons) biomass in the BS. Arrows indicate climate (in 0 black) and trophic (in gray) forcing on species dynamics. Water temperature in the BS 140 (correlated with NAO index) has a positive 120 effect on cod and herring recruitment. Herring 100 (mostly age 1–2) and cod (mostly age 3–6) 80 feed on capelin larvae, and adult, respectively, 60 and consequently have a negative effect on 40 capelin biomass. In turn, capelin feed on 20 northern zooplankton, of which P. libellula Amphipod density 0 and P. abyssorum are dominant 1980 1982 1984 1986 1988 1990 1992 1994 1996 components. drowned in fish nets (Haug and Nilssen 1995). capelin stock, with the ultimate effect of replacing Also, observations of increased zooplankton bio- a large capelin biomass with a small herring mass during capelin collapses (Figure 12.12) are biomass. indicative of the capelin impact at lower trophic Based on these observations we conclude that levels. In the ‘‘collapse’’ years, herring replaced the BS pelagic ecosystem appears, to some extent, capelin as the main zooplankton feeder. By feeding to be a ‘‘-waist’’ ecosystem, a term originally also on capelin larvae, a low biomass of juvenile coined for upwelling regions such as the Benguela herring is able to block the rebuilding of the ecosystem (Cury et al. 2000). In such ecosystems, CLIMATE FORCING, FOOD WEB STRUCTURE, COMMUNITY DYNAMICS 167

1970–84 1963). May (1973) showed, through mathematical modeling, that the ‘‘complexity begets stability’’ Cod Common guillemot idea might not necessarily be valid, although Maynard Smith (1974) cautioned about drawing Capelin Herring too firm conclusions on the basis of pure theoret- ical analyses. It bears note that the set of empirical observations gathered in this review are not sufficient to address in full details of the issue of Northerly zooplankton Southerly zooplankton resilience in marine pelagic systems. In addition, the anthropogenic forcing (i.e. fishing), active in 1986–88 all three systems, may have synergic effect with climate, complicating the issue of community Cod Common guillemot resilience even further. However, we feel that addressing the issue of community stability is Capelin Herring particularly appropriate within the context of this book, and may offer the unique opportunity to test the traditional theoretical knowledge of Northerly zooplankton Southerly zooplankton complexity and stability in systems (i.e. marine pelagic) where such theories have, to date, been Figure 12.13 Conceptual model of changes in BS trophic links unexplored. By focusing on three ecological during years of high capelin abundance (1970–84) and capelin systems of different and dynamics we collapse (1986–88, ‘‘collapse’’ years). During ‘‘collapse’’ years cod diet switches from one based primarily on capelin, to another based can ask to what extent they exhibit different on zooplankton and herring. Capelin decline, also has negative degrees of resilience. effects on seabird populations, such as the common guillemot. A summary of selected metrics for the three Herring replace capelin as a central forage species, and can inspected systems is presented in Table 12.1. Of exercise a large predation impact on capelin larvae (further delaying the three, TP is the one where the ocean variability, their recovery) and zooplankton, particularly from the Atlantic waters (southerly zooplankton). Because herring are not distributed associated with climate forcing, is most extreme. as far north as capelin, most of the northerly zooplankton remains Community changes in the TP ecosystem occur ungrazed during ‘‘collapse’’ years. Also, because herring live in likely with climate fluctuations, but high species the BS only for a limited period of their life cycle (see text), diversity, high degree of omnivory, high con- a large portion of the BS originated biomass is exported out of nectivity, and weak species interactions contribute the system. likely to its resilience and stability through time. In the GOA system, within the period in which community dynamics was monitored, we have the crucial intermediate level of pelagic fish is witnessed only a single shift of the dominant dominated by a few species (capelin, juvenile cod, climate forcing: a raise of the PDO index in the and herring in the BS), exerting top-down control 1976, a brief return to pre-1976 after 1989, and on zooplankton and bottom-up control of predators what appears to be a more stable return of the (Figure 12.13). PDO in recent years. Also, in recent years, it appears that the North Pacific climate is shifting Conclusions toward a new mode of variability (Bond et al. 2004). The community has clearly changed after The relation between structure and complexity of a the 1976 climate change, but after a transitory food web and its stability has been a much debated shift in 1989, there was no sign of a return to the issue within the field of ecology, starting with pre-1976 community state: piscivorous macro- Mac Arthur (1955) and Elton (1958) who claimed nekton kept on rising (e.g. arrowtooth flounder), that complexity begets stability—a view conveyed while forage species (e.g. shrimps and capelin), to many ecology students (see, for example, Odum and apex predators (SSL and seabirds) kept on 168 AQUATIC FOOD WEBS

Table 12.1 Descriptive metrics of the three pelagic ecosystems included in this review

Metrics TP Western GOA BS

Extension (millions km2) 80.0 0.38 1.4 Primary production (gC m2 per year) 50–300 50–300 110 Dominant period of climate forcing (year) 4–10 (ENSO) 20–30 (PDO) 4–10 (Nin˜ o North) Interannual with a weak period of about 8 years Nodal micronekton Epipelagic macronekton Juvenile gadids Capelin Nodal macronekton Small-size scombrids Large flatfish Cod Nodal apex Large-size scombrids, blue shark, Sea birds and pinnipeds Sea birds and baleen toothed whales whales Nodal trophic links Small-size scombrids—epipelagic Large flatfish-micronekton Cod-capelin-herring micronekton Large-size scombrids—epi and mesopelagic micronekton Climate trace in food web Bottom-up and middle-down Bottom-up and top-down Middle-out Ranked complexity (including diversity, High Average Low number of links and connectance) Ranked resilience High Low Average

decreasing. In recent years, immediately following tinuously shifting among two states, albeit whose the alleged 1999 change of PDO regime (warm to magnitude may vary, mainly depending on cold), pandalid shrimps, eulachon, and sandfish the abundance of the main forage species of the appear to be recovering, but there is no sign of systems. change in other forage (e.g. capelin) or apex pre- Our survey does suggest that the more complex dator (e.g. arrowtooth flounder) dynamics. Thus, system (TP) is more stable (or resilient)—thus the evidence gathered so far indicates that the 1976 supporting the ‘‘complexity begets stability’’ climate regime shift has led the GOA community concept. However, much work is needed before toward a new equilibrium state. However, given we can reach firm and general conclusions for the top-down control of apex predators on forage large marine ecosystems. Such work should cover species and the relatively long life cycle of large both the empirical and the analytical facets of macronekton, we may have to wait few more years community dynamics studies. Within the empir- to fully understand whether the GOA community ical framework, to overcome the objective difficulty can ever return to a pre-1976 state. of manipulating marine pelagic ecosystems, To a larger degree than the GOA, and certainly climate forcing should be seen as a natural ‘‘large- than other sub-Arctic systems (e.g. northwest scale perturbation experiment’’ (Carpenter 1990). Atlantic), the dynamics of dominant BS species Within the analytical framework, we stress the has a tendency to recover after collapsing. This importance of statistical models with structure seems to be a system dominated by three species: inferred from the observed patterns of population cod, capelin, and herring. Changes of capelin and variability (e.g. Hjermann et al., 2004). We also herring abundance (via top-down and bottom-up stress the importance of model simulations in forcing), have led the entire community to pro- relation to different levels of internal structures found phase-shifts in the food web structure. (i.e. food web) and external forcing (e.g. Watters During the last 30 years we have witnessed two of et al. 2003). Our hope is that what we have these transitions, one in 1985–89 and another in done here might serve as a spark leading to the 1994–99, with a possible third transition occurring development of more comprehensive analysis of at the time of writing (Hjermann et al. 2004). food web and community dynamics in large Thus the BS community appears to be con- marine pelagic ecosystems. CLIMATE FORCING, FOOD WEB STRUCTURE, COMMUNITY DYNAMICS 169

Acknowledgements for inviting us to contribute with a chapter on marine pelagic food webs. Comments from Matt We are indebted to a number of people who pro- Wilson and Kevin Bailey improved an earlier ver- vided data and information used in this chapter. sion of this chapter. This research is contribution They are: Paul Anderson, Mei-Sun Yang, Nate FOCI-0495 to NOAA’s Fisheries-Oceanography Mantua, Steven Hare, Ann Hollowed, and Matt Coordinated Investigations. Wilson. We are grateful to the editors of this book CHAPTER 13 Food-web theory provides guidelines for marine conservation

Enric Sala and George Sugihara

Introduction misery careful the urgency for better under- standing of marine ecosystems and for the need to Vast, unfathomable, and the major of our connect this knowledge to conservation and better planet, the oceans had an assumed capacity that resource management. Given this background, defied limitation. Unfortunately, this old pre- it is the particular and immediate challenge of sumption turned out to be terribly wrong as ecology and of food-web theory in general to con- technological advancement reduced the world’s tribute knowledge that can be used to assess, and oceans to mortal size by amplifying man’s inputs ultimately manage, marine ecosystems toward (pollution) and magnifying his outputs (fishing). desirable states and away from disaster. At the very And the fact of this magnification combined with least, one would like to have available some early evidence of man’s early impacts (Jackson and Sala quantitative indicators of when systems are being 2001; Jackson et al. 2001; Pandolfi et al. 2003) has threatened. rapidly overtaken our historical myth of the infinite Our objective here is to briefly review some of sea. The last couple of decades in particular have this very costly data with the aim of illustrating been hard on marine ecosystems and our myths ecological generalities that should be useful as about their robustness have all but faded. Witness guidelines for detecting and understanding human the vast hypoxic zone (a 20,000 km2 ecological impacts on marine ecosystems. That is, we shall desert in the Gulf of Mexico), the numerous other attempt to summarize generalities concerning coastal fiascos (e.g. Dayton et al. 1998; Jackson et al. human impacts on marine food webs (broadly 2001), and the recently reported collapse of many defined here to include properties of communities major fisheries to 10% of preindustrialized fishing and ecosystems) as they relate to our theoretical levels (Myers and Worm 2003). understanding of these webs, and illustrate the Much to our dismay, ecological disasters in generalities with specific examples. The time for marine systems are in full bloom. These disasters scholasticism is past. It is time to step up, move are unintended and uncontrolled experiments, and beyond ivory tower theory, and state what we can ironically, further our understanding of the know (however humble that may be) in a way that functioning and structure of marine systems, and can be useful for marine conservation. may eventually help to save them. Indeed examples Although the theme of this chapter concerns food of fisheries can yield important webs, and the latest initiative in fisheries manage- ecological data that has been extremely useful to ment is the ecosystems perspective, we wish to point ecologists and marine fisheries scientists and may out that it is important not to ignore the simple contribute to a general understanding of marine results from single-species management and con- populations and food webs. These are valuable but servation. For example, three simple patterns that costly data for marine science as some of these occur regularly with overfishing and are emblematic changes may not be easily reversible. Continuing on of impacts in the fishery are: (1) a decrease in the the negative side, this acceleration in environmental

170 FOOD-WEB THEORY FOR MARINE CONSERVATION 171 average size of fish caught, (2) a decrease in the and in the marine cases to be discussed below. It is average age of fish caught, and (3) a decrease in the remote-control aspect of this that is worrying the percentage abundance of super-spawners. We in its ambiguity in complex networks where emphasize that guidelines based on facts such as simple linear chains are the exception and not the these coming from single-species management rule. Nonetheless, when the very strongest inter- should not be ignored, while plans are being made actions line up in a linear chain, food-web theory to manage from the ecosystem and food-webs provides a reasonable cautionary guideline as to perspective. That is to say, although the current the expected propagation of indirect outcomes momentum in fisheries management is toward the resulting from manipulation (harvesting) of some ecosystems perspective, we should not loose sight component. Thus, if a top predator is harvested, of what we have uncovered from single-species and there is a supporting linear chain of strong studies. interactions below it, one should expect abund- ances in successive links in this chain to be alter- Dynamically propagated effects natively augmented and reduced. in food webs: interaction strength, and cascades Topology and dynamics

Simple cause and effect is comforting. Pull lever A The main caveat of the basic model result is that and get result B. A world of simple direct effects is the chain is a simple linear one, rather than part of safe and knowable. Things get scary when the a more reticulated network. The latter describes a levers are connected with hidden wires. Marine more realistic situation where the indirect effects ecosystems are known to contain such hidden are less clear. Studies with simple models that wires, and the most well-studied configuration of include omnivory links that bridge trophic levels wires is described by the so-called trophic cascade. and knit the simple chains into more complicated This is a problem whose origins trace back to a networks, confirm the intuitive hypothesis that classic paper by Hairston et al. (1960), which gave these alternative pathways will tend to reduce the an explanation for why the earth is green, and not likelihood and magnitude of trophic cascades (e.g. an overgrazed desert. Simply put, the idea is that McCann and Hastings 1997; McCann et al. 1998). overgrazing by herbivores is prevented by the These alternative pathways must be sufficiently action of predatory carnivores keeping herbivore weak so as to not be destabilizing in themselves, numbers in check. This simple notion of the but strong enough to matter. This suggests that remote control of plants by top carnivores (via more highly connected systems with more herbivores) was sharpened by Carpenter and omnivory links can be more resistant to trophic Kitchell (1974) and became enshrined as the cas- cascades. Note that this stabilizing effect of cade effect. Typically, food chains with three or connectivity only superficially runs counter to more nodes of odd length (length measured as the classical dynamic arguments about stability– number of nodes starting with primary producers) complexity (May 1973), where large numbers of are dominated by primary producers while chains species, higher connectivity, and strong interaction of even length are not. Thus, in fresh waters, when terms can be destabilizing. What harmonizes the secondary predatory fish are present in ponds result is the need for the secondary interactions to (chains of even length (4)) the water tends to be be sufficiently weak so as to not be destabilizing in clearer than when such top predators are removed themselves. That is, they cannot lead to Gershgorin leaving only primary predators, herbivores, and disks with large radii. Recent analysis of the most autotrophs (odd length (3)). Likewise, if predators- detailed web to date for a Caribbean coral reef of-top-predators are introduced (e.g. fishing: pro- system (Bascompte et al. submitted) containing ducing chains of odd length with humans at the several hundred species and several thousand top), the water turns greener. These things seem to interaction coefficients shows that interaction be true, both in models and in field experiments, strengths are often lognormally distributed so that 172 AQUATIC FOOD WEBS relatively few interactions are strong and the vast there may be some structural buffering operating majority are weak (supporting previous results against cascades in marine systems. This buffering obtained from smaller webs; for example, Paine may promote community persistence and stability 1992; Wootton 1997). One can assume that most (Fagan 1997; McCann and Hastings 1997; McCann real ecological systems are built this way. That is, et al. 1998). Nonetheless, as Bascompte et al. point they are the product of dynamic selection, in that out, the strong predatory chains or potential the configurations we see are ones that can persist cascades that exist are nonrandomly associated long enough to be seen. Likewise, system configu- with commercial top predators (sharks, in their rations that are dynamically fragile are rarely if Caribbean web), and thus cascades remain a con- ever seen. Although not well publicized, this idea cern for . This is, perhaps a was demonstrated in models fairly early on heavier concern in temperate marine food webs as (Sugihara 1982), where it was shown that dynamic these webs have been shown to be simpler and selection is capable of reproducing some difficult have lower connectivity than tropical marine webs. food-web regularities, including rigid circuits (tri- Therefore, it is not surprising that trophic cascades angulation), intervality, and tree-like guild patterns. appear to be more ubiquitous in temperate marine That is to say, more often than not, randomly food webs as compared to tropical webs (such as assembled simple model systems (large, random, those described in Hughes 1994; McClanahan 1995; and initially unstable) tend to decay dynamically Dulvy et al. 2004). into nonrandom configurations having the peculiar Thus, the configuration of the strongest interac- topologies found in natural food webs. That random tions in natural food webs is not random, and model assemblages settle down by dynamic selec- although current information suggests that in tion (via extinctions) to smaller stable assemblages general it is biased toward more stable configura- having many topological attributes (triangulation, tions, the danger remains that human activity in etc.) that are seen in nature is not surprising marine webs can be selective in a way that targets perhaps. It is an example of self-organization. fragile nodes and links. Awareness of known In similar vein one may speculate that in real principles relating dynamics to the configuration systems, potential trophic cascades (linear chains of strong linkages in webs is essential to informed of strongly interacting species) would be rare, and conservation and management, and speaks to the would be more commonly associated with acknowledged need for the ecosystems perspective omnivory than otherwise expected. This was in marine fisheries management. found to be true in the highly resolved Caribbean coral reef system cited above (Bascompte et al. in Examples of dynamically propagated preparation). The cooccurrence of strong interac- effects in marine food webs tions in tri-trophic food chains (three trophic levels connected hierarchically by two strong predatory We illustrate the relevance of these principles to links) in this Caribbean coral reef food web was marine conservation, with several cases of propa- less frequent than expected by chance. Moreover, a gated effects involving strong interactors in marine significant proportion of these strongly interacting food webs. chains had a strong omnivory link between the top predator and the basal resource. That is, tri-trophic Coastal–intertidal chains containing two successive strong predatory Starfishes (mostly Pisaster) in rocky intertidal food links were far rarer than expected by chance; and webs in the northeastern Pacific provide a para- when they did occur, they had a much higher digmatic example of cascades. They demonstrate likelihood of being associated with omnivory than the importance of strong interactors in the web and expected by chance. This, combined with the the consequences of their removal. In the coast of observation that marine food webs are thought to , Pisaster is a top predator feeding have a slightly higher degree of connectivity than mainly on the Mytilus californianus, but also webs from other (Link 2002), suggests that on barnacles and chitons. In the presence of FOOD-WEB THEORY FOR MARINE CONSERVATION 173

Pisaster, benthic dominance is shared among a otter by humans can result in increases of sea number of species, including barnacles and algae. urchin abundance beyond a point where they leave The classic experimental field work in this system shelter and eat the themselves, eventually conducted by Paine (1966, 1980) consisted of the resulting in the destruction of the kelp canopy and removal of Pisaster, and showed that the removal the formation of a barren dominated by encrusting of the top predator resulted in a successive replace- (Estes and Duggins 1995). ment by more efficient occupiers of space. The Two major patterns that emerge for coastal endpoint of these successional changes was a com- marine food webs following the removal of strong munity dominated overwhelmingly by Mytilus. interactors are as follows. Dramatic food-web-wide Although the removal of the top predator was effects can ensue if: (1) the strong interactor is a experimental, Paine’s work was an early warning predator on a food chain feeding upon a strong about the ecological risks and food-web impacts of interactor which in turn feeds upon an archi- species depletion. tectural species (such as kelp), or (2) if the strong Fishing provides many examples of propagated interactor is a predator feeding upon the dominant effects due to the removal of strong interactors primary producer or architectural species. This from the web. A global-scale example is the often should occur even though the strong interactor has seen explosion in populations that low abundance. result from the commercial removal of a variety of large- and medium-sized sea urchin predators, Coastal–Pelagic such as sea otters and fishes. These cases show that In some cases, dramatic changes in the abundance following the removal of sea urchin predators, the of the strongest interactors may not be caused by abundant urchins can subsequently turn complex fisheries targeting them directly, but may result algal forests into marine barrens. Again, this is a more indirectly from harvest of their prey. Here global phenomenon which has happened, among the effects may not be seen immediately and can other places, in the Mediterranean rocky sub- involve significant time lags, especially with the littoral (Sala et al. 1998; Hereu 2004), and kelp tendency for longer response times (e.g. longer forests in the north Atlantic (Witman and Sebens generation times) at higher trophic levels. An 1992; Vadas and Steneck 1995), Alaska (Estes and intriguing example is the delayed effect of the Duggins 1995; Estes et al. 1998), and New Zealand dramatic depletion (removal) of gray whales by (Babcock et al. 1999; see also review by Pinnegar oceanic fishing fleets in the 1960s and 1970s. These et al. 2000). In the relatively simple food webs in whales are a major prey item of transient orcas, temperate seas, these trophic cascades generally both of which normally reside in the open ocean of result in a decrease of many other measures of the north Pacific. Decades after intensive hunting ecological complexity (e.g. lower species richness of gray whales, transient orcas moved to nearshore and evenness). areas and shifted their feeding behavior to other Indeed, the hallmark example of trophic cas- marine mammals including sea otters (Estes et al. cades in nature are the kelp forests in the Pacific 1998; Springer et al. 2003). The reduction of sea northeast. In this system, sea otters are top pre- otters are believed to have triggered the trophic dators, and these top predators feed upon benthic cascade described above where kelp forests were grazers, such as sea urchins and abalones. The eliminated and replaced by barrens (Estes et al. consumption of sea urchins is especially import- 1998). ant, since sea urchins are avid consumers of kelp. Kelps are the most important architectural species, Some generalities providing a canopy that serves as refuge and When the food web contains an embedded three- provides microhabitats for an array of subordinate species food chain (defined by two strong links) species. In the presence of sea otters, kelps will there is a higher likelihood of trophic cascades. thrive and sea urchins generally shelter in crevices, For example, in a recent study of a Caribbean coral feeding upon drift algae. The removal of the sea reef food web, where the top predators are sharks, 174 AQUATIC FOOD WEBS groupers, and jacks; medium consumers are they can partly compensate for the reduced groupers; and basal species are herbivorous fishes, predation (Navarrete and Menge 1996). Other such as parrotfish and surgeonfish (Bascompte experimental work has indicated the potential et al. in preparation), it was found that sharks and importance of non-keystone predators in the other top predators are overrepresented in absence of the keystone species (Dayton 1971; strongly interacting food chains. Humans tend to Paine 1971). Indeed, some weak interactors may selectively target these strong interactors first, have strong effects due to great field densities or to because of their higher economic value (e.g. Pauly their link frequency, and their removal could also et al. 1998; Sala et al. 2004); and this has implica- cause food-web instability (McCann et al. 1998; tions for food web structure and dynamics. The Kokkoris et al. 1999). How to distinguish these removal of sharks may have resulted in a decline weak interactors with potentially destabilizing of the parrotfish populations via trophic cascade effects is another question. No matter what, pre- (Bascompte et al. in preparation). The decline of tending that the removal of small, seemingly the parrotfish has been implicated in the ecological insignificant species will have no significant effects shift of Caribbean reefs from coral to algal domin- on food webs is, at this point, an act of hubris ated (Hughes 1994). Thus the removal of reef and eventually risky. More studies are needed sharks by fishing may have contributed to a that explore the role of weak interactors in real, trophic cascade that extends to lower trophic levels speciose food webs. and involves competitive interactions between The effects of the removal of weak interactors algae and . In this example, the selective are not as predictable as the effects of the removal depletion of the strongest interactors increased of strong interactors (Paine 1980). On a per capita the likelihood of propagating trophic cascades basis, the absence of a weak interactor in a species- (Bascompte et al. in preparation). Thus, although as rich food web should not have a significant effect a group marine food webs may have some struc- on the web. From a real-world perspective and on tural buffering (higher connectance and omnivory), a population basis, the removal of weak interactors fishing remains a threat that promotes instability by can have unpredictable effects, these being basi- virtue of its tendency toward targeted removal of cally a function of numbers. This uncertainty is species at higher trophic levels. enhanced by the fact that weak interactions gen- As previously mentioned, the effects of removing erally show greater variance than strong interac- a strongly interacting species is more evident and tions (Berlow 1999). Moreover, there is consensus immediate in simpler, less connected food webs that weak interactors may have important stabil- (specifically webs having lower connectance). The izing roles in food webs (e.g. McCann et al. 1998). intuitive explanation for the vulnerability of simple Although small species often have per capita webs is their lack of ecological redundancy, with interaction strengths similar or smaller than larger the caveat that the redundant secondary interac- species, their tendency to have greater densities in tions cannot be too strong. This most likely accounts the field increases their potential food-web impacts for the relative ubiquity of trophic cascades (Sala and Graham 2002). There is clear evidence of in temperate marine food webs as compared to significant food-web-wide impacts due to striking tropical food webs. increase in the population of weak interactors. Not all food web components are equally Such increase can be caused by release from pre- important with regard to web dynamics. Strong dation after the elimination of a strong predator, or interactors can maintain food-web stability, and to other factors such as environmental fluctuations. their removal should be averted if the goal is to For example, amphipods are weak interactors in preserve complex food webs. Experimental work California kelp forests, and under usual densities in the eastern Pacific intertidal dealing with only their grazing has insignificant effects on the kelps four species interactions suggests that non- (Sala and Graham 2002), but during El Nin˜o events keystone species have only minor effects on the amphipod populations can exhibit explosive food web, although after keystone species removal increases and eat out entire sections of the kelp FOOD-WEB THEORY FOR MARINE CONSERVATION 175 forest (Graham 2000). Another example of the key planktonic algal blooms, and the removal of large role of species with relatively low biomass and herbivores in tropical coastal food webs can facil- weak per capita interaction strength is symbiont itate the appearance of sea grass disease (Jackson zooxanthellae in reef corals (Knowlton and Rower et al. 2001). One of the few examples of another top 2003). The loss of just a few species could enhance predator reducing the complexity of a food web is during prolonged warming events that of transient orcas in the north Pacific which, and shift the ecological state of coral reefs. by reducing the abundance of sea otters, enhance Although these examples do not involve trophic an increase in sea urchin abundance and, by cascades, they highlight the potential for trophic cascading effects, a decrease in kelp cover and cascades when a species interacting with an a reduction of overall food-web complexity (Estes architectural species is not a strong interactor but et al. 1998). Perhaps the reason in this case is that has a strong impact on the basis of extraordinary transient orcas, like humans, moved into coastal abundance. ecosystems relatively recently (Springer et al. 2003). Humans also have the advantage of agri- cultural subsidies, in turn subsidized by oil. This Humans versus other marine top predators might be the single most important reason why In the northwest Mediterranean subtidal, it has been humans are unlike marine top predators and their estimated that humans account for 11% of the effects on food webs are remarkably different. trophic links for a food web with relatively low Lotka-Volterra dynamics and population regula- taxonomic resolution (Sala 2004). In the same web, tion can operate for low-mobility marine pre- top predators such as monk seals and sharks dators, but human populations so far do not account for 24% of the links. Unlike other top pre- appear to be regulated by changes in the abund- dators, however, humans have the capability to ance of lower trophic level marine prey. Ironically, reduce the complexity of food webs by (1) effectively humans are using fossil food webs to exploit eliminating other top predators, (2) reducing the present marine food webs. abundance of many other species as a consequence of direct trophic effects, and (3) reducing the number Overfishing and inverting food-web structure of functional trophic levels (i.e. trophic levels con- taining species at abundances such that they still Ecological textbooks portray the classic pyramidal perform their ecological roles) (e.g. Dayton et al. food-web structure, where the biomass of a trophic 1998; Jackson et al. 2001; Pandolfi et al. 2003). level is always lower than the biomass of the Indeed, trophic levels are reduced to the point trophic level immediately underneath. The reasons where even fish parasites are less diverse and are purely energetic. While this seems to be a abundant in fished areas than in marine reserves general pattern in terrestrial food webs, in aquatic (C. F. Boudouresque, personal communication). food webs the biomass of an upper trophic level Food webs in the presence of marine top pre- can be higher than the biomass of a lower trophic dators generally exhibit greater complexity than in level upon which it feeds, if the lower trophic level their absence (e.g. Estes et al. 1998; Sala et al. 1998). has sufficiently rapid turnover. This is the case of Food webs that have been degraded by overfishing the coral reef food web of the northwest Hawaiian have lost the top-down regulation exerted by other archipelago, a large marine where predators and are increasingly influenced by the biomass of top predatory fishes is 54% of total bottom-up factors (Figure 13.1) (Jackson et al. 2001). fish biomass (Friedlander and DeMartini 2002). In Although humans exercise a formidable top-down contrast, in the main Hawaiian Islands, subject to control on marine food webs, unlike other top intense fishing, the biomass of top predators is predators we increase the likelihood of large fluc- only 3%. Top predatory fishes are not exclusively tuations and instability in food-web structure. For piscivorous, hence the differences in biomass instance, the removal of herbivores in estuarine between top predators and all prey will be smaller food webs subject to eutrophication can result in than those for fishes alone. An interesting result 176 AQUATIC FOOD WEBS

Before fishing After fishing Kelp forests

(a) Alaska/California Gulf of Maine (b) Alaska/California Gulf of Maine

Killer whales Killer whales People ?

Sea Sheep- Sea Sea Sheep- Sea Cod Cod otters head ? mink otters head mink ? Lobster

Sea ? Sea Sea cows Abalones Sea cows Abalones urchin urchin

Kelp Kelp

Coral reefs People (c)Monk (d) Monk Birds Sharks Birds Sharks Crocodiles seals seats

Pred Pred Pred Pred inverts fish inverts fish

Grazing Grazing Sea Sea Grazing Grazing Sea Sea inverts fish cows turtles inverts fish cows turtles

Sponges Macroalgae Corals Sea grass Macroalgae Corals Sea grass

Estuaries People

(e)Whales Sharks Seals Crocodiles (f) Whales Sharks Seals Crocodiles

Birds Pred fish Turtles Birds Pred fish Turtles

Pred Pred inverts Jellyfish inverts

Zoo- Grazing Sea Worms/ Oysters Zoo- Grazing Sea Worms/ Oysters plankton fish cows Amphipods plankton fish cows Amphipods

Benthic Benthic Phytoplankton Phytoplankton algae Sea grass algae Sea grass

Microbial Detritus Microbial Detritus loop loop

Figure 13.1 Schematical representation of selected marine food webs before and after intense fishing from Jackson et al. (2001). FOOD-WEB THEORY FOR MARINE CONSERVATION 177 from that study is that, although the biomass of oranges). This reflects the many more-or-less top predators is overwhelmingly greater than that independent ways of making a living, which is an of lower trophic levels in the northwest Hawaiian expression of the underlying heterogeneity in the archipelago, the biomass of its lower trophic levels ‘‘realized’’ or ‘‘functional’’ structure of the food is still greater than that in fished areas. This sup- web. Thus, the connectivity of the ‘‘effective’’ ports the idea that humans are often the strongest linkages has a more spread-out look as in scale- interactors in marine food webs. These extra- free networks, as opposed to uniformly dense as in ordinary differences in food-web structure between random networks. The heterogeneity implied by areas with and without fishing suggest that our scale-free structure in food webs corresponds to knowledge of marine food webs is probably biased more evenly branched niche similarity dendro- toward simplified food webs, as our baselines for grams. Using the approach advocated earlier, along ‘‘pristine’’ systems shift. When top predators are succession and for a particular food web belonging eliminated, increases in abundance of weaker to a relatively stable environment, we would expect interacting prey may cause significant population dendrograms to become more evenly branched to impacts. The organization of the food web also reflect the partial ordering of niches (apples and shifts: the strength of top-down control is dimin- oranges aspect), abundance distributions to become ished and the importance of bottom-up control more equitable, and food webs to become more enhanced. We do not claim that this is necessarily a complex (less homogeneously connected). However, general pattern in marine food webs; nonetheless, it this may not be true for high-energy systems where illustrates the danger of generalizing missing the endpoint of succession is characterized by the important functional components from webs. dominance of one or few species. Evenness declines when the underlying func- Species relative abundance and tional portrait simplifies, and the niches become evenness as an indicator of homogenized in that they revert from a ‘‘partial food-web complexity ordering’’ to a simplified ‘‘perfect ordering.’’ Such homogenization occurs when a single over- The distribution of species relative abundance in whelmingly strong factor is imposed on the system. food webs can also provide a simple means of Thus, if a previously complex system is subjected to measuring the health and vulnerability of eco- a strong homogenizing force (e.g. intense pollution), systems. Species abundance distributions reflect a formerly complex partially ordered system can the complexity of the underlying food web and become a perfect ordering. Species can now be its associated dendrogram of niche similarities appropriately ranked in relation to how they (Sugihara et al. 2003). Simpler food webs (e.g. two respond to the single dominant structuring force. trophic levels based on single ecological resources) The predicted decrease in equitability in abund- are expected to have very uneven abundance dis- ances that comes from simplification of the tributions and are dominated by a few species; underlying niche space via homogenization of whereas more complex systems will have niches is a phenomenon that has enormous more equitable relative abundance distributions empirical support. The reduction of evenness that (Sugihara et al. 2003). This applies to both between- accompanies human disturbance (such as nutrient and within-systemcomparisons.Thus,theoverlying enrichments or other forms of homogenizing abundance pattern is a reflection of the functional stress) is one of the most robust generalizations organization of food webs. Equitable abundances in ecology. imply symmetrical dendrograms of niche similar- ity. Such symmetrical or evenly branched den- Acceleration and homogenization drograms correspond to an underlying niche space of marine food webs that is complex and has many different structuring forces (e.g. a diverse resource base) that give rise to Odum (1969) and Margalef (1997) suggested the a partial ordering of niche similarities (apples and general pattern that average growth rates of 178 AQUATIC FOOD WEBS organisms will decrease with successional maturity, availability is going to be a crucial step in fisheries and that mature communities will have smaller management and marine conservation in general. production : biomass (P : B) ratios. In this scheme, The decline of top predators, the rise of high early successional stages, characterized by fast turnover species, and the microbial loop represent growing opportunistic species, are replaced by an acceleration and functional homogenization of slower growing, tough competitors. There are the food web. An increasingly important factor in exceptions of course, but this is a good generality. this homogenization of nature is the spread of Fishing acts the opposite way, selectively removing invasive species. Biological invasions have the the top predators first, and gradually moving down potential to homogenize entire food webs, and to target lower levels in the food web (Pauly et al. shift species abundance distributions to more 1998; Sala et al. 2004). The removal of predators skewed ones. Although connectivity and number accelerates the turnover of food webs by (1) redu- of trophic links in an invaded food web might be cing the biomass in species with lower turnover maintained, the community-wide ecological effects (P : B ratio), and (2) eventually triggering an of affected species can be dramatically altered. increase in the biomass of prey with higher turn- Invasive species including crabs (Grosholz et al. over. The end result is an increase in the turnover of 2000), snails (Steneck and Carlton 2001), and algae the entire food web. (Meinesz 2002; Boudouresque and Verlaque 2002) Accelerated growth rates often lead to destabi- can affect all trophic levels. An extreme example of lization of systems. A very general and robust the homogenization and acceleration of a marine result from ecological theory concerns the desta- food web after a biological invasion is the small bilizing effect of accelerating growth rates. This planktonic ctenophore in the Black Sea. idea was discussed early on by Rosenzweig (1971) The introduction of this exotic species caused in his , and has emerged as dramatic reductions in fish biomass and explosive one of the more robust generalities of ecological increases of gelatinous zooplankton (Shiganova theory (May 1974). Resource enrichment (e.g. and Bulgakova 2000) resulting in overall reduced nutrient loading) provoke higher growth rates that equitability in abundances in the web. destabilize the system and lead to species loss and In conclusion, humans universally accelerate reduced equitability. In the process of selectively marine food webs by (1) eliminating food-web culling species with lower growth rates, fishing components with slow dynamics (low turnover) increases the average growth rates in food webs, and enhancing dominance of high-turnover spe- which are also more likely to be regulated by cies, and (2) accelerating the dispersal of locally bottom-up processes, and exhibit greater fluctua- weak interactors (but potentially strong some- tion and instability. Chronic fishing pressure does where else), which probably would have failed not allow marine food webs to exhibit lengthy to disperse and colonize on their own during successional dynamics, and it can lock food-web ecological or evolutionary timescales. We are structure into one dominated by high turnover also accelerating evolution by increasing by several species. The endpoint of the degradation dynamics orders of magnitude the number of natural experi- may be (as it has been already observed in ments on species interactions in ecological time- estuarine systems) a food web overwhelmingly scales (Palumbi 2002). In other words, enhancing dominated by planktonic organisms, such as the links between food webs entails the accele- jellyfish and (Jackson et al. 2001). The ration of the dynamics within the webs. Inevitably, functional structure of the web is lost after the the accretion of structure and information in of entire trophic levels. While human webs (mostly urban) has to cause an on land higher turnover means greater agricultural acceleration and homogenization of the biosphere, production, in the sea it is not clear whether higher including marine food webs, besides the accelera- turnover always means higher production avail- tion of the oxidation of the necrosphere (Margalef able to humans. We believe that understanding the 1991). Such are the consequences of ecological relationship between food web turnover and food . FOOD-WEB THEORY FOR MARINE CONSERVATION 179

Human impacts reverse not exhibit significant seasonal changes, although successional trends their prey do (Garcı´a-Rubies 1996; Sala and Zabala 1996). Although the species composition and There are regularities commonly found in eco- topology of this food web does not change, the logical successions. Some general changes occur- biomass of many of its components changes dra- ring from early to mature successional stages are matically throughout the year. We would expect increases in species richness, number of trophic that changes in prey items cause subsequent levels, biomass of higher trophic levels, total bio- changes in interaction strengths. Therefore the mass, and three-dimensional biogenic structure structure of the food web exhibits seasonal chan- (Margalef 1997). Although these regularities have ges in structure and complexity. While these been well documented in phytoplankton, we know changes occur at a scale of months, longer-term very little for most other marine systems (but see changes can occur due to invasions of exotic spe- Grigg and Maragos 1974; Grigg 1983). The old cies and interannual variations in recruitment of diversity concept, understood as the distributions important species, such as sea urchins, among of abundances into species in food webs (also other factors (e.g. Sala 2004). called ‘‘ecodiversity,’’ to distinguish it from Similar changes in food-web structure occur at ‘‘biodiversity’’), often increases along terrestrial pluriannual scales in Californian giant kelp forests. succession as well (Margalef 1997). In marine food The most mature successional stage of these kelp webs, ecodiversity may show a unimodal rela- forests is virtually a monoculture of the giant kelp, tionship. In an underwater lava flow in Hawaii, (Dayton et al. 1984, 1992). ecodiversity of colonizing coral communities Every few years, strong storms or El Nin˜o South- increased with time but decreased before reaching ern Oscillation episodes virtually destroy the giant the successional endpoint, due to extreme compe- kelp canopy, resetting the food web to early tition for space, which led to monopolization by a successional stages dominated by undercanopy few dominant species (Grigg and Maragos 1974). or turf algae (e.g. Dayton et al. 1992). In both In any case, the increase in species richness in a the Mediterranean and Californian food webs, food web over successional time will inevitably successional trajectories involve an increase in result in the multiplication of the number of spe- three-dimensional biogenic structure, a recreation cies dependent or subordinate to those already of ecological niches that allows the recruitment of present, and therefore in an increase of food-web new species in the food web, and an increase in complexity. total biomass and total production. Regardless Although we know very little of changes in of what triggers these asymmetrical changes of marine food-web properties in successions, there is food-web complexity (e.g. disturbances or seasonal evidence of short-term (annual) successions in environmental cycles), complexity clearly increases planktonic and benthic communities. In a Medi- along succession. terranean rocky sublittoral food web, algal Human impacts (as well as other kinds of assemblages undergo striking seasonal changes in catastrophic disturbance) nearly always reverse structure (Ballesteros 1991; Sala and Boudouresque successional trends. In the temperate food webs 1997). Although algal species composition does not described above, fishing can cause sea urchin change significantly, total biomass shifts from an population explosions and the virtual elimination annual low where biomass is partitioned quite of complex algal forests (Estes et al. 1998; Sala et al. evenly among a large number of species (high 1998; Steneck 1998). In the Mediterranean food ecodiversity), to a peak where biomass is mono- web, increased sea urchin grazing reduces the polized by a few large species (low ecodiversity) biomass and diversity of the benthic community, (Ballesteros 1991). These seasonal changes in algal and reduces the differences between the endpoints biomass are immediately followed by similar of the annual succession: less total algal biomass, changes in epifaunal invertebrates (Sala 1997). The less structural complexity, and much smaller epi- biomass of sea urchins and fishes, in contrast, does faunal biomasses (Verlaque 1987; Hereu 2004). 180 AQUATIC FOOD WEBS

Most of the benthic dynamics in sea urchin barrens communities to move along succession and locks lay on a turf of microscopic primary producers, them in early successional states as long as pres- such as benthic diatoms, which is continuously sure is applied. But the removal of the disturbance grazed by the urchins. Fishing marine food webs may not be sufficient to allow the community to thus accelerates the system by increasing the global exhibit significant successional changes. This is a P : B ratio. The number of species, the number of key point that has tremendous consequences for trophic links, and complexity in general also the conservation of marine food webs, as we will decrease with increased fishing. In addition, three- discuss below. dimensional biogenic structure can drop dramati- cally because of trophic cascades. These cascades Environmental gradients and human impacts have the strongest food-web-side impacts when the removal of a predator results in the elimination of Environmental gradients can also produce changes architectural species. The removal of the archi- in the structure and dynamics of marine food webs tectural species in turn has effects on the recruit- and hence determine the strength and scale of ment and feeding behavior of many other species. human impacts. We would expect the complexity In the case of coral reefs, a decline in coral cover has of food webs to be greater in oligotrophic systems been shown to cause striking declines in reef fish (e.g. coral reefs) than in systems subject to high- species richness and abundance (Jones et al. 2004). energy/nutrient inputs (e.g. upwelling areas). The Invasive species homogenize food webs by number of species and trophic levels, and thus the truncating the frequency distribution of species number of trophic links and functional subwebs abundance, eventually turning it into an extremely are expected to be greater in low-energy systems, skewed distribution dominated by the invader. while the likelihood of monocultures or dom- Despite the introduction of one more species, inance of a few architectural species is greater in changes arising from invasion generally end up in high-energy systems. Food-web structural diver- local extinctions and decline in species richness. sity and evenness will hence be higher in low- A paradigmatic example is the transformation of energy systems. For example, food web complexity Mediterranean sublittoral habitats after the intro- is lower in a high-nutrient Californian duction of the tropical green alga Caulerpa taxifolia. overwhelmingly dominated by the biomass of one In the absence of the invader, Mediterranean sub- species, the giant kelp, than in an oligotrophic littoral food webs are composed of an extremely Mediterranean algal community with greater spe- diverse (up to >100 algal species in only 400 cm2) cies richness and equitability (Graham 2004; Sala and dynamic benthos, exhibiting annual and 2004). There also exist differences within systems. pluriannual successional dynamics (see above). For instance, the lower trophic levels of Caribbean After the invasion, the same habitats become a coral reefs were dominated by single species of green carpet, and diversity plummets to very low Acropora (a coral with relatively high growth rate) values (Meinesz 2002). The dominance of space in shallow habitats subject to strong wave energy, inhibits significant successional changes on the whereas in deeper, calmer habitats coral abund- local communities. Finally, because C. taxifolia is a ance was shared more evenly among tens of chemically defended species, the number and the species (e.g. Goreau 1959). Anthropogenic dis- diversity of trophic links also decreases. turbances have turned the shallow reefs previously The difference between nonanthropogenic and dominated by Acropora into algal beds also domin- anthropogenic disturbance is that the former are ated by a few species with yet higher turnover generally pulse disturbances and seldom cause (Hughes 1994; Knowlton 2001). In the Mediterra- local extinctions, allowing the food web to restart nean sublittoral, benthic communities exhibit an post-disturbance. In contrast, amazing gradient in structure and dynamics over a most human disturbances are chronic and cause mere 30-m distance along a vertical wall. Shallow local and even global extinctions. In many cases, food webs are dominated by a few species of algae this chronic pressure does not allow the ecological with strong seasonal dynamics, while deeper webs FOOD-WEB THEORY FOR MARINE CONSERVATION 181 are dominated by a diverse community of potential impacts of human activities and help suspension feeders (including sponges, ascidians, prevent them. and cnidarians) with slow, pluriannual dynamics (Garrabou et al. 2002). This strong biological gra- Conservation of marine food webs dient is explained by a strong gradient in physico- chemical conditions: shallow communities are It should be said that conservation (or manage- exposed to higher-energy inputs (more light and ment) is a value-laden concept. Nonexploitative wave motion) than deeper ones. uses of marine ecosystems are optimal with com- Fluctuations in food-web complexity are expec- plex food webs (e.g. a diver will spend a vacation in ted to be higher in high-energy systems. The above an unfished coral reef rather than in an overfished differences in food-web structure, specially for reef ). In contrast, social groups with sole interest in benthic webs, partly occur because high-energy industrial exploitative activities, such as fishing, systems allow for dominance of species with high may aim at less complex food webs in order to turnover, whereas in low-energy systems surface- target single species with large P : B ratios. The dependent strategies are based mainly on the slow simplest approach is to eliminate marine top pre- accretion of biomass and nonfunctional structures, dators to reduce competition for prey (Yodzis 2001), such as the biogenic matrix of a coral reef (Zabala following the disastrous example of previous and Ballesteros 1989). Food webs in higher-energy in the continental US. systems have a faster turnover (P : B ratio) than A downside is that simpler food webs are also more these in low-energy systems. This links structure prone to instability and fluctuations. Moreover, and dynamics of food webs within a particular even though simplified webs may not be a goal in set of environmental conditions: complex food itself, under current fishing practices marine food webs are ‘‘decelerated’’ and exhibit slow dynam- webs will inevitably be simplified and degraded. It ics, while degraded food webs have simpler is still not clear which food-web complexity levels structure and are ‘‘accelerated’’ (higher turnover produce optimal catches. Will fishing be more and instability). productive by exploiting accelerated food webs or What systems are thus more susceptible to decelerated ones? Accelerated food webs are greater rates of change and acceleration due to dominated by high P : B ratios, but total production human activities? We would expect that human available to humans may be lower, than in complex activities of similar intensity cause greater webs composed of more trophic levels. This is an damage in systems with less complexity because unexplored area that deserves serious attention. they have reduced functional redundancy. Do we want to preserve homogeneous ecosys- Therefore, tropical oligotrophic food webs such as tems with complex food webs, or mosaics of pat- coral reefs would be more resistant to disturbance ches at different successional stages? Conservation and less prone to fluctuation due to the reasons could be interpreted as the preservation of some- explained above. For instance, tropical systems thing stable, but food webs may be stable only at would be less likely to exhibit trophic cascades. short temporal scales. Thus the goal of marine However, based on the environmental constraints, conservation should not be the preservation of a low-energy systems will suffer a greater rate of particular food-web structure, because this struc- change than high-energy systems under a similar ture will inevitably change over time, and the cost: disturbance. In other words, the loss of com- benefit expectations of the public may be lowered. plexity and information will be relatively greater For instance, a coral reef can be restored to former in a ‘‘decelerated’’ system. Therefore, it seems levels of complexity by an expensive and time that the best question to ask is not what food consuming management process, with the invol- webs can we exploit, but ‘‘how are food webs vement of local communities whose goal is to going to respond to human activities on the basis enjoy both the intrinsic and instrumental values of of their intrinsic and environmental constraints?’’ the reef. However, the reef can be torn to pieces by The answer to this question will help determine a hurricane, and unless the local species pool and 182 AQUATIC FOOD WEBS the environmental conditions are right, it will not species for colonization at the right successional recover without further intervention. The public time, including strong interactors and keystone may be disappointed, blame managers and scien- species. For instance, species with relatively low tists, and be skeptical of further restoration efforts. dispersal cannot recover locally (e.g. in a marine Regardless of the goals of conservation efforts, the reserve) after being eliminated if the regional pool public and the decisionmakers need to understand is exhausted. However, this simple fact has been the dynamic nature of ecological communities. ignored in many conservation works, and used A better, albeit complex, approach for identifying as an argument by opponents to marine conservation goals is to integrate ecological conclude that reserves do not work and should be succession and food-web dynamics, as we have eliminated. The metapopulation and metacom- attempted here. Anthropogenic degradation of munity aspects of marine food webs are essential for food webs is similar to catastrophic disturbances, understanding successional dynamics in degraded although anthropogenic disturbances are chronic. webs. In some specific cases, the recovery of top In any event, anthropogenic disturbances dis- predators can be enhanced through human inter- assemble marine food webs, and conservation vention (e.g. sea otter reintroductions in central efforts that reduce or eliminate these disturbances California), but in many others these interventions eventually result in the reassembly of the webs. will not prove cost-effective. Future studies should The process can be viewed as a typical ecological explore the rates of recovery of marine food webs, succession where a food web acquires complexity not just single species, embedded in a spatial over long-time periods, asymptotically decreasing mosaic of webs at varying levels of complexity. the rate of change, to lose it more or less catastro- Meanwhile, we should aim at ensuring that strong phically, following anthropogenic disturbance in interactors are available, and to preserve functional shorter timescales. When the disturbance is past, trophic levels or food-web modules (sensu Paine the food web slowly regains complexity, the time- 1980). Diagnosing health and measuring success scale depending on the post-disturbance starting of management can be carried out using simple point. A good example is the changes in fish bio- measures of food-web complexity. We believe mass in the presence and the absence of fishing. The developing a based on succes- recovery of predatory fish involves much longer sional dynamics, and with practical implications timescales than their removal by fishing (Russ and at relevant scales, is an important challenge for Alcala 1996). And, although small species with fast food-web students and conservation biologists. turnover can recover to former unexploited biomass in coastal reserves within a few years, the complete Future directions: the linkage recovery of higher trophic level predators can take between marine food webs and more than 25 years (Micheli et al. in press). Marine human networks conservation science, as successional dynamics, is asymetrical: we know a great deal about the loss Science alone will not solve the conservation pro- of complexity, but relatively little about the slow blems of marine ecosystems. Food-web theory and recovery of that complexity. practice can provide decisionmakers with useful The goal of marine conservation should therefore be to information and recommendations vis-a`-vis the preserve the global initial conditions so that food webs impacts of anthropogenic activities on marine can self-organize and respond naturally to environ- ecosystems. However, these data and recommen- mental change. Global initial conditions include dations will be weighted against socioeconomic availability of subsidies, favorable environments, considerations before decisions are made. A major and low level of chronic disturbance, among others. problem, which in many cases has resulted in Food webs isolated from all possible subsidies are counterproductive actions, is that food webs less resistant to disturbance and indeed more and human societal networks have seldom been likely to be simplified. These subsidies include analyzed in an integrated way. In most food-web trophic subsidies as well as the availability of studies which include humans, the socioeconomic FOOD-WEB THEORY FOR MARINE CONSERVATION 183 intricacies that regulate the interaction strength Acknowledgments among humans tend to be ignored. In other words, We are grateful to J. Bascompte, L. Bersier, although economies and marine food webs are C. Boudouresque, R. Bradbury, F. Courchamp, linked, food-web models and economic models are P. Dayton, M. Graham, A. Hobday, J. Jackson, not linked. We urgently need to develop an inte- N. Knowlton, C. Melia´n, R. Paine, and R. Southwood grated network science that links human dynamic for interesting discussions on many aspects of actions and sociopolitical networks to ecological food webs. networks and other systems. CHAPTER 14 Biodiversity and aquatic food webs

Helmut Hillebrand and Jonathan B. Shurin

Introduction (prediction 1) and how prey density affects con- sumer diversity (prediction 2) (Figure 14.1(a)). Many fundamental ideas about food webs and the Second, we reverse the question and review the control of species diversity have roots in aquatic evidence on how consumer and prey diversity ecology. These include the trophic control of bio- affects the ‘‘strength’’ of the trophic interaction (i.e. mass (Forbes 1887; Lindeman 1942), the import- the biomass effect of predators on resources, and ance of top-down versus bottom-up regulating vice versa, predictions 3, 5) and diversity at adja- mechanisms (Lubchenco 1978; McQueen et al. cent trophic levels (predictions 4, 6) (Figure 1989; Leibold et al. 1997), the mechanics of com- 14.1(b)). We then continue by adopting a more petition (Tilman 1982; Sommer 1985; Keddy 1989), holistic view, asking how effects from single and the topology of food webs (Martinez 1991; trophic interactions propagate through food webs Havens 1992). Studies of food webs and species (prediction 7) and how diversity at one level diversity diverge in that food-web ecologists often affects the diversity at other trophic levels aggregate groups of species that share resources (prediction 8) (Figure 14.1(c)). Finally, we extend and consumers into guilds. Even very detailed our scope and look at the regional setting of the food-web studies with information on many food web (Figure 14.1(d)), addressing the impor- species contain a number of unresolved nodes tance of species additions by dispersal and inva- (e.g. ‘‘bacteria,’’ Martinez 1991). By contrast, species sion on trophic interactions (prediction 9). diversity is often defined monotrophically as the number of species within a guild such as primary producers or herbivores (Duffy 2002). While this The effect of single trophic division has historical and conceptual roots, it interactions on diversity is becoming apparent that studies of food webs The effect of consumer presence on prey and species diversity have much relevance to one diversity (prediction 1) another. For instance, diversity and species heterogeneity within guilds can affect biomass The effect of consumers on the coexistence of partitioning and the control of trophic structure competing prey species has been a central focus in (Leibold et al. 1997). Similarly, trophic architecture community ecology (Paine 1966; Lubchenco 1978; may have major effects on diversity within guilds Gurevitch et al. 2000; Chase et al. 2002). Two (e.g. Worm et al. 2002). Here we discuss avenues counteractive mechanisms mediate the effects for integrating the two approaches, and the of consumption: first, consumers reduce prey relationships among concepts pertaining to food biomass by increasing prey mortality and/or webs and species diversity. suppressing prey reproduction. In so doing, We begin by analyzing both the causes and consumers potentially drive prey species locally consequences of diversity in simple consumer— extinct, thus reducing prey diversity. Second, prey systems (Figure 14.1(a,b)). First, we investi- consumers also prevent competitive exclusion of gate how consumer presence affects prey diversity prey species and thus maintain local diversity.

184 iest acd onad n rydniyefcso iest rpgt pad,()dvriyi osmro rylvl fetthe affect levels prey size. or effect consumer consumer in local diversity affects (8) richness upwards, species propagate (d) Regional di diversity cascade. consumer (9) on trophic affects effects diversity the density prey of (6) prey strength size, and effect downwards consumer affects interactions cascade diversity trophic diversity prey single (5) in diversity, prey level affects (c) trophic diversity adjacent consumer (4) the (b) size, of diversity. effect consumer diversity affects or density biomass prey the (2) diversity, prey affects presence -hpdted.Ga ybl niaels motn rls elsuidrltos h rdcin r ragdi lsesa outlined as clusters 4 in arranged are predictions 9 The relations. well-studied less or (a) important chapter. less the indicate in symbols Gray trends. U-shaped 14.1 Figure rdcin ntepoaaino fet n(ro)dvriyi utpetohcinteractions trophic multiple in diversity of) (or on effects of propagation the on Predictions (c) (b) (a) rdcin nterlto ewe iest n rpi neatos fet r iulzda oiie eaie nmdl or unimodal, negative, positive, as visualized are Effects interactions. trophic and diversity between relation the on Predictions 8 consumer consumer diversity presence Consumer Consumer Consumer Consumer diversity diversity presence presence

Strength of Top Top rdcin nteefc fsnl rpi neatoso osmro rydiversity prey or consumer on interactions trophic single of effect the on Predictions trophic cascade Consumer diversity Prey diversity consumer mediate Inter- rdcin nteiprac frgoa iest ndvriyfo-e relations diversity–food-web in diversity regional of importance the on Predictions diversity diversity biomass biomass Basal prey diversity Basal prey Prey Prey Prey Prey biomass (d) 34 1 5 Basal prey 7 Consumer Consumer diversity effect size effect size Prey diversity Consumer diversity Consumer presence Consumer presence Local foodweb Prey diversity Regional Species rdcin nteefc fcnue rpe iest on diversity prey or consumer of effect the on Predictions Pool IDVRIYADAUTCFO WEBS FOOD AQUATIC AND BIODIVERSITY 3 osmrdvriyafcsconsumer affects diversity Consumer (3) .

9 Top Consumer Consumer Consumer 2 Consumer 6 effect size diversity Prey diversity diversity diversity Consumer diversity 7 Tp osmrefcson effects consumer (Top) (7) . Regional species Prey diversity pool presence Prey density Prey density 1 Consumer (1) . . versity. 185 186 AQUATIC FOOD WEBS

These contrasting forces result in widely divergent competitive may be less prone to key- consumer effects on prey diversity, ranging from stone predation effects (Chase et al. 2002). The positive to negative effects (Chase et al. 2002; probability of consumer-mediated coexistence is Figure 14.1(a)). Positive effects of consumer pre- moreover affected by intrinsic factors such as sence were found in plant—herbivore interactions consumer-to-prey size ratio, which determines the in both pelagic (Proulx et al. 1996) and benthic consumer effect on the spatial structure (and habitats in lakes (Hillebrand 2003), for large therewith diversity) of the prey assemblage carnivores such as fish (Diehl 1992; Shurin 2001) as (Steinman 1996; Hillebrand 2003). well as for small ones such as predatory nema- Additionally, consumer presence can weaken todes (Michiels et al., 2003). But even within the the competitive interactions between prey species same study systems, consumer presence decreased by indirect effects on the maintenance of resource prey diversity under different circumstances diversity (Abrams 2001), the regeneration of (Proulx et al. 1996; Shurin 2001; Hillebrand 2003). nutrients (Attayde and Hansson 1999), or on prey It is thus of major importance to describe the cir- behavior (Eklo¨v and VanKooten 2001; Werner and cumstances under which consumer-mediated Peacor 2003). These indirect effects promote prey coexistence occurs (Chase et al. 2002). These factors diversity if they reduce the impact of the compet- pertain to the consumer–prey interaction itself itively dominant prey species. (intrinsic factors) or to abiotic or biotic features of Prey diversity also increases with consumer the habitat (extrinsic factors). presence if the effect of consumers is distributed heterogeneously in space and time (Levins 1979; Intrinsic factors Pacala and Crawley 1992). Prey can persist by Consumers are expected to promote prey diversity repeated dispersal to open spaces in a meta- when their impact is greater on the dominant community (see prediction 9 for details), depend- competitor. Actively selecting consumers can ing on their dispersal rates and effects on prey switch toward the most abundant prey and act as extinction (Shurin and Allen 2001). Consumers also a stabilizing factor maintaining prey diversity remove the requirement for a trade-off between (Chesson 2000). The same outcome—maintenance colonization and competition for coexistence of prey diversity—also prevails when consumers between competing prey species: a prey that is an are specialists and consumers of abundant prey are inferior colonizer and competitor can persist if the more common, whereas rare prey species suffer consumer is sufficiently abundant. low per capita-loss to consumers (Pacala and Crawley 1992). Extrinsic factors Consumers can also prevent prey exclusion by Consumer density and prey productivity are two feeding mainly upon the dominant competitor biotic factors that mediate consumer effects on species and limiting its growth (Chase et al. 2002), prey diversity. Consumer density and feeding rate an effect known as keystone predation (Paine regulates how strongly consumers affect prey 1966; Lubchenco 1978; Leibold 1996). Keystone mortality. In analogy to the intermediate dis- predation does not require actively selecting turbance hypothesis (Connell 1978; Flo¨der and consumers, but may occur when dominance cor- Sommer 1999), the highest prey diversity has relates to higher susceptibility to consumption been observed at intermediate consumer density due to growth form or chemical constitution (Lubchenco 1978; Sommer 1999). Rare or inefficient (Hillebrand et al. 2000; Chase et al. 2002). Promi- consumers do not prevent competitive exclusion of nent examples of keystone predation are from their prey, whereas abundant or efficient con- marine intertidal areas, where primary space sumers can increase mortality to levels that inhibit holders are often dominated by a single species or the local existence of prey. Second, productivity growth form in the absence of consumers (Paine affects the rate of competitive exclusion, which 1966; Lubchenco 1978; Worm et al. 1999; Hillebrand determines whether consumers promote prey et al. 2000). Other systems with more symmetric coexistence (Huston 1994). Positive consumer BIODIVERSITY AND AQUATIC FOOD WEBS 187 effects on prey diversity are most likely when affects consumer–prey dynamics in experiments competitive exclusion is fast (i.e. at high pro- and models. For instance, the presence of vegeta- ductivity). Productivity increases the population tion in ponds alters the effect of predatory fish on growth rates of competing species and intensifies invertebrate diversity (Diehl 1992). Moreover, competitive dominance (see below for more on consumer presence often depends on the presence productivity–diversity relationships). Consumers of vegetation structure within the habitat (e.g. and disturbances play similar roles in maintaining macroalgae for , Worm et al. diversity in that the response depends on whether 1999) or outside the habitat (e.g. trees as landmarks their effects fall on dominant or subordinate for swarming insects, whose larvae remain close to species, and the rate at which processes which the egg deposition site, Harrison and Hildrew limit diversity (e.g. competitive exclusion) occur. 2001). Temporal heterogeneity (i.e. periodic dis- The roles of productivity and disturbance have turbance) also mediates the effects of consumers been incorporated in both nonequilibrium (Huston on prey diversity (Weithoff et al. 2000). Adverse 1979, 1994) and equilibrium (Kondoh 2001) models. physical conditions such as frequent disturbances Both predict interactive unimodal relationships or stress were proposed to reduce the importance between diversity and both disturbance and pro- of biotic interactions (Menge and Sutherland 1976; ductivity. These models offer concise predictions but see Chesson and Huntly 1997 for a critique). for consumer–prey dynamics: at high productivity, consumers enhance prey diversity, which suffers The effect of prey density on consumer from fast exclusion rates, by preventing exclusion. diversity (prediction 2) At low productivity, consumers reduce prey diversity by adding mortality in a situation where Here, we ask—in analogy to the effects of con- few prey species are able to exist. sumers on prey diversity—how resource density These predictions have been tested in two affects consumer diversity (Figure 14.1(a)). This quantitative reviews, both of which corroborated question is central to the debate about the rela- the predicted patterns. In studies from freshwater, tionships between productivity and diversity. marine, and terrestrial systems, a preponderance Increasing productivity has been proposed to of positive effects of consumers on prey diversity generate a unimodal or monotonic increasing was found in highly productive systems, whereas response of diversity (Rosenzweig and Abramsky negative effects of consumers dominated at low 1993; Abrams 1995). The increasing part of the productivity (Proulx and Mazumder 1998). Worm productivity–diversity relationship has a strong et al. (2002) conducted a quantitative meta-analysis theoretical basis. Increasing productivity results in on freshwater and marine experiments that larger population sizes and thus lower extinction manipulated both consumer presence and resource risk (Rosenzweig and Abramsky 1993; Abrams availability. Corroborating the model predictions, 1995). Additionally, diversity increases with pro- consumers promoted prey diversity at high nutri- ductivity if higher productivity promotes higher ent availability and decreased prey diversity at density of rare prey, allowing more specialized low nutrients (Worm et al. 2002). The interaction consumers to exist, or if productivity results in term between consumption and nutrient avail- higher intraspecific density-dependence (Abrams ability was stronger than the main effect of either 1995). Theoretical predictions for the negative part factor, suggesting that the feedback between con- of the productivity–diversity curve are less sumers and resources is of major importance for straightforward. Rosenzweig and Abramsky (1993) regulating prey diversity. reviewed nine hypotheses for the decrease of In addition to productivity, other physical diversity at highest productivity and proposed characteristics of the habitat that influence the that exclusion of consumer species increased at intensity of species interactions moderate the highest productivity because of reduced temporal effects of consumers on prey diversity. Spatial and spatial heterogeneity in prey supply at higher heterogeneity in the physical habitat strongly prey density. By contrast, Abrams (1995) predicted 188 AQUATIC FOOD WEBS that monotonic increasing functions should prevail significant U-shaped functions were observed too and argued that empirical evidence for reduced (Horner-Devine et al. 2003). prey heterogeneity at high supply rates was poor. This lack of generality suggests that diversity– Recently, Wilson et al. (2003) showed that enrich- productivity patterns depend on system-specific ment may decrease evenness and species richness environmental conditions. The effect of produc- if interspecific variance in tivity on diversity at one trophic level depends on increases with productivity (i.e. mean carrying the presence of a consumer (Worm et al. 2002). capacity). In that case, few species come to dom- Coexistence at one level is also enhanced if pro- inate the community at high productivity (Wilson ductivity fluctuates in time (Levins 1979; Sommer et al. 2003). Thus, the relationship between divers- 1985) or is heterogeneously distributed in space ity and productivity depends on how the effects of (Amarasekare and Nisbet 2001; Chase et al. 2001). abiotic factors (e.g., nutrients) are distributed The history of community assembly can also among species. influence the shape of diversity–productivity pat- The diverse theoretical predictions are reflected terns, with different invasion histories showing by a lack of any general empirical pattern relating distinct relationships (Fukami and Morin 2003). diversity and productivity. Surveys of the literat- Chase and Leibold (2002) found that different ure have found that unimodal, positive relations, relations operate on different scales in ponds, with and nonsignificant patterns are common, but a unimodal relationship between system pro- negative and U-shaped relationships are also ductivity and local richness, and a monotonic observed (Waide et al. 1999; Dodson et al. 2000; positive relationship with regional richness. This Mittelbach et al. 2001). The measures of pro- pattern indicates that productivity effects on ductivity used in these studies varied broadly and species coexistence are scale dependent; increasing comprise direct measures of productivity (such as nutrient supply enhances local diversity in oligo- actual or potential , primary trophic systems and decreases it under eutrophic productivity, or resource availability) and indirect conditions, while greater productivity always measures (such as biomass, rainfall), which may promotes regional diversity. Thus, more product- already affect the shape of the relationship (Groner ive ‘‘regions’’ have higher or more and Novoplansky 2003). turnover in species composition between local Aquatic systems harbor mostly unimodal func- sites. Chase and Leibold (2002) propose a number tions of diversity and productivity. Waide et al. of plausible explanations for this observation, (1999) found preponderance of unimodal functions including greater variability in abiotic conditions for aquatic vertebrates and invertebrates. Plants among sites in productive regions or local inter- showed fewer significant relationships in general, actions giving rise to alternative stable states only but the patterns in aquatic systems were more often under high productivity. unimodal than on land. In a survey of 33 lakes, Dodson et al. (2000) found unimodal relationships The effect of diversity on trophic interactions between species richness of different taxonomic (predictions 3–6) groups and lake productivity. Mittelbach et al. (2001) used original data (rather than the published Diversity is under strong direct control by trophic results as in Waide et al. 1999) for a meta-analysis of interactions, but evidently also from other local the productivity–diversity relation and found that and historical factors. It is therefore warranted to the average linear regression term was positive reverse the question and to ask whether variation and the average quadratic regression coefficient in diversity has consequences for the control of was negative. This analysis indicated unimodal trophic structure and partitioning of biomass in nonlinearity, and unimodality was the pattern most communities. This question is highly relevant in often observed, followed by monotonic increases. the face of globally increasing rates of biodiversity Among different groups of aquatic bacteria, diversity loss. These two aspects are not independent, was related unimodally to productivity, although but tightly linked by feedback loops, that is, the BIODIVERSITY AND AQUATIC FOOD WEBS 189 relation between species diversity and trophic By contrast, a meta-analysis of trophic cascade interactions is bidirectional (Worm and Duffy experiments across ecosystems found that produ- 2003). Recent experimental and theoretical cer diversity did not explain variability in the advances revealed that productivity and com- strength of top-down control (Borer et al., in munity stability may change with diversity within press). Thus, high prey diversity may dampen top- trophic groups (Loreau 2000; Tilman et al. 2001). down control over community biomass within Notwithstanding the much debated difficulties in systems, however these effects may be obscured by the design and proper analysis of these experi- other sources of variability among systems. ments, these studies raise the question of how the Effect on consumer diversity (prediction 4) diversity of prey or consumers affects (1) the Prey diversity is generally predicted to increase strength of trophic interactions and (2) diversity at consumer diversity through greater opportunities the corresponding higher or lower trophic level for niche specialization and differentiation. (Figure 14.1(b)). This prediction applies to both nonsubstitutable resources (nutrients, light) and substitutable The effect of prey diversity resources such as biological prey for heterotrophic consumers. However, the mechanisms may differ Effect on consumer control of prey biomass for these two prey types. (prediction 3) The impact of prey diversity on consumer control Non-substitutable resources. Classical competition over prey biomass has received surprisingly little theory predicts that the number of coexisting theoretical attention, and experimental manipula- species at equilibrium cannot exceed the number of tions are scarce (Duffy 2002). Theory predicts that limiting resources in homogenous environments high prey diversity corresponds to higher variance (Tilman 1982), a prediction that has been corrob- in edibility and thus higher probability of includ- orated in controlled experiments (Tilman 1982; ing inedible species (Duffy 2002). Alternatively, Sommer 1985). Consequently, Interlandi and consumers can derive nutritional benefits from Kilham (2001) found that phytoplankton diversity in a diverse prey community (the balanced diet lakes was greatest when more resources (light and hypothesis, DeMott 1998), and therefore be more nutrients) were below limiting levels. Equilibrium- abundant in areas with more varied resources. dynamics thus suggest a linear link between Thus, two opposite predictions emerge. Either resource diversity and consumer diversity. Analyses increasing prey diversity enhances consumption of nonequilibrium models show that nonlinear resistance (and reduce the consumer efficiency, dynamics allow many species to coexist on few Figure 14.1(b)), or increasing prey diversity resources (Armstrong and McGehee 1980). Using increases consumer abundance (and enhance models with three limiting resources and a large consumer effects). number of competing species, nonequilibrium Steiner (2001) found weaker consumer control dynamics with oscillations and chaotic fluctuations over algal biomass by zooplankton at higher algal reduced competitive exclusion and maintained high diversity, which he attributed to the presence of diversity (Huisman and Weissing 1999). inedible phytoplankton at high diversity that were Substitutable resources. By ingesting organic par- able to compensate for the reduction in vulnerable ticles, heterotrophic consumers acquire resource taxa. In a quantitative meta-analysis of grazer– packages, which are (partly) substitutable but experiments, grazer effects decreased differ in quality. While producer species vary significantly with increasing algal diversity tremendously in nutrient content, invertebrate and (Hillebrand and Cardinale 2004). This relation of vertebrate consumers are less variable in their consumer effects to prey diversity remained con- biochemical body composition. Therefore, the sistent even after carefully accounting for several growth and survivorship of consumers is deter- confounding variables such as habitat type, mined by the match or mismatch between experiment design, or consumer and prey biomass. consumer nutrient demand and prey nutrient 190 AQUATIC FOOD WEBS composition (Sterner and Elser 2002). Loladze et al. more effective at reducing prey biomass (Holt and (2004) showed theoretically that these stoichi- Loreau 2001; Figure 14.1(b)). The proposed ometric constraints can result in coexistence of mechanisms parallel biodiversity effects on two consumer species on one prey. In addition, process rates of primary producers (Loreau 2000; increased prey diversity may increase consumer Tilman et al. 2001). Consumer diversity is posi- survivorship, growth, and reproduction due to a tively related to consumer biomass and production more balanced diet (DeMott 1998; Anderson and (Moorthi 2000; Duffy et al. 2003) due to selection or 2000). Finally, higher diversity of sub- complementarity effects (Loreau 2000). Selection stitutable prey allows more consumers to coexist, effects increase the chance that a certain trait (here given that consumers show specialization for one consumer efficiency) is present at higher diversity prey type and the proportion of such specialists (Loreau 2000). Complementarity mechanisms remains constant or increases with prey diversity. (Loreau 2000) may prevail when different con- sumer species have different food requirements More detailed theoretical or empirical studies on or when positive interactions among consumer the effect of prey diversity on consumer diversity species increase consumer efficiency. are scarce. Petchey (2000) proposed three mechan- The available empirical evidence for evaluating isms promoting consumer persistence at higher these predictions has yielded generally supportive, prey diversity, ‘‘prey composition,’’ which is iden- but also some mixed results. In aquatic micro- tical to a more balanced diet, ‘‘prey biomass’’ and cosms, phytoplankton biomass decreased with ‘‘prey reliability.’’ In addition to providing a more increasing consumer diversity (Naeem and Li balanced diet, higher prey diversity may produce 1998). Downing and Leibold (2002) manipulated more abundant prey through sampling effects species richness across three functional groups in or over-yielding at higher diversity (more prey pond food webs (invertebrate grazers, invertebrate biomass) or reduced temporal variability of prey predators, macrophytes) and analyzed the biomass (more reliable prey). Petchey (2000) tested response of phyto-, zooplankton, and periphyton these predictions in aquatic microcosms and found biomass. Phytoplankton biomass increased, and that the persistence of consumers was enhanced by zooplankton and periphyton biomass decreased prey diversity via the prey reliability mechanism. with increasing overall richness, consistent A number of empirical tests in terrestrial systems with higher consumption of periphyton and also point to positive effects of resource diversity on zooplankton by either invertebrate grazers or the number of coexisting consumers (Siemann et al. predators. Suspension feeding invertebrates in 1998; Haddad et al. 2001), although other results streams showed significantly increased filtration are more equivocal (Koricheva et al. 2000; Hawkins rates at higher species diversity, which was also and Porter 2003). Also the diversity of marine higher than expected from monoculture perform- phytoplankton and zooplankton were uncorrelated ance (Cardinale et al. 2002). Sommer et al. (2003) at the global scale (Irigoien et al. 2004). Whereas found increased consumption of phytoplankton in high prey diversity promotes consumer coexistence mesocosms stocked with copepods after invasion by providing more opportunities for niche differ- of cladocerans into the mesocosms. Convincing entiation, the importance of this mechanism relative evidence for positive effects of consumer species to others in regulating heterotrophic diversity richness on process rates comes from a series of remains unknown in most systems. well-connected experimental and field studies on shredder species in streams. In a field survey of streams in Sweden, shredder species richness was The effect of consumer diversity positively associated with consumption (leaf litter Effect on consumer control of prey breakdown) rates (Jonsson et al. 2001). Laboratory biomass (prediction 5) experiments showed that high species richness Most theory and experimental work indicates that increased leaf processing rates (Jonsson and more diverse consumer assemblages should be Malmqvist 2000). Replacing individuals of one BIODIVERSITY AND AQUATIC FOOD WEBS 191 species with individuals from a different species over filamentous algae. In experimental containers, also increased process rates (Jonsson and snail presence had positive effects on tadpole Malmqvist 2003b). development, growth, and weight (Bro¨nmark et al. However, not all studies have found diverse 1991). Since tadpoles were competitively dominant consumers to be more efficient at reducing prey over snails, snails were forced to feed on low- biomass. Equivocal effects of diversity on process quality algae. By doing so, they increased nutrient rates were found for filter-feeding collectors in turnover rates, which enhanced microalgal pro- streams, whereas process rates even decreased ductivity and food supply to tadpoles. with increasing species richness for both grazers However, species interactions provide mechan- and predators (Jonsson and Malmqvist 2003a). In isms not only for positive, but also for negative experiments with multispecies algal assemblages, effects of consumer diversity on the strength of manipulating the number of Daphnia species did consumer effects. Increasing the number of con- not affect phytoplankton biomass (Norberg 2000). sumer species can decrease consumption if direct However, consumer richness was low (maximum behavioral interactions, territoriality, or aggression four species) and the species were closely related. play important roles. Amarasekare (2003) pro- The importance of the range of diversity manip- posed that decreasing consumption rates can be ulated becomes evident from the only marine study expected if consumer species compete not only by on consumer diversity to date: grazer diversity and exploitation of prey but also engage in interference identity in eelgrass communities affected neither competition. When inferior exploiters are superior epiphyte grazing nor eelgrass biomass when interference competitors, consumer efficiency will 1–3 consumer species were manipulated (Duffy be reduced at higher diversity—a prediction cor- et al. 2001). Subsequent experiments with up to six roborated by experimental results on a terrestrial species resulted in a clear decline in algal biomass system (Amarasekare 2003). Finke and at high grazer diversity, driven mainly by the Denno (2004) provide another example where highest diversity treatment (Duffy et al. 2003). intraguild predation and aggression among dif- Examples of species interactions that can result ferent spider species reduced their cascading in strong top-down effects at high consumer effects on herbivorous insects and plants. Thus, diversity are physical habitat modification and diverse predator assemblages exerted weaker top- behavioral flexibility. Cardinale et al. (2002) found down control over lower trophic levels. that increasing diversity of suspension-feeding caddisfly larvae led to reduced current shading by Effect on prey diversity (prediction 6) increasing biophysical complexity and thus Whereas the effect of prey diversity on consumer increased the fraction of suspended material diversity is relatively well known (see above), available to filter feeders. In cases of prey species there is only scarce information whether a single able to seek refuges, the presence of two con- consumer species has qualitatively different sumers may facilitate consumption (Eklo¨v and effects than more diverse consumer assemblages. VanKooten 2001): if only one predatory fish is We showed above (prediction 1) that consumers present in either pelagic or littoral habitats, prey enhance prey diversity if competitively dominant move to the less risky habitat, whereas the pre- prey suffer higher mortality or if consumers sence of both pelagic and littoral predators reduces maintain resource diversity for the prey. Both the possibility for predation avoidance and mechanisms are potentially affected by consumer increases predation success of both species. Similar diversity. Higher consumer diversity can result predator facilitation has also been shown between in more complete utilization of the prey spectrum. trout and stonefly predators in streams that mod- In that case, added consumer species reduce ify prey (mayfly) behavior (Soluk and Richardson prey diversity by increasing the consumption of 1997). Such positive effects or indirect subdominant prey species, which reduces may be highly asymmetric. Both snails and tad- the probability of consumer-mediated coexistence. poles consume periphyton and prefer microalgae Alternatively, higher consumer diversity can 192 AQUATIC FOOD WEBS increase the diversity of the prey’s resources The cascading effects of predator presence on through nutrient regeneration, space opening, or basal prey diversity were shown in simple aquatic increased spatial heterogeneity. These mechanisms food webs found in the pitcher plant Sarracenia are not mutually exclusive, but may interactively purpurea, where the addition of a top predator shape the response of prey diversity to increased reduced the abundance of the intermediate con- consumer diversity. sumer, cascading into increased bacterial abund- ance and diversity (Kneitel and Miller 2002). The bottom-up propagation of increasing basal prey Assembling the food web: cascades and density on predator diversity has mainly been propagating effects (predictions 7–8) shown in correlational studies suggesting non- The scenarios described above for the effects of linear relationships. Leibold (1999) analyzed consumer and resource control over diversity 21 fishless ponds and found hump-shaped curves become more complex once we imagine a world of primary producer (phytoplankton) diversity with more than two trophic levels, and the and consumer (zooplankton) diversity with potential for omnivorous feeding links. Extending resource level (either nitrogen or phosphorus). the previous sections, we ask below the questions Jeppesen et al. (2000) surveyed Danish lakes in five how top-consumer presence cascades down to productivity (total P) classes and found that basal prey diversity or conversely, how basal species richness of zooplankton and submerged resource enrichment propagates up to higher macrophytes declined with increasing productiv- trophic levels (prediction 7) and how diversity ity, whereas unimodal relationships were observed changes on one or several trophic levels affect for phytoplankton, fish, and floating macrophytes. propagating trophic interactions (prediction 8) In a third study, lake productivity was unimodally (Figure 14.1(c)). related to the diversity of both primary producers and consumers (Dodson et al. 2000). A critical factor relating propagating resource Propagating trophic effects on effects and predator diversity in chain-like inter- the diversity of basal prey or top consumers (prediction 7) actions is the increasing preponderance of non- edible prey at high resource supply (Leibold 1989; Predictions are still straightforward for chain-like Grover 1995). These predictions correspond well to interactions with predator, intermediate consumer, experimental results on pelagic communities and basal prey. Such chains represent a cornerstone indicating that the degree of resource and con- of community ecology, implemented by concepts sumer control depends on the proportion of such as the Trophic Cascade Model (Carpenter et al. inedible algae (Bell 2002; Steiner 2003). Thus, 1987), the Exploitation (Oksanen et al. enrichment may inhibit rather than benefit top 1981), or the Bottom-up Top-down Model (McQueen consumers due to interactions with inedible prey. et al. 1989). The main focus of these concepts has However, different enrichment–edibility trade-offs been to explain the distribution and control of bio- may be important in other aquatic communities. mass on different trophic levels (Leibold 1989; In benthic communities, species profiting most Leibold et al. 1997), whereas the control of diversity from enrichment are also most susceptible to received less attention. From the previous sections herbivores (Hillebrand et al. 2000). we derive that predator presence will alter the bio- The possible effects of top consumers on basal mass and diversity of the intermediate consumers, prey diversity become much more complex once which in turn affects the diversity of the basal prey. more reticulate feeding relations such as omnivory Similarly, basal prey density will also affect predator are considered. Omnivory is common in fresh- diversity via changes in the intermediate consumer water food webs (Diehl 1992, 1995), both as life- level. The sign of these indirect effects depends long or life-history omnivory (Polis and Strong on the mechanisms and environmental factors out- 1996; Persson 1999). Omnivory may uncouple lined above (predictions 1–6). consumer and prey abundance (Polis and BIODIVERSITY AND AQUATIC FOOD WEBS 193

Strong 1996), or lead to complex population Cardinale 2004) and therewith reduce the dynamics (Persson 1999). Intraguild predation propagation of enrichment effects through the may lead to nonlinear consumer–prey dynamics food webs. To our knowledge, this prediction and nonlinear effects of productivity on the food has not been tested. Moreover, the propagation of web (Diehl and Feissel 2001). During ontogeny of effects is complicated by indirect effects such as large consumers, size varies and so does prey size, apparent competition (Holt et al. 1994) and pre- resulting in complex physiology–population rela- ference of associated prey species (Wahl et al. 1997; tions, mutual predation, or intraspecific predation Karez et al. 2000). Both patterns may result in (cannibalism) between size classes (Woodward increasing consumer density at higher prey and Hildrew 2002; Persson et al. 2003). Moreover, diversity. indirect consumer effects such as nutrient regen- Much of the diversity–ecosystem functioning eration also propagate through food webs (Persson debate is motivated by the current high rate of 1999). Thus, while predictions from chain-like anthropogenic species loss and it is important to trophic interactions allow addressing the possible ask whether the diversity within the food web effects of top-consumers on basal prey diversity, affects the probability of secondary losses follow- more realistic and reticulate food webs prevent ing the deletion of one or several species. Borrvall simple assessments of cascading effects on basal et al. (2000) found that more species per functional prey diversity. group reduced the risk of cascading extinctions in model food webs. The probability of secondary losses was also affected by the trophic position of Cascading effects of diversity and diversity the species lost and the distribution of interaction loss (prediction 8) strengths. However, a subsequent study with more Propagating effects of diversity in top or basal complex food webs addressing connectance as trophic groups on the biomass and diversity on the well as diversity revealed that the probability of other trophic level are not well studied in multi- secondary extinctions decreased with connectance, trophic assemblages (Raffaelli et al. 2002). Theor- but was independent of species richness (Dunne etical predictions for diversity effects on biomass et al. 2002a). in multitrophic situations are highly complex (Holt and Loreau 2001; Thebault and Loreau 2003). This Emergent properties of food webs complexity arises already in simple food-web models with 2–3 trophic levels without omnivory The obvious complexity of predictions on diversity (Abrams 1993): increasing diversity within a effects in reticulate food webs (predictions 7–8) trophic level simply from 1 to 2 produced virtually resulted in early attempts to describe general any outcome for propagating effects depending on properties of entire food webs in order to derive food-web structure. general conclusions on food-web function and Experimental tests of these complex predictions assembly (Cohen 1978; Martinez 1991; Havens are scarce and focus on the role of the intermediate 1992). Several of these emerging properties of food consumer. Duffy (2002) suggested that increasing webs are supposed to change with diversity, such diversity at intermediate consumer levels would as the overall number of links (Havens 1992; reduce the strength of trophic cascades since prob- Martinez 1993), the number of links per species ability of resistant species being present increases (Martinez 1991, 1993; Havens 1992; Schmid-Araya with diversity. A terrestrial study including plants, et al. 2002b) and the mean and maximum length of herbivores, and their corroborated this food chains (Martinez 1991; Schmid-Araya et al. suggestion (Montoya et al. 2003), whereas Shurin 2002b). Other properties have been more con- (2001) found that high grazer diversity did not troversial. It remains unclear whether connectance weaken the strength of the trophic cascade. (the proportion of possible links realized) is Increasing diversity of basal prey can also insensitive to web size (Martinez 1991, 1993) enhance consumption resistance (Hillebrand and or if it decreases with diversity (Havens 1992; 194 AQUATIC FOOD WEBS

Schmid-Araya et al. 2002b). The general question processes and local interactions as mutually whether more diverse food webs are more stable exclusive alternative explanations for community (persistent, resistant against perturbations and structure (e.g. Cornell and Lawton 1992). Much extinctions) has a long history rich in reversals of effort has been directed at estimating the rela- paradigms (see McCann 2000 for a recent review). tive importance of local interactions and regional- More recent food-web analyses predicted the scale processes in shaping assemblages (Hilleb- topology of food webs from species number and rand and Blenckner 2002; Shurin and Srivastava, in connectance (Williams and Martinez 2000), ana- press). Recent results indicate several ways in lyzed the relationship between diversity and the which local and regional forces jointly influence distance of nodes over which trophic effects pro- communities and interact in generating patterns in pagate (Williams et al. 2002) and the division of food webs (Leibold et al. 2004). Integrating the two food webs into compartments (Krause et al. 2003). schools of thought has led to important novel Brose et al. (2004) unified the scaling of diversity to insights into dynamics and patterns in commu- spatial and food-web properties, which allows nities. Here we discuss the implications of regional investigating the impact of spatial patterns on dynamics for the structure and function of aquatic food-web structure. New analysis methods come food webs. from network theory, which has become an One way to view the dichotomy between local important tool to analyze food webs (Dunne et al. interactions and regional-scale processes is that 2002b, Garlaschelli et al. 2003). Whereas the ana- dispersal and speciation provide the supply of lysis of emergent properties may not be suitable to species to colonize habitat patches (Figure 14.1(d)), elude mechanisms of the relation between food- while local interactions serve as ‘‘filters’’ selecting web structure and biodiversity, it results in the which species coexist. By this view, diversity description of important patterns which give rise within a habitat can be limited either by the supply to testable hypotheses on the response and effects of or local exclusion by biotic or abiotic of diversity in trophic interactions. Moreover, there interactions. A variety of evidence suggests that is clear potential to increase the strength of such diversity within a guild or trophic level can have analyses (Borer et al. 2002). major implications for energy flow and trophic structure. If dispersal constrains local diversity, then it may in turn affect food-web processes. Extending the scope For instance, phytoplankton species richness was found to interact with microbial richness in The regional perspective (prediction 9) determining community production in laboratory Much of our thinking about aquatic food webs is microcosms (Naeem et al. 2000). Here, the focused on processes that take place within com- mechanism of diversity effects was a mutualism munities. This perspective dates back to Forbes’ between producers and decomposers driven by (1887) notion of lakes as ‘‘microcosms,’’ or isolated the abilities of species to utilize and produce communities, separate and apart from the sur- different organic and inorganic compounds. rounding landscape. Indeed, the important roles of Species-specific mutualism via nutrient cycles many local biotic and abiotic constraints on food is one possible route to diversity effects. Thus, webs have been unambiguously demonstrated. diversity within guilds of organisms at the same However, considerable evidence has also accu- trophic position can have major implications for mulated that all ecological communities are influ- food-web structure. enced by forces that operate over broader scales Another potential role for dispersal in shaping than just within habitats. These regional forces food webs is in limiting the length of food chains. include dispersal among habitats, speciation, and Several lines of evidence suggest that organisms at subsequent geographic spread, and large-scale higher trophic positions in lakes are more con- climatic effects. Early thinking about the roles of strained in their distributions by dispersal than local and regional processes treated regional more basal species. Producers, decomposers, and BIODIVERSITY AND AQUATIC FOOD WEBS 195 basal consumers like rotifers and protozoans opportunities present themselves (e.g. Knapp et al. are likely to be relatively effective dispersers 2001), this suggests that metapopulation behavior because of their abilities to reproduce asexually of fish promotes regional diversity among insects, and utilize dormant life stages that may disperse and possibly other prey as well. Shurin and Allen through space or time (e.g. diapause, Bohonak and (2001) used metacommunity models to show that Jenkins 2003; Havel and Shurin 2004). By contrast, predators can promote regional coexistence among many top predators such as fishes, , competing prey, and either increase or decrease and insect predators are obligately sexual (and local diversity depending on the traits of the therefore subject to Allee effects), have lower local species. Homogenizing predator distributions can population sizes (which should produce fewer have the effect of regionally excluding species that propagules), and lack resting stages to withstand rely on spatial refugia for persistence. For instance, transport through the terrestrial matrix. If organ- trout stocking reduced the number of fishless lakes isms at higher trophic positions are generally poor in the Sierra Nevada , with major dispersers, then the number of trophic levels in a implications for other species, particularly amphi- given habitat may often be limited by colonization bians such as Red Legged (Knapp et al. of top predators. For instance, Magnuson et al. 2001). Habitat connectivity can also influence top- (1998) found that lake isolation and the presence of down control of prey diversity and biomass by stream corridors had major effects on fish com- predators. Shurin (2001) found that introduction of munities, indicating that the inability to colonize fish and insect predators reduced zooplankton excludes some fishes. Another indication that dis- diversity in isolated ponds, but that these effects persal ability declines with increasing trophic were either reduced or reversed when colonization position is that many phytoplankton and zoo- from the regional pool was allowed. Similarly, plankton assemblages show rapid species turnover responses of marine macroalgae to grazers and and high resilience in response to perturbation nutrients depend on the availability of the seed such as acidification (Vinebrooke et al. 2003) or bank. The presence of propagules allowed for predator introduction (Knapp et al. 2001). New earlier recruitment which resulted in a seasonal species invade rapidly when environmental con- escape from grazing (Lotze et al. 2000). Also ditions change, indicating a large supply of competition between algal species was affected by potential colonists from the region or dormant supply, a process altered by herbivore pool. By contrast, fishes and crustacean zoo- presence due to a trade-off between profiting from plankton at higher trophic positions showed less the seed bank and being susceptible to grazing species turnover in response to acidification, indi- (Worm et al. 2001). These observations indicate cating lower ability to disperse (Vinebrooke et al. that local processes such as keystone predation 2003). Cottenie et al. (in preparation) derived at interact with dispersal among habitats in shaping a similar conclusion from the stronger distance- local diversity and composition. dependent decay of similarity in larger organisms Another potential indication of the role for on higher trophic positions compared to smaller regional dispersal in food-web structure is the organisms representing the base of the food web. relationship between food-chain length (as meas- Thus, the identity of species at high trophic posi- ured by the trophic position of top predators) and tions may often be sensitive to dispersal at the lake surface area (Vander Zanden et al. 1999; Post regional scale. et al. 2000). These stable isotope studies found that Strong dispersal limitation among predators also piscivorous fishes feed at higher trophic positions has the potential to influence local and regional in lakes with large surface areas. One possible diversity of organisms at lower positions in the explanation for this pattern is that larger lakes food webs that support them. For instance, species contain more diverse prey assemblages of of odonates often segregate between lakes with zooplankton, phytoplankton, and benthic prey and without fish (McPeek 1998). Since many fish- (Dodson et al. 2000). A greater number of prey less lakes are invasible by fish when dispersal species provides more potential trophic pathways 196 AQUATIC FOOD WEBS from producers to predators than in less speciose showed that biodiversity and food-web ecology assemblages. If large lakes contain more species are tightly coupled. Nevertheless, several areas due to higher colonization rates (bigger surface demand increased research efforts. Here, we out- areas provide larger targets to propagules), line four major issues extending the approach of then this suggests that biogeographic processes this chapter. influence energy flow throughout a food web. This explanation for the relationship between lake Diversity of the small surface area and trophic position of fishes remains to be tested. The results presented here are based on a limited subset of communities from aquatic habitats. Most Conclusions research has been done in ‘‘classical food webs’’ comprising phytoplankton, zooplankton, plankti- An ongoing challenge for both studies of bio- vorous, and piscivorous fish in the pelagial or diversity within trophic levels and of food webs benthic algae, macrozoobenthos, and fish in the has been an apparent lack of generalities that benthos. However, much diversity in aquatic sys- apply across ecosystems or organisms. On the tems occurs in the pelagic microbial food web scale of two interacting trophic levels (consumer (Berninger et al. 1993; Finlay et al. 1997) and in the and prey), the response of diversity to changes in meio- and microbenthos (Schmid-Araya et al. the biomass of the adjacent trophic level can be 2002a; Traunspurger 2002). This lack of informa- predicted reasonably well according to a few basic tion is astonishing in the light of clear trophic principles. Empirical studies have identified a links between macroscopic and microscopic food number of conditions, particularly consumer pre- webs (Wickham 1995; Adrian et al. 2001). For ference, productivity, and consumer density, that instance, cascading effects of predatory fish on influence the types of effects on biodiversity that invertebrate detritivores and fungal biomass and are observed. At the same time, model and detritivore diversity have been observed in lakes empirical studies predict pathways for diversity on (Mancinelli et al. 2002). Invertebrates grazing on one trophic level to affect the interaction between biofilm consume a complete microbial food web consumers and prey. By contrast, few generalities and first experimental results indicate that apply across multiple systems. The relation of invertebrates have different effects on the diversity biodiversity and food-web structure increases of the different autotrophic and heterotrophic dramatically in complexity as soon as the food web components (Hillebrand 2003; Wickham et al. comprises more than two trophic levels and reti- 2004). Including the microbial and detritus-based culate interactions. Here, the topological analyses food webs in studies of biodiversity will therefore of emergent properties of food webs may help to change the perception of patterns between divers- reveal relevant and testable hypotheses. As studies ity and trophic interactions. are performed in more ecosystems, and as theory develops in parallel, more regularities will likely Evolution and food webs emerge. However, we argue that recognizing the considerable heterogeneity among members of In order to understand the relation between a guild, and the inherently multitrophic nature of diversity and food webs, we have to extend the all communities, are essential steps if our under- temporal framework to include evolutionary pro- standing of biodiversity and food webs are to cesses. The importance of speciation and micro- progress. evolution is highlighted by two recent aquatic examples. Eklo¨v and Svanba¨ck (submitted) found Outlook that predators in different habitats separated single-species prey populations (perch) into sub- Our review on biodiversity and freshwater food populations, which developed different morpho- webs spanned several organizational levels and logy and exhibited contrasting fitness parameters BIODIVERSITY AND AQUATIC FOOD WEBS 197 in different habitats. This study suggests that not speciation). Brose et al. (2004) presented a first only competition but also trophic interactions may unification of diversity scaling to macroecological drive speciation (see also Rundle et al. 2003). The and food-web patterns. second example is on simple consumer–prey microcosms, where consumers showed clear Anthropogenic effects on food webs microevolutionary trends (Fussmann et al. 2003) and diversification of algal prey into several Most ecosystems are dominated by humans and clones changed dramatically the consumer–prey the anthropogenic effect on species within these dynamics (Yoshida et al. 2003). Evolutionary ecosystems is rarely evenly distributed over dynamics clearly drive patterns of diversity by trophic levels. Cultural eutrophication of lake providing new species, and may occur on short ecosystems may shift the balance from benthic enough timescales to influence the outcome of toward pelagic food webs and result in the loss of ecological interactions. benthic pathways of organic matter processing (Vadeboncoeur et al. 2003). Human activities that decrease biological diversity are most strongly felt and food webs at top predator levels. Fisheries are prominent The description of large-scale diversity patterns examples where humans reduce the number of and their underlying causes has received wide large fish species with unforeseen consequences attention, emerging into the field of macroecology for diversity (Myers and Worm 2003). Even more (Gaston and Blackburn 1999). There is an aston- complex anthropogenic effects such as climate ishing lack of consideration from trophic ecology change will probably affect especially top and in the discussion of macroecological patterns. For intermediate consumers (Petchey et al. 1999), example, the importance of ecosystem size on because the generally lower overall diversity and food-chain length corresponds well to theoretical smaller population sizes at higher trophic posi- analysis of species area relationships (SARs). Holt tions skew any proposed biodiversity loss toward et al. (1999) modeled closed communities of spe- stronger effects on top species (Duffy 2002). cialized consumers and found that the SAR will be Therefore, food-web ecology has to provide steeper in slope with increasing trophic rank. answers how the anticipated further loss in Although empirical evidence is scarce and model important consumer species affects food-web outcomes may be altered by system openness and structure and ecosystem functioning. generalist consumption, this would in turn mean that richness of top predators increases faster with ecosystem size than for basal species, resulting in Acknowledgments an increase in mean chain length (cf. previous results on food-chain length). A recent meta- This review was read, corrected, and commented analysis of nearly 600 latitudinal gradients of upon by U.-G. Berninger, A. Ehlers, M. Feiling, diversity revealed that the decline of diversity A. Longmuir, H. K. Lotze, B. Matthiessen, from the equator to the poles varied with trophic S. Moorthi, R. Ptacnik, U. Sommer, B. Worm. All of position (Hillebrand 2004). The gradient was them tremendously improved this chapter. HH weakest for autotrophs and strongest for carni- acknowledges financial support from the Erken vores, suggesting that trophic structure alters the Laboratory (Uppsala University, Sweden) and importance of possible causes for the latitudinal Institute for Marine Research (Kiel University, gradient (such as area, productivity, and net Germany). This page intentionally left blank PART IV Concluding remarks This page intentionally left blank CHAPTER 15 Ecological network analysis: an escape from the machine

Robert E. Ulanowicz

Introduction to old, habitual ways. McLuhan’s favorite example was that of International Business Machine, Inc, The scientific community is abuzz over the use of which began its existence building machinery for networks to describe complex systems (Watts 1999; businesses. The corporation floundered in unex- Baraba´si 2002). Recently, the leading journals have ceptional fashion, until it dawned upon someone reported the rediscovery of the fact that collections that the nature of their enterprise was really of processes and relationships in complex systems more like the transfer of information. Thereupon, often deviate from conventional statistics (Jeong the company fortune exploded with such vigor et al. 2000; Montoya and Sole 2002). Many net- that its consequent meteoric rise is studied till works of natural systems are said to be ‘‘scale-free’’ today by business students and stockbrokers alike. in that their elements are distributed according to (Eventually, the association with machinery was non-normal power laws (Ulanowicz and Wolff hidden within their acronym, IBM, which one now 1991; Baraba´si and Albert 1999). Others appear to associates more with the technology of compu- be nested in hierarchical fashion, while still others tation than with machines per se). are dominated by chains of what Almaas and If some readers remain unimpressed by Baraba´si (2004) have called ‘‘hot links’’ (see also McLuhan’s admonition, perhaps they might heed Ulanowicz and Wolff 1991). Because scientists the castigating words of the mathematician, John have been conditioned for over 300 years to regard Casti (2004). Casti builds upon a children’s story, nature as a grand clockwork, the race is now ‘‘Little Bear,’’ by Else Minarik (1957) wherein the on to elucidate the ‘‘mechanisms’’ behind why principal character tries to get to the by the elements of natural systems are arranged in climbing a tree. Casti contends that using the these ways. conventional methods of physics to improve one’s Before getting too caught up in the search for understanding of complex systems is akin merely mechanisms, it might be wise for investigators to to climbing a taller tree. He cites how, when consider if they are pointing their flashlights in the complex systems are approached using conven- right direction (Popper 1977), for as the late media tional tools, they almost invariably give qualit- pundit, Mashall McLuhan (1964) once wrote, atively contradictory prognoses (and not ones that The or the meeting of two media is a moment of are merely quantitatively inaccurate). truth and revelation from which new form is born ...The If some readers should find Casti’s admonition ... moment of the meeting is a moment of freedom and a little too pessimistic, they would do well to recall release from the ordinary trance and numbness imposed how Eugene and Howard Odum, Robert Rosen, by them on our senses. Stanley Salthe, this author and others have labored By this he meant that when confronted by new to point out how the dynamics of is ways of seeing things, observers too often are qualitatively distinct from that of purely physical numbed into interpreting what they see according systems. In fact, if one examines closely the

201 202 AQUATIC FOOD WEBS fundamental postulates upon which science has fundamental postulates about nature according to operated for the past 300 years, one discovers that which Newtonian investigations were pursued: each axiom in its turn is violated by one or another Newtonian systems are causally closed. Only ecosystem behavior (Ulanowicz 1999.) Given such mechanical or material causes are legitimate. disparity, it comes as no surprise that conventional Newtonian systems are deterministic. Given precise approaches to ecosystems behavior have not initial conditions, the future (and past) states of yielded any more progress than might come from a system can be specified with arbitrary precision. ‘‘climbing a taller tree.’’ Partly out of frustration, Newtonian systems are reversible. Laws governing a number of investigators have embarked upon a behavior work the same in both temporal phenomenological search for new ways to quantify directions. ecosystem dynamics and have keyed on quanti- Newtonian systems are atomistic. They are strongly fying networks as a possibly fruitful approach decomposable into stable least units, which can be built up and taken apart again. (Ulanowicz 1986; Wulff et al. 1989; Higashi and Burns 1991). In light of these considerations, In addition, Prigogine and Stengers (1984, see also it should prove helpful to study in more detail Ulanowicz 1999) alluded to a fifth article of faith, exactly how ecosystem dynamics transcend namely that the usual scientific metaphysic and to explore Physical laws are universal. They apply every- more fully how quantifying ecosystem networks where, at all times and scales. provides a completely new perspective on the natural world. Ecosystem dynamics Normal science Although it might at first seem somewhat removed The problem with writing about the ‘‘conven- from the subject of networks, determinism is the tional’’ approach to science is that no single image most convenient assumption with which to begin exists. Rather, as Kuhn (1962) has suggested, each the discussion of ecosystem dynamics. Every individual weights differently the various ecologist is aware of the significant role that the criteria that he or she uses to delimit legitimate aleatoric plays in ecology. Chance events occur science. To deal with such diversity it is helpful to everywhere in ecosystems. Stochasticity is hardly focus on a set of fundamental postulates that once unique to ecology, however, and the entire dis- formed a broad consensus about nature around the cipline of probability theory has evolved to cope turn of the nineteenth century (ca. 1800). This with contingencies. Unfortunately, few stop to ‘‘strawman’’ is not intended to describe the beliefs consider the tacit assumptions made when invok- of scientists today—no one still adheres to the ing probability theory—namely that chance events truth of all the classical postulates. On the other are always simple, generic, and recurrent. If an hand, virtually every contemporary approach to event is not simple, or if it occurs only once for all natural problems still depends upon one or more time (is truly singular), then the of of these assumptions. The argument made here is probabilities would not apply. that none of the postulates remains inviolate within It may surprise some to learn that ecosystems the domain of ecosystem dynamics, and it is the appear to be rife with singular events. To see why, magnitude of such discrepancies that has forced it helps to recall an argument formulated by phy- the current phenomenological turn toward sicist Walter Elsasser (1969). Elsasser sought describing ecosystems in terms of networks. to delimit what he called an ‘‘enormous’’ number. While descriptions of the scientific method are By this he was referring to numbers so large that legion, one rarely encounters attempts to enumer- they should be excluded from physical consid- ate the fundamental assumptions upon which eration, because they greatly exceed the number of the method is based. One exception is that of physical events that possibly could have occurred Depew and Weber (1994), who articulated four since the Big Bang. To estimate a threshold for ECOLOGICAL NETWORK ANALYSIS 203 enormous numbers Elsasser reckoned the number is positive. Without loss of generality, one may of simple protons in the known universe to be focus attention on a serial, circular conjunction of about 1085. He then noted as how the number of three processes A, B, and C (Figure 15.1). Any nanoseconds that have transpired since the increase in A is likely to induce a corresponding beginning of the universe have been about 1025. increase in B, which in turn elicits an increase in C, Hence, a rough estimate of the upper limit on the and whence back to A. number of conceivable events that could have A didactic example of autocatalysis in ecology is occurred in the physical world is about 10110. Any the community that builds around the aquatic number of possibilities much larger than this value macrophyte, Utricularia (Ulanowicz 1995). All simply loses any meaning with respect to physical members of the genus Utricularia are carnivorous reality. plants. Scattered along its feather-like stems Anyone familiar with combinatorics immedi- and leaves are small bladders, called utricles ately will realize that it does not take very many (Figure 15.2(a)) Each utricle has a few hair-like distinguishable elements or processes before the triggers at its terminal end, which, when touched number of their possible configurations becomes by a feeding zooplankter, opens the end of the enormous. One does not need Avagadro’s number bladder, and the animal is sucked into the utricle of particles (1023) to produce combinations in by a negative that the plant had excess of 10110—a system with merely 80 or so maintained inside the bladder. In nature the sur- distinct components will suffice. In probabilistic face of Utricularia plants is always host to a film of terms, any event randomly comprising more than algal growth known as periphyton. This peri- 80 separate elements is virtually certain to never phyton in turn serves as food for any number have occurred earlier in the history of the universe. of species of small zooplankton. The autocatalytic Such a constellation is unique over all time past. It cycle is closed when the Utricularia captures and follows, then, that in ecosystems with hundreds or absorbs many of the zooplankton (Figure 15.2(b)). thousands of distinguishable organisms, one must In chemistry, where reactants are simple and reckon not just with occasional unique events, but fixed, autocatalysis behaves just like any other rather with a legion of them! Unique, singular events are occurring all the time, everywhere! In the face of this reality, all talk of determinism as a + A + universal characteristic is futile, and the argument C + B for reversibility collapses as well. Despite the challenge that rampant singularities Figure 15.1 A simple example of autocatalysis. pose for the Baconian pursuit of science, it still can be said that a degree of regularity appears to characterize such ecological phenomena as suc- (a) (b) cession. The question then arises as to the origins and maintenance of such order? An agency that both creates and maintains regularities is embed- ded in the patterns of processes that are repres- ented by trophic networks. In particular, the key to how living systems act differently from purely physical systems appears to reside in the adjunc- tion of autocatalytic loops (or cycles of mutualism, that can be found in ecosystem networks) with frequent aleatoric events (Ulanowicz 1997a). Here autocatalysis will be defined as any manifestation of a positive feedback loop whereby the direct Figure 15.2 (a) Utricularia, a . (b) The cycle of effect of every link on its downstream neighbor rewards in the Utricularia system. 204 AQUATIC FOOD WEBS mechanism. As soon as one must contend possible will converge (Figure 15.3). That is, an with organic macromolecules and their ability to autocatalytic loop embedded in a network defines undergo small, incremental alterations, however, itself as the focus of centripetal flows. the game changes. Especially when the effect of It is important to note as how autocatalytic any catalyst on the downstream element is fraught selection pressure is exerted in top–down fashion— with contingencies (rather than being obligatorily that is, action by an integrated cluster of processes mechanical), a number of decidedly nonmechanical upon its constituent elements. Centripetality, in its behaviors can arise (Ulanowicz 1997a). These turn, is best described as an agency that acts at the emergent attributes of complex systems render the focal level. Both selection and centripetality violate remaining Newtonian postulates inappropriate for the restriction of causal closure, which permits only ecosystem dynamics (Ulanowicz 2004). mechanical actions at smaller levels to ramify up Perhaps most importantly, autocatalysis is cap- the hierarchy of scales. In autocatalytic selection, able of exerting selection pressure upon its ever- causal action resembles the final causality of changing, malleable constituents. To see this, one , which was explicitly excluded from considers a small spontaneous change in process B. Newtonian discourse, while centripetality bears all If that change either makes B more sensitive to the trappings of Aristotelian formal cause (by vir- A or a more effective catalyst of C, then the tran- tue of the agency being exerted by a configuration sition will receive enhanced from A. of processes), which concept likewise atrophied in Conversely, if the change in B either makes it less the wake of Newton. sensitive to the effects of A or a weaker catalyst Centripetality also guarantees that whenever of C, then that perturbation will likely receive two or more autocatalytic loops exist in the same diminished support from A. That is to say that there network and draw from the same pool of finite is a preferred direction inherent in autocatalysis— resources, competition among the loci will neces- that of increasing autocatalytic participation. This sarily ensue. In particular, whenever two loops preferred direction can be interpreted as a breaking share pathway segments in common, the result of of symmetry, and such asymmetry, like the sin- this competition is likely to be the exclusion or gular events just discussed, also transcends the radical diminution of one of the nonoverlapping assumption of reversibility. Furthermore, with such sections. For example, should a new element D increasing autocatalytic engagement, or mutual happen to appear and to connect with A and C in adaptation, elements lose their capability of acting parallel to their connections with B, then if D is on their own; or, should they remain capable of more sensitive to A and/or a better catalyst of C, persisting in isolation, it would be with behavior the ensuing dynamics should favor D over B to the radically different from that exhibited as part of extent that B will either fade into the background the autocatalytic scheme. That is, the full cycle or disappear altogether (Figure 15.4). That is, the manifests an organic nature that is refractory to the assumption of atomism. To see how another very important attribute of Centripetality living systems can arise, one notes in particular that any change in B is likely to involve a change in the amounts of material and energy that are required to sustain process B. As a corollary to selection pressure one immediately recognizes the tendency to reward and support any changes that serve to bring ever more resources into B. Because this circumstance pertains to any and all members of the feedback loop, any autocatalytic cycle becomes the epi-center of a centripetal flow of Figure 15.3 Centripetal action as engendered resources toward which as many resources as by autocatalysis. ECOLOGICAL NETWORK ANALYSIS 205

(a) (b) (c) A new metaphysic A A A It begins to appear as though Casti was not exaggerating after all. As Popper (1990) suggested, C B C B C D a wholly new perspective on how things happen in D nature may be required in order to achieve an Figure 15.4 The selection of new element D to replace B. adequate understanding of development and evolution. The topsy-turvy realm of pressure and centripetality generated dynamics must seem strange indeed to those by complex autocatalysis (as embedded in the educated as biologists always to look to smaller macroscopic network of processes) is capable of scales for the causes behind phenomena, but in influencing the replacement of elements. hindsight the appeal of now seems The reader will note the above emphasis upon but a chimera. a causality that arises out of a configuration of pro- In order for reductionism to work, the simplest cesses, and which is able to influence significantly and most enduring elements must all be at the very which objects will remain and which will pass bottom of the spatio-temporal scales, so that the more from the scene. Unlike the conventional attitude complicated and less durable objects can be built up that all events are the consequences of actions by from them. Such is the case with the nested hierarchy objects, the reverse now becomes possible: objects of quarks, mesons, electrons, atoms, and chemical themselves can become the products of constel- compounds. When one reaches the ecosystem, lations of processes. In other words, networks however, one encounters a significant inversion of themselves can become legitimate agents of these assumptions. The ecosystem per se is not as change. Hence, to describe network dynamics, it is well-organized as the individual organisms that no longer mandatory that one search for con- comprise it (Ulanowicz 2001). Furthermore, the net- stituent mechanisms that are facilitated by objects work constituents (and mechanisms) come and go, residing in the nodes. One is now free to develop while the configuration of ecosystems processes what might be called process ecology (Ulanowicz endures (and some would even say, preceded the 2004). current forms of its constituents (Odum 1971)). Finally, it is worthwhile to note how auto- Such inversion notwithstanding, no one should catalytic selection can act to stabilize and regular- regard the causal agency inherent in networks as ize behaviors across the hierarchy of scales. Unlike a triumph for Holism as it was once depicted. with Newtonian universality, a singular event Certainly, no one is contending that configurations anywhere rarely will ramify up and down the of processes fully determine the fate and nature of hierarchy without attenuation. The effects of noise each constituent. One must always bear in mind at one level are usually subject to autocatalytic that singular events loom significantly in the selection at higher levels and to energetic culling at dynamical picture. In the overwhelming number lower levels. Nature as a whole exhibits regular- of cases, however, singular events occur and dis- ities, but in place of the universal effectiveness of appear, leaving no trace upon the overall system all natural laws, one discerns instead a granularity makeup. Occasionally, they will exert detrimental about the real world. That is, models of events at effects upon autocatalytic action, and the system any one scale can explain matters at another scale will respond by reconfiguring itself so as to only in inverse proportion to the remoteness ameliorate such disturbance. In a very small number between them. Obversely, the domain within of instances, a singular event can serve to enhance which irregularities and perturbations can damage the autocatalytic functioning of the system and a system is usually circumscribed. Chance does will become incorporated as an enduring (historical) not necessarily unravel a system, which is held change to the larger network structure (which together by the (flexible) lattice of network inter- thenceforth will exert somewhat different selection actions itself. pressure upon subsequent singular events). 206 AQUATIC FOOD WEBS

The realm of ecosystem behavior is certainly Hirata and Ulanowicz (1984) showed how the different from that of classical mechanical ascendency can then be expressed as dynamics. Instead of a world that is closed, X TijT:: atomistic, reversible, deterministic, and universal; A ¼ Tij log 0, T: T : one now perceives a domain that is (respectively) i;j j i open, organic, historical, contingent, and granular. and Zorach and Ulanowicz (2003) showed how the geometric mean number of roles in the network A network dynamic can be estimated as bA, where b is the base used to calculate the logarithms in the formula for A. If the reader studies closely the scenario described In opposition to this drift toward increasing above, he or she will discern the interplay of two ascendency is the spontaneous tendency to antagonistic tendencies. In one direction there is increase what has been called the network ‘‘over- what might be called a probabilistic drift that head,’’ F. Overhead is the encapsulation of all ratchets the system in a direction of ever-greater ambiguity, incoherence, redundancy, inefficiency, autocatalytic activity. Opposing this drift is the and indeterminacy inherent in the network entropic tendency resulting from the unpredictable (Ulanowicz and Norden 1990). It can be quantified occurrences of singular events, which, on one hand, by an information-theoretic property called the act to disrupt system organization, but on the other ‘‘conditional entropy,’’ which is complementary could also provide a source for diversity and novel to the mutual information that was used to quan- behaviors. Fortunately, these two tendencies can F tify the ascendency. In terms of the Tij, can be both be tracked as changes in quantitative network written as properties. "# X 2 The probabilistic drift toward greater organiza- Tij F ¼ Tij log : tion has long been characterized as ‘‘increasing : : i;j T jTi network ascendency’’ (Ulanowicz 1980, 1986, 1997a). The ascendency of a network is defined as As with the ascendency, Zorach and Ulanowicz the product of its total activity (as measured by the (2003) have demonstrated how the logarithm of sum of all the arc weights) times the average the geometric mean of the network link-density, mutual information inherent in the linkage struc- LD, is equal to one-half of the overhead. That is, f/2 ture (Rutledge et al. 1976; Hirata and Ulanowicz LD ¼ b . (Link-density is the effective number of 1984). This mutual information of the flow struc- arcs entering or leaving a typical node. It is one ture measures, on the average, how definitively measure of the connectivity of the network.) transfers act in the system. That is, if a transfer is Experience shows that the effective numbers of but one of a number of similar, parallel processes, roles and the connectivities of real ecosystems are it contributes little to the mutual information; but not arbitrary. It has long been known, for example, if a process plays a unique role in sustaining that the number of trophic roles (levels) in eco- another node or subgraph, then the contribution of systems is generally fewer than 5 (Pimm and that key link to the mutual information becomes Lawton 1977). Similarly, the effective link-density significant. Zorach and Ulanowicz (2003) showed of ecosystems (and a host of other natural systems) how this latter attribute is captured by the net- almost never exceeds 3 (Pimm 1982; Wagensberg work’s mutual information, which turns out to be et al. 1990). Regarding this last stricture, the logarithm of the effective number of distinct Ulanowicz (2002) suggested how the May–Wigner roles embedded in the network. stability criterion (May 1972) could be reinter- To quantify the ascendency, one must know the preted in information-theoretic terms to identify e/3 magnitude, Tij, of each flow from arbitrary node i a threshold of stability at e , or ca. 3.015 links per to any other node j. The total activity then becomes node. the sum of all the Tij,orT.., where a dot in place of Both limits appear quite visibly when one plots the a subscript indicates summation over that index. number of roles versus the effective connectivity of ECOLOGICAL NETWORK ANALYSIS 207

12

10

8

6

Link density 4 Figure 15.5 Combinations of link densities 2 and numbers of roles pertaining to random networks (open circles) and actual 0 024681012ecosystems networks (solid squares). Number of roles The ‘‘window of vitality’’ is indicated by the dotted lines.

a collection of 44 estimated ecosystem flow networks As one follows the historical dance of an eco- (Figure 15.5). Whereas the pairs of numbers gener- system within the window of vitality, it is import- ated by randomly constructed networks are scat- ant to hold firmly in mind that any description of a tered broadly over the positive quadrant, those trajectory solely in terms of mechanisms and the associated with actual ecosystem networks are con- actions of individual organisms will perforce fined to a small rectangle near the origin. Ulanowicz remain inadequate. Rather, the prevailing agencies (1997b) has labeled this rectangle the ‘‘window of at work are the tendency of configurations of vitality,’’ because it appears that the entire drama of processes (subgraphs) to increase in ascendency ecosystem dynamics plays out within this small acting in opposition to the entropic tendency theatre: as mentioned above, the endogenous generated by complex, singular events. It is only tendency of ecosystems is to drift toward the right by focusing on these larger actors that one can within this window (i.e. toward ever-increasing discover, as Karl Popper (1990) once put it, ascendency, or higher system performance). At any we are not things, but flames. Or a little more prosaically, time, however, singular events can appear as exo- we are, like all cells, processes of metabolism; nets of genous perturbations that shift the network abruptly chemical processes ...(Italics by Popper.) to the left. (Whether the link-density rises or falls during this transition depends upon the nature and That is, by embarking upon a serious examination severity of the disturbance). In particular, whenever of the nature of ecological networks, ecologists are the system approaches one of the outer edges of the not simply climbing trees; they are attempting to window, the probability increases that it will fall go beyond rocket science! back toward the interior. Near the top, horizontal barrier (LD ¼3.015 links per node) the system lacks Acknowledgments sufficient cohesiveness and disintegrates spon- taneously. As the system approaches the right-hand The author was supported in part during the frame (# roles ¼ 4.5 to 5.0), it presumably undergoes writing of this essay by the National Science something like a ‘‘self-organizing catastrophe’’ Foundation’s Program on (Contract (Bak 1996) as described by Holling (1986, see also No. DEB-9981328.) Mr Michael J. Zickel helped Ulanowicz 1997a). with the drawing of Figure 15.5. AFTERWORD A prospectus for future aquatic food web studies

Mathew A. Leibold

Aquatic ecologists have been food web ecologists Many aquatic ecologists (though maybe not all of from an early date. In limnology, Forbes viewed them) would be more puzzled by the lack of lakes as microscosms in which food web interac- indirect effects in aquatic food webs, than they tions determined how entire ecosystems were would be by their presence. structured. Thieneman, Ivlev, Hutchinson and The prevalence and significance of indirect Lindeman continued this perspective in limnology effects, and especially of feedback mediated via as did others who made more comparative studies indirect effects, means that the study of food webs of lakes such as Birge and Juday. Other areas of is almost necessarily the study of complex systems. aquatic ecology have similarly strong traditions Food web scientists are in a continuous struggle to related to food web interactions. Many of the early identify ways to resolve questions about this ideas have been modified, but food web interactions complexity with answers that almost invariably still remain as one of the strongest threads in the simplify potentially crucial elements of it. Thus study of lakes, streams, estuaries, intertidals, and food web diagrams conventionally focus on oceans as well as in the smaller phytotelmata and in tropho-species (groups of organisms potentially the lab studies with . involving multiple species, that have unique sets For most researchers who have studied food of links to other such tropho-species). This sim- web interactions in aquatic systems, it is hard not plifies many theoretical and empirical aspects of to be conscious of just how important food web food web research but theory tells us that quanti- interactions can be in shaping communities and tative differences among the species that are ecosystems. Not only are the magnitude of some members of the same tropho-species could be just food web interactions in all of these systems large, as important in affecting food web phenomena as but there are numerous cases where any given are the qualitative differences implied by the con- trophic interaction has strong and obvious ramifica- cept of tropho-species. Similarly, trophic relations tions that alter entire suites (or cascades) of are not the only important ones in real commu- ecological processes and the populations of nities that affect food web phenomena. Never- organisms that are often quite distant (in causal theless food web studies are among those that terms) from the modified interaction. While there most directly and realistically confront the com- may be reasons to think that indirect effects may plexity that exists in most natural ecosystems. have limits on their propagation, it is clear from Work to date on aquatic food webs shows that it is many studies, especially those conducted in possible to gain insights about important ecologi- aquatic systems, that they are an important part of cal phenomena without so drastically simplifying the overall dynamics of ecosystems. Regardless of this complexity that our theories seem like pale the form that it takes, the study of food webs is a caricatures of natural systems. reflection on the causes and consequences of these As an example, food web studies on the indirect indirect effects in networks of interacting species. interactions and feedback between ‘bottom-up’

208 AFTERWORD 209 and ‘top-down’ food web interactions seem to be – How do food webs and consequently their critical to our current explanations of trophic effects on ecosystems respond to environmental structure in aquatic systems. Surveys of plant and changes? herbivore biomass in lakes supported a ‘bottom- As illustrated by the contributions to this book, up’ perspective because they showed that both the answers to these questions in aquatic systems increased in lakes with higher levels of limiting are starting to resolve themselves with novel ideas nutrients and consequent productivity. This strong about these issues, novel ways to evaluate these pattern however was in conflict with numerous ideas, and novel insights about how these ideas experiments and other studies that showed strong relate to other phenomena in ecology, biology, and top-down regulation of both plant and herbivore environmental sciences as a whole. biomass when fish stocks were altered. Theory Nevertheless food web ecology seems to be an indicated that these two observations were largely area that continuously generates as many ques- incompatible in simple food chains with only one tions as it answers. An important aspect of this species per trophic level. More recent work indi- book is that the contributions suggest many new cates that it is the more complex set of interactions directions that could or should be taken in future and feedbacks that occur in food webs that can work. There seem to be at least three important resolve this apparent contradiction where there are issues that emerge: multiple species involved in trophic transfers at First, how much can we extrapolate from aquatic any single trophic level. As in many other cases systems to other systems? Some authors have where paradoxical patterns have been found in argued that such extrapolation is unwarranted for aquatic ecosystems, it was the food web perspect- example in the study of trophic cascades. While ive that allowed for the synthesis and resolution of they recognized how significant the finding of the problem. trophic cascades in aquatic systems were, they Additionally, many other aspects of food web argued that trophic cascades were less likely to be structure and dynamics are perhaps better docu- important in terrestrial systems because terrestrial mented in aquatic systems than in terrestrial ones. food webs were claimed to be more complex, with While terrestrial ecologists may (arguably) still be individual populations that were more highly able to make significant progress taking strict structured, and with more complications arising ‘bottom up’ perspectives that ignore how ‘top- from factors such as omnivory etc. Clearly, such down’ effects work, the simultaneous action and claims would come as a surprise to many aquatic interaction of top-down and bottom-up interac- ecologists to begin with, but additionally such tions in aquatic systems is hard to deny. Aquatic claims beg the question of why such differences systems provide some of the best examples for might exist and why they would negate the import- many of the ideas that have emerged from con- ance of trophic cascades. Since then, examples of sideration of these effects. The contributions in this trophic cascades in terrestrial systems have been book continue in this vein and address some of the found. And while it is clear that the effects can most important current questions about food differ in magnitude from some of the findings in webs: lakes, it is also clear that there is tremendous – How do structural properties of food webs affect variation in trophic cascades in aquatic systems as the dynamic properties of the ecosystems they are well. Somehow, there is a need to better investigate a part of? how and why some patterns extrapolate well from – How can we explain the amazing complexity of aquatic systems to terrestrial ones and others do food webs? not. Part of resolving this has to be in identifying – How are food webs assembled by processes of the ways complex processes interact and this can colonization and extinction? really only be resolved with a strong theoretical – How does this assembly affect other dependent framework. This book goes a long way in this properties of the ecosystem? direction by having a strong basis in ecological 210 AFTERWORD theory but this is an area that still requires further might be important. The first involves adaptive development. plasticity, the second involves adaptive genetic Second, at what scale do different interpreta- changes within populations of a single species, and tions of food web dynamics apply? Clearly, food the third involves changes in species composition web models that are built on locally interacting that result in adaptation in the sense that they yield an populations in closed systems can make predic- improved correspondence between the traits of the tions that differ strongly from food web models organisms found in a food web and the environment that allow for spatial structure or for other meta- in which the food web occurs. These three types of community level processes to occur. There is, for adaptive dynamics are all likely to be important, at example, a much smaller problem in explaining least under some scenarios, and they are likely to the diversity and complexity of food webs at larger interact with each other. While the importance of spatio-temporal scales where such metacommu- non-random organization of food webs has been nity dynamics exist than at the smaller scale typi- recognized since the work of May and Pimm, it seems fied by models of locally interacting populations. likely that more careful thinking about how these A number of the contribution to this book address three processes affect food webs will improve our some elements of this issue, but we are still a long understanding of food web dynamics. way from understanding the dynamics of food Aquatic ecologists as a group are likely to make webs at larger spatio-temporal scales even though among the strongest contributions to food web this is maybe of greater ultimate relevance. Fur- thinking. In part this is because some aquatic thermore, there still seem to be significant differ- systems are ‘model systems’ for the study of food ences in how food web scientists interpret the links web dynamics (e.g., streams, intertidal, lakes and between food web attributes and principles that ponds), in part this is because it is possible to apply to ecological processes that operate over conduct experiments that involve major aspects of different scales. realistic food webs (e.g., work with protists in Finally, one element that is only beginning to be microcosms, mesocosms in lakes and streams, cage addressed is the role of evolutionary or adaptive experiments in intertidals), and in part because dynamics in food webs. A number of the con- aquatic systems span the gamut of complexity tributions here deal either directly or indirectly ranging from relatively simple phytotelmata to with some form of adaptive dynamics (in the entire oceans. This book illustrates that answers to broadest sense of the word) and it seems clear that complex questions that arise from food web ecol- there are important implications for food web ogy are most satisfying when they appear useful in theory and for future food webs studies. There are the interpretation of phenomena across such at least three types of adaptive dynamics that diverse systems and using such diverse methods. References

Introduction Williams, R. J., and N. D. Martinez. 2000. Simple rules yields complex food webs. Nature 404: 180–183. Azam, F., and A. Z. Worden. 2004. Microbes, molecules, Williams, R. J., E. L. Berlow, J. A. Dunne, A.-L. Baraba´si, Science 303 and marine ecosystems. : 1622–1624. and N. D. Martinez. 2002. Two degrees of separation Azam, F. 1998. Microbial control of oceamic carbon flux: in food webs. Proceedings of the National Academy of the plot thickens. Science 280: 694–696. Sciences USA 99: 12913–12917. Food Webs and Niche Space Cohen, J. E. 1998. . Princeton Wulff, F., J. G. Field, and K. H. Mann. (eds). 1989. Network University Press, Princeton, NJ. Analysis in Marine Ecology. Springer-Verlag, New York, Christensen, V. and D. Pauly. (eds). 1993. Trophic Models p. 284. of Aquatic Ecosystems. ICLARM, Manila, p. 390. Dunne, J. A., R. J. Williams, and N. D. Martinez. 2002a. Network structure and biodiversity loss in food webs: Chapter 1 robustness increases with connectance. Ecology Letters Andersen, T. 1997. Pelagic Nutrient Cycles: Herbivores as 5 : 558–567. Sources and Sinks. Springer-Verlag, Berlin, Heidelberg, Dunne, J. A., R. J. Williams, and N. D. Martinez. 2002b. New York. Food-web structure and network theory: the role Andersen, T., and D. O. Hessen. 1991. Carbon, nitrogen, of connectance and size. Proceedings of the National and phosphorus content of freshwater zooplankton. 99 Academy of Sciences USA : 12917–12922. Limnology and Oceanography 36: 807–814. Garlaschelli, D., G. Caldarelli, and L. Pietronero. 2003. Baudouin-Cornu, P., Y. Surdin-Kerjan, P. Marliere, Universal scaling relations in food webs. Nature and D. Thomas. 2001. Molecular evolution of protein 423: 165–168. atomic composition. Science 293: 297–300. MacArthur, R. H. 1955. Fluctuations of animal popula- Carrillo, P., M. Villar-Argaiz, and J. M. Medina-Sanchez. tions, and a measure of community stability. Ecology 2001. Relationship between N : P ratio and growth 36: 533–536. rate during the life cycle of copepods: an in situ May, R. M. 1973. Stability and Complexity in Model measurement. Journal of Plankton Research 23: 537–547. Ecosystems, 2nd edn. Princeton University Press, Cebrian, J. 1999. Patterns in the fate of production in Princeton, NJ. plant communities. American Naturalist. 154: 449–468. Paine, R. T. 1966. Food web complexity and species Cebrian, J., and C. M. Duarte. 1995. Plant growth-rate diversity. American Naturalist 100: 65–75. dependence of detrital carbon storage in ecosystems. Pimm, S. L. 1982. Food Webs. Chapman and Hall, London. Science 268: 1606–1608. Polis, G. A., and K. O. Winemiller. 1996. Food Webs: Cebrian, J., M. Williams, J. McClelland, and I. Valiela. Integration of Patterns and Dynamics. Chapman and Hall. 1998. The dependence of heterotrophic consumption Ulanowicz, R. E. 1997. Ecology, The Ascendent Perspective. and C accumulation on autotrophic nutrient content in Columbia University Press, New York, 224 pp. ecosystems. Ecology Letters 1: 165–170. Ulanowicz, R. E., and L. G. Abarca-Arenas. 1997. Darchambeau, F., P. J. Faerøvig, and D. O. Hessen. 2003. An informational synthesis of ecosystem structure How Daphnia copes with excess carbon in its food. and function. Ecology Modelling 95: 1–10. Oecologia 136: 336–346. Ulanowicz, R. E. 1996. The propensities of evolving DeMott, W. R. 2003. Implications of element deficits for systems. In: E. L. Khalil and K. E. Boulding (eds), zooplankton growth. Hydrobiologia 491: 177–184. Evolution, Order and Complexity. Routledge Publishers, DeMott, W. R., and R. D. Gulati. 1999. Phosphorus London, pp. 217–233. limitation in Daphnia: evidence from a long-term

211 212 REFERENCES

study of three hypereutrophic Dutch lakes. Limnology Fussmann, G. F., S. P. Ellner, and J. Hairston, N. G. 2003. Oceanography 44: 1557–1564. Evolution as a critical component of plankton DeMott, W. R., R. D. Gulati, and K. Siewertsen. 1998. dynamics. Proceedings of the Royal Society of London Effects of phosphorus-deficient diets on the carbon and Series B 270: 1015–1022. phosphorus balance of . Limnology and Gorokhova, E., T. A. Dowling, L. J. Weider, T. Crease, Oceanography 43: 1147–1161. and J. J. Elser. 2002. Functional and ecological Diehl, S. 2003. The evolution and maintenance of significance of rDNA IGS variation in a clonal omnivory: dynamic constraints and the role of food organism under divergent selection for production quality. Ecology 84: 2557–2567. rate. Proceedings of the Royal Society of London B 269: Dykhuizen, D. E., A. M. Dean, and D. L. Hartl. 1987. 2373–2379. Metabolic flux and fitness. 115: 25–31. Hairston, J., N. G., and N. G. Hairston, Sr. 1993. Eickbush, T. H., W. D. Burke, D. G. Eickbush, and Cause–effect relationships in energy flow, trophic W. C. Lathe. 1997. Evolution of R1 and R2 in the structure, and interspecific interactions. American rDNA units of the genus Drosophila. Genetica 100: Naturalist 142: 379–411. 49–61. Hassett, R. P., B. Cardinale, L. B. Stabler, and Elser, J. J., and J. Urabe. 1999. The stoichiometry of J. J. Elser. 1997. Ecological stoichiometry of N and consumer-driven nutrient cycling: theory, observa- P in pelagic ecosystems: comparison of lakes and tions, and consequences. Ecology 80: 735–751. oceans with emphasis on the zooplankton–- Elser, J. J., K. Acharya, M. Kyle, J. Cotner, W. Makino, phytoplankton interaction. Limnology and Oceanography T. Markow, T. Watts, S. Hobbie, W. Fagan, J. Schade, 41: 648–662. J. Hood, and R. W. Sterner. 2003. Growth Hastings, A., and M. Conrad. 1979. Length and rate–stoichiometry couplings in diverse biota. Ecology evolutionary stability of food chains. Nature 282: Letters 6: 936–943. 838–839. Elser, J. J., D. Dobberfuhl, N. A. MacKay, and Hessen, D. O. 1992. Nutrient element limitation of J. H. Schampel. 1996. Organism size, life history, zooplankton production. American Naturalist 140: and N : P stoichiometry: towards a unified view of 799–814. cellular and ecosystem processes. BioScience 46: Hessen, D. O., and B. Bjerkeng. 1997. A model approach 674–684. to planktonic stoichiometry and consumer–resource Elser, J. J., W. F. Fagan, R. F. Denno, D. R. Dobberfuhl, stability. 38: 447–471. A. Folarin, A. Huberty, S. Interlandi, S. S. Kilham, Hessen, D. O., and A. Lyche. 1991. Inter- and intraspecific E. McCauley, K. L. Schulz, E. H. Siemann, and variations in zooplankton element composition. R. W. Sterner. 2000a. Nutritional constraints in terrestrial Archives of 121: 343–353. and freshwater food webs. Nature 408: 578–580. Hessen, D. O., and J. Skurdal. 1987. Food consumption, Elser, J. J., H. Hayakawa, and J. Urabe. 2001. Nutrient turnover rates and assimilation in the noble crayfish limitation reduces food quality for zooplankton: (Astacus astacus). Freshwater Crayfish 7: 309–317. Daphnia response to seston phosphorus enrichment. Hessen, D. O., B. A. Faafeng, and P. Brettum. Ecology 82: 898–903. 2003. Autotroph: herbivore biomass ratios; carbon Elser, J. J., R. W. Sterner, E. Gorokhova, W. F. Fagan, deficits judged from plankton data. Hydrobiologia 491: T. A. Markow, J. B. Cotner, J. F. Harrison, S. E. Hobbie, 167–175. G. M. Odell, and L. J. Weider. 2000b. Biological stoi- Hessen, D. O., P. J. Faerovig, and T. Andersen. 2002. chiometry from genes to ecosystems. Ecology Letters Light, nutrients, and P : C ratios in algae: Grazer 3: 540–550. performance related to food quality and quantity. Faerovig, P. J., T. Andersen, and D. O. Hessen. 2002. Ecology 83: 1886–1898. Image analysis of Daphnia populations: non-destructive Hobbie, S., Q. Hong, D. A. Beard, and S. Liang. 2003. determination of demography and biomass in cultures. Stoichiometric network theory for nonequilibrium Freshwater Biology 47: 1956–1962. biochemical systems. European Journal of Biochemistry Fagan, W. F., E. H. Siemann, R. F. Denno, C. Mitter, 270: 415–421. A. Huberty, H. A. Woods, and J. J. Elser. 2002. Nitrogen Holt, R. D. 1995. Linking species and ecosystems: where’s in insects: implications for trophic complexity Darwin? In: C. G. Jones and J. H. Lawton (eds), Linking and species diversification. American Naturalist Species and Ecosystems. Chapman and Hall, New York, 160: 84–802. pp. 273–279. REFERENCES 213

Jaenike, J., and T. A. Markow. 2002. Comparative Munch, S. B., and D. O. Conover. 2003. Rapid growth elemental stoichiometry of ecologically diverse results in increased susceptibility to predation in Drosophila. 17: 115–120. Menidia menidia. Evolution 57: 2119–2127. Lavigne, D. M. 1996. Ecological interactions between Murdoch, W. W., C. J. Briggs, and R. M. Nisbet. 2003. marine mammals, commercial fisheries, and their prey: Consumer–Resource Dynamics. Princeton University unravelling the tangled web. In: W. A. Montevecchi Press, Princeton, NJ. (ed.) Studies of High-Latitude Seabirds. 4. Trophic Perkins, M. C., H. A. Woods, J. F. Harrison, and J. J. Elser. Relationships and Energetics of Endotherms in Cold Ocean 2003. Dietary phosphorus affects the growth of larval Systems. Occasional Paper Number 91. Canadian Manduca sexta. Archives of Insect Biochemistry and Wildlife Service, St John’s Newfoundland. Physiology 55: 153–168. Lima, S. L., and L. M. Dill. 1990. Behavioral decisions Plath, K., and M. Boersma. 2001. limitation of made under the risk of predation—a review and zooplankton: stoichiometric constraints and optimal prospectus. Canadian Journal of Zoology—Reviews in foraging. Ecology 82: 1260–1269. Canadian Zoology 68: 619–640. Schade, J., M. M. Kyle, S. S. Hobbie, W. W. Fagan, and Loladze, I., Y. Kuang, and J. J. Elser. 2000. Stoichiometry J. J. Elser. 2003. Stoichiometric tracking of soil nutrients in producer–grazer systems: linking energy flow and by a desert insect herbivore. Ecology Letters 6: 96–101. element cycling. Bulletin of Mathematical Biology Schulz, K. L., and R. W. Sterner. 1999. Phytoplankton 62: 1137–1162. phosphorus limitation and food quality for Bosmina. Loladze, I., Y. Kuang, J. J. Elser, and W. F. Fagan. 2004. Limnology Oceanography 44: 1549–1556. Competition and stoichiometry: coexistence of two Sommer, U. 1992. Phosphorus-limited Daphnia: intra- predators on one prey. Journal of Theoretical Biology specific facilitation instead of competition. Limnology 65: 1–15. Oceanography 37: 966–973. Lotka, A. J. 1925. Elements of Physical Biology. Williams Sterner, R. W. 1986. Herbivores’ direct and indirect and Wilkins, Baltimore, MD. effects on algal populations. Science 231: 605–607. Loxdale, H. D., and G. Lushai. 2003. Rapid changes in Sterner, R. W. 1995. Elemental stoichiometry of species in clonal lines: the death of a ‘‘sacred cow.’’ Biological ecosystems. In: C. Jones and J. Lawton (eds), Linking Journal of the Linnean Society 79: 3–16. Species and Ecosystems. Chapman and Hall, New York, Lushai, G., and H. D. Loxdale. 2002. The biological pp. 240–252. improbability of a clone. Genetics Research 79: 1–9. Sterner, R. W., and J. J. Elser. 2002. Ecological Stoichiometry: Lushai, G., H. D. Loxdale, and J. A. Allen. 2003. The The Biology of Elements from Molecules to the Biosphere. dynamic clonal genome and its adaptive potential. Princeton University Press, Princeton, NJ. Biological Journal of the Linnean Society 79: 193–208. Sterner, R. W., and D. O. Hessen. 1994. Algal nutrient Main, T., D. R. Dobberfuhl, and J. J. Elser. 1997. N : P limitation and the nutrition of aquatic herbivores. stoichiometry and ontogeny in crustacean zooplankton: Annual Review of Ecological Systems 25: 1–29. a test of the growth rate hypothesis. Limnology and Sterner, R. W., J. J. Elser, E. J. Fee, S. J. Guildford, and Oceanography 42: 1474–1478. T. H. Chrzanowski. 1997. The light : nutrient ratio Makino, W., and J. B. Cotner. 2003. C : N : P stoichio- in lakes: the balance of energy and materials affects metry of a heterotrophic bacterial community in ecosystem structure and process. American Naturalist a freshwater lake: implications for substrate- and 150: 663–684. growth-dependent variations. Aquatic Sullender, B. W., and T. J. Crease. 2001. The behavior of a 34: 33–41. Daphnia pulex transposable element in cyclically and Makino, W., J. B. Cotner, R. W. Sterner, and J. J. Elser. obligately parthenogenetic populations. Journal of 2003. Are bacteria more like plants or animals? Growth Molecular Evolution 53: 63–69. rate and substrate dependence of bacterial C : N : P Sutcliffe, W. H. 1970. Relationship between growth rate stoichiometry. Functional Ecology 17: 121–130. and ribonucleic acid concentration in some inverte- Markow, T. A., B. Raphael, D. Dobberfuhl, brates. Journal of the Fisherics Research Board of Canada C. M. Breitmeyer, J. J. Elser, and E. Pfeiler. 1999. 27: 606–609. Elemental stoichiometry of Drosophila and their hosts. Tilman, D. 1982. Resource Competition and Community Functional Ecology 13: 78–84. Structure. Princeton University Press, Princeton, NJ. Muller, E., R. M. Nisbet, S. A. L. M. Kooijman, J. J. Elser, Turchin, P. 2003. Complex Population Dynamics. A Theore- and E. McCauley. 2001. Stoichiometric food quality tical/Empirical Synthesis. Princeton University Press, and herbivore dynamics. Ecology Letters 4: 519–529. Princeton, NJ. 214 REFERENCES

Urabe, J., and R. W. Sterner. 1996. Regulation of Caswell, H., and J. E. Cohen. 1993. Local and regional herbivore growth by the balance of light and nutrients. regulation of species–area relations: a patch-occupancy Proceedings of the National Academy of Sciences, USA model. In: R. E. Ricklefs and D. Schluter (eds), Species 93: 8465–8469. Diversity in Ecological Communities. The University of Urabe, J., J. J. Elser, M. Kyle, T. Sekino, and Z. Kawabata. Chicago Press. Chicago, IL. 2002a. Herbivorous animals can mitigate unfavorable Cottenie, K., E. Michelis, N. Nuytten, and L. De Meester. ratios of energy and material supplies by enhancing 2003. Zooplankton metacommunity structure: regional nutrient recycling. Ecology Letters 5: 177–185. vs. local processes in highly interconnected ponds. Urabe, J., M. Kyle, W. Makino, T. Yoshida, T. Andersen, Ecology 84: 991–1000. and J. J. Elser. 2002b. Reduced light increases herbivore Cottenie, K., and L. De Meester. 2004. Metacommunity production due to stoichiometric effects of light: structure: synergy of biotic ineractions as selective nutrient balance. Ecology 83: 619–627. agents and dispersal as fuel. Ecology 85: 114–119. Vadstein, O., B. O. Harkjerr, A. Jensen, Y. Olsen, and Dunne, J. A., R. J. Williams, and N. D. Martinez. 2002. H. Reinertsen. 1989. Cycling of organic carbon in Food web structure and network theory: the role the of a eutrophic lake with special of connectance and size. Proceedings of the National reference to the heterotrophic bacteria. Limnology and Academy of Sciences, USA 99: 12917–12922. Oceanography 34: 840–855. Fisher, W.(ed). 1978. FAO Species Identification Sheets for Varma, A., and B. O. Palsson. 1994. Metabolic flux Purposes. Western Central Atlantic (Fishing area balancing: basic concepts, scientific and practical use. 31). Vol I–VII. Food and Agriculture Organization of Bio/Technology 12: 994–998. the United Nations, Rome. Vrede, T., J. Persson, and G. Aronsen. 2002. The influence Froese, R., and D. Pauly. 2003. FishBase World Wide Web of food quality (P : C ratio) on RNA : DNA ratio electronic publication. http://fishbase.org, version and somatic growth rate of Daphnia. Limnology and 24 Septemper 2003. Oceanography 47: 487–494. Gotelli, N. J. 1991. Metapopulation models: the rescue Weider, L. J., K. L. Glenn, M. Kyle, and J. J. Elser. 2004. effect, the propagule rain, and the core-satellite hypo- Associations among ribosomal (r)DNA intergenic thesis. American Naturalist 138: 768–776. spacer length, growth rate, and C : N : P stoichiometry Hanski, I. 1983. Coexistence of competitors in a patchy in the genus Daphnia. Limnology and Oceanography 49: environment. Ecology 64: 493–500. 1417–1423. Hanski, I., and M. E. Gilpin. (eds). 1997. Metapopulation Bio- Wildermuth, M. C. 2000. Metabolic control analysis: logy: Genetics, and Evolution. Academic Press, San Diego. biological applications and insights. Genome Biology Hastings, A. 1980. Disturbance, coexistence, history and 1: Reviews 1031.1–Reviews 1031.5. the competition for space. Theoretical Yoshida, T., L. E. Jones, S. P. Ellner, G. F. Fussmann, 18: 363–373. and J. Hairston, N. G. 2003. Rapid evolution drives Holt,R.D.1996.Foodwebsinspace:anislandbiogeo- ecological dynamics in a predator–prey system. Nature graphic perspective. In: G. A.Polis,andK.O. Winemiller 424: 303–306. (eds), Food Webs: Integration of Patterns and Dynamics. Chapman and Hall, London. Chapter 2 Holt, R. D. 1997. From metapolulation dynamics to com- munity structure: some consequences of spatial hetero- Bascompte, J., and R. V. Sole´. (eds). 1998. Modeling geneity. In: I. Hanski, and M. Gilpin (eds), Metapopulation Spatiotemporal Dynamics in Ecology. Springer, Berlin. Biology. Academic Press, San Diego, pp. 149–164. Bascompte, J., C. J. Melia´n, and E. Sala. 2004. Interaction Hori, M., and T. Noda. 2001. Spatio-temporal variation strength motifs and the overfishing of marine food of avian foraging in the rocky intertidal food web. webs. (submitted). Journal of Animal Ecology 70: 122–137. Caldarelli, G., P. G. Higgs, and A. J. McKane. 1998. Hubbell, S. P. 2001. The Unified Neutral Theory of Bio- Modelling coevolution in multispecies communities. diversity and . Princeton University Press, Journal of Theoretical Biology 193: 345–358. Princeton, NJ. Carr, M. H., T. W. Anderson, and M. A. Hixon. Jennings, S., and S. Mackinson. 2003. Abundance–body 2002. Biodiversity, population regulation, and the mass relationships in size-structured food webs. stability of coral-reef fish communities. Proceedings Ecology Letters 6: 971–974. of the National Academy of Sciences, USA, 99: Kareiva, P. 1987. and the stability 11241–11245. of predator–prey interactions. Nature 326: 388–390. REFERENCES 215

Karlson, R. H., and H. V. Cornell. 2002. Species richness Tilman, D. 1994. Competition and biodiversity in of coral assemblages: detecting regional influences at spatially structured habitats. Ecology 75: 2–16. local spatial scales. Ecology 83: 452–463. Tilman, D., and P. Kareiva. (eds). 1997. . Kneitel, J. M., and T. E. Miller. 2003. Dispersal rates The Role of Space in Population Dynamics and inter- affect species composition in metacommunities of specific Interactions. Princeton University Press, Sarracenia purpurea . American Naturalist Princeton, NJ. 162: 165–171. Wilson, D. S. 1992. Complex interactions in meta- Levins, R. 1969. Some demographic and genetic con- communities, with implications for biodiversity and sequences of environmental heterogeneity for bio- higher levels of selection. Ecology 73: 1984–2000. logical control. Bulletin of the Entomological Society of America 15: 237–240. Chapter 3 Loreau, M., Mouquet, N., and R. D. Holt. 2003. Meta- ecosystems: a theoretical framework for a spatial eco- Almunia, J., G. Basterretxea, J. Arı´stegui, and system ecology. Ecology Letters 6: 673–679. R. E. Ulanowicz. 1999. Benthic–pelagic switching in Margalef, R. 1963. On certain unifying principles in a coastal subtropical lagoon. Estuarine, Coastal and Shelf ecology. American Naturalist 97: 357–374. Science 49: 363–384. Melia´n, C. J., and J. Bascompte. 2002. Food web structure Baird, D. 1998. Orientors and ecosystem properties and habitat loss. Ecology Letters 5: 37–46. in coastal zones. In: F. Muller and M. Leupelt (eds), Melia´n, C. J., and J. Bascompte, 2004. Food web cohesion. Ecotargets, Goal Functions, and Orientors. Springer- Ecology 85: 352–358. Verlag, Heidelberg. pp. 232–242. Miklo´s, I., and J. Podani. 2004. Randomization of Baird, D. and J. J. Heymans. 1996. Assessment of presence–absence matrices: comments and new the ecosystem changes in response to freshwater algorithms. Ecology 85: 86–92. inflow of the Kromme River Estuary, St Francis Bay, Mouquet, N., and M. Loreau. 2002. Coexistence in South Africa: a network analysis approach. Water SA metacommunities: the regional similarity hypothesis. 22: 307–318. American Naturalist 159: 420–426. Baird, D., J. M. Mcglade, and R. E. Ulanowicz. 1991. Mouquet, N., and M. Loreau. 2003. Community patterns The comparative ecology of six marine ecosystems. in source–sink metacommunities. American Naturalist Philosphical Transactions of the Royal Society London B 162: 544–557. 333: 15–29. Mumby, P. J., A. J. Edwards, J. E. Arias-Gonza´lez, Baird, D., and R. E. Ulanowicz. 1989. Seasonal dynamics K. C. Lindeman, P. G. Blackwell, et al. 2004. of the Chesapeake Bay ecosystem. Ecological Mono- Mangroves enhance the biomass of coral reef fish graphs 59: 329–364. communities in the Caribbean. Nature 427: 533–536. Baird, D. and R. E. Ulanowicz. 1993. Comparative study Opitz, S. 1996. Trophic Interactions In Caribbean Coral on the trophic structure, cycling and ecosystem pro- Reefs. Technical Report 43. International Centre for perties of four tidal estuaries. Marine Ecology Progress Living Aquatic Resource Management (now World- Series 991: 221–237. Fish Center), Penang, Malaysia. Baird, D., R. E. Ulanowicz, and W. R. Boynton. 1995. Polis, G. A., W. B. Anderson, and R. D. Holt. 1997. Seasonal nitrogen dynamics in the Chesapeake Bay: Toward an integration of landscape and food web a Network Approach. Estuarine, Coastal and Shelf ecology: the dynamics of spatially subsidized food Science 41: 137–162. webs. Annual Review of Ecology and Baird, D., J. Luczkovich, and R. R. Christian. 1998. 28: 289–316. Assessment of spatial and temporal variability in eco- Randall, J. E. 1967. Food habits of reef fishes of the West system attributes of the St Marks National Wildlife Indies. Studies in Tropical Oceanography 5: 665–847. Refuge, Apalachee Bay, Florida. Estuarine, Coastal and Ricklefs, R. E. 1987. Community diversity: relative roles Shelf Science 47: 329–349. of local and regional processes. Science 235: 167–171. Baird, D., R. R. Christian, C. H. Peterson, and G. Johnson. Shurin, J. B. 2001. Interactive effects of predation and 2004. Consequences of hypoxia on estuarine ecosystem dispersal on zooplankton communities. Ecology function: energy diversion from consumers to 82: 3404–3416. microbes. Ecological Applications 14: 805–822. Thompson, J. N. 1999. Specific hypotheses on the Batagelj, V., and A. Mrvar. 2002. Pajek 0.85 Network geographic mosaic of coevolution. American Naturalist Software. Available at: http://vlado.fmf.uni-lj.si/pub/ 153: S1–S14. networks/pajek/ 216 REFERENCES

Borgatti, S. P., M. G. Everett, and L. C. Freeman. 2002. Finn, J. T., and T. Leschine. 1980. Does ferti- Ucinet for Windows: Software for Analysis. lization enhance shellfish production? An application Analytic Technologies, Harvard. of flow analysis. Environmental Management 4: 193–203. Brando, V. E., R. Ceccarelli, S. Libralato, and G. Ravagnan. Gordon, D. C., P. R., Boudreau, K. H., Mann, J.-E. Ong, 2004. Assessment of environmental management W. L. Silvert, S. V. Smith, G. Wattayakom, F. Wulff, effects in a shallow water basin using mass-balance and T. Yanagi. 1996. LOICZ Biogeochemical Modelling models. Ecological Modelling 172: 213–232. Guidelines. LOICZ Reports and Studies No. 5, 96 p. Christensen, V. 1995. Ecosystem maturity—towards Heymans, J. J., and D. Baird. 2000. Network analysis quanitification. Ecological Modelling 77: 3–32. of the northern Benguela ecosystem by means of Christensen, V., and D. Pauly. 1992. ECOPATH II—a NETWRK and ECOPATH. Ecological Modelling software for balancing steady-state ecosystem models 131: 97–119. and calculating network characteristics. Ecological Holligan, P. M., and H. de Boois. (eds). 1993. Land-Ocean Modelling 61: 169–185. Interactions in the Coastal Zone [LOICZ] Science Plan. Christensen, V., and D. Pauly. (eds). 1993. Trophic Models IGBP Report No. 25, 50 pp. of Aquatic Ecosystems. ICLARM Conference Proceedings Johnson, J. C., S. P. Borgatti, J. J. Luczkovich, and 26, Manila, Phillipines. M. G. Everett. 2001. Network role analysis in the study Christensen, V., and D. Pauly. 1995. Fish production, of food webs: an application of regular role coloration. catches and the carrying capacity of the world oceans. Journal of 2(3). Naga, The ICLARM Quarterly 18: 34–40. Johnson, J. C., J. Boster, and L. Palinkas. 2003. Social roles Christian, R. R., and J. J. Luczkovich. 1999. Organizing and and the evolution of networks in isolated and extreme understanding a winter’s seagrass food web network through environments. The Journal of Mathematical Sociology effective trophic levels. Ecological Modelling 117: 99–124. 27: 89–122 Christian, R. R., and R. E. Ulanowicz. 2001. Network Kay, J.J., L. A.Graham, and Ulanowicz, R. E. 1989. A detailed ecology. In: A. El-Shaarawi and W. W. Pierogorsch guide to network analysis. In: F. Wulff, J. G. Field, and (eds), Encyclopedia of Environmetrics, Vol. 3. John Wiley K. H. Mann (eds), Network Analysis in Marine Ecology. and Sons Ltd, Chichester, UK, pp. 1393–1399. Springer Verlag, Berlin, Heidelberg, pp. 15–61. Christian, R. R., and C. R. Thomas. 2003. Network ana- Kremer, J. N. 1989. Network information indices with lysis of nitrogen inputs and cycling in the Neuse River an estuarine model. In: F. Wulff, J. G. Field, and estuary, North Carolina, USA. Estuaries 26: 815–828. K. H. Mann (eds), Network Analysis in Marine Ecology: Christian, R. R., E. Fore`s, F. Comı´n, P. Viaroli, M. Naldi, Methods and Applications. Coastal and Estuarine Studies and I. Ferrari. 1996. Nitrogen cycling networks of coastal 32, Springer-Verlag, Heidelberg, pp. 119–131. ecosystems: influence of trophic status and primary Lin, H-.J., J-.J. Hung, K-.T. Shao, and F. Kuo. 2001. Trophic producer form. Ecological Modelling 87: 111–129. functioning and nutrient flux in a highly productive Christian, R. R., J. K. Dame, G. Johnson, C. H. Peterson, tropical lagoon. Oegologia 129: 395–406. and D. Baird. 2004. Monitoring and Modeling of the Luczkovich, J. J., S. P. Borgatti, J. C. Johnson, and Neuse River Estuary, Phase 2: Functional Assessment of M. G. Everett. 2003. Defining and measuring trophic Environmental Phenomena through Network Analysis. role similarity in food webs using regular equivalence. Final Report, UNC Research Institute, The Journal of Theoretical Biology 220: 303–321. Raleigh, North Carolina, USA. Report No. 343-E. Monaco, M. E., and R. E. Ulanowicz. 1997. Comparative Christensen, V., C. J. Walters, and D. Pauly. 2000. ECOPATH ecosystem trophic structure of three U. S. mid-Atlantic with ECOSIM: A Users Guide, October 2000 edn. Fisheries estuaries. Marine Ecology Progress Series 161: 239–254. Centre, University of British Columbia, Vancouver, Odum, E. P. 1969. The strategy of ecosystem develop- Canada, and International Center for Living Aquatic ment. Science 164: 262–270. Resources Management, Penang, Malaysia, 130 pp. Pauly, D., and V. Christensen. 1995. Primary produc- Elmgren, R. 1989. Man’s impact on the ecosystem of the tion required to sustain global fisheries. Nature Baltic Sea: energy flows today and at the turn of the 374: 255–257. century. Ambio 18: 326–332. Pernetta, J. C., and J. D. Milliman. (eds). 1995. Land– Fath, B. D., and B. C. Patten. 1999. Review of the Ocean Interactions in the Coastal Zone [LOICZ] foundations of network analysis. Ecosystems 2: 167–179. Implementation Plan. IGBP Report No. 33, 215 pp. Finn, J. T. 1976. Measures of ecosystem structure and Polovina, J. J. 1984. Model of a coral reef ecosystem. Part I. function derived from analysis of flows. Journal of The ECOPATH model and its applications to French Theoretical Biology 56: 363–380. Frigate Shoals. Coral Reefs 3: 1–11. REFERENCES 217

Richardson, D. C., and J. S. Richardson. 1992. The Wulff F, J. G. Field, and K. H. Mann. (eds). 1989. Network kinemage: a tool for scientific communication. Protein Analysis in Marine Ecology: Methods and Applications. Science 1, 3–9. Available at: http://kinemage.biochem. Coastal and Estuarine Studies 32, Springer-Verlag, duke.edu/ Heidelberg. Scharler, U. M., and D. Baird. 2003. The influence of Wulff, F., and R. E. Ulanowicz. 1989. A comparative catchment management on salinity, nutrient stochio- of the Baltic Sea and Chesapeake Bay metry, and phytoplankton biomass of Eastern Cape ecosystems. In: F. Wulff, J. G. Field, and K. H. Mann estuaries, South Africa. Estuarine, Coastal and Shelf (eds), Network Analysis in Marine Ecology. Coastal Science 56: 735–748. and Estuarine Studies Series. Springer-Verlag, Berlin, Scharler, U. M., and D. Baird. A comparison of selected pp. 232–256. ecosystem attributes of three South African estuaries with different freshwater inflow regimes, using Chapter 4 network analysis. Journal of Marine Systems (in press). Smith, S., D. Swaney, L. Talaue-McManus, S. Bartley, P. Adrian, R., N. Walz, T. Hintze, S. Hoeg, and R. Rusche. Sandhei, C. McLacghlin, V. Dupra, C. Crossland, 1999. Effects of ice duration on the plankton succession R. Buddemeier, B. Maxwell, and F. Wulff. 2003. during spring in a shallow . Freshwater Humans, , and the distribution of inorganic Biology 41: 621–623. nutrient loading to the ocean. Bioscience 93: 239–245. Anderson, W. L., D. M. Robertson, and J. J. Magnuson. Szyrmer, J., and R. E. Ulanowicz. 1987. Total flows in 1996. Evidence of recent warming and El-Nin˜o-related ecosystems. Ecological Modelling 35: 123–136. variations in ice break-up of Wisconsin lakes. Limnology Ulanowicz, R. E. 1984. Community measures of marine Oceanography 41: 815–821. food networks and their possible applications. Anneville, O., S. Soussi, S. Gammeter, and D. Straile. 2004. In: M. J. R. Fasham (ed.), Flows of Energy and Materials Seasonal and inter-annual scales of variability in phyto- in Marine Ecosystems. Plenum, London, pp. 23–47. plankton assemblages: comparison of phytoplankton Ulanowicz, R. E. 1986. Growth and Development: Ecosystem dynamics in three peri-alpine lakes over a period of Phenomenology. Springer-Verlag, New York. 28 years. Freshwater Biology 49: 98–115. Ulanowicz, R. E. 1995a. Network growth and develop- Anneville, O., S. Soussi, F. Iban˜ez, V. Ginot, J.-C. Druart, ment: ascendency. In: B. C. Patten and S. E. Jørgensen and N. Angeli. 2002. Temporal mapping of phyto- (ed), Complex Ecology: The Part–Whole Relation in plankton assemblages in Lake Geneva: annual and Ecosystems. Prentice Hall, PTR, New Jersey, pp. interannual changes in their patterns of succession. 643–655. Limnology Oceanography 47: 1355–1366. Ulanowicz, R. E. 1995b. Ecosystem trophic founda- Ba¨uerle, E., and U. Gaedke. (eds). 1998. Lake constance— tions: Lindeman Exonerata. In: B. C. Patten and characterization of an ecosystem in transition. Archives of S. E. Jorgensen (eds), Complex Ecology: The Part–Whole Hydrobiology Special Issues Advances Limnology 53: 1–610. Relation in Ecosystems. Prentice Hall, Englewood Cliffs, Beardall, J., and J. A. Raven. 2004. The potential effects of NJ, pp. 549–560. global climate change on microalgal photosynthesis, Ulanowicz, R. E., and D. Baird. 1999. Nutrient controls on growth and ecology. Phycologia 43: 26–40. ecosystem dynamics: the Chesapeake Bay mesohaline Benndorf, J., H. Kneschke, K. Kossatz, and E. Penz. 1984. community. Journal of Marine Science 19: 159–172. Manipulation of the pelagic food web by stocking with Ulanowicz, R. E., and J. H. Tuttle. 1992. The trophic predacious fishes. Internationale Revue der Gesamten consequences of stock rehabilitation in Chesa- Hydrobiologie 69: 407–428. peake Bay. Estuaries 15: 298–306. Brett, M. T., and C. R. Goldman. 1996. A meta-analysis Ulanowicz, R. E., C. Bondavalli, and M. S. Egnotovich. of the freshwater trophic cascade. Proceedings of the 1999. Network Analysis of Trophic Dynamics in South National Academy of Sciences USA 93: 7723–7726. Florida Ecosystem, FY 97: The Florida Bay Ecosystem. Briand, J. F., C. Leboulanger, J. F. Humbert, C. Bernard, Annual Report to the United States Geological Service and P. Dufour. 2004. Cylindrospermopsis raciborskii Biological Resources Division University of Miami (Cyanobacteria) invasion at mid-latitudes: selection, Coral Gables, FL 33124. wide physiological tolerance, or global warming? Walters, C., V. Christensen, and D. Pauly. 1997. Struc- Journal of 40: 231–238. turing dynamic models of exploited ecosystems from Ca´ceres, C. 1998. Seasonal dynamics and interspecific mass-balance assessments. Reviews in Fish Biology and competition in Oneida Lake Daphnia. Oecologia Fisheries. 7: 139–172. 115: 233–244. 218 REFERENCES

Carpenter, S. R., and J. F. Kitchell. 1993. The Trophic dynamic conditions on the plankton development Cascade in Lakes. Cambridge University Press. in Lake Constance in spring and summer. Archives Carpenter, S. R., S. G. Fisher, N. B. Grimm, and J. F. Kitchell. in Hydrobiology Special Issues in Advances Limnology 1992. Global change and freshwater ecosystems. 53: 565–585. Annual Review of Ecological Systems 23: 119–139. Gaedke, U., D. Straile, and C. Pahl-Wostl. 1996. Trophic Carpenter, S. R., J. F. Kitchell, and J. R. Hodgson. 1985. structure and carbon flow dynamics in the pelagic Cascading trophic interactions and lake productivity. community of a large lake. In: G. A. Polis and Bioscience 35: 634–639. K. O. Winemiller (eds), Food Webs: Integration of Pattern Chen, C. Y., and C. L. Folt. 1996. Consequences of fall and Dynamics. Chapmann and Hall, pp. 60–71. warming for zooplankton overwintering success. George, D. G. and G. P. Harris. 1985. The effect of climate Limnology and Oceanography 41: 1077–1086. on long-term changes in the crustacean zooplankton Cohen, J. E., T. Jonsson, and S. R. Carpenter. 2003. biomass of Lake Windermere, UK. Nature 316: 536–539. Ecological community description using the food web, George, D. G. and D. P. Hewitt. 1999. The influence of species abundance, and body size. Proceedings of the year-to-year variations in winter weather on the National Academy of Sciences USA 100: 1781–1786. dynamics of Daphnia and Eudiaptomus in Estwaite Conover, D. O. 1992. Seasonality and the scheduling of Water, Cumbria. Functional Ecology 13(Suppl.1): 45–54. life-history at different latitudes. Journal of Fish Biology George, D. G., and A. H. Taylor. 1995. UK lake plankton 41B: 161–178. and the Gulf stream. Nature 378: 139. Cushing, D. H. 1990. Plankton production and year- Gerten, D., and R. Adrian. 2000. Climate-driven changes class strength in fish populations: an update of the in spring plankton dynamics and the sensitivity match/mismatch hypotheses. Advances of shallow polymictic lakes to the North Atlantic 26: 249–293. Oscillation. Limnology and Oceanography 45: 1058–1066. Eckmann, R., U. Gaedke, and H. J. Wetzlar. 1988. Effects of Gerten, D. and R. Adrian. 2002a. Effects of climate climatic and density-dependent factors on year-class warming, North Atlantic Oscillation, and El Nin˜o- strength of Coregonus lavaretus in Lake Constance. Cana- Southern Oscillation on thermal conditions and dian Journal of Fisherics and Aquatic Sciences 45: 1088–1093. plankton dynamics in northern hemispheric lakes. Gaedke, U. 1992. The size distribution of plankton The Scientific World Journal 2: 586–606. biomass in a large lake and its seasonal variability. Gerten, D., and R. Adrian. 2002b. Species-specific changes Limnology and Oceanography 37: 1202–1220. in the phenology and peak abundance of freshwater Gaedke, U. 1993. Ecosystem analysis based on biomass copepods in response to warm summers. Freshwater size distributions: a case study of a plankton com- Biology 47: 2163–2173. munity in a large lake. Limnology and Oceanography Gillooly, J. F. 2000. Effect of body size and temperature 38: 112–127. on generation time in zooplankton. J Plankt Research Gaedke, U. 1995. A comparison of whole-community and 22: 241–251. ecosystem approaches (biomass size distributions, food Goldman, C. R., A. Jassby, and T. Powell. 1989. Inter- web analysis, network analysis, simulation models) to annual fluctuations in primary production: meteoro- study the structure, function and regulation of pelagic logical forcing at two subalpine lakes. Limnology and food webs. Journal of Plankton Research 17: 1273–1305. Oceanography 34: 310–323. Gaedke, U. and D. Straile. 1994a. Seasonal changes of the Gu¨ ss, S., D. Albrecht, H.-J. Krambeck, D. C. Mu¨ller- quantitative importance of protozoans in a large lake. Navarra, and H. Mumm. 2000. Impact of weather on An ecosystem approach using mass-balanced carbon a , assessed by cyclo-stationary MCCA flow diagrams. Marine Microbial Food Webs 8: 163–188. of long-term observations. Ecology 81: 1720–1735. Gaedke, U. and D. Straile. 1994b. Seasonal changes of Hairston, N. G. Jr. 1998. Time travelers: what’s timely trophic transfer efficiencies in a plankton food web in diapause reserach? Archives in Hydrobiology Special derived from biomass size distributions and network Issues in Advances in Limnology 52: 1–15. analysis. Ecological Modelling 75/76: 435–445. Hairston, N. G. Jr., and C. Ca´ceres. 1996. Distribution of Gaedke, U., S. Hochsta¨dter, and D. Straile. 2002. Interplay crustacean diapause: micro- and macroevolutionary between energy limitation and nutritional deficiency: pattern and process. Hydrobiology 320: 27–44. empirical data and food web models. Ecological Hairston, N. G. Jr., and N. G. Sr. Hairston. 1993. Monographs 72: 251–270. Cause–effect relationships in energy flow, trophic Gaedke, U., D. Ollinger, E. Ba¨uerle, and D. Straile. 1998. strcture, and interspecific interactions. American The impact of the interannual variability in hydro- Naturalist 142: 379–411. REFERENCES 219

Hairston, N. G. Jr., A. M. Hansen, and W. R. Schaffner. Mittelbach, G. G., A. M. Turner, D. J. Hall, J. E. Rettig, 2000. The effect of diapause on the seasonal and C. W. Osenberg. 1995. Perturbation and resilience: dynamics of a zooplankton assemblage. Freshwater a long-term, whole-lake study of predator extinction Biology 45: 133–146. and reintroduction. Ecology 71: 2241–2254. Hairston, N. G., F. E. Smith, and L. B. Slobodkin. 1960. Mu¨ ller, H., and C. Wu¨ nsch. 1999. Seasonal dynamics of Community structure, population control, and com- cyst formation of pelagic strombidiid ciliates in a deep petition. American Naturalist 94: 421–425. prealpine lake. Aquatic Microbial Ecology 17: 37–47. Hansson, L. 1996. Algal recruitment from lake sediments Murdoch, W. M. 1966. ‘‘Community structure, popu- in relation to grazing, sinking, and dominance patterns lation control, and competition’’—a critique. American in the phytoplankton community. Limnology and Naturalist 100: 219–226. Oceanography 41: 1312–1323. Nicholls, K. H. 1998. El Nino, ice cover, and Great Houghton, J. T., Y. Ding, D. J. Griggs, M. Noguer, Lakes phosphorus: implications for climate warming. P. J. van der Linden, X. Da, K. Maskell, and Limnology and Oceanography 43: 715–719. C. A. Johnson. 2001. Climate Change 2001—The Oksanen, L., S. D. Fretwell, J. Arruda, and Niemel. 1981. Scientific Basis. Contribution of Working Group I to Exploitation ecosystems in gradients of primary pro- the Third Assessment Report of the Intergovernmental ductivity. American Naturalist 118: 240–261. Panel on Climate Change. Cambridge University Press. Persson, L., A. M. De Roos, D. Claessen, P. Bystro¨m, Hrba´cˇek, J. 1958. Typologie und Produktivita¨t der J. Lo¨vgren, S. Sjo¨gren, R. Svanba¨ck, E. Wahlstro¨m, teichartigen Gewa¨sser. Verhein Internationale Verein and E. Westman. 2003. Gigantic cannibals driving a Limnologie 13: 394–399. whole-lake trophic cascade. Proceedings of the National Hurrell, J. W. 1995. Decadal trends in the North Atlantic Academy of Sciences, USA 100: 4035–4039. Oscillation: regional temperatures and precipitation. Persson, L., S. Diehl, L. Johansson, G. Andersson, and Science 269: 676–679. S. F. Hamrin. 1992. Trophic interactions in temperate Jankowski, T. and D. Straile. 2004. Allochronic differen- lake ecosystems: a test of food chain theory. American tiation among Daphnia species, hybrids and back- Naturalist 140: 59–84. crosses: the importance of sexual reproduction for Platt, T., C. Fuentes-Yaco, and K. T. Frank. 2003. population dynamics and genetic architecture. Journal Spring and larval fish survival. Nature of Evolutionary Biology 17: 312–321. 423: 398–399. Lampert, W. 1978. Climatic conditions and planktonic Post, J. R., and D. O. Evans. 1989. Size-dependent interactions as factors controlling the regular succes- overwinter mortality of young-of-the-year yellow sion of spring algal bloom and extremely clear water in perch (Perca flavescens)—Laboratory, in Situ enclosure, Lake Constance. Verhein Internationale Verein Limnologie and field experiments. Canadian Journal of Fisheries and 20: 969–974. Aquatic Sciences 46: 1958–1968. Levine, S. 1980. Several measures of trophic structure Post, D. M., S. R. Carpenter, D. L. Christensen, applicable to complex food webs. Journal of theoretical K. L. Cottingham, J. F. Kitchell, D. E. Schindler, and Biology 83: 195–207. J. R. Hodgson. 1997. Seasonal effects of variable Livingstone, D. M. 1997. An example of the simultaneous recruitment of a dominant on pelagic food- occurrence of climate-driven ‘‘sawtooth’’ deep-water web structure. Limnology and Oceanography 42: 722–729. warming/cooling episodes in several Swiss lakes. Reynolds, C. S. 1997. Vegetation Processes in the Pelagic: A Verhein Internationale Verein Limnologie 26: 822–826. Model for Ecosystem Theory. Excellence in Ecology Livingstone, D. M. 2000. Large-scale climatic forcing Series. Vol 9, Ecology Institute, Oldendorf-Luhe, 371 pp. detected in historical observations of lake ice break-up. Reynolds, C. S., and E. G. Bellinger. 1992. Patterns of Verhein Internationale Verein Limnologie 27: 2775–2783. abundance and dominance of the phytoplankton of Livingstone, D. M. 2003. Impact of secular climate change Rostherne Mere, England: evidence from an 18-year on the thermal structure of a large temperate central data set. 54: 10–36. European lake. Climatic Change 57: 205–225. Rodionov, S., and R. A. Assel. 2003. Winter severity in Magnuson, J. J., D. M. Robertson, B. J. Benson, the region: a tale of two oscillations. R. H. Wynne, D. M. Livingstone, T. Arai, R. A. Assel, Climatic Research 24: 19–31. R. G. Barry, V. Card, E. Kuusisto, N. G. Granin, Sanderson, B. L., T. R. Hrabik, J. J. Magnuson, and T. D. Prowse, K. M. Stewart, and V. S. Vuglinski. 2000. D. M. Post. 1999. Cyclic dynamics of a yellow perch Historical trends in lake and river ice cover in the (Perca flavescens) population in an oligotrophic lake: Northern Hemisphere. Science 289: 1743–1746. evidence for the role of intraspecific interactions. 220 REFERENCES

Canadian Journal of Fisheries and Aquatic Sciences Straile, D. 2002. North Atlantic Oscillation synchronizes 56: 1534–1542. food-web interactions in central European lakes. Sanford, E. 1999. Regulation of keystone predation by small Proceedings of the Royal Society of London B 269: 391–395. changes in ocean temperature. Science 283: 2095–2097. Straile, D. 2004. A fresh (water) perspective on Scheffer, M., S. Carpenter, C. Foley, and B. Walker. 2001a. NAO ecological impacts in the North Atlantic. Catastrophic shifts in ecosystems. Nature 413: 591–596. In: N. C. Stenseth, G. Ottersen, J. W. Hurrell, and Scheffer, M., S. H. Hosper, M. L. Meijer, B. Moss, and A. Belgrano (eds), Ecological Effects of Climate Variations E. Jeppesen. 1993. Alternative equilibria in shallow in the North Atlantic. lakes. TREE 8: 275–279. Straile, D., and R. Adrian. 2000. The North Atlantic Scheffer, M., D. Straile, E. H. van Nes, and H. Hosper. Oscillation and plankton dynamics in two European 2001b. Climatic warming causes regime shifts in lake lakes—two variations on a general theme. Global food webs. Limnology and Oceanography 46: 1780–1783. Change Biology 6: 663–670. Schindler, D. W. 1997. Widespread effects of climatic Straile, D. and W. Geller. 1998. Crustacean zooplankton warming on freshwater ecosystems in North America. in Lake Constance from 1920. to 1995: response to Hydrological Processes 11: 1043–1067. eutrophication and reoligotrophication. Archives in Schindler, D. W., K. G. Beaty, E. J. Fee, D. R. Cruikshank, Hydrobiology. Special Issues in Advances Limnology E. R. DeBruyn, D. L. Findlay, G. A. Linsey, J. A. Shearer, 53: 255–274. M. P. Stainton, and M. A. Turner. 1990. Effects of Straile, D. and A. Ha¨lbich. 2000. Life history and multiple climatic warming on lakes of the central boreal forest. anti-predator defenses of an invertebrate predator: Science 250: 967–970. Bythotrephes longimanus, in a large and deep lake. Schindler, D. W., P. J. Curtis, B. R. Parker, and Ecology 81: 150–163. M. P. Stainton. 1996. Consequences of climate warming Straile, D., K. Joehnk, and H. Rossknecht. 2003. Complex and lake acidification for UV-B penetration in North effects of winter warming on the physico-chemical American boreal lakes. Nature 379: 705–708. characteristics of a deep lake. Limnology and Oceano- Schmitz, O. J., E. Post, C. E. Burns, and K. M. Johnston. graphy 48: 1432–1438. 2003. Ecosystem responses to global climate Straile, D., D. M. Livingstone, G. A. Weyhenmeyer, change: moving beyond color mapping. Bioscience and D. G. George. 2003. The response of freshwater 53: 1199–1205. ecosystems to climate variability associated with Shapiro, J., and D. I. Wright. 1984. Lake restoration the North Atlantic Oscillation. In: J. W. Hurrell, by biomanipulation—Round Lake, Minnesota, the 1st Y. Kushnir, G. Ottersen, and M. Visbeck (eds), 2 Years. Freshwater Biology 14: 371–383. The North Atlantic Oscillation: Climatic significance and Sommer, U. 1989. Plankton Ecology. Springer. Environmental Impact, pp. 263–279. Sommer, U., Z. M. Gliwicz, W. Lampert, and A. Duncan. Tonn, W. M. and C. A. Paszkowski. 1986. Size-limited 1986. The PEG-model of seasonal succession of plank- predation, winterkill, and the organization of Umbra— tonic events in fresh waters. Archives in Hydrobiology Perca fish assemblages. Canadian Journal of Fisheries and 106: 433–471. Aquatic Sciences 43: 194–202. Stich, H. B., and W. Lampert. 1981. Predator evasion Ulanowicz, R. E. 1986. Growth and Development: Ecosystems as an explanation of diurnal vertical migration by Phenomenology. Springer. zooplankton. Nature 293: 396–398. Urabe, J., J. Togari, and J. J. Elser. 2003. Stoichiometric Straile, D. 1995. Die saisonale Entwicklung des Kohlen- impacts of increased on a planktonic stoffkreislaufes im pelagischen Nahrungsnetz des herbivore. 9: 818–825. Bodensees—Eine Analyse von massenbilanzierten van Donk, E., L. Santamaria, and W. M. Mooij. 2003. Flußdiagrammen mit Hilfe der Netzwerktheorie. PhD Climate change causes regime shifts in lake food Thesis, University of Konstanz. Hartung-Gorre Verlag, webs: a reassessment. Limnology and Oceanography Konstanz. 48: 1350–1353. Straile, D. 1998. Biomass allocation and carbon flow in the Visser, P. M., B. W. Ibelings, B. van der Veer, J. Koedood, and pelagic food web of Lake Constance. Archives in L. R. Mur. 1996. Artificial mixing prevents nuisance Hydrobiology Special Issues in Advances in Limnology blooms of the cyanobacterium Microcystis in Lake Nieuwe 53: 545–563. Meer, the Netherlands. Freshwater Biology 36: 435–450. Straile, D. 2000. Meteorological forcing of plankton Walther, G.-R., E. Post, P. Convey, A. Menzel, dynamics in a large and deep continental European C. Parmesan, T. J. C. Beebee, J. M. Fromentin, lake. Oecologia 122: 44–50. O. Hoegh-Guldberg, and F. Bairlein. 2002. REFERENCES 221

Ecological responses to recent climate change. Nature effects on periphyton and macroinvertebrates. Journal 416: 389–395. of The North American Benthological Society 15: 551–563. Weyhenmeyer, G. A., T. Blenckner, and K. Pettersson. Clements, W. H., and M. C. Newman. 2002. Community 1999. Changes of the plankton spring outburst related Ecotoxicology. Wiley, Chichester, 336 pp. to the North Atlantic Oscillation. Limnology Oceano- Closs, G. P. 1996. Effects of a predatory fish (Galaxias graphy 44: 1788–1792. olidus) on the structure of intermittent Williams, R. J., and N. D. Martinez. 2004. Limits to communities in southeast Australia. Australian Journal trophic levels and omnivory in complex food webs: of Ecology 21: 217–223. theory and data. American Naturalist 163: E458–E468. Closs, G. P., and P. S. Lake. 1994. Spatial and temporal Williamson, C. E., G. Grad, H. J. De Lange, S. Gilroy, and variation in the structure of an intermittent-stream D. M. Karapelou. 2002. Temperature-dependent ultra- food-web. Ecological Monographs 64: 1–21. violet responses in zooplankton: implications of climate Cohen, J. E. 1978. Food Webs and Niche Space. Monographs change. Limnology and Oceanography 47: 1844–1848. in Population Biology, No. 11. Princeton University Press, Princeton, NJ. Chapter 5 Cohen, J. E. and C. M. Newman. 1988. Dynamic basis of food web organization. Ecology 69: 1655–1664. Allan, J. D. 1982. Feeding habits and prey consumption of Cohen, J. E., R. A. , S. H. Cousins, D. A. DeAngelis, three setipalpian stoneflies (Plecoptera) in a mountain L. Goldwasser, K. L. Heong, R. D. Holt, A. J. Kohn, stream. Ecology 63: 26–34. J. H. Lawton, N. Martinez, R. O’Malley, L. M. Page, Allan, J. D. 1995. Stream Ecology: Structure and Function of B. C. Patten, S. L. Pimm, G. A. Polis, M. Rejma´nek, Running Waters. Chapman and Hall, London. T. W. Schoener, K. Schoenly, W. G. Sprules, J. M. Teal, Allen, K. R. 1951. The Horokiwi Stream: A Study of a R. E. Ulanowicz, P. H. Warren, H. M., Wilbur, and Trout Population. Fisheries Bulletin 10. New Zealand P. Yodzis. 1993b. Improving food webs. Ecology Marine Department, Wellington, New Zealand. 74: 252–258. Benke, A. C., and J. B. Wallace. 1997. Trophic basis of Cohen, J. E., T. Jonsson, and S. R. Carpenter. 2003. production among riverine caddisflies: implications for Ecological community description using food web, food web analysis. Ecology 78: 1132–1145. species abundance, and body-size. Proceedings of the Benke, A. C., J. B. Wallace, J. W. Harrison, and National Academy of Sciences USA 100: 1781–1786. J. W. Koebel. 2001. Food web quantification using Cohen, J. E., S. L. Pimm, P. Yodzis, and J. Saldan˜a. 1993a. secondary production analysis: predaceous inverte- Body sizes of animal predators and animal prey in brates of the habitat in a subtropical river. food webs. Journal of Animal Ecology 62: 67–78. Freshwater Biology 46: 329–346. Collier, K. J., S. Bury, and Max Gibbs. 2002. A stable Berlow, E. L., A. M. Neutel, J. E. Cohen, P. de Ruiter, isotope study of linkages between stream and terres- B. Ebenman, M. Emmerson, J. W. Fox, V. A. A. Jansen, trial food webs through spider predation. Freshwater J. I. Jones, G. D. Kokkoris, D. O. Logofet, A. J. McKane, Biology 47: 1651–1659. J. M. Montoya, and O. Petchey. 2004. Interaction Cousins, S. H. 1996. Food webs: from the Lindeman strengths in food webs: issues and opportunities. paradigm to a taxonomic general theory of ecology. Journal of Animal Ecology 73: 585–598. In: G. A. Polis, and K. O. Winemiller (eds), Food Webs: Bird, G. A., and W. Schwartz. 1996. Effect of microbes on Integration of Patterns and Dynamics. Chapman and the uptake of Co-60, Sr-85, Tc-95M, I-131 and Cs-134 by Hall, chapter 23, pp. 243–254. decomposing elm leaves in aquatic microcosms. Cross, W. F., J. P. Benstead, A. D. Rosemond, and Hydrobiologia 333: 57–62. J. B. Wallace. 2003. Consumer-resource stoichiometry Bradley, D. C., and S. J. Ormerod. 2001. Community in detritus-based streams. Ecology Letters 6: 721–732. persistence among stream invertebrates tracks the Dangles, O., and B. Malmqvist. 2004. Species richness- Northern Atlantic Oscillation. Journal of Animal Ecology decomposition relationships depend on species 70: 987–996. dominance. Ecology Letters 7: 395–402. Bunn, E. B., Davies, P. M., and Winning, M. 2003. Dangles, O., M. O. Gessner, F. Guerold, and E. Chauvet. Sources of organic carbon supporting the food web of 2004. Impacts of stream acidification on litter an arid zone floodplain river. Freshwater Biology breakdown: implications for assessing ecosystem 48: 619–635. functioning. Journal of Applied Ecology 41: 365–378. Charlebois, P. M., and Lamberti, G. A. 1996. Invading Death, R. G., and M. R. Winterbourn. 1995. crayfish in a Michigan stream: direct and indirect Diversity patterns in stream benthic invertebrate 222 REFERENCES

communities: the influence of habitat stability. Ecology Gessner, M. O., and E. Chauvet. 1994. Importance of 76: 1446–1460. stream microfungi in controlling breakdown rates of De Ruiter, P. C., A.-M. Neutel, and J. C. Moore. 1995. leaf-litter. Ecology 75: 1807–1817. Energetics, patterns of interaction strengths, and Hall, R. O., J. B. Wallace, and S. L. Eggert. 2000. Organic stability in real ecosystems. Science 269: 1257–1260. matter flow in stream food webs with reduced detrital Dick, J. T. A., W. I. Montgomery, and R. W. Elwood. 1993. resource base. Ecology 81: 3445–3463. Replacement of the indigenous amphipod Gammarus Hall, R. O., G. E. Likens, and H. M. Malcolm. 2001. duebeni celticus by the introduced G. pulex: differential Trophic basis of invertebrate production in two streams cannibalism and mutual predation. Journal of Animal at the Hubbard Brook Experimental Forest. 62: 79–88. the North American Benthological Society 20: 432–447. Dodds, W. K., M. A. Evans-White, N. M. Gerlanc, L. Gray, Hall, S. J., and D. Raffaelli. 1993. Food webs: theory and D. A. Gudder, M. J. Kemp, A. L. Lopez, D. Stagliano, reality. Advances in Ecological Research 24: 187–239. E. A. Strauss, J. L. Tank, M. R. Whiles, and Ha¨ma¨la¨inen, H., and P. Huttunen. 1996. Inferring the W. M. Wollheim. 2000. Quantification of the nitrogen minimum pH of streams from macroinvertebrates cycle in a stream. Ecosystems 3: 574–589. using weighted averaging regression and calibration. Downing, A. L., and M. Leibold. 2002. Ecosystem Freshwater Biology 36: 697–709. consequences of species richness and composition in Hieber, M., and M. O. Gessner. 2002. Contribution of pond food webs. Nature 416: 837–841. stream detrivores, fungi, and bacteria to leaf breakdown Edling, H., and L. R. Tranvik. 1996. Effects of pH on based on biomass estimates. Ecology 83: 1026–1038 -glucosidase activity and availability of DOC to bac- Hildrew, A. G. 1992. Food webs and species interactions. teria in lakes. Archiv fu¨r Hydrobiologie/Advances in In: P. Calow, and G. E. Petts (eds), The Rivers Handbook Limnology 48: 123–132. Blackwell Sciences, Oxford, pp. 309–330. Eggert, S., and J. B. Wallace. 2003. Litter breakdown Hildrew, A. G., and S. J. Ormerod. 1995. Acidification: causes, and invertebrate detritivores in a resource-depleted consequences and solutions. In D. M. Harper, and Appalachian stream. Archiv fu¨r Hydrobiologie 156: A. J. D. Ferguson (eds), The Ecological Basis for River Man- 315–338. agement. John Wiley and Sons Ltd., London, pp. 147–160. Elser, J. J., and J. Urabe. 1999. The stoichiometry of Hildrew, A. G., and C. R. Townsend. 1977. The consumer-driven nutrient recycling: theory, observa- influence of substrate on the functional response of tions and consequences. Ecology 80: 735–751. Plectrocnemia conspersa (Curtis) larvae (Trichoptera: Elton, C. S. 1927. Animal Ecology. Sedgwick and Jackson, Polycentropodidae). Oecologia 31: 21–26. London. Hildrew, A. G., and C. R. Townsend. 1982. Predators and Emmerson, M. C., and D. G. Raffaelli. 2004. Predator–prey prey in a patchy environment: a freshwater study. body size, interaction strength and the stability of a Journal of Animal Ecology 51: 797–815. real food web. Journal of Animal Ecology 73: 399–409. Hildrew, A. G., C. R. Townsed, and J. E. Francis. 1984. Emmerson, M. C., M. Solan, C. Emes, D. M. Paterson, Community structure in some southern English and D. Raffaelli. 2001. Consistent patterns and the streams: the influence of species interactions. idiosyncratic effects of biodiversity in marine eco- Freshwater Biology 14: 297–310. systems. Nature 411: 73–77. Hildrew, A. G., C. R. Townsend, and A. Hasham. 1985. Fahey, B., and R. Jackson. 1997. Hydrological impacts The predatory Chironomidae of an iron-rich stream: of converting native forests and grasslands to pine feeding ecology and food web structure. Ecological plantations, South Island, New Zealand. Agricultural 10: 403–413. and Forest Meteorology 84: 69–82. Hildrew, A. G., G. Woodward, J. H. Winterbottom, and Flecker, A. S. 1996. Ecosystem engineering by a dominant S. Orton. 2004. Strong density-dependence in a pre- detritivore in a diverse tropical stream. Ecology datory insect: larger scale experiments in a stream. 77: 1845–1854. Journal of Animal Ecology 73: 448–458. Flecker, A. S., and C. R. Townsend. 1994. Community- Hillebrand, H. 2002. Top-down versus bottom-up control wide consequences of trout introduction in of autotrophic biomass—a meta-analysis on experi- New Zealand streams. Ecological Applications 4: 798–807. ments with periphyton. Journal of the North American Friberg, N., A. Rebsdorf, and S. E. Larsen. 1998. Effects of Benthological Society 21: 349–369. afforestation on acidity and invertebrates in Danish Hogg, I. D., and D. D. Williams. 1996. Response of stream streams and implications for freshwater communities invertebrates to a global-warming thermal regime: in Denmark. Water, Air, and Soil Pollution 101: 235–256. an ecosystem-level manipulation. Ecology 77: 395–407. REFERENCES 223

Holt, R. D. 1977. Predation, apparent competition, and the Ledger, M. E. and A. G. Hildrew. 2000b. Resource structure of prey communities. Theoretical Population depression by a trophic generalist in an acid stream. Biology 12: 197–229. Oikos 90: 271–278. Huryn, A. D. 1996. An appraisal of the Allen Paradox in a Ledger, M., and A. G. Hildrew. 2001. Growth of an acid- New Zealand trout stream. Limnology and Oceanography tolerant stonefly on biofilm from streams of contrasting 41: 243–252. pH. Freshwater Biology 46: 1457–1470. Huryn, A. D. 1998. Ecosystem-level evidence for top- Leeks, G. J. L. 1992. Impact of plantation on sedi- down and bottom-up control of production in a ment transport processes. In: P. Billi, R. D. Hey, grassland stream system. Oecologia 115: 173–183. C. R. Thorne, and P. Tacconi (eds), Dynamics of Gravel- Hynes, H. B. N. 1975. The stream and its valley. Bed Rivers. John Wiley and Sons, Chichester, pp. 651–670. Verhandlungen der Internationalen Vereinigung fu¨r Lehman, C. L., and D. Tilman. 2000. Biodiversity, stability, Theoretische und Angewandte Limnologie 19: 1–15. and productivity in competitive communities. American Irons, J. G., M. W. Oswood, R. J. Stout, and C. M. Pringle. Naturalist 156: 534–552. 1994. Latitudinal patterns in leaf-litter breakdown—is Lindeman, R. L. 1942. The trophic-dynamic aspect of temperature really important? Freshwater Biology ecology. Ecology 23: 399–417 32: 401–411. MacArthur, R. H. 1955. Fluctuations of animal popu- Jenkins, B., R. L. Kitching, and S. L. Pimm. 1992. lations and a measure of community stability. Ecology Productivity, disturbance and food web structure at a 36: 533–536. local spatial scale in experimental container habitats. Malmqvist, B., S. D. Rundle, A. P. Covich, A. G. Hildrew, Oikos 65: 249–255. C. T. Robinson, and C. R. Townsend. Prospects for Jonsson, M., and B. Malmqvist. 2003. Importance of Streams and rivers: an ecological perspective (in press). species identity and number for process rates within McCann, K. S. 2000. The diversity–stability debate. Nature stream functional feeding groups. Journal of Animal 405: 228–233. Ecology 72: 453–459. McCann, K. S., A. Hastings, and G. R. Huxel. 1998. Weak Jonsson, M., O. Dangles, B. Malmqvist, and F. Guerold. trophic interactions and the . Nature 2002. Simulating species loss following perturbation: 395: 794–798. assessing the effects on process rates. Proceedings of the McIntosh, A. R., and C. R. Townsend. 1995. Impacts of Royal Society of London B 269: 1047–1052. an introduced predatory fish on mayfly grazing in Jonsson, M., B. Malmqvist, and P. O. Hoffsten. 2001. Leaf New Zealand streams. Limnology and Oceanography 40: litter breakdown rates in boreal streams: does shredder 1508–1512. species richness matter? Freshwater Biology 46: 161–171. Matthews, W. J., and E. G. Zimmerman. 1990. Potential Junk, W. J., P. B. Bayley, and R. E. Sparks. 1989. The effects of global warming on native fishes of the sou- Pulse Concept in river–floodplain systems. thern Great Plains and the Southwest. Fisheries 15: 26–32. In: D. P. Dodge (ed.) Canadian Fisheries and Aquatic May, R. M. 1972. Will a large complex system be stable? Sciences 106: (Spec. Publ.) 110–127. Nature 238: 413–414. Keleher, C. J., and F. J. Rahel. 1996. Thermal; limits to May, R. M. 1973. Stability and Complexity in Model salmonid distributions in the rocky mountain region Ecosystems. Princeton University Press, Princeton, NJ. and potential habitat loss due to global warming: Mihuc, T. B. 1997. The functional trophic role of lotic a geographic (GIS) approach. primary consumers: generalist versus specialist Transactions of the American Fisheries Society 125: 1–13. strategies. Freshwater Biology 37: 455–462. Lancaster, J. 1996. Scaling the effects of predation and Milner, A. M., E. E. Knudssen, C. Soiseth, A. L. Robertson, disturbance in a patchy environment. Oecologia D. Schell, L. T. Phillips, and K. Magnusson. 2000. 107: 321–331. Colonization and development of stream communities Lancaster, J., and A. L. Robertson. 1995. Microcrustacean across a 200-year gradient in Bay National prey and macroinvertebrate predators in a stream food Park, Alaska, USA. Canadian Journal of Fisheries and web. Freshwater Biology 34: 123–134. Aquatic Sciences 57: 2319–2355. Lancaster, J., M. Real, S. Juggins, D. T. Monteith, Mithen, S. J., and J. H. Lawton. 1986. Food-web models R. J. , and W. R. C. Beaumont. 1996. Monitoring that generate constant predator–prey ratios. Oecologia temporal changes in the biology of acid waters. Fresh- 69: 542–550. water Biology 36: 179–201. Mohseni, O., H. G. Stefan, and J. G. Eaton. 2003. Global Ledger, M. E., and Hildrew, A. G. 2000a. Herbivory in an warming and potential changes in fish habitat in acid stream. Freshwater Biology 43: 545–556. U.S. streams. Climatic Change 15: 389–409. 224 REFERENCES

Molander, S. H., H. Blanck, and M. So¨derstro¨m. 1990. Schmid-Araya, J. M., A. G. Hildrew, A. Robertson, Toxicity assessment by pollution-induced community P. E. Schmid, and J. Winterbottom. 2002a. The tolerance (PICT) and identification of metabolites importance of meiofauna in food webs: evidence from in periphyton communities after exposure to an acid stream. Ecology 83: 1271–1285. 4,5,6-trichloroguaiacol. Aquatic 18: 115–136. Schmid-Araya, J. M., P. E. Schmid, A. Robertson, Mulholland, P. J., A. V. Palumbo, J. W. Elwood, and J. Winterbottom, C. Gjerlov, and A. G. Hildrew. 2002b. A. D. Rosemond. 1987. Effect of acidification on leaf Connectance in stream food webs. Journal of Animal decomposition in streams. Journal of the North American Ecology 71: 1056–1062. Benthological Society 6: 147–158. Schmid, P. E., M. Tokeshi, and J. M. Schmid-Araya. 2000. Nakano, S., H. Miyasaka, and N. Kuhara. 1999. Terrestrial– Relation between population density and body size in aquatic linkages: Riparian inputs alter trophic stream communities. Science 289: 557–1560. cascades in a stream food web. Ecology 80: 2435–2441. Schreiber, E. S. G., G. P. Quinn, and P. S. Lake. 2003. Naeem, S. 1998. Species redundancy and ecosystem Distribution of an alien aquatic snail in relation to reliability. 12: 39–45. flow variability, human activities and . Neutel, A. -M., J. A. P. Heesterbeek, and P. C. de Ruiter. Freshwater Biology 48: 951–961. 2002. Stability in real food webs: weak links in long Simon, K. S., E. F. Benfield, and S. A. Macko. 2003. loops. Science 296: 1120–1123. Food web structure and the role of epilithic biofilms in Ormerod, S. J., and S. J. Tyler. 1991. Exoloitaion of prey by cave streams. Ecology 84: 2395–2406. a river bird, the dipper Cinclus cinclus, along acidic and Speirs, D. C., W. S. C. Gurney, J. H. Winterbottom, and circumneutral streams in upland Wales. Freshwater A. G. Hildrew. 2000. Long-term demographic balance Biology 25: 105–116. in the Broadstone Stream insect community. Journal of Paine, R. T. 2002. Trophic control of production in a rocky Animal Ecology 69: 45–58. intertidal community. Science 296: 736–739. Stead, T. K., J. M. Schmid-Araya, and A. G. Hildrew. Petchey, O. L., A. L. Downing, G. G. Mittelbach, L. Persson, All creatures great and small: patterns in the stream C. F. Steiner, P. H. Warren, and G. Woodward. 2004. benthos across the whole range of metazoan body size. Species loss and the structure and functioning of Freshwater Biology 48: 532–547. multitrophic aquatic systems.Oikos104: 467–478. Steinmetz, J., S. L. Kohler, and D. A. Soluk. 2003. Birds Poff, N. L., and J. D. Allan. 1995. Functional organiza- are overlooked top predators in aquatic food webs. tion of fish assemblages in relation to hydrological Ecology 84: 1324–1328. variability. Ecology 76: 606–627. Sterner, R. W. 1995. Elemental stoichiometry of species Polis, G. A. 1998. Stability is woven by complex webs. in ecosystems. In: C. Jones and J. Lawton (eds), Linking Nature 395: 744–745. Species and Ecosystems. Chapman and Hall, New York. Post, D. M., M. L. Pace, and N. G. Hairston. 2000. Stone, M. K. and J. B. Wallace. 1998. Long term recovery Ecosystem size determines food-chain length in lakes. of a mountain stream from clear-cut : the Nature 405: 1047–1049. effects of forest succession on benthic invertebrate Power, M. E. 1990. Effects of fish in river food webs. community structure. Hydrobiologia 353: 107–119. Science 250: 811–814. Sutcliffe, D. W., and A. G. Hildrew. 1989. Invertebrate com- Puckridge, J. T., K. F. Walker, and J. F. Costelloe 2000. munities in acid streams. In: R. Morris, E. W. Taylor, D. J. Hydrological persistence and the ecology of dryland rivers. A. Brown, and J. A. Brown (eds), Acid Toxicity and Aquatic Regulated Rivers—Research and Management 16:385–402. Animals. Seminar Series of the Society for Experimental Rosemond, A. D., C. M. Pringle, A. Ramirez, and M. J. Paul. Biology, Cambridge University Press, pp. 13–29. 2001. A test of top-down and bottom-up control in a Tavares-Cromar, A. F., and D. D. Williams. 1996. detritus-based food web. Ecology 82: 2279–2293. The importance of temporal resolution in food web Rosi-Marshall, E. J., and J. B. Wallace. 2002. Invertebrate analysis: Evidence from a detritus-based stream. food webs along a stream resource gradient. Freshwater Ecological Monographs 66: 91–113. Biology 47: 129–141. Thompson, R. M., and C. R. Townsend. 1999. The effect Rounick, J. S., and B. J. Hicks. 1985. The stable carbon of seasonal variation on the community structure and isotope ratios of fish and their invertebrate prey in four food-web attributes of two streams: implications for New Zealand rivers. Freshwater Biology 15: 207–214. food-web science. Oikos 87: 75–88. Schmid, P. E., and J. M. Schmid-Araya. 2000. Trophic Thompson, R. M., and C. R. Townsend. 2003. Impacts relationships: integrating meiofauna into a realistic on stream food webs of native and exotic forest: an benthic food web. Freshwater Biology 44: 149–163. intercontinental comparison. Ecology 84: 145–161. REFERENCES 225

Thompson, R. M., and C. R. Townsend. Energy Warren, P. H. 1996. Structural constraints on food web availability, spatial heterogeneity and ecosystem size assembly. In: M. E. Hochberg, J. Clobert, and R. Barbault predict food-web structure in streams. Oikos (in press). (eds), Aspects of the Genesis and Maintenance of Tokeshi, M. 1999. Species Coexistence: Ecological and Biological Diversity. Oxford University Press, Oxford, Evolutionary Perspectives. Blackwell Science. pp. 142–161. Townsend, C. R. 1989. The concept Webb, B. W. 1996. Trends in stream and river tempera- of stream community ecology. Journal of the North ture. Hydrological Processes 10: 205–226. American Benthological Society 8: 36–50. Wikramanayake, E. D., and P. B. Moyle. 1989. Ecological Townsend, C. R. 1996. Invasion biology and ecological structure of tropical fish assemblages in wet-zone impacts of brown trout (Salmo trutta) in New Zealand. streams of Sri Lanka. Journal of Zoology (London) Biological Conservation 78: 13–22. 218: 503–526. Townsend, C. R. 2003. Individual, population, commun- Williams, R. J., and N. D. Martinez. 2000. Simple rules ity, and ecosystem consequences of a fish invader in yield complex food webs. Nature 404: 180–183. New Zealand streams. Conservation Biology 17: 38–47. Winemiller, K. O. 1990. Spatial and temporal variation in Townsend, C. R., and A. G. Hildrew. 1979. Resource tropical fish trophic networks. Ecological Monographs partitioning by two freshwater invertebrate predators 60: 331–367. with contrasting foraging strategies. Journal of Animal Winemiller, K. O. 1996. Dynamic diversity: fish com- Ecology 48: 909–920. munities of tropical rivers. In: M. L. Cody and Townsend, C. R., S. Dole´dec, and M. R. Scarsbrook. 1997. J. A. Smallwood (eds), Long-term Studies of Vertebrate Species traits in relation to temporal and spatial hetero- Communities. Academic Press, pp. 99–134. geneity in streams: a test of habitat templet theory. Winner, R. W., M. W. Boesel, and M. P. Farrell. 1980. Freshwater Biology 37: 367–387. Insect community structure as an index of heavy- Townsend, C. R., R. M. Thompson, A. R. McIntosh, metal pollution in lotic ecosystems. Canadian Journal of C. Kilroy, E. Edwards, and M. R. Scarsbrook. 1998. Fisheries and Aquatic Sciences 37: 647–655. Disturbance, resource supply and food-web archi- Wipfli, M. S., and J. Caouette. 1998. Influence of salmon tecture in streams. Ecology Letters 1: 200–209. carcasses on stream productivity: response of biofilm Tuchman, N. C., K. A. Wahtera, R. G. Wetzel, and and benthic macroinvertebrates in southeastern Alaska,

J. A. Teeri. 2003. Elevated atmospheric CO2 alters leaf USA. Canadian Journal of Fisheries and Aquatic Sciences litter quality for stream ecosystems: an in situ leaf 55: 1503–1511. decomposition study. Hydrobiologia 495: 203–211. Woodward, G., and A. G. Hildrew. 2001. Invasion of Ulanowicz, R. E., and C. J. Puccia. 1990. Mixed trophic a stream food web by a new top predator. Journal of impacts in ecosystems. Coenoses 5: 7–16. Animal Ecology 70: 273–288. Usio, N. 2000. Effects of crayfish on leaf processing and Woodward, G., and A. G. Hildrew, 2002a. Body- invertebrate of leaves in a headwater stream: size determinants of niche overlap and intraguild decoupling of a trophic cascade. Oecologia 124: 608–614. predation within a complex food web. Journal of Animal Usio, N. and C. R. Townsend. 2001. The significance of Ecology 71: 1063–1074. the crayfish Paranephrops zealandicus as shredders in a Woodward, G., and A. G. Hildrew. 2002b. Differential New Zealand headwater stream. Journal of Crustacean vulnerability of prey to an invading top predator: Biology 21: 354–359. integrating field surveys and laboratory experiments. Vanni, M. J. 2002. Nutrient cycling by animals in Ecological Entomology 27: 732–744. freshwater ecosystems. Annual Review of Ecology and Woodward, G., and A. G. Hildrew. 2002c. Food web Systematics 33: 341–370. structure in riverine . Freshwater Biology Vannote, R. L., G. W. Minshall, K. W. Cummins, 47: 777–798. J. R. Sedell, and C. E. Cushing. 1980. The river Woodward, G., and A. G. Hildrew. 2002d. The impact of continuum concept. Canadian Journal of Fisheries and a sit-and-wait predator: separating consumption and Aquatic Sciences 37: 130–137. prey emigration. Oikos 99: 409–418. Walker, B. H. 1992. Biodiversity and ecological Woodward, G., J. I. Jones, and A. G. Hildrew. 2002. redundancy. Conservation Biology 6: 18–23. Community persistence in Broadstone Stream (U.K.) Ward, J. V., G. Bretschko, M. Brunke, D. Danielopol, over three decades. Freshwater Biology 47: 1419–1435. G. Gibert, T. Gonser, and A. G. Hildrew. 1988. Woodward, G., D. C. Speirs, and A. G. Hildrew. Quanti- A metazoan perspective of riparian/groundwater fication and resolution of a complex, size-structured ecology. Freshwater Biology 40: 531–569. food web. Advances in Ecological Research (in press). 226 REFERENCES

Wootton, J. T. 1997. Estimates and test of per capita interac- web structure: testing the equiprobability assumption tion strength: diet, abundance and impact of intertidally of the cascade model. Oecologia 123: 241–251. foraging birds. Ecological Monographs 67: 45–64. Pimm, S. L. 1982. Food Webs. Chapman and Hall, London. Wootton, J. T., M. S. Parker, and M. E. Power. 1996. The Pruesse, G., and F. Ruskey. 1994. Generating linear effect of disturbance on river food webs. Science 273: extensions fast. SIAM Journal of Computing 23: 373–386. 1558–1561. Schmid-Araya, J. M., P. E. Schmid, A. Robertson, Wright, J. F., D. W. Sutcliffe, and M. T. Furse. eds. 2000. J. Winterbottom, C. Gjerlov, and A. G. Hildrew. 2002. Assessing the Quality of Fresh Waters: Rivpacs and Connectance in stream food webs. Journal of Animal Other Techniques. Freshwater Biological Association, Ecology 71: 1056–1061. Ambleside, 373 pp. Solow, A. R. 1996. On the goodness of fit of the cascade Young, R. G., and A. D. Huryn. 1997. Longitudinal pat- model. Ecology 77: 1294–1297. terns of organic matter transport and turnover along a Solow, A. R., and A. R. Beet. 1998. On lumping species in New Zealand grassland river. Freshwater Biology 38: food webs. Ecology 79: 2013–2018. 93–107. Varol, Y., and D. Rotem. 1981. An algorithm to generate all topological sorting arrangements. Computer Journal 24: 83–84. Chapter 6 Williams, D. A. 1982. Extrabinomial variation in logistic Azzalini, A. 1996. Statistical Inference. Chapman and Hall, linear models. Applied Statistics 31: 144–148. London. Williams, R. J., and N. D. Martinez. 2000. Simple rules Camacho, J., R. Guimera`, and L. A. Nunes Amaral. 2002. yield complex food webs. Nature 404: 180–183. Robust patterns in food web structure. Physical Review Winemiller, K. O., and G. A. Polis. 1996. Food webs: Letters 88:1–4. what do they tell us about the world? In: G. A. Polis and Cohen, J. E. 1990. A theory of community food K.O. Winemiller (eds), Food webs: Integration of Pattern webs. VI. Heterogeneous alternatives to the cascade and Dynamics. Chapman and Hall, London, pp. 1–22. model. Theoretical Population Biology 37: 55–90. Cohen, J. E., and C. Newman. 1985. A stochastic theory of Chapter 7 community food webs. I. Models and aggregated data. Proceedings of the Royal Society B224: 421–448. Asaeda, T., and L. H. Nam. 2002. Effects of rhizome age Cohen, J. E., F. Briand, and C. Newman. 1990. Community on the decomposition rate of Phragmites australis Food Webs: Data and Theory. Springer, Berlin. rhizomes. Hydrobiologia 485: 205–208. Cohen, J. E., S. L. Pimm, P. Yodzis, and J. Saldan˜a. 1993. Chapin, F. S., E. S. Zaveleta, V. T. Eviner, R. L. Naylor, Body size of animal predators and animal prey in food P. M. Vitousek, H. L. Reynolds, D. U. Hoooper, webs. Journal of Animal Ecology 62: 67–78. S. Lavorel, O. E. Sala, S. E. Hobbie, M. C. Mack, and Forbes, S. A. 1887. The lake as a microcosm. Bulletin of the S. Dı´az. 2000. Consequences of changing biodiversity. Peoria Scientific Association 537–550. Nature 405: 234–242. Link, J. 2002. Does food web theory work for marine Christian, R. R., and J. J. Luczkovich. 1999. Organizing ecosystems? Marine Ecology Progress Series 230: 1–9. and understanding a winter’s seagrass foodweb Martinez, N. D. 1992. Constant connectance in commu- network through effective trophic levels. Ecological nity food webs. The American Naturalist 139: 1208–1218. Modelling 117: 99–124. Martinez, N. D. 1993. Effects of resolution on food web Cohen, J. E., and 23 others. 1993. Improving food webs. structure. Oikos 60: 403–412. Ecology 74: 252–258. May, R. M. 1973. Stability and Complexity in Model Gordon, M. S., G. A. Bartholomew, A. D. Grinnell, Ecosystems. Princeton University Press, Princeton, NJ. C. B. Jorgensen, and F. N. White. 1972. Animal Murtaugh, P. A. 1994. Statistical analysis of food webs. Physiology. Macmillan Co., New York, 592 pp. Biometrics 50: 1199–1202. MacArthur, R. H. 1957. On the relative abundance of Murtaugh, P. A., and D. R. Derryberry. 1998. Models of bird species. Proceedings of the National Academy connectance in food webs. Biometrics 54: 754–761. of Sciences, USA 43: 293–295. Murtaugh, P. A., and J. P. Kollath. 1997. Variation in Lambers, H. 1985. Respiration in intact plants and tissues: trophic fractions and connectance in food webs. Ecology its regulation and dependence on environmental fac- 78: 1382–1387. tors, metabolism and invaded organisms. In: R. Douce Neubert, M. G., S. C. Blumenshine, D. E. Duplisea, and D. A. Day (eds), Encyclopedia of , T. Jonsson, and B. Rashleigh. 2000. Body size and food Vol. 18. Springer Verlag. Berlin, pp. 418–473. REFERENCES 227

Landsberg, J. J. 1986. Physiological Ecology of Forest Ulanowicz, R. E., and J. Norden. 1990. Symmetrical Production. Academic Press, London, 198 pp. overhead in flow networks. International Journal of Martinez, N. 1991. Artifacts or attributes? Effects of Systems Sciences 21: 429–437. resolution on the Little Roke Lake food web. Ecological Ulanowicz, R. E., C. Bondavalli, and M. S. Egnotovich. Monographs 61: 367–392. 1998. Network Analysis of throphic dynamics in South May, R. 1973. Stability and Complexity in Model Ecosystems. Florida Ecosystem FY97: The Florida Bay Ecosystem. Princeton University Press, Princeton, NJ. Annual Report to the United States Geological Service McCann, K., A. Hastings, and G. R. Huxel. 1998. Weak Biological Resource Division, University of Miami, trophic interactions and the balance of nature. Nature Coral Gablel, FL. 395: 794–798. Van Santvoort, P. J., G. J. De Lange, J. Thomson, S. Colley, McGrady-Steed, J., P. M. Harris, and P. J. Morin. 2001. F. J. Meysman, and C. P. Slomp. 2002. Oxidation Biodiversity regulates ecosystem predictability. Nature and origin of organic matter in surficial eastern 412: 34–36. mediterranean hemipelagic sediments. Aquatic Geo- Milo, R., S. Itzkovitz, N. Kashtan, R. Levitt, S. Shen-Orr, chemistry 8: 153–175. I. Ayzenshtat, M. Sheffer, and U. Alon. 2004. Super- Zorach, A. C., and R. E. Ulanowicz. 2003. Quantifying the families of evolved and designed networks. Science complexity of flow networks: how many roles are 303: 1538–1542. there? Complexity 8: 68–76. Morris, J. T., and K. Lajtha. 1986. Decomposition and nutrient dynamics of litter from four species of fresh water emergent macrophytes. Hydrobiology Chapter 8 131: 215–223. Beddington, J. R., and Basson, M. 1994. The limits to Odum, E. P. 1969. The strategy of ecosystem develop- exploitation on land and sea. Philosophical Transactions ment. Science 164: 262–270. of the Royal Society B343: 87–92. Qualls, R. G., and C. J. Richardson. 2003. Factors Belgrano, A., A. P. Allen, B. J. Enquist. and J. F. Gillooly. controlling concentration, export, and decomposition 2002. Allometric scaling of maximum population of dissolved organic nutrients in the everglades of density: a common rule for marine phytoplankton and Florida. Biogeochemistry 62(2): 197–229. terrestrial plants. Ecology Letters 5: 611–613. Paine, R. T. 1988. Food webs: road maps of interactions Benoit, E., and M. J. Rochet. 2004. A continuous model of or grist for theoretical development. Ecology 69: biomass size spectra governed by predation and the 1648–1654. effects of fishing on them. Journal of Theoretical Biology Pimm, S. L. 1982. Food Webs. Chapman and Hall, London, 226: 9–21. 219 pp. Beverton, R. J. H., and S. J. Holt. 1959. A review of the Salovius, S., and E. Bonsdorff. 2004. Effects of depth, lifespan and mortality rates of fish in nature and sediment and grazers on the degradation of drifting their relationship to growth and other physiological filamentous algae (Cladophora glomerata and Pilayella characteristics. Ciba Foundation Colloquim on Ageing littoralis). Journal of Experimental Marine Biology and 5: 142–180. Ecology 298: 93–109. Blazka, P., T. Backiel, and F. B. Taub. 1980. Trophic Schomberg, H. H., and J. L. Steiner. 1997. Estimating crop relationships and efficiencies. In: E. D. Le Cren and residue decomposition coefficients using substrate- R. H. Lowe-McConnell (eds), The Functioning of Fresh- induced respiration. and Biochemistry water Ecosystems. Cambridge University Press, Cam- 29: 1089–1097. bridge, pp. 393–410. Ulanowicz, R. E. 1986a. Growth and Development: Eco- Boudreau, P. R., and L. M. Dickie. 1989. Biological model systems Phenomenology. ToExcel Press, 203 pp. of fisheries production based on physiological and Ulanowicz, R. E. 1986b. A phenomenological perspective ecological scalings of body size. Canadian Journal of of ecological development. In: T. M. Poston and R. Fisheries and Aquatic Sciences 46: 614–623. Purdy (eds), and Environmental Fate, Boudreau, P. R., and L. M. Dickie. 1992. Biomass spectra Vol. 9. ASTM STP 921. American Society for Testing of aquatic ecosystems in relation to fisheries yield. and Materials, Philadelphia, PA, pp. 73–8l. Canadian Journal of Fisheries and Aquatic Sciences Ulanowicz, R. E. 1997. Ecology, the Ascendent Perspective. 49: 1528–1538. Columbia University Press, 201 pp. Boyle, P. R., and S. V. Bolettzky. 1996. Ulanowicz, R. E. 2002. The balance between adaptability populations: definition and dynamics. Philosophical and adaptation. BioSystems 64: 13–22. Transactions of the Royal Society 351: 985–1002. 228 REFERENCES

Brey, T. 1990. Estimating productivity of macrobenthic Duplisea, D. E., and S. R. Kerr. 1995. Application of a invertebrates from biomass and mean individual biomass size spectrum model to demersal fish data weight. Meeresforschung 32: 329–343. from the Scotian shelf. Journal of Theoretical Biology 177: Brey, T. 1999. Growth performance and mortality in 263–269. aquatic macrobenthic invertebrates. Advances in Marine Fenchel, T. 1974. Intrinsic rate of natural increase: the Biology 35: 153–223. relationship with body size. Oecologia 14: 317–326. Brown, J. H. 1995. Macroecology. University of Chicago, France, R., M. Chandler, and R. Peters. 1998. Mapping Press, Chicago, IL. trophic continua of benthic food webs: body size d15N Brown, J. H., and J. F. Gillooly. 2003. Ecological food relationships. Marine Ecology Progress Series 174: 301–306. webs: high-quality data facilitate theoretical unification. Fry, B., and R. B. Quinones. 1994. Biomass spectra and Proceedings of the National Academy of Sciences, USA stable-isotope indicators of trophic level in zooplankton 100: 1467–1468. of the northwest Atlantic. Marine Ecology Progress Series Brown, J. H., and G. B. West. 2000. Scaling in Biology. 112: 201–204. Oxford University Press, Oxford. Gaedke, U. 1993. Ecosystem analysis based on biomass Charnov, E. L. 1993. Life History Invariants: Some size distributions: a case study of a planktonic com- Explorations of Symmetry in . munity in a large lake. Limnology and Oceanography Oxford University Press, Oxford. 38: 112–127. Chase, J. M. 2000. Are there real differences among Gaston, K. J., and T. M. Blackburn. 2000. Pattern and aquatic and terrestrial food webs? Trends in Ecology and Process in Macroecology. Blackwell Science, Oxford. Evolution 15: 408–412. Gislason, H., and J. C. Rice. 1998. Modelling the response Cohen, J. E., T. Jonsson, and S. R. Carpenter, 2003. of size and diversity spectra of fish assemblages to Ecological community description using the food changes in exploitation. ICES Journal of Marine Sciences web, species abundance and body size. Proceedings 55: 362–370. of the National Academy of Sciences, USA 100: Gunderson, D. R., and P. H. Dygert. 1988. Reproductive 1781–1786. effort as a predictor of natural mortality rate. Journal du Cohen, J. E., S. L. Pimm, P. Yodzis, and J. Saldan˜a. 1993. Conseil, Conseil International pour l’Exploration de la Mer Body sizes of animal predators and animal prey in 44: 200–209. food webs. Journal of Animal Ecology 62: 67–78. Hall, S. J., and D. G. Raffaelli. 1993. Food webs: theory Cushing, D. H. 1975. Marine Ecology and Fisheries. and reality. Advances in Ecological Research 24: 187–239. Cambridge University Press, Cambridge. Harvey, P. H., and M. D. Pagel. 1991. The Comparative Cyr, H. 2000. Individual energy use and the of Method in Evolutionary Biology. Oxford University population density. In: J. H. Brown and G. B. West Press, Oxford. (eds), Scaling in Biology. Oxford University Press, Jennings, S., and J. L. Blanchard. 2004. Fish abundance Oxford, pp. 267–295. with no fishing: predictions based on macroecological Damuth, J. 1981. Population density and body size in theory. Journal of Animal Ecology 73: 632–642. mammals. Nature 290: 699–700. Jennings, S., and S. Mackinson. 2003. Abundance–body Denney, N. H., S. Jennings, and J. D. Reynolds. 2002. mass relationships in size structured food webs. Life history correlates of maximum population growth Ecology Letters 6: 971–974. rates in marine fishes. Proceedings of the Royal Society: Jennings, S., and K. J. Warr. 2003. Smaller predator–prey Biological Sciences 269: 2229–2237. body size ratios in longer food chains. Proceedings of the Dickie, L. M. 1976. Predation, yield and ecological Royal Society Biological Sciences 270: 1413–1417. efficiency in aquatic food chains. Journal of the Fisheries Jennings, S., J. K. Pinnegar, N. V. C. Polunin, and T. Boon. Research Board of Canada 33: 313–316. 2001. Weak cross-species relationships between body Dickie, L. M., S. R. Kerr, and P. R. Boudreau. 1987. Size- size and trophic level belie powerful size-based trophic dependent processes underlying regularities in structuring in fish communities. Journal of Animal ecosystem structure. Ecological Monographs 57: 233–250. Ecology 70: 934–944. Dinmore, T. A., and S. Jennings. 2004. Predicting body Jennings, S., J. K. Pinnegar, N. V. C. Polunin, and size distribution in benthic infaunal communities. K. J. Warr. 2002a. Linking size-based and trophic Marine Ecology Progress Series 276: 289–292. analyses of benthic community structure. Marine Duplisea, D. E. 2000. Benthic organsism size-spectra in Ecology Progress Series 226: 77–85. the Baltic Sea in relation to the sediment environment. Jennings, S., K. J. Warr, and Mackinson, S. 2002b. Use of Limnology and Oceanography 45: 558–568. size-based production and stable isotope analyses to REFERENCES 229

predict trophic transfer efficiencies and predator–prey Rice, J., and H. Gislason. 1996. Patterns of change body mass ratios in food webs. Marine Ecology Progress in the size spectra of numbers and diversity of the Series 240: 11–20. North Sea fish assemblage, as reflected in surveys Kerr, S. R. 1974. Theory of size distribution in ecological and models. ICES Journal of Marine Sciences 53: communities. Journal of the Fisheries Research Board of 1214–1225. Canada 31: 1859–1862. Rice, J. C., N. Daan, J. G. Pope, and H. Gislason. 1991. The Kerr, S. R., and L. M. Dickie. 2001. The Biomass Spectrum: stability of estimates of suitabilities in MSVPA over A Predator–Prey Theory of Aquatic Production. Columbia four years of data from predator stomachs. ICES University Press, New York. Marine Science Symposia 193: 34–45. Li, W. K. W. 2002. Macroecological patterns of phyto- Schwinghamer, P. 1981. Characteristic size distributions plankton in the northwestern Atlantic Ocean. Nature of integral benthic communities. Canadian Journal of 419: 154–157. Fisheries and Aquatic Sciences 38: 1255–1263. Murawski, S. A., and J. S. Idoine. 1992. Multispecies size Schwinghamer, P., B. Hargrave, D. Peer, and composition: a conservative property of exploited C. M. Hawkins. 1986. Partitioning of production and fishery systems. Journal of Northwest Atlantic Fishery respiration among size groups of organisms in an Science 14: 79–85. intertidal benthic community. Marine Ecology Progress Nee, S., A. F. Read, J. J. D. Greenwood, and P. H. Harvey. Series 31: 131–142. 1991. The relationship between abundance and body Sheldon, R. W., and S. R. Kerr. 1972. The population size in British birds. Nature 351: 312–313. density of monsters in Loch Ness. Limnology and Olive, P. J. W., J. K. Pinnegar, N. V. C. Polunin, Oceanography 17: 769–797. G. Richards, and R. Welch. 2003. Isotope trophic-step Sheldon, R. W., A. Prakash, and W. H. Sutcliffe. 1972. fractionation: a dynamic equilibrium model. Journal of The size distribution of particles in the Ocean. Animal Ecology 72: 608–617. Limnology and Oceanography 17: 327–340. Pauly, D. 1980. On the interrelationships between natural Slobodkin, L. B. 1960. Ecological energy relationships mortality, growth parameters and mean environmental at the population level. American Naturalist 94: 213–236. temperature in 175 fish stocks. Journal du Conseil, Sprules, W. G., and J. D. Stockwell. 1995. Size based bio- Conseil International pour l’Exploration de la Mer 39: mass and production models in the St Lawrence Great 175–192. Lakes. ICES Journal of Marine Sciences 52: 705–710. Peters, R. H. 1983. The Ecological Implications of Body Size. Stephens, D. W., and J. R. Krebs. 1986. Foraging Theory. Cambridge University Press, Cambridge. Princeton University Press, Princeton, NJ. Pimm, S. L. 1991. The balance of nature? Ecological Issues in Thiebaux,M.L.,andL.M.Dickie.1992.Modelsofaquatic the Conservation of Species and Communities. University biomass size spectra and the common structure of their of Chicago Press, Chicago, IL. solutions. Journal of Theoretical Biology 159: 147–161. Polunin, N. V. C., and J. K. Pinnegar. 2002. Ecology of Thiebaux, M. L., and L. M. Dickie. 1993. Structure of the fishes in marine food-webs. In: P. J. Hart and body size spectrum of the biomass in aquatic eco- J. D. Reynolds (eds), Handbook of Fish and Fisheries. systems: a consequence of allometry in predator–prey Blackwell, Oxford. interactions. Canadian Journal of Fisheries and Aquatic Pope, J. G., J. G. Shepherd, and J. Webb. 1994. Successful Sciences 50: 1308–1317. surf-riding on size spectra: the secret of survival in Ursin, E. 1973. On the prey size preferences of cod and the sea. Philosophical Transactions of the Royal Society dab. Meddelelser fra Danmarks Fiskeri-og Havundersogelser 343: 41–49. 7: 85–98. Pope, J. G., T. K. Stokes, S. A. Murawski, and S. I. Iodoine. Ware, D. M. 2000. Aquatic ecosystems: properties 1988. A comparison of fish size composition in and models. In: P. J. Harrison and T. R. Parsons (eds), the North Sea and on Georges Bank. In: W. Wolff, Fisheries Oceanography: An Integrative Approach to C. J. Soeder, and F. R. Drepper (eds), Ecodynamics: Fisheries Ecology and Management. Blackwell Science, Contributions to Theoretical Ecology. Springer Verlag, Oxford, pp. 161–194. Berlin, pp. 146–152. Post, D. M. 2002a. The long and short of food chain Chapter 9 length. Trends in Ecology and Evolution 17: 269–277. Post, D. M. 2002b. Using stable isotopes to estimate Abarca-Arenas, L. G., and R. E. Ulanowicz. 2002. The trophic position: models, methods and assumptions. effects of taxonomic aggregation on network analysis. Ecology 83: 703–718. Ecological Modelling 149: 285–296. 230 REFERENCES

Abarca-Arenas, L. G., and E. Valero-Pacheco. 1993. Baird, D., J. M. Glade, and R. E. Ulanowicz. 1991. Toward a trophic model of Tamiahua, a coastal lagoon The comparative ecology of six marine ecosystems. in Mexico. In: V. Christensen and D. Pauly (eds), Philosophical Transactions of the Royal Society of London B Trophic Models of Aquatic Ecosystems. ICLARM 333: 15–29. Conference Proceedings 26, pp. 181–185. Barrera-Oro, E. 2002. The role of fish in the Antarctic Ainley, D. G., W. R. Fraser, W. O. Smith, T. L. Hopkins, and marine food web: differences between inshore and J. J. Torres. 1991. The structure of upper level pelagic offshore waters in the southern Scotia Arc and west food webs in the Antarctic: effect of phytoplankton on Antarctic Peninsula. Antarctic Science 14: 293–309. distribution. Journal of Marine Systems 2: 111–122. Beattie, A., and M. Vaconcellos. 2002. Previous ecosystem Ainsworth, C., B. Ferriss, E. Leblond, and S. Guenette. models of Northern BC. University of British Columbia, 2001. The Bay of Biscay France; 1998 and 1970 models. Fisheries Centre Research Reports 10(1): 78–80. University of British Columbia, Fisheries Centre Research Bengtsson, J. 1994. Confounding variables and independ- Reports 9(4): 271–313. ent observations in comparative analyses of food webs. Alin˜o, P. M., L. T. McManus, J. W. McManus, C. Nan˜ola, Ecology 75: 1282–1288. M. D. Fortes, G. C. Trono, and G. S. Jacinto. 1993. Initial Bersier, L. F., C. Banasek-Richter, and M. F. Cattin. 2002. parameter estimations of a coral reef flat ecosystem Quantitative descriptors of food-web matrices. Ecology in Bolinao, Pangasinan, Northwestern Philippines. In: 83: 2394–2407. V. Christensen and D. Pauly (eds), Trophic Models of Bjornsson, H. 1998. Calculating capelin consumption Aquatic Ecosystems. ICLARM Conference Proceedings by Icelandic cod using a spatially disaggregated 26, pp. 252–258. simulation model. In: Fishery Stock Assessment Andersen, K. P., and E. Ursin. 1977. A multispecies Models. University of Alaska Sea Grant, AK-SG-98-01: extension to the Beverton and Holt theory, with 703–718. accounts of phosphorus circulation and primary Bogstad, B., K. H. Hauge, and O. Ulltang. 1997. production. Meddelelser fra Danmarks Fiskeri-og MULTSPEC—a multispecies model for fish and Havundersogelser, NS 7: 319–435. marine mammals in the Barents Sea. Journal of Arias-Gonzalez, J. E., B. Delesalle, B. Slavat, and R. Galzin. Northwest Atlantic Fishery Science 22: 317–341. 1997. Trophic functioning of the Tiahura reef sector, Borga, K., G. W. Gabrielsen, and J. U. Skaare. 2001. Moorea Island, French Polynesia. Coral Reefs 16: 231–246. Biomagnification of organochlorines along a Barents Arreguin-Sanchez, F., and L. E. Calderon-Aguilera. 2002. Sea food chain. Environmental Pollution 113: 187–198. Evaluating harvesting strategies for fisheries in the Bradford-Grieve, J., and S. Hanchet. 2002. Understanding central Gulf of California ecosystem. University of ecosystems: a key to managing fisheries? Water and British Columbia, Fisheries Centre Research Reports Atmosphere 10: 16–17. 10(2): 135–141. Bradford-Grieve, J. M., P. K. Probert, S. D. Nodder, Arreguin-Sanchez F., J. C. Seijo, and E. Valero-Pacheco. D. Thompson, J. Hall, S. Hanchet, P. Boyd, J. Zeldis, 1993. An application of ECOPATH II to the north central A. N. Baker, H. A. Best, N. Broekhuizen, S. Childerhouse, shelf ecosystem of Yucatan, Mexico. In: V. Christensen M. Clark, M. Hadfield, K. Safi, and I. Wilkinson. 2003. and D. Pauly (eds), Trophic Models of Aquatic Ecosystems. Pilot trophic model for subantarctic water over the ICLARM Conference Proceedings 26, pp. 269–278. Southern Plateau, New Zealand: a low biomass, high Atwell, L., K. A. Hobson, and H. E. Welch. 1998. transfer efficiency system. Journal of Experimental Biomagnification and of mercury in Marine Biology and Ecology 289: 223–262. an arctic marine food web: insights from stable Briand, F., and J. E. Cohen. 1984. Community food webs nitrogen isotope analysis. Canadian Journal of Fisheries have scale-invariant structure. Nature 307: 264–267. and Aquatic Sciences 55: 1114–1121. Briand, F., and J. E. Cohen. 1987. Environmental correlates Aydin, K. Y., V. V. Lapko, V. I. Radchenko, and of food chain length. Science 238: 956–960. P. A. Livingston. 2002. A comparison of the eastern Brodeur, R. D., and W. G. Pearcy. 1992. Effects of envir- and western Bering Sea shelf and slope ecosystems onmental variability on trophic interactions and food through the use of mass-balance food web models. web structure in a pelagic upwelling ecosystem. Marine NOAA Technical Memorandum NMFS-AFSC-130. Ecology Progress Series 84: 101–119. Azam, F., T. Fenchel, J. G. Field, J. S. Gray, L. A. Browder, J. A. 1993. A pilot model of the Gulf of Mexico Meyer-Reil, and F. Thingstad. 1983. The ecological role continental shelf. In: V. Christensen, and D. Pauly (eds), of water-column microbes in the sea. Marine Ecology Trophic Models of Aquatic Ecosystems. ICLARM Progress Series 10: 257–263. Conference Proceedings 26, pp. 279–284. REFERENCES 231

Buchary, E. 2001. Preliminary reconstruction of the Clarke, T. A., A. O. Flechsig, and R. W. Grigg. 1967. Icelandic marine ecosystem in 1950. University of Ecological studies during Project Sealab II. Science 157: British Columbia, Fisheries Centre Research Reports 9(4): 1381–1389. 198–206. Closs, G. P., S. R. Balcombe, and M. J. Shirley. 1999. Buchary, E., J. Alder, S. Nurhakim, and T. Wagey. 2002. Generalist predators, interaction strength and food- The use of ecosystem-based modelling to investigate web stability. Advances in Ecological Research 28: 93–126. multi-species management strategies for capture Cohen, J. E. 1989. Food webs and community structure. fisheries in the Bali Strait, Indonesia. University of In: J. Roughgarden, R. M. May, and S. A. Levin (eds), British Columbia, Fisheries Centre Research Reports Perspectives in Ecological Theory. Princeton University 10(2): 24–32. Press, Princeton, NJ, pp. 181–202. Bulman, C., F. Althaus, X. He, N. J. Bax, and A. Williams. Cohen, J. E. 1994. Marine and continental food webs: 2001. Diets and trophic guilds of demersal fishes of the three paradoxes? Philosophical Transactions of the Royal south-eastern Australian shelf. Marine and Freshwater Society of London B 343: 57–69. Research 52: 537–548. Cohen, J. E., and F. Briand. 1984. Trophics links of Bundy, A. 2001. Fishing on ecosystems: the interplay of community food web. Proceedings of the National fishing and predation in Newfoundland–Labrador. Academy of Sciences, USA 81: 4105–4109. Canadian Journal of Fisheries and Aquatic Sciences 58: Cohen, J. E., and C. M. Newman. 1988. Dynamic basis of 1153–1167. food web organization. Ecology 69: 1655–1664. Bundy, A., and D. Pauly. 2000. Selective harvesting by Cohen, J. E., F. Briand, and C. M. Newman. 1990. Community small-scale fisheries: ecosystem analysis of San Miguel Food Webs: Data and Theory. Springer-Verlag, New York. Bay, Philippines. Fisheries Research 53: 263–281 Cohen, J. E., R. A. Beaver, S. H. Cousins, D. L. DeAngelis, Bundy, A., G. R. Lilly, and P. A. Shelton. 2000. A Mass L. Goldwasser, K. L. Heong, R. D. Holt, A. J. Kohn, Balance Model of the Newfoundland-Labrador Shelf. J. H. Lawton, N. D. Martinez, R. O’Malley, L. M. Page, Canadian Technical Report of Fisheries and Aquatic B. C. Patten, S. L. Pimm, G. A. Polis, M. Rejmanek, Sciences, No. 2310, 157 p. T. W. Schoener, K. Schoenly, W. G. Sprules, J. M. Teal, Carr, M. H., T. W. Anderson, and M. A. Hixon. 2002. R. E. Ulanowicz, P. H. Warren, H. M. Wilbur, and Biodiversity, population regulation, and the stability of P. Yodzis. 1993. Improving food webs. Ecology 74: coral-reef fish communities. Proceedings of the National 252–258. Academy of Sciences, USA 99: 11241–11245. Constable, A. J. 2001. The ecosystem approach to man- Carr, M. H., J. E. Neigel, J. A. Estes, S. Andelman, aging fisheries: achieving conservation objectives for R. R. Warner, and J. L. Largier. 2003. Comparing predators of fished species. CCAMLR Science 8: 37–64. marine and terrestrial ecosystems: implications for the Constable, A. J., W. K. de la Mare, D. J. Agnew, I. Everson, design of coastal marine reserves. Ecological Applications and D. Miller. 2000. Managing fisheries to conserve the 13: S90-S107. Antarctic marine ecosystem: practical implementation Cheung, W.-L., R. Watson, and T. Pitcher. 2002. Policy of the Convention on the Conservation of Antarctic simulation of fisheries in the Hong Kong marine Marine Living Resources. ICES Journal of Marine ecosystem. University of British Columbia, Fisheries Sciences 57: 778–791. Centre Research Reports 10(2): 46–53. Cox, S. P., T. E. Essington, J. F. Kitchell, S. J. D. Martell, Cho, B. C., and F. Azam. 1988. Major role of bacteria in C. J. Walters, C. Boggs, and I. Kaplan. 2002. Recon- biogeochemical fluxes in the ocean’s interior. Nature structing ecosystem dynamics in the central pacific 332: 441–443. Ocean, 1952–1998. II. A preliminary assessment of the Christensen, V. 1995. A model of trophic interactions in the trophic impacts of fishing and effects on tuna North Sea in 1981, the year of the stomach. Dana 11: 1–28. dynamics. Canadian Journal of Fisheries and Aquatic Christensen, V. 1998. Fishery-induced changes in a marine Sciences 59: 1736–1747. ecosystem: insight from models of the Gulf of Thailand. Cousins, S. H. 1996. Food webs: from the Lindeman Journal of Fish Biology 53(A): 128–142. paradigm to a taxonomic general theory of ecology. Christensen, V., and D. Pauly. 1992. ECOPATH II—A In: G. A. Polis and K. O. Winemiller (eds), Food Webs: software for balancing steady-state models and cal- Integration of Patterns and Dynamics. Chapman and culating network characteristics. Ecological Modelling Hall, pp. 243–254. 61: 169–185. Crowder, L. B., D. P. Reagan, and D. W. Freckman. 1996. Clarke, G. L. 1946. Dynamics of production in a marine Food web dynamics and applied problems. In: area. Ecological Monographs 16: 323–335. G. A. Polis and K. O. Winemiller (eds), Food Webs: 232 REFERENCES

Integration of Patterns and Dynamics. Chapman and Godinot, O., and V. Allain. 2003. A preliminary Hall, New York, pp. 327–336. ECOPATH model of the warm pool pelagic ecosystem. de Ruiter, C. P., A. Neutal, and J. C. Moore. 1995. 16th Meeting of the Standing Committee on Tuna and Energetics, patterns of interaction strengths, and Billfish, Mooloolaba, Queensland, Australia, 9–16 July stability in real ecosystems. Science 269: 1257–1260. 2003. URL, available at: http://spc.org.nc/OceanFish/ Dommasnes, A., V. Christensen, B. Ellertsen, C. Kvamme, Html/SCTB/SCTB16/BBRG5.pdf. W. Melle, L. Nottestad, T. Pedersen, S. Tjelmeland, and Goldwasser, L., and J. Roughgarden. 1993. Construction D. Zeller. 2001. Ecosystem model of the Norwegian and analysis of a large Caribbean food web. Ecology 74: and Barents Seas. University of British Columbia, Fisheries 1216–1233. Centre Research Reports 9(4): 213–240. Goldwasser, L., and J. Roughgarden. 1997. Sampling Dunne, J. A., R. J. Williams, and N. D. Martinez. 2002a. effects and the estimation of food-web properties. Network structure and biodiversity loss in food webs: Ecology 78: 41–54. robustness increases with connectance. Ecology Letters Gomes, M. D. C. 1993. Predictions Under Uncertainty: 5: 558–567. Fish Assemblages and Food Webs on the Grand Banks Dunne, J. A., R. J. Williams, and N. D. Martinez. 2002b. of Newfoundland. Social and Economic Studies Food-web structure and network theory: the role of No. 51, ISER, Memorial University of Newfoundland, connectance and size. Proceedings of the National St John’s, Nfld. Academy of Sciences, USA 99: 12917–12922. Greenstreet, S. P. R., A. D. Bryant, N. Broekhuizen, Dunne, J. A., R. J. Williams, and N. D. Martinez. 2004. S. J. Hall, and M. R. Heath. 1997. Seasonal variation in Network structure and robustness of marine food the consumption of food by fish in the North Sea and webs. Marine Ecology Progress Series 273: 291–302. implications for food web dynamics. ICES Journal of Elton, C. S. 1958. The Ecology of Invasions by Animals and Marine Sciences 54: 243–266. Plants. Methuen, London. 181 p. Gribble, N. A. 2001. A model of the ecosystem and Emery, A. R. 1978. The basis for fish community structure: associated penaeid prawn community, in the far marine and freshwater comparisions. Environmental northern Great Barrier Reef. In: E. Wolanski (ed.), Biology of 3: 33–47. Oceanographic Process and Coral Reefs, Physical and Essington, T. E., J. F. Kitchell, C. Boggs, D. E. Schindler, Biological Links in the Great Barrier Reef. CRC Press, R. J. Olson, and R. Hilborn. 2002. Alternative fisheries New York, pp. 189–207. and the predation rate of yellowfin tuna in the eastern Guenette, S., and T. Morato. 2001. The Azores Archipe- Pacific Ocean. Ecological Applications 12: 724–734. lago in 1997. University of British Columbia, Fisheries Etter, R. J., and L. S. Mullineaux. 2001. Deep sea commu- Centre Research Reports 9(4): 241–270. nities. In: M. D. Bertness, S. D. Gaines, M. E. Hay, and Hall, S. J., and D. Raffaelli. 1991. Food-web patterns: M. A. Sunderland (eds), Marine Community Ecology. lessons from a species-rich web. Journal of Animal Sinauer Associates, Inc., pp. 724–734. Ecology 60: 823–842. Field, J. G., R. J. M. Crawfod, P. A. Wickens, Hall, S. J., and D. G. Raffaelli. 1993. Food webs: theory C. L. Moloney, K. L. Cochrane, and C. A. Villacastin- and reality. Advances in Ecological Research 24: 187–236. Herrero. 1991. Network analysis of Benguela pelagic Hardy, A. C., 1924. The herring in relation to its animate food webs. Benguela Ecology Programme, Workshop environment, Part 1. The food and feeding habits of the on Seal–Fishery Biological Interactions. University of herring. Fisheries Investigations London, Series II 7: 1–53. Cape Town, 16–20 September, 1991, SW91/M5a. Harvey, C. J., S. P. Cox, T. E. Essington, S. Hansson, and Fisk, A. T., K. A. Hobson, and R. J. Norstrom. 2001. J. F. Kitchell. 2003. An ecosystem model of food web Influence of chemical and biological factors on trophic and fisheries interactions in the Baltic Sea. ICES Journal transfer of persistent organic pollutants in the north- of Marine Sciences 60: 939–950. water polynya marine food web. Environmental Science Havens, K., 1992. Scale and structure in natural food and Technology 35: 732–738. webs. Science 257: 1107–1109. Fulton, B., and T. Smith. 2002. ECOSIM case study: Haydon, D., 1994. Pivotal assumptions determining Port Phillip Bay, Australia. University of British the relationship between stability and complexity: an Columbia, Fisheries Centre Research Reports 10(2): analytical synthesis of the stability–complexity debate. 83–93. American Naturalist 144: 14–29. Gardner, M. R., and W. R. Ashby. 1970. Connectance of Heymans, J., and D. Baird. 2000. Network analysis of the large dynamical (cybernetic) systems: critical values northern Benguela ecosystem by means of NETWRK for stability. Nature 228: 784. and ECOPATH. Ecological Modelling 131: 97–118. REFERENCES 233

Heymans, J. J., L. J. Shannon, and A. Jarre. 2004. Changes in Kauzinger, C. M. K., and P. J. Morin. 1998. Productivity the northern Benguela ecosystem over three decades: controls food-chain properties in microbial commun- 1970s, 1980s, 1990s. Ecological Modelling 172: 179–195. ities. Nature 395: 495–497. Hobson, K. A., and H. E. Welch. 1992. Determination of Kenny, D., and C. Loehle. 1991. Are food webs randomly trophic relationships within a high Arctic marine connected? Ecology 72: 1794–1797. food web using 13C and 15N analysis. Marine Ecology Kitchell, J. F., C. H. Boggs, X. He, and C. J. Walters. 1999. Progress Series 84: 9–18. Keystone predators in the Central Pacific. In: Ecosystem Hobson, K. A., W. G. Ambrose, and P. E. Renaud. 1995. Approaches for . University of Sources of primary production, benthic–pelagic Alaska Sea Grant, AK-SG-99-01: 665–684. coupling, and trophic relationships within the Kitchell, J. F., T. E. Essington, C. H. Boggs, D. E. Schindler, Northeast Water Polynya: insights from 13C and 15N and C. J. Walters. 2002. The role of sharks and longline analysis. Marine Ecology Progress Series 128: 1–10. fisheries in a Pelagic ecosystem of the Central Pacific. Hogetsu, K. 1979. Biological productivity of some regions Ecosystems 5: 202–216. of Japan. In: M. J. Dunbar, (ed), Marine Production Krause, A. E., K. A. Frank, D. M. Mason, R. E. Ulanowicz, Mechanisms. Cambridge University Press, Cambridge, and W. M. Taylor. 2003. Compartments revealed in pp. 71–87. food-web structure. Nature 426: 282–285. Hollowed, A. B., N. Bax, R. Beamish, J. Collie, M. Fogarty, Lindeman, R. L. 1942. The trophic-dynamic aspect of P. Livingston, J. Pope, and J. C. Rice. 2000. Application ecology. Ecology 23: 399–418. of multispecies models in assessment of impacts of Link, J. 2002. Does food web theory work for marine commercial fishing. Symposium proceedings on the ecosystems? Marine Ecology Progress Series 230: 1–9. ecosystem effects of fishing, ICES Journal of Marine Lotka, A. J. 1925. Elements of Physical Biology. Williams Sciences 57: 707–719. and Watkins, Baltimore, MD. Hop, H., K. Borga, G. W. Gbrielsen, L. Kleivane, and Lyons, K. G., and M. W. Schwartz. 2001. Rare species loss J. U. Skaare. 2002. Food web magnification of persist- alters ecosystem function—invasion resistance. Ecology ent organic pollutants in and home- Letters 4: 358–365. otherms from the Barents Sea. Environmental Science MacArthur, R. H. 1955. Fluctuation of animal populations and Technology 36: 2589–2597. and a measure of community stability. Ecology 36: Jackson, J. B. C., M. X. Kirby, W. H. Berger, K. A. Bjorndal, 533–536. L. W. Botsford, B. J. Bourque, R. H. Bradbury, R. Cooke, Manickchand-Heileman, S., F. Arreguin-Sanchez, A. Lara- J. Erlandson, J. A. Estes, T. P. Hughes, S. Kidwell, Dominguez, and L. A. Soto. 1998b. Energy flow and C. B. Lange, H. S. Lenihan, J. M. Pandolfi, C. H. Peterson, network analysis of Terminos Lagoon, SW Gulf of R. S. Steneck, M. J. Tegner, and R. R. Warner. 2001. Mexico. Journal of Fish Biology 53(A): 179–197. Historical overfishing and the recent collapse of coastal Manickchand-Heileman, S., L. A. Soto, and E. Escobar. ecosystems. Science 293: 629–638. 1998a. A preliminary trophic model of the continental Jarre, A., P. Muck, and D. Pauly. 1991. Two approaches shelf, South-western Gulf of Mexico. Estuarine, Coastal for modelling fish stock interactions in the Peruvian and Shelf Science 46: 885–899. upwelling ecosystem. ICES Marine Science Symposia Mann, K. H., and J. R. N. Lazier. 1991. Dynamics of Marine 193: 178–184. Ecosystems: Biological–Physical Interactions in the Oceans. Jarre-Teichmann, A., and D. Pauly. 1993. Seasonal Blackwell Scientific Publishers, Oxford. changes in the Peruvian Upwelling ecosystem. In: Martinez, N. D. 1991. Artifacts or attributes? Effects of V. Christensen and D. Pauly (eds), Trophic Models of resolution on the Little Rock Lake food web. Ecological Aquatic Ecosystems. ICLARM Conference Proceedings Monographs 61: 367–392. 26, pp. 307–314. Martinez, N. D. 1992. Constant connectance in community Jarre-Teichmann, A., L. Shannon, C. Moloney, and food webs. American Naturalist 139: 1208–1218. P. Wickens. 1998. Comparing trophic flows in the Martinez, N. D. 1993. Effects of resolution of food web southern Benguela to those in other upwelling ecosys- structure. Oikos 60: 403–412. tems. South African Journal of Marine Science 19:391–414. Martinez, N. D. 1994. Scale-dependent constraints on Kaehler, S., E. A. Pakhomov, and C. D. McQuaid. 2000. food-web structure. American Naturalist 144: 935–953. Trophic structure of the marine food web at the Prince Martinez, N. D., B. A. Hawkins, H. A. Dawah, and Edward Islands (Southern Ocean) determined by B. P. Feifarek. 1999. Effects of sampling effort on 13C and 15N analysis. Marine Ecology Progress Series characterization of food-web structure. Ecology 80: 208: 13–20. 1044–1055. 234 REFERENCES

May, R. M. 1972. Will a large complex system be stable? Olson, R. J., and G. M. Watters. 2003. A model of the Nature 238: 413–414. pelagic ecosystem in the eastern Tropical Pacific Ocean. May, R. M. 1973. Stability and Complexity in Model Inter-American Tropical Tuna Commission Bulletin 22(3): Ecosystems. Princeton University Press, Princeton, NJ. 135–218. McCann, K., A. Hastings, and G. R. Huxel. 1998. Weak Opitz, S. 1993. A quantitative model of the trophic trophic interactions and the balance of nature. Nature interactions. in a Caribbean coral reef ecosystem. In: 395: 794–798. V. Christensen and D. Pauly (eds), Trophic Models of Mendoza, J. J. 1993. A preliminary biomass budget for the Aquatic Ecosystems. ICLARM Conference Proceedings northeastern Venezuela shelf. In: V. Christensen and 26, pp. 269–278. D. Pauly (eds), Trophic Models of Aquatic Ecosystems. Ortiz, M., and M. Wolff. 2002. Dynamical simulation of ICLARM Conference Proceedings 26, pp. 285–297. mass-balance trophic models for benthic communities Mendy, A. N., and E. Buchary. 2001. Constructing the of north-central Chile: assessment of resilience time Icelandic marine ecosystem model for 1997 using a under alternative management scenarios. Ecological mass balance modeling approach. University of British Modelling 148: 277–291. Columbia, Fisheries Centre Research Reports 9(4): 182–197. Paine, R. T. 1966. Food web complexity and species Mohammed, E. 2001. A preliminary model for the Lan- diversity. American Naturalist 100: 65–75. caster Sound Region in the 1980s. University of British Paine, R. T. 1988. Food webs: road maps of interactions Columbia, Fisheries Centre Research Reports 9(4): 99–110. or grist for theoretical development? Ecology 69: Murphy, E. J. 1995. Spatial structure of the Southern 1648–1654. Ocean ecosystem: predator-prey linkages in Southern Pauly, D., and V. Christensen. 1993. Stratified models of Ocean food webs. Journal of Animal Ecology 64: 333–347. large marine ecosystems: a general approach and an National Marine Fisheries Service (NMFS). 1999. application to the South China Sea. In: K. Sherman, Ecosystem-Based Fishery Management. A Report to L. M. Alexander, and B. D. Gold (eds), Large Marine Congress by the Ecosystems Principles Advisory Ecosystems: Stress, Mitigation, and Sustainability. AAAS Panel. US DOC, Silver Spring, MD. Press, Washington, DC, pp. 148–174. Niquil, N., J. E. Arias-Gonzalez, B. Delesalle, and Pauly, D., V. Christensen, J. Dalsgaard, R. Froese, and R. E. Ulanowicz. 1999. Characterization of the plank- F. Torres. 1998. Fishing down marine food webs. tonic food web of Takapoto Atoll lagoon, using Science 279: 860–863. network analysis. Oecologia 118: 232–241. Pedersen, S. A. 1994. Multispecies interactions on the Odum, E. P. 1960. Organic production and turnover in offshore West Greenland shrimp grounds. Journal of old field succession. Ecology 41: 34–49. Marine Sciences, ICES C.M. Documents 1994/P:2. Odum, H. T. 1957. Trophic structure and productivity of Pedersen, S., and D. Zeller. 2001. Multispecies inter- Silver Springs, Florida. Ecological Monographs 27: 55–112. actions in the West Greenland ecosystem: importance Okey, T. A. 2001. A ‘‘straw-man’’ Ecopath model of of the shrimp fisheries. University of British Columbia, the Middle Atlantic Bight continental shelf, United Fisheries Centre Research Reports 9(4): 111–127. States. University of British Columbia, Fisheries Centre Petersen, C. G. J. 1918. The sea bottom and its production of Research Reports 9(4): 151–166. fish food. Reports of the Danish Biological Station 25: 1–62. Okey, T. A., and D. Pauly. 1999. A mass-balanced Pimm, S. L. 1982. Food Webs. Chapman and Hall, London. model of trophic flows in Prince William Sound: Pimm, S. L., and J. C. Rice. 1987. The Dynamics of Mul- de-compartmentalizing ecosystem knowledge. In: tispecies, Multi-Life-Stage Models of Aquatic Food Ecosystem Approaches for Fisheries Management. Univer- webs. Theoretical Population Biology 32: 303–325. sity of Alaska Sea Grant, AK-SG-99-01, pp. 621–635. Pimm, S. L., J. H. Lawton, and J. E. Cohen. 1991. Food web Okey, T. A., and R. Pugilese 2001. A preliminary patterns and their consequences. Nature 350: 669–674. Ecopath model of the Atlantic continental shelf adja- Polis, G. A. 1991. Complex trophic interactions in : cent to the southeastern United States. University an empirical critique of food-web theory. American of British Columbia, Fisheries Centre Research Reports Naturalist 138: 123–155. 9(4): 167–181. Polis, G. A., and D. R. Strong. 1996. Food web complexity Olivieri, R. A., A. Cohen, and F. P. Chavez. 1993. An and community dynamics. American Naturalist 147: ecosystem model of Monterey Bay, California. In: 813–846. V. Christensen and D. Pauly (eds), Trophic Models of Polovina, J. J. 1984. Model of a coral reef ecosystem. Aquatic Ecosystems. ICLARM Conference Proceedings I. The ECOPATH model and its application to French 26, pp. 315–322. Frigate Shoals. Coral Reefs 3: 1–11. REFERENCES 235

Pomeroy, L. R. 2001. Caught in the food web: complexity Aquatic Ecosystems. ICLARM Conference Proceedings made simple? Scientia Marina 65: 31–40. 26, pp. 300–306. Power, M. E., and L. S. Mills. 1995. The keystone cops Solow, A., and A. Beet. 1998. On lumping species in food meet in Hilo. Trends in Ecology and Evolution 10: 182–184. webs. Ecology 79: 2013–2018. Purcell, J. E. 1986. Predation on fish eggs and larvae by Stanford, R., K. Lunn, and S. Guenette. 2001. A pre- pelagic cnidarians and ctenophores. Bulletin of Marine liminary ecosystem model for the Atlantic coast Science 37: 739–755. of Morocco in the mid-1980s. University of British Qiyong, Z., L. Qiumian, L. Youtong, and Z. Yueping. Columbia, Fisheries Centre Research Reports 9(4): 314–344. 1981. Food web of fishes in Minnan-Taiwanchientan Steele, J. H. 1985. A comparison of terrestrial and marine fishing ground. Acta Oceanologica Sinica 3: 275–290. ecological systems. Nature 313: 355–358. Raffaelli, D. 2000. Trends in research on shallow water Steele, J. H. 1991. Can ecological theory cross the food webs. Journal of Experimental Marine Biology and land–sea boundary? Journal of Theoretical Biolgoy Ecology 250: 223–232. 153: 425–436. Raffaelli, D. G., and S. J. Hall 1996. Assessing the relative Sugihara, G., K. Schoenly, and A. Trombla. 1989. Scale importance of trophic links in food webs. In: G. A. Polis invariance in food web properties. Science 245: 48–52. and K. O. Winemiller (eds), Food Webs: Integration of Tittlemier, S. A., A. T. Fisk, K. A. Hobson, and Patterns and Dynamics. Chapman and Hall, New York, R. J. Norstrom. 2002. Examination of the bioaccumu- pp. 185–191. lation of halogenated dimethyl bipyrroles in an Arctic Reagan, D. P., G. R. Camilo, and R. B. Waide. 1996. The marine food web using stable nitrogen isotope analy- community food web: major properties and patterns of sis. Environmental Pollution 116: 85–93. organization. In: D. P. Reagan and R. B. Waide (eds), Tjelmeland, S., and B. Bogstad. 1998. MULTSPEC—a The Food Web of a Tropical Rain Forest. University of review of a multispecies modelling project for the Chicago Press, Chicago, IL, pp. 461–510. Barents Sea. Fisheries Research 37: 127–142. Reid, K., and J. P. Croxall. 2001. Environmental response Trites, A. W., P. A. Livingston, S. Mackinson, of upper trophic-level predators reveals a system M. C. Vasconcellos, A. M. Springer, and D. Pauly. 1999. change in an Antarctic marine ecosystem. Proceedings of Ecosystem change and the decline of marine mammals the Royal Society of London B 268: 377–384. in the eastern Bering Sea: testing the ecosystem Rysgaard, S., T. G. Nielsen, and B. W. Hansen. 1999. shift and commercial hypotheses. University Seasonal variation in nutrients, pelagic primary pro- of British Columbia, Fisheries Centre Research Reports duction and grazing in a high-Arctic coastal marine 7(1): 106 p. ecosystem, Young Sound, Northeast Greenland. Marine Vega-Cendejas, M. E., F. Arreguin-Sanchez, and Ecology Progress Series 179: 13–25. M. Hernandez. 1993. Trophic fluxes on the Campeche Sandberg, J., R. Elmgren, and F. Wulff. 2000. Carbon Bank, Mexico. In: V. Christensen and D. Pauly (eds), flows in Baltic Sea food webs—a re-evaluation using Trophic Models of Aquatic Ecosystems. ICLARM a mass balance approach. Journal of Marine Systems Conference Proceedings 26, pp. 206–213. 25: 249–260. Venier, J. M., and D. Pauly. 1997. Trophic dynamics of Schlesinger, W. H. 1997. Biogeochemistry, An Analysis of a Florida Keys coral reef ecosystem. Proceedings of the Global Change. Academic Press, San Diego. 8th International Coral Reef Symposium 1: 915–920. Schoener, T. W. 1989. Food webs from the small to the Verity, P. G. and V. Smetacek. 1996. Organism life large. Ecology 70: 1559–1589. cycles, predation, and the structure of marine pelagic Shannon, L. J., and A. Jarre-Teichmann. 1999. A model of ecosystems. Marine Ecology Progress Series 130: 277–293. trophic flows in the northern Bengula upwelling Volterra, V. 1926. Fluctuations in the abundance of a system during the 1980s. South African Journal of Marine species considered mathematically. Nature 118: 558–560. Science 21: 349–366. Walters, C., V. Christensen, and D. Pauly. 1997. Struc- Sherr, E. B., B. F. Sherr, and G. A. Paffenhoeffer. turing dynamic models of exploited ecosystems from 1986. Phagotrophic protozoa as food for metazoans: trophic mass-balance assessments. Reviews in Fish A ‘‘missing’’ trophic link in marine pelagic food webs? Biology and Fisheries 7: 139–172. Marine Microbial Food Webs 1: 61–80. Warren, P. H. 1994. Making connections in food webs. Silvestre, G., S. Selvanathan, and A. H. M. Salleh. 1993. Trends in Ecology and Evolution 9: 136–141. Preliminary trophic model of the coastal fishery Welch, H. E., M. A. Bergmann, T. D. Siferd, K. A. Martin, resources of Brunei Darussalam, South China Sea. In: M. F. Curtis, R. E. Crawford, R. J. Conover, and H. Hop. V. Christensen and D. Pauly (eds), Trophic Models of 1992. Energy flow through the marine ecosystem of 236 REFERENCES

the Lancaster Sound region, arctic Canada. Arctic Christensen, V., and C. J. Walters. 2004. Ecopath 45: 343–357. with Ecosim: methods, capabilities and limitations. Whipple, S., J. S. Link, L. P. Garrison, and M. J. Fogarty. Ecological Modelling 172: 109–139. 2000. Models of predation and fishing mortality in Cohen, J. E., F. Briand, and C. M. Newman. 1990. aquatic ecosystems. Fish and Fisheries 1: 22–40. Community Food Webs: Data and Theory. Springer- Williams, R. J., and N. D. Martinez. 2000. Simple rules Verlag, New York. yield complex food webs. Nature 6774: 180–183. Daily, G. C. (ed.). 1997. Nature’s Services: Societal Dependence Winemiller, K. O. 1990. Spatial and temporal variation in on Natural Ecosystems. , Washington DC. tropical fish trophic networks. Ecological Monographs De Angelis, D. L. 1975. Stability and connectance in food 60: 27–55. web models. Ecology 56: 238–243. Winemiller, K. O., and G. A. Polis. 1996. Food webs: what de Ruiter, P. C., A.-M. Neutel, and J. C. Moore. 1995. can they tell us about the world? In: G. A. Polis and Energetics, patterns of interaction strengths, and K. O. Winemiller (eds), Food Webs: Integration of stability in real ecosystems. Science 269: 1257–1260. Patterns and Dynamics. Kluwer Academic Publishers, Diehl, S., and M. Feissel. 2001. Intraguild prey suffer from Boston, pp. 1–22. enrichment of their resources: a microcosm experiment Wolff, M. 1994. A trophic model for Tongoy Bay—a with ciliates. Ecology 82: 2977–2983. system exposed to suspended scallop culture (Northern Drossel, B., P. G. Higgs, and A. J. McKane. 2001. Chile). Journal of Experimental Marine Biology and Ecology The influence of predator–prey population dynamics 182: 149–168. on the long-term evolution of food web structure. Yodzis, P. 1998. Local trohpodynamics and the interaction Journal of Theoretical Biology 208: 91–107. of marine mammals and fisheries in the Benguela Dunne, J. A., R. J. Williams, and N. D. Martinez. 2002a. ecosystem. Journal of Animal Ecology 67: 635–658. Food web structure and network theory: the role of Yodzis, P. 2000. Diffuse effects in food webs. Ecology 81: connectance and size. Proceedings of the National 261–266. Academy of Sciences, USA 99: 12917–12922. Zeller, D., and K. Freire. 2001. A preliminary North-east Dunne, J. A., R. J. Williams, and N. D. Martinez. 2002b. Atlantic marine ecosystem model: Faroe islands and Network structure and biodiversity loss in food webs: ICES area Vb. University of British Columbia, Fisheries robustness increases with connectance. Ecology Letters Centre Research Reports 9(4): 207–212. 5: 558–567. Dunne, J. A., R. J. Williams, and N. D. Martinez. 2004. Chapter 10 Network structure and robustness of marine food webs. Marine Ecology Progress Series 273: 291–302. Albert, R., J. Jeong, and A.-L. Barabasi. 2000. Error and Elton, C. S. 1927. Animal Ecology. Sidgwick and Jackson, attack tolerance of complex networks. London. Baird, D., and R. E. Ulanowicz. 1989. The seasonal Elton, C. S. 1958. Ecology of Invasions by Animals and dynamics of the Chesapeake Bay ecosystem. Ecological Plants. Chapman and Hall, London. Monographs 59: 329–364. Field, J. G., R. J. M. Crawford, P. A. Wickens, Berlow, E. 1999. Strong effects of weak interactions in C. L. Moloney, K. L. Cochrane, and C. A. Villacastic- ecological communities. Nature 398: 330–334. Herraro. 1991. Network analysis of Benguela pelagic Brose, U., N. D. Martinez, and R. J. Williams. 2003a. food webs. Benguela Ecology Programme, Workshop Estimating species richness: sensitivity to sample on Seal-Fishery Biological Interactions. University of coverage and insensitivity to spatial patterns. Ecology Cape Town, 16–20 September, 1991, BEP/SW91/M5a. 84: 2364–2377. Fussmann, G. F., and G. Heber. 2002. Food web Brose, U., R. J. Williams, and N. D. Martinez. 2003b. complexity and chaotic population dynamics. Ecology Comment on ‘‘Foraging adaptation and the relationship Letters 5: 394–401. between food-web complexity and stability.’’ Science Garlaschelli, D., G. Caldarelli, and L. Pietronero. 2003. 301: 918b. Universal scaling relations in food webs. Nature 423: Caldarelli, G., P. G. Higgs, and A. J. McKane. 1998. 165–168. Modelling coevolution in multispecies communities. Girvan, M., and M. E. J. Newman. 2002. Community Journal of Theoretical Biology 193: 345–358. structure in social and biological networks. Proceedings Camacho, J., R. Guimera, and L. A. N. Amaral. 2002. of the National Academy of Sciences, USA 99: 8271–8276. Analytical solution of a model for complex food webs. Hairston, N. G., J. D. Allen, R. K. Colwell, D. J. Futuyma, Physical Review E 65: 030901(R). J. Howell, M. D. Lubin, J. Mathias, and REFERENCES 237

J. H. Vandermeer. 1968. The relationship between May, R. M. 1973. Stability and Complexity in Model species diversity and stability: and experimental Ecosystems, 2nd edn. Princeton University Press, approach with protozoa and bacteria. Ecology 49: Princeton, NJ. 1091–1101. May, R. M. 1974. Biological populations with non- Hassell, M. P. 1978. The Dynamics of Arthropod overlapping generations: stable points, stable cycles, Predator–Prey Systems. Princeton University Press, and chaos. Science 186: 645–647. Princeton, NJ. May, R. M. 1976. Simple mathematical models with very Hastings, A., and T. Powell. 1991. Chaos in a three- complicated dynamics. Nature 261: 459–467. species food chain. Ecology 72: 896–903. May, R. M. 2001. Stability and Complexity in Model Hastings, A., C. L. Hom, S. Ellner, P. Turchin, and Ecosystems. Princeton Landmarks in Biology Edition. H. C. J. Godfray. 1993. Chaos in ecology: is mother Princeton University Press, Princeton, NJ. nature a strange attractor. Annual Review of Ecology and McCann, K. 2000. The diversity–stability debate. Nature Systematics 24: 1–33. 405: 228–233. Hector, A. et al. (34 authors). 1999. Plant diversity and McCann, K., and A. Hastings. 1997. Re-evaluating the productivity experiments in European grasslands. omnivory–stability relationship in food webs. Proceed- Science 286: 1123–1127. of the Royal Society London B 264: 1249–1254. Holling, C. S. 1959. Some characteristics of simple types McCann, K., and P. Yodzis. 1994a. Biological conditions of predation and . Canadian Entomology for chaos in a three-species food chain. Ecology 75: 91: 385–399. 561–564. Holling, C. S. 1966. The strategy of building models McCann, K., and P. Yodzis. 1994b. Nonlinear dynamics of complex ecological systems. In: K. E. F. Watt (ed.), and population disappearances. American Naturalist in Ecology. Academic Press, New 144: 873–879. York, pp. 195–214. McCann, K., A. Hastings, and G. R. Huxel. 1998. Weak Hutchinson, G. E. 1959. Homage to Santa Rosalia, or why trophic interactions and the balance of nature. Nature are there so many kinds of animals? American Naturalist 395: 794–798. 93: 145–159. Moore, J. C., and H. W. Hunt. 1988. Resource compart- Kondoh, M. 2003a. Foraging adaptation and the relation- mentation and the stability of real ecosystems. Nature ship between food-web complexity and stability. 333: 261–263. Science 299: 1388–1391. Murdoch, W. W., and A. Oaten. 1975. Predation and Kondoh, M. 2003b. Response to Comment on ‘‘Foraging population stability. Advances in Ecological Research adaptation and the relationship between food-web 9: 1–131. complexity and stability.’’ Science 301: 918c. Naeem, S., L. J. Thompson, S. P. Lawlor, J. H. Lawton, Krause, A. E., K. A. Frank, D. M. Mason, R. E. Ulanowicz, and R. M. Woodfin. 1994. Declining biodiversity can and W. W. Taylor. 2003. Compartments revealed in alter the performance of ecosystems. Nature 368: food-web structure. Nature 426: 282–285. 734–737. Lawlor, L. R. 1978. A comment on randomly constructed Neutel, A.-M., J. A. P. Heesterbeek, and P. C. deRuiter. model ecosystems. American Naturalist 112: 445–447. 2002. Stability in real food webs: weak links in long Lindeman, R. L. 1942. The tropic-dynamic aspect of loops. Science 296: 1120–1123. ecology. Ecology 23: 399–418. Odum, E. 1953. Fundamentals of Ecology. Saunders, MacArthur, R. H. 1955. Fluctuations of animal popu- Philadelphia, PA. lations and a measure of community stability. Ecology Paine, R. T. 1966. Food web complexity and species 36: 533–536. diversity. American Naturalist 100: 65–75. Martinez, N. D. 1991. Artifacts or attributes? Effects of Pimm, S. L., and J. H. Lawton. 1978. On feeding on more resolution on the Little Rock Lake food web. Ecological than one trophic level. Nature 275: 542–544. Monographs 61: 367–392. Pimm, S. L., and J. H. Lawton. 1977. Number of trophic Martinez, N. D. 1992. Constant connectance in commu- levels in ecological communities. Nature 268: 329–331. nity food webs. American Naturalist 140: 1208–1218. Polis, G. A. 1991. Complex desert food webs: an empirical Martinez, N. D., B. A. Hawkins, H. A. Dawah, and critique of food web theory. American Naturalist 138: B. Feifarek. 1999. Effects of sample effort on character- 123–155. ization of food-web structure. Ecology 80: 1044–1055. Post, D. M., M. E. Conners, and D. S. Goldberg. 2000. May, R. M. 1972. Will a large complex system be stable? Prey preference by a top predator and the stability of Nature 238: 413–414. linked food chains. Ecology 81: 8–14. 238 REFERENCES

Skalski, G. T., and J. F. Gilliam. 2001. Functional responses webs. In: G. A. Polis and K. O. Winemiller (eds), Food with predator interference: viable alternatives to the Webs: Integration of Patterns and Dynamics. Chapman Holling type II model. Ecology 82: 3083–3092. and Hall, New York, pp. 371–395. Strogatz, S. H. 2001. Exploring complex networks. Nature Bolker, B., M. Holyoak, V. Krivan, L. Rowe, and 410: 268–276. O. Schmitz. 2003. Connecting theoretical and empirical Tilman, D., P. B. Reich, J. Knops, D. Wedin, T. Mielke, studies of trait-mediated interactions. Ecology 84: and C. Lehman. 2001. Diversity and productivity in a 1101–1114. long-term grassland experiment. Science 294: 843–845. Brose, U., R. J. Williams, and N. D. Martinez. 2003. Ulanowicz, R. E. 2002. The balance between adaptability Comments on ‘‘foraging adaptation and the relation- and adaptation. BioSystems 64: 13–22. ship between food-web complexity and stability.’’ Watts, D. J., and S. H. Strogatz. 1998. Collective dynamics Science 301: 917c. of ‘small-world’ networks. Nature 393: 440–442. Chen, X., and J. E. Cohen. 2001a. Transient dynamics and Williams, R. J., and N. D. Martinez. 2000. Simple rules food-web complexity in the Lotka-Volterra cascade yield complex food webs. Nature 404: 180–183. model. Proceedings of The Royal Society of London, Series B Williams, R. J., and N. D. Martinez. 2004a. Trophic levels 268: 869–877. in complex food webs: theory and data. American Chen, X., and J. E. Cohen. 2001b. Global stability, local Naturalist 163: 458–468. stability and permanence in model food webs. Journal Williams, R. J., and N. D. Martinez. 2004b. Stabilization of Theoretical Biology 212: 223–235. of chaotic and non-permanent food web dynamics. DeAngelis, D. L. 1975. Stability and connectance in food European Physics Journal B 38: 297–303. web models. Ecology 56: 238–243. Williams, R. J., and N. D. Martinez. 2004c. Diversity, De Ruiter, P. C., A. Neutel, and J. C. Moore. 1995. complexity, and persistence in large model ecosystems. Energetics, patterns of interaction strengths, and Santa Fe Institute Working Paper 04–07–022. stability in real ecosystems. Science 269: 1257–1260. Yodzis, P. 1981. The stability of real ecosystems. Nature Dunne, J. A., R. J. Williams, and N. D. Martinez. 2002. 289: 674–676. Network structure and biodiversity loss in food webs: Yodzis, P. 1998. Local trophodynamics and the inter- robustness increases with connectance. Ecology Letters action of marine mammals and fisheries in the Benguela 5: 558–567. ecosystem. Journal of Animal Ecology 67: 635–658. Emlen, J. M. 1966. The role of time and energy in food Yodzis, P. 2000. Diffuse effects in food webs. Ecology 81: preference. American Naturalist 100: 611–617. 261–266. Emmerson, M. C., and D. G. Raffaelli. 2004. Predator– Yodzis, P. 2001. Must top predators be culled for the sake prey body size, interaction strength and the stability of of fisheries? Trends in Ecology and Evolution 16: 78–84. a real food web. Journal of Animal Ecology 73: 399–409. Yodzis, P., and S. Innes. 1992. Body-size and consumer– Endler, J. A. 1986. Defence against predators. In: resource dynamics. American Naturalist 139: 1151–1173. M. E. Feder and G. V. Lauder (eds), Predator–Prey Relationships. University of Chicago Press, Chicago, IL, pp. 109–134. Chapter 11 Fryxell, L. M., and P. Lundberg, 1997. Individual Behavior Abrams, P. A. 1982. Functional responses of optimal and Community Dynamics. Chapman and Hall, New York. foragers. The American Naturalist 120: 382–390. Gilpin, M. E. 1975. Stability of feasible predator–prey Abrams, P. A. 1984. Foraging time optimization and inter- systems. Nature 254: 137–139. actions in food webs. American Naturalist 124: 80–96. Goldwasser, L., and J. Roughgarden. 1993. Construction Abrams, P. A. 1992. Predators that benefit prey and prey of a large Caribbean food web. Ecology 74: 1216–1233. that harm predators: unusual effects of interacting Haydon, D. T. 2000. Maxmully stable model ecosystems foraging adaptations. American Naturalist 140: 573–600. can be highly connected. Ecology 81: 2631–2636. Abrams, P. A. 2000. The evolution of predator–prey Holling, C. S. 1959. Some characteristics of simple types interactions: theory and evidence. Annual Review of of predation and parasitism. Canadian Entomologist 91: Ecological Systems 31: 79–105. 385–398. Abrams, P. A., and H. Matsuda. 1996. Positive indirect Holt, R. D. 1977. Predation, apparent competition, and the effects between prey species that share predators. structure of prey communities. Theoretical 77: 610–616. Biology 12: 197–229. Abrams, P. A., B. A. Menge, G. G. Mittelbach, D. Spiller, Holt, R. D. 1983. Optimal foraging and the form of the and P. Yodzis. 1996. The role of indirect effects in food predator isocline. American Naturalist 122: 521–541. REFERENCES 239

Holt, R. D., and G. A. Polis. 1997. A theoretical framework May, R. M. 1973. Stability and Complexity in Model for intraguild predation. American Naturalist 149: Ecosystems. Princeton University Press, Princeton, NJ. 745–754. Martinez, N. D. 1991. Artifacts or attributes? Effects of Hughes, R. N. 1990. Behavioural Mechanisms of Food resolution on the Little Rock Lake food web. Ecological Selection, NATO-ASI Series, Vol. G20. Springler-Verlag, Monographs 61: 367–392. Berlin. McCann, K. and A. Hastings. 1997. Re-evaluating the Ives, A. R., and A. P. Dobson. 1987. Antipredator behavior omnivory-stability relationship in food webs. Proc. R. and the population dynamics of simple predator–prey Soc. Lond. B. 264: 1249–1254. systems. American Naturalist 130:431–437. McCann, K., A. Hastings, and G. R. Huxel. 1998. Weak Keitt, T. H. 1997. Stability and complexity on a lattice: trophic interactions and the balance of nature. Nature coexistence of species in an individual-based food web 395: 794–798. model. Ecological Modelling 102: 243–258. Menge, B. A. 1994. Indirect effects in marine rocky Kondoh, M. 2003a. Foraging adaptation and the relation- intertidal interaction webs: pattern and importance. ship between food-web complexity and stability. Ecological Monographs 65: 21–74. Science 299: 1388–1391. Neutel, A., J. A. P. Heesterbeek, and P. C. de Ruiter. 2002. Kondoh, M. 2003b. Response to comment on ‘‘foraging Stability in real food webs: weak links in long loops. adaptation and the relationship between food-web Science 296: 1120–1123. complexity and stability.’’ Science 301: 918c. Nunney, L. 1980. The stability of complex model Kr˘ivan, V. 1998. Effects of optimal antipredator behavior ecosystems. American Naturalist 115: 639–649. of prey on predator–prey dynamics: the role of refuges. Paine, R. T. 1988. Food webs: road maps of interactions or Theoretical Population Biology 53: 131–142. grist for theoretical development? Ecology 69: 1648–1654. Kr˘ivan, V. 2000. Optimal intraguild foraging and popu- Pelletier, J. D. 2000. Are large complex ecosystems more lation stability. Theoretical Population Biology 58: 79–94. unstable? A theoretical reassessment with predator Law, R., and R. D. Morton. 1996. Permanence and the switching. Mathematical Biosciences 163: 91–96. assembly of ecological communities. Ecology 77: 762–775. Petchey, O. L. 2000. Prey diversity, prey composition, Lawlor, L. R. 1978. A comment on randomly constructed and predator population dynamics in experimental model ecosystems. American Naturalist 112: 445–447. microcosms. Journal of Animal Ecology 69: 874–882. Lima, S. L. 1992. Life in a multi-predator environment: Pimm, S. L. 1991. The Balance of Nature?: Ecological Issues some considerations for antipredatory vigilance. in the Conservation of Species and Communities. University Annals of Zoology—Fennica 29: 217–226. of Chicago Press, Chicago, IL. Lima, S. L., and L. M. Dill. 1996. Behavioral decisions Pimm, S. L., and J. H. Lawton. 1977. On the number of made under the risk of predation: a review and pro- trophic levels. Nature 268: 329–331. spectus. Candian Journal of Zoology 68: 619–640. Pimm, S. L., and J. H. Lawton, 1978. On feeding on more MacArthur, R. H., and E. R. Pianka. 1966. On optimal than one trophic level. Nature 275: 542–544. use of a patchy environment. American Naturalist 100: Polis, G. A. 1991. Complex desert food webs: an empirical 603–609. critique of food web theory American Naturalist 138: MacNamara, J. M., and A. I. Houston. 1987. Starvation 123–155. and predation as factors limiting population size. Post, W. M., and S. L. Pimm. 1983. Community assembly Ecology 68: 1515–1519. and food web stability. Mathematical Bioscience 64: Matsuda, H., P. A. Abrams, and M. Hori. 1993. The effect 169–192. of adaptive antipredator behavior on exploitative Post, D. M., M. E. Conners, and D. S. Goldberg. 2000. competition and mutualism between predators. Oikos Prey preference by a top predator and the stability of 68: 549–559. linked food chains. Ecology 81: 8–14. Matsuda, H., M. Hori, and P. A. Abrams. 1994. Effects of Roxburgh, S. H. and J. B. Wilson. 2000. Stability and predator-specific defense on predator persistence and coexistence in a lawn community: mathematical pre- community complexity. Evolutionary Ecology 8: 628–639. diction of stability using a Community matrix with Matsuda, H., M. Hori, and P. A. Abrams. 1996. Effects of parameters derived from competition experiments. predator-specific defense on biodiversity and commun- Oikos 88: 395–408. ity complexity in two-trophic-level communities. Evo- Schmitz, O. J., and G. Booth. 1997. Modelling food web lutionary Ecology 10: 13–28. complexity: The consequences of individual-based, May, R. M. 1972. Will a large complex system be stable? spatially explicit behavioural ecology on trophic Nature 238: 413–414. interactions. Evolutionary Ecology 11: 379–398. 240 REFERENCES

Schoener, T. W. 1993. On the relative importance of direct Anderson, P. J., J. F. Piatt. 1999. Community reorgan- versus indirect effects in ecological communities. In: ization in the Gulf of Alaska following ocean climate H. Kawanabe, J. E. Cohen, and K. Iwasaki (eds), regime shift. Marine Ecology Progress Series 189: Mutualism and Community Organization: Behavioral, 117–123. Theoretical and Food Web Approaches. Oxford University Angliss, R. P., and K. L. Lodge. 2002. Alaska marine Press, Oxford, pp. 366–411. mammals stock assessment, 2002. U.S. Department Sih, A. 1984. Optimal behavior and density-dependent Commerce, NOAA Technical Memorandum NMFS- predation. American Naturalist 123: 314–326. AFSC-133, 224 p. Sole, R. V., and J. M. Montoya. 2001. Complexity and Anker-Nilssen, T., R. T. Barret, and J. V. Krasnov. 1997. fragility in ecological networks. Proceedings of The Royal Long- and short-term responses of seabirds in the Society of London, Series B 268: 2039–2045. Norwegian and Barents Seas to changes in stocks of Sutherl, W. 1996. From Individual Behaviour to Population prey fish. In: Forage Fishes in Marine Ecosystems. Alaska Ecology. Oxford University Press, Oxford. Sea Grant College Program Publication AK-SG-97–01. Stephens, D. W., and J. R. Krebs. 1987. Foraging Theory. University of Alaska, Fairbanks, pp. 683–698. Princeton University Press, Princeton, NJ. Anthony, J. A., D. D. Roby, and K. R. Turco. 2000. Sterner, R. W., A. Bajpai, and T. Adams. 1997. The enigma content and energy density of forage fishes from the of food chain length: absence of theoretical evidence northern Gulf of Alaska. Journal of Experimental Marine for dynamic constraints. Ecology 78: 2258–2262. Biology and Ecology 248: 53–78 Tansky, M. 1978. Switching effect in prey–predator Bailey, K. M. 2000. Shifting control of recruitment of system. Journal of Theoretical Biology 70: 263–271. walleye pollock Theragra chalcogramma after a major Teramoto, E., K. Kawasaki, and N. Shigesada. 1979. climatic and ecosystem change. Marine Ecology Progress Switching effect of predation on competitive prey Series 198: 215–224. species. Journal of Theoretical Biology 79: 305–315. Bailey, K. M., and S. A. Macklin. 1994. Analysis of pat- Vos, M., S. M. Berrocal, F. Karamaouna, L. Hemerik, and terns in larval Walleye pollock theragra-chalcogramma L. E. M. Vet. 2001. Plant-mediated indirect effects and survival and wind mixing events in Shelikof the persistence of parasitoid–herbivore communities. Strait, Gulf of Alaska. Marine Ecology Progress Series Ecology Letters 4: 38–45. 113: 1–12 Warren, P. H. 1989. Spatial and temporal variation in Bailey, K. M., and S. J. Picquelle. 2002. Larval distribution tropical fish trophic networks. Oikos 55: 299–311. of offshore spawning flatfish in the Gulf of Alaska: Wilmers, C. C., S. Sinha, and M. Brede. 2002. Examining potential transport pathways and enhanced onshore the effects of species richness on community stability: transport during ENSO events. Marine Ecology Progress an assembly model approach. Oikos 99: 364–368. Series 236: 205–217 Wilson, D. S., and J. Yoshimura. 1994. On the coexistence Ballance, L. T., and R. L. Pitman. 1999. Foraging ecology of specialists and generalists. American Naturalist 144: of tropical seabirds. In: N. J. Adams, and R. H. Slotow 692–707. (eds), Proceedings of the 22nd International Ornithology Winemiller, K. O. 1990. Spatial and temporal variation in Congress, Durban, 2057–2071. BirdLife South Africa, tropical fish trophic networks. Ecological Monographs Johannesburg. 60: 331–367. Behrenfeld, M. J., and Z. S. Kolber, 1999. Widespread iron Yodzis, P. 1981. The stability of real ecosystems. Nature limitation of phytoplankton in the South Pacific Ocean. 289: 674–676. Science 283: 840–843. Yodzis, P. 2000. Diffuse effects in food webs. Ecology Benson, A. J., and A. W. Trites. 2002. Ecological effects of 81: 261–266. regime shifts in the Bering Sea and North Pacific Ocean. Fish and Fisheries 3: 95–113. Berenboim, B. I., A. V., Dolgov, V. A., Korzhev, and Chapter 12 N. A. Yaragina. 2000. The impact of cod on the Alaska Sea Grant. 1993. Is it food? Addressing Marine dynamics of Barents Sea shrimp (Pandalus borealis) Mammals and Seabirds Declines. Alaska Sea Grant as determined by multispecies models. Journal of Report No. 93–01. University of Alaska Fairbanks, Northwest Atlantic Fisheries Science 27: 69–75 Fairbanks. Bidigare, R. R., and M. E. Ondrusek. 1996. Spatial and Anderson, P. J. 2000. Pandalid shrimp as indicators of temporal variability of phytoplankton pigment distri- ecosystem regime shift. Journal of Northwest Atlantic bution in the central equatorial Pacific Ocean. Deep-Sea Fisheries Science 27: 1–10. Research II 43: 809–833. REFERENCES 241

Blindheim, J., V. Borovkov, B. Hansen, S-A. Malmberg, Cohen, J. E., F. Briand, and C. M. Newman. 1990. Com- W. R. Turrell, and S. Østerhus. 2000. Upper layer cooling munity Food Webs. Springer-Verlag, 308 pp. and freshening in the Norwegian Sea in relation to Compagno, L. 1984. Sharks of the world. FAO Species atmospheric forcing. Deep-Sea Research I 47: 655–680. Catalogue 4: 655. Bogstad B., T. Haug, and S. Mehl. 2000. Who eats whom Cury P., A. Bakun, R. J. M. Crawford, A. Jarre, in the Barents Sea? NAMMCO Scientific Publication 2: R. A. Quinones, L. J. Shannon, and H. M. Verheye. 98–119. 2000. Small pelagics in upwelling systems: patterns of Bond, N. A., J. E. Overland, M. Spillane, and P. Stabeno. interaction and structural changes in ‘‘wasp-waist’’ 2004. Recent shifts in the state of the North Pacific. ecosystems. ICES Journal of Marine Sciences 57: 603–618. Geophysical Reasearch Letters 30: 2183–2186. Cushing, D. H. 1982. Climate and Fisheries. Academic Boyd, P. W. et al. 2000. A mesoscale phytoplankton Press, London. bloom in the polar Southern Ocean stimulated by iron Cushing, D. H. 1990. Plankton production and year-class fertilization. Nature 407: 695–702. strength in fish populations: An update of the match/ Brodeur, R. D. 1998. In situ observations of the associations mismatch hypothesis. Advances in Marine Biology 26: between juvenile fishes and schyphomedusae in the 249–294. Bering Sea. Marine Ecology Progress Series 163:11–20. Dalpadado, P., and H. R. Skjoldal. 1996. Abundance, Brodeur, R. D., and D. M. Ware. 1992. Long-term varia- maturity and growth of krill species Thysanoessa inermis bility in zooplankton biomass in the sub-arctic Pacific and T. longicaudata in the Barents Sea. Marine Ecology Ocean. Fisheries Oceanography 1: 32–38. Progress Series 144: 175–183. Brodeur, R. D., B. W. , S. R. Hare, R. C. Francis, and Dalpadado, P., N. Borkner, B. Bogstad, and S. Mehl. 2001. W. J. Jr Ingraham. 1999. Interannual variations in Distribution of Themisto spp () in the zooplankton biomass in the Gulf of Alaska, and Barents Sea and predator–prey interactions. ICES covariation with California Current zooplankton bio- Journal of Marine Sciences 58: 876–895. mass. In: K. Sherman and Q. Tang (eds), Large Marine Dalpadado, P., B. Bogstad, H. Gjøsæter, S. Mehl, and Ecosystems of the Pacific Rim: Assessment, Sustainability, H. R. Skjoldal. 2002. Zooplankton–fish interactions in and Management. Blackwell Science Inc., Malden, MA, the Barents Sea. In: K. Sherman and H. R. Skjoldal pp. 106–137. (eds), Large Marine Ecosystems of The North Atlantic. Burek, K. A., F. M. D. Gulland, G. Sheffield, D. Calkins, Elsevier, Amsterdam, pp. 449 þ xiv. E. Keyes, T. R. Spraker, A. W. Smith, D. E. Skilling, Dickson, R. R., T. J. Osborn, J. W. Hurrell, J. Meincke, J. Evermann, J. L. Stott, and A. W. Trites. 2003. Disease J. Blindheim, B. A˚ dlandsvik, T. Vinje, G. Alekseev, and agents in Steller sea lions in Alaska: a review and W. Maslowski. 2000. The Arctic Ocean response to analysis of serology data from 1975–2000. Fisheries the North Atlantic Oscillation. Journal of Climate 13: Centre Reports 11(4): 26. 2671–2696. Carpenter, S. R. 1990. Large-scale perturbations: oppor- Dolgov, A. V. 2002. The role of capelin (Mallotus villosus) tunities for innovation. Ecology 71: 2038–2043. in the food web of the Barents Sea. ICES Journal of Carpenter, S. R., and J. F. Kitchell. 1993. The trophic cas- Marine Sciences 59: 1034–1045 cade in lakes. Cambridge University Press, Cambridge. Dragesund, O., J. Hamre, and O. Ulltang. 1980. Biology and Chavez, F. P., J. Ryan, S. E. Lluch-Cota, and M. Niquen. population dynamics of the Norwegian spring-spawning 2003. From anchovies to sardines and back: Multidecadal herring. Rapports et Proce`s-verbaux des Re´unions Conseil change in the Pacific Ocean. Science 299: 217–221. international pour l’Exploration de la Mer 177: 43–71. Choi, S. J., K. T. Frank, W. C. Legget, and K. Drinkwater. Dragoo, D. E., G. V. Byrd, and D. B. Irons. 2000. Breeding 2004. Transition to an alternate state in a continental status and population trends of seabirds in Alaska in shelf ecosystem. Canadian Journal of fisheries and Aquatic 1999. U.S. Fish and Wildlife Service Report AMNWR Sciences 61: 505–510. 2000/02. Ciannelli, L., K. S. Chan, K. M. Bailey, and N. C. Stenseth. Ellertsen, B., P. Fossum, P. Solemdal, and S. Sundby. 2004. Non-additive effects of the environment on the 1989. Relation between temperature and survival of survival of a large marine fish population. Ecology eggs and first-feeding larvae of northeast Arctic cod 85(12). (Gadus morhua L.). Rapports et Proce´s-Verbaux des Coale, K. H. et al. 1996. A massive phytoplankton bloom Re´unions Conseil international pour l’Exploration de la Mer induced by an ecosystem-scale 191: 209–219. experiment in the Equatorial Pacific Ocean. Nature 383: Elton, C. 1958. The Ecology of Invasions by Animals and 495–501. Plants. Methuen and Co., London. 242 REFERENCES

Estes, J. A., M. T. Tinker, T. M. Williams, and D. F. Doak. Hassel, A., H. R. Skjoldal, H. Gjøsæter, H. Loeng, 1998. Killer whale predation on sea otters linking and L. Omli. 1991. Impact of grazing from capelin oceanic and nearshore ecosystems. Science 282:473–476. (Mallotus villosus) on zooplankton—a case study in the Food and Agricultural Organization (FAO). 2002. The Northern Barents Sea in August 1985. Polar Research state of world fisheries and aquaculture. Available at: 10: 371–388. www.fao.org. Hasting, A. 1988. Food web theory and stability. Ecology Fossum, P. 1996. A study of first-feeding herring (Clupea 69(6): 1665–1668. harengus L) larvae during the period 1985–1993. ICES Haug, T., and K. T. Nilssen. 1995. Ecological mplications Journal of Marine Science 53: 51–59. of harp seal Phoca groenlandica invasions in northern Francis, R. C., S. R. Hare, A. B. Hollowed, and Norway. In: A. S. Blix, L. Walløe, and Ø. Ulltang (eds), W. S. Wooster. 1998. Effects of interdecadal climate Whales, Seals, Fish and Man. Developments in Marine variability on the oceanic ecosystems of the NE Pacific. Biology 4. Elsevier, Amsterdam, pp. 545–556. Fisheries Oceanography 7: 1–21. Hjermann, D. Ø., N. C. Stenseth, and G. Ottersen. 2004. Furevik, T. 2001. Annual and interannual variability of Competition among fisherman and fish causes the Atlantic Water temperatures in the Norwegian and collapse of the Barents Sea capelin. Proceedings of the Barents Seas: 1980–1996. Deep Sea Research I 48: 383–404. National Academy of Sciences USA 101(32): 11679–11684. Gillett, N. P., H. F. Graf, and T. J. Osborn. 2003. Climate Hoerling, M. P., J. W. Hurrell, and T. Y. Xu. 2001. Tropical changeandtheNorthAtlantic Oscillation.In:J.W.Hurrell, origins for recent North Atlantic climate change. Science Y. Kushnir, G. Ottersen, and M. Visbeck (eds), The North 292:90–92. Atlantic Oscillation: Climatic Significance and Environmental Hollowed, A. B., S. R. Hare, and W. S. Wooster. 2001. Impact. American Geophysical Union, Washington DC, Pacific Basin climate variability and patterns of pp. 193–209. Northeast Pacific marine fish production. Progress in Gjøsæter, H., and B. Bogstad. 1998. Effects of the presence Oceanography 49: 257–282. of herring (Clupea harengus) on the stock-recruitment Hunt, G. L., P. Stabeno, G. Walters, E. Sinclair, relationship of Barents Sea capelin (Mallotus villosus). R. D. Brodeur., Napp, J. M., and N. A. Bond, 2002. Cli- Fisheries Research 38: 57–71. mate change and control of the southeastern Bering Sea Gjøsæter, H., P. Dalpadado, and A. Hassell. 2002. Growth pelagic ecosystem. Deep-Sea Research II 49: 5821–5853. of Barents Sea capelin (Mallotus villosus) in relation to Hurrell, J. W. 1995. Decadal trends in the North Atlantic zooplankton abundance. ICES Journal of Marine Sciences Oscillation: regional temperatures and precipitation. 59: 959–967. Science 169: 676–679. Grandperrin R., 1975. Structures trophiques aboutissant Hurrell, J. W., Y. Kushnir, G. Ottersen, and M. Visbeck. aux thons de longues lignes dans le Pacifique sud- 2003. An overview of the North Atlantic Oscillation. In: ouest tropical. The`se de Doctorat, Universite´ de Mar- J. W. Hurrell, Y. Kushnir, G. Ottersen, and M. Visbeck seille. 296 pp. (eds), The North Atlantic Oscillation: Climatic Significance Gueredrat, J. A. 1971. Evolution d’une population de and Environmental Impact. Geophysical Monograph copepodes dans le systeme des courants equatoriaux Series. American Geophysical Union, pp. 1–35. de l’ocean Pacifique. Zoogeographie, ecologie et Washington DC, diversite specifique. Marine Biology 9: 300–314. Huse, G., and R. Toresen. 1996. A comparative study Hampton, J., and D. A. Fournier. 2001. A spatially dis- of the feeding habits of herring (Clupea harengus, aggregated, length-based, age-structured population Clupeidae, L.) and capelin (Mallotus villosus, model of yellowfin tuna (Thunnus albacares) in the Osmeridae, Mu¨ller) in the Barents Sea. Sarsia 81: western and central Pacific Ocean. Marine and Fresh- 143–153. water Research 52: 937–963. Huse, G., and R. Toresen. 2000. Juvenile herring prey on Hamre, J. 1994. Biodiversity and exploitation of the main Barents Sea capelin larvae. Sarsia 85: 385–391. fish stocks in the Norwegian–Barents Sea ecosystem. Ingvaldsen, R., H. Loeng, G. Ottersen, and B. A˚ dlandsvik. Biodiversity and Conservation 3: 473–492 2003. Climate variability in the Barents Sea during the Hamre, J. 2000. Effects of Climate and Stocks Interactions on 20th century with focus on the 1990s. ICES Marine the Yield of North-east Arctic Cod. Results from Multi- Sciences Symposium 219: 160–168. species Model Run. Report No. ICES CM 2000/V:04. Jackson, J. B. C., M. X. Kirby, W. H. Berger, K. A. Bjorndal, Hamre, J. 2003. Capelin and herring as key species for L. W. Botsford, B. J. Bourque, R. H. Bradbury, R. Cooke, the yield of north-east Arctic cod. Results from multi- J. Erlandson, J. A. Estes, T. P. Hughes, S. Kidwell, species model runs. Scientia Marina 67: 315–323. C. B. Lange, H. S. Lenihan, J. M. Pandolfi, C. H. Peterson, REFERENCES 243

R. S. Steneck, M. J. Tegner, and R. R. Warner. 2001. Loeng, H. 1989. The influence of temperature on some Historical overfishing and the recent collapse of coastal fish population parameters in the Barents Sea. Journal ecosystems. Science 293: 629–638. of Northwest Atlantic Fisheries Science 9: 103–113. Jorgensen, T. 1990. Long-term changes in age at sexual Loeng, H. 1991. Features of the physical oceanographic maturity of Northeast Arctic cod (Gadus morhua L). conditions of the Barents Sea. Polar Research 10: 5–18. Journal Du Conseil 46: 235–248. Loeng, H., J. Blindheim B., A˚ dlandsvik and G. Ottersen. Kendall, A. W., J. D. Schumacher, and S. Kim. 1996. 1992. Climatic variability in the Norwegian and Barents Walleye pollock recruitment in Shelikof Strait: applied Seas. ICES Marine Sciience Symposium 195: 52–61. fisheries oceanography. Fisheries Oceanography 5: 4–18. MacArthur, R. 1955. Fluctuations of animal populations and Krasnov, Y. V., and R. T. Barrett. 1995. Large-scale a measure of community stability. Ecology 36: 533–536. interactions between seabirds, their prey and man Mackas, D. L., R. E. Thomson, and M. Galbraith. 2001. in the southern Barents Sea. In: H. R. Skjoldal, Changes in the zooplankton community of the British C. C. E. Hopkins, K. E. Erikstad, and H. P. Leinaas Columbia continental margin, 1985–1999, and their (eds), Ecology of Fjords and Coastal Waters. Elsevier covariation with oceanographic conditions. Canadian Science, Amsterdam, 623, pp. 443–456. Journal of Fisheries and Aquatic Sciences 58: 685–702. Krause, A. E., K. A. Frank, D. M. Mason, R. E. Ulanowicz, Mann, K. H. 1993. Physical Oceanography, food-chains, and W. W. Taylor. 2003. Compartments revealed in and fish stocks—a review. ICES Journal of Marine food-web structure. Nature 426: 282–285. Science 50: 105–119. Landry, M. R., and D. L. Kirchman. 2002. Microbial Mantua, N. J., and S. R. Hare. 2002. The Pacific decadal community structure and variability in the tropical oscillation. Journal of Oceanography 58: 35–44. Pacific. Deep-Sea Research II 49: 2669–2693. Mantua, N. J., S. R. Hare, Y. Zhang, J. M. Wallace, and Landry, M. R., J. Constantinou, and J. Kirshtein. 1995. R. C. Francis. 1997. A Pacific interdecadal climate Microzooplankton grazing in the central equatorial oscillation with impacts on salmon production. Bulletin Pacific during February and August, 1992. Deep-Sea of the American Meteorological Society 78: 1069–1079. Research 42: 657–671. Mauchline, J. 1998. The biology of calanoid copepods. Law, R. 2000. Fishing, selection, and phenotypic evo- Advances in Marine Biology. Academic Press, London:. lution. ICES Journal of Marine Sciences 57: 659–668. May, R. M. 1973. Stability and Complexity in Model Le Borgne, R., and M. Rodier. 1997. Net zooplankton and Ecosystems. Princeton University Press, Princeton, NJ. the : a comparison between the oligo- Maynard Smith, J. 1974. Models in Ecology. Cambridge trophic and mesotrophic equatorial Pacific. Deep-Sea University Press. Research 44: 2003–2024. McFarlane, G. A., J. R. King, and R. J. Beamish. 2000. Legand, M., P. Rourret, P. Fourmanoir, R. Grandperrin, Have there been recent changes in climate? Ask the J. A. Gueredrat, A. Mitchel, P. Rancurel, R. Repelin, fish. Progress in Oceanography 47: 147–169. and C. Roger. 1972. Relations trophiques et distribu- McGowan, J. A., D. R. Cayan, and L. M. Dorman. 1998. tions verticales en milieu pe´lagique dans l’ocean Climate-Ocean variability and ecosystem response in Pacifique intertropical, centre ORSTOM de Noumea, the Northeast Pacific. Science 281: 210–217. New Caledonia. McPhaden, M. J., and J. Picaut. 1990. El Nin˜o-Southern Lehodey, P. 2001. The pelagic ecosystem of the tropical Oscillation index displacements of the Western Pacific Ocean: dynamic spatial modelling and bio- Equatorial Pacific Warm Pool. Science 50: 1385–1388. logical consequences of ENSO. Progress in Oceanography Mehl, S. 1991. The Northeast Arctic cod stock’s place in 49: 439–468. the Barents Sea ecosystem in the 1980s: an overview. Lehodey, P., M. Bertignac, J. Hampton, A. Lewis, and Polar Research 10: 525–534. J. Picaut. 1997. El Nino Southern Oscillation and tuna Mehlum, F., and Gabrielsen, G. W. 1995. Energy expen- in the western Pacific. Nature 389: 715–718. diture and food consumption by seabird populations Lehodey, P., F. Chai, and J. Hampton. 2003. Modelling in the Barents Sea region. In: H. R. Skjoldal, C. Hopkins, climate-related variability of tuna populations from a K. E. Erikstad, and H. P. Leinaas (eds), Ecology of Fjords coupled ocean–biogeochemical–populations dynamics and Coastal Waters. Elsevier Science, Amsterdam, pp. model. Fisheries Oceanography 12: 483–494. 457–470. Lima, M., N. C. Stenseth, and F. M. Jaksic. 2002. Food Melle, W., and H. R. Skjoldal. 1998. Reproduction, life web structure and climate effects on the dynamics of cycles and distributions of Calanus finmarchicus and small mammals and in semi-arid Chile. Ecology C. hyperboreus in relation to environmental conditions Letters 5: 273–284. in the Barents Sea. Dr. Scient Department of Fisheries 244 REFERENCES

and Marine Biology, University of Bergen, Bergen, Parsons, T. R. 1986. Ecological relations. In: The Gulf of Norway. Alaska: Physical Environment and Biological Resources. Merati, N., and R. D. Brodeur. 1996. Feeding Habits and D. W. Hood and S. T. Zimmerman (eds), pp. 561–570. Daily Ration of Juvenile Walleye Pollock, Theragra Management Service, Springfield, VA. chalcogramma, in the Western Gulf of Alaska. NOAA Parsons, T. R. 1996. The impact of industrial fisheries Technical Report NMFS 126: 65–79. on the trophic structure of marine ecosystems. In: Mueter, F. J., and B. L. Norcross. 2000. Changes in species G. A. Polis, and K. O. Winemiller (eds), Food Webs. composition of the demersal fish community in near- Chapman and Hall, New York, pp. 352–357. shore waters of Kodiak Island, Alaska. Canadian Journal Pauly D., V. Christensen, J. Dalsgaard, R. Froese, and of Fisheries and Aquatic Sciences 57: 1169–1180. F. Torres. 1998. Fishing down marine food webs. Murtugudde, R. G., S. R. Signorini, J. R. Christian, Science 279: 860–863. A. J. Busalacchi, C. R. McClain, and J. Picaut 1999. Philander, S. G. H. 1990. El Nin˜o, La Nin˜a, and the Southern variability of the tropical Indo-Pacific Oscillation., Vol., Academic Press, New York. Basin observed by SeaWiFS during 1977–1998. Journal Piatt, J. P., and P. Anderson. 1996. Response of common of Geophysical Research 104: 18351–18366. murres to the Exxon Valdez and long-term Nesterova, V. N. 1990. Plankton Biomass Along the Drift changes in the Gulf of Alaska marine ecosystem. In: Route of Cod Larve (in Russian). PINRO Press, S. D. Rice, R. B. Spies, D. A. Wolfe, and B. A. Wright Murmansk, 64 pp. (eds), Proceedings of Exxon Valdez Oil Spill Symposium. National Research Council. 2003. Decline of Steller Sea American Fisheries Society Symposium 18, Bethesda, lion in Alaskan Waters. National Academy Press, MD, pp. 720–737. Washington DC. Pimm, S. L. 1982. Food Webs. Chapman and Hall, Nilssen, K. T., O. P. Pedersen, L. P. Folkow, and T. Haug. 219 pp. 2000. Food consumption estimates of Barents Sea harp Polovina, J. J., G. T. Mitchum, and G. T. Evans. 1995. Decadal seals. NAMMCO Scientific Publications 2: 9–28. and basin-scale variation in mixed layer depth and the Odum, E. P. 1963. Ecology. Holt, Rinehart and Winston, impact on biological production in the central and North New York. Pacific, 1960–1988. Deep Sea Research 42: 1201–1716. Odum, E. P. 1985. Trends expected in stressed ecosystems. Pozo-Vazquez, D., M. J. Esteban-Parra, F. S. Rodrigo, and BioScience 35: 419–422. Y. Castro-Diez. 2000. An analysis of the variability of Olsen, A., L. G. Anderson, and T. Johannessen. 2002. The the North Atlantic Oscillation in the time and the impact of climate variations on fluxes of oxygen in the frequency domains. International Journal of Climatology Barents Sea. Continental Shelf Research 22: 1117–1128. 20: 1675–1692. Orensanz, J. M., J. Armstrong, D. Armstrong, and R. Raffaelli, D. 2000. Trends in research on shallow water Hilborn. 1998. Crustaceans resources are vulnerable to fiid webs. Journal of Experimental Biology and Ecology serial depletion—the multifaceted decline of crab and 250: 223–232. shrimps fisheries in the Greater Gulf of Alaska. Reviews Reed, R. K., and J. D. Schumacher. 1987. Physical oceano- in Fish Biology and Fisheries 8: 117–176. graphy. In: D. W. Z. S. Hood, (ed.), The Gulf of Alaska: Ottersen, G., and H. Loeng. 2000. Covariability in early Physical Environment and Biological Resources.US growth and year-class strength of Barents Sea cod, Department of Commerce, NOAA, p. 655. haddock, and herring: the environmental link. ICES Rey F., H. R. Skjoldal, and D. Slagstad. 1987. Primary pro- Journal of Marine Sciences 57: 339–348. duction in relation to seasonal changes in the Barents Ottersen, G., and N. C. Stenseth. 2001. Atlantic climate Sea. In: H. Loeng (ed.), The Effect of Oceanographic governs oceanographic and ecological variability in the Conditions on Distribution and Population Dynamics of Barents Sea. Limnology and Oceanography 46: 1774–1780. Commercial in the Barents Sea. Institute of Paine, R. T. 1980. Food webs—linkage, interaction Marine Research, Bergen, Norway, pp. 29–46. strength and community infrastructure—the 3rd Rogers, J. C. 1984. The assosiation between the North Tansley Lecture. Journal of Animal Ecology 49: 667–685. Atlantic Oscillation and the Southern Oscillation in the Panasenko, L. D. 1984. Feeding of the Barents Sea capelin. northern hemisphere. Monsoon and Weather Reviews Journal of Marine Sucuess, ICES CM Documents 1984/H: 112: 1999–2015. 6, 16. Roman, M. R., H. G. , R. Le Borgne, and X. Zhang. Parsons, L. S., and W. H. Lear, 2001. Climate variability 2002. Latitudinal comparisons of equatorial Pacific and marine ecosystem impacts: a North Atlantic per- zooplankton. Deep-Sea Research Part II-Topical Studies in spective. Progress in Oceanography 49: 167–188. Oceanography 49: 2695–2711. REFERENCES 245

Stock Assessment and Fishery Evaluation (SAFE). 2003. Stenseth, N. C., G. Ottersen, J. W. Hurrell, A. Mysterud, Report for the Ground Fish Resources of the Gulf of M. Lima, K. S. Chan, N. G. Yoccoz, and B. A˚ dlandsvik Alaska. North Pacific Fishery Management Council, 2003. Studying climate effects on ecology through Anchorage, Alaska. Available at: http://afsc.noaa. the use of climate indices: the North Atlantic Oscilla- gov/refm/stocks/assessments.htm. tion, El Nin˜o Southern Oscillation and beyond. Pro- Sakshaug, E. 1997. Biomass and productivity distribu- ceedings of the Royal Society of London Series B 270: tions and their variability in the Barents Sea. ICES 2087–2096. Journal of Marine Sciences 54: 341–350. Strasburg, D. W. 1958. Distribution, abundance, and Sakshaug, E., A. Bjørge, B. Gulliksen, H. Loeng and habits of pelagic sharks in the central Pacific Ocean. F. Mehlum (eds), 1992. Økosystem Barentshavet. Fisheries Bulletin 58: 335–361. Lillehammer. Sturdevant, M. V. 1999. Diet Overlap, Sambrotto, R. N., and C. J. Lorenzen. 1987. Phytoplankton 1994–1996. EXXON VALDEZ Oil Spill Restoration and primary production. In: D. W. Z. S. Hood (ed.), The Project. Final Report (Restoration Project 97163C). Gulf of Alaska: Physical Environment and Biological Toresen, R., and O. J. Østvedt. 2000. Variation in the Resources. US Department of Commerce, NOAA, p. 655. abundance of Norwegian spring-spawning herring Schindler, D. W. 1985. Detecting ecosystem responses to (Clupea harengus, Clupeidae) throughout the 20th anthropogenic stress. Canadian Journal of Fisheries and century and the influence of climatic fluctuations. Fish Aquatic Sciences. 44(Suppl. 1): 6–25. and Fisheries 1: 231–251. Sekino, T., and N. Yamamura. 1999. Trenberth, K. E., and J. W. Hurrell, 1994. Decadal of zooplankton: optimum migrating schedule based on Atmosphere-Ocean variations in the Pacific. Climate energy accumulation. Evolutionary Ecology 13: 267–282. Dynamics 9: 303–319. Sinclair, E. H., and T. K. Zeppelin. 2003. Seasonal and Tricas, T. C. 1979. Relationships of the blue shark, spatial differences in diet in the western stock of Steller Prionace glauca, and its prey species near Santa Catalina sea lions (Eumetopias jubatus). Journal of Island, California. Fisheries Bulletin 77: 175–182. 83(4): 973–990. Trites, A. W., and L. P. Donnelly, 2003. The decline of Springer, A. M. 1993. Report of the seabird working Steller sea lions in Alaska: A review of the nutritional group. In: Is it food?: Addressing and stress hypothesis. Mammal Review 33: 3–28. Seabird Declines Workshop Summary. University of Ushakov, N. G., and D. V. Prozorkevich. 2002. The Barents Alaska, Alaska Sea Grant Report 93–01, Fairbanks, AK, Sea capelin—a review of trophic interrelations and pp. 14–29. fisheries. ICES Journal of Marine Sciences 59: 1046–1052. Springer, A. M., J. A. Estes, G. B. van Vliet, T. M. Williams, Verity, P. G., V. Smetacek, and T. J. Smayda. 2002. Status, D. F. Doak, E. M. Danner, K. A. Forney, and B. Pfister trends and the future of the marine pelagic ecosystem. 2003. Sequential megafaunal collapse in the North Environmental Conservation 29: 207–237. Pacific Ocean: An ongoing legacy of industrial whal- Vinogradov, M. E. 1981. Ecosystems of equatorial upwell- ing? Proceedings of the National Academy of Sciences, USA ings. In: A. R. Longhurst (ed.), Analysis of Marine 100: 12223–12228. Ecosystems. Academic Press, pp. 69–93. Stabeno, P., N. A. Bond, A. J. Hermann, C. W. Mordy, Watters, G. M., R. J. Olson, R. C. Francis, P. C. Fiedler, and J. E. Overland. 2004. Meteorology and oceano- J. J. Polovina, S. B. Reilly, K. Y. Aydin, C. H. Boggs, graphy of the northern Gulf of Alaska. Continetal Shelf T. E. Essington, C. J. Walters, and J. F. Kitchell. 2003. Research 24: 859–897. Physical forcing and the dynamics of the pelagic Steele, J. H. 1974. The Structure of Marine Ecosystems. ecosystem in the eastern tropical Pacific: simulations Harvard University Press, Cambridge, MA, 128 pp. with ENSO-scale and global-warming climate drivers. Steele, J. H., and M. Schumacher. 2000. Ecosystem Canadian Journal of Fisheries and Aquatic Sciences 60: structure before fishing. Fisheries Research 44: 201–205. 1161–1175. Stenseth, N. C., W. Falck, O. N. Bjornstad, and C. J. Krebs. Worm, B., and R. A. Myers. 2003. Meta-analysis of 1997. Population regulation in and cod-shrimp interactions reveals top-down control in Canadian lynx: Asymmetric food web configurations oceanic food webs. Ecology 84: 162–173. between hare and lynx. Proceedings of the National Yang, M. S., and M. W. Nelson. 2000. Food Habits of the Academy of Sciences, USA 94: 5147–5152. Commercially Important Groundfishes in the Gulf of Stenseth, N. C., A. Mysterrud, G. Ottersen, J. W. Hurrel, Alaska in 1990, 1993, and 1996. US Department of K. S. Chan, and M. Lima. 2002. Ecological effects of Commerce, NOAA Technical Memornadum, NMFS- climate fluctuations. Science 297: 1292–1296. AFSC-112, 174 pp. 246 REFERENCES

Yang, M. 2004. Diet changes of Pacific cod (Gadus macro- disturbance and recovery in a kelp forest community. cephalus) in Pavlof Bay associated with climate Ecological Monographs 62: 421–445. changes in the Gulf of Alaska between 1980 and 1995. Dulvy, N. K., R. P. Freckleton, and N. V. C. Polunin. 2004. Fishery Bulletin. Coral reef cascades and the indirect effects of predator Zenkevitch, L. 1963. Biology of the Seas of the USSR. removal by exploitation. Ecology Letters 7: 410–416. George Allen and Unwin, London. Estes, J. A., and D. O. Duggins. 1995. Sea otters and Zhang, Y., J. M. Wallace, and D. S. Battisti. 1997. kelp forests in Alaska—Generality and variation in a ENSO-like interdecadal variability. Journal of Climate community ecological paradigm. Ecological Monographs 10: 1004–1020. 65: 75–100. ˚ Adlandsvik, B., and H. Loeng. 1991. A study of the Estes, J. A., M. T. Tinker, T. M. Williams, and D. F. Doak. climatic system in the Barents Sea. Polar Research 10: 1998. Killer whale predation on sea otters linking 45–49. oceanic and nearshore ecosystems. Science 282: 473–476. Fagan, W. F. 1997. Omnivory as a stabilizing feature of Chapter 13 natural communities. American Naturalist 150: 554–567. Friedlander, A. M., and E. E. DeMartini. 2002. Contrasts Babcock, R. C., S. Kelly, N. T. Shears, J. W. Walker, and in density, size, and biomass of reef fishes between the T. J. Willis. 1999. Changes in community structure in northwestern and the main Hawaiian islands: the temperate marine reserves. Marine Ecology-Progress effects of fishing down apex predators. Marine Ecology Series 189: 125–134. Progress Series 230: 253–264. Ballesteros, E. 1991. Els vegetals i la zonacio´ litoral: Garcı´a-Rubies, A. 1996. Estudi ecolo`gic de les poblacions espe`cies, comunitats i factors que influeixen en la seva de peixos sobre substrat roco´s a la Mediterra`nia Occi- distribucio´. Arxius de la Seccio´ de Cie`ncies C. I., dental: efecte de la fonda`ria, el substrat, l’estacionalitat Institut d’Estudis Catalans, Barcelona. i la proteccio´. PhD Thesis, Universitat de Barcelona, Bascompte, J., C. Melian, and E. Sala. The structure of Spain. species interaction strength and the overfishing of Garrabou, J., E. Ballesteros, and M. Zabala. 2002. Structure marine food webs (submitted). and dynamics of north-western Mediterranean rocky Berlow, E. L. 1999. Strong effects of weak interactions in benthic communities along a depth gradient. Estuarine ecological communities. Nature 6725: 330–334. Coastal and Shelf Science 55: 493–508. Berlow, E. L., S. A. Navarrete, C. J. Briggs, M. E. Power, Goreau, T. 1959. The ecology of Jamaican coral reefs. and B. A. Menge. 1999. Quantifying variation in I. Species composition and zonation. Ecology 40: the strengths of species interactions. Ecology 80: 67–90. 2206–2224. Graham, M. H. 2000. Planktonic Patterns and Processes in the Boudouresque, C. F., and M. Verlaque. 2002. Biological Giant Kelp Macrocystis Pyrifera. Ph D Thesis, University pollution in the : invasive versus of California, San Diego introduced macrophytes. Marine Pollution Bulletin Graham, M. H. 2004. Effects of local deforestation on the 44: 32–38. diversity and structure of Southern California giant Carpenter, S. R., and J. F. Kitchell. 1974. The trophic cascade kelp forest food webs. Ecosystems 7: 341–357. in lakes. Cambridge University Press, Cambridge. Grigg, R. W. 1983. Community structure succession and Dayton, P. K. 1971. Competition, disturbance, and com- development of coral reefs in Hawaii, USA. Marine munity organization: the provision and subsequent Ecology Progress Series 11: 1–14. utilization of space in a rocky intertidal community. Grigg, R. W., and J. E. Maragos. 1974. Recolonization of Ecological Monographs 41: 351–389. hermatypic corals on submerged lava flows in Hawaii. Dayton, P. K., V. Currie, T. Gerrodette, B. D. Keller, Ecology 55: 387–395. R. Rosenthal, and D. Ven Tresca. 1984. Patch dynamics Grosholz, E. D., G. M. Ruiz, C. A. Dean, K. A. Shirley, and stability of some California kelp communities. J. L. Maron, and P. G. Connors. 2000. The impacts of a Ecological Monographs 54: 253–289. nonindigenous marine predator in a California bay. Dayton, P. K., M. J. Tegner, P. B. Edwards, and K. L. Riser. Ecology 81: 1206–1224. 1998. Sliding baselines, ghosts, and reduced expecta- Hairston, N. G., F. E. Smith, and L. B. Slobodkin. 1960. tions in kelp forest communities. Ecological Applications Community structure, population control, and com- 8: 309–322. petition. American Naturalist 94: 421–425. Dayton, P. K., M. J. Tegner, P. E. Parnell, and Hereu, B. 2004. The role of trophic interactions between P. B. Edwards. 1992. Temporal and spatial patterns of fishes, sea urchins and algae in the northwestern REFERENCES 247

Mediterranean rocky infralittoral. PhD thesis, Micheli, F., B. S. Halpern, L. W. Botsford, and Universitat de Barcelona. R. R. Warner. Trajectories and correlates of com- Hughes, T. P. 1994. Catastrophes, phase shifts, and large- munity change in no-take marine reserves. Ecological scale degradation of a caribbean coral reef. Science 265: Applications (in press). 1547–1551. Myers, R. A., and B. Worm. 2003. Rapid worldwide Jackson, J. B. C., and E. Sala. 2001. Unnatural oceans. depletion of predatory fish communities. Nature 423: Scientia Marina 65: 273–281. 280–283. Jackson J. B. C., M. X. Kirby, W. H. Berger, K. A. Bjorndal, Navarrete, S. A., and B. A. Menge. 1996. Keystone pre- L. W. Botsford, B. J. Bourque, R. H. Bradbury, R. Cooke, dation and interaction strength—interactive effects of J. Erlandson, J. A. Estes, T. P. Hughes, S. Kidwell, predators on their main prey. Ecological Monographs 66: C. B. Lange, H. S. Lenihan, J. M. Pandolfi, C. H. Peterson, 409–429. R. S. Steneck, M. J. Tegner, and R. R. Warner. 2001. Odum, E. 1969. The strategy of ecosystem development. Historical overfishing and the recent collapse of coastal Science 164: 262–270. ecosystems. Science 293: 629–638. Paine, R. T. 1966. Food web complexity and species Jones, G. P., M. I. McCormick, M. Srinivasan, and diversity. American Naturalist 100: 65–75. J. V. Eagle. 2004. Coral decline threatens fish bio- Paine, R. T. 1971. A short term experimental investigation diversity in marine reserves. Proceedings of the National of resource partitioning in a New Zealand rocky Academy of Sciences, USA 101: 8251–8253. intertidal habitat. Ecology 52: 1096–1106. Knowlton, N. 2001. The future of coral reefs. Procee- Paine, R. T. 1980. Food webs: Linkage, interaction dings of the National Academy of Sciences, USA 98: strength and community infrastructure. Journal of 5419–5425. Animal Ecology 49: 669–685. Knowlton, N., and F. Rohwer 2003. Multispecies Paine, R. T. 1992. Food-web analysis through field microbial mutualisms on coral reefs: The host as a measurement of per capita interaction strength. Nature habitat. American Naturalist 162: S51-S62. 355: 73–75. Kokkoris, G. D., A. Y. Troumbis, and J. H. Lawton. 1999. Palumbi, S. R. 2002. The evolution explosion: how humans Patterns of species interaction strength in assembled cause rapid evolutionary change. W.W. Norton & theoretical competition communities. Ecology Letters 2: Company. 70–74. Pandolfi, J. M., R. H. Bradbury, E. Sala, T. P. Hughes, Link, J. 2002. Does food web theory work for marine K. A. Bjorndal, R. G. Cooke, D. McArdle, ecosystems? Marine Ecology Progress Series 230: 1–9. L. McClenachan, M. J. H. Newman, G. Paredes, Margalef, R. 1991. Teoria de los sistemas ecolo´gicos. R. R. Warner, and J. B. C. Jackson 2003. Global trajec- Publicacions de la Universitat de Barcelona, Barcelona. tories of the long-term decline of coral reef ecosystems. Margalef, R. 1997. Our Biosphere. Ecology Institute, Science 301: 955–958. Oldendorf/Luhe. Pauly, D., V. Christensen, J. Dalsgaard, R. Froese, and May, R. M. 1973. Stability and Complexity in Model F. Torres. 1998. Fishing down marine food webs. Science Ecosystems. Princeton University Press, Princeton, NJ. 279: 860–863. May, R. M. 1974. Biological populations with non- Pinnegar, J. K., N. V. C. Polunin, P. Francour, overlapping generations: stable points, stable cycles, F. Badalamenti, R. Chemello M. L. Harmelin-Vivien, and chaos. Science 186: 645–647. B. Hereu, M. Milazzo, M. Zabala, G. D’Anna, and McCann, K., and A. Hastings. 1997. Re-evaluating the C. Pipitone. 2000. Trophic cascades in benthic marine omnivory–stability relationship in food webs. Proceed- ecosystems: lessons for fisheries and protected-area ings of the Royal Society of London Series B: Biological management. Environmental Conservation 27: 179–200. Sciences 264: 1249–1254. Rosenzweig, M. L. 1971. Paradox of enrichment: desta- McCann, K., A. Hastings, and G. R. Huxel. 1998. Weak bilization of exploitation ecosystems in ecological time. trophic interactions and the balance of nature. Nature Science 171: 385–387. 395: 794–798. Russ, G. R., and A. C. Alcala. 1996. Marine reserves— McClanahan, T. R. 1995. A coral reef ecosystem-fisheries rates and patterns of recovery and decline of large model—impacts of fishing intensity and catch selection predatory fish. Ecological Applications 6: 947–961. on reef structure and processes. Ecological Modelling Sala, E. 1997. The role of fishes in the organization of a 80: 1–19. Mediterranean sublittoral community. 2. Epifaunal Meinesz, A. 2002. Killer Algae. University of Chicago communities. Journal of Experimental Marine Biology and Press, Chicago, IL. Ecology 212: 45–60. 248 REFERENCES

Sala, E. 2004. The past and present topology and struc- In: H. R. Skjoldal, C. Hopkins, K. E. Eirkstad, and ture of Mediterranean subtidal food webs. H. P. Leinaas (eds), Ecology of Fjords and Coastal Waters, Ecosystems 7: 333–340. Elsevier Science, London, pp. 509–524. Sala, E., and C. F. Boudouresque. 1997. The role of fishes Verlaque, M. 1987. Relations entre Paracentrotus lividus in the organization of a Mediterranean sublittoral (Lamarck) et le phytobenthos de Mediterranee community. 1. Algal communities. Journal of Experi- occidentale. In: C. F. Boudouresque (ed.), Colloque mental Marine Biology and Ecology 212: 25–44. International sur Paracentrotus lividus et les oursins Sala, E., C. F. Boudouresque, and M. HarmelinVivien. comestibles. GIS Posidonie, Marseille, France, pp. 5–36. 1998. Fishing, trophic cascades, and the structure of Witman, J. D., and K. P. Sebens. 1992. Regional variation algal assemblages: evaluation of an old but untested in fish predation intensity: a historical perspective in paradigm. Oikos 82: 425–439. the Gulf of Maine. Oecologia 90: 305–315. Sala, E., and M. H. Graham. 2002. Community-wide Wootton, J. T. 1997. Estimates and tests of per capita inter- distribution of predator–prey interaction strength in action strength: diet, abundance, and impact of inter- kelp forests. Proceedings of the National Academy of tidally foraging birds. Ecological Monographs 67: 45–64. Sciences, USA 99: 3678–3683. Yodzis, P. 2001. Must top predators be culled for the sake Sala, E., and M. Zabala, 1996. Fish predation and the of fisheries? Trends in Ecology and Evolution 16: 78–84. structure of the sea urchin Paracentrotus lividus popu- Zabala, M., and E. Ballesteros. 1989. Surface-dependent lations in the NW Mediterranean. Marine Ecology strategies and energy flux in benthic marine commun- Progress Series 140: 71–81. ities or, why corals do not exist in the Mediterranean. Sala, E., O. Aburto-Oropeza, M. Reza, G. Paredes, and Scientia Marina 53: 3–17. L. G. Lopez-Lemus. 2004. Fishing down coastal food webs in the Gulf of California. Fisheries 29: 19–25. Shears, N. T., and R. C. Babcock. 2003. Continuing trophic Chapter 14 cascade effects after 25 years of no-take marine reserve protection. Marine Ecology Progress Series 246: 1–16. Abrams, P. A. 1993. Effect of increased productivity on Shiganova, T. A., and Y. V. Bulgakova. 2000. Effects of the abundance of trophic levels. American Naturalist gelatinous plankton on Black Sea and Sea of Azov fish 141: 351–371. and their food resources. ICES Journal of Marine Science Abrams, P. A. 1995. Monotonic or unimodal diversity– 57: 641–648. productivity gradients: what does competition theory Springer, A. M., J. A. Estes, G. B. van Vliet, T. M. Williams, predict? Ecology 76: 2019–2027. D. F. Doak, E. M. Danner, K. A. Forney, and B. Pfister. Abrams, P. A. 2001. The effect of density-independent 2003. Sequential megafaunal collapse in the North mortality on the coexistence of exploitative competitors Pacific Ocean: an ongoing legacy of industrial whal- for renewing resources. American Naturalist 158: 459–470. ing? Proceedings of the National Academy of Sciences, USA Adrian, R., S. A. Wickham, and N. M. Butler. 2001. 100: 12223–12228. Trophic interactions between zooplankton and the Steneck, R. S. 1998. Human influences on coastal microbial community in contrasting food webs: the ecosystems: does overfishing create trophic cascades? and of a Trends in Ecology and Evolution 13: 429–430. mesotrophic lake. Aquatic Microbial Ecology 24: 83–97. Steneck, R. S., and J. Carlton. 2001. Human alterations of Amarasekare, P. 2003. Diversity–stability relationships in marine communities: Students beware! In: M. Bertness, multitrophic systems: an empirical exploration. Journal S. Gaines, and M. E. Hay (eds), Marine Community of Animal Ecology 72: 713–724. Ecology. Sinauer, Sunderland, MA, p 445–468. Amarasekare, P., and R. M. Nisbet. 2001. Spatial hete- Sugihara, G. 1982. Niche hierarchy: Structure assembly rogeneity, source–sink dynamics, and the local coex- and organization in natural communities. PhD thesis, istence of competing species. American Naturalist 158: Princeton University, Princeton NS. 572–584. Sugihara G., L. F. Bersier, T. R. E. Southwood, S. L. Pimm, Anderson, T. R., and D. W. Pond. 2000. Stoichiometric and R. M. May. 2003. Predicted correspondence theory extended to micronutrients: Comparison of the between species abundances and dendrograms of roles of essential fatty acids, carbon and nitrogen in the niche similarities. Proceedings of the National Academy of nutrition of marine copepods. Limnology and Oceano- Sciences, USA 100: 5246–5251. graphy 45: 1162–1167. Vadas, R. L., and R. S. Steneck. 1995. Overfishing Armstrong, R. A., and R. McGehee. 1980. Competitive and inferences in kelp-sea urchin interactions. exclusion. American Naturalist 115: 151–170. REFERENCES 249

Attayde, J. L., and L. A. Hansson. 1999. Effects of nutrient Chesson, P., and N. Huntly. 1997. The roles of harsh and recycling by zooplankton and fish on phytoplankton fluctuating conditions in the dynamics of ecological communities. Oecologia 121: 47–54. communities. American Naturalist 150: 519–553. Bell, T. 2002. The ecological consequences of unpalatable Cohen, J. E. 1978. Food Webs and Niche Space. Princeton prey: phytoplankton response to nutrient and predator University Press, Princeton, NJ. additions. Oikos 99: 59–68. Connell, J. H. 1978. Diversity in tropical rain forests and Berninger, U. G., S. A. Wickham, and B. J. Finlay. 1993. coral reefs. Science 199: 1302–1310. Trophic coupling within the microbial food web: Cornell, H. V., and J. H. Lawton. 1992. Species inter- a study with fine temporal resolution in a eutrophic actions, local and regional processes, and limits to the . Freshwater Biology 30: 419–432. richness of ecological communities: a theoretical per- Bohonak, A. J., and D. G. Jenkins. 2003. Ecological and spective. Journal of Animal Ecology 61: 1–12. evolutionary significance of dispersal by freshwater DeMott, W. R. 1998. Utilization of a cyanobacterium and invertebrates. Ecology Letters 6: 783–796. a phosphorus-deficient green alga as complementary Borer, E. T., K. Anderson, C. A. Blanchette, B. Broitman, resources by daphnids. Ecology 79: 2463–2481. S. D. Cooper, and B. S. Halpern. 2002. Topological Diehl, S. 1992. Fish predation and benthic community approaches to food web analyses: a few modifications structure: the role of omnivory and habitat complexity. may improve our insights. Oikos 99: 397–401. Ecology 73: 1646–1661. Borer, E. T., E. W. Seabloom, J. B. Shurin, K. E. Anderson, Diehl, S. 1995. Direct and indirect effects of omnivory in a C. A. Blanchette, B. Broitman, S. D. Cooper, B. S. Halpern. littoral lake community. Ecology 76: 1727–1740. In press. What determines the strength of a trophic Diehl, S., and M. Feissel. 2001. Intraguild prey suffer from cascade? Ecology. enrichment of their resources: A microcosm experi- Borrvall, C., B. Ebenman, and T. Jonsson. 2000. Bio- ment with ciliates. Ecology 82: 2977–2983. diversity lessens the risk of cascading extinction in Dodson, S. I., S. E. Arnott, and K. L. Cottingham. 2000. The model food webs. Ecology Letters 3: 131–136. relationship in lake communities between primary Bro¨nmark, C., S. D. Rundle, and A. Erlandsson. 1991. productivity and species richness. Ecology 81: 2662–2679. Interactions between fresh-water snails and tadpoles— Downing, A. L., and M. A. Leibold. 2002. Ecosystem competition and facilitation. Oecologia 87: 8–18. consequences of species richness and composition in Brose, U., A. Ostling, K. Harrison, and N. D. Martinez. pond food webs. Nature 416: 837–841. 2004. Unified spatial scaling of species and their Duffy, J. E. 2002. Biodiversity and ecosystem function: the trophic interactions. Nature 428: 167–171. consumer connection. Oikos 99: 201–219. Cardinale, B. J., M. A. Palmer, and S. L. Collins. 2002. Duffy, J. E., K. S. Macdonald, J. M. Rhode, and J. D. Parker. Species diversity enhances ecosystem function- 2001. Grazer diversity, functional redundancy, and ing through interspecific facilitation. Nature 415: productivity in seagrass beds: an experimental test. 426–429. Ecology 82: 2417–2434. Carpenter, S. R., J. F. Kitchell, J. R. Hodgson, Duffy, J. E., J. P. Richardson, and E. A. Canuel. 2003. P. A. Cochran, J. J. Elser, M. M. Elser, D. M. Lodge Grazer diversity effects on ecosystem functioning in et al. 1987. Regulation of lake primary productivity sea grass beds. Ecology Letters 6: 637–645. by food web structure. Ecology 68: 1863–1876. Dunne, J. A., R. J. Williams, and N. D. Martinez. 2002a. Chase, J. M., P. A. Abrams, J. P. Grover, S. Diehl, Network structure and biodiversity loss in food webs: P. Chesson, R. D. Holt, S. A. Richards et al. 2002. robustness increases with connectance. Ecological Letters The interaction between predation and competition: 5: 558–567. a review and synthesis. Ecology Letters 5: 302–315. Dunne, J. A., R. J. Williams, and N. D. Martinez. 2002b. Chase, J. M., and M. A. Leibold. 2002. Spatial scale Food-web structure and network theory: the role of dictates the productivity–biodiversity relationship. connectance and size. Proceeding of the National Academy Nature 416: 427–430. of Sciences, USA 99: 12917–12922. Chase, J. M., W. G. Wilson, and S. A. Richards. 2001. Eklo¨v, P., and T. VanKooten. 2001. Facilitation among Foraging trade-offs and resource patchiness: theory piscivorous predators: effects of prey habitat use. and experiments with a freshwater snail community. Ecology 82: 2486–2494. Ecology Letters 4: 304–312. Finke, D. L. and R. F. Denno. 2004. Predator diversity Chesson, P. 2000. Mechanisms of maintenance of species dampens trophic cascades. Nature 429: 407–410. diversity. Annual Review of Ecological and Systemstic Finlay, B. J., S. C. Maberly, and J. I. Cooper. 1997. Micro- 31: 343–366. bial diversity and ecosystem function. Oikos 80: 209–213. 250 REFERENCES

Flo¨der, S., and U. Sommer. 1999. Diversity in planktonic Hillebrand, H., B. Worm, and H. K. Lotze. 2000. Marine communities: an experimental test of the intermediate microbenthic community structure regulated by disturbance hypothesis. Limnology and Oceanography 44: nitrogen loading and grazing pressure. Marine Ecology 1114–1119. Progress Series 204: 27–38. Forbes, S. A. 1887. The lake as a microcosm. Bulletin of the Holt, R. D., and M. Loreau. 2001. Biodiversity and Peoria Scientific Association 77–87. ecosystem functioning: the role of trophic interactions Fukami, T., and P. J. Morin. 2003. Productivity– and the importance of system openness. In: A. P. Kinzig, biodiversity relationships depend on the history of S. W. Pacala, and D. Tilman (eds), The Functional community assembly. Nature 424: 423–426. Consequences of Biodiversity. Princeton University Press, Fussmann, G. F., S. P. Ellner, and N. G. Hairston. 2003. Princeton, NJ, pp. 246–262. Evolution as a critical component of plankton dynamics. Holt, R. D., J. Grover, and D. Tilman. 1994. Simple rules for Proceedings of the Royal Society of London B 270: 1015–1022. interspecific dominance in systems with exploitative and Garlaschelli, D., G. Caldarelli, and L. Pietronero. 2003. apparent competition. American Naturalist 144: 741–771. Universal scaling relations in food webs. Nature 423: Holt, R. D., J. H. Lawton, G. A. Polis, and N. D. Martinez. 165–168. 1999. Trophic rank and the species–area relationship. Gaston, K. J., and T. M. Blackburn. 1999. A critique for Ecology 80: 1495–1505. macroecology. Oikos 84: 353–363. Horner-Devine, M. C., M. A. Leibold, V. H. Smith, and Groner, E., and A. Novoplansky. 2003. Reconsidering B. J. M. Bohannan. 2003. Bacterial diversity patterns diversity–productivity relationships: directness of pro- along a gradient of primary productivity. Ecology ductivity estimates matters. Ecology Letters 6: 695–699. Letters 6: 613–622. Grover, J. P. 1995. Competition, herbivory, and Huisman, J., and F. J. Weissing. 1999. Biodiversity of enrichment—nutrient-based models for edible and plankton by species oscillations and chaos. Nature inedible plants. American Naturalist 145: 746–774. 402: 407–410. Gurevitch, J., J. A. Morrison, and L. V. Hedges. 2000. Huston, M. 1979. A general hypothesis of species The interaction between competition and predation: diversity. American Naturalist 113: 81–101. a meta-analysis of field experiments. American Nat- Huston, M. A. 1994. Biological Diversity: The Coexistence of uralist 155: 435–453. Species in Changing Landscapes. Cambridge University Haddad, N. M., D. Tilman, J. Haarstad, M. Ritchie, and Press, Cambridge. J. M. H. Knops. 2001. Contrasting effects of plant Interlandi, S. J., and S. S. Kilham. 2001. Limiting resources richness and composition on insect communities: and the regulation of diversity in phytoplankton a field experiment. American Naturalist 158: 17–35. communities. Ecology 82: 1270–1282. Harrison, S. S. C., and A. G. Hildrew. 2001. Epilithic Irigoien, X., J. Huisman, and R. P. Harris. 2004. Global communities and habitat heterogeneity in a lake biodiversity patterns of marine phytoplankton and littoral. Journal of Animal Ecology 70: 692–707. zooplankton. Nature 429: 863–867. Havel, J. E., and J. B. Shurin. 2004. Mechanisms, effects Jeppesen, E., J. P. Jensen, M. Sondergaard, T. Lauridsen, and scales of dispersal in freshwater zooplankton. and F. Landkildehus. 2000. Trophic structure, species Limnology and Oceanography 49: 1229–1238. richness and biodiversity in Danish lakes: changes along Havens, K. 1992. Scale and structure in natural food a phosphorus gradient. Freshwater Biology 45: 201–218. webs. Science 257: 1107–1109. Jonsson, M., and B. Malmqvist. 2000. Ecosystem process Hawkins, B. A., and E. E. Porter. 2003. Does herbivore rates increases with animal species richness: evidence diversity depend on plant diversity? The case of from leave-eating aquatic insects. Oikos 89: 519–523. California butterflies. American Naturalist 161: 40–49. Jonsson, M., and B. Malmqvist. 2003a. Importance of Hillebrand, H. 2003. Opposing effects of grazing and species identity and number for process rates within nutrients on diversity. Oikos 100: 592–600. different stream invertebrate functional feeding Hillebrand, H. 2004. On the generality of the latitudinal groups. Journal of Animal Ecology 72: 453–459. gradient of diversity. American Naturalist 163: 192–211. Jonsson, M., and B. Malmqvist. 2003b. Mechanisms Hillebrand, H., and T. Blenckner. 2002. Regional and behind positive diversity effects on ecosystem local impact on species diversity—from pattern to functioning: testing the facilitation and interference processes. Oecologia 132: 479–491. hypotheses. Oecologia 134: 554–559. Hillebrand, H., and B. J. Cardinale 2004. Consumer effect Jonsson, M., B. Malmqvist, and P. O. Hoffsten. 2001. Leaf size declines with prey diversity. Ecology Letters 7: litter breakdown rates in boreal streams: does shredder 192–201. species richness matter? Freshwater Biology 46: 161–171. REFERENCES 251

Karez, R., S. Engelbert, and U. Sommer. 2000. predators on one prey. Theoretical Population Biology Co-consumption and protective coating: two new 65: 1–15. proposed effects of epiphytes on their macroalgal hosts Loreau, M. 2000. Biodiversity and ecosystem functioning: in mesograzer–epiphyte–host interactions. Marine recent theoretical advances. Oikos 91: 3–17. Ecology Progress Series 205: 85–93. Lotze, H. K., B. Worm, and U. Sommer. 2000. Propagule Keddy, P. A. 1989. Competition. Chapman and Hall, banks, herbivory, and nutrient supply control popu- London. lation development and dominance patterns in Knapp, R. A., K. R. Matthews, and O. Sarnelle. 2001. macroalgal blooms. Oikos 89: 46–58. Resistance and resilience of alpine lake fauna to fish Lubchenco, J. 1978. Plant species diversity in a marine introductions. Ecological Monographs 71: 401–421. intertidal community: importance of herbivore food Kneitel, J. M., and T. E. Miller. 2002. Resource and top- preference and algal competitive abilities. American predator regulation in the pitcher plant (Sarracenia Naturalist 112: 23–39. purpurea) community. Ecology 83: 680–688. Magnuson, J. J., W. M. Tonn, A. Banerjee, J. Toivonen, Kondoh, M. 2001. Unifying the relationships of species O. Sanchez, and M. Rask. 1998. Isolation vs. extinction richness to productivity and disturbance. Proceedings of in the assembly of fishes in small northern lakes. the Royal Society of London B. 268: 269–271. Ecology 79: 2941–2956. Koricheva, J., C. P. H. Mulder, B. Schmid, J. Joshi, and Mancinelli, G., M. L. Costantini, and L. Rossi. 2002. K. Huss-Danell. 2000. Numerical responses of different Cascading effects of predatory fish exclusion on the trophic groups of invertebrates to manipulations of detritus-based food web of a lake (Lake plant diversity in grasslands. Oecologia 125: 271–282. Vico, central Italy). Oecologia 133: 402–411. Krause, A. E., K. A. Frank, D. M. Mason, R. E. Ulanowicz, Martinez, N. D. 1991. Artifacts or attributes? Effects of and W. W. Taylor. 2003. Compartments revealed in resolution on the little rock lake food web. Ecological food-web structure. Nature 426: 282–285. Monographs 61: 367–392. Leibold, M. A. 1989. Resource edibility and the effects of Martinez, N. D. 1993. Effect of scale on food web struc- predators and productivity on the outcome of trophic ture. Science 260: 242–243. interactions. American Naturalist 134: 922–949. McCann, K. S. 2000. The diversity–stability debate. Nature Leibold, M. A. 1996. A graphical model of keystone 405: 228–233. predators in food webs: Trophic regulation of McQueen, D. J., M. R. S. Johannes, J. R. Post, T. J. Stewart, abundance, incidence, and diversity patterns in com- and D. R. Lean. 1989. Bottom-up and top-down munities. American Naturalist 147: 784–812. impacts on freshwater pelagic community structure. Leibold, M. A. 1999. Biodiversity and nutrient enrich- Ecological Monographs 59: 289–309. ment in pond plankton communities. Evolution and McPeek, M. A. 1998. The consequences of changing Ecological Research 1: 73–95. the top predator in a food web: a comparative experi- Leibold, M. A., J. M. Chase, J. B. Shurin, and mental approach. Ecological Monographs 68: 1–23. A. L. Downing. 1997. Species turnover and the regu- Menge, B. A., and J. P. Sutherland. 1976. Species diversity lation of trophic structure. Annual Review of Ecology and gradients: synthesis of the roles of predation, compe- Systematics 28: 467–494. tition, and temporal heterogeneity. American Naturalist Leibold M. A., M. Holyoak, N. Mouquet, P. Amarasekare, 110: 351–369. J. M. Chase, M. F. Hoopes, R. D. Holt, J. B. Shurin, Michiels, I. C., S. Matzak, and W. Traunspurger. 2003. Main- R. Law, D. Tilman, M. Loreau, and A. Gonzalez. 2004. tenance of biodiversity through predation in freshwater The metacommunity concept: a framework for multi- ? Nematology Monographs and Perspectives 2: 1–15. scale community ecology. Ecology Letters 7: 601–613. Mittelbach, G. G., C. F. Steiner, S. M. Scheiner, K. L. Gross, Leiden, N. L., and Traunspurger, W. 2002. Nematoda. In: H. L. Reynolds, R. B. Waide, M. R. Willig et al. 2001. S. D. Rundle, A. L. Robertson, and J. M. Schmid-Araya What is the observed relationship between species (eds), Freshwater Meiofauna: Biology and Ecology. Back- richness and productivity? Ecology 82: 2381–2396. huys Publishers, pp. 63–104. Montoya, J. M., M. A. Rodriguez, and B. A. Hawkins. Levins, R. 1979. Coexistence in a variable environment. 2003. Food web complexity and higher-level ecosystem American Naturalist 114: 765–783. services. Ecology Letters 6: 587–593. Lindeman, R. L. 1942. The trophic-dynamic aspect of Moorthi, S. 2000. Beziehung zwischen Diversita¨t ecology. Ecology 23: 399–418. und Stabilta¨t am Beispiel benthischer Protistenge- Loladze, I., Y. Kuang, J. J. Elser, and W. F. Fagan 2004. meinschaften, Masters Thesis, Christian-Albrechts- Competition and stoichiometry: coexistence of two Universita¨t zu Kiel. 252 REFERENCES

Myers, R. A., and B. Worm. 2003. Rapid worldwide and P. Inchausti (eds), Biodiversity and Ecosystem depletion of predatory fish communities. Nature 423: Functioning. Oxford University Press Oxford, pp. 280–283. 147–154. Naeem, S., D. R. Hahn, and G. Schuurman. 2000. Rosenzweig, M. L., and Z. Abramsky. 1993. How are Producer- co-dependency influences diversity and productivity related? In: R. E. Ricklefs biodiversity effects. Nature 403: 762–764. and D. Schluter (eds), Species Diversity in Ecological Naeem, S., and S. Li. 1998. Consumer species richness Communities. University of Chicago Press, Chicago, and autotrophic biomass. Ecology 79: 2603–2615. USA, pp. 52–65. Norberg, J. 2000. Resource-niche complementarity and Rundle, H. D., S. M. Vamosi, and D. Schluter 2003. autotrophic compensation determines ecosystem-level Experimental test of predation’s effect on divergent responses to increased cladoceran species richness. selection during character displacement in sticklebacks. Oecologia 122: 264–272. Proceedings of the National Academy of Sciences, USA 100: Oksanen, L., S. D. Fretwell, J. Arruda, and P. Niemela¨. 14943–14948. 1981. Exploitation ecosystems in gradients of primary Schmid-Araya, J. M., A. G. Hildrew, A. Robertson, productivity. American Naturalist 118: 240–261. P. E. Schmid, and J. Winterbottom. 2002a. The import- Pacala, S. W., and M. J. Crawley. 1992. Herbivores and ance of meiofauna in food webs: evidence from an acid plant diversity. American Naturalist 140: 243–260. stream. Ecology 83: 1271–1285. Paine, R. T. 1966. Food web complexity and species Schmid-Araya, J. M., P. E. Schmid, A. Robertson, diversity. American Naturalist 100: 65–75. J. Winterbottom, C. Gjerlov, and A. G. Hildrew. 2002b. Persson, L. 1999. Trophic cascades: abiding heterogeneity Connectance in stream food webs. Journal of Animal and the trophic level concept at the end of the road. Ecology 71: 1056–1062. Oikos 85: 385–397. Shurin, J. B. 2001. Interactive effects of predation and Persson, L., A. M. De Roos, D. Claessen, P. Bystrom, dispersal on zooplankton communities. Ecology 82: J. Lovgren, S. Sjogren, R. Svanback et al. 2003. Gigantic 3404–3416. cannibals driving a whole-lake trophic cascade. Shurin, J. B., and E. G. Allen. 2001. Effects of competi- Proceedings of the National Academy of Sciences, USA 100: tion, predation, and dispersal on species richness at 4035–4039. local and regional scales. American Naturalist 158: Petchey, O. L. 2000. Prey diversity, prey composition, 624–637. and predator population dynamics in experimental Shurin, J. B., and D. S. Srivastava. New perspectives on microcosms. Journal of Animal Ecology 69: 874–882. local and regional diversity: beyond saturation. In: Petchey, O. L., P. T. McPhearson, T. M. Casey, and M. Holyoak, R. D. Holt, and M. A. Leibold (eds), P. J. Morin. 1999. Environmental warming alters food- Metacommunities (in press). web structure and ecosystem function. Nature 402: Siemann, E., D. Tilman, J. Haarstad, and M. Ritchie. 1998. 69–72. Experimental test of the dependence of arthropod Polis, G. A., and D. R. Strong. 1996. Food web complexity diversity on plant diversity. American Naturalist 152: and community dynamics. American Naturalist 147: 738–750. 813–846. Soluk, D. A., and J. S. Richardson. 1997. The role of Post, D. M., M. L. Pace, and N. G. Hairston, Jr. 2000. stoneflies in enhancing growth of trout: a test of the Ecosystem size determines food-chain length in lakes. importance of predator–predator facilitation within a Nature 405: 1047–1049. stream community. Oikos 80: 214–219. Proulx, M., and A. Mazumder. 1998. Reversal of Sommer, U. 1985. Comparison between steady state and grazing impact on plant species richness in non-steady state competition: experiments with nutrient-poor vs. nutrient-rich ecosystems. Ecology 79: natural phytoplankton. Limnology and Oceanography 2581–2592. 30: 335–346. Proulx, M., F. R. Pick, A. Mazumder, P. B. Hamilton, Sommer, U. 1999. The impact of herbivore type and and D. R. S. Lean. 1996. Experimental evidence for grazing pressure on benthic microalgal diversity. interactive impacts of human activities on lake algal Ecology Letters 2: 65–69. species richness. Oikos 76: 191–195. Sommer, U., F. Sommer, B. Santer, E. Zo¨llner, K. Ju¨ rgens, Raffaelli, D., W. H. Van der Putten, L. Persson, C. Jamieson, M. Boersma et al. 2003. Daphnia versus D. A. Wardle, O. L. Petchey, J. Koricheva, M. van copepod impact on summer phytoplankton: functional der Heijden et al. 2002. Multi-trophic dynamics compensation at both trophic levels. Oecologia 135: and ecosystem processes. In: M. Loreau, S. Naeem, 639–647. REFERENCES 253

Steiner, C. F. 2001. The effects of prey heterogeneity and interactions link the microbial and classic food Webs. consumer identity on the limitation of trophic-level Limnology and Oceanography 40: 1173–1181. biomass. Ecology 82: 2495–2506. Wickham, S., S. Nagel, and H. Hillebrand. 2004. Control Steiner, C. F. 2003. Keystone predator effects and grazer of epibenthic communities by grazers and control of planktonic primary production. Oikos 101: nutrients. Aquatic Microbial Ecology 35: 153–162. 569–577. Williams, R. J., and N. D. Martinez. 2000. Simple rules Steinman, A. D. 1996. Effects of grazers on benthic yield complex food webs. Nature 404: 180–183. freshwater algae. In: R. J. Stevenson, M. L. Bothwell, Williams, R. J., E. L. Berlow, J. A. Dunne, A. L. Barabasi, and R. L. Lowe (eds), Algal Ecology—Freshwater Benthic and N. D. Martinez. 2002. Two degrees of separation in Ecosystems, Academic Press, pp. 341–373. complex food webs. Proceedings of the National Academy Sterner, R. W., and J. J. Elser. 2002. Ecological Stoichio- of Sciences, USA 99: 12913–12916. metry. Princeton University Press, Princeton, NJ. Wilson, W. G., P. Lundberg, D. P. Vazquez, J. B. Shurin Thebault, E., and M. Loreau 2003. Food-web constraints M. D. Smith, W. Langford, K. L. Gross, and on biodiversity–ecosystem functioning relationships. G. G. Mittelbach. 2003. Biodiversity and species inter- Proceedings of National Academy of Sciences, USA 100: actions: extending Lotka–Volterra community theory. 14949–14954. Ecology Letters 6: 944–952. Tilman, D. 1982. Resource Competition and Community Woodward, G., and A. G. Hildrew. 2002. Body-size Structure, Princeton University Press, Priceton, NJ. determinants of niche overlap and intraguild predation Tilman, D., J. Knops, D. Wedin, and P. Reich. 2001. Experi- within a complex food web. Journal of Animal Ecology mental and observational studies of diversity, pro- 71: 1063–1074. ductivity and stability. In: A. P. Kinzig, S. W. Pacala, and Worm, B., and J. E. Duffy. 2003. Biodiversity, productivity D. Tilman (eds), The Functional Consequences of Biodiversity. and stability in real food webs. Trends in Ecology and Princeton University Press, Princeton, NJ, pp. 42–70. Evolution 18: 628–632 Vadeboncoeur, Y., E. Jeppesen, M. J. Vander Zanden, Worm, B., H., K. Lotze, C. Bostro¨m, R. Engkvist, H. H. Schierup, K. Christoffersen, and D. M. Lodge. V. Labanauskas, and U. Sommer. 1999. Marine 2003. From Greenland to green lakes: cultural eutro- diversity shift linked to interactions among grazers, phication and the loss of benthic pathways in lakes. nutrients and propagule banks. Marine Ecology Progress Limnology and Oceanography 48: 1408–1418. Series 185: 309–314. Vander Zanden, M. J., B. J. Shuter, N. Lester, and Worm, B., H. K. Lotze, and U. Sommer. 2001. Algal J. B. Rasmussen. 1999. Patterns of food chain length in propagule banks modify competition, consumer and lakes: a stable isotope study. American Naturalist 154: resource control on Baltic rocky shores. Oecologia 128: 406–416. 281–293. Vinebrooke, R. D., D. W. Schindler, D. L. Findlay, Worm, B., H. K. Lotze, H. Hillebrand, and U. Sommer. M. A. Turner, M. Paterson, and K. H. Milis. 2003. 2002. Consumer versus resource control of species Trophic dependence of ecosystem resistance and diversity and ecosystem functioning. Nature 417: species compensation in experimentally acidified lake 848–851. 302S (Canada). Ecosystems 6: 101–113. Yoshida, T., L. E. Jones, S. P. Ellner, G. F. Fussmann, and Wahl, M., M. E. Hay, and P. Enderlein. 1997. Effects of epibio- N. G. Hairston. 2003. Rapid evolution drives ecological sis on consumer-prey interactions. Hydrobiologia 355: 49–59. dynamics in a predator-prey system. Nature 424: Waide, R. B., M. R. Willig, C. F. Steiner, G. Mittelbach, 303–306. L. Gough, S. I. Dodson, J. P. Juday et al. 1999. The rela- tionship between productivity and species richness. Chapter 15 Annual Review of Ecology and Systematics. 30: 257–300. Weithoff, G., A. Lorke, and N. Walz. 2000. Effects of Almaas, E., and A.-L. Baraba´si, and A.-L. 2004. Power water-column mixing on bacteria, phytoplankton, and laws in biological networks. Available at: ArXiv rotifers under different levels of herbivory in a shallow . eutrophic lake. Oecologia 125: 91–100. Bak, P. 1996. How Nature Works: The Science of Self- Werner, E. E., and S. D. Peacor. 2003. A review of trait- Organized Criticality. Copernicus, New York, 212 p. mediated indirect interactions in ecological commun- Baraba´si, A.-L. 2002. Linked: The New Science of Networks. ities. Ecology 84: 1083–1100. Perseus Publications, Cambridge, MA, 280 p. Wickham, S. A. 1995. Trophic Relations between Baraba´si, A.-L. and R. Albert. 1999. Emergence of scaling cyclopoid copepods and ciliated protists—complex in random networks. Science 286: 509–512. 254 REFERENCES

Casti, J. L. 2004. Why the future happens. The Second Prigogine, I., and I. Stengers. 1984. Order out of Chaos: Man’s International Biennial Seminar on the Philosophical, New Dialogue with Nature. Bantam, New York, 349 p. Methodological and Epistemological Implications of Rutledge, R. W., B. L. Basorre, and R. J. Mulholland. 1976. Complexity Theory. International Convention Center, : an information theory viewpoint. La Habana, Cuba, January 7–11, 2004. Journal of Theoretical Biology 57: 355–371. Depew, D. J., and B. H. Weber. 1994. Evolving: Ulanowicz, R. E. 1980. An hypothesis on the develop- Systems Dynamics and the Geneology of . ment of natural communities. Journal of Theorectical MIT Press, Cambridge, MA, 588 p. Biology 85: 223–245. Elsasser, W. M. 1969. Acausal phenomena in physics and Ulanowicz, R. E. 1986. Growth and Development: Ecosystems biology: a case for reconstruction. American Scientist Phenomenology. Springer-Verlag, New York, 203 p. 57(4): 502–516. Ulanowicz, R. E. 1995. Utricularia’s secret: the advantages Higashi, M., and T. P. Burns. 1991. Theoretical Studies of positive feedback in oligotrophic environments. of Ecosystems: The Network Perspective. Cambridge Ecological Modelling 79: 49–57. University Press, Cambridge, 364 p. Ulanowicz, R. E. 1997a. Ecology, the Ascendent Perspective. Hirata, H., and R. E. Ulanowicz. 1984. Information Columbia University Press, New York, 201p. theoretical analysis of ecological networks. International Ulanowicz, R. E. 1997b. Limitations on the connectivity of Journal of 15: 261–270. ecosystem flow networks. In: A. Rinaldo and Holling, C. S. 1986. The resilience of terrestrial eco- A. Marani (eds), Biological Models. Istituto Veneto de systems: local surprise and global change. In: W. C. Scienze, Lettere ed Arti. Venice, pp. 125–143. Clark and R. E. Munn (eds), Sustainable Develoment of Ulanowicz, R. E. 1999. Life after Newton: an ecological the Biosphere. Cambridge University Press, Cambridge, metaphysic. BioSystems 50: 127–142. pp. 292–317. Ulanowicz, R. E. 2001. The organic in ecology. Ludus Jeong, H., B. Tombor, R. Albert, Z. N. Oltvai, and Vitalis 9(15): 183–204. A.-L. Baraba´si. 2000. The large-scale organization of Ulanowicz, R. E. 2002. The balance between adaptability metabolic networks. Nature 407, 651–654. and adaptation. BioSystems 64: 13–22. Kuhn, T. S. 1962. The Structure of Scientific Revolutions. Ulanowicz, R. E. 2004. Order and fluctuations in ecosystem University of Chicago Press, Chicago, IL, 172 p. dynamics. Emergence. May, R. M. 1972. Will a large complex system be stable. Ulanowicz, R. E. 2004. Ecosystem dynamics: a natural Nature 238: 413–414. middle 2(2): 231–253. McLuhan, H. M. 1964. Understanding Media: The Extension Ulanowicz, R. E., and J. Norden. 1990. Symmetrical of Man. McGraw-Hill, New York. overhead in flow networks. International Journal of Minarik, E. M. 1957. Little Bear. Harper and Row, Systems Science 21(2): 429–437. New York, 63 p. Ulanowicz, R. E., and W. F. Wolff. 1991. Ecosystem flow Montoya, J. M., and R. V. Sole. 2002. Small-world patterns networks: loaded dice? Mathematical Biosciences 103: in food webs. Journal of Theoretical Biology, 214:405–412. 45–68. Odum, H. T. 1971. Environment, Power and Society. Wiley, Wagensberg, J., A. Garcia, and R. V. Sole. 1990. Con- New York, p. 331. nectivity and information transfer in flow networks: Pimm, S. L. 1982. Food Webs. Chapman and Hall, London, two magic numbers in ecology? Bulletin of Mathematical 219 p. Biology 52: 733–740. Pimm, S. L., and J. H. Lawton. 1977. Number of tro- Watts, D. J. 1999. Small Worlds: The Dynamics of Networks phic levels in ecological communities. Nature 268: Between Order and Randomness. Princeton University 329–331. Press, Princeton, NJ, 262 p. Popper, K. R. 1977. The bucket and the searchlight: Wulff, F., J. G. Field, and K. H. Mann (eds). 1989. Net- two theories of knowledge. In: M. Lippman (ed.), work Analysis in Marine Ecology. Springer-Verlag, Discovering Philosophy. Prentice Hall, Englewood Cliffs, Berlin, Heidleberg, New York. NJ, pp. 328–334 Zorach, A. C., and R. E. Ulanowicz. 2003. Quantifying the Popper, K. R. 1990. A World of Propensities. Thoemmes, complexity of flow networks: how many roles are Bristol 51 p. there? Complexity 8(3): 68–76. Index

Page numbers in italics refer to Figures. abundance-body mass relationship to Shannon flow biomass turnover rate (P/B) 27 relationships 67, 86–7, diversity 83 Blenniidae sp. 21 92, 93, 94–5 blue sharks 153 acceleration of marine food body size see size of organisms webs 178, 181 bacteria, seasonal changes 39 Bosmina sp., stoichiometric acidification of streams 62 Baltic Sea niche 13 Acropora sp. 180 effect of eutrophication 30 bottom-up control, effect on stability adaptation 133–4 interecosytem comparisons 32, 120–1 see also foraging adaptations 35,36 boundaries adaptive food web hypothesis Barents Sea (BS) 144, 145, 146, in marine environment 101 140–2 147, 168 in streams 59 adaptive food web model 135–7 community dynamics 161–2 Broadstone Stream food web 51, adaptive networks 126 food web structure 151, 156–7 52,53 Alaska coastal current (ACC) 147 North Atlantic Oscillation Cordulegaster boltonii invasion 62 Alaska Current (Alaska Stream) (NAO) 149–50, 165–7, 166 long-term stability 61 145, 147 basal prey diversity 192–3 prey overlap graphs 55 algae, seasonal changes 45, 179 bay anchovy, seasonal changes 39 quantification 54 algal mats, spatial structure Benguela food web dynamics 127 seasonal variability 60 study 20–4 benthic processes, role 31–2 brown trout (Salmo trutta) Allen’s paradox 59 bigeye tuna 152–3 Allen’s paradox 59 allochthonous material, importance biodemographic(interaction) food invasion of stream food webs 62 in streams 59 webs 103 buffering 172 amphipods, population biodiversity 184 Bythotrephes longimanus 45 explosions 174–5 consumer diversity 185, 187–8, anchovies 152 190–2 bay anchovies, seasonal effect on trophic interactions 185, Calanus sp. 156 changes 39 188–9 Calanus finmarchicus 165 apex predators of Tropical Pacific future research areas 196–7 cannibalism 46, 99, 100 153 prey diversity 184, 185, 186–7, capelin 154, 156–7 Arctic Current 147 189–90 abundance in GOA 159, 164 arrowtooth flounder 155, 159, 160, propagating effects 193 in Barents Sea 161, 162 164 regional effects 194–6 effects of NAO 166 ascendency of networks 83, 84, 205 in streams 63–6 effects of population As/Cd 27, 85 see also species richness fluctuation 165–6, 167, 168 definition 73 bioenergetic dynamics model 123, Carangidae (scads) 152 asymptotic local stability 130 125 carbon Atlantic Current 147 bioenergetic food webs 103 C:P ratio 11–12 autocatalysis 202–3 biogeochemical budgeting models, pools and fluxes 10 autotroph-grazer limit cycles 12 LOICZ 33–6 carbon-based flow charts 9 average mutual information (AMI) biological stoichiometry see stoichio- carbon flow models 42 73, 76, 82–3, 84 metry Carcharhinidae sp. 21 effect of rules of hypothetical biomanipulation 42 Caribbean coral reef food web 171–2 webs 85 biomass dynamics 142 trophic cascades 173–4

255 256 INDEX

Caribbean fish, study of spatial community persistence, effect of consumers structure 19–24 species richness 138 control of prey biomass, effect of cascade model 69, 70–1, 99, 125 comparative ecosystem ecology 25 prey diversity 189 Caulerpa taxifolia, introduction comparisons between food webs effect on prey diversity 184, 185, to Mediterranean 180 26–7, 39–40 186–7 cellular metabolism 7, 8 interecosystem comparisons 30–2 diversity 185, 187–8, 189–92 centripetality 203–4 temporal comparisons 27–30, 28 see also predators Ceratium sp. 49 compartmentalization 122 consumption, relaxation at low chaotic population dynamics 121, competition-colonization trade-off 19 resource density 125 123–4 competition theory 189 copepods chemical stress 32 complexity 7–8, 18, 98–9 in Barents Sea 156 Chesapeake Bay food web of marine food webs 113 temperature effects 48, 49 analysis 27, 28,29 relationship to energy input in Tropical Pacific 152 interecosystem comparisons 31–2, of systems 180, 181 coral reefs 35,36 and species persistence 125–6 conservation 181–2 regular equivalence model 36–9 and species relative effect of intense fishing 176 seasonal changes 26 abundance 177–8 human activities, impact of 180 cladocerans, temperature effects 49 of stream food webs 51–2 spatial structure study 20–4 clear-water phase 45 and susceptibility to damage trophic cascades 180 effect of temperature 48 by human activity 181 zooxanthellae 175 climate change 197 complexity-stability relationship Cordulegaster sp. effect on lakes 41 56, 98, 99, 118–20, 126, Cordulegaster boltonii, invasion of effect on plankton food webs 46–50 130–2, 171–2 Broadstone Stream 62 effect on streams 63 adaptive food web hypothesis prey overlap graphs 55 climate forcing 47 137–42 cormorants 161 El Nin˜o Southern Oscillation adaptive food web model 135–7 65 (ENSO) 146, 147–9, 148 community matrix analysis 120–3 life-cycle omnivory 14 effects on tuna 162–4, 163 community-population crustaceans in Gulf of Alaska 164–5 interactions 140 carnivorous, seasonal variation in North Atlantic Oscillation effect of food web flexibility 132–3 diet 45 (NAO) 149–50 effect of foraging adaptations seasonal changes 39 Pacific inter-Decadal Oscillation 134–5, 138–40 Crystal River, trophic efficiency (PDO) 147, 148 in marine ecosystems 167, 168 study 31 Clupeidae 152 May’s findings 119, 130–2 cyanobacteria, effect of mild see also herring monocultures, vulnerability 118 winters 48 clustering 122 conditional entropy 205 cycling 99, 100 coastal marine food webs, trophic connectance 99 cycloid copepods, seasonal variation cascades 172–3 constant 100 in diet 45 coastal systems, Land Ocean effect of adaptive foraging 136–7 Interactions in the Coastal effect on population dab 157, 162 Zone (LOICZ) project 33 persistence 138 Daphnia sp. cod in dynamical food web models Daphnia pulicaria, growth rate Atlantic 157 126 hypothesis 16,17 in Barents Sea 161, 162 relationship to diversity 193 effect of high NAO winters 47, 48 effect of capelin population snapshot 136 effect of phosphorus-limitation 12 fluctuations 167 connectance food webs 51–3, 52 phosphorus limitation 11, 13 effect of NAO 165, 166 connections, quantification 77–8, temperature effects 48, 49 Pacific 154, 155, 160, 164 79, 80 dark respiration, in hypothetical food polar 156 see also interaction strength web construction 76 cod fishing moratorium 144 conservation of marine deepwater redfish 157, 162 Coleoptera, body nitrogen food webs 181–2 defence adaptations 133–4 content 14 Constance, Lake deforestation, effects on stream food combinatorics 202 pelagic food web 42–3 webs 61 community approach to seasonal changes 45 Delaware Bay, interecosystem com- quantification 55, 56–7 temperature effects 48, 49 parisons 31–2 community matrices 20 constant connectance model 69, determinism 201 stability analysis 120–3 100, 111 detritus, role in estuaries 30 INDEX 257 detritus-based streams 56 ecological stoichiometry 8, 9 extensive metric, definition 73 biodiversity 64 application to streams 57–8 external forcing 143 seasonal variability 60 Ecopath/Ecosim systems 26, 27, 111 detritus chain 42 ecosystem approach to feasibility 130 developmental capacity (Cd) 81, 82 quantification 55 feeding, relaxation at low resource As/Cd 27, 85 ecosystem size hypothesis 58 density 125 definition 73 ECOWeB database 69, 72 see also foraging adaptations ‘devious strategies’ (May) 119, eggs, effect of climate change on Finn Cycling Index (FCI) 27 120, 129 hatching 46 as indicator of stress 32 diapause 46 El Nin˜ o Southern Oscillation fish diatoms, effects of high NAO 48–9 (ENSO) 47, 146, 147–9, 148 absence from ponds 192 diet, seasonal variation 45 effects on tuna species 158, 162–4, in Barents Sea 156–7, 161, 162 dietary preferences 13 163 bias towards in marine food see also foraging adaptations; emergent properties of webs 110, 111 preference switching food webs 193–4 Caribbean, spatial structure 19–24 dimensions of food webs 77–9 empirical food web analysis 76–7 effects in stream food webs 65 Diptera, body nitrogen content 14 empirical food webs in Gulf of Alaska 154–5 dispersal effects 194–5 relationship to real webs 85 phosphorus requirements 13 diversity see biodiversity; complexity stability 122 relationship of body size to trophic diversity-stability debate 118 Enchrasicholinus punctifer 152 level 89 see also complexity-stability energy equivalence rule 91–2 seasonal changes ion lakes 45 relationship Engraulidae (anchovies) 152 size-based analysis 95–6, 97 dolphins 153, 157 entropic tendency 205 in Tropical Pacific 152 dominance of food chains 171 environmental gradients and human winter, effect of 48 dormancy 46 impact 180–1 fishing dragonflies 65 Epischura lacustris, egg-hatching 46 in Benguela ecosystem 127–8 Drosophilasp., C:N:P equal weighting of species 65 impact 178, 179–80 stoichiometry 14–15 equatorial currents 144, 146 overexploitation 170–1, 175, 176 dynamic metacommunity model equatorial undercurrent (EUC) 146 recovery of fish populations 182 21–3 equilibrium point 120 flathead sole 160 dynamically propagated effects , protein amino acid flexible food web structure 132–3 171–2 structure 15 floodplains 59 in marine food webs 172–5 Esthwaite Water, effect of high NAO Florida Bay, food web analysis 27, 28, dynamics modelling 117–18, 128–9 winters 47 29 Benguela ecosystem 127–8 estuaries flounder, arrowtooth 155, 159, 160, complex food webs, species persis- effect of eutrophication 26 164 tence 125–6 effect of freshwater flow reduc- flow diversity see Shannon flow non-linear, non-equilibrium tion 26 diversity (SI) dynamics 123–4 effect of intense fishing 176 flows types of persistent dynamics 121 interecosystem comparisons quantification 77–8, 79, 80 dynamics of networks, 30–2 randomization in hypothetical food visualization 36–9 temporal comparisons 27–30 webs 75 Eudiaptomus sp., effect of high NAO flyingfish 152 ecodiversity 179 winters 47 food chain length 91 eulachon 154 effect of dispersal 194–5 ecological network analysis abundance in GOA 159 influence of PPMR 93–4 (ENA) 25, 39–40 Eurycercus lamellatus, temperature relationship to lake surface autocatalysis 202–4 effects 49 area 195–6 centripetality 203–4 eutrophication food chain theory 42 comparison with LOICZ of Baltic Sea 30 foraging adaptations 133, 134–5, 141 approach 33–6 in estuaries 26 effect on complexity-stability rela- dynamics 205–6 effect on metrics 73 tionship 138–40 visualization of 36–9 evolution, responses to stoichiometric effect on food web structure 136 food web comparisons 26–7 imbalance 14–17 see also dietary preferences; prefer- interecosystem comparisons 30–2 evolutionary food web models 126 ence switching software systems 26 evolutionary processes 196–7 foraging effort dynamics 142 temporal comparisons 27–30, 28 Exocœtidae (flyingfish) 152 forest food webs 61, 62 258 INDEX fractional flows, quantification 77–8 herbivores, effect of removal 175 insects freshwater input, effect in herring 152, 156–7 body nitrogen content 14 estuaries 30 in Barents Sea 161, 162 in streams 60 full developmental capacity see devel- effect of capelin population fluctua- intensive metric, definition 73 opmental capacity (Cd) tions 167 interaction (functional/ biodemo- function of webs 73–4 effect of NAO 165, 166 graphic) food webs 103 functional (interaction) food web 1, 2 interaction strength 54–5, 56, 171–2 food webs 103 Pacific 154 and complexity-stability fur seal culling 127–8 heterotrophic nanoflagellates relationship 132 (HNF) 42 estimation in marine food galaxiids 62 high nutrient, low-chlorophyll webs 111 gelatinous zooplankton 102, 152 (HNLC) situation 152 measurement 57–8 genome, evolutionary change 17 history of food webs 98–100 weak interactions 124, 129 global asymptotic stability 130 homogenization 177 interannual changes 28, 29–30 global warming in marine food webs 178, 180 interecosystem comparisons 30–2 effect on lakes 41 hot links 200 International Geosphere-Biosphere effect on streams 63 human activity 197 Program (IGBP) 32–3 see also climate change assessment of impact on marine intersystem comparisons 27 grassland food webs 61, 62 ecosystem 95 invasion 118–19, 178 gray whales, effects of hunting 173 effect on environmental homogenization 180 grazing chain 42 gradients 180–1 stream food webs 61–2 ‘green world’ hypothesis (Hairston, effect on marine successional iron limitation, in Tropical Smith and Slobodokin trends 179–80 Pacific 152 [HSS]) 42 effect on stream food webs 61–3 growth rate hypothesis (GRH) 15 humans as top predators 175 kelp forests testing in Daphnia pulicaria 16,17 hypothetical food web amphipods 174–5 growth rate, relationship to body construction 74–6 effect of intense fishing 176 size 87 connections 77 successional changes 179 guillemot 157 respiration rates 76 trophic cascades 173 population decline 162, 165, 167 rules, effect on information keystone predation 186 Gulf of Alaska (GOA) 144, 145, metrics 84–5 keystone species 102 146–7, 167–8 transfer coefficients 76 in stream food webs 65 community dynamics 158–61 killer whales (orcas) 153 effects of PDO 164 ice cover duration 47 effect of altered feeding effect of sea surface temperature immigration behaviour 173 change 143 incorporation into metacommunity effect on food web complexity 175 food web structure 151, 154–6 model 22 predation on Stellar sea lions 165 Pacific inter-Decadal Oscillation see also dispersal effects; invasion kittiwakes, black-legged 160–1 (PDO) 147, 164 indices in ecological network analy- krill 156 Gulf of Mexico, hypoxic zone 170 sis 27 Kromme Estuary Gulf Stream variations 48 indirect diet 31 food web analysis 27, 28, 29–30 gut content analysis (GCA) 57 indirect effects in food webs 130 interecosystem comparisons 30–1, inedible prey 192 32, 35,36 haddock 157 information metrics 73 in Barents Sea 162 ascendency 83, 84, 205 lakes hake fishery, Benguela average mutual information effect of climate change 46–50 ecosystem 127–8 (AMI) 82–3 effect of temperature increase 41 halibut 159, 160, 164 developmental capacity (Cd) 81 seasonal succession 44–6 Greenland 157, 162 empirical food web analysis trophic structure of food webs Pacific 155 76–7 42–4 Halodule wrightii, network food web dimensions 77–9 Land Ocean Interactions in the analysis 27, 28 hypothetical network analysis 76 Coastal Zone (LOICZ) 33–6 hammerhead sharks 153 rules of organization of hypotheti- land-use change, effects on stream harbour seals 155, 160 cal webs, effects 84–5 food webs 61 harp seal 157 Shannon flow diversity (SI) 80–1 ‘laws’ of food webs 99 Hawaiian food webs 175, 177 total system throughput (TST) Leptodora kindtii 45 Hemiptera, body nitrogen content 14 78, 79–80 life cycle omnivory 14, 46 INDEX 259 life cycles, effect of climate marlin 153 Newtonian systems, postulates change 46 Maspalomas Lagoon, food web (Depew and Weber) 201 lifespan, relationship to body size 87 analysis 27, 28,29 niche model 125 light, effect of high levels 11 mature communities 178 and marine food webs 111–12 likelihood methods 70–2 Mediterranean 180–1 nitrogen, body content of limiting nutrients 11–14 mesozooplankton 152 insects 14 linear extension of partial mesquite trees, effect on phosphorus nitrogen stable isotope ordering 72 availability 16 analysis 88 link scaling law 99 metabolic rate, relationship to non-linear, non-equilibrium link-density 205–6 body size 87 dynamics 123–4, 128 link-species scaling law 99 metacommunity models 19, 21–3 North Atlantic Oscillation (NAO) local (neighbourhood) stability metaphoetesis 101 149–50 analysis 120–3 meteorological forcing 47–9 effects in Barents Sea 165, 166 Loch Ness monsters metrics effects of variations 47, 48–9 biomass estimation 87 in marine food webs 112 North Equatorial Current (NEC) 144, population density prediction 95 see also information metrics 146 loops 122–3 microalgae, role in Sundays North Sea fish, size-based Estuary 30 analysis 89, 96,97 MacArthur hypothesis 1 microbes, role of 3 northwest Atlantic foodweb 7, 8 mackerel, atka 155 microbial loop 2, 102, 154 Norwegian Atlantic Current 150 Macrocystis pyrifera 179 Microcystis sp. 49 Norwegian Coastal Current 147 macroecological approach 87, 196–7 Notholca caudata, temperature 96–7, 197 micronekton 150 effects 49 macronekton 150 in Barents Sea 156 numbers, delimitation 201–2 in Barents Sea 157 in Gulf of Alaska 154–5, 159 nutrient availability, effect of in Gulf of Alaska 155, 159–60 in Tropical Pacific 152 meteorological forcing 48 in Tropical Pacific 152–3 microzooplankton 152 nutrient limitation 11–14 macrozooplankton 152 minke whale 157 nutrient loading 178 mako sharks 153 Mnemiopsis sp. 178 LOICZ project variables 33 mammals, marine 153, 157 modelling see also killer whales; seals; whales adaptive food web 135–7 oceanic anchovy 152 Manduca sexta, nutrient limitation 13 community matrices 120–3 offshore reefs, spatial structure mangrove habitat, spatial structure dynamic metacommunity 21–3 study 20–4 study 20–4 hypothetical food web construc- omnivory 13–14, 172, 192–3 marine ecosystems tion 74–7 effect on stability 122, 124 assessment of human impact 95 likelihood methods 70–2 life-cycle omnivory 46 ecological disasters 170 non-linear dynamics 123–4 in marine species 101 history of food webs 98–100 software 26 study in Caribbean 20–4, 21, 23 marine environment, scale 100–1 see also specific models ontogenetic shifts in marine marine food webs 112–13 monocultures, vulnerability 118 species 101 ancillary information sources mortality, relationship to body orcas see killer whales 110–11 size 87 organism-focussed reasoning 7–8 bias towards fish 110 murres 161 Oscillatoria sp. 49 categorization 103 Mytilus californianus 172–3 see fishing: compilation 104–9 overexploitation conservation 181–2 Narragansett Bay, interecosystem overhead of networks 205 Ecopath/Ecosim models 111 comparisons 31–2 oysters, effect of over-harvesting 26 effect of intense fishing 176 Neoptera, body nitrogen content 14 and food web theory 111–12 nested hierarchy 53 Pacific inter-Decadal Oscillation metrics 103, 110 network analysis (NA) 25, 118 (PDO) 147, 148, 167–8 representation of marine see also ecological network analysis effects in Gulf of Alaska 159, 164 ecosystems 112 (ENA) effects on tuna species 158 trophic cascades 172–5 network motifs 84 pandalid shrimps, abundance in marine pelagic ecosystems 144–6 NETWRK4 26, 27 GOA 159, 164 marine phytoplankton, abundance- Neuse River Estuary, food web parrotfish, population decline 174 body-mass relationships 94 analysis 28–9, 35,36 PEG (Plankton Ecology Group) marine species 101–2 neutral stability 120 model 44 260 INDEX pelagic food webs 10–11 population/community approach see also trophic efficiency marine 144–6 to quantification 55, proteins, amino acid structure 15 seasonal changes 44–6 56–7 Peridinium sp. 49 porpoise 157 quantification of food webs 54–6 permanence 130 power-law networks 129 measurement of interaction persistence predation strength 57–8 in adaptive food webs 137 reciprocal 53 see also information metrics in complex food webs 125–6 size-based 86 effect of connectance 138–9 predation matrix 70 relationship to species predation probability 71 random communities, stability analy- richness 137–8, 140 predator dominance model 71–2 sis 120 persistent dynamics, types 121 predator presence, cascading random food web models, variation phosphorus effects 192 in stability 131–2 association of body content with predator-prey interactions, in size ranking of species 71–2 growth rate 11 spectrum 90–1 rare species, significance 102 C:P ratio 11–12 predator-prey mass ratio (PPMR) 88, rDNA variation 15, 17 growth rate hypothesis 89–90, 96 reciprocal predation 53 (GRH) 15 influence of food chain length recruitment limitation 19 low availability 11–12 93–4 recurring circuit elements 84 pools and fluxes 10 predator-prey models, lack of reductionism 204 phosphorus-based flow charts 9 stability 119 redundancy in stream food webs 64 photorespiration in hypothetical food predator-prey relationship refugia, stabilization of food web construction 76 adaptations 133–4 webs 56–7 physical stress 32 and complexity-stability REGE algorithm 37–8 phytoplankton relationship 132 regional forces in food webs 194–6 abundance-body-mass predators regular equivalence model 36–9 relationships 94 effect of removal 172–3, 174, regularities in food webs 70 temperature effects 47, 48–9 177, 178 reproductive output, relationship to pinnipedia 155 humans 175 body size 87 in Barents Sea 157 nutrient limitation 13 research, new approaches 1, 3, 196–7 see also seals of Tropical Pacific 153 resilience 130 Pisaster sp. 172–3 see also consumers resource enrichment 178 pitcher plant (Sarracenia purpurea) preference switching 64, 124 LOICZ variables 33 192 effect on stability 186 respiration rates in hypothetical food plankton in Gulf of Alaska macronekton web construction 76 effect of climate change 46–50 155 resting stages 46 PEG model 44 see also foraging adaptations River Continuum Concept (RCC) 58 resting stages 46 presence/absence community RNA variation 15, 16,17 see also phytoplankton; matrices 20 rockfish 155 zooplankton prey, inedible 192 roles 84 planktonic systems, trophic prey biomass, effect of consumer running water see streams efficiency 10 diversity 190–1 Planktothrix, effect of mild prey density, effect on consumer sablefish 155 winters 48 diversity 185, 187–8 Saccaromyces cerevisiae, protein amino plasticity 64 prey diversity 184, 185, 186–7, acid structure 15 Plectrocnemia sp., prey overlap 189–90, 191–2 sailfish 153 graphs 55 prey overlap graphs 55 saithe 157 pollock 154, 155, 160, 164 primary producers in Barents Sea 162 pollution 32 in Barents Sea 161 saltmarsh grass fertilization, link with in streams 62 dominance of food chains 171 production 25 ponds, fishless 192 in Gulf of Alaska 159 sample size 51, 52 population adaptation 133 in Tropical Pacific 158 sampling techniques 90 population dynamics, time series probabilistic drift 205 in marine environment 100 121 process ecology 204 sand, spatial structure study 20–4 population growth restriction, effect productivity sandfish on stability 120 effect on consumer-prey abundance in GOA 159 population persistence see persistence dynamics 186–7, 187–8 Pacific 154 INDEX 261 sandlance 154 abundance-body-mass complex food webs, species sardines 152 relationships 67, 86–7, persistence 125–6 Sarracenia purpurea 192 92, 93, 94–5 methodologies 128 scads 152 food-chain length 93–4 non-linear, non-equilibrium scale-free systems 177, 200 practical applications 95–6 dynamics 123–4 scale invariance laws and marine predator-prey interactions 90–1 Stability and Complexity in Model food webs 112 predator-prey mass ratio Ecosystems! (May, 1972, scale of marine environment 100–1 (PPMR) 88, 89–90 1973) 119, 120 scaling relationships 87 relationship of size to trophic stable isotope analysis (SIA) 57 Scaridae sp. 21 level 87–9 172–3 Scenedesmussp. 49 size spectrum dynamics 90 statistical analysis, likelihood scientific approach 201 size of organisms methods 70–2 scombrids 153 and biological properties 87 Stellar sea lions 155, 160, 161 sea otters relationship to food web population decline 165 effect of removal 173 patterns 53–4 stoichiometric growth rate population decline 165 relationship to trophic level 44, hypothesis 11 sea surface temperature change, effect 87–9 stoichiometric imbalance 9–11 in Gulf of Alaska 143 size of webs 73 evolutionary responses 14–17 sea urchins, population effect on information metrics 84 stoichiometric niches 13 explosions 173, 179–80 skipjack tuna (Katsuwonus stoichiometry 8–9 seabirds 153–4 pelamis) 152 and nutrient limitation 11–13 in Barents Sea 157, 162 snapshot connectance 136 and omnivory 13–14 in Gulf of Alaska 155–6, 160, 164–5 solar energy see light stratification of lakes 49 seagrass beds/algal mats, spatial sole, flathead 155 effect of climate change 46 structure study 20–4 South Equatorial Current (SEC) 144, seasonal variation 47 seals 146 streams 51–3, 52,66 in Barents Sea 157, 162, 165 Southern Oscillation (SO) see El Nin˜o anthropogenic effects 61–3 culling 127–8 Southern Oscillation (ENSO) approaches to quantification 54–6 in Gulf of Alaska 155, 160 Spartina alterniflora marshes, stabi- population/community seasonal changes 30 lity 119 approach 56–7 in algal biomass 179 spatial population dynamic model measurement of interaction in Gulf of Alaska 154 (SEPODYM) of tuna 163–4 strength 57–8 in lakes 41, 44–6, 49 spatial structure, study in biodiversity 63–6 regular equivalence model for Caribbean 19–24 spatial variation 58–60 Chesapeake Bay 36–9 spatial variation, in streams 58–60 temporal variation 60–1 in streams 60, 61 species numbers, in marine food stress 32 secondary production 31 webs 103, 110 strict regional similarity (SRS) 22 selection pressure 203–4 species relative abundance and strict regional trophic similarity self-sustainability 140 complexity 177–8 (SRTS) 22 Shannon flow diversity (SI) 76, 80–1 species richness structural analysis 69 effect of web size 84 relationship to population structured webs relationship to average mutual persistence 137–8, 140 ascendency 83, 84 information 83 see also biodiversity average mutual information shark fishing, effect on Caribbean species scaling law 99 (AMI) 82–3 reefs 174 Sphyrnidae sp. 21 construction 74–5 sharks 153 stability 26–7, 118 developmental capacity (Cd) 81, 82 shell fish production, link with of marine food webs 111, 112 Shannon flow diversity 80–1 fertilization of saltmarsh measurement of 120 total system throughput 79–80 grass 25 in natural food webs 129 sub-Arctic gyre 146–7 153 preference-switching, effect successional changes shrimps 152, 154 of 186 in lakes 44–6 abundance in GOA 159, 164 refugia, effect of 56–7 in marine food webs 179–80 Sialis sp., prey overlap graphs 55 see also complexity-stability summer Sibinia setosa effect of low C:P relationship; persistence climate forcing 47 ratio 15, 16 stability analysis meteorological forcing 48 silky sharks 153 Benguela ecosystem 127–8 Sundays Estuary, interecosystem size-based analysis 86–7, 96–7 of community matrices 120–3 comparisons 30–1, 32, 35,36 262 INDEX superlinear-link-scaling model 71 mean effective number 79 Shannon flow diversity 80–1 suspension feeders, seasonal relationship to stability 122, 123 total system throughput 79, 80 changes 39 relative biomasses 175, 177 upper triangularity 53 Swartkops Estuary, interecosystem relationship to body size 86, 87–9 comparisons 30–1, 32, 35,36 trophic links, flexibility 136–7 variables in ecological network 153 effect on complexity-stability analysis 27 symmetry of dendrograms 177 relationship 133 variables in LOICZ project 33 system ascendency see ascendency trophic plasticity 64 vertical zonation, Tropical Pacific 150 see also preference switching visualization of dynamics 36–9, 40 Tanypodinae trophic redundancy 64 sp., prey overlap ‘wasp-waist’ ecosystems 166–7 55 trophic species 73 graphs weak interactions, stabilising taxa, quantification 77 definition 70 in marine environment 101–2 effect 171–2 temporal comparisons in food weak interactors, effect of removal 28 Tropical Pacific (TP) 144, 146, webs 27–30, from food web 174–5 temporal variation in stream 167, 168 web size 73 food webs 60–1 community dynamics 158 effect on information metrics 84 Themisto El Nin˜o Southern Oscillation sp. 156 weevils, effect of low C:P ratio 15, 16 (ENSO) 146, 147–9 thresher sharks 153 weighted foodwebs, analysis 26 Thysanoessa sp. 156 effects on tuna 162–4, 163 food web structure 150–4, 151 weighted interactions 25 topological food webs 103 whales 153 topology models 125 Pacific inter-Decadal Oscillation (PDO) 147 in Barents Sea 157 total system throughput (TST) 27, in Gulf of Alaska 155 78 80 trout, brown 59, 62 , 79– see also trout-stocking, effects 195 killer whales definition 73 whitetip sharks 153 transfer coefficients 78–9 Tuesday Lake 44 tuna 152–3 Windermere, Lake, summer in hypothetical food web meteorological forcing 48 construction 76 effects of ENSO cycles 158, 162–4, 163 window of vitality 206 transfer efficiencies 31, 89, 90, 96 winter turnover of marine food webs 178 tri-trophic food chainsstudy in growth restriction of algae 45 Caribbean 20–4, 21, 23 meteorological forcing 47, 48, trophic cascades 42, 171, 172, 180 UCINET 26 49–50 effect of diversity 193 ultraviolet radiation increase, effect winterkill 48 in marine food webs 172–5 in lakes 41 trophic efficiency (TE) 9–10, Ultricularia sp., example of autocata- yellowfin tuna 152–3 27, 31 lysis 202 of pelagic food webs 11 uncommon species, significance 102 zooplankton see also productivity unexploited size spectrum slope in Barents Sea 156, 161 trophic interactions, relation to prediction 95–6 gelatinous 102, 152 biodiversity 185, 188–9 unique events 202 in Gulf of Alaska 154 trophic levels unstructured food web seasonal changes in lakes 45 in lakes 42–4 construction 74, 75–6 temperature effects 47, 49 in marine food webs 110 ascendency 83, 84 in tropical Pacific 152 number average mutual information zooxanthellae 175 determinants of 13 (AMI) 82 Zurich, Lake, cyanobacterial maximum number 205 developmental capacity (Cd) 81, 82 blooms 48