Global-Change Scenarios: Their Development and Use

University of Nebraska - Lincoln DigitalCommons@University of Nebraska - Lincoln

US Department of Energy Publications U.S. Department of Energy

7-2007

Edward Parson University of Michigan

Virginia Burkett U.S. Geological Survey

Karen Fisher-Vanden Dartmouth College

David Keith University of Calgary

Linda Mearns National Center for Atmospheric Research

See next page for additional authors

Follow this and additional works at: https://digitalcommons.unl.edu/usdoepub

Part of the Bioresource and Agricultural Engineering Commons

Parson, Edward; Burkett, Virginia; Fisher-Vanden, Karen; Keith, David; Mearns, Linda; Pitcher, Hugh; Rosenzweig, Cynthia; and Webster, Mort, "Global-Change Scenarios: Their Development and Use" (2007). US Department of Energy Publications. 7. https://digitalcommons.unl.edu/usdoepub/7

This Article is brought to you for free and open access by the U.S. Department of Energy at DigitalCommons@University of Nebraska - Lincoln. It has been accepted for inclusion in US Department of Energy Publications by an authorized administrator of DigitalCommons@University of Nebraska - Lincoln. Authors Edward Parson, Virginia Burkett, Karen Fisher-Vanden, David Keith, Linda Mearns, Hugh Pitcher, Cynthia Rosenzweig, and Mort Webster

This article is available at DigitalCommons@University of Nebraska - Lincoln: https://digitalcommons.unl.edu/ usdoepub/7 Global-Change Scenarios Their Development and Use

U.S. Climate Change Science Program Synthesis and Assessment Product 2.1b

July 2007 FEDERAL EXECUTIVE TEAM

Acting Director, Climate Change Science Program ...... William J. Brennan

Acting Director, Climate Change Science Program Office ...... Peter A. Schultz

Lead Agency Principal Representative to CCSP; Division Director, Department of Energy, Office of Biological and Environmental Research ...... Jerry W. Elwood

Product Lead; Department of Energy, Office of Biological and Environmental Research ...... John C. Houghton

Synthesis and Assessment Product Advisory Group Chair, Associate Director, EPA National Center for Environmental Assessment...... Michael W. Slimak

Synthesis and Assessment Product Coordinator, Climate Change Science Program Office ...... Fabien J.G. Laurier

OTHER AGENCY REPRESENTATIVES

Department of Agriculture ...... Jan Lewandrowski Environmental Protection Agency ...... Francisco de la Chesnaye National Aeronautics and Space Administration ...... Phil DeCola National Oceanic and Atmospheric Administration ...... Hiram Levy National Oceanic and Atmospheric Administration ...... Larry W. Horowitz

This document does not express any regulatory policies of the United States or any of its agencies, or provide recommendations for regulatory action. Further information on the process for preparing Synthesis and Assessment products and the CCSP itself can be found at www.climatescience.gov. Global-Change Scenarios Their Development and Use

Synthesis and Assessment Product 2.1b Report by the U.S. Climate Change Science Program and the Subcommittee on Global Change Research

AUTHORS: Edward A. Parson, University of Michigan Virginia R. Burkett, U.S. Geological Survey Karen Fisher-Vanden, Dartmouth College David W. Keith, University of Calgary Linda O. Mearns, National Center for Atmospheric Research Hugh M. Pitcher, Pacific Northwest National Laboratory Cynthia E. Rosenzweig, National Aeronautics and Space Administration Mort D. Webster, Massachusetts Institute of Technology CLIMATE CHANGE SCIENCE PROGRAM PRODUCT DEVELOPMENT ADVISORY COMMITTEE Eight members of the Climate Change Science Program Product Development Advisory Committee (CPDAC) wrote this Climate Change Science Program Synthesis and Assessment Product at the request of the Department of Energy. The entire CPDAC has accepted the contents of the product. Recommendations made in this report regarding pro- grammatic and organizational changes, and the adequacy of current budgets, reflect the judgment of the report’s authors and the CPDAC and are not necessarily the views of the U.S. Government.

Chair Designated Federal Officer Robert M. White, The Washington Advisory Group Anjuli S. Bamzai, Department of Energy Office of Biological and Environmental Research Vice Chair Soroosh Sorooshian, University of California-Irvine

Members

David C. Bader Isaac M. Held William A. Pizer Lawrence Livermore National Geophysical Fluid Dynamics Resources for the Future Laboratory Laboratory John M. Reilly *Virginia R. Burkett Henry D. Jacoby Massachusetts Institute of U.S. Geological Survey Massachusetts Institute of Technology Technology Antonio J. Busalacchi Richard G. Richels University of Maryland *David W. Keith Electric Power Research Institute University of Calgary Leon E. Clarke *Cynthia E. Rosenzweig Pacific Northwest National Kenneth E. Kunkel National Aeronautics and Space Laboratory Illinois State Water Survey Administration

Curtis C. Covey Richard S. Lindzen Robin T. Tokmakian Lawrence Livermore National Massachusetts Institute of Naval Postgraduate School Laboratory Technology *Mort D. Webster James A. Edmonds *Linda O. Mearns Massachusetts Institute of Pacific Northwest National National Center for Atmospheric Technology Laboratory Research Julie A. Winkler *Karen Fisher-Vanden Ronald L. Miller Michigan State University Dartmouth College National Aeronautics and Space Administration Gary W. Yohe Brian P. Flannery Wesleyan University Exxon-Mobil Corporation *Edward A. Parson University of Michigan Minghua H. Zhang William J. Gutowski Stony Brook University Iowa State University *Hugh M. Pitcher Pacific Northwest National David G. Hawkins Laboratory Natural Resources Defense Council *Authors for Global Change Scenarios: Their Development and Use July 2007

Members of Congress:

On behalf of the National Science and Technology Council, the U.S. Climate Change Science Pro- gram (CCSP) is pleased to transmit to the President and the Congress this report, Scenarios of Green- house Gas Emissions and Atmospheric Concentrations and Review of Integrated Scenario Development and Application. This is the second in a series of Synthesis and Assessment Products produced by the CCSP. This series of 21 reports is aimed at providing current evaluations of climate change science to inform public debate, policy, and operational decisions. These reports are also intended to help inform CCSP’s consideration of future program priorities. This second Synthesis and Assessment Product issued pursuant to Section 106 of the Global Change Research Act of 1990 and has two components: “Development of New Scenarios of Greenhouse Gas Emissions and Atmospheric Concentrations” (Part A) and a “Review of Integrated Scenario Development and Application” (Part B). Here we transmit to you Part B.

CCSP’s guiding vision is to provide the Nation and the global community with the science-based knowledge to manage the risks and opportunities of change in the climate and related environmental systems. The Synthesis and Assessment Products are important steps toward that vision, helping translate CCSP’s extensive observational and research base into informational tools directly address- ing key questions that are being asked of the research community.

This product will contribute to and enhance the ongoing and iterative international process of producing and refining climate-related scenarios and scenario tools. It was developed with broad scientific input and in accordance with the Guidelines for Producing CCSP Synthesis and Assessment Products, Section 515 of the Treasury and General Government Appropriations Act for Fiscal Year 2001 (Public Law 106-554), and the Information Quality Act guidelines issued by the Department of Energy pursuant to Section 515. The CCSP Interagency Committee relies on Department of Energy certifica- tions regarding compliance with Section 515 and the Guidelines for Producing CCSP Synthesis and Assessment Products.

We commend the report’s authors for both the thorough nature of their work and their adherence to an inclusive review process.

Samuel W. Bodman Carlos M. Gutierrez John H. Marburger III, Ph.D. Secretary of Energy Secretary of Commerce Director, Office of Science and Technology Policy Chair, Committee on Vice-Chair, Committee on Executive Director, Committee on Climate Change Climate Change Climate Change Science and Technology Integration Science and Technology Integration Science and Technology Integration ACKNOWLEDGEMENT This report has been peer reviewed in draft form by individuals chosen for their diverse perspectives and technical expertise. The expert review and selection of reviewers followed the OMB’s Information Quality Bulletin for Peer Review. The purpose of this independent review is to provide candid and critical comments that will assist the Climate Change Science Program in making this published report as sound as possible and to ensure that the report meets institutional standards. The peer review comments, draft manuscript, and response to the peer review comments are publicly available at: www.climatescience.gov/Library/sap/sap2-1/default.php.

We wish to thank the following individuals for their peer review of this report: Frans Berkhout, Free University of Amsterdam Garry Brewer, Yale University Ged Davis, World Economic Forum Nebojsa Nakicenovic, International Institute for Applied Systems Analysis Robert Lempert, RAND Corporation M. Granger Morgan, Carnegie-Mellon University John Robinson, University of British Columbia Thomas J. Wilbanks, Oak Ridge National Laboratory

We would also like to thank the numerous individuals who provided their comments during the public comment period. The public review comments, draft manuscript, and response to the public comments are publicly available at: www.climatescience.gov/Library/sap/sap2-1/default.php.

The writing team is grateful for research assistance by Kristen Cleary, Emily Kelly, Paul Porter, Gautam Rao, and Nora Van Horssen.

Editor ...... Elizabeth L. Malone Technical Advisor ...... David Dokken Graphic Production ...... DesignConcept

Recommended Citations The Entire Volume - CCSP Synthesis and Assessment Product 2.1 CCSP, 2007: Scenarios of Greenhouse Gas Emissions and Atmospheric Concentrations (Part A) and Review of Inte- grated Scenario Development and Application (Part B). A Report by the U.S. Climate Change Science Program and the Subcommittee on Global Change Research [Clarke, L., J. Edmonds, J. Jacoby, H. Pitcher, J. Reilly, R. Richels, E. Par- son, V. Burkett, K. Fisher-Vanden, D. Keith, L. Mearns, C. Rosenzweig, M. Webster (Authors)]. Department of Energy, Office of Biological & Environmental Research, Washington, DC., USA, 260 pp. This Sub-Report (2.1B) Parson, E., V. Burkett, K. Fisher-Vanden, D. Keith, L. Mearns, H. Pitcher, C. Rosenzweig, M. Webster, 2007. Global Change Scenarios: Their Development and Use. Sub-report 2.1B of Synthesis and Assessment Product 2.1 by the U.S. Climate Change Science Program and the Subcommittee on Global Change Research. Department of Energy, Office of Biological & Environmental Research, Washington, DC., USA, 106 pp. The Companion Sub-Report (2.1A) Clarke, L., J. Edmonds, H. Jacoby, H. Pitcher, J. Reilly, R. Richels, 2007. Scenarios of Greenhouse Gas Emissions and Atmospheric Concentrations. Sub-report 2.1A of Synthesis and Assessment Product 2.1 by the U.S. Climate Change Science Program and the Subcommittee on Global Change Research. Department of Energy, Office of Biological & Environmental Research, Washington, DC., USA, 154 pp. TABLE OF CONTENTS Executive Summary References Acronyms C 5 C 4 Review ofMajorClimate-ChangeScenarioExercises 3 Scenarios inGlobal-Change Analysis andDecisionSupport 2 Scenarios, theirCharacteristicsandUses 1 SECTIONS Introduction onclusions andRecommendations hallenges andControversies inScenariosfor ClimateChange ...... 83 ...... 51 ...... 31 ...... 21 ...... 13 ...... Global-Change Scenarios-TheirDevelopment andUse 97 95 9 1 The U.S. Climate Change Science Program Global-Change Scenarios - Their Development and Use

A scenario is a description of potential future conditions produced to inform decision-making under uncertainty. Scenarios can help inform decisions that involve high stakes and poorly characterized uncertainty, which may thwart other, conventional forms of analysis or decision support. Orig- inally developed to study military and security problems, scenarios are now widely used for strategic planning and assessment in businesses and other organizations, and increasingly to inform planning, analysis, and decision-

EXECUTIVE SUMMARY making for environmental issues, including climate change.

Scenarios can serve many purposes. They can help inform specific decisions, or can provide inputs to assessments, models, or other decision-support activities when these activities need specification of potential future conditions. They can also provide various forms of indirect decision support, such as clarifying an issue’s importance, framing a decision agenda, shaking up habitual thinking, stimulating creativity, clarifying points of agreement and disagreement, identifying and engaging needed participants, or providing a structure for analysis of potential future decisions.

SCENARIOS FOR CLIMATE CHANGE: FIVE TYPES

Developing a scenario exercise involves many design choices, of which the most important involve choosing the few key uncertainties to represent in alternative scenarios. Five types of scenarios have been developed to address different aspects of the climate-change issue; these are distinguished by where they fall along a simple linear causal chain extending from the socio-economic determinants of greenhouse-gas emissions through the impacts of climate change as shown in Figure ES-1. (This figure does not represent the complete causal structure of the climate issue, which has many linkages and feedbacks. Rather, this simple structure only illustrates how scenarios have been used to fit within the simplest and most prominent causal pathway of the issue.)

SOCIO- ATMOSPHERE Figure ES-1. ECONOMIC EMISSIONS IMPACTS & CLIMATE CONDITIONS Scenarios of • greenhouse gases • sea level rise • carbon cycle • population • aerosols • fresh water anthropogenic • temperature • GDP • other drivers, • ecosystems • precipitation climate change: • energy system e.g. land-cover • agriculture • humidity • industry change • human health simple linear causal • soil moisture • technology • property, chain • agriculture, • extreme events infrastructure land-use

1 The U.S. Climate Change Science Program Executive Summary

Emissions Scenarios for Climate Simulations: trends. The factors that influence particular Emissions scenarios present future paths of impacts and vulnerabilities are likely to greenhouse-gas emissions or other climate be widely variable, and may include perturbations. A major use of these is to demographic, economic, technological, provide needed inputs to climate models. institutional, and cultural characteristics. Such scenarios may specify simple arbitrary Consequently, scenarios may have to be gen- perturbations of emissions or concentrations erated in an exploratory manner in the con-

(e.g., doubling atmospheric CO2), or time- text of attempting to assess specific local paths reflecting specified assumptions for and regional impacts. evolution of socio-economic drivers such as population, economic growth, and technological change. SCENARIOS FOR CLIMATE CHANGE: Emissions Scenarios for Exploring Alternative MAJOR EXAMPLES Energy/Technology Futures: Another use of emissions scenarios involves specifying The report reviews four major exercises pro- an environmental or emissions target, arbi- ducing or using scenarios for climate-change trarily or based on normative or political applications. The examples include national goals, to examine what patterns of socio- and international activities, produced by differ- economic change, energy resources, and ent sets of actors for different purposes. technology development are consistent with the target and/or what interventions might The Intergovernmental Panel on Climate Change be needed to meet it. Such scenarios have (IPCC) has produced three sets of scenarios of examined conditions for stabilizing atmos- 21st-century greenhouse-gas emissions, of which pheric CO2 concentration at various levels the most ambitious and important were pro- and the implications of stabilizing radiative duced for the Special Report on Emissions Sce- forcing for multi-gas reduction strategies. narios (SRES) between 1997 and 1999. SRES Climate-Change Scenarios: Climate scenarios produced four qualitative storylines on which specify potential future climate conditions six “marker” scenarios were based – one model to inform assessments of impacts, vulnera- quantification of each storyline plus two tech- bilities, and adaptation options, and inform nological variants of one storyline that stressed decision-making for adaptation or mitigation. fossil-intensive and low-carbon energy supply They can be produced by arbitrary perturba- technologies – each produced by a different tion of present conditions, by using climates energy-economic model. Other models’ repli- from elsewhere or the past as a proxy for po- cations of each other’s marker scenarios plus a tential future climate in a given location, or few additional explorations yielded 40 scenarios by climate-model simulations driven by some in total. These scenarios highlighted several in- specified scenario of future emissions. sights, including the ability of alternative paths with similar emissions in 2100 to differ widely Scenarios of Direct Biophysical Impacts, e.g., in their interim pathways and thus in atmos- Sea Level Rise: Scenarios can specify al- pheric concentrations; the ability of alternative ternative trajectories for some important technological assumptions alone to generate as form of climate impact that influences many wide a range of emissions futures as substan- other impacts. For example, scenarios of sea tially divergent socio-economic pathways; and level rise can capture the most important im- the fact that similar emissions paths can come pact pathways in many coastal regions, in- from widely different combinations of underly- cluding the large uncertainties associated ing socio-economic factors and so pose distinct with potential loss of continental ice sheets mitigation problems. A widely publicized cri- in Greenland and Antarctica. tique of the SRES scenarios alleged over-esti- Multivariate Scenarios for Impact Assessment: mation of future emissions growth due to the Assessing climate-change impacts requires metric used to compare incomes in rich and not just considering climate in isolation, but poor nations, but the overestimation was later other linked changes and stresses, including found to be insignificant. More serious and il- both environmental and socio-economic luminating challenges associated with these

2 Global-Change Scenarios - Their Development and Use scenarios concerned how to balance and inte- authority over separate assessments. The re- grate qualitative and quantitative scenarios; how liance on climate scenarios from just one fam- to use and how much to coordinate multiple ily of climate models may pose risks of models to generate the most useful insights; and incomplete representation of key uncertainties. whether, when, and how it is appropriate to assign explicit probability judgments to alterna- The Millennium Ecosystem Assessment (MEA) tive scenarios or associated ranges of examined the status, present trends, and longer- quantitative variables. term challenges to the world’s ecosystems, including climate change and other stresses. One The US National Assessment was a compre- of the assessment’s four working groups con- hensive assessment of potential impacts of cli- structed scenarios of global ecosystems to 2050 mate change and variability on the United and beyond, largely independently of the group States, focusing on major regions and sectors examining current status and trends. All as- (agriculture, water, human health, coastal areas sessment components used a common concep- and marine resources, and forests). The Na- tual framework, which distinguished indirect tional Assessment needed scenarios of 21st-cen- drivers of ecosystem change (e.g., population tury US climate and socio-economic changes. and economic growth, technological change, For climate scenarios, it relied principally on policies and lifestyles), direct drivers (e.g., cli- climate-model scenarios produced by the UK mate change, air pollution, and land-use and Hadley Centre and the Canadian Centre for Cli- land-cover change), ecosystem indicators, mate Modeling and Analysis, each driven by a ecosystem services, measures of human well- single emissions scenario, with statistical down- being, and response options. The Scenarios scaling based on detailed local conditions and group applied this framework to characterizing present patterns of fine-scale climate variation. potential ecosystem stresses in 2050, with more Other proposed types of climate scenario, in- limited projections to 2100. The four scenarios cluding historical scenarios and inverse meth- were based on two dimensions of uncertainty: ods to probe for key vulnerabilities, were less degree of globalization, and predominance of used. For socio-economic scenarios, a novel ap- proactive vs. reactive response to ecosystem proach was proposed that combined specified stresses. The qualitative storylines underlying scenarios for a few key national-level variables these scenarios were more richly developed than such as population and economic growth, and a in other climate-change scenario exercises. common process to elaborate and document ad- Concerns with these scenarios pertained to the ditional socio-economic assumptions as needed degree of integration and consistency among for specific regional and sector analyses. The qualitative and quantitative scenario compo- National Assessment was criticized for relying nents; risks of logical circularity within scenar- on just two climate-model runs and one emis- ios; and unexplained similarity of projected sions scenario, although these choices were dic- ecosystem effects among scenarios. tated by time limits and availability of climate-model runs. Limited use was made of SCENARIOS the socio-economic approach, principally due FOR CLIMATE CHANGE: to time limits and communication problems. CHALLENGES AND CONTROVERSIES The UK Climate Impacts Programme (UKCIP) provides common datasets, tools, and support, Scenarios and Decisions including scenarios, for climate-impact assessments for UK regions and sectors by re- Scenarios can inform climate-change mitigation searchers and stakeholders. The program and adaptation decisions, but most uses so far produced climate scenarios in 1998 and 2002, have had relatively indirect connections to such all based on the Hadley Centre climate models, decisions. Although there is no single global and socio-economic scenarios in 2001. The climate-change decision-maker, scenarios can program stresses building a sustained assess- inform the many decision-makers with diverse ment capability by acting as a motivator, re- responsibilities that will affect and be affected source, and light coordinator with little central by climate change. Three groups of decision-

3 The U.S. Climate Change Science Program Executive Summary

makers with distinct information needs can be bearing on high-stakes policy decisions, but such distinguished: mitigation policy-makers, who attempts to restrict scenarios should be resisted, are mostly but not exclusively national officials; principally through prominent communication of impacts and adaptation managers, including na- the reasoning, assumptions, and treatment of par- tional officials and others who are responsible ticular uncertainties underlying scenarios. for particular climate-sensitive assets, resources, or interests; and energy resource and Scenario Development Process: technology managers, who include owners, de- Expert-Stakeholder Interactions velopers, and investors in energy resources and energy-related capital stock and new technologies. Scenario developers must decide how and how much to involve scenario users and stakeholders A key issue in creating scenarios for all in scenario development. In other fields – where decision-makers is how to represent decisions users are clearly identified – relatively few and within scenarios. In general, decisions by the homogeneous, intensive collaboration between scenario user should be explicitly examined rel- scenario developers and users or their represen- ative to baseline conditions specified in scenar- tatives is desirable. Close user involvement is ios, while decisions by others outside their also advantageous in developing scenarios for control should be treated like any exogenous un- climate change, but potential users of these sce- certainty. The issue is most important in the narios are more numerous and diverse, may not treatment of mitigation decisions: scenarios to be clearly identified, and may have contending inform mitigation should allow explicit exami- material interests in the scenarios’ content or nation of the entire relevant range of mitigation use. This situation calls for delicate decisions decisions, while scenarios to inform impacts about participation and representation to keep and adaptation should specify the likely range scenarios tuned to practical users’ needs while of mitigation efforts – usually yielding a nar- keeping the development process small enough rower range of emissions futures than is con- to be manageable. sidered in scenarios to inform mitigation. Communication of Scenarios Scenarios in Assessments and Policy Debates Climate change scenarios must be communicated to multiple audiences with diverse inter- In climate-change assessments, scenarios can ests and information needs. In addition to the provide required inputs to other parts of the scenarios’ content, sufficient information must analysis and help to organize multiple compo- be provided about the process and reasoning by nents of the assessment. When scenarios are which the scenarios were developed, to allow used in a prominent assessment, they may sub- users to scrutinize the underlying data, models, sequently be adopted in planning or decision- and reasoning; judge their confidence in the support processes outside the original scenarios; and have opportunities to critique the assessment. Scenarios can also help frame pub- scenarios and suggest alternative approaches. lic and policy debate, in part by providing an ag- Effective communication can help engage gregate metric of the issue’s severity. They a broad user community in updating and im- consequently may gain prominence in con- proving scenarios. Open communication of the tentious policy debates, and so become subject decisions, assumptions, and uncertainties un- to political attempts to influence their content and derlying scenarios is likely to both increase political criticism based on their perceived im- users’ confidence that the scenarios have rea- plications for policy action. The unavoidable sonably represented current knowledge and key judgments underlying construction of scenarios uncertainties, and help them develop alterna- provide opportunity for partisan efforts to make tives if they are unconvinced. scenarios policy prescriptive, and for claims that only certain scenarios are plausible (e.g., high- or low-emissions scenarios, depending on the critic’s motivation). These claims are unavoidable, since scenarios represent key uncertainties

4 Global-Change Scenarios - Their Development and Use

Consistency and Integration nent controversy in treatment of uncertainty in in Scenarios scenarios has concerned whether or not to explicitly assign probabilities to scenarios or as- Scenario developers should strive for internal sociated ranges of quantitative outcome consistency. At one level, this means avoiding variables. The debate rests in part on different clear contradictions with well-established views of the typical contents of scenarios, since knowledge and not moving inadvertently out- subjective probabilities can readily be assigned side bounds of historical experience – although to ranges of one or two quantitative variables. such sharp departures from experience may be Explicit probability assignment in such simple useful if pursued intentionally to examine low- cases offers clear benefits for assessing alterna- probability risks or broaden decision-makers’ tive choices and avoids the risk of users assign- perceptions. Perceptions of internal consistency ing their own, perhaps less informed, or coherence in scenarios ultimately rest on sub- probability judgments. Assigning probabilities jective judgments, which pose well-known risks to rich multivariate scenarios, particularly if of bias if not carefully structured and controlled. these include narrative elements, is much more Potential inconsistencies grow when scenario problematic, since there is no clearly defined in- exercises use multiple models and attempt to terval “between” such scenarios and their harmonize them, particularly when some key boundaries are not clearly defined. quantities are externally specified for some models and calculated within others. Attempt- CONCLUSIONS AND ing to avoid such inconsistency by standardizing RECOMMENDATIONS model outputs, however, can carry more serious risks by obscuring interpretation of results and Use of Scenarios precluding use of model variation to illuminate in Climate-Change Decisions uncertainty. Attempts to connect qualitative and • Scenarios can make valuable contributions quantitative aspects of scenarios have been par- to climate-change decision-making. There ticularly challenging for pursuit of consistency. is a big gap between the use of scenarios in Different narrative scenarios often reflect dif- current practice and their potential contri- ferent assumptions about how the world works, butions, but interest in using scenarios is in- which correspond more closely to different creasing. model structures than to parameter variation. Better integrating the two approaches will re- • Scenarios of global emissions and resultant quire developing ways to connect narrative sce- climate change are required by many diverse narios to model structures, rather than merely to climate-related decision-makers, but beyond target values for a few variables that models are these common requirements decision-makers’ then asked to reproduce. needs from climate-change scenarios are highly diverse. Treatment of Uncertainty • Impacts and adaptation managers include in Scenarios both national officials and others responsible for more specific domains of impact. They A scenario exercise can represent a few key un- need climate-change scenarios, driven by certainties by variation among scenarios. Ex- specified global emissions scenarios, to rep- treme economy is required in choosing which resent potential climate-related stresses on uncertainties to represent, what variation (in- their areas of responsibility, plus other envi- cluding potential extremes) to represent for each, ronmental and socio-economic information and how to combine them in a manageable num- at appropriate scales. Their combined needs ber of scenarios. Complex narrative scenarios – for centrally produced climate scenario in- pose special problems in representing and com- formation, associated tools and support, and municating uncertainty, usually addressed by a capability to develop and apply additional seeking underlying structural uncertainties – scenario information related to their respon- e.g., deep societal trends such as globalization sibilities – suggest the need for a cross-scale or values shifts – that are judged to influence organizational structure to provide scenario many other factors of concern. The most promi- information.

5 The U.S. Climate Change Science Program Executive Summary

• Mitigation policy-makers, who are mainly bias in scenario production and focus subse- but not exclusively national officials, need quent argument on underlying uncertainties. scenarios of global and national emissions trends, resultant climate change, and aggre- What Should Centrally Provided gate impacts. In addition, they need scenario Emissions and Climate Scenarios information about the potential policy envi- Look Like? ronment for their choices, including alterna- • Centrally provided scenarios of emissions tive scenarios of other nations’ mitigation and resultant climate change should be strategies, international mitigation decisions, global in scope, with major climate-relevant and implementation and compliance. In emissions and other perturbations specified some cases, they can usefully employ target- at least for major world regions. They driven scenarios for backcasting analysis. should have a time horizon of a century or Mitigation decisions require scenario devel- longer, with interim results at roughly opment capacity at the national level. decadal resolution. • Scenarios for mitigation decisions should in- • Centrally provided scenarios of global emis- clude a wide range of baseline emissions as- sions and climate change can help inform sumptions and should not pre-judge the mitigation and adaptation decisions at na- likely level of mitigation effort, while sce- tional and sub-national scales, but such de- narios for impact and adaptation managers cisions require additional information at should be based on emissions assumptions these scales. that include the range of mitigation interventions they judge likely. • Emissions scenarios of several types are needed to serve diverse uses, including al- • Energy resource and technology managers, ternative baselines, alternative levels of in- who are mainly private-sector actors, prima- cremental stringency of mitigation effort, rily need scenarios that represent alternative and specified future targets to support back- policy regimes over the 30- to 50-year time casting and feasibility analysis. Some emis- horizons relevant for investment and tech- sions scenarios should be coupled to explicit nology-development decisions. Scenarios scenarios of wide-ranging alternative socio- of emissions and climate change may pro- economic futures, but this is not necessary vide background, but do not capture the for all uses. Scenarios should reflect vari- most important uncertainties for these ous explicit degrees of coordination, includ- decision-makers. ing simple fully standardized scenarios for evaluating and comparing downstream mod- Use of Scenarios els, multi-model scenarios using common in Climate-Change Assessments input assumptions, and non-standardized • Large-scale, official assessments are cur- scenarios to explore alternative assumptions rently the main users of scenarios and will or meet specific user needs. likely remain major users. Scenarios in as- • Some scenarios of socio-economic condi- sessments mostly support further analysis, tions should include qualitative and quanti- modeling, and assessment. They can also tative elements and sustained analytic efforts help frame the climate issue for the public to link the two. These elements can provide and policy-makers. Presentation of scenar- a vehicle to explore major historical uncer- ios in assessments leads to additional un- tainties with large implications for climate foreseen uses. change and vulnerability; provide a logical • Scenarios contain unavoidable elements of structure to connect assumed trajectories for judgment in their production and use. This multiple variables; and provide guidance to makes them vulnerable both to attempts at other analysts or users to extend scenarios by bias and to partisan attack. The most produc- elaborating additional detail. Alternative qual- tive response lies in transparency about the itative and narrative elements should be linked process, reasoning, and assumptions used to not just to alternative parameter values in produce scenarios, which can both help limit quantitative models, but also to alternative forms of causal relations and model struc- 6 Global-Change Scenarios - Their Development and Use

tures.

Scenario Process: Treatment of Uncertainty Developer-User Interactions in Scenarios • There is value in close collaboration be- • More explicit characterization of probabil- tween scenario developers and users, partic- ity judgments should be included in some ularly at the beginning and ending stages of future scenario exercises than has been prac- a scenario exercise. ticed so far. Means available to express these judgments are of widely varying speci- • The ease of achieving such collaboration ficity, ranging from agreed terminology to and its value are likely to be greater when explicitly quantified probability distribu- scenario users are clearly identified, few in tions. All such judgments should include number, and similar in their interests explicit acknowledgement of their inevitable and perspectives. subjective elements and appropriate caveats. Communication of Scenarios • Explicit probability judgments are easiest to produce and least controversial in scenarios • Effective communication of scenarios is es- generated using quantitative models of cli- sential, in forms useful to audiences of di- mate change or specific impact domains. verse interests and technical skills. In These can be conditioned on specific as- addition to scenario contents, communica- sumptions for socio-economic inputs such as tion should include associated documenta- emissions, and can represent explicitly and tion, tools, and support. quantitatively the effects of specified varia- • Transparency of underlying reasoning, as- tion in initial conditions or unknown param- sumptions, and major uncertainties is cru- eter values. These devices are also available, cial. Such transparency is necessary to although in less widespread use, in economic support the credibility of scenarios, to alert models used to project emissions. potential users to conditions under which • Including explicit probability judgments is they might wish to use or modify them, and likely to be most useful when key variables to inform criticism and improvement are few, quantitative outcomes are needed, of scenarios. and potential users are numerous and diverse. It is likely to be least useful when sce- Consistency and Integration narios specify multiple characteristics, in Scenarios including prominent qualitative elements; • Any scenario should be internally consistent when the purpose is sensitivity analysis or in its assumptions and reasoning, to the ex- heuristic exploration; and when potential tent this can be established. users are few, similar, and known. • In scenario exercises that use multiple mod- • Because of their large and diverse set of po- els to explore potential uncertainties in fu- tential users, centrally provided scenarios of ture conditions, consistency among models global emissions and climate change should should be pursued primarily through coor- attempt to include some explicit probability dination of inputs, not outputs, except when judgments for ranges of key quantitative coordinated outputs represent common outputs, including global emissions and goals for policy evaluation. global-average temperature change. These should span a wide range of judged uncer- • Transparency in reporting scenario and tainty on these variables, e.g., 95 to 99 per- model differences as well as underlying as- cent. Providing such explicit likelihood sumptions and reasoning can help mitigate statements lets users choose whether to use the effects of inconsistencies among scenarios. them or not. • Scenario exercises should give more attention to low-probability, high-consequence extreme cases, such as loss of a major con-

7 The U.S. Climate Change Science Program Executive Summary

tinental ice sheet or changes in meridional expertise, and analytic and modeling ca- ocean circulation. With these, it is especially pability. crucial to be explicit and transparent about • The program should integrate and bal- the reasoning and assumptions underlying ance goals and criteria related to scien- each scenario, including developers’ judg- tific and technical quality, and those ments of relative likelihoods. related to utility and relevance to users. • The program should be insulated from Expanding and Sustaining Capacity political control. for Production and Use of Scenarios • The program should strive for maximum • Present scenario capacity is inadequate. To transparency in its own activities, in ad- help fulfill these presently unmet needs, the dition to demanding it from activities it CCSP should establish a program to: supports. • Commission scenarios for use in assess- • The program will require the authority ments and decision-support activities. and resources necessary to articulate and • Disseminate scenarios with associated promulgate standards for transparency, documentation, tools, and guidance consistency, and quality control. materials. • The program will require adequate sus- • Commission various groups to evaluate tained resources level of effort. scenarios and their applications, and to develop improved methods. • Archive results and documentation related to all these tasks, to provide historical perspective and institutional memory for future scenario-related activities. • Design and management conditions of this new program should include six elements. • The program should build and maintain strong connections with outside relevant

8 Global-Change Scenarios - Their Development and Use

Introduction INTRODUCTION

This report examines the development and use of scenarios in global climate change applications. It considers scenarios of various types – including but not limited to emissions scenarios – and reviews how they have been developed, what uses they have served, what consistent challenges they have faced, what controversies they have raised, and how their development and use might be made more effective. The report is Synthesis & Assessment Product 2.1b of the US Climate Change Science Program. By synthesizing available literature and critically reviewing past experience, the report seeks to assist those who may be conducting, using, or commissioning scenarios related to global climate change.

Scenarios are used to support planning and decision-making when issues have deep or poorly characterized uncertainty and high stakes, often accompanied by long time horizons. These conditions apply to the major decisions about how to respond to global climate change. As scientific research advances our knowledge of the climate’s present state and trends, its patterns of variability, and its responses to external forcings, we are gaining an increasingly clear view of risks that may be realized late this century or beyond. These future risks are linked to near-term socio-economic trends and decisions in both public and private sectors. Some near-term decisions – such as investment in long-lived capital equipment, new resources, or new technologies in the energy sector – can influence long-term trends in the emissions contributing to climate change. Other near-term decisions – such as investment in water resources infrastructure or coastal development – can influence how adaptable and how vulnerable future society will be to the impacts of climate change.

9 The U.S. Climate Change Science Program Introduction

Although such decisions are being made now, be adequately represented as wider distributions making them responsibly requires considering of recognized uncertain quantities. In others, their potential consequences over the longer they may represent events whose character or term, including associated uncertainties. This even possibility we have not yet imagined. requires thinking about the future conditions that will shape their consequences, not just next Despite pervasive uncertainties, people must month or next year but 10, 30, 50, or 100 years make near-term decisions related to climate in the future – longer periods than the horizon of change that have long-term consequences, in- conventional methods of planning or analysis.2 cluding potential irreversibilities. Scenarios are Attempting to describe potential future condi- tools to help inform these decisions by gathering tions over this long time horizon presents a and organizing relevant knowledge, organizing seeming paradox. On the one hand, conditions associated uncertainties, and structuring and dis- Despite pervasive this far in the future, and the factors and actors ciplining associated speculation. This report as- that may influence them, are deeply uncertain.3 sesses experience to date in developing and using uncertainties, people On the other hand, we have a great deal of scenarios for global climate change. must make near-term knowledge that can help make informed as- decisions related to sumptions about future conditions, even over Early climate-change debates mainly concerned climate change that such long horizons. This includes well-established scientific questions such as whether and how have long-term scientific knowledge about physical, chemical, the climate is changing, how much change is consequences, and biological processes; more weakly, rela- caused by human activities, and how sensitive including potential tively well-established causal mechanisms in the climate system is. Scenarios did not figure irreversibilities. economics, sociology, and politics; and more prominently in these debates. But as advancing weakly still, certain seemingly robust empirical climate science has increasingly shifted the de- patterns of historical change in population, eco- bate from confirming and describing the cli- nomics, and technology. All of these give some mate-change problem toward deciding what to guidance to support judgments about future do about it, the need for long-term decision-sup- conditions that are more or less likely, virtually port tools like scenarios has increased, as have certain, or virtually impossible. In some ways the scrutiny and criticism these have attracted.4 we might be highly confident that the future will In a contentious public-policy area like climate resemble the present, e.g., in the radiative prop- change, controversy over scenarios is to be ex- erties of atmospheric trace gases. In others, we pected: scenarios are a method to structure and might judge it likely that future conditions will communicate the most important uncertainties, lie within some envelope extrapolated from and conflicting judgments about uncertainties present and past trends, e.g., in projecting rates are a major source of disagreements over what of change in fertility, mortality, or labor pro- to do. Consequently, we expect the trend of sce- ductivity. Still other areas, such as the develop- narios’ increasing prominence and contentious- ment and social consequences of major ness to continue – particularly for emissions technological advances, or large-scale political scenarios, since these are the relevant metric of events such as wars, political realignments, or human environmental burden and the point of epidemics, may hold more fundamental uncer- most contested proposed intervention. tainties. In some cases, such uncertainties may

2 Morgan et al. 1998. 4 See, e.g., Lomborg 2001; Michaels 2003b; Castles 3 Lempert et al. 2006. and Henderson 2003a, b; UK House of Lords 2005.

10 Global-Change Scenarios - Their Development and Use

In this report, we try to cast some light on cur- Our review of this experience has not been en- rent and coming debates over climate-change tirely independent, since members of this writ- scenarios. These debates presently exhibit basic ing team were involved as participants, confusion about the definition, purposes, and reviewers, and critics in two of the scenario ex- potential uses of scenarios. We aim to provide ercises we review, the IPCC SRES process and clarification and practical advice to two related the US National Assessment. While we have audiences: those conducting assessments or drawn on the experience of these team mem- analyses that develop or use scenarios; and bers, we have attempted to limit the risk of idio- those commissioning, using, and interpreting syncratic interpretations and bias by drawing on such assessments or analyses. For the first other sources as well and by engaging all team group, we seek to provide an organized sum- members in developing our summary and dis- mary of relevant experience in similar efforts, cussions of these exercises. Moreover, our pur- discussion and clarification of key choices and pose is not to either attack or defend these past We aim to provide challenges, and – to the extent present knowl- exercises, but to seek to understand the deci- clarification and edge allows – practical guidance about pitfalls, sions made in conducting them, the factors that practical advice to two challenges, and opportunities in particular ap- influenced them, and the constraints under related audiences: proaches. For the second group, we seek to pro- which they operated, in order to assess their ex- vide guidance on what to ask for, how and how perience, identify successes and pitfalls, and to those conducting much to participate in its production, how to in- the extent possible, provide guidance to advance assessments or terpret the results, and what questions to ask. scenario methods for climate change and other analyses that use similar environmental issues. Because the expe- scenarios; and those Because the charge of this report is unlike those rience we review does not provide a sufficiently commissioning, using, of other Synthesis and Assessment products, the large or random sample to support strong scien- and interpreting such approach we have taken to producing it is nec- tific inference, the diagnoses, interpretations, and assessments or essarily different as well. We were tasked with recommendations we present rely on our collec- analyses. reviewing, interpreting, and evaluating experi- tive judgment. We have endeavored to follow our ence with scenario methods in global climate own advice, and be as transparent as possible change applications. This is not a narrowly fo- about the foundation and reasoning underlying cused question, and there is not a well-devel- our conclusions and recommendations. oped scientific literature on which we can draw for answers. While we have reviewed the exist- The report is organized as follows. Drawing on ing literature on scenarios, most of it concerns the broader literature on scenarios – most of scenarios in other decision domains than global which concerns domains other than climate climate change. In addition, we have examined change – Section 1 introduces scenarios, sharp- several major scenario exercises in global-change ens their definition, and outlines a few major di- applications. In this, we have drawn on published mensions of variation and decisions that must materials, both from the exercises themselves and be made in developing a scenario exercise. Sec- from commentary and criticism, as well as doc- tion 2 focuses specifically on scenarios for umentary materials and records, interviews with global climate change, and outlines the types of participants and users, and the experience and decisions that could use scenarios and the main judgments of team members. types of scenarios that have been developed for this issue. Section 3 reviews four major expe-

11 The U.S. Climate Change Science Program Introduction

riences in developing and using global-change scenarios. Section 4 discusses several issues that have posed key challenges in climate-change scenarios and that are likely to require particular attention in designing new scenario exercises. In addition to drawing on Section 3, this discussion also makes use of briefer discussions of eight other examples of global-change scenarios that illustrate particular issues or challenges; these examples are presented as short boxes in Section 4. Section 5 provides our conclusions and recommendations for future development and use of global climate-change scenarios.

12 Global-Change Scenarios - Their Development and Use

Scenarios, 1 Their Characteristics and Uses SECTION

1.1 DEFINING SCENARIOS

A scenario is a description of potential future conditions, developed to inform decision-making under uncertainty. The decisions in question can be faced by individuals, groups, organizations, or governments, and may pertain to any subject matter. While many writers on scenarios give no explicit definition, others have offered a wide range of definitions, often substantially more complex and restrictive than this one. The published definitions gathered in Box 1.1 give a sense of both the broad commonalities among many analysts’ conceptions and the significant differences among them.

BOX 1.1 Scenarios: a Sampling of Published Definitions.

A scenario is a coherent, internally consistent, and plausible description of a possible future state of the world.5

A scenario is a story that describes a possible future. It identifies some significant events, the main actors and their motivations, and it conveys how the world functions. Building and using scenarios can help people explore what the future might look like and the likely challenges of living in it.6

Scenarios are images of the future, or alternative futures. They are neither predictions nor forecasts. Rather, each scenario is one alternative image of how the future might unfold. A set of scenarios as- sists in the understanding of possible future developments of complex systems. Some systems, those that are well understood and for which complete information is available, can be modeled with some certainty, as is frequently the case in the physical sciences, and their future states predicted. However, many physical and social systems are poorly understood, and information on the relevant variables is so incomplete that they can be appreciated only through intuition and are best communicated by images and stories. Prediction is not possible in such cases.7

continued on next page

5 IPCC 2001b:149. 6 Shell International 2003. 7 Nakicenovic and Swart 2000:62.

13 The U.S. Climate Change Science Program Section 1 - Scenarios, Their Characteristics and Uses

BOX 1.1, continued from previous page.

A climate scenario is a plausible representation of future climate that has been constructed for explicit use in investigating the potential impacts of anthropogenic climate change. Climate scenarios often make use of climate projections (descriptions of the modeled response of the climate system to scenarios of greenhouse gas and aerosol concentrations), by manipulating model outputs and combining them with observed climate data.8

[Scenarios] are created as internally consistent and challenging descriptions of possible futures. They are intended to be representative of the ranges of possible future developments and outcomes in the external world. What happens in them is essentially outside our own control.9

Scenarios are coherent, internally consistent and plausible descriptions of possible future states of the world, used to inform future trends, potential decisions, or consequences. They can be considered as a convenient way of visioning a range of possible futures, constructing worlds outside the normal timespans and processes covering the public policy environment.10

Scenarios are plausible, challenging, and relevant sets of stories about how the future might unfold. They are generally developed to help decision-makers understand the wide range of potential futures, confront critical uncertainties, and understand how decisions made now may play out in the future. They are intended to widen perspectives and illuminate key issues that might otherwise be missed or dismissed. The goal of developing scenarios is often to support more informed and rational decision-making that takes both the known and the unknown into account.11

The historical uses of scenarios for planning and oped at the Rand Corporation, scenarios were analysis lie in war games, exercises of simulated sketches of challenging but plausible situations conflict used for military training, planning, and to which participants had to respond, allowing operational decision-making. The roots of war exploration of associated threats and opportu- games extend to antiquity, although the first for- nities. They adopted the term “scenario” from malized war games were developed for officer film and theatre, where it denotes a brief sketch training in 19th-century Prussia.12 In the 1940s of a story that includes only enough detail to and 1950s, exercises resembling war games convey broad points of plot and character. As in began to be applied outside the purely military classic war games, scenarios in these exercises domain, to study potential international crises served to help organizations and their leaders that included both high-level political decision- prepare for novel, complex challenges that they making and the potential for military conflict. might not anticipate, and which – if they did These exercises were informed by the then-new arise – would likely develop too fast to allow ad- field of game theory, which promised new for- equate reflection or analysis in real time.14 mal insights into situations of conflict and strategic decision-making,13 motivated by the Over the past few decades, the use of scenarios recognition that the new nuclear age had both has moved outside the realm of military and raised the stakes of international diplomacy and diplomatic activity. Beginning with strategic created profound new uncertainties over how to planning at the Royal Dutch/Shell oil com- proceed. In these exercises, principally devel- pany,15 scenarios are now widely used for strategic planning, analysis, and assessment by

8 IPCC 2001a:741. 9 van der Heijden 1996:5. 14 Levine 1964a,b; Schelling 1964; DeWeerd 1967, 10 Berkhout et al. 2001:i. 1975; Brewer and Shubik 1979. 11 MEA 2006:xvii. 15 For relevant history, see Hausrath 1971, Shubik 1975, Greenberger et al. 1983, Huss 1988, Schoemaker 1995, 12 Brewer and Shubik 1979. Schultz and Sullivan 1972, Schwartz 1991, Shell In- 13 Von Neumann and Morgenstern 1944, Nash 1950. ternational 2003.

14 Global-Change Scenarios - Their Development and Use

businesses and other organizations. They have scenarios. For example, models do not need sce- also figured increasingly prominently in plan- nario-based inputs when used to reconstruct past ning, analysis, and policy debate for long-term conditions or study causal processes. environmental issues, in particular global climate change. Because the total body of experi- The distinction between assessments and sce- ence with scenarios provides useful insights into narios is perhaps clearest in conventional as- their use in any particular domain, this section sessments based on deliberations of expert elaborates the meaning, characteristics, and po- panels, such as the IPCC, US National Assess- tential uses of scenarios in general. The next ment, or Millennium Ecosystem Assessment section turns to their specific use for global en- (MEA). Such assessments often construct rep- vironmental issues. resentations of future development of an issue, usually based on formal models. These repre- 1.1.1 Distinguishing scenarios from sentations require scenario-based inputs and Scenarios have figured assessments, models, and analyses may produce outputs that are themselves used increasingly prominently as scenarios in other activities. But the sce- in planning, analysis, and Confusion is widespread in discussions of scenario-related activities are frequently not the policy debate for long- narios, because their form and usage are highly central focus of the overall assessment, which diverse, and because writers’ uses of the term may examine many additional things, e.g., the term environmental are often imprecise and occasionally contradic- state of knowledge in particular scientific areas, issues, in particular tory. Scenarios must be distinguished, on the the status of and trends in particular environ- global climate change. one hand, from assessments and various types mental conditions, the evidence attributing par- of decision support activity that often use sce- ticular environmental changes to human inputs, narios; and, on the other hand, from other types or particular policy-relevant scientific ques- of statements about future conditions, such as tions. Assessments may also include evalua- predictions, projections, or forecasts. tions of proposed actions or proposed criteria for conducting such evaluations. Scenarios thus An assessment is any process that reviews and may provide required inputs to assessments, but synthesizes scientific or other expert knowledge are distinct from them. to provide information of relevance to policy- or decision-makers.16 The most common meth- 1.1.2 Distinguishing scenarios from ods of assessment are deliberations of expert projections, predictions, and panels and formal models, but other methods forecasts combine human deliberations with formal analysis or modeling, including war games or Scenarios must also be distinguished from other other simulation games, policy exercises, types of statements about the future, such as political-military exercises, constructing future predictions, projections, and forecasts. All of histories, backcasting, and others.17 These meth- these satisfy the basic definition above: they ods may use specifications of potential future are descriptions of potential future conditions conditions – i.e., scenarios – as inputs to or whose primary purpose in most cases is to sup- components of their work. Scenarios may even port decisions. How can scenarios be distin- be essential for some of these methods. For ex- guished? Examining the ways scenarios are ample, a war or crisis gaming exercise needs a used and discussed by practitioners and re- scenario to specify the nature of the threat or cri- searchers suggests four characteristics that dis- sis, while a formal model used to represent fu- tinguish them from other types of future ture development of some issue of concern needs statements. Although these characteristics are a scenario to specify future values of those inputs not essential, they are all more likely to be pres- not explicitly calculated within the model. But ent in scenarios than in other types of future these methods are broader than and distinct from statement, so they help to sharpen and delimit the definition of a scenario.

16 Parson 2003:9; Mitchell et al. 2006. First, scenarios are multi-dimensional: they de- 17 NRC 1996; Hausrath 1971; Brewer 1986; Shubik scribe multiple characteristics that collectively 1975; Svedin and Aniansson 1987; Schultz and Sulli- make up a coherent representation of future van 1972; Jones 1985; Parson 1996, 1997.

15 The U.S. Climate Change Science Program Section 1 - Scenarios, Their Characteristics and Uses

conditions. To achieve this, scenarios assemble fidence level and may be explicitly contingent and organize available knowledge, information, on specified assumptions about other future and assumptions from diverse bodies of conditions. Calling a future statement a “sce- research and expert judgment. The elements of nario” usually implies still less confidence and a scenario can be of diverse types: quantitative more associated contingencies. Any use of a or qualitative, precise or fuzzy, based on scenario for serious planning or analysis does, well-established research or informed specula- however, presume some minimal threshold of tion. Effective scenarios integrate their diverse likelihood. The situation described must be elements in a way that is coherent, communi- judged likely enough to merit attention, and to cates a clear theme or organizing principle, and justify expending resources and effort to study avoids internal contradiction. its implications and potential responses to it. There may also be a time ordering among these Effective scenarios Second, scenarios are schematic: that is, they three types of statements – predictions or fore- are multidimensional but not without limit. casts tend to describe nearer-term futures and integrate their diverse Scenarios do not seek to describe potential scenarios longer-term futures – but there are ex- elements in a way that future conditions with complete precision or de- ceptions, and the meaning of near term and long is coherent, tail. Rather, they highlight essential character- term depend on the particular context. communicates a clear istics and processes with enough detail that theme or organizing knowledgeable observers perceive them as re- 1.2. CREATING A principle, and avoids alistic and relevant, but not so much as to dis- SCENARIO EXERCISE: internal contradiction. tract from large-scale patterns. Indeed, one KEY CHARACTERISTICS potential use of scenarios is to stimulate creative AND CHOICES thinking and insights, for which they must leave something to the imagination. How much detail Beyond these general characteristics, scenarios and precision is appropriate is a judgment that vary greatly in their use, production, and con- depends on the particular application. tents. Extensive scholarly effort has gone into providing alternative scenario taxonomies.19 Third, scenarios usually come in groups. To be Scenarios can be distinguished, for example, by a useful tool to inform decision-making under whether they present a snapshot of a future state uncertainty, scenarios must represent uncer- or a dynamic account of changes over time to tainty. This is most often done by providing reach that state; by their degree of complexity; multiple scenarios, each presenting an alterna- by the relative balance of deliberation and intu- tive realization of uncertain future conditions.18 ition versus formal analysis used in producing The number of scenarios depends on the appli- them; or by their temporal and spatial scale. cation. Scenario exercises usually use between The set of characteristics on which scenarios two and seven, depending on the stakes of the could be sorted is long and open-ended, so we issue, the resources invested in the exercise, and make no attempt to provide an exhaustive list. the depth of analysis devoted to each scenario. Instead, we summarize the main dimensions of The most frequently proposed numbers are scenario variation in the form of a list of poten- three or four. tially open-ended design choices.

Finally, scenarios tend to claim less confidence 1.2.1 Variation among scenarios: than other types of future statements. Although three basic dimensions different authors’ usage is not consistent, “prediction” and “forecast” usually denote state- Three dimensions of variation, concerned with ments for which the highest confidence is the purpose of a scenario exercise, have far- claimed. “Projection” denotes a less confident reaching implications for its design and use and statement, which may have some specified con- so merit separate discussion. First, the intended use of a scenario exercise can vary from the more predictive to the more exploratory or 18 Crisis-response exercises are often an exception, presenting one scenario at a time showing a novel challenge, to which participants must respond and which 19 See, e.g., Duncan and Wack 1994, Godet and Roube- is implicitly contrasted to the status quo. lat 1996, van Notten et al. 2003.

16 Global-Change Scenarios - Their Development and Use

heuristic. It is of course a fundamental error to crete decisions and decision-makers is indirect. take a scenario’s illustrative description of po- Scenarios may be used for risk assessment, con- tential future conditions as a confident predic- tingency planning, identification of potential tion of what will actually happen – in our threats or actions to be considered, or early char- terminology, to take a scenario as a projection or acterization of a poorly understood issue. In even a prediction.20 Still, a scenario must be such applications, exploratory uses dominate. judged likely enough to merit the attention of Scenario exercises that are closer to decisions busy people. Exploratory uses of scenarios may with significant stakes operate under very dif- presume only this low threshold, yet have great ferent requirements, usually driven by specific value. For example, scenarios can be used to user needs. Their uses tend to be more predic- probe and challenge the mental models, thought tive – constrained by limits of available knowl- Scenarios can be used habits, and unrecognized assumptions of edge – so they might be expected, for example, to probe and decision-makers; to clarify points of agreement to provide more explicit and complete charac- challenge the mental and disagreement; to identify and engage terization of major uncertainties. They are also models, thought needed participants; to provide a preliminary likely to be more integrated with methods to habits, and structure for advance analysis of potential fu- evaluate alternative choices and identify pre- unrecognized ture decisions; or to seek insights into unrecog- ferred ones. nized opportunities, risks, causal linkages, or assumptions of uncertainties.21 Such insights can arise not just A third basic dimension of variation concerns decision-makers; to from examination of uncertainties, but also whether scenarios are defined primarily norma- clarify points of from meticulous critical examination of future tively, on the basis of their perceived desirabil- agreement and factors that are essentially certain (e.g., strongly ity or undesirability, or primarily on the basis of disagreement; to determined demographic trends such as the their perceived plausibility or likelihood. Al- identify and engage aging of industrialized-country populations) or though all scenarios include both positive and needed participants; even of present conditions whose significance normative elements, it is important to keep as to provide a 22 has been overlooked. Still, the predictive con- clear as possible which elements are included preliminary structure fidence accorded to scenarios is a matter of de- based on perceived likelihood or plausibility for advanced analysis gree: carefully developed scenarios that are and which because of perceived desirability or of potential future judged to have captured the most important un- undesirability. The most frequent use of ex- decisions; or to seek certainties may claim some moderate degree of plicitly normative scenarios involves construct- insights into confidence, and reasonable distinctions may be ing some hypothetical future state primarily on drawn between scenarios that represent conven- the basis of its desirability. Such a future might unrecognized tional versus surprising futures, best and worst be constructed to embody participants’ general opportunities, risks, cases, etc. intuitions about desirable social trends, or to causal linkages, or achieve specific environmental, development, uncertainties. A related dimension of variation among sce- or other goals.24 The scenario exercise then con- nario exercises is their proximity to decision- sists primarily of backcasting – attempting to making.23 Scenario exercises may involve construct paths that connect present conditions actual decision-makers and seek to directly ad- to the specified future conditions, to examine vise a specific, identified, near-term decision, the feasibility of the target, and identify costs, but more frequently their relationship to con- tradeoffs, and conditions associated with meeting it.25 Similarly, one can posit an undesirable future state and then reason through conditions 20 Several such errors are collected and discussed in associated with avoiding it. This approach is Bracken 1977, 1990; and Brewer 1990. 21 Brewer 1990. 22 Shell International 2001, 2003. For example, in a 24 See, for example, the simple scenario exercise in 1960s crisis exercise on a Soviet invasion of Iran, one NRC (1999:161-176) that posited specific targets to re- participant realized the local supply of jet fuel avail- duce world hunger and greenhouse-gas emissions by able to support a rapid US response was ten times year 2050, or the scenarios of the Global Scenario larger than had been thought, because kerosene – an Group, which included some defined by specified acceptable substitute – was used for domestic cooking trends and others back-cast from normatively specified and heating (Schelling 1994). targets for 2050 (Kemp-Benedict et al. 2002, Raskin et al. 2002). 23 This dimension is presented by Van Notten et al. 2003 as “exploration” versus “decision support.” 25 Robinson 1982, 2003.

17 The U.S. Climate Change Science Program Section 1 - Scenarios, Their Characteristics and Uses

sometimes proposed to reduce the risks of hid- The most basic choices in developing scenarios, den bias in construction of scenarios which, like which include the three dimensions of variation any decision support tool, can be misused to called out above, involve identifying the main provide support for a decision already made for focus of the exercise: what issues are to be ad- other reasons, rather than to inform a decision dressed or what decisions informed, for whom? not yet made. By bundling normative assump- The decision to conduct a scenario-based exer- tions into the future target state or boundary cise does not necessarily imply that these mat- conditions, analysts hope to reduce their pene- ters are clearly understood. The closer a tration into the subsequent instrumental reason- scenario exercise is to concrete decisions, the ing about actions and conditions to reach the more likely it is that these definitional issues target.26 will be understood clearly, in part through discipline on the process imposed by the involve- But whether the 1.2.2 Developing scenarios: ment of decision-makers. But most often, the main dimensions of choice coupling of scenarios to decisions is relatively relationship of a weak.27 In some applications (e.g., corporate scenario exercise to Table 1.1 extends the preceding discussion, strategic planning, responding to a novel mili- decisions is near or far, summarizing the main areas of variation and tary threat) the relevant decision-makers may be direct or indirect, clear choice involved in constructing a scenario ex- clearly identified at the outset, but the issues to understanding of its ercise. This is a highly simplified representa- be addressed and relevant choices may not be. focus and purpose is tion of a complex set of choices. In any In other applications, scenarios may be devel- important, and particular scenario exercise some of these oped to address some broad issue or concern infrequently achieved. choices may be made by default, without ex- (e.g., climate change, emerging infectious dis- plicit consideration, perhaps because the pre- eases, or terrorism), but the potential users and ferred choice is obvious in context. Although decisions to be informed might both be unspec- we present these choices in simple sequential ified. But whether the relationship of a scenario order, this is not necessary: choices might be exercise to decisions is near or far, direct or in- made in some other order, or iteratively ad- direct, clear understanding of its focus and pur- justed. But while the process and sequence of pose is important, and infrequently achieved: choices may be idealized, the set of choices is many scenario exercises muddle through with not: creating a scenario requires explicit or im- vagueness, confusion, or disagreement regard- plicit choices on all these dimensions. ing the focus, purpose, and intended user of the exercise. Clarifying the overall focus of a scenario exercise may require broad consultations or scoping workshops involving many poten- Table 1.1. Idealized Sequence of Major Choices in Scenario Development. tially interested decision-makers, other stakeholders, and analysts and researchers. Main focus, framing, users, question(s) to be addressed Process and participation A second basic set of decisions concerns the process by which scenarios are developed. Like Key uncertainties to explore: how many, over what range the focus of the exercise, decisions about the Narrative, quantitative, or both process of developing scenarios often receive Level of complexity (number of quantitative variables, detail of narrative) little thought, or are not even explicitly recognized as choices, but are nonetheless highly Specific variables and factors to specify consequential. What expertise must be included Time horizon and spatial extent to ensure the scenarios adequately reflect the best available scientific knowledge, data, and Temporal and spatial resolution models? What decision-makers, stakeholders, or their surrogates must be involved to keep the scenarios relevant, plausible, and credible? For

26 This approach does not preclude such misuse: if a goal is strongly desired, scenarios are at risk of con- scious or unconscious bias to make it look easy. Japan- 27 E.g., note the predominance of scenarios on the “ex- ese war-games of the Battle of Midway provide striking ploration,” rather than the “decision support” side in examples (Bracken 1977). the survey of Van Notten et al. 2003.

18 Global-Change Scenarios - Their Development and Use scenario exercises that must integrate knowl- represented by four scenarios, often illustrated edge from diverse domains, individual partici- by a two-by-two matrix.28 Several alternative pants’ knowledge, flexibility, and imagination scenarios might seek to span a plausible range can be as important as the disciplines or stake- for some key quantitative variable, or present holder groups they represent. How intensively, distinct qualitative outcomes for a single uncer- for how long, and by what means will partici- tainty, e.g., three types of competitive threat, or pants interact? Will the process be open to out- three political futures for a region in turmoil. side observers or participants? How and when Alternatively, scenarios can represent plausible will feedback on the scenarios be sought, and extreme or “worst-case” scenarios, to assess the how will it be used? How and to whom will re- robustness of decisions or strategies. sults be communicated? And crucially, how will the be process be led, and how will dis- How rich and complex should each scenario be? agreements be resolved? With good leadership, Defining scenarios as we have, as multivariate Many dimensions of resolving differences in a scenario exercise can but synoptic, still leaves a wide range of levels uncertainty may be be less arbitrary and more illuminating than in of complexity to choose from. At one extreme, relevant to the issue other group tasks; when disagreements persist scenarios may specify time-paths for just a few being examined, but only after careful examination, they can be treated as quantitative variables, or even just one. Such important uncertainties to be retained as alter- scenarios are common, e.g., in applications such a few can be examined native scenarios, not suppressed by picking a as analyzing a firm’s profitability under alter- explicitly in any winner, splitting the difference, or retreating to native scenarios for oil prices, or projecting tax scenario exercise. vague language. revenues under alternative scenarios of productivity growth and inflation, often in a standard Whatever process is chosen, a series of substan- “high, middle, low” format. A scenario can active choices must be made about what goes into commodate more complexity by projecting ad- the scenarios. The most important of these conditional quantitative variables, but as the cern what key uncertainties will be explored, and number of variables increases, so also does the how much richness and detail should be included need for an organizing principle or gestalt to tie in scenarios to illuminate these. them together in a non-arbitrary way.

What uncertainties are to be explored, and how? At the other extreme, the core of a set of sce- Many dimensions of uncertainty may be rele- narios can be a set of rich, coherent narratives, vant to the issue being examined, but only a few an approach frequently called the Shell ap- can be examined explicitly in any scenario ex- proach.29 Each narrative, described principally ercise. Defining these is a crucial choice that in text, reflects a distinct conception of how the shapes much of what follows in a scenario ex- world might develop with a persuasive underly- ercise. For those uncertainties judged most im- ing causal logic. A narrative scenario can stand portant, alternative outcomes are usually alone but may also include specifications of im- represented in alternative scenarios. For exam- portant quantitative variables, e.g., of popula- ple, scenarios might present high- and low- tion or economic growth, consistent with the growth futures, or alternative forms that a broad causal logic underlying the scenario. The competitive threat might take. Other uncertain- narrative provides the context and explanatory ties judged less crucial are usually suppressed logic that tie together the time-paths of quanti- by presenting a single “best guess” or “refer- tative variables, although particular time-paths ence case.” The few key uncertainties chosen are regarded as illustrating, not defining, the can be represented in the number and character scenario – i.e., a different scenario would pres- of scenarios, depending on the intended use. A ent substantially different time-paths or rela- particular uncertainty might be represented by tionships among them. high and low values of some quantity, or by a reference case supplemented with high and/or low variants. If two or more uncertainties interact, they can be represented by scenarios that 28 Alternative interpretations of this matrix structure are combine different outcomes of each: in the sim- discussed in van’t Klooster and van Asselt 2006. plest form, two interacting uncertainties can be 29 Van der Heijden 1996; Wack 1985a,b; Schwartz 1991; Shell International 2003.

19 The U.S. Climate Change Science Program Section 1 - Scenarios, Their Characteristics and Uses

The choice of how rich and complex to make The remaining substantive choices in specify- scenarios has far-reaching implications for the ing a scenario follow from the preceding large- process of developing the scenarios, what can scale choices. They include specifying the time be done with them, and the uses they can serve. horizon and spatial extent of the scenarios, de- The two extreme approaches imply large differ- ciding the particular elements to include, and ences in how uncertainty is treated, what aspects the temporal and spatial resolution at which sce- of the problem receive attention, and the rela- nario outputs are stated. Decisions about tem- tionship between scenarios and their users, poral resolution (e.g., hourly to multi-decadal) which we discuss for climate-change scenarios and spatial resolution (e.g., regional, national, in Section 4. Richer and more complex scenar- continental scales) are particularly important ios require more time and effort to develop, when – as is often the case in global-change ap- whether based on quantitative models, narra- plications – scenarios are produced or used by tives, or both. Complex, narrative-based sce- quantitative models. Such models may have very narios may require many person-months of precise requirements for the specification and development to become realistic, relevant, and resolution of inputs and outputs, creating the pos- persuasive, with consistent relationships among sibility for serious mismatches between what scenario elements. In return for the extra effort, users need or expect, and what scenario develop- this approach allows great flexibility in the way ers feel comfortable and competent providing. potential futures are described. Narratives can convey different aspects of a future situation This section has discussed the uses, types, and with varying degrees of salience or specificity, characteristics of scenarios broadly, in any ap- and they can compactly convey the tone or char- plication area. The next section narrows the acter of a future situation by allusion, where a focus to climate change and related areas of precise specification would appear arbitrary or global environmental change, summarizing the labored. The narrative approach avoids limit- types of scenarios that have been used and pro- ing the defining characteristics of a scenario to posed, and that might be required, to explore any particular set of pre-specified variables, but and inform decision-making in this area. attempts to be alert to a wide range of potentially important characteristics and mechanisms of causal influence. Proponents of this approach argue that a coherent narrative at the core of a scenario is necessary to avoid arbitrariness in specifying multiple variables, and to make the exercise useful to decision-makers: “Most scenarios merely quantify alternative outcomes of obvious uncertainties (for example, the price of oil may be $20 or $40 a barrel in 1995). Such scenarios are not helpful to decision-makers”.30

30 Wack 1985a:74.

20 Global-Change Scenarios - Their Development and Use

Scenarios in 2 Global-Change Analysis and Decision Support SECTION

Analysts have tried to develop scenarios to support understanding of and decision-making for global environmental issues, beginning with the global models of the mid-1970s and early assessment of acid rain and stratospheric ozone in the late 1970s to early 1980s.31 The reasons for using scenarios in global change are similar to those that apply in other decision domains: high-stakes decisions that must be made under deep uncertainty about the conditions that will determine their consequences, the values at stake, or the relevant set of choices and actors. As in other domains, well-designed scenario exercises can provide a structure for assessing alternative choices and help focus on the nature of the issue, the relevant choices and actors, the values that might be at stake, and the types of research or analysis that might help clarify preferred choices.

For climate-change applications, scenario exercises have been conducted and sponsored by governments, international organizations, non-governmental organizations, and collaborative groups. While these have been diverse in form, details, and purposes, they have tended to focus more on heuristic and exploratory uses than on supporting specific decisions. The boundaries of the climate-change issue are not sharply defined, however: climate change implicates and connects to many other areas of policy, including energy, agriculture, hazard protection, and broad questions of economic development. Consequently, there is substantial uncertainty about what all the relevant decisions, decision-makers, and potentially affected values are. While some decisions are clearly of primary relevance to climate change, many other decisions that appear to be connected have not yet incorporated consideration of climate change or even recognized the connection. Reflecting these fuzzy boundaries of the issue, scenario exercises developed for climate change have overlapped with other exercises primarily focused on ecosystems, energy, and broad issues of world development. The fuzziness of the climate issue’s definition increases the challenge of developing useful scenarios, but also increases the potential value of well-crafted and executed scenario exercises.

31 See, e.g., Meadows et al. 1972, Barney 1981; summary of early ozone assessments in Parson 2003; and summary history of scenarios in global-change applications in Swart et al. 2004. What was the earliest scenario work in global change depends, of course, on how the boundaries of global change are defined. Kahn and Wiener 1967 might be considered an early example.

21 The U.S. Climate Change Science Program Section 2 - Scenarios in Global-Change Analysis and Decision Support

The decisions most directly related to climate have been embedded in larger exercises of change are conventionally sorted into two cate- assessment, modeling, or characterization of gories, mitigation and adaptation.32 Mitigation the issue. These uses have included formal consists of actions that reduce the human per- integrated-assessment models,33 comprehensive turbations of the climate system, by reducing assessments conducted by multi-disciplinary ex- net anthropogenic greenhouse-gas emissions. pert bodies (e.g., IPCC), and more narrowly fo- Adaptation consists of actions to reduce the cused assessment exercises targeting specific harm or increase the benefit from climate aspects of the climate-change issue. In these change and its impacts. Despite uncertainty uses, scenarios represent components of the about the precise decision agenda, we can iden- climate-change issue that are required inputs to tify in general terms the type of information sce- an assessment or model. narios might provide that would be useful to each To date, most climate- type of decision. The causal logic of the climate-change issue is complex, including multiple two-way causal related uses of Scenarios can help inform adaptation decisions links and feedbacks among socio-economic, scenarios have not by characterizing the nature and severity of rel- geophysical, and ecological systems. examined decisions evant potential impacts; identifying key vulner- Integrated-assessment models seek to represent directly, but have been abilities, particularly those that might not many of these linkages and feedbacks explic- embedded in larger otherwise have been recognized; identifying re- itly; Figure 2.1 shows a typical example of the exercises of search or monitoring priorities that might give “wiring diagrams” that illustrate the increas- assessment, modeling, advance warning about impacts, particularly ingly dense linkages and feedbacks represented or characterization of acute vulnerabilities; expanding the perceived in these models. But while such diagrams the issue. set of potential responses; and providing a might be taken to indicate that all relationships framework for evaluating alternative adaptation are represented explicitly within the model – en- measures. They may also help to clarify the dogenously – this is not the case. All models of time structure of relevant decisions, identifying the climate-change issue rely on scenarios to those near-term decisions that might have im- specify some future quantities exogenously, and portant but under-recognized connections to fu- in virtually all cases, scenario-specified inputs ture impacts and vulnerability. are not modified to account for results of the subsequent analysis: i.e., they are truly exoge- Similarly, scenarios can help inform mitigation nous, and the causal logic does not close. decisions by characterizing the potential impacts of climate change and their severity, since When scenarios are used to specify exogenous these provide the motivation for mitigation. inputs to a model of some aspect of the climate- But, in addition, mitigation decisions can bene- change issue, the causal logic of the analysis can fit from information about potential emissions be greatly simplified from that shown in Figure trends, which determine the nature of the chal- 2.1. Instead, the logic can be represented by a lenge of limiting emissions; about potential simple linear structure that extends from human pathways of the extraction and depletion of cur- activities to emissions to climate change to im- rent energy resources and development of new pacts. Figure 2.2 shows this highly simplified ones; and about potential pathways of techno- structure. This representation is even more suit- logical development. Mitigation decisions may able for the uses of scenarios in other types of also benefit from scenarios representing the po- global-change assessments, which have been or- tential policy context in which they are made. ganized around much simpler causal structures than those that integrated-assessment models To date, most climate-related uses of scenar- seek to represent. Note that we are not claiming ios have not examined decisions directly, but this simple logical structure adequately represents the true structure of the climate-change issue: only that it illustrates the ways that scenar- 32 While this categorization has frequently been criti- ios have been used to provide exogenous inputs cized for neglecting actions with overlapping effects to global-change models and assessments. and the third category of direct interventions in the climate system (Schelling 1983, Keith 2000, Keith et al. 2006, Parson 2006), it remains a useful approximation for most currently proposed responses. 33 Weyant et al. 1996, Parson and Fisher-Vanden 1997. 22 Global-Change Scenarios - Their Development and Use

Figure 2.1. Wiring Diagram for Integrated Assessment models of climate change. (Source: Weyant et al. 1996)

ATMOSPHERIC COMPOSITION CLIMATE & SEA LEVEL

Atmospheric Climate Chemistry

Ocean Ocean • temperature Carbon Cycle • sea level

HUMAN ACTIVITIES ECOSYSTEMS

Energy Other Human Terrestrial Unmanaged System Systems Carbon Cycle Ecosystem

Agriculture, Coastal Crops & Hydrology Livestock, & System Forestry Forestry

This linear logical structure allows a simple, as land-use change, are the best known type of practical categorization of five types of scenar- climate-change related scenario. Emissions ios that have been developed for the climate- scenarios provide required inputs to model cal- change issue. These types are defined by what culations of future climate change, as shown in quantities they specify and what primary area Figure 2.3. As the focus and intended use of cli- of analysis they provide input to. Their differ- mate-model studies have shifted over time, so ences can be represented by where they cut the has the role of emissions scenarios. Early re- causal chain in Figure 2.2, with the scenario search studies examined the climate system’s re- specifying quantities lying on one side of the sponse to potential (rather than projected) cut, and the assessment or other activity using emissions inputs in individual model studies or the scenario lying on the other side. The next standardized model comparisons. In such exer- five subsections discuss these five types of cli- cises, the purpose of a scenario is to provide a mate-change-related scenarios in turn. known, consistent perturbation big enough to generate an informative model response. Such 2.1. EMISSIONS SCENARIOS scenarios must be standardized, so differences FOR FUTURE CLIMATE between model runs can be traced to scientific SIMULATIONS uncertainties and model differences, but they can be simple and arbitrary, making no claim to Scenarios of greenhouse-gas emissions, some- being realistic. The earliest such scenarios times including other human perturbations such showed a “step-change” increase in atmospheric

SOCIO- Figure 2.2. ATMOSPHERE ECONOMIC EMISSIONS IMPACTS & CLIMATE Anthropogenic CONDITIONS • greenhouse gases • sea level rise climate change: • carbon cycle • population • aerosols • fresh water • temperature Simplified linear • GDP • other drivers, • ecosystems • precipitation • energy system e.g. land-cover • agriculture causal chain. • humidity • industry change • human health • soil moisture • technology • property, • agriculture, • extreme events infrastructure land-use

23 The U.S. Climate Change Science Program Section 2 - Scenarios in Global-Change Analysis and Decision Support

concentration of CO2 from its pre-industrial energy-related CO2, representing emissions in value, to either two or four times that value.34 terms of underlying drivers such as population,

Models’ equilibrium responses to doubled CO2 economic growth, and technological change and provided a standard benchmark of model re- projecting these drivers using some combina- sponsiveness, which has remained around the tion of modeling and trend extrapolation. range of 1.5 to 4.5°C for more than 20 years. Driven by such scenarios, climate models for This range of modeled equilibrium responses to the first time can claim to be reasonable esti- a standardized perturbation does not predict ac- mates of how the climate might actually change. tual climate changes under human perturba- In addition, comparisons using multiple models tions, although it has often mistakenly been and emissions scenarios have allowed partition- taken as such. ing of uncertainty in future climate change into roughly equal shares attributed to uncertainty in The next generation of climate-model studies, climate science and models, and in emissions beginning in the early 1990s, specified a time- trends.36 These comparisons have also allowed path of atmospheric concentrations rather than estimation of the climate-change benefits from a one-time perturbation. These studies for the specified emissions reductions. first time allowed comparison of models’ transient responses, by examining not just how much As this shift occurred, advances in climate mod- the cimate changes, but how fast it changes. They els – e.g., improved representations of atmos- still used a simple, highly idealized standard sce- pheric aerosols, tropospheric ozone, and nario, most frequently a 1 percent per year in- atmosphere-surface interactions – produced crease in atmospheric concentration, expressed mismatches between emissions scenarios and

as CO2-equivalent. Only two such transient the input needs of climate models. For example, simulations had been conducted by the first climate models now require emissions of sev- IPCC assessment (1990), but by the time of the eral types of aerosols and reactive gases (prin- second assessment (1996), most modeling cipally the ozone precursors, hydrocarbons, CO 35 groups had produced at least one. and NOx), explicit estimates of black carbon and organic carbon, and some disaggregation of dif- Since the mid-1990s, climate modelers have in- ferent types of volatile organic compound creasingly sought to produce realistic pictures (VOC) emissions. Moreover, because these of how the climate may actually change, requir- emissions act locally and regionally rather than ing a new approach to emissions scenarios. globally, they must be specified at the spatial Scenarios must now present well-founded judg- scale of a model grid-cell, about 150 sq. km. ments or guesses of actual future emissions Models of atmospheric chemistry and transport trends and their consequences for atmospheric then use these emissions to generate the con- concentrations. The required emissions scenar- centrations and radiative forcings used by the ios have been constructed either by extrapolat- climate model. Since emissions scenarios often ing recent emissions trends, or, particularly for do not provide the required detail, climate mod-

Figure 2.3. SOCIO- ATMOSPHERE ECONOMIC EMISSIONS IMPACTS Emissions & CLIMATE CONDITIONS scenarios for • greenhouse gases • sea level rise • carbon cycle • population • aerosols • fresh water climate • temperature • GDP • other drivers, • ecosystems • precipitation simulations. • energy system e.g. land-cover • agriculture • humidity • industry change • human health • soil moisture • technology • property, • agriculture, • extreme events infrastructure land-use

Content of Scenarios Use of Scenarios

35 Washington and Meehl 1989, Manabe et al. 1991, IPCC 1996a. 34 e.g., Manabe and Wetherald 1967, Manabe and Stouffer 1979. 36 Cubash et al. 2001.

24 Global-Change Scenarios - Their Development and Use elers meet these input needs through various ad Alternatively, scenarios can be used to examine hoc approaches. what changes in policies, technologies, or other factors would be required to shift emissions Changes in standard emissions scenarios pose from some assumed baseline onto a specified challenges for maintaining comparability with lower path, and to estimate the cost of such a past model results. For example, the IPCC’s shift. To be used in this way, an emissions sce-

IS92 scenarios projected that future SO2 emis- nario might be specified arbitrarily, or might sions would roughly double, then stabilize, specify some environmental target based on while the later SRES scenarios projected sharp normative criteria as discussed in Section 1.2. decreases, giving 2100 emissions about Such scenarios have been most frequently used one-quarter the IS92 value. This change caused to examine emissions trajectories that stabilize significant increases in projected warming that atmospheric CO2 concentrations at specified were not due to any changed scientific under- levels. More recent exercises have instead taken Models’ equilibrium standing. To help maintain backward compara- stabilization of radiative forcing as the target, to responses to doubled bility, many climate-model groups have examine the role of non-CO greenhouse gases in 2 CO provided a continued to run simulations using older stan- meeting stabilization goals.37 2 standard benchmark of dardized scenarios, to provide benchmarks for comparisons both among current models and An important early example is the Wigley, model responsiveness, between current and previous-generation models. Richels, Edmonds (WRE) scenarios, which pre- which has remained sented emissions pathways that stabilized at- around the range of

2.2. EMISSIONS SCENARIOS mospheric CO2 concentration at five levels, 1.5 to 4.5˚C for more FOR EXPLORING ALTERNATIVE ranging from 450 to 1000 ppm.38 Developed than 20 years. ENERGY AND TECHNOLOGY heuristically from a simple model of the global FUTURES carbon cycle and two energy-economic models, these scenarios illustrated the large cost savings In addition to providing needed inputs to cli- attainable by approaching stable concentrations mate models, emissions scenarios have also through emission paths that initially rise and been produced to examine alternative then decline steeply, rather than by beginning a socio-economic, energy, and technological fu- more gradual decline immediately. tures, as shown in Figure 2.4. As in Figure 2.3 the content of the scenario is emissions, but the Several other sets of stabilization scenarios have scenario is now used to examine the socio-eco- been proposed and used for similar explo- nomic implications of alternative emission rations. For example, the Energy Modeling paths, which lie upstream or to the left in the Forum (EMF) has convened several multi- causal chain. A scenario specifying a particular model scenario exercises focusing on emissions, emissions time-path can be used to explore what emissions constraints, and their socio-economic patterns of demographic and economic change, effects. These have studied decision-making energy resource availability, and technology de- under uncertainty, international distribution of velopment are consistent with that trajectory. costs and benefits, the costs and benefits of the

SOCIO- ATMOSPHERE Figure 2.4. ECONOMIC EMISSIONS IMPACTS & CLIMATE CONDITIONS emissions • greenhouse gases • sea level rise • carbon cycle scenarios for • population • aerosols • fresh water • temperature • GDP • other drivers, • ecosystems energy/technology • precipitation • energy system e.g. land-cover • agriculture • humidity futures. • industry change • human health • soil moisture • technology • property, • agriculture, • extreme events infrastructure land-use

Content of Scenarios Use of Scenarios

37 de la Chesnaye and Weyant 2006, EMF 2006, CCSP 2007. 38 Wigley et al. 1997. 25 The U.S. Climate Change Science Program Section 2 - Scenarios in Global-Change Analysis and Decision Support

Kyoto Protocol, the implications of potential fu- nental regions, to climate model grid-cells or ture energy technologies and technological finer scales. They may project these at time res- change for emissions, and the implications of in- olutions ranging from annual or seasonal aver- 41 cluding non-CO2 gases and carbon sequestration ages to daily or even finer-scale weather. in mitigation targets and policies.39 Three major types of climate scenarios are dis- In a recent scenario exercise of this type spon- tinguished by how they are produced: incre- sored by the CCSP, three modeling teams con- mental scenarios, analog scenarios, and structed separate reference-case scenarios to climate-model scenarios.42 Incremental scenar- examine the implications of stabilizing radiative ios change current conditions by plausible but

forcing at levels roughly corresponding to CO2 arbitrary amounts. For example, a region’s tem- concentrations of 450, 550, 650, and 750 ppm. perature might be warmed by 2, 3, or 4°C from They examined the energy system, land-use, present conditions, or its precipitation increased and economic implications of moving to stabi- or decreased by 5, 10, or 20 percent. Such ad- lization. This project explored the role of mul- justments can be made to annual or seasonal av- tiple gases and alternative multi-gas control erages, to finer-period measurements of current strategies in pursuing atmospheric stabilization. conditions, or to the variability of temperature These scenarios may also provide a basis for fu- or precipitation over days, months, or years.43 ture analyses by the CCSP, the Climate Change Like the simple emissions scenarios used for Technology Program (CCTP), or others.40 climate-model comparisons, incremental climate scenarios are simple to produce but make 2.3. CLIMATE CHANGE no claim to represent actual future conditions. SCENARIOS They are used for initial exploratory studies of climate impacts and to test the sensitivity of im- Climate scenarios describe potential future cli- pact models. mate conditions (Figure 2.5). They are used to provide inputs to assessments of climate-change Analog climate scenarios represent potential fu- impacts, vulnerabilities, and associated options ture climates by the observed climate regime at for adaptation, and to inform decision-making another place or time. A spatial analog imposes related to either adaptation or mitigation. De- the climate of one location on another, e.g., rep- pending on their specific use, climate scenarios resenting the potential climate of New York in may include multiple variables, such as temper- the 2050s by that of Atlanta today or that of Illi- ature, precipitation, cloudiness, humidity, and nois in the 2050s by that of Kansas today.44 A winds. They may describe these at spatial scales temporal analog imposes climate conditions ob- ranging from the entire globe, through broad served in the past, in the historical record or ear- latitude bands, large continental and sub-conti- lier paleoclimatic observations, e.g., using the

Figure 2.5. Climate- SOCIO- ATMOSPHERE ECONOMIC EMISSIONS IMPACTS change scenarios. & CLIMATE CONDITIONS • greenhouse gases • sea level rise • carbon cycle • population • aerosols • fresh water • temperature • GDP • other drivers, • ecosystems • precipitation • energy system e.g. land-cover • agriculture • humidity • industry change • human health • soil moisture • technology • property, • agriculture, • extreme events infrastructure land-use

Content of Scenarios Use of Scenarios 41 IPCC – TGCIA 1999, Barrow et al. 2004. 42 Mearns et al. 2001. 39 See, e.g., Weyant and Hill 1999; Weyant 2004; de la Chesnaye and Weyant 2006; EMF 2006. 43 e.g., Mearns et al. 1992, 1996; Semenov and Porter 1995. 40 CCSP 2007. 44 E.g., Kalkstein and Greene 1997.

26 Global-Change Scenarios - Their Development and Use hot, dry period of the 1930s to study impacts of climate model, because the models are sensitive potential future hot, dry climates.45 Like incre- to small differences in starting conditions. And mental scenarios, analog climate scenarios are some of the uncertainty can be seen in differences more useful for exploratory studies of the cli- between calculations by different models, mainly mate sensitivity of particular resources or sys- caused by differences in the computational meth- tems than for projecting likely impacts. While ods they use to handle errors introduced by finite they represent climate states that are known to grid-cells, and the parameterizations they use to be physically possible, they are limited as rep- represent small-scale processes. resentations of potential future states since they do not consider the changes in greenhouse-gas Just as modeling future climate change requires concentrations that are the principal driver of specification of future emissions trends, assess- climate change. ments of future climate-change impacts require specification of future climate change. Data Just as modeling future Climate-model scenarios use computers to pro- from a climate-change scenario might be used climate change duce a physically consistent representation of as input to impact assessments of freshwater requires specification the movement of air, water, energy, and radia- systems, agriculture, forests, or any other of future emissions tion through the atmosphere. Climate models, climate-sensitive system or activity. Impact also called General Circulation Models or studies can use various methods, including trends, assessment of GCMs, approximate this calculation by divid- quantitative models such as hydrologic and crop future climate-change ing the atmosphere into thousands of grid-cells, models, threshold analyses that examine quali- impacts requires roughly 150 km. square in today’s models, with tative disruptions in the behavior of climate-sen- specification of future a dozen vertical layers, treating conditions as uni- sitive systems, or expert judgments that climate change. form within each cell and representing finer- integrate various pieces of scientific knowledge. scale processes by numerical relationships, called “parameterizations,” that are defined at the scale As with all scenarios, the usefulness of climate of a grid cell. Climate models are used to study scenarios depends on how well they meet users’ the present climate and its responses to past per- information needs. The highly specific climate- turbations like variation in the sun’s output or vol- data needs of impact analyses may not readily canic eruptions, and to construct scenarios of be provided by climate-model outputs, or may future climate change under any specified sce- include results of whose validity climate mod- nario of emissions and other disturbances. elers are not confident. For example, a common need of impact analyses is for data at Unlike incremental and analog scenarios, substantially finer scales than the relative coarse climate-model scenarios use emissions scenar- grid of a climate model, which might have only ios as inputs. Model-based scenarios have a 60 to 100 cells over the continental United greater claim than the other types to being real- States. One advantage of incremental and ana- istic descriptions of how the climate might ac- log scenarios is that they can typically provide tually change, because they are based on data at substantially finer scales. “Downscal- specified assumptions of future emissions ing” techniques seek to combine the benefits of trends acting on modeled representations of model-based scenarios – physical realism and known physical processes. Even with a given explicit emissions-scenario drivers – and data at emissions scenario, model-based climate sce- finer scales. The two major approaches are narios are uncertain. Since climate models are statistical downscaling and nested regional driven by the radiative effects of atmospheric con- modeling.46 Statistical downscaling involves es- centrations of relevant species, some of this un- timating statistical relationships between certainty comes from the carbon-cycle and large-scale variables of observed climate, such chemical processes by which specified emission as regional-average temperature, and local vari- paths determine concentrations and radiative forc- ables such as site-specific temperature and ings. Some of the uncertainty can be seen in the precipitation.47 These relationships are then as- slight differences among different runs of the same sumed to remain constant under global climate

46 Giorgi et al. 2001. 45 E.g., Rosenberg et al. 1993. 47 Wilby and Wigley 1997.

27 The U.S. Climate Change Science Program Section 2 - Scenarios in Global-Change Analysis and Decision Support

change. A nested regional climate model pro- local sea level rise.50 vides an explicit physical representation of climate for a specific region, including local In addition to its gradual impacts, sea level rise factors such as mountain ranges, complex coast- is subject to large uncertainties from the poten- lines, and surface vegetation patterns, with initial loss of continental ice sheets in Greenland tial and boundary conditions provided by a and West Antarctica. The consequences of these GCM. Regional climate models can provide events for global sea level rise are well known projections at scales as small as 10 to 20 km. because they can be calculated quite precisely Although downscaled results are anchored to from the volume of the ice sheets – roughly 7 local features with well-understood climatic ef- meters rise from complete loss of the West fects, downscaling introduces uncertainties beyond Antarctic Ice Sheet and 5 meters from Green- those already present in GSM results.48 land – but the probabilities of these events and their likely speed of occurrence are both highly 2.4. SCENARIOS OF DIRECT uncertain. One recent study has suggested a BIOPHYSICAL IMPACTS: probability of a few per cent that the West SEA LEVEL RISE Antarctic Ice Sheet will contribute an additional one meter per century beyond that calculated Although climate-change scenarios can provide from gradual warming.51 inputs to studies of any impact, scenarios can also be constructed of particularly important There are several reasons to call out sea level forms of impact, such as sea level rise – one of rise from other climate-change impacts to be the more costly and certain consequences of cli- represented in separate scenarios. First, sea mate warming (Figure 2.6).49 Changes in global level rise is a powerful driver of other forms of mean sea level as the climate warms can be cal- climate-change impact, probably the most im- culated using a GCM with a coupled ocean and portant driver of impacts in coastal regions. atmosphere, which can simulate the transfer of Since it is a direct physical impact of climate heat to the ocean and the variation of ocean tem- change that can be described precisely and com- perature with depth. To construct sea level rise pactly, a sea level rise scenario is an efficient scenarios for particular coastal locations, way to transmit the most important information model-derived projections of global mean sea about climate change to coastal impact assess- level rise must be combined with projections of ments. Moreover, since sea level rise does not local subsidence or uplift of coastal lands, as depend on socio-economic processes and can- well as local tidal variations derived from his- not be significantly influenced by human ac- torical tide-gauge data. tions (other than by limiting climate change itself), it may be reasonably treated as exoge- Sea level rise will increase circulation and nous for purposes of impact assessment. For all change salinity regimes in estuaries, threaten these reasons, sea level rise is a good proxy for coastal wetlands, alter shorelines through in- the most important causal routes by which climate creased erosion, and increase the intensity of change will affect coastal regions. coastal flooding associated with normal tides and storm surge. Scenarios of sea level rise are Finally, because sea level rise is subject to large consequently needed to assess multiple linked uncertainties with known consequences but un- impacts on coastal ecosystems and settlements. known probabilities, it is a useful variable for In specific locations, these impacts will depend exploratory analysis of worst-case scenarios in on many characteristics of coastal topography, long-range planning. Other forms of climate ecosystems, and land use – e.g., coastal eleva- impact might also merit being called out in sep- tion and slope, rate of shoreline erosion or ac- arate scenarios: changes in snowpack in moun- cretion, tide range, wave height, local land use tain regions, seasonal flow regimes in major and coastal protection, salinity tolerance of river basins, or the structure and function of coastal plant communities, etc. – in addition to major ecosystem types. Based on present

48 Mearns et al. 2001, Giorgi et al. 2001. 50 Burkett et al. 2005. 49 IPCC 2001a. 51 Vaughan and Spouge 2002.

28 Global-Change Scenarios - Their Development and Use

SOCIO- ATMOSPHERE DIRECT OTHER ECONOMIC EMISSIONS & CLIMATE BIOPHYSICAL IMPACTS Figure 2.6.: CONDITIONS IMPACTS • greenhouse gases Scenarios of • carbon cycle • coastal erosion • population • aerosols • sea level rise • temperature • estuaries direct biophysical • GDP • other drivers, • energy system e.g. land-cover • precipitation • ecosystems impacts: sea level • industry change • humidity • property, infrastructure rise. • technology • soil moisture • fresh water • agriculture, • extreme events land-use

Content of Scenarios Use of Scenarios knowledge, however, only sea level rise has ronmental changes are imposed – e.g., how shown these characteristics strongly enough to many people there are; where and how they live; motivate construction of separate scenarios. how wealthy they are; how they gain their liveli- hoods; and what types of infrastructure, institu- 2.5. MULTIVARIATE SCENARIOS tions, and policies they have in place.53 FOR ASSESSING IMPACTS, ADAPTATION, AND Assessment of climate impact on ecosystems VULNERABILITY that are intensively managed for human use, such as agriculture, managed forests, range- Many potentially important impacts of climate lands, and hydrologic systems, must consider change cannot be adequately assessed by con- human management as a factor in impacts. The sidering only how the climate might change. non-climatic factors that influence these man- These impacts require multivariate scenarios agement decisions – e.g., changes in market that include climate change and other charac- conditions, technologies, or cultural practices – teristics likely to influence impacts. This is the must be considered for inclusion in scenarios if case, for different reasons, for both ecosystems they are sufficiently important in mediating cli- and socio-economic systems, although the na- mate impacts. ture of the multivariate scenarios that are required – i.e., the number and identity of the In other domains, socio-economic factors can characteristics that must be specified – will vary mediate climate impacts by influencing vulner- widely among particular impacts. ability and adaptive capacity. No general model of the socio-economic determinants of adaptive Ecosystems are affected by climate change, but capacity exists. Important factors are likely to also by many other changes in environmental vary across specific types of impact, locations, conditions that are influenced by human activi- and cultures, and may include many demo- ties, such as nitrogen and sulfur deposition, tro- graphic, economic, technological, institutional, pospheric ozone and smog, and changes in and cultural characteristics. erosion, runoff, loadings of other pollutants, land use, land cover, and coastal-zone charac- Some socio-economic characteristics that are teristics. Consequently, realistic assessments of likely to be relevant for many impact assess- potential future impacts on ecosystems require ments – e.g., the size and sometimes the age specifying the most important forms of human- structure of population, the size and sometimes driven stresses jointly, not just climate.52 the sectoral mix of GDP – are normally generated in the course of producing emissions sce- In addition, many important forms of climate- narios. Consequently, when current emissions change impact depend not just on climate scenarios exist for the region for which an im- change, its direct biophysical impacts such as pact assessment is being conducted, it makes sea level rise, and perhaps other forms of envi- sense to strive for consistency with them.54 ronmental stress, but also on the nature of the Even for these variables, however, there may be society on which these climate and other envi-

53 Parson et al. 2001, 2003; Arnell et al. 2004. 52 MEA 2005. 54 Berkhout et al. 2001, citing UNEP 1994 guidelines.

29 The U.S. Climate Change Science Program Section 2 - Scenarios in Global-Change Analysis and Decision Support

Figure 2.7: SOCIO- ATMOSPHERE DIRECT OTHER Multivariate ECONOMIC EMISSIONS & CLIMATE BIOPHYSICAL IMPACTS scenarios for CONDITIONS IMPACTS • greenhouse gases • carbon cycle • agriculture impact assessment. • population • aerosols • sea level rise • temperature • human health • GDP • other drivers, • fresh water • energy system e.g. land-cover • precipitation • ecosystems • property, infrastructure • industry change • humidity • technology • soil moisture • agriculture, • extreme events land-use

Content of Scenarios Use of Scenarios

significant problems of incompatible spatial may be crucial determinants of vulnerability to scale. Impact assessments often examine climate impacts.56 Impact assessments have re- smaller spatial scales than emissions projec- sponded to this dilemma in two broad ways.57 tions, so they may need these socio-economic First, constructing scenarios of relevant socio- data at finer scale than is available. Downscal- economic conditions has been delegated to local ing future socio-economic projections has or regional teams with expertise in the impacts proved challenging thus far. There is no gener- being assessed, subject to constraints to main- ally accepted method for doing so, and several tain consistency with other assessments. Sec- research groups are exploring development of ond, since local or regional scenario groups may alternative methods.55 not have access to all knowledge relevant to understanding the main determinants of impacts, In contrast to the few clearly identified aggre- more open-ended approaches have been em- gate characteristics needed to construct emis- ployed – e.g., exploratory analyses that iterate sions scenarios, the socio-economic factors that between considering particular characteristics most strongly shape adaptive capacity and vul- that might be important, examining their impli- nerability for particular impacts may be de- cations for impacts using the data and models tailed, subtle, and location-specific. It may not available, then re-assessing what variables are even be clear what characteristics are most im- most important. portant before doing a comprehensive analysis of potential causal pathways shaping impacts. This section has sketched a typology of global- The most important characteristics may interact change scenarios and identified major types of strongly with each other or with other economic decision-makers who might use global-change or social trends, or may not be readily quantifi- scenario-based information. The next section able. All these factors make the development turns to current experience with global-change of socio-economic scenarios for impact assess- scenarios, summarizing the development, con- ment a much more difficult endeavor than contents, and uses of four major exercises. In- structing emissions scenarios. formed by these cases plus additional short scenario examples presented in text boxes, Sec- Because scenarios are schematic, not all factors tion 4 will summarize and discuss the major that might be important for impacts can be in- challenges for making and using scenarios that cluded. Details are typically not included or are raised by this experience, providing the basis treated as merely illustrative. But particular de- for the conclusions and recommendations pre- tails, which cannot be identified in advance, sented in Section 5.

55 Toth and Wilbanks 2004, Pitcher 2005. 56 Berkhout et al. 2002. 57 Berkhout et al. 2001, Parson et al. 2001.

30 Global-Change Scenarios - Their Development and Use

Review of 3 Major Climate-Change Scenario Exercises SECTION

In this section, we review experience to date developing and using scenarios for global climate change applications. Because little literature on these activities yet exists, our selection of cases has inevitably been both limited by time and resources at our disposal and reliant in part on the knowledge and experience of team members. We discuss four exercises in detail, in an attempt to cover the largest-scale and most important activities. Section 3.1 reviews the IPCC emission scenarios, with particular detail on the most recent and important exercise, the Special Report on Emissions Scenarios (SRES). Section 3.2 considers the US National Assessment, which developed and used scenarios of both climate and socio-economic conditions. Section 3.3 considers the UK Climate Impacts Programme, which has also both developed and used scenarios, following a different approach from the US National Assessment. Section 3.4 reviews the ambitious scenario-generating exercise conducted as part of the Millennium Ecosystem Assessment (MEA), in which climate change was one of several dimensions of stress considered on global ecosystems.

For each exercise, we consider only the development and use of scenarios, rather than examining the larger assessment processes of which the scenarios were part. We consider how the scenarios were developed, including both methods of reasoning and managerial process; how and by whom they were used; and subsequent evaluations, when these are available. General issues and challenges that emerge from these experiences are discussed in Section 4.

To provide more illustrative variation, we also provide shorter summaries of eight additional scenario activities, some of them related to the four we consider in detail. Presented in text boxes throughout Section 4, these are intended to provide additional information to highlight particular issues. We have particularly sought experiences that illuminate potential relationships between scenarios and decision-making.

All these scenario exercises represent early work in an immature field. Our aim is not to criticize particular exercises, but to seek insights from their experience into the general problems of making useful global-change scenarios.

31 The U.S. Climate Change Science Program Section 3 - Review of Major Climate-Change Scenario Exercises

3.1. IPCC EMISSIONS SCENARIOS IS92a scenario for assuming increasing divergence in the per capita emissions of industrial- Since its establishment in 1989, the IPCC has ized and developing regions, arguing that this organized three exercises to develop scenarios represented a strong bias in favor of already de- of greenhouse-gas emissions, of increasing veloped regions.63 scale and complexity. For its first report, IPCC’s Working Group 3 on “Response Strate- In response, the 1996 plenary session of the gies” included a sub-group on emissions sce- IPCC requested a new set of emissions scenar- narios. Four scenarios were produced but little ios. These new scenarios were to improve treat- used in this assessment because of time limits ment of sulfur aerosols and emissions from and because, with one exception, only doubled- land-use change, and were not to rely on a sin-

CO2 equilibrium climate-model runs were avail- gle model or expert team, but instead to draw able at the time.58 The next exercise produced on the existing literature and invite any group six new scenarios, called IS92a through IS92f.59 with relevant expertise to participate in an These were the first global emissions scenarios “open process.”64 They were also charged to with a full suite of greenhouse gases and at least serve more uses than climate-model inputs, some explicit calculation underlying each. The such as supporting impact analyses, but to as- IS92a scenario, one of the central scenarios in sume no new climate-policy interventions. Al- this group, was used in climate-model compar- though not explicitly in the terms of reference, isons conducted for the 1996 IPCC assessment, it was also clearly understood that the scenarios along with the simpler transient scenario of 1 would address the criticism of the IS92 scenar-

percent annual increase in equivalent-CO2 con- ios by focusing on convergent development centration and further equilibrium runs.60 paths between North and South.

The third and most ambitious IPCC scenario In January 1997 the IPCC established a writing team, process was established in 1997 and worked for including members of several energy-economic two years to produce the Special Report on modeling groups and experts in related areas Emissions Scenarios (SRES).61 In part, this such as population, technological change, and process was established in response to two scenario development methods. The process widely circulated criticisms of the IS92 scenar- ran under tight time pressure to provide prelim- ios. The first of these criticized the 1992 sce- inary scenarios by early 1998 for climate-model narios for inconsistency with other published runs in the IPCC Third Assessment. scenarios of energy and carbon intensity for major world regions; failing to reflect important Prior scenarios were compiled in a web-based recent trends, including the collapse of the So- database,65 and any researcher was invited to viet Union and increasing restrictions on sulfur submit new ones. By mid-1998 the database emissions worldwide; relying inappropriately contained more than 400 scenarios. Most of

on a single model; and only being useful as cli- these projected only energy-related CO2 emis- mate-model inputs, not for other purposes such sions, but they were highly diverse in their cov- as mitigation studies or supporting climate- erage and resolution, the variables included, and change negotiations.62 The second criticized the their methodologies. The usefulness of these scenarios in constructing new ones was limited by several weaknesses, however. Many were in- 58 The scenarios were mentioned in a 1-page Appendix complete, lacked documentation of inputs, or to the Working Group 1 report. The one non-equilib- made inconsistent assumptions. Few included rium run available was a preliminary transient run sulfur or land-use emissions, which were specif- using 1 percent annual CO2 concentration increase. See Mitchell et al. 1990, Bretherton et al. 1990, IPCC 1990. 59 Leggett et al. 1992. 62 Alcamo et al. 1995. 60 The 1 percent scenario was similar to IS92a, but gave 63 Parikh 1992, 1998. total radiative forcing about 20 percent greater by 2100. Washington and Meehl 1989; Stouffer et al. 1989; 64 Nakicenovic and Swart 2000: 324, Appendix I (terms Bretherton et al. 1990:180-182. of reference). 61 Nakicenovic and Swart 2000. 65 Morita and Lee 1998.

32 Global-Change Scenarios - Their Development and Use

ically requested of the new scenarios. Many growth with weaker convergence, and moderate were unclear on whether they assumed mitiga- improvements in energy efficiency and devel- tion efforts, while the new scenarios were in- opment of non-carbon energy sources.68 The structed to exclude them. Consequently, the storylines were elaborated in short text descrip- development of new scenarios had to proceed tions with some preliminary numbers attached largely independently of the collection of exist- in fall 1997.69 ing scenarios through the literature review and open process. Modeling teams were asked to produce initial quantifications of these scenarios in fall 1997, Early on, participants decided to use narrative to match specified 2100 target values within 10 scenarios in addition to quantitative models, and percent. In February 1998, the preliminary to include experts in this approach on the writ- quantitative targets were re-confirmed and mod- ing team. This decision drew on recent suc- elers asked to continue work on quantifications, The SRES interim cesses using such scenarios for energy and now including a breakdown of economic output marker scenarios were environmental applications,66 and responded to into four world regions.70 In April, one model’s used to provide the charge to make the scenarios more inte- quantification was chosen as a “marker sce- emissions scenarios to grated and more broadly useful. Participants in nario” for each of the four scenarios – a partic- an April 1997 workshop chose two key uncer- ular scenario that would provide the basis for climate models tainties to explore in the scenarios: whether interim reporting to climate modelers, some of participating in the world values would mainly stress economic whose results other participating models would IPCC third assessment. prosperity or balance economic and ecological be asked to replicate. The specifications and concerns (labeled “A” vs. “B” scenarios); and models for these marker scenarios are shown in whether the organization of economies and in- Table 3.1. stitutions would continue shifting toward global integration, or reverse and move toward regional These interim marker scenarios were used to fragmentation (labeled “1” vs. “2” scenarios).67 provide emissions scenarios to climate models participating in the IPCC third assessment. An Combined, these gave four scenarios, which IPCC climate modelers’ meeting in June 1998 were sketched in preliminary terms at the work- agreed to use SRES scenarios and asked for shop. In the A1 (economic, global) scenario, three cases, central emissions, stabilization, and economic growth and inter-regional income high emissions.71 The writing team initially dis- convergence continue strongly worldwide – all cussed meeting this request by identifying sce- developing countries grow like Japan and Korea narios corresponding to each of these requested from the 1950s to the 1980s – while world pop- cases,72 but decided to provide only the marker ulation peaks at 9 billion by 2050. Rapid inno- scenarios and recommend that all four be used vation yields many advanced energy sources, without identifying any as “central.” while acid rain and other local and regional environmental problems are aggressively controlled. In contrast, the A2 (economic, regional) scenario has higher population growth, lower 68 Arnell et al. 2004; Minutes, Lead Authors Meeting, economic growth with more continuing re- Paris, April 13-15, 1997. gional disparities, slower innovation, and 69 Minutes, informal modelers meeting, Berkeley, Feb weaker international institutions. B1 (ecologi- 7-8. cal, global) has low population growth, moder- 70 Draft minutes, informal modelers meeting, Berkeley, ate economic growth with strong convergence, Feb 7-8:4. and strong reductions in per capita energy use, 71 Minutes of the Laxenburg meeting, July 2-3, 1998, mostly through higher efficiency, while B2 has reporting results of June 29-July 1 IPCC Scoping intermediate population growth, low economic Meeting, Bonn. 72 In July 1998, members decided that A1F or A2 could be the requested high-emissions scenario (with emissions of ~ 30 GtC in 2100), B2 or A1B a central case (~15 GtC in 2100, with two different SO2 profiles), 66 See, e.g., WEC/IIASA 1995, WBCSD 1997. and B1 or an A1 variant called A1R a stabilization case 67 Minutes, Lead Authors Meeting, Paris, April 13-15, (at about 550 ppm) (Laxenburg report, July 2-3, 1997. 1998:1).

33 The U.S. Climate Change Science Program Section 3 - Review of Major Climate-Change Scenario Exercises

explicit mitigation, including one stabilization Table 3.1. SCENARIO A1B A2 B1 B2 Target Values scenario.75 At the final IPCC approval meeting, for 2100 in Population 7.1 15.1 7.1 10.4 it was decided at the request of the Saudi dele- Initial Scenario gation to reduce the two fossil-intensive vari- Quantifications GDP (trillion) $530 $250 $340 $235 ants of A1 to one, a variant of the gas-intensive Final Energy (EJ) ~1,700 870 770 950 scenario which was renamed A1FI (for “fossil- intensive”).76 Model for AIM ASF IMAGE MESSAGE Marker scenario 3.1.1 Significance and use Source: Minutes of Laxenburg meeting, July 2-3, 1998. The SRES scenarios have been the most comprehensive, ambitious, and carefully docu- The marker scenarios also provided the basis for mented emissions scenarios produced to date. coordination of subsequent scenario develop- They represented a substantial advance from ment. Up to this point, there had been substan- prior scenarios, and contributed to assessments tial discrepancy between different models’ and subsequent research on climate impacts and quantifications of the same scenarios, particu- responses. The SRES scenarios formed the larly at regional level. With the adoption of the basis for climate-model comparisons in the markers, other groups were asked to replicate IPCC Third Assessment (2001) and continuing The SRES scenarios (within 5-10 percent) the marker results on pop- work in the Fourth Assessment. Most subse- ulation, GDP, and final energy for the four quent climate-model work has used only a few have been the most world regions, for 2100 and several interim of the marker scenarios – typically A2 and B2, comprehensive, years.73 Achieving this requested replication sometimes with A1B added. SRES scenarios ambitious, and posed significant challenges for modelers.74 also provided baselines for analysis of mitiga- carefully documented tion scenarios in the Third Assessment.77 emissions scenarios With a further year of work, modeling teams produced to date. produced a total of 40 scenarios that were re- Several significant insights were illuminated by tained in the report, of which 26 replicated one the SRES scenarios. of the marker scenarios. Although a few of the 14 non-replicates reflected a model’s inability • Scenarios with similar emissions in 2100 to match the results of a marker scenario, most can follow markedly different paths in the were produced because a modeling team inten- interim, giving wide differences in cumula- tionally sought to explore alternative assump- tive emissions and concentrations. tions. For example, the A1 scenario, which originally balanced fossil and non-fossil energy • Technology and energy-resource assump- sources, was augmented by variants with dif- tions can strongly perturb future emissions, ferent assumptions about fossil resources and even with constant socio-economic assump- non-fossil technology development, giving tions. For example, the three A1 variants widely divergent emissions paths stressing coal, show that changing these assumptions alone gas, and non-fossil energy technology. Modifi- can generate as wide a range of emissions cations of the scenario set continued until late futures as substantial variation of demo- in the process. For example, it was decided in graphic and economic futures. October 1998 to drop several B variants with

75 SRES modelers meeting report, Beijing, October 6- 7, 1998:4. At this meeting, B1 was also proposed for 73 Because markers were produced by different models removal, but was retained based on a decision that none with different time steps, the interim years to be har- of the many policy interventions it presumed was an monized differed for each scenario. explicit greenhouse-gas limitation, so it was consistent 74 For example, discussions in Beijing re-confirmed that with the terms of reference. allowed deviation from markers at 4-region level would 76 A1FI was the gas-intensive scenario, A1G, with re- be 5 percent for GDP and 10 percent for final energy, visions to methane emissions and additional non-CO but substantial discrepancies in base-year energy could 2 gases added from the A1 run of the MESSAGE model. not be harmonized due to time constraints (SRES modelers meeting report, Beijing, October 6-7, 1998:2). 77 Morita et al. 2001.

34 Global-Change Scenarios - Their Development and Use

• Highly distinct combinations of demo- They were sharply criticized for this decision.80 graphic, socio-economic, and energy-mar- Critics argued that there were no technical ob- ket conditions can produce similar stacles to assigning probabilities to emissions emissions trajectories, suggesting that a par- ranges bounded by the marker scenarios; that ticular emissions trajectory can pose very scenario developers must have made proba- different types of mitigation problems, de- bilistic judgments in generating and evaluating pending on what combination of driving fac- the scenario quantifications and that not mak- tors underlies the emissions. ing these judgments explicit would withhold relevant information; and that if scenario devel- 3.1.2 Criticisms and controversies opers decline to assign probabilities, others who are less informed will do so. Indeed, many The SRES experience raised issues of great sig- probabilistic emissions calculations have been nificance for subsequent attempts to develop produced since the SRES, using various meth- It may well be more useful climate-change scenarios: the de- ods such as assigning uniform or other speci- unhelpful to assign sirability of and appropriate methods for char- fied distributions over the emissions range of probabilities to rich, acterizing probabilities associated with the marker scenarios, counting scenarios lying multi-dimensional scenarios; the quantitative representation of the in specified intervals in the larger SRES set, un- relationship between North and South; methods bundling and recombining alternative values of narrative scenarios, for developing and using narrative scenarios the drivers underlying SRES emissions figures, yet still useful to and integrating them with quantitative model re- or sampling over parameter distributions within assign interval sults; alternative modes for coordinating use of a single model. In response to these criticisms, probabilities when multiple models and their implications for the SRES authors argued that attempting to assign scenarios principally interpretation and use of scenarios; and the re- probabilities to scenarios would require assign- represent uncertainty lationship between scenario exercises and their ing joint distributions to the underlying driving in one or two users, including the need for clarity about spe- factors, and that this would lead to an explosion quantitative variables cific intended uses, appropriate methods for en- of combinatoric possibilities over which any at- gaging users in scenario development, and how tempt to assign probabilities would be spurious to improve utility of scenarios when not all po- and arbitrary.81 tential user groups are specifically identified. These are discussed in Sections 4 and 5. The situation of the SRES scenarios is in fact more nuanced than the arguments of either their The first two of these issues were the subjects of authors or critics would suggest. It may well be forceful public criticisms. We discuss these, fol- unhelpful to assign probabilities to rich, multi- lowed by several other issues that have received dimensional narrative scenarios, yet still useful less attention but which in our view pose more to assign interval probabilities when scenarios central and instructive challenges for future sce- principally represent uncertainty in one or two nario exercises. quantitative variables. And while the SRES scenarios began their lives like the former type of Assigning explicit probabilities storyline scenario, they finished more like the latter. For many users, the scenarios are their The SRES team decided at the outset to make projections of greenhouse-gas emission trends. no probabilistic statements about the scenarios. When they are viewed in this way, a potential Their report used great care in its language to user may reasonably ask, how likely are emis- avoid any suggestion that one scenario might be sions to be higher than this – a distinct and bet- more central or more likely than any other.78 ter-posed question than, what is the probability This decision was consistent both with standard of an A1 world? The uncertainty issue has no practice in developing narrative scenarios, and clear resolution in this case, and poses hard de- with the instruction in their terms of reference sign problem for scenarios and assessments not to favor any model.79 more broadly. Although the SRES exercise has

78 E.g., Minutes of London meeting, March 1999. 80 Schneider 2001, 2002; Pittock et al. 2001; Allen et al. 2001; Reilly et al. 2001. 79 Draft minutes of the Washington, DC, meeting, April 29-30, 1998:6. 81 Grubler and Nakicenovic 2001.

35 The U.S. Climate Change Science Program Section 3 - Review of Major Climate-Change Scenario Exercises

raised this controversy most explicitly to date, the A related, more serious concern is that all SRES problem is a general one that any scenario exercise scenarios assumed varying degrees of real in- must confront. We discuss it further in section 4.6. come convergence between North and South; this was done in response to criticisms that the Exchange rates: PPP versus MER IS92 scenarios were biased in favor of the North. But an exercise to construct potential The most prominently publicized criticism of climate-change futures may need to consider SRES focused on the fact that all but one of the less optimistic and less desirable futures in participating models compared GDP across re- which some currently poor regions fail to solve gions using market exchange rates (MER), in- the development problem. Not considering less stead of the more correct purchasing-power fortunate futures, including ones that might parity (PPP) approach. PPP comparisons cor- challenge the adequacy of current responses, in- …while MER rect for price differences among countries, provid- stitutions, and decision-making capacity, may ing a more accurate comparison of real incomes. limit scenarios’ usefulness in supporting long- overstates future Because lower-income countries have lower price term risk assessment and planning for the soci- income growth in poor levels, MER-based comparisons overstate the in- etal response to climate change. countries, it also come gap between rich and poor countries. overstates future Underdevelopment reductions in energy In a series of letters to the IPCC chairman and of narrative scenarios and emissions intensity. several subsequent publications, two critics argued that the use of MER caused SRES scenar- Although the SRES storylines were produced ios to over-estimate future income growth in first and featured prominently in publications, developing countries (because they over-esti- they remained underdeveloped and underused mated the initial income gap), and consequently throughout the process. In part due to time to over-estimate future emissions growth. Their pressure, in part due to the predominance of criticism was widely circulated and repeated by quantitative modelers in the process, little at- prominent climate-change skeptics.82 tention was given to further development of the storylines once initial quantifications were es- But, although using MER does overstate future tablished and modeling work began. Nor was income growth, it does not necessarily follow significant effort devoted to integration and that future projections of emissions growth are cross-checking between storylines and quanti- also overstated. MER is universally recognized tative scenarios, although a major purpose of as a flawed measure of income, whose use in the narratives was to give coherent structure to global-change scenarios is only justified by bet- quantifications.84 Concerns raised about the ter availability of current and historical data, storylines included lacking specification of and the fact that international emissions trades characteristics other than those needed to gen- in any future mitigation regime will likely be erate emissions; imbalance between the story- made at market exchange rates. But changing lines, with A1 much more developed than the the measure of income also changes the rela- others and B2, the least developed, likely to be tionship between income and such physical heavily used as the median scenario for emis- quantities as energy and food consumption, sions; apparent inconsistencies within A2; and which determine emissions. Consequently, lack of clarity regarding the distinctions be- while MER overstates future income growth in tween A2 and B2 – a serious enough concern poor countries, it also overstates future reduc- that merging them was repeatedly considered tions in energy and emissions intensity. These until late in the process.85 opposing errors are likely to be similar in size, in which case any error in emissions projections from using MER will be small.83 84 Minutes of the Beijing meeting, October 6-7, 82 Castles and Henderson 2003a, 2003b; The Economist 1998:10. 2003a,b; Michaels 2003b. 85 Draft minutes of the Bilthoven meeting, September 83 Nakicenovic et al. 2003, McKibben et al. 2004, 17-19, 1997:7-8; draft minutes of the Berkeley meet- Holtsmark and Alfsen 2005, Manne et al. 2005, Gr bler ing, February 7-8, 1997:6; draft minutes of the Wash- et al. 2004. ington, DC, meeting, April 29-30, 1998.

36 Global-Change Scenarios - Their Development and Use

Moreover, participants disagreed over the mean- process, only population was exogenous for all ing of some of the scenarios, as indicated by the participating models. Because GDP and final persistent difficulty they had in agreeing on de- energy were endogenous for some or all partic- scriptive names.86 These names were dropped ipating models, matching their specified values late in the project, in the context of a broad re- required manipulating other internal model treat from attempting to flesh out the storylines. characteristics. Once one model run was cho- That so little integration of qualitative and sen as the marker for each scenario, subsequent quantitative components was achieved in spite attempts by other models to replicate the results of serious and persistent efforts suggests the posed the same problem more acutely, since magnitude of the analytical and methodological more outputs were specified at this point. challenges involved. The problems associated with attempting to har- Harmonizing scenarios, interpreting monize model outputs are related to the under- the results development of narrative scenarios and limited integration of qualitative and quantitative com- Scenario exercise that use multiple models can ponents. The storylines were associated with coordinate them in several ways: choosing one relatively restrictive numerical targets even or a few illustrative scenarios as coordinating though the storylines did not develop the rich- devices for subsequent analyses, as was done ness or coherence that would carry implications with the SRES marker scenarios; fixing values for additional characteristics. The preliminary of a small set of exogenous inputs to multiple targets were only slightly modified throughout models, to characterize resultant uncertainties the project, despite subsequent discovery of and examine their origins through focused significant problems. For example, the United model intercomparisons; or fixing key outputs Nations 1998 population projections, with sub- as targets, to reason backwards and examine re- stantial reductions in projected fertility, were quirements for achieving them. completed while the scenario development work was underway but not incorporated.87 Choosing a few quantitative variables as the initial link between storylines and models makes Clarity about uses, involving users: these variables serve as a framework to capture the storylines’ basic logical structure. Although The SRES scenarios were charged with serving these choices are not obvious, the variables cho- uses beyond driving climate models but given sen here appear reasonable. But the causal little guidance on what specific additional uses structure of a model will not generally mirror or users to serve, or how the scenarios might the presumed causal logic of a narrative, so a best serve them, neither of which is obvious.88 model cannot be expected to calculate values Providing climate-model inputs remained the for other variables that flesh out the storyline most prominent and most clearly specified use, logic. Moreover, the few key variables so cho- as well as the only use that had an early dead- sen may not be exogenous inputs for every line. But climate modelers were not involved in model used in the subsequent quantification. the scenario development process, and there Of the three variables specified in the SRES was substantial divergence between their needs and the outputs and capabilities of the SRES process. A September 1997 briefing identified 86 While names proposed for the “1” storylines suggest substantial common understanding (A1 was called the principal needs of climate modelers as early “High Growth,” “Productivity,” and “Golden Eco- availability of scenarios and greater emissions nomic Age,” B1 was “Green” and “Sustainable devel- detail.89 They wanted separate emissions tra- opment”), names proposed for the “2” scenarios, jectories for major greenhouse gases, not just particularly B2, do not (A2 was called “Regional Con- solidation,” Divided World,” and “Clash of Civiliza- tions”; B2, “Regional Stewardship,” “Small Is Beautiful,” “Dynamics as Usual,” “Gradually Better,” 87 Minutes of the Bilthoven meeting, September 17-19, and “Muddling Through”) (draft minutes of the 1997:11. Bilthoven meeting, September 17-19:7-8; draft min- 88 Alcamo et al. 1995. utes of the Berkeley meeting, February 7-8, 1997; UKCIP 1998 report summarizing SRES progress; 89 Draft minutes of the Bilthoven meeting, September Pitcher 1998 presentation slides. 17-19, 1997:5.

37 The U.S. Climate Change Science Program Section 3 - Review of Major Climate-Change Scenario Exercises

CO2-equivalent, including regional detail for 3.2. THE US NATIONAL some emissions such as sulfur – even suggest- ASSESSMENT ing that it would be desirable to have sulfur emissions disaggregated by stack height, to dis- The US National Assessment was the most tinguish dispersed emissions from large point comprehensive attempt to date to assess climate sources. Although SRES provided gridded sul- impacts on the United States over 25-year and fur data by post-processing model outputs, in 100-year horizons, and to consider both major most cases the emissions included and their spa- sub-national regions and sectors.93 Responding tial detail (not to mention stack height) were to a requirement in the 1990 Global Change Re- limited by the capabilities and structures of par- search Act, the National Assessment was or- ticipating models. ganized by federal agencies participating in the US Global Change Research Program. Work The US National Other uses received less attention, and repre- began in 1997, with various components com- sentatives of other potential uses were even less pleted between 1999 and 2003. Regional im- Assessment was the involved than climate modelers in the process. pacts were initially considered in 20 regional most comprehensive Supporting assessment of mitigation strategies workshops, followed by more extended analy- attempt to date to was largely deferred to the post-SRES scenarios ses of impacts, leading to published assessments assess climate impacts prepared for the IPCC Third Assessment Re- for 12 regions, conducted by university-based on the United States port, although ambiguity about the degree of teams. Sectoral impacts were examined by na- over 25-year and 100- mitigation effort implied by some SRES sce- tional teams focusing on agriculture, water, year horizons, and to narios complicated that task. Impact and vul- human health, coastal areas and marine re- consider both major nerability assessments depend on diverse, sources, and forests. A federal advisory com- sub-national regions small-scale socio-economic and ecological fac- mittee, the National Assessment Synthesis Team and sectors. tors that a global exercise centered on energy- (NAST), provided direction for the assessment economic models cannot provide.90 For the and synthesized its results in two published re- population and economic projections that were ports.94 Roughly two thousand experts and provided in the course of generating emissions stakeholders participated. scenarios, the key issue for impacts and adaptation was the degree of spatial detail provided. As an assessment focused on climate impacts, For consistency among scenarios, and to avoid the National Assessment needed both climate base-year discrepancies with national and re- scenarios and scenarios of potential future gional datasets, SRES scenario results were re- socio-economic conditions over the 21st century, ported only for four large world regions. since substantial changes are likely over this pe- Greater regional detail was available from indi- riod in socio-economic conditions that might in- vidual models, but with inconsistent regional fluence vulnerability to climate and adaptive boundaries. Providing the greater regional de- capacity. tail desired for impact assessments would generate discrepancies between the global-model 3.2.1. Emission and climate results represented in scenarios and the more de- scenarios tailed data and projections available at national and regional levels.91 Developing valid methods For climate scenarios, the National Assessment to downscale socio-economic scenario informa- relied predominantly on data and model results tion and integrate it with national and regional previously produced. Study teams conducted datasets remains a key challenge for producing additional checking, processing, documenta- useful scenarios for impact assessment.92 tion, and dissemination as needed to make these

93 There had been two previous assessments of US cli- 90 See, e.g., discussion with Mike Hulme on behalf of mate impacts. The US EPA (1989) did a preliminary TGICA, draft minutes of the Washington, DC, meet- assessment for five representative US regions and five ing, April 29-30, 1998:9. sectors (agriculture, forests, water resources, health, 91 January 1998, meeting with Richard Moss, WG2 and coasts), while the US OTA (1993) examined im- Technical Support Unit, described in draft meetings of pacts for six sectors – coasts, water, agriculture, wet- the Berkeley meeting, February 7-8, 1997. lands, protected areas, and forests. 92 Pitcher 2005. 94 NAST 2000, 2001.

38 Global-Change Scenarios - Their Development and Use usable. The assessment encouraged the use of and interpolated to the finer time and spa- three types of climate scenarios: historical sce- tial scales needed for impact studies narios produced by extrapolating observed trends or re-imposing historical climate vari- 7. Be based on well-documented models par- ability or extremes; an inverse approach using ticipating in the IPCC Third Assessment (for sensitivity analyses to explore the responses of comparability between US and international climate-sensitive systems, with particular em- efforts) phasis on thresholds defining key vulnerabilities; and climate model simulations of future 8. Be able to interface results with higher-res- climate conditions.95 olution regional model studies

Of these three approaches, the climate-model 9. Provide a comprehensive array of results scenarios were the most precisely specified and openly over the internet. Of the three the most widely used. The National Assessment approaches, the did not have the resources or time to commis- To ensure timely dissemination, the National climate-model sion new climate model runs and so had to rely Assessment chose climate-model scenarios to scenarios were the on those completed and published when it be used in its analyses in mid-1998. At that began its work. A set of criteria was developed time, only two groups had completed runs that most precisely by the NAST for the climate model scenarios to met most of the key criteria: the UK Hadley specified and the most be used in the assessment. Climate-model sce- Centre (Model Version 2) and the Canadian widely used. narios used in the assessment should, to the Centre for Climate Modeling and Analysis greatest extent possible:96 (Model Version 1).97 All participating regional and sector teams were asked to use these sce- 1. Include comprehensive representations of narios. The climate sensitivity of these models the atmosphere, oceans, and land surface, was 2.5°C (UK Hadley) and 3.6°C (Canadian), and key feedbacks among them lying in the middle of the 1.7 to 4.2°C range of sensitivities represented by models participat- 2. Simulate the climate from 1900 to 2100, ing in the IPCC Third Assessment.98 based on a well-documented emissions scenario that includes greenhouse gases and aerosols These two models were limited in their ability to reproduce observed patterns of inter-annual and 3. Have the finest practicable spatial and tem- inter-decadal climate variability. But other cli- poral resolution, with grid cells of less than mate-model runs available at the time failed to 5˚ latitude x longitude meet essential requirements of the ecosystem models that were the basis for an important part 4. Include the daily cycle of solar radiation, to of the assessment: availability of documented allow projections of daily maximum and results, projections to 2100, standard/compara- minimum temperatures ble emissions scenarios, and explicit treatment of the day-night cycle. 5. Be able to represent significant aspects of climate variability such as the El Niño- For these two climate models, model runs using Southern Oscillation (ENSO) cycle only one emissions scenario were available, and only one ensemble run was used for each.99 The 6. Be completed in time to be quality-checked

97 Johns et al. 1997; Boer et al. 1999a, 1999b; Mac- Cracken et al. 2003. 95 NAST 2001:25. It is arguable whether or not the inverse approach involves scenarios by the definition we 98 Cubasch et al. 2001, Table 9.1:538-540; and Table have adopted here, because it does not stipulate speci- 9A.1:577. fied future climate conditions, but attempts to identify 99 Ensembles of climate-model runs are repeated simu- them from presumed thresholds or breakpoints. How- lations with small variations in initial conditions which ever, we are following the usage of the NAST reports improve the characterization of climate variability. The in calling these approaches three types of scenarios. Canadian group had completed only one ensemble run 96 NAST 2001:31-32; MacCracken et al. 2001; Mac- at this time. The Hadley Center had completed three, Cracken et al. 2003:1714. but the National Assessment was only able to use one.

39 The U.S. Climate Change Science Program Section 3 - Review of Major Climate-Change Scenario Exercises

emissions scenario was IS92a, which repre- warming by 2100 of 5.0°C (Canadian) and 2.6°C sented the middle of the range of IPCC’s 1992 (Hadley), at the high end and below the middle, scenarios.100 In addition to greenhouse gases, respectively, of the range of projections in the the scenario included atmospheric loadings of IPCC Third Assessment.105 In their projections of sulfate aerosols, which were assumed to in- precipitation change over the United States, these crease sharply through 2050 and then level off scenarios both lay at the high end – the Hadley for the rest of the 21st century.101 scenario projected the highest precipitation in 2100 and the Canadian the second-highest106 – but The applicability of these two scenarios was the Canadian model’s greater warming offset the tested by checking the models’ ability to repli- effect of this precipitation increase on soil mois- cate broad patterns of US climate change over ture, which was projected to decrease over most of the 20th century when driven by historical green- the continental United States.107 house-gas forcings. Model results were compared against the VEMAP (Vegetation- Although only the Hadley and Canadian cli- Ecosystem Mapping and Analysis Project) mate-model scenarios were used throughout the dataset, a corrected climatic dataset for the 20th assessment, several others that met some or all century. This comparison showed reasonable of the assessment’s needs became available dur- accuracy in reproducing the spatial distribution ing its work. Several region and sector teams of average temperatures and century-long tem- were able to use these additional scenarios. In perature trends, but significantly weaker repro- some cases, the additional scenarios allowed duction of observed patterns of precipitation, groups to strengthen their conclusions. For ex- mainly because the spatial distribution of pre- ample, an analysis of future Great Lakes water cipitation depends on topographic detail that is levels under climate change using eleven cli- too fine-scale to be captured even by the 0.5- mate models found that ten of these showed degree VEMAP grid.102 lower levels and only one higher.108 In other cases, using multiple models allowed more de- With the specified scenario of future emissions, tailed characterization of uncertainties in future the two climate-model scenarios projected regional changes. For example, the Pacific global warming by 2100 of 4.2°C (Canadian) Northwest team presented distributions of re- and 2.6°C (Hadley).103 These projections were gional temperature and precipitation change in at the high end and in the middle, respectively, the 2030s and 2090s using seven GCMs.109 of the range of warming projected for this emissions scenario by models participating in the Despite the National Assessment’s aim of ex- IPCC Third Assessment Report.104 For the con- ploring future climate using three distinct types tinental United States, the two models projected of scenario, historical scenarios and sensitivity analyses were less extensively used than GCM scenarios and featured less prominently in the 100 The IS92a scenario is described in section 3.1. There were small differences among climate-modeling 105 The seven models for which these results were avail- groups in the way they converted emissions trajecto- able clustered at the top and the bottom. Three of them ries into atmospheric concentrations and radiative forc- – the Canadian, GFDL, and Hadley 3 models – lay very ings, making the actual scenarios driving each model close together at the high end, the Canadian the high- run very close, but not quite identical. est by a fraction of a degree; three others lay close together at the low end, Hadley 2 the highest of them by 101 See www.usgcrp.gov/usgcrp/nacc/background/sce- somewhat less than a degree. A seventh model, narios/emissions.html for further detail on emissions ECHAM4, tracked the high group through 2050, the scenarios used in the National Assessment. last year for which its results were available. Since 102 VEMAP members 1995, Kittel et al. 1995. these comparisons usually reflect only one ensemble run of each model, small differences between runs may 103 NAST 2001:36, Table 2. reflect consistent inter-model differences, or noise re- 104 Cubasch et al. 2001, Figure 9.5a:541. While the flected in a single ensemble run. See NAST 2001:547, Canadian model lies at the high end, it is not an outlier. Figure 7. The GFDL model (which was more responsive than 106 NAST 2001:545, Figure 8. the Canadian model, with a climate sensitivity of 4.2° C) projected higher global warming than the Canadian 107 NAST 2001:552, Figure 16 and 18. model in this scenario for the first few decades of the 108 Lofgren et al. 2000; NAST 2001:175. century, but only had results through 2060 in time for the TAR. 109 NAST 2001:256.

40 Global-Change Scenarios - Their Development and Use

assessment’s publications. Two uses of histori- 3.2.2. Socio-economic scenarios cal climate data – describing observed impacts of climate variability and using observed his- As discussed in Section 2.5 above, assessing torical extremes as benchmarks to compare pro- impacts of future climate change can require jected future changes – were made by all specifying not just scenarios of future climate, groups. To support systematic use of historical but also socio-economic characteristics of the scenarios, the VEMAP 20th-century dataset was future society that will experience the changed provided to all groups, but no further guidance climate. Specifying future socio-economic con- was provided on how to generate climate sce- ditions might be necessary for two reasons. narios from these historical data, e.g., on what First, socio-economic conditions may influence periods to choose or how to use them to assess the demands placed on particular resources that potential future impacts. Several groups used are also sensitive to climate change, the value these historical data to describe the impacts of assigned to them, and the non-climatic stresses particular recognized patterns of climate vari- imposed on them. For example, future flow ability, such as ENSO or the Pacific Decadal regimes in river systems will be influenced by Oscillation (PDO).110 Many groups examined upstream demands for municipal and irrigation past climate extremes, but only in qualitative water use, in addition to the changes caused by ways; most did not follow the approach, taken in climate. Socio-economic scenarios are also some previous impact studies, of using histori- needed to assess climate-change impacts on Socio-economic cal extreme periods as quantitative proxies for human communities – e.g., economic impacts scenarios are also potential future climate.111 and their distribution, human health effects, and needed to assess vulnerability to extreme events – because socio- climate-change The third approach, vulnerability analysis, was economic characteristics of a community expe- impacts on human the least used in the National Assessment. This riencing a changed climate will strongly communities. “inverse” approach involves describing the influence the community’s vulnerability to properties of a climate-sensitive system, speci- changes and its capacity to adapt. fying some important change or disruption, and asking what climate changes would be required In contrast to climate scenarios, little prior to bring about that disruption and how likely – information or experience was available on con- based on historical data and model calculations structing scenarios of socio-economic condi- – such climate changes appear to be. Given the tions for impact assessment. Consequently, the complex dynamics of climate-sensitive systems assessment developed new methods, using an and models of these systems, and the multiple approach that combined centralized and decen- dimensions of climate on which these can tralized elements. Centralization was needed depend, this approach requires a substantial pro- because a few variables, such as population, gram of new research, analysis, and method- economic growth, and employment, are likely ological development.112 In part because of the to be important in all regions and sectors. For intrinsic difficulty of this task – and in part due these, consistent assumptions are required to to management and resource problems – this allow comparison of impacts across regions and approach was not pursued. The NAST pro- sectors, and to aggregate from separate assess- posed it, but more tractable approaches to ments up to overall national impacts. A NAST analyzing climate impacts dominated the as- sub-group developed high, medium, and low- sessment’s work. This remains an important growth scenarios of these variables at the na- area for further work in development of assess- tional level. These followed the US Census ment and modeling methods. Bureau high, middle, and low scenarios for fertility and mortality through 2030, but assumed a wider range of values for net immigration to 113 110 E.g., Mote et al. 2003, Southeast Regional Assess- account for possible illegal immigration. ment Team 2002. Over this period, national population, GDP, and 111 Rosenberg et al. 1993. employment were disaggregated among regions 112 For an example of such efforts, see the AIACC (As- and sectors using a commercial regional eco- sessments of Impacts and Adaptations to Climate Change) project, information at http://www.aiaccpro- ject.org. 113 Parson et al. 2001:102-103.

41 The U.S. Climate Change Science Program Section 3 - Review of Major Climate-Change Scenario Exercises

nomic model.114 Beyond 2030, national projec- Several factors contributed to this limited use of tions of these variables followed OECD growth socio-economic scenarios. In addition to vari- rates in the SRES marker scenarios.115 ous managerial and communication problems, many participants were reluctant to use socio- Decentralization was also needed because the economic scenarios, especially the proposed de- particular socio-economic characteristics that centralized approach. Some preferred to avoid most strongly influence climate impacts and any socio-economic projections, implicitly pre- vulnerability may differ among regions, activi- suming either that socio-economic conditions ties, and resources. For example, major socio- did not matter for impacts, or that those that did economic determinants of climate impacts on matter would remain similar to present condi- Great Plains agriculture may include the crops tions. Others objected to specific contents grown, the extent of irrigation, and the tech- of the scenarios or the methods used to gen- More useful nologies used to provide it, while the main de- erate them, or judged that their team lacked terminants of coastal-zone impacts may be the expertise required to evaluate them. Still assessments of impacts patterns of coastal development, zoning, infra- others objected that uncertainties in future and vulnerability will structure, and local property values. The NAST socio-economic conditions made any attempt to require more extensive judged that those assessing regional or sector construct scenarios for more than a few years in use of socio-economic impacts were likely to know more about such the future unacceptably speculative.117 Conse- scenarios, improved factors than a central body. Consequently, to quently, while the assessment attempted to ad- integration of socio- support decentralized scenario development, the vance scenario methods, weak implementation economic with climatic NAST proposed a consistent template for as- of these methods limited its ability to identify and environmental sessment teams to follow in creating their own key vulnerabilities. More useful assessments of scenarios, and scenarios. Teams were asked to identify two impacts and vulnerability will require more ex- substantial further socio-economic factors they judged most im- tensive use of socio-economic scenarios, im- investment in portant for their impacts of concern; to identify proved integration of socio-economic with development and a range of these factors to represent roughly 90 climatic and environmental scenarios, and sub- testing of new percent confidence; and to create socio-eco- stantial further investment in development and nomic scenarios by combining high and low testing of new methods.118 methods. values of these factors, plus middle or best- guess values if they so chose. 3.2.3. Criticisms and controversies

Implementation of socio-economic scenarios in The National Assessment was the object of sub- the National Assessment was weak. Few as- stantial political and scientific controversy. sessment teams used the proposed approach. Here, we summarize the major criticisms that Many made no socio-economic projections at pertain to the development and use of scenar- all, but rather projected only biophysical im- ios. Criticisms focused predominantly on the pacts based on GCM results. One assessment climate scenarios, especially those derived from team found the socio-economic scenarios were GCMs, probably because these were more pre- inconsistent with superior local estimates of cisely defined, widely used in the analyses, and current population, and so decided not to use featured in the assessment’s publications. Three them.116 The teams that did use the socio-eco- criticisms of these were advanced. nomic scenarios used only aggregate projections of population and economic growth, or in The first, widely circulated during 2000, was some cases assumed continuation of present that the use of non-American climate models conditions in the assessment period. None used for climate scenarios was inappropriate and po- the proposed template for identifying and pro- tentially injurious to national interests.119 While jecting additional important socio-economic this criticism indicates a dimension of political characteristics.

117 Morgan et al. 2005. 114 Terleckyj 1999a,b. 118 Lorenzoni et al. 2000, Berkhout and Hertin 2000, 115 The high-growth scenario was roughly comparable Parson et al. 2003. with A1, medium with B1, and low with A2 and B2. 119 Congressional Record, June 16, 2001, Statements of 116 Rosenzweig and Solecki 2001. Senators Hagel (page S5292) and Craig (page S5294).

42 Global-Change Scenarios - Their Development and Use vulnerability of the assessment, it does not ad- range of future climate-change uncertainty ex- dress the assessment’s technical quality. Since tends both below the Hadley scenario and – in climate models represent the physics of the a long, thin tail – above the Canadian. global atmosphere, they contain no representation of political or economic factors. The A related criticism of the climate scenarios Hadley and Canadian global climate models claimed that the emissions scenario driving were extensively documented in peer-reviewed them was implausibly high. The issues bearing scientific literature – and, moreover, were the on choice of emission scenarios are similar to only models that met the most critical of the as- those for choice of climate models. It would be sessment’s criteria. That they were developed preferable to have a wide and relevant range of by scientific groups outside the United States emissions scenarios driving an impact assess- has no significance for their ability to provide ment – at least for the post-2050 period. Using scenarios to assess US impacts. Using US mod- a wide range of emissions scenarios would also els would have avoided this criticism, but at the allow comparison of projected impacts under cost of either weakening the analysis by using high and low emissions futures, and so give in- scenarios that did not meet the assessment’s sights into what degree of impacts could be needs, or delaying the work by one to two years. avoided by what degree of mitigation effort. In deciding to proceed with non-US models, as- Model runs with this emissions scenario were sessment organizers judged that these costs all that were available, however. Moreover, were too high. there is no clear basis to reject this particular scenario, since it was the scenario most widely The second major criticism was that the two cli- used in climate-model runs at the time and lies mate-model scenarios used were at the extreme near the middle of the range of both the 1992 end of available models in their projected cli- and the 2001 IPCC scenarios. Finally, there is mate change. This is partially correct. When no support for the claim that this scenario was temperature and precipitation factors are con- chosen with the aim of making 21st-century cli- sidered together (because high precipitation in mate change appear as frightening as possi- some cases may offset the impacts of high tem- ble.121 But, although using just two climate perature), the Canadian scenario lies at the high- models with one emissions scenario was un- impact end – although not an outlier, as other avoidable in this assessment, it still represented IPCC model projections lie close to it – while a serious limitation. With more model simula- the Hadley lies at or somewhat below the mid- tions testing a range of emission scenarios al- dle for most analyses. ready available, future assessments will be able to remedy this deficiency. The National Assessment’s organizers and its critics agreed that using more models would In contrast with the preceding criticisms that the have been preferable, but the assessment was scenarios used in the assessment understated limited by its schedule and its technical re- uncertainty, another criticism focused on the quirements. Given a limit of only two, there can disparities between the two scenarios’ projec- be good reasons to choose one scenario in the tions. Some critics argued that such disparities middle of current projections and one near the – e.g., the Canadian scenario projects the South- top that provides a plausible upper-bound, but eastern states becoming much drier than the the significance of the results must then be com- Hadley model does – show that our limited municated with great care. Some critics sug- knowledge of regional climate change makes gested that presentation of results based on the any attempt to assess future impacts and vul- relatively high Canadian scenario should be nerabilities irresponsible.122 This criticism im- more carefully qualified to highlight its position near the top of current projections.120 Such 121 Michaels 2003a:171-192. qualifications must be crafted very subtly, how- 122 ever, lest they imply these results may safely be Disparities between the two models’ projections were the basis of an unsuccessful lawsuit brought against the ignored, when most analyses suggest the full Assessment under the Federal Data Quality Act (See Competitive Enterprise Institute, “Complaint for De- 120 MIT Integrated Assessment project, comments on clarative Relief,” http://www.cei.org/pdf/3595.pdf, at National Assessment, Aug 11, 2000:15. paragraph 24.)

43 The U.S. Climate Change Science Program Section 3 - Review of Major Climate-Change Scenario Exercises

plies that impact assessment should wait until This combination of centralized and decentral- precise, high-confidence regional climate pro- ized information requirements suggests the jections are available. Since a major purpose of need for a cross-scale organizational structure the assessment was to represent current uncer- for developing and applying scenarios, includ- tainty about climate change and its impacts, ing scenarios of both climate and socio-eco- such discrepancies between model projections nomic conditions. served a valuable purpose, as indications of the uncertainty of projections at the regional scale – 3.3. THE UK CLIMATE IMPACTS particularly when the model disparities had a PROGRAMME clear origin, such as differences in projected jet- stream location. The UK Climate Impacts Programme (UKCIP) was established in April 1997 as one element of Future assessments In sum, the National Assessment’s use of cli- a broad program of scientific research, assess- mate-change scenarios was hampered by the ment, and support for policy-making on climate will need to invest lack of available relevant runs, but reflected an change. The UKCIP supports research and substantial resources adequate attempt to represent then-understood analysis of impacts for particular regions, sec- in developing the state variation in climate projections for the United tors, and activities in the UK. The program pro- of underlying States. Future assessments will need to use vides common datasets and tools, as well as knowledge, models, more climate-model projections – including ongoing support to university researchers and and assessment multiple ensemble runs – informed by a wider organized stakeholder groups in all UK regions. methods for range of relevant emissions scenarios. The Na- As part of its role in stimulating, supporting, integrating socio- tional Assessment attempted to advance the and coordinating decentralized and stakeholder- economic state of the art in using socio-economic scenar- driven impact analyses, the UKCIP has pro- considerations into ios, but achieved only limited success in imple- duced and disseminated three sets of scenarios: assessments of menting its plans. Future assessments will need climate scenarios in 1998 and 2002, and socio- climate impacts. to invest substantial resources in developing the economic scenarios in 2001. state of underlying knowledge, models, and assessment methods for integrating socio-eco- The 1998 climate scenarios provided informa- nomic considerations into assessments of tion only at the rather coarse scale of the Hadley climate impacts. This includes further develop- Centre’s HadCM2 climate model, with four ment of novel approaches to link climate and grid-cells over the entire UK. Four scenarios, socio-economic scenarios, such as the proposed called “high,” “medium-high,” “medium-low,” “inverse” approach to vulnerability analysis. and “low,” combined variation in emissions assumptions with variation in assumed climate The experience of the National Assessment sensitivity. The medium-high and medium-low raises three significant issues for future climate- scenarios both used the HadCM2 model, with a change scenario exercises. First, like several of sensitivity of 2.5°C.123 The medium-high sce- the experiences reviewed here, it illustrates the nario was driven by a 1 percent per year equiv-

difficulty and scale of effort involved in pro- alent-CO2 transient scenario, similar to IS92a. ducing scenario-based assessments. Second, The medium-low scenario was driven by a 0.5

the large required start-up effort and time to percent per year equivalent-CO2 transient sce- build the capacity to conduct such an exercise nario, similar to the lowest IS92 scenario, illustrates the great value of sustaining analytic IS92d. The high and low scenarios used the and institutional capacity over time, rather than same two emissions scenarios driving a simpler relying on separate projects. Such continuity of climate model, whose sensitivity was set at capacity will avoid wasteful repetition of start- 4.5°C for the high scenario and 1.5°C for the up efforts, support accumulation of learning and low. These scenarios were used in an initial im- experience, and develop and maintain the re- pact assessment focusing predominantly on di- quired collaborative networks. Finally, the as- rect biophysical impacts.124 The scenarios did sessment’s experience illustrates both the need not include any explicit statements of probabil- for consistency in large-scale assessments, and the great specificity of information needs within 123 UKCIP 1998:13-15. particular impact and adaptation assessments. 124 UKCIP 2000.

44 Global-Change Scenarios - Their Development and Use

ity, although their documentation suggested that economic characteristics qualitatively, includ- the medium-high and medium-low scenarios ing multiple aspects of economic development, “in one sense … may be seen as being equally settlement and planning, values and policy, agri- likely,”125 while the high and low scenarios cap- culture, water, biodiversity, coastal zone devel- tured part of the tails of the distribution. Nor opment, and the built environment.128 did they include any potential extreme climate events such as those associated with large Each scenario was also realized in projections changes in the North Atlantic circulation. of multiple quantitative variables, at the national scale only. For the 2020s, these provided detail The UKCIP’s socio-economic scenarios drew on population, GDP (including the governmen- on the Foresight Program, a broader exercise of tal share and the sector split between industry, the UK Department of Trade and Industry to de- agriculture, and services); household numbers velop scenarios for long-range planning in sev- and average household size; land use and rates The UKCIP’s socio- eral policy areas, with additional detail in areas of change; total transport and modal split; agri- economic scenarios relevant to greenhouse-gas emissions and cli- cultural production (including such details as drew on the Foresight mate impacts.126 As in several other scenario chemical and financial inputs, subsidies, yields, Program, a broader exercises, developers identified two fundamen- and organic area); freshwater supply, demand, tal uncertainties and combined two alternative and quality; and several indicators of biodiver- exercise of the UK outcomes of each to produce four scenarios. sity and coastal vulnerability. For the 2050s a Department of Trade The two core uncertainties they chose were sim- smaller set of quantitative variables was pro- and Industry to ilar to those used in the SRES exercise: social jected, describing population, GDP, land use, develop scenarios for and political values, which varied from an in- and transport. The plausibility of projections long-range planning in creased focus on individual consumption and was checked, mainly by comparing projected several policy areas, personal freedom (“consumerism”) to a wide- future rates of change to historical experience. with additional detail in spread elevation of concern for the common The scenarios were published with a detailed areas relevant to good (“community”); and governance, which guidance document, which provided sugges- greenhouse-gas varied from authority and power concentrated tions on how to use them together with climate emissions. at the national level (“autonomy”), to power scenarios for impact studies.129 increasingly flowing to global institutions, downward to local ones, and outward to non- As of 2005, the socio-economic scenarios had governmental institutions and civil society been used in six impact studies.130 There has (“interdependence”). The two dimensions of been some difficulty applying the national-level uncertainty, values and governance, were as- scenarios in specific, smaller-scale regions. sumed to be independent of each other. Other The most ambitious use has been a preliminary major uncertainties such as demographic integrated assessment of climate impacts change, the rate and composition of economic and responses in two regions of England, the growth, and the rate and direction of technolog- Northwest and East Anglia.131 This study pro- ical change, were treated largely as conse- duced four integrated scenarios of regional quences of alternative realizations of the two climate impacts, by pairing each of the four core dimensions of values and governance.127 socio-economic scenarios with one climate scenario based on a rough correspondence between The four scenarios built around these two the socio-economic scenario and the IPCC dimensions of variation were called “National emissions scenario underlying the climate sce- Enterprise,” “World Markets,” “Local Steward- nario132 Based on these four scenarios, the ship,” and “Global Sustainability.” Each was initially developed as a qualitative narrative of 128 Berkhout et al. 2001. future conditions in UK society intended to 129 Berkhout and Hertin 2001. apply broadly to both the 2020s and 2050s. 130 UKCIP 2005. Each scenario specified several dozen socio- 131 Holman et al. 2002. 132 Regional (National) Enterprise was taken as UKCIP 125 UKCIP 1998:iv. High (IPCC A2); Global Markets as UKCIP Medium- High (A1B); Regional (Local) Stewardship UKCIP 126 UKCIP 2001. Medium-Low (B2); and Global Sustainability UKCIP 127 UKCIP, 2001. Low (B1).

45 The U.S. Climate Change Science Program Section 3 - Review of Major Climate-Change Scenario Exercises

study elaborated preliminary regional scenarios threat assessment, a set of scenarios of flood corresponding to the four national socio-eco- risk, and a set of policy recommendations. An nomic scenarios, and conducted an assessment evaluation of this study’s effects one year later of coastal-zone impacts and responses using found that it was being used by several public these scenarios and a formal land-use model. and private actors to inform decision-making.133

Four new climate scenarios were produced in The UKCIP, in contrast to the US National As- 2002, based on the SRES marker scenarios and sessment, has built a sustained assessment ca- new versions of Hadley Center climate models. pability. In addition, the central program has These new scenarios differed only in their emis- less authority over the separate assessments, in- sions assumptions, not climate sensitivity. The stead acting more as motivator, resource, and high, medium-high, medium-low, and low sce- light coordinator. Access to scenarios is to li- The UK program’s narios were driven by the A1FI, A2, B2, and B1 censed users, of whom there are about 130 – marker scenarios, respectively. These were used roughly half in universities, the rest about experience highlights to drive the HadCM3 global climate model equally split among private sector and all levels some of the same (with a grid-scale of 250-300 km), generating of government. Most active users have been na- issues for future climate-change projections for 30-year future tional officials responsible for climate-sensitive scenario exercises as periods centered on the decades of the 2020s, resources, with less participation from the pri- the US National 2050s, and 2080s. For some emissions scenar- vate sector and local governments.134 Assessment, in ios and time periods, climate projections were particular the processed through a nested hierarchy of three The program has invested in generating, dis- importance of Hadley Center climate models: the HadCM3 seminating, and documenting useful climate continuity of model at global scale, the HadAM3H model at scenarios for impacts users. The jury appears institutional and intermediate scale, with a horizontal resolution to still be out on whether the level of effort and analytic capability and of about 120 km, and the HadRM3 model for success is similar for socio-economic scenarios, the desirability of high-resolution climate projections in the which have not yet been either downscaled or developing and United Kingdom and Europe, with a horizontal repeated. Getting scenarios used is a slow resolution of about 50 km. This nested pro- process, but the scenarios produced by this pro- supporting scenarios cessing was done for the baseline period (1960- gram are starting to be used by decision-makers using an organizational 1990), and for the most distant projection period in support of their practical responsibilities. A structure that (2070-2100) to produce three ensemble runs for significant limitation of the program, however, combines centralized the medium-high (A2) emissions scenario and is its exclusive reliance on just one family and decentralized one for the medium-low (B2). For the other of climate models. This may pose risks of elements. emissions scenarios and the intervening projec- under-estimating future climate uncertainty and tion periods, results of the global-scale model over-confidence in assessments of potential cli- were downscaled using statistical patterns of mate impacts and responses. Although the UK fine spatial-scale climate variation derived from program followed a substantially different or- full runs using scenario A2. These scenarios ganizational model from the US National As- were widely distributed and supported through sessment, its experience highlights some of the a web-based interface, including map-based same issues for future scenario exercises, in par- graphical display of projected changes in more ticular the importance of continuity of institu- than a dozen climate indicators on a fine-scale tional and analytic capacity and the desirability (50 km) grid of the United Kingdom. of developing and supporting scenarios using an organizational structure that combines cen- Several analyses are continuing to use the 2002 tralized and decentralized elements. climate scenarios in conjunction with the socio- economic scenarios. For example, a 2004 integrated analysis of flood risk and erosion control over a 30-100 year time horizon produced a

133 UK Office of Science and Technology 2002. 134 West and Gawith 2005.

46 Global-Change Scenarios - Their Development and Use

3.4. THE MILLENNIUM environment such as climate change, air pollu- ECOSYSTEM ASSESSMENT tion, land-use and land-cover change, resource consumption, and external inputs to ecosystems The Millennium Ecosystem Assessment (MEA) such as irrigation and synthetic fertilizer use. was a large, United Nations (UN)-sponsored as- Indirect drivers included underlying socio-eco- sessment of the current status, present trends, nomic factors such as population, economic and longer-term challenges to the world’s growth, technological change, policies, atti- ecosystems, including climate change and other tudes, and lifestyles.137 sources of stress. Conducted between 2001 and 2005, the MEA sought to assess changes in The scenarios working group sought to apply ecosystems in terms of the services they provide this conceptual framework to long-term trends to people and the effects of ecosystem change in ecosystems, looking ahead to 2050 with more on human well-being. It also sought to identify limited projections to 2100. They developed the The MEA sought to and assess methods to mitigate and respond to structure of the scenarios in an iterative process, assess changes in ecosystem change, for various private and pub- including consultations with potential scenario ecosystems in terms of lic-sector decision-makers, including those re- users and experts in a wide range of decision- the services they sponsible for the several international treaties making positions around the world.138 Like sev- that deal with ecosystems.135 More than 1350 eral other major scenario exercises, they initially provide to people and authors from 95 countries participated in the sought to identify two basic dimensions of un- the effects of global assessment’s four working groups, and certainty in long-term ecosystem stresses, ecosystem change on hundreds more in about 30 associated sub- which together would produce four scenarios.139 human well-being. global assessments. The assessment’s goals For the first dimension, like SRES they chose were broad, ranging from providing a bench- globalization: continuation and acceleration of mark for future assessments and guiding future present global integration trends, versus reversal research to identifying priorities for action.136 of these trends to increasing separation and isolation of nations and regions. For the second di- Results of the global assessment were presented mension, in contrast to the broad value-based in a March 2005 synthesis report, and in addi- uncertainties used in the SRES and UKCIP sce- tional volumes presenting the output of the as- narios, they chose one more specifically related sessment’s four working groups, “Current State to ecosystems: whether responses to increasing and Trends,” “Scenarios,” “Policy Responses,” ecosystem stresses are predominantly reactive and “Multi-Scale Assessments.” The current – waiting until evidence of deterioration and state and trends group examined ecosystem loss of services is clear – or predominantly trends over the past 50 years and projections to proactive, taking protective measures in ad- 2015; the scenarios group took a longer view to vance of their clear need. The combination of 2050 and beyond. Because of time limitations, two polar values of each of these uncertainties the work of these two groups proceeded yielded four scenarios, summarized in Table 3.2. largely independently. The Global Orchestration (global, reactive) sce- All components of the assessment used a com- nario presented a globally integrated world with mon large-scale conceptual framework, which low population growth, high economic growth, distinguished indirect drivers of ecosystem and strong efforts to reduce poverty and invest change, direct drivers, ecosystem indicators, in public goods such as education. In this sce- ecosystem services, measures of human well- nario, society focuses on liberal economic val- being, and response options. Direct drivers in- ues, follows an energy-intensive lifestyle with cluded direct human perturbations of the no explicit greenhouse-gas mitigation policy,

135 E.g., the Convention on Biological Diversity, the Convention to Combat Desertification, the Convention 137 MEA 2006:153 (Table 6.1) and 304 (Table 9.2). on Migratory Species, and the Ramsar Convention on Wetlands. 138 MEA 2006:152. 136 MEA 2006:xii. 139 MEA 2006, Figure 5.2.

47 The U.S. Climate Change Science Program Section 3 - Review of Major Climate-Change Scenario Exercises

Table 3.2. WORLD DEVELOPMENT Millennium ECOSYSTEM MANAGEMENT Ecosystem Global Regional Assessment Scenarios Reactive Global Orchestration Order from Strength

Proactive TechnoGarden Adapting Mosaic

and takes a reactive approach to ecosystem tion, with world population in 2050 ranging problems.140 In Order from Strength (regional, from 8.1 billion (Global Orchestration) to 9.6 reactive) the reactive approach to ecosystem billion (Order from Strength).143 No statements problems takes place in a fragmented world pre- of probability or likelihood were made about the occupied with security and less attentive to pub- scenarios. lic goods.141 This scenario exhibits the highest population growth and lowest economic growth. From the indirect drivers, a more specific and Economic growth is particularly low in the de- quantified set of direct drivers was developed, veloping countries, and it decreases over time. using formal models where possible. Species In Adapting Mosaic (regional, proactive), polit- introduction and removal was the only unquan- ical and economic activity are concentrated at tified direct driver.144 Separate pre-existing mod- the regional ecosystem scale. Societies invest els were used of the world energy-economy, heavily in protection and management of greenhouse-gas emissions and climate change, ecosystems in locally organized and diverse ef- air pollution, land-use change, freshwater, ter- forts. Population growth is nearly as high as in restrial ecosystems, biodiversity, and marine Order from Strength, and economic growth is and freshwater fisheries. To the extent possi- initially slow but increases after 2020. Finally, ble, these quantitative models were used to rea- TechnoGarden (global, proactive) presents a son from indirect and direct drivers to world that is both focused on ecosystem man- ecosystem effects, changes in ecosystem serv- agement and globally connected, with strong ices, and effects on human well-being.145 In development of environmentally friendly tech- some cases this was achieved by soft-linking nology. Population growth is moderate, and eco- models, using outputs from one as inputs to an- nomic growth is relatively high and increasing.142 other, but this was limited by different variable definitions, spatial and temporal resolution, and Each scenario was initially constructed as a other model incompatibilities.146 Not all sce- qualitative description. Population and GDP nario elements could be modeled quantitatively, were specified quantitatively, while all other in- so expert judgments were also extensively used. direct drivers – including social, political, and The qualitative scenario process proceeded in cultural factors – were qualitative. Population parallel with quantitative modeling – elaborat- scenarios were derived from the International ing aspects of the scenarios that were not Institute for Applied Systems Analysis’ amenable to modeling, filling gaps, and stipu- (IIASA’s) 2001 probabilistic projections, cap- lating feedbacks between ecosystem services turing the middle 50-60 percent of the distribu- and human well-being and behavior.147

140 MEA 2006, Ch 5.5.1 141 This scenario was originally named “Fortress World” (report of first meeting of MEA global modeling group, Jan 7, 2003). The later name reflected partici- 143 MEA 2006:182. pants’ judgments that in such a decentralized world preoccupied with security concerns, maintaining global 144 MEA 2006:304, Table 9.2. order would require democratic nations to be militarily 145 MEA 2006, Table S3. strong – i.e., it is a world of “realist” international affairs (MEA 2006:133) 146 MEA 2005, Table S2. 142 MEA 2006:131. 147 MEA 2006:155.

48 Global-Change Scenarios - Their Development and Use

The groups attempted to check for consistency In sum, the MEA scenarios project invested between quantitative and qualitative scenario el- substantially more effort in developing rich ements through periodic consultations, qualitative and narrative scenarios than the particularly for feedbacks that could not be SRES, but also fell short on integrating quali- modeled analytically. Some of these were tative and quantitative components. In part be- interactions between direct drivers and ecosys- cause of the greater elaboration of the tems, but the most difficult occurred in scenar- qualitative components, this limited coordina- ios that assumed strong socio-economic tion resulted in significant inconsistencies and feedbacks and regulating mechanisms. Adapt- requirements to resolve conflicts between the ing Mosaic, for example, assumed strong feed- two components. These inconsistencies arose backs from new ecosystem observations and even with just one model used for several com- knowledge to changes in human behavior that ponents of the assessment, so the challenges of could not be incorporated into the models used. harmonization among models – and the associ- Representing these required allowing qualita- ated possibility to explore model-structure un- tive scenario logic to override both the quanti- certainty – did not arise. A related problem was tative results and the structure of models. that for many factors it was difficult to generate Unfortunately, time limits prevented this con- the desired level of variation between scenar- sistency checking from being done thoroughly, ios.150 This raises issues of potential method- so unexamined disparities between qualitative ological interest, such as how to distinguish and quantitative aspects of the scenarios re- robust results from inadvertent convergence of mained a significant weakness. assumptions or model structures, which remain to be investigated. Finally, the great breadth of Many of the conclusions developed from the conditions represented in the scenarios, as well scenarios are common to all four scenarios, as possible concerns with logical circularity while in others Order from Strength is the ex- between their presumptions and results,151 ception. For example, one major conclusion is make interpreting the significance of the that rapid conversion of ecosystems for use in results difficult. agriculture, cities, and infrastructure will continue, and that habitat loss will continue to The experience of this scenario exercise pro- contribute to biodiversity loss. However, if vides a different perspective on some of the ecosystem services increase as projected, some same key challenges for future scenarios high- ecosystem services – although not biodiversity lighted by the other activities reviewed. The – may be decoupled from ecosystem stresses. quite distinct difficulties faced here in attempt- Food security is projected to remain out of ing to combine quantitative and qualitative sce- reach for many people. Extreme, spatially narios highlight the central importance and the diverse changes are projected for freshwater re- difficulty of developing new methods to inte- sources, with general deterioration in developing grate these two approaches. In addition, this ex- countries under both “reactive” scenarios. In- perience highlights the value of clarity about the creasing demands for fishery products are pro- intended uses of scenarios, including clarity jected to increase risks of regional marine fishery about whether they are intended to address spe- collapses.148 In sum, ecosystem services show cific questions, guide decisions, or explore long- mixes of improving and worsening trends in all term conditions. The risk of scenarios becoming scenarios except Order from Strength, in which nearly all ecosystem services are projected to be more impaired in 2050 than in 2000. The same 150 Report of the First Meeting of the MEA Global Mod- three scenarios also suggest that significant eling Group, 7 Jan 2003, at www.usf.uni-kassel.de/ma- changes in policies, institutions, and practices gmgroup/dl/first_report.doc; Second Report of the MEA Global Modeling Group, 7 March 2003, avail- 149 can mitigate some negative consequences. able at www.usf.uni-kassel.de/ma-gmgroup/dl/sanjose _report.doc. 151 This concern is particularly present regarding implications of the assumption that ecosystem management is either proactive or reactive (See, e.g., MEA 2006, Ch 148 MEA 2006, Table S3. 8.4.2.1 and Ch 9). 149 MEA 2006:127.

49 The U.S. Climate Change Science Program Section 3 - Review of Major Climate-Change Scenario Exercises

less useful due to breadth and vagueness may be particularly acute for scenarios that attempt to capture multiple stresses on some system – even though such multi-stress assessment is repeatedly advocated for climate-change and other forms of environmental assessment.152

152 NAST 2001.

50 Global-Change Scenarios - Their Development and Use

Challenges and 4 Controversies in Scenarios for Climate Change SECTION

This section discusses several issues that have arisen in multiple scenario exercises related to climate change, issues that pose challenges for expanding the usefulness of scenarios to climate change analysis, assessment, and decision support. Section 4.1 examines the type of information needs of specific types of decisions related to climate change and considers the requirements and challenges of crafting scenarios to serve these needs. Section 4.2 considers the use of scenarios that has been more common thus far, in structuring climate-change assessments and framing broad policy debates, and identifies the distinct challenges in enhancing the value of scenarios in these purposes. The remaining sub-sections examine particular design challenges in crafting scenario exercises: how to structure interactions between experts and stakeholders in developing scenarios; how to communicate scenarios to potential users not involved in their creation; how to pursue the two, not perfectly aligned goals of consistency and integration in scenarios; and how to represent and interpret uncertainty in scenarios. Throughout this section, we present illustrative examples of scenario activities in text boxes. These examples shed additional light on various challenges, especially relating to scenarios’ use in decision-making.

4.1. SCENARIOS AND DECISIONS

As discussed in Section 1, the general purpose of scenarios is to inform decisions, but their connection to specific, identified decisions can be more or less close and direct. In interpreting and evaluating present experience with scenarios and identifying key challenges in making them more useful, it is important to distinguish scenario exercises by their major characteristics, including their specificity, their proximity to decisions, the degree of normative presumptions embedded in them, and where they lie in the causal chain outlined in Section 2. To consider how scenarios can help inform climate-change decisions, we must first specify the relevant decisions and decision- makers more sharply. This section considers the major concrete decisions that comprise a response to climate change. Decisions related to assessment, modeling, and research are considered in Sec- tion 4.2. This discussion must be somewhat hypothetical, extending from rather thin current practice to reasonable speculation about future decisions and likely information needs.

Because the dynamics of climate change operate on multiple spatial scales from the local to the global, there is no single global climate-change decision-maker. Rather, many distinct decision- makers with diverse responsibilities will affect and be affected by climate change. Because of climate’s recent appearance on policy agendas and its dense connections to other issues, many of these decision-makers’ primary responsibilities are defined as something other than climate change. Some of them are already considering how climate change might affect their responsibilities, but many are not.

51 The U.S. Climate Change Science Program Section 4 - Challenges and Controversies in Scenarios for Climate Change

Section 2 described climate-change decisions preparedness or public health officials; officials using the conventional dichotomy of mitigation making zoning or coastal development policy; versus adaptation. To consider potential contri- or firms in insurance or financial markets who butions of scenarios in more detail, we discuss may bear secondary risks from impacts or seek three types of decision-maker: national officials, to develop new instruments to exchange these impacts and adaptation managers, and energy risks. Unlike national officials, these actors’ resource and technology managers. These can decisions are purely responses to climate often be identified as particular programs, divi- change, realized or anticipated: they have little sions, agencies, organizations, or individuals, influence over how the climate will change. each with different responsibilities and types of Their responsibilities will often connect with information they might consider in making their the impacts-related responsibilities of national decisions. All three groups face decisions under officials, but are narrower in scale or scope. Im- All three groups face uncertainty with long-term consequences related pacts and adaptation managers would be con- to climate change, and so might benefit from sce- cerned not with aggregate climate-change decisions under narios providing structured information and as- impacts on the United States, but with more uncertainty with long- sumptions about the values at stake, the available specific impacts such as those on seasonal flows term consequences choices, and their consequences under alterna- and water-management operations on the related to climate tive climate-change futures. Upper Mississippi. change, and so might benefit from scenarios National officials’ responsibilities are the broad- Energy resource and technology managers in- providing structured est and the most likely to be explicitly related to clude developers and operators of fossil or non- information and climate change. They develop national policies fossil energy resources, investors in long-lived assumptions about the that target greenhouse-gas emissions and moti- energy-dependent capital stock such as electri- values at stake, the vate investment in technologies that will influ- cal utilities, and researchers, innovators, and in- available choices, and ence future emissions trends. They negotiate vestors in new energy-related technologies. their consequences policies internationally with officials from other These decision-makers are mostly but not ex- under alternative nations, and with sub-national officials who clusively in the private sector. Their decisions may share mitigation responsibilities or under- may have consequences that interact with vari- climate-change futures. take mitigation measures at their own initiative. ous processes operating over multiple time- They also have responsibilities to anticipate and scales, from short-term market responses, to respond to climate-change impacts in their na- decadal-scale processes of investment, resource tions. Their climate-change responsibilities are development and depletion, and penetration of open-ended, not limited to mitigation and new technologies, to century-scale processes of adaptation: these decision-makes will deter- climate change.153 These actors’ decisions will mine the extent to which other responses such strongly influence society’s ability to control as geoengineering are considered, and the de- greenhouse-gas emissions. This group also in- sign of systems and institutions for assess- cludes energy consumers such as firms or pub- ment. They are also responsible for overall lic agencies considering mitigation actions in their national welfare, including not just the environ- own operations. While their areas of responsibil- mental effects of their decisions but also other ity may be vulnerable to climate change and its linked national interests such as economic pros- impacts, the largest climate-related risks for this perity and security. group are likely to come not from climate change itself, but from climate-change policies: Impacts and adaptation managers have respon- national mitigation policies, and other market sibility for particular assets, resources, or inter- and regulatory decisions that shape the outcomes ests that might be sensitive to climate change. of private mitigation activities. They must decide how to anticipate, prepare for, and respond to the threat, minimize its harm, At greatly varying levels of precision and speci- and maximize any associated benefit. They ficity, scenarios can present two types of infor- may be private or public actors – e.g., owners mation to support decisions by these three types or managers of long-lived assets such as ports or water-management facilities; managers of lands, forests, or protected areas; emergency 153 Shell International 2001, Davis 2003.

52 Global-Change Scenarios - Their Development and Use of actors. Scenarios can represent potential fu- These needs introduce a dimension of com- ture developments that may threaten decision- plexity into mitigation scenarios, sometimes makers’ interests or values, call for decisions, called “reflexivity.” Because mitigation policies or challenge conventional thinking and prac- seek to reduce future emissions by altering the tices. And they can provide a structure to as- socio-economic drivers of emissions growth, sess potential consequences of alternative the analysis of mitigation policies and their con- decisions for things that matter to the decision- sequences must be coupled to the causal logic of maker. Beyond this generalization, the three emissions scenarios. Whereas climate scenarios types of decision-makers will differ substan- can be treated as exogenous when assessing tially in the specific types of information they adaptation decisions, emissions scenarios can- need, the time horizons of their decisions, and not be treated as exogenous in assessing miti- the type and extent of causal connections be- gation decisions. Any emissions scenario tween their decisions and the conditions embeds some assumptions about mitigation Any emissions scenario specified in scenarios. policies, assumptions that may have to be embeds some changed to assess particular mitigation policies. assumptions about 4.1.1. Scenario needs: This effect will be strongest when emissions mitigation policies, national officials projections and mitigation options are being considered at the same spatial scale, e.g., na- assumptions that may As national officials have the broadest respon- tional mitigation policies are being assessed rel- have to be changed to sibilities related to climate change, they are also ative to national emissions projections. The assess particular likely to have the broadest information needs. effect of national mitigation strategies on global mitigation policies. In their responsibilities to build national adap- emissions will be weaker: no nation controls tation capacity and manage key vulnerabilities, global emissions trends, and the effects of small their needs are similar to those of impacts and nations’ mitigation strategies on global trends adaptation managers: scenarios of potential fu- can be very small. ture climate change under specified emissions assumptions and resultant impacts on particular Scenarios to inform mitigation decisions are resources and communities in their nation, with also likely to require considering alternative as- particular focus on areas of greatest vulnerabil- sumptions about the policy context in which ity. They will likely have less need for fine spa- these decisions are made. The effects of na- tial and sectoral detail in potential impacts, but tional mitigation strategies – including how more need for consistent scenarios that allow much they reduce national emissions, as well as comparison and aggregation across sub-national their costs and other consequences – will de- regions and sectors. These will help to priori- pend on the economic, technological, and policy tize, identify key areas of vulnerability, and es- context, including related decisions by other timate aggregate costs for planning purposes. major nations, individually and through international coordination. Assumptions about the In their responsibilities for national mitigation policy context will be less important in scenar- policy, national decision-makers will also need ios to inform international mitigation decisions, information about the aggregate impacts of since when decisions are globally coordinated climate change, since the more severe climate there is no “elsewhere” – but alternative as- impacts are likely to be, the greater the justifi- sumptions about nations’ degrees of compliance cation and likely political support for mitigation and form of implementation of international measures. But mitigation decisions also require commitments may still be needed. additional information – including projections of future emissions in the absence of explicit Scenarios of emissions, climate change, and im- mitigation efforts, and the consequences of alter- pacts inform mitigation decisions by helping to native mitigation policies, in their effects on emis- characterize the potential severity of climate sions, their cost, and their implications for other change and therefore how important it is to con- dimensions of national interest. trol emissions. This support is indirect, serving primarily to elevate or moderate the general level of concern on the issue. More focused work on mitigation has been done using con-

53 The U.S. Climate Change Science Program Section 4 - Challenges and Controversies in Scenarios for Climate Change

structed scenarios of limited emissions, often ables are generated and saved, and that data and aiming at stabilizing atmospheric concentrations documentation are transferred accurately. or radiative forcing at various levels, and examining the configurations of technology, energy re- This is the area of climate-related decisions for sources, and economic and population growth which the provision of information from cli- that are consistent with the specified scenario. mate-change scenarios is most advanced. Still, Some studies have used quantitative models to further progress is needed in the development estimate costs of such scenarios, relative to an as- and use of scenarios of socio-economic condi- sumed baseline emissions scenario.154 tions, and in creation of methods and tools to augment centrally provided scenario infor- 4.1.2. Scenario needs: impacts and mation with information tailored to specific adaptation managers impact assessments. In addition, many impacts- Impacts and adaptation related decisions will require scenarios of cli- To assess the threats and opportunities they face mate change in the context of other linked managers need and evaluate responses, impacts and adaptation stresses and changes. scenarios of potential managers need scenarios of potential future cli- future climate change, mate change, its impacts in their areas of re- 4.1.3. Scenario needs: energy its impacts in their sponsibility, and the factors that influence resource and technology managers areas of responsibility, vulnerabilities. With few exceptions, these ac- and the factors tors’ decisions will have no effect on the climate Energy and technology managers will most ben- that influence change to which they must respond, so scenar- efit from scenarios that explore alternative pol- vulnerabilities. Energy ios of climate-change stresses can be con- icy regimes and their consequences for the and technology structed independently of the assessment of value of energy and technology assets and in- managers will most potential adaptation decisions, without concern vestments. For some, the predominant concern benefit from scenarios for feedbacks that may modify the conditions may be overall policy stringency, perhaps sum- that explore specified in the scenario. marized as alternative emissions-price trajecto- alternative policy ries over time; for others, specific details of Particular decision-makers’ needs will be highly policy design and implementation may need to regimes and their specific in the variables they require, and their be considered. Scenarios of emissions, climate consequences for the time and spatial scale and resolution. A planner change, and impacts only matter for decisions value of energy and of water-management infrastructure may need via their likely influence on policy stringency, technology assets and monthly or finer-scale rain and snow projec- and so do not need to be explicitly represented investments. tions over a watershed; a designer of coastal in- in scenarios. These actors may have some infrastructure may need probabilistic projections fluence on policy, but probably not such strong of sea level, storm intensity and frequency, influence that climate-policy scenarios would storm surge, or saltwater intrusion. But in their have to incorporate feedbacks from their own climatic elements, these information needs all advocacy efforts. rest on a common core of scenarios of global climate change and emissions drivers. This dual Unlike the other two types of decision-makers, structure of information – highly particular cli- these actors are likely to compete with each mate variables, based on a set of common “core other. If, for example, they are investors allo- scenarios” – suggests a cross-scale organiza- cating research effort between higher and lower- tional structure for providing scenario infor- emitting energy sources, those who better mation: commonly produced scenarios of anticipate future policy will benefit relative to climate change and other components requir- those who do worse. If they use scenarios, they ing consistency, specialized expertise, or high- may consequently choose to produce them pri- cost resources; development of decentralized vately, perhaps coupled with other analyses to capabilities in impact assessments to adapt these generate practical guidelines for investments.155 core scenario elements and develop assessment- As for the other types of decision-makers, these specific extensions; and close communication specialized scenarios could be based on general between these groups to ensure that useful vari- scenarios of global emissions and climate

155 Ged Davis, personal communication. (posted expert 154 Morita et al. 2001, CCSP 2007. review comments). 54 Global-Change Scenarios - Their Development and Use change. Published scenarios produced to date Following the general reasoning above, how on the climate-change issue, however, have not these are treated should depend on what type of considered mitigation policies with the speci- decision is being informed. In climate scenar- ficity necessary to inform these actors’ decisions. ios to inform impact assessments and related decisions, the scenario users are not consider- 4.1.4. Representing decisions ing mitigation decisions and have little influ- in scenarios ence over them, so emissions scenarios should include assumptions about the likely or plausi- A major challenge to developing scenarios to ble range of mitigation efforts. The range of fu- support decisions is reflexivity, that is, how to ture climate change considered may thus be represent decisions within scenarios without narrowed to reflect the possibility of negative making scenarios either circular or contradic- social and political feedbacks: sustained rapid tory. In meeting this challenge, the most basic emissions growth may generate pressure for ag- In climate scenarios to distinction to draw is between decisions by the gressive mitigation, due to increasing signs of inform impact scenario’s targeted users and decisions by other climate change, alarming projections of future assessments and actors. From the users’ perspective, decisions change, or other environmental harms from related decisions … by others over which they have no influence are rapid expansion of coal or synthetic fuels. indistinguishable from non-choice events. If the emissions scenarios factors influencing these decisions are confi- Such a negative-feedback mechanism may not should include dently understood, they might be represented be effective, of course. Many factors could in- assumptions about the deterministically, like well-understood biophys- tervene: mitigation measures may not gain likely or plausible ical or economic processes. In the more fre- enough support to be adopted, socio-political range of mitigation quent case that others’ choices cannot be capacity to enact stringent policies may be di- efforts. confidently predicted, they might be represented minished, policies adopted may be ineffective, as uncertainties – again, just like uncertain bio- or early technology or policy decisions may physical or economic processes. As with all un- lock in high-emitting future paths. But to the certainties, how to treat them depends on their extent that such a negative-feedback mechanism judged importance for the users’ decisions: if it does operate, persistence of the highest emis- is high, they can be represented in alternative sions paths beyond a few decades would be- scenarios; if not, they can be fixed at some best- come unlikely. guess value for all scenarios. In either case, these decisions are treated exogenously. Parallel reasoning may apply to extremely low emission paths, if sustaining such low emissions Representing decisions by the scenario users is requires continued costly mitigation efforts that fundamentally different. Since the scenarios are come to be seen as unnecessary. This negative- intended to inform these decisions, alternative feedback mechanism would likely be weaker choices should not be represented as exogenous than that operating at the high end of the emis- uncertainties but be stipulated independently sions distribution, however, because long time- from the scenarios. Users can then explore their constants mean that increasing signs of climate implications under conditions imposed by sce- change are likely to continue through most of narios, including representation of major un- the 21st century even if we follow a low-emis- certainties. Various degrees of coupling can be sions path. If impacts assessors and managers required between the logic of scenarios and the judge these negative feedbacks to make extreme analysis of consequences of the users’ decisions. emissions paths sufficiently unlikely, particu- In scenarios for impacts, these can usually be larly high ones, they may reasonably decide not separate; in scenarios for mitigation, they may to consider these extreme emissions futures in have to be closely coupled, since emissions sce- their planning for adaptation. narios may change under alternative assumptions about mitigation decisions. For scenarios to inform mitigation decisions, particularly at the international level, the situa- In scenarios to inform global climate-change tion is different. Informing these choices re- decisions, the sharpest question is how to rep- quires information about potential emissions resent mitigation decisions within scenarios. paths and their consequences under all levels of

55 The U.S. Climate Change Science Program Section 4 - Challenges and Controversies in Scenarios for Climate Change

mitigation effort that decision-makers might relied on to act responsibly at all can easily lead plausibly consider – including no additional to decisions that incur excessive costs, by try- measures, or even reversal of previous measures ing to achieve rapid mitigation immediately or if this is on the agenda. Consequently, in con- tie future decision-makers’ hands. trast to scenarios for impacts, extreme emissions futures should not be excluded when Two approaches appear promising for integrat- assessing mitigation decisions. For example, if ing future mitigation decisions into scenarios to scenarios that truncate high-emissions futures inform current decisions. Scenarios could pre- by assuming stringent mitigation are used to sume that today’s decision-makers choose the support a decision that stringent mitigation is future path of mitigation, allowing them to as- not necessary, the result is contradictory: a con- sess and contribute to a trajectory of effort that clusion supporting a decision is based on the considers the welfare of both current and future …in contrast to presumption of the contrary decision. To avoid citizens. Alternatively, scenarios could treat fu- such contradictions, scenarios to inform miti- ture large-scale mitigation choices as uncer- scenarios for impacts, gation decisions must consider alternative mit- tainties represented in alternative scenarios, extreme emissions igation choices explicitly, not embed them while also considering how current choices can futures should not be implicitly in the underlying logic of the scenario. seek to influence the opportunities and incen- excluded when tives faced by future decision-makers. assessing mitigation Moreover, national officials act only for their decisions. … Scenarios own nations in the near term, even when they In sum, the importance of connecting scenarios to inform mitigation negotiate global mitigation. They may make to actual decisions is widely recognized, but decisions must choices for long-term planning and institutional there is a large gap between, on the one hand, consider alternative design for future mitigation as well, but it is the value scenarios could provide to climate- mitigation choices their successors who will decide whether to change decisions and the aspirations of scenario explicitly, not embed continue, strengthen, or otherwise change meas- producers to provide that value, and current them implicitly in the ures adopted today. From the perspective of practice on the other hand. There has been lit- underlying logic of the current national officials, mitigation decisions tle use of scenarios to directly inform climate- by other nations and in the future fall between change related decisions, although there appears scenario. the two cases discussed above: they are not con- to be a sharp increase in the interest of decision- trolled by the scenario user, but can be influ- makers and early attempts. The rapid increase enced to some degree. For policy choices by in interest is particularly evident for informing other nations, national officials may need to be decisions related to climate-change impacts and advised in two modes, reflecting their dual re- adaptation. There are fewer indications of sim- sponsibilities to make national policy and to ne- ilarly direct use of scenarios to inform mitigation gotiate international agreements. In the latter decisions, perhaps in part because nearly all cur- capacity, alternative approaches to global miti- rent mitigation decisions have been near-term. gation strategy should be represented as choices. But when they consider national deci- Mitigation decisions at the national and inter- sions separate from globally coordinated strat- national level have taken scenarios into account egy, relevant decisions of other major nations indirectly. Most scenarios have been con- should be represented as uncertainties. This structed to provide inputs to assessments, mod- may require use of two distinct types of scenar- els, or other analyses. This has included serving ios to advise development of different aspects as inputs to the production of other types of sce- of national mitigation policy. narios, which then describe other potential future conditions that depend on those specified How to represent future mitigation decisions in the scenario, as for example a model-based poses a still harder dilemma. On the one hand, climate scenario depends on inputs from an it appears risky or even irresponsible to assume emissions scenario. While these uses can be that the bulk of mitigation efforts can be left to characterized as supporting decisions (i.e., de- future decision-makers, even if we assume this cisions about assessments, modeling, and re- will be easier for them because of greater wealth search), their connection to concrete decisions or technological prowess. On the other hand, of mitigation and adaptation is indirect, assuming that future decision-makers cannot be achieved through contributions such as sup-

56 Global-Change Scenarios - Their Development and Use porting strategic planning and risk assessment, public. This description applies to the major providing advance analysis for potential future scenario exercises discussed in this report, in- decisions, exploring plausible extreme cases, cluding the IPCC emissions and climate sce- helping to characterize and prioritize key un- narios, the US and UK assessments of climate certainties, or educating decision-makers or the impacts, and the MEA scenarios.

BOX 4.1. Scenarios for Climate-Change Adaptation in the New York Metropolitan Region

Three linked activities – the Metropolitan East Coast (MEC) assessment of the US National As- sessment, the New York Climate and Health project (NYCHP), and the New York City Department of Environmental Protection (NYCDEP) Task Force on Climate Change – have used or are using scenarios to assess impacts of climate change on the New York Metropolitan Region, identify areas of vulnerability, and inform regional planning and decision-making.156 The MEC assessment began with a regional workshop in April 1998, involving about 150 participants, including public agencies at all levels of government as well as climate researchers. The subsequent assessment was conducted by sector teams of researchers and officials from agencies responsible for the study sectors. Teams developed regional scenarios of climate change and sea level rise based on the downscaled climate-model scenarios provided by the US National Assessment, plus two additional scenarios based on extrapolation of recent regional climate trends and historical extremes. The scenarios were used to project climate-change impacts on beach nourishment, 100 and 500- year flood heights, wetland aggregation and loss, adequacy of the water supply system under droughts and floods, illnesses from acute air-pollution episodes, and peak energy loads. These impact projections were used for preliminary assessment of adaptation strategies and policies. Following the MEC Assessment, the NYCHP created updated regional climate scenarios in consultation with an expert-stakeholder Advisory Board. This study further analyzed public health impacts, focusing on air quality and extreme heat events. The updated climate scenarios used the IPCC A2 and B2 emissions scenarios driving global and regional climate models to create downscaled scenarios for the region. These were augmented with newly developed scenarios of future regional land use and population growth based on the IPCC A2 and B2 storylines. In response to the widespread public attention received by the MEC Assessment Report, the Com- missioner of the NYCDEP established the Climate Change Task Force, a collaboration among regional researchers and the agency that manages the water system. The Task Force uses the latest climate-model simulations from the IPCC Fourth Assessment Report, as well as additional global and regional climate models, to develop new regional scenarios. These will include probability distributions of average and extreme temperature and precipitation change, as well as sea level rise. The Task Force is also developing qualitative scenarios of extreme sea level rise in the region. DEP is using these results to develop a comprehensive adaptation strategy for the New York City water system, including assessment of many specific adaptation options, that considers both uncertainties in future climate change and managerial factors such as the time horizon of different adaptation responses and capital turnover cycles. This is a successful example of scenario-based assessment of climate impacts and adaptation options. The scenarios are connected with the concrete responsibilities and concerns of stakeholders, who were involved in their design from the outset. Although officials have found the wide range of uncertainty in climate scenarios difficult to incorporate into infrastructure design specifications, particularly for precipitation, the exercise has effectively conveyed the challenges posed by future climate uncertainty to current decisions of planning and infrastructure design. Stakeholders’ will- ingness to support and participate in three separate phases of these activities and NYCDEP’s in- corporation of them into a strategic planning exercise provides clear evidence of the practical utility of the exercises.

156 Rosenzweig and Solecki 2001; Kinney et al. 2005, 2006; Rosenzweig and Major 2006. 57 The U.S. Climate Change Science Program Section 4 - Challenges and Controversies in Scenarios for Climate Change

BOX 4.2. Scenarios of Sea Level Rise along the Gulf Coast

Sea-level rise is one of several factors that contributed to the decline of coastal ecosystems along the US Gulf of Mexico coast in the 20th century (Figure 4.1).157 In southeastern Louisiana, where the local rate of land surface subsidence is as high as 2.5 cm per year, rise in local “relative sea level” may be the most important factor in the rapid loss of coastal zone wetlands over the past several decades.158 Despite the importance of sea level rise in historical losses of coastal lands, planning projections of future changes in coastal Louisiana used by both federal and state agencies prior to the devastating impact of Hurri- canes Katrina and Rita in 2005 were based on just one scenario: no change in the rate of sea level rise. No alternative sea level scenario was considered in the plans then being developed to restore and protect the Louisiana coastal zone.159 This assumption sharply contrasts with the IPCC projections, which state that the global average rate of sea level rise in the 21st century may increase 2- to 4-fold over that of the 20th. Such increases will exacerbate wetland losses throughout the Gulf Coast region and obstruct restoration plans that do not take account of likely increases in water levels and salinity. The ecosystem modeling team working for the State of Louisiana and the US Army Corps of Engineers is presently integrating accelerated sea level rise scenarios into planning exercises that will aid federal and state agencies in evaluating restoration alternatives.160 The State of Louisiana is consulting with the Rand Corpora- tion to obtain probability estimates for various scenarios of sea level change to help guide engineering decisions and the design of projects aimed at restoring levees and coastal landforms that protect coastal communities.161 Sea level rise scenarios are also being used to assess the impacts of climate change and variability on the Gulf Coast transportation sector. To assess transportation impacts, a sea level rise simulation model developed by the US Geological Survey generates scenarios of sea level change using over a dozen GCMs and six SRES emission scenarios. Sea level rise scenarios are important not just in regions like Louisiana. The Big Bend region of the Florida pan- handle is experiencing very little vertical movement of the land surface, so sea level there has been rising at approximately the global average rate of 1 to 2 mm per year. But even here, coastal wetlands positioned on flat limestone surfaces may be subject to highly nonlinear effects as sea level reaches a threshold at which large areas are subject to increased salinity or inundation.162 Regional scenarios of potential sea level rise are needed to support coastal management and protection activities, as well as plans for wetland restoration and post-hurricane reconstruction. Absent consideration of such scenarios, restoration and rebuilding programs are likely to lock in errors that result in wasted resources and avoidable increases in future vulnerability.

2 Figure 4.1. Output from a Gulf Coast sea level rise scenario tool 1.5 Historical sea level change and projected sea level rise under three greenhouse-gas emissions scenarios, in meters, are shown for Galveston, 1 Texas. Both historical data and future projections are smoothed from monthly data using a

12-month moving average. (Source: Thomas W. Doyle, National Wetlands Research Center, United .5 States Geological Survey.)

0 Historical Data Low (B1) Scenario Middle (A1B) Scenario -.5 High (A1FI) Scenario 1910 1930 1950 1970 1990 2010 2030 2050 2070 2090 Linear Trend from 1909−2003 Monthly Data

157 Gosselink 1984, Williams et al. 1999, Burkett et al. 2005, Morton et al. 2002. 161 158 Shinkle and Dokka 2004, Barras et al. 2003. Presentation by Randy Hanchey, Louisiana Depart- ment of Natural Resources, to Governor’s Advisory 159 US Army Corps of Engineers 2004. Commission on Coastal Protection, Restoration and 160 See, e.g., http://www.clear.lsu.edu/clear/web-content Conservation, Baton Rouge, LA, June 22, 2006. /index.html. 162 Cook 1939, Doyle et al. 2003. 58 Global-Change Scenarios - Their Development and Use

BOX 4.1.3. Scenarios in the California Water Plan

The California Department of Water Resources (DWR) produces an updated Califor- nia Water Plan every five years. The plan projects water supplies and demands, and eval- uates current and proposed demand-management programs and supply investments, to “provide a framework for water managers, legislators, and the public to consider options and make decisions regarding California’s water future.”163 In contrast to prior plans that constructed only one future scenario, the 2005 plan explicitly considered uncertainty in supply and demand projections. Three alternative scenarios of supply and demand conditions were constructed through 2030: one extending current trends in population and economic growth, agricultural production, environmental restrictions on water use, and water conservation occurring without policy initiatives (e.g., through equipment replacement, technological change, and revised building codes); and two presenting higher and lower increases in demands. The report of the 2005 plan discusses global climate change and the potential challenges it poses to water supply and demand in California, but climate change is not explicitly represented in the plan’s three scenarios. In addition to adopting these scenarios, the State of California is developing data and analytic capacity to enrich the treatment of uncertainty and climate change in future plans. In parallel with development of the three principal scenarios in this plan update, DWR sponsored development of several analytic tools to strengthen the treatment of uncertainty in future plans. In addition, the California Climate Change Research Center with co-sponsorship from DWR is developing fine-scale regional climate-model scenarios to support analysis of climate-change impacts on water resources.164 The DWR plans to incorporate these climate-change scenarios explicitly in the next plan update in 2010.

4.2. SCENARIOS IN Similarly, the US National Assessment and UK ASSESSMENTS AND POLICY Climate Impacts Programme have both at- DEBATES tempted to identify a small set of climate and socio-economic scenarios, to coordinate and Within large-scale assessments of climate gain comparability across multiple studies and change or other environmental issues, scenarios allow aggregate assessment of impacts and vul- can serve several roles. Most straightforwardly, nerabilities at the national level. they can provide required inputs to other parts of the analysis, as the IPCC emissions scenarios In a broad assessment including many teams support the controlled comparison of climate- considering separate questions of climate- model runs. They can also serve as devices to change, impacts, mitigation, and adaptation, organize and coordinate the multiple compo- simple coordinating devices are needed to make nents of a large-scale assessment, particularly teams’ work comparable and allow synthesis to when much of the assessment is forward-look- produce aggregate conclusions. Emissions sce- ing. In the IPCC assessments, for example, narios are natural devices to provide such coor- emissions scenarios have not just been used to dination, both because emissions hold the drive coordinated climate-model projections, clearest near-term opportunities for interven- but have also increasingly been followed tion, and because they have clear and recog- through to coordinate characterization of cli- nized connections both directions in the causal mate impacts and adaptation opportunities, and chain, to every aspect of the climate-change used in a more preliminary way to organize as- issue. However, in part due to management is- sessments of the economic and technological sues, these efforts to use scenarios as broad co- implications of alternative mitigation strategies. ordinating devices have not been wholly

163 California Water Plan home page, http://www.water- plan.water.ca.gov. 164 California DWR 2005:4-32.

59 The U.S. Climate Change Science Program Section 4 - Challenges and Controversies in Scenarios for Climate Change

satisfactory in practice. To serve as coordinat- normative biases in scenarios can pose the risk ing devices, scenarios must be developed and of excluding consideration of futures that are disseminated early in the process, preferably be- judged undesirable or that pose sharp decision- fore the work of assessment teams even begins. making challenges. Such biases can be difficult Moreover, they must be documented with de- both to detect and to correct. Beyond exhorting tailed information about the process and rea- developers to scrutinize scenarios critically to soning used to generate them, including explicit avoid bias, the best protection against such biases identification of underlying assumptions and lies in transparency about the assumptions and supporting data, models, and arguments. In information underlying scenarios and associated practice, timely, detailed, and transparent dis- judgments of likelihood. semination of scenario information has rarely been achieved. Other political pressures come onto scenarios in Scenarios inevitably the broader use, debate, and criticism that they Scenarios used in large-scale assessments can encounter after release. For impartial support become political also make other contributions that are related to of policy decisions, scenarios should fairly pres- objects, in two senses. the prominent dissemination a major assess- ent knowledge and uncertainty about potential They are subject to ment receives. They may, for example, be used variation on important dimensions. This typi- political forces that as inputs to planning or decision-support cally requires consideration of a wide range of seek to influence their processes that were not part of the original as- potential futures – often a wider range than rel- development, and sessment. In such use, they may gain a more di- evant decision-makers might initially think political reactions to rect connection to decision-making than they plausible, due to well-known habits of conven- them once developed. had in their original production or use. Scenar- tional thinking and excessive confidence. ios of global emissions and the model-based climate scenarios based on them especially lend Sometimes a scenario’s implications for deci- themselves to such derivative uses, informing sions may be obvious. For example, a scenario many different decisions by diverse actors. might represent developments so severe that most people would judge it to demand inter- Scenarios in prominent assessments can also vention. Another might represent developments contribute to the framing of public and policy that most people would judge inconsequential debates. In this role, scenarios inevitably be- or beneficial, so not meriting any intervention. come political objects, in two senses. They are A wide-ranging set of scenarios may include ex- subject to political forces that seek to influence amples of both such extremes. Consequently, their development, and political reactions to such a wide range of potential futures in a set them once developed. These pressures pose of scenarios – even if this is faithful representa- challenges and risks that differ quite markedly tion of present knowledge and uncertainty – from those that apply in using scenarios to in- provides opportunity for partisan distortion and form decision-making, where we tend to as- efforts to make scenarios policy-prescriptive. sume a greater degree of commonality of knowledge, perspective, and interest in the In global change scenarios, conflicts and op- process among participants and some group of portunities for bias arise most acutely over relatively well-defined users. emissions scenarios. Since much of the uncertainty about climate change beyond 2050 comes Within scenario exercises, various actors may from uncertainty in future emissions trends, ac- seek to bias scenarios’ content to help advance tors with strong policy preferences can highlight their policy preferences or their broader politi- emissions scenarios that lend support to their cal objectives, by limiting consideration to fu- views. Those who advocate aggressive mitiga- tures they judge desirable or showing some tion may highlight the highest-emissions sce- problem in an acute state that would appear to narios to emphasize the elevated risk of demand a response. While it is not possible to climate change that would follow. Those who eliminate biases in scenarios, unacknowledged oppose mitigation may highlight the lowest-

60 Global-Change Scenarios - Their Development and Use

emission scenarios to suggest that no action to too confident that the future will extend recent limit emissions is warranted. Because scenarios trends.165 The threshold any single scenario are used when knowledge of causal processes is must pass is to appear sufficiently plausible or weak, it is easy to make any scenario appear instructive to merit consideration in planning salient and likely, even if it is extreme. It is and analysis, and this is a judgment to be made equally easy to probe inside the details of any by developers and users – with enough trans- scenario to find inconsistent or implausible im- parency about underlying assumptions and rea- plications, particularly when a scenario is rich in soning that users can make an informed detail. judgment. A set of scenarios should be constructed so that the range of conditions they rep- But, although political actors may have legiti- resent encompasses present knowledge and mate reasons to highlight one extreme scenario, relevant uncertainties that might influence it is not appropriate for any scenario to domi- mitigation or adaptation decisions. Since A claim that only a nate assessment or consideration of decisions. subjective judgments cannot be avoided in single scenario is A claim that only a single scenario is plausible constructing scenarios, the range provided plausible – especially – especially one near the top or bottom of the should err on the side of being broad rather than one near the top or present range – is a claim to predict the future, narrow, at least initially. Identifying problems which can be readily dismissed. Claims that a with one scenario does not necessarily impugn bottom of the present particular scenario is implausible cannot be so the credibility even of that one scenario, cer- range – is a claim to easily dismissed, however, since scenarios rep- tainly not the whole set, because scenarios can- predict the future, resent only the imperfect judgment of the team not be consistent in every underlying detail. which can be that produced them. Leaving aside scenarios readily dismissed. that violate clear principles of science (e.g., one In subsequent revisions as knowledge advances, whose energy assumptions violate the laws of scenarios can continue to play their role coordi- thermodynamics) or economics (e.g., one that nating assessments and framing policy debates presumes a large new capital stock in a few with more focus and less arbitrariness. Contin- decades without the investments needed to cre- uing research and analysis might come to ate it), it is possible to construct pictures of the identify some scenarios as severe in their con- next century so extreme or unprecedented that sequences and others as inconsequential, or most observers would agree they do not merit might revise the initial characterization of the serious consideration. But short of such an ex- determinants and feasibility of particular sce- treme – which describes no global-change scenarios, including suggesting that some are too nario discussed here or known to us – claims unlikely to merit serious consideration. These that a broad class of potential futures is implau- judgments can feed into decisions about con- sible should have to pass a high hurdle. Identi- tinuing analysis of scenarios: which ones can be fying specific extreme or implausible elements dropped and what new ones should be added. within a scenario does not suffice to make this One major basis for updates in scenarios will be case, since virtually any scenario will be found policies adopted, which can set a baseline to to contain these if scrutinized closely enough. focus further deliberations. Perfect attainment Nor does identifying ways that a scenario of fu- of targets and success of policies should not be ture change diverges from some established assumed, but scenarios can focus subsequent trend or pattern, since established trends can debate by posing such questions as “What if we and do change. just meet this target; what if we fall short by this much; and what if we exceed it by this much, or Historical studies of forecasting exercises such adopt these additional measures?” as energy forecasts have repeatedly found them

165 Smil 2005, Greenberger et al. 1983.

61 The U.S. Climate Change Science Program Section 4 - Challenges and Controversies in Scenarios for Climate Change

BOX 4.4. Scenarios of Ozone Depletion in International Policy-making166

Emission scenarios of chlorofluorocarbons (CFCs) and other ozone-depleting chemicals substantially influenced policy debates over control of these chemicals to protect the ozone layer. Until the early 1980s, these policy debates used a convention to project ozone losses that originally served as a simplifying research assumption: constant emissions forever. This convention has obvious research benefits, like the simple doubled-

CO2 equilibrium scenarios used in climate models. It standardized input assumptions, allowing exploration of scientific and modeling uncertainties without the confounding effect of different emissions assumptions. Moreover, this convention made no claim to realism, and so avoided distracting arguments over whether one emissions projection or another was more realistic. Nevertheless, the resultant calculations were frequently mistaken for projections of realistic future trends. The question of what future emissions trends were likely only became prominent in policy debates around 1983. World CFC production had dropped nearly a third in the late 1970s due to both regulatory and market-driven reductions in their largest use, aerosol spray propellants, and declined further in the early 1980s recession. It was widely argued that further restrictions were unnecessary—CFCs’ major markets were saturated and further growth was highly unlikely. The resumption of sharp growth in 1983 undermined this claim, making it clear for the first time that managing the ozone risk required considering scenarios of CFC growth as well as steady-state and decline. How much emissions might grow and what that would mean for the atmosphere remained highly controversial, however. Emissions of other chemicals complicated the picture. Advances in stratospheric chemistry showed that future ozone loss depended not just on CFCs, but also on emissions of several other gases, including carbon dioxide, methane, and nitrous oxide. But the knowledge and computing capacity to credibly model interactions among all these pollutants only began to appear in the early 1980s. In 1984, a major scientific assessment conducted the first standardized comparison of multiple stratospheric models using a few simple scenarios of emissions trends for CFCs and other chemicals. This exercise had the striking result that under a wide range of trends in other emissions, constant CFC emissions would lead to only very small ozone losses, while CFC growth above about 1 percent per year would lead to large losses. This result, together with resumed growth in CFC production, had a powerful effect in breaking the deadlock in international negotiations that had persisted since the mid- 1970s. Although not the only factor that mattered, this result was crucial in persuading long-standing opponents of CFC controls to accept limits on their future growth. This decisively shifted the agenda for the subsequent negotiations that in 1987 yielded agreement on the Montreal Protocol, which cut CFCs by 50 percent. In this debate, scenarios used in model-based projections of ozone loss identified divergent trends in future risk that were robust to a wide range of assumptions about trends in other emissions over which there was disagreement. By parsing projected futures into high-risk and low-risk cases, scenarios served to coordinate and simplify a policy debate and so help to focus an agenda for collective decision-making.

166 This example is drawn from Parson 2003.

62 Global-Change Scenarios - Their Development and Use

BOX4.2.2. Climate-Change Scenarios for the Insurance Industry

The insurance and reinsurance industries face large financial risks from climate change. These are present in many business lines, but the clearest risk is in insurance for property damage from weather-related events, especially windstorms and floods. In the past two decades, weather-related insurance losses have increased rapidly. By some estimates losses have doubled, even controlling for population and insured value – a much faster increase than that in losses from non-weather events. Climate change is likely to increase insurance risks in multiple ways, increasing the frequency and severity of loss events and also their correlation. Historically based pricing, which is often required by regulations or market conditions, can compound insurers’ vulnerability by preventing them from anticipating and adapting to a changed risk environment. Insurance companies do not use scenarios of future climate change in pricing decisions, because property and casualty contracts are written for short periods, usually one year. Since 1992’s Hurricane Andrew, these have mostly been priced using historically based Catastrophic Event Risk Models (Cat models). These estimate losses using a simulated distribution of storm conditions based on historical experience, together with estimates of the durability of insured property. While future climate change poses no risk for these short-term pricing decisions, insurers are concerned that climate change may already have invalidated the historical distributions on which these models are based, by increasing either the probability of severe events or the correlation among them. Two published exercises have used climate-change scenarios to explore longer-term risks to the insurance industry. The first, conducted for the Association of British In- surers in June 2005, examined potential impacts of climate change on the costs of extreme weather events (both insured and total economic costs) under the six SRES marker scenarios, as well as IS92a and CO2 stabilization at 550 ppm. The analysis calculated changes in losses due to US hurricanes, Japanese typhoons, and European windstorms. The second scenario exercise, conducted by Harvard Medical School researchers with sponsorship by Swiss Re and the United Nations Development Program, used two scenarios of 21st-century climate change to examine potential impacts on human and ecosystem health, and associated economic costs, not limited to the insurance industry.

The two climate scenarios both assumed CO2 doubling by approximately mid-century, one with continued incremental climate changes and one with hypothesized nonlinear impacts and abrupt events. The exercise examined potential changes in infectious and water-borne diseases, asthma, agricultural productivity, marine ecosystems, freshwater availability, and natural disasters (including heat waves and floods). The analysis was based primarily on qualitative judgments. The first scenario showed increased property losses and business interruptions relative to recent trends, emergence of new types of health-related losses, and increasing difficulty in underwriting. The second scenario was qualitatively similar but more severe, with substantial increases in both average losses and variability leading to large premium increases and withdrawal of insurers from many markets, particularly along coastlines. As many insurance firms succumbed to mounting losses, those remaining established strict limits on coverage, shifting more exposure back to individuals and businesses. Neither of these exercises was connected to any specific, near-term business decision faced by insurance companies. Both could serve longer-term decision-making, however, including planning for reserve accumulation, providing supporting analysis for advocat- ing mitigation and adaptation policies, and supporting changed regulations to allow more flexible pricing of risks experiencing long-term increases. Such exercises can also serve to inform firms’ long-term risk-avoidance strategies, including decisions to exit certain areas of business.

63 The U.S. Climate Change Science Program Section 4 - Challenges and Controversies in Scenarios for Climate Change

BOX 4.6. Scenarios of Climate Impacts in the Columbia River Basin

In conjunction with the US National Assessment, researchers at the University of Wash- ington studied climate impacts on the Columbia River system, which is the primary source of energy and irrigation water for the Northwest and one of the most intensively managed river systems in the world.167 The project examined the response of annual and seasonal flows both to existing patterns of climate variability – the El Niño Southern Os- cillation (ENSO) and the Pacific Decodal Oscillation (PDO) – and to projected 21st century climate change. The study projected climate change through 2050 using eight climate models driven by

one emissions scenario (1 percent per year CO2 concentration increase), which on average projected 2.3°C regional warming by the 2040s with precipitation increases of 10 percent in winter and a few percent in summer. In the Columbia, these changes are projected to increase flows in winter (because there is more winter precipitation and more of it falls as rain) and to decrease flows in summer (because there is less snowpack and it melts earlier in the spring). The impact of summer decreases is likely to be substantially more serious than that of winter increases. Because the Columbia is a snowmelt-dominated system, winter flows could double or even triple and remain below the present spring peak. Assessing the impacts of these flow changes requires assumptions about water demands and system management. The study used a reservoir operations model that calculated the combined effects of flow changes and alternative system-operation rules on the reliability of different water-management objectives, such as electrical generation, flood control, irrigation supply, and preserving flows for salmon. Under historical climate variability, all objectives can achieve high reliability in high-flow years (i.e., in the cool phase of ENSO or PDO), but conflict between them occurs in low-flow (warm) years, when only one top-priority objective can be maintained at or near 100 percent reliability and other uses suffer substantial risks of shortfall. Different operating rules distribute this risk among uses. Under the projected climate and flows of the 2040s, this model showed a pattern of competition between uses similar but additional to what already prevails in low-flow years, suggesting increases in already sharp conflict among uses over flow allocations. One objective could be maintained near full reliability, but other uses suffered reliability losses up to 10 percent from the changed climate, in addition to effects of continued climate variability. In this analysis, scenarios helped to illustrate interactions between management decisions and climate change and variability, and to explore opportunities and limits for adaptation through management changes alone, with no change in infrastructure or larger-scale policies. This analysis has not been incorporated into any operational decisions, but has been integrated into the Fifth Conservation Plan issued by the Northwest Power and Conservation Council.168 More detailed assessment of climate-change impacts would require extending this analysis to include projected changes in water demands, both through direct climate effects and through scenarios of regional economic and population growth, allowing a more realistic assessment of potential effects of new water-management investments and changes in large-scale policies to alter water demand, balance competing uses, or improve coordination among the multiple organizations involved in managing the river system.

167 Mote et al. 2003, Payne et al. 2004. 168 www.nwcouncil.org/energy/powerplan/plan.

64 Global-Change Scenarios - Their Development and Use

4.3. THE PROCESS OF narios. While high-level decision-makers are DEVELOPING SCENARIOS: not usually involved in the detailed work of sce- EXPERT-STAKEHOLDER nario construction, they or their surrogates may INTERACTIONS be intensively involved in problem definition, identification and elaboration of key uncertain- Scenario exercises are collaborative activities ties, large-scale scenario design, evaluation and that need to be managed. As Section 1 dis- criticism of scenario outputs, and deliberation cussed, managing a scenario exercise includes over lessons and implications. Their level of in- deciding who participates, what jobs they are volvement must be high for results to be useful, assigned and how these jobs fit together, how particularly if a major purpose of the exercise disagreements are resolved, and how much time is to challenge decision-makers’ assumptions and money are dedicated to the exercise. These and promote creative thinking. matters can be decisive for the success of an ex- Managing a scenario ercise. For some of them, the nature of chal- In these areas, scenarios typically serve a clearly exercise includes lenges and tradeoffs they pose are fairly identified, relatively small and homogeneous set deciding who obvious. For example, scenario exercises need of users who have some degree of agreement on participates, what jobs enough time to build a team, research scenario what values they are trying to advance, what is- components, consult with users, and dissemi- sues are relevant, and what choices are feasible, they are assigned and nate results, but often too little time is available, acceptable, and within their power and author- how these jobs fit so various compromises must be made. Adding ity. This is most clearly the case when scenar- together, how participants expands the expertise and the range ios are developed for a single organization, but disagreements are of views represented, but increases the time also applies to scenario exercises for larger resolved, and how needed for team building and internal commu- groups that are sufficiently homogeneous in much time and money nication. Delegating parts of the exercise to their interests and perspectives, e.g., scenarios are devoted to the smaller groups can overcome this tradeoff, but for property and casualty insurers, for organized exercise. These matters can introduce coordination problems and in- labor in the United States, or for European en- can be decisive for the consistencies among groups. Accepting exter- vironmental groups. In such context, the prob- success of an exercise. nal direction on scenario exercises increases the lems of deciding participation are likely to be chance that decision-makers will take the sce- manageable. narios seriously, but also increases the risk that they are seen as biased or simply reflect con- Intensive user involvement has also been advo- ventional wisdom. These issues pose various cated in developing scenarios for climate challenges, but the challenges are not unique to change. This is obviously correct when climate- scenario exercises. change scenario exercises serve specific, clearly identified user groups. The strongest examples The more central process problems for scenar- are scenarios to support narrowly targeted as- ios concern the relationship between experts sessments of impacts and adaptation in particu- and stakeholders in the design, creation, evalu- lar industries, resources, or regions, e.g., ation, and application of scenarios. There has scenarios for coastal managers considering the been substantial experience and research in establishment or revision of setback lines for processes for involving stakeholders in envi- coastal-zone construction as sea level rises,170 ronmental decisions, in the United States and for rangeland managers considering the pur- other regions.169 In the most well-established chase of conservation lands or easements for the areas of scenario use – e.g., strategic planning purpose of providing migration corridors, or for for corporations or other organizations, and mil- insurance and reinsurance firms examining the itary and security planning – it is widely under- nature of climate-change risks they may face stood that there should be close, intensive and potential responses. In such cases, inten- collaboration between developers and users in sive participation of users is relatively easy to the production, revision, and application of sce- achieve and provides access to valuable expertise and assurance of practical utility.

169 Chess and Purcell 1999; Gregory and McDaniels 2005; Holling 1978; NRC 1996, 2005; Renn et al. 1995. 170 McLean et al. 2001.

65 The U.S. Climate Change Science Program Section 4 - Challenges and Controversies in Scenarios for Climate Change

But climate-change scenarios typically serve such diverse users, especially in the initial de- larger and more diverse sets of users and stake- sign of a scenario exercise, and in the evalua- holders. This is especially true for scenarios tion and refinement of scenarios for relevance, produced in large-scale, official assessments practicality, and utility. In principle, the re- such as the IPCC or US National Assessment. quired approach is to involve a reasonably di- Climate-change stakeholders – defined by the verse and representative group of users and CCSP as “individuals or groups whose interests stakeholders, as well as an appropriate range of (financial, cultural, value-based, or other) are disciplinary and modeling experts, while keep- affected by climate variability, climate change, ing the size of the scenario team manageable. or options for adapting to or mitigating these But the judgments about participation and rep- phenomena”171 – are an enormous group, di- resentation needed to carry out this approach in verse in their interests and responsibilities. any particular scenario exercise will be complex Climate-change and challenging. Even when the set of all potential users is nu- stakeholders are an merous and diverse, there may be some types of Can a scenario process be completely open? In enormous group, users who are clearly identified – e.g., climate political settings, some insulation from users diverse in their modelers who need input from emissions sce- may be needed to insure consistency across par- interests and narios or impact assessors who need input from ticipating models and analyses. Whatever ap- responsibilities. climate-change scenarios – and who have highly proach to stakeholder participation is adopted, specific scenario needs, including such prosaic numbers must be kept manageable. Despite re- factors as data format and resolution. Close cent progress in scenario methods allowing a consultation with such users is clearly impor- substantial increase in the number of partici- tant, especially when their desires exceed what pants, there are still practical limits. Although scenario developers can confidently provide, requirements for expertise external to the core e.g., when climate modelers need emissions scenario team increase with scenario complex- data at fine spatial resolution and for specific ity, a scenario process is unlikely to work with gases or aerosols, which are not readily avail- a hundred people in the room. A few scenario able from the energy-economic models used for processes have engaged much larger numbers emissions scenarios. These situations call for of participants, but these have greatly reduced particularly close and sustained consultation, so the complexity of the scenario-creation process the two sides can understand each other’s needs by limiting it to specifying inputs to a single in- and capabilities in enough detail to develop teractive model, or have involved large numbers workable resolutions. of participants in independent, parallel sessions interacting with a computer-based model or sce- Other users, however, may be numerous, diverse nario construction system.172 These tensions be- in their disciplinary foundations, methods, and tween representational realism, participation, tools, and not clearly identified. Their infor- and managerial feasibility pose challenges for mation needs may have some commonalities design of processes of representation and con- but substantial differences. They may even have sultation in scenario development, on which fur- points of conflicting interest in the construction ther progress is needed. and use of scenarios. The general case for stakeholder involvement remains strong with

172 See, e.g., Envision Sustainability Tools 1999, Roth- man et al. 2003, Stockholm Environment Institute 171 CCSP 2003:112. 1999.

66 Global-Change Scenarios - Their Development and Use

BOX 4.7. Scenarios in Acid-Rain Assessments: Two Approaches

Two programs, one in the United States and one in Europe, developed scenarios in integrated-assessment models of acid rain to inform policy decisions over sulfur emissions. Among many other differences, the two programs differed strongly in their approaches to involving stakeholders and in their effectiveness at informing decision-making. The US National Acid Precipitation Assessment Program (NAPAP) was created in 1980 as a 10-year program to study all aspects of acid deposition: emissions, transport and deposition, impacts, and control strategies.173 Managed by a committee of six government agencies and supported by a full- time staff office, the program involved roughly 2,000 researchers.174 Although charged to conduct both scientific research and assessment, NAPAP strongly emphasized research. Its assessment report was opaque on the origin and interpretation of its scenarios, and did not use the scenarios to integrate across the issue or examine implications of alternative policies. Overall, NAPAP is regarded as having succeeded as a research program, but fallen critically short in providing useful information for decision-making.175 An alternative approach to acid-rain assessment was taken in Europe as part of the policy debates under the Convention on Long-Range Transboundary Air Pollution (LRTAP). The core of this assessment was a cooperative program to monitor and model acid emissions, transport, deposition, and impacts. In contrast to NAPAP, this program focused more on assessment than research, being specifically established to inform the policy process.176 Scientific models of components of the acid-rain issue were chosen to contribute to a simplified integration of the problem; scenarios of emissions and controls were chosen in consultation with officials, in an attempt to replicate the policy alternatives under consideration. The culmination of this pursuit of simple, accessible, and policy-relevant analyses was the RAINS model, developed at the International Institute for Applied Systems Analysis (IIASA) in Austria. As a result of its flexibility, ease of use, and relevance to policies under consideration, the RAINS model was used extensively by policy-makers in the negotiation of sulfur-control agreements under the Convention, and had substantial influence over the controls adopted.177 The contrast in approach and outcome between these two programs suggests the potential value of close interaction between experts and stakeholders in producing scenarios, at least when the stakeholders are relatively expert officials responsible for a specific set of decisions. In the European case, such close interaction helped to ensure the credibility of baseline emissions scenarios and the relevance of proposed control scenarios, despite the diverse and sometimes contending interests of the participating officials. The contrast between the two programs also suggests that there can be significant tradeoffs between scientific and assessment objectives in programs that seek to integrate the two activities.

4.4. COMMUNICATION But, in all likelihood, most users to whom sce- OF SCENARIOS narios must be communicated did not participate directly in scenario development. Scenarios related to climate change must be communicated to multiple audiences, with di- Although specific needs will vary from case to verse interests and information needs. Involv- case, any communication of scenario-based ining users in scenario development can aid formation to a large, diverse public audience is subsequent communication in various ways – likely to require a few common elements. First, e.g., by ensuring that scenarios are understand- in addition to the scenarios’ content, informa- able and practically oriented, and helping to dis- tion should be provided about the process and seminating scenarios to their constituencies. reasoning by which the scenarios were developed. This allows users and stakeholders to un-

173 NAPAP 1982, Herrick 2002. 174 Herrick 2002. 176 Gough et al. 1998. 175 Roberts 1991, Cowling 1992, Russell 1992, Miller 1990, Perhac 1991, Rubin 1991. 177 Levy 1995.

67 The U.S. Climate Change Science Program Section 4 - Challenges and Controversies in Scenarios for Climate Change

derstand and critique scenarios, and to deter- as well as various graphical and tabular for- mine their own levels of confidence. Second, mats.178 In the US National Assessment, cli- scenario developers should identify the uncer- mate scenarios and other related information tainties considered. A particularly important were provided to participating assessment teams distinction to communicate clearly is between in several formats (e.g., tabular summaries, scientific uncertainty and scenario uncertainty, models, graphic representations), through web- including explicit statements of when and how sites backed up with workshop presentations. In

scenarios change (e.g., the reduced SO2 projec- the IPCC, the Task Group on Data and Scenario tions in the IPCC SRES scenarios), and clear Support for Impact and Climate Analysis explanations of the effects of such changes. (TGICA) was established in 1997 to facilitate Third, related to uncertainty, developers should distribution of climate scenario data, model re- acknowledge the unavoidable elements of sub- sults, and baseline and scenario information on Developers should jective judgment in developing scenarios, and other environmental and socio-economic con- be prepared to explain and defend the judg- ditions, for use in climate impact and adaptation acknowledge the ments they made. Fourth, when particular sce- assessments. Data, scenarios, and supporting unavoidable elements narios were constructed to have specific information are distributed over the internet by of subjective judgment meanings – e.g., a reference case, a plausible the IPCC Data Distribution Center (DDC).179 in developing scenarios, worst-case, or the exploration of a particular and be prepared to causal process taken to its extreme – these To compactly communicate uncertainty in cli- explain and defend the should be clearly conveyed. Fifth, if scenario mate scenarios, the TGICA and several national judgments they made. developers have articulated any indicators of the scenario efforts have developed various graph- confidence they place on scenarios or distribu- ical methods, including scattergrams showing tions of associated variables, this information the range of projected temperature and precipi- and any supporting reasoning should also be tation changes generated by several climate made available. models using four SRES marker scenarios, and comparing these projected changes to estimates A communication strategy should attempt to of natural variability.180 In Figure 4.2, each data steer users away from certain common pitfalls, point represents one climate-model projection such as choosing one scenario and treating it as associated with a given SRES emissions sce- a highly confident prediction, or taking the nario. Efforts to develop similarly compact rep- range spanned by a set of scenarios as encom- resentations of the distribution of scenarios for passing all that can possibly happen. An effec- extremes as well as annual and seasonal aver- tive strategy of communicating scenarios and ages are underway. their underlying reasoning can help to engage users in the process of updating and improving To help users select climate scenarios for scenarios. Providing transparency rather than impact assessments, an alternative to summa- claiming authoritative status for scenarios is rizing climate-model scenarios in such scatter- likely to increase users’ confidence that the sce- grams is to combine various climate-model narios have reasonably represented current results using statistical methods to construct ex- knowledge and key uncertainties. It also pro- plicit probability distributions for important cli- vides users with the tools to develop alternative mate variables.181 Figure 4.3 shows one such representations if they are unconvinced. method, which assigns weights to model results

In large and complex assessments such as the IPCC and US National Assessment, communi- 178 See, e.g., Svedin and Aniansson 1987. cation of scenarios and underlying information 179 both to various groups within the assessment Information on the TGICA is at ipcc- wg1.ucar.edu/wg1/wg1_tgica.html. The DDC is jointly and to potential outside users poses representa- operated by the UK Climatic Research Unit and the tional and managerial challenges. Scenario de- Deutsches Klimarechenzentrum, with several mirror velopers have experimented with various visual sites around the world. Data are provided via the web or CD-ROM. All data distributed are in the public do- techniques for conveying complex information main. in vivid and understandable form, including 180 Ruosteenoja et al. 2003. landscape representations, maps, and pictures, 181 Raisanen and Palmer 2001; Tebaldi et al. 2004, 2005.

68 Global-Change Scenarios - Their Development and Use

Figure 4.2. Regional scattergram for eastern North America, 2040-2069. The x-axis shows temperature changes in °C, the y-axis precipitation changes in percent. Each point shows one model’s projection under one emissions scenario. A point’s color denotes the corresponding emissions scenario, its shape the corresponding model (per legends in upper left figure). Ovals show 95 percent confidence bounds for natural 30-year climate variability, calculated from unforced 1000-year runs of the models CGCM2 (orange) and HadCM3 (blue). Points outside the ellipses indicated projected climate change significantly outside the range of natural variability, most frequently due to changes in temperature rather than precipitation. (Source: Ruosteenoja et al. 2003.)

1.2

Figure 4.3. 1.0 Constructed probability distributions of model- 0.8 simulated temperature change in 2080-2099 The x axis shows projected 0.6 temperature change in Eastern North America from

Density the 1980-1999 historical average, using 19 climate 0.4 models participating in the IPCC Fourth Assessment Report driven by the SRES 0.2 A2 (red) and B1 (blue) emissions scenarios. Each point on the x axis shows the result from one model. The 0.0 curves above the axis show probability distributions 1 2 3 4 5 6 constructed from these individual model results. Change in Temperature (deg. C) (Source: Tebaldi et al. 2005.)

69 The U.S. Climate Change Science Program Section 4 - Challenges and Controversies in Scenarios for Climate Change

based on their bias in simulating the current cli- vertently implausible assumptions can arise, for mate (smaller biases are assigned higher example, when multiple elements of a scenario weight) and their correspondence with other are specified without cross-checking; e.g., end- model results (outliers are assigned lower year specifications of a region’s population and weights). This method compactly communi- GDP without checking the implied growth rate cates multiple model results, clearly conveying in GDP per capita, or specifying energy-related which ones fall at the top and bottom of the dis- emissions trajectories without checking what tribution (“unlikely to be higher/lower than they imply for resource availability. Avoiding this”), and which fall in the middle of the range. these pitfalls requires thorough cross-comparisons of related values with each other, of ter- This current focus on collections and intercom- minal values with implied time-trends in the parisons of model-based projections with vari- intervening period, and of values within and The current focus on ous emission scenarios represents a new among regions. Scenario developers should not approach for communicating scenario-driven always and necessarily avoid extreme or un- collections and model output to users engaged in assessment precedented outcomes, however. Presenting intercomparisons of and adaptation activities. It has enabled users extreme or seemingly implausible future con- model-based to consider a broader range of emission scenar- ditions intentionally, with an explanation of projections with ios and climate models than was feasible in the how they could in fact arise, can contribute to various emissions US National Assessment and previous IPCC as- several of the major purposes of scenarios, e.g., scenarios represents a sessments. It allows users to consider all avail- shaking up habitual thinking and broadening new approach for able model and scenario combinations to span expectations of what future developments communicating the literature, or to select only scenarios that ex- are plausible. scenario-driven model ceed some threshold of interest or fall within output to users some specified probability range. Future as- But statements about internal consistency in engaged in assessment sessments should benefit from this type of scenarios usually claim much more than the and adaptation multi-model, multi-scenario approach, which mere absence of gross contradictions and inad- activities. allows users more effective and informed choice vertently implausible values. They tend to over scenarios to consider. claim that the multiple elements of a scenario are related in a way that reflects reasonable, 4.5. CONSISTENCY AND well-informed judgments about causal relations, INTEGRATION IN SCENARIOS suggesting that some events or trends are more likely to occur together, some less. Expressing One of the most often stated requirements for the goal as “coherence” rather than “internal scenarios is that they be “coherent” or “inter- consistency” suggests a higher level of per- nally consistent.” This is clearly an important ceived affinity among scenario elements, evok- goal. Scenarios usually specify multiple char- ing normative or even aesthetic aspects. acteristics of an assumed future, whether as multiple elements of a narrative or multiple Expressed in probabilistic terms, statements quantitative variables, so these elements should about internal consistency may be interpreted fit together. Difficulties arise in the pursuit of as claims that alignments of factors similar to such consistency, however, and in some scenario those in the scenario are more likely than other, exercises the pursuit of consistency, together dissimilar alignments. One might, for example, with the goal that scenarios integrate many claim that a scenario with rapid growth in both components of a broad issue such as climate the economy and energy use is more internally change, may jeopardize the validity and useful- consistent than one in which the economy grew ness of the scenarios. rapidly but energy use did not. But where do these perceptions of greater or lesser likelihood Certain elements of internal consistency in sce- come from, and how valid are they? In some narios are unproblematic, such as avoiding cases a well-founded theory or model might say gross contradictions with well known principles that certain outcomes tend to be related. Alter- of behavior of biophysical or socio-economic natively, explicit analyses might connect the systems, and not inadvertently moving far out- claim to underlying assumptions that are open side the bounds of historical experience. Inad- to scrutiny and criticism. But in the absence of

70 Global-Change Scenarios - Their Development and Use such transparent foundations for judgments of ogenous inputs, together with model results, can what scenario conditions are consistent and jointly comprise a scenario that is generated for what are not, these claims can only rest on more some alternative use. diffuse judgments by scenario developers, refined and tested through various deliberative Consistency problems grow when scenario ex- processes – e.g., arguing about the claims, ercises involve multiple models and attempts working through their implications relative to are made to achieve model harmonization. those of alternative specifications, and identi- When scenarios are constructed partly out of fying additional bodies of research and scholar- exogenous inputs provided by a scenario (made ship that can be brought to bear. consistent as much as possible through qualitative or intuitive causal reasoning) and partly out These difficulties can be compounded when, in of models, multiple models are often used. addition to consistency, a goal of scenario “in- Using multiple models in parallel can allow for In the absence of tegration” is also pursued (although the precise more extensive exploration of causal relations, transparent meaning of “integrated” can be difficult to as- and helps to characterize uncertainty in scenar- foundations for certain). The integration of a scenario is a func- ios since different models embody different rep- judgments of what tion of its complexity or breadth, which is resentations of causal processes. It may also related to the number of characteristics it jointly enhance the credibility of the process. But scenario conditions are specifies. In global climate-change scenario ap- models of the same broad set of phenomena – consistent and what plications, integration typically refers to in- e.g., models of the economy and energy sector are not, claims can only cluding all major elements of the causal chain of – frequently differ in which variables they require rest on more diffuse the issue, i.e., multiple dimensions of emissions as exogenous inputs and which ones they calcu- judgments by scenario and their socio-economic drivers, climate, im- late endogenously. In this case, some variables developers, refined and pacts of climate change, and responses. must be specified exogenously for some models, tested through various but are calculated endogenously by others. deliberative processes. Asking a scenario to be integrated in this way imposes on the scenario the burden of captur- This creates various problems for consistency. ing all relevant elements of the future. Such an In general, when scenario exercises are con- expansive scenario may occasionally be needed ducted in this way, some elements are assumed – e.g., for preliminary assessment of a threat for and others are model-calculated. Attempting to which no relevant data or current research ex- avoid this poses even more serious problems, ists. However, the risks of error, bias, and arbi- however. It is not usually possible to arbitrarily trariness in such a scenario are greatly increased, perturb the exogenous input variables so all in- because so much of reality (with whatever un- puts and outputs match across all models, since known causal processes by which it actually op- such perturbations will influence other variables erates) is being stuffed into the scenario. in the model. Consequently, avoiding these inconsistencies will require manipulating internal More likely, an integrated scenario would be relationships within models to make their out- constructed by combining exogenous assump- puts match the specified values, given the com- tions about some elements with model-calcu- mon inputs. But such reverse-engineering of lated values for others. This approach does not internal model relationships to match specified avoid increasing risks of inconsistency and con- outputs, in addition to being exceedingly cum- tradiction as a scenario is expanded, particularly bersome and arbitrary, can corrupt the internal when multiple models are used. Since models logic of models, obscure the interpretation and embody specific, quantitative causal relations significance of results, and make it impossible among variables, they do not require – or indeed to use model variation to illuminate uncertainty. allow – all variables to be specified. Scenarios provide only those exogenous inputs that the For example, in an exercise to generate model does not produce. These scenario-based non-intervention scenarios of potential future inputs should be consistent with each other, but emissions, little insight is likely to be gained to a lesser extent than the precise standard that from defining scenarios in terms of the resulting defines consistency in a scenario. These ex- emissions and forcing the different models to

71 The U.S. Climate Change Science Program Section 4 - Challenges and Controversies in Scenarios for Climate Change

generate these emissions targets.182 It may be explicitly distinguish between three classes of equally fruitless to define scenarios in terms of variables: (1) a minimal set, exogenous to all; GDP and energy consumption trajectories and (2) those specified exogenously for some mod- to force multiple models to reproduce these. els, but generated by others; (3) model outputs, For this reason, multi-model exercises such as whose variation reflects partly model and partly the Energy Modeling Forum usually seek to parameter uncertainty. harmonize only a few of the most essential and commonly used inputs.183 When multiple mod- An alternative way to use multiple models is to els are used to generate scenarios, the most use- let each model produce one scenario, as was ful way to pursue consistency may be to develop done in the selection of the SRES marker sce- common assumptions for the variables furthest narios. With this approach, each scenario rep- back in the causal chain, but the wide variation resents a particular realization of uncertainty When multiple models of model structure can make even this approach over both exogenous inputs and model struc- to harmonization challenging. ture. But this approach confounds model un- are used to generate certainty with parameter uncertainty. It may be scenarios, the most In addition to consistency within a scenario, preferable to cross exogenous inputs with mod- useful way to pursue consistency across scenarios within an exercise els to produce a larger number of scenarios from consistency may be to also requires attention. Ideally, factors not ex- which subsets can be extracted as needed, per- develop common plicitly recognized as the basis for inter-scenario haps organizing these as a nested hierarchy of assumptions for the differences should be consistent across scenar- scenarios similar to the SRES: six marker sce- variables furthest back ios. Or alternatively, all bases for differences narios, 40 SRES scenarios in total, and hun- in the causal chain, but between scenarios should be explicitly recog- dreds of scenarios in the literature review. the wide variation of nized and stated. model structure can There are good reasons to combine narrative make even this When models are used in a scenario exercise, with quantitative approaches, as scenario exer- approach to significant variation in model structures sug- cises have increasingly sought to do. But the harmonization gests less mature underlying knowledge, or at connection between qualitative and quantitative least greater recognition of knowledge gaps, aspects of global-change scenarios has been in- challenging. than when model structures converge and all re- adequate, diminishing the usefulness of the ex- maining uncertainty is over exogenous input pa- ercises due to inconsistencies within each type rameters. For scenarios to provide faithful of scenario and between the two types. This representation of present knowledge and uncer- problem has partly been due to limited time and tainty, this variation should not be suppressed or resources, but has also reflected substantive dif- concealed. Consequently, when scenarios are de- ficulties in linking the two types of scenario, fined over variables that include outputs of some difficulties that have not been understood or participating models as well as inputs, it is crucial managed well. Narrative scenarios typically not to pursue false consistency by forcing mod- specify deep structural characteristics like so- els to match the target outputs through manipu- cial values and the nature of institutions, which lation of their internal causal processes. This is are associated with structural characteristics of suppressing model uncertainty. models such as the determinants of fertility trends, labor-force participation, savings and in- One preferable alternative would be for the re- vestment decisions, and substitutability in the sults of scenario exercises involving both ex- economy. Consequently, the differences among ogenous inputs and multiple models to alternative narrative scenarios, reflecting different basic assumptions about how the world works, correspond more closely to variation of 182 Note that this is not the case if the purpose of sce- model structure than to variation of parameters. narios is to explore the implications of specified limits Better integrating the two approaches will re- on future emissions. If an emission constraint is assumed to be imposed by policy, then different models quire developing ways to connect narrative sce- can be used to explore the implications of that con- narios to model structures, rather than merely straint for costs, technologies, and other impacts. In to target values for a few variables that models this case, caution is needed in deciding what other are then asked to reproduce. This has not hap- model variables, if any, should be constrained. 183 Weyant and Hill 1999.

72 Global-Change Scenarios - Their Development and Use pened because scenario exercises have not had 4.6.1. Uncertainty in simple the capability or resources to direct new model quantitative projections: development, or to induce modelers to under- basic approaches take substantial structural changes to their models. This would require substantial efforts, How these choices are made and their implica- including getting modelers to interact with sce- tions for scenario use and effectiveness are nario exercises in a new way, but might hold closely related to the large-scale decisions in de- more promise for allowing scenarios to usefully signing a scenario exercise outlined in Section inform discussions about large-scale policy 2.1. In particular, the role of uncertainty in a choices for mitigation and adaptation. scenario exercise is strongly linked to scenario complexity, richness, and use. In the simplest 4.6. TREATMENT OF case, a scenario exercise may be dominated by UNCERTAINTY IN SCENARIOS a single quantitative variable, so all uncertainty Representing and could be represented by alternative future lev- communicating Representing and communicating uncertainty is els or time-paths of that variable. This case is so uncertainty is perhaps perhaps the most fundamental purpose of sce- simple that many scholars and practitioners the most fundamental narios. This section discusses how scenarios argue it should not be considered a scenario at represent uncertainties, how these methods con- all.185 Still, even this simple and extreme case purpose of scenarios. nect scenario exercises to simpler formal exer- raises significant issues. We begin here and cises in the analysis of decisions under then move to more complex cases. uncertainty, and what challenges are posed in how uncertainty is represented. It also ad- If we also assume that the probability distribu- dresses several important debates in the treat- tion is known, the situation reduces to a formal ment of uncertainties. exercise in analysis of decision-making under uncertainty. Given a known set of choices and In most scenario exercises, uncertainty is repre- outcomes of each choice under each uncertain sented not in a single scenario, but in variation outcome, alternative choices can be evaluated across several scenarios considered together.184 by formal methods such as seeking the best out- The choices to be made in deciding how to rep- come on average or under some risk-averse val- resent uncertainty include the following: uation scheme, or seeking robust strategies. This decision-analytic approach can be extended to • What characteristics are varied situations of a few uncertain variables with a • By how much these characteristics are var- known joint distribution, multiple decision- ied, separately and together (e.g., should ex- makers who evaluate outcomes differently, or treme values of multiple characteristics be (with somewhat more difficulty) decision mak- combined, or extremes of some combined ers with different probability distributions. with the middle cases of others?) Further relaxation of these simplifying assump- • How many scenarios to create and consider tions moves us toward activities that are more together widely recognized as scenario exercises. First, • What description, documentation, or other if a scenario exercise is addressed to more than information is attached – including whether just a few decision-makers with known choice and how specifically measures of likelihood sets and outcome valuations, scenarios can no are assigned. longer simply be inputs to an analytic exercise, but rather become descriptions of potential future states that must be communicated directly or indirectly to decision-makers for their reflection and deliberation. Second, if distributions of important quantities are unknown, it is nec- 184 When a scenario exercise uses just one scenario, this essary to exercise judgment regarding how to usually presents some specific threat or challenge posed to existing procedures or decision-makers. In these cases, uncertainty is still represented by differ- 185 E.g., Wack (1985a:74) states that such a scenario is ences among scenarios, but the single scenario is im- just “quantification of a clearly recognized uncer- plicitly contrasted to the status quo. tainty.”

73 The U.S. Climate Change Science Program Section 4 - Challenges and Controversies in Scenarios for Climate Change

draw on relevant knowledge to construct and assessment and decision-making. The input describe alternative future values of the quanti- probability distributions are not known with ties, and how to represent these values to users certainty, nor are the structural assumptions that with a manageable number of scenarios. determine the mapping of inputs onto outputs within any particular model. Uncertainty analy- Since scenarios describe future conditions, the sis can embrace this additional level of uncer- distributions of quantities in scenarios cannot tainty, sometimes called “meta-uncertainty,” by be known in the same sense that the distribution stepping up one more level of abstraction – of current characteristics – e.g., the November considering not just uncertain quantities, but daily high temperature at O’Hare Airport – can uncertainty about their uncertainty, or alterna- be known through repeated observations. Prob- tively, probability distributions over probability abilistic statements about future conditions al- distributions of unknown quantities. Methods Probabilistic ways incorporate elements of subjective to represent and process such meta-uncertainty judgment. Many forms of current knowledge – mirror those used for first-order uncertainty. statements about including data, models, and expert judgments – This is an active area of research, but its impor- future conditions are relevant to forming these judgments about tance for assessment methods and their appli- always incorporate future conditions. In constructing scenarios of cation is unclear. This level of abstraction elements of subjective population growth, for example, the distribution increases the difficulty of communicating sce- judgment. Many of observed past growth rates can be used to narios and their underlying reasoning transpar- forms of current construct a range or distribution of plausible fu- ently and comprehensibly to non-specialists. knowledge – including ture values. But while scenarios can draw on, Moreover, since any step of analysis represents data, models, and and be made conditional on, such knowledge, an act of potentially fallible judgment, taking expert judgments – this does not overcome their unavoidable re- the step to meta-uncertainty still does not cap- are relevant to liance on subjective judgments as well. ture all possible uncertainty. It is not clear forming these whether, for purposes of constructing and using judgments about Scenarios can also be based on model repre- scenarios, the explicit separation of uncertainty future conditions. sentations of knowledge of causal processes. in outcomes from uncertainty in probability dis- For example, instead of simply extrapolating tributions brings more benefit than could be past population growth rates, one could use a gained from simple heuristic guidance to as- demographic model that represents trends in sume distributions are wider than initially seems fertility rates, lifespan, and migration to calcu- necessary. late a resultant population trend. Formal modeling can represent the structural relationships A major risk in all scenarios is subjective bias, transparently, reducing the risk of generating in- which can be reduced but not eliminated consistent projections. Structural models can through use of existing data and formal model- possibly also perform better in extrapolating to ing. Judgment is an essential element in con- conditions beyond the observed range of be- structing scenarios, both to apply relevant data havior. Because models represent causal re- and models when these are available, and to lationships among multiple variables, these build future descriptions using less formal models can extend the range of current and his- methods when they are not. The expert judg- torical data that are relevant to projections, al- ments supporting such less formal projections though this may result in an expansion of data may be better founded than mere uninformed needs. Models can also help characterize un- speculation, since there is typically much rele- certainty in future quantities of interest, by al- vant knowledge available beyond what is ex- lowing the uncertainty to be attributed to input plicitly captured in present datasets and models. parameters – explored through sensitivity analysis or simulation techniques such as Monte Approaches to developing expert-judgment Carlo – or to model structure. based projections vary widely in their structure and formality, from simply asking one or more Estimating output distributions based on assumed experts to state their best estimate of some un- distributions of uncertain input parameters does known quantity, to highly structured elicitation not capture all uncertainty of importance for exercises that provide multiple cross-checked

74 Global-Change Scenarios - Their Development and Use

estimates of the same quantity.186 Such meth- highly confident projection, suppressing the un- ods must attend to risks of overconfidence and certainty that scenario developers tried to com- bias, which are well documented in experts as municate by providing, and carefully spacing, well as laypeople. Carefully designed elicita- three scenarios. The same risk applies to any tion protocols can reduce the effects of such bi- odd number of scenarios, leading many devel- ases, e.g., by prompting experts to broaden their opers of quantitative scenarios to the informal estimates of uncertain quantities, but cannot guideline that the number provided should al- eliminate them.187 An additional challenge to ways be even, so there is no “middle” scenario these methods is that there is no generally ac- that users can inappropriately fix on. cepted method for selecting or aggregating estimates from multiple experts. More specific guidance on the appropriate number and range of scenarios must reflect both 4.6.2. How many scenarios, over scenario developers’ sense of the underlying dis- Similarly for global what range? tribution from which scenarios are drawn, and environmental change, their intended use. One must consider whether low-probability risks In communications of scenarios, limited time, departures in both directions from the middle are might need to be resources, and attention usually require that of similar importance, or whether only departures only a few discrete values or time-paths are in one direction need be represented. For exam- considered if their specified, not a complete distribution. Scenario ple, one might judge that in an assessment of im- consequences or their developers must decide how many scenarios to pacts of climate change a scenario drawn from effects on preferred provide and how to space them. the lower tail of potential climate change is likely decisions are large to provide little substantive insight, since in most enough. How many scenarios to provide rests on a judg- cases the impacts of a small-change scenario are ment of the value provided by each additional predictably small. point from the underlying distribution relative to the burden of producing and using each new One must also consider how far a set of scenar- scenario. If the use made of each scenario is ex- ios should extend toward including extreme or pensive – e.g., consuming large quantities of unlikely futures. In estimating unknown quan- time of busy senior people, or running a large tities, many fields of empirical research draw model – then the number of scenarios that can intervals to capture from 90 percent to 99 per- be adequately treated may be very few. The cent probability, but in constructing scenarios 1992 IPCC scenario exercise provided six sep- to inform decisions there may be good reasons arate scenarios, of which nearly all subsequent to consider more extreme and less likely possi- analyses used just one or two. Of the 40 scenarios bilities, whether these likelihood judgments are produced by the SRES process, only 6 (initially 4) expressed quantitatively or qualitatively. As- were highlighted as “marker” scenarios, while sessments and policies in both regulation of most subsequent analyses used just 2 or 3.188 health and safety risks and national security, for example, routinely focus on high-consequence Deciding how many scenarios to provide also risks that are judged much less than 1 percent involves some element of attempting to avoid likely. Similarly for global environmental predictable errors in their use. While the most change, low-probability risks might need to be obvious and frequent choice in providing sce- considered if their consequences or their effects narios of a quantitative variable has been to pro- on preferred decisions are large enough. vide three – one high, one low, and one in the middle – it has been widely noted that this prac- It is often suggested that a set of scenarios tice runs the risk that users will ignore the top should “span the literature” of prior scenarios and bottom, pick the middle, and treat it as a or projections of the same quantities. However, there may be good reasons for a wider or different range, or even a narrower range – al- 186 Morgan and Keith 1995. though developers should be cautious about a 187 Tversky and Kahnemann 1974, Wallsten and Whit- set of scenarios that spans a much narrower field 1986. range than published estimates of the same 188 Initially A2 and B2 were most widely used. More re- quantities. A published scenario may have been cent work has used A2 and B1, sometimes with A1B.

75 The U.S. Climate Change Science Program Section 4 - Challenges and Controversies in Scenarios for Climate Change

constructed to serve various purposes other than difference between a half-meter and a five- providing an independent new estimate of a meter sea level rise over this century – and the quantity of interest. Previous scenarios devel- well-identified mechanism by which such a oped to serve some particular purpose may or large rise could occur – may suggest the impor- may not be relevant to a new scenario exercise, tance of explicitly considering a scenario in- depending on the relationship between their involving loss of one of the major continental ice tended purposes. Moreover, previously pub- masses. But including such a scenario runs the lished scenarios can be highly self-referential, risk that users will assign it a much higher prob- since many published analyses use prominent ability than is appropriate, either because of its pre-existing scenarios as inputs to a new study, vividness and extremity or because they pre- or examine a new model by forcing it to repro- sume that developers’ decision to include the duce some pre-existing scenario. For all these scenario meant that they assigned high proba- Previously published reasons, previously published scenarios are bet- bility to it. When such a scenario is included, ter regarded as one input to the judgment of de- scenario developers have a serious responsibil- scenarios are better velopers of new scenarios than an authoritative ity to communicate, loudly and consistently, its regarded as one input picture of present knowledge that new scenarios different status. to the judgment of must follow. developers of new A further challenge in representing large-scale scenarios than an 4.6.3. Bifurcations and major or discrete changes in scenarios is that there authoritative picture state changes might be many such possibilities, all of them of present knowledge high-consequence but believed to be unlikely. that new scenarios While many uncertainties may be treated as a Including any particular one may mislead both must follow. continuous range of possible values, some un- by exaggerating its likelihood and by strength- certainties may capture large-scale bifurcations ening users’ tendency to ignore others, when or abrupt changes. For climate change, poten- these all represent “unknown unknowns” that tial abrupt changes include melting of major should receive some consideration. The more continental ice sheets or shifts to some new there are, the more the right approach might be mode of ocean circulation.189 Large-scale bi- to shift all scenarios further out to reflect the furcations may also arise from breakthroughs in various mechanisms by which conventional un- energy technology. Such possibilities are typi- derstanding may under-represent the tail of the cally not captured in either historical data (be- distribution, rather than highlighting any par- cause they are by assumption novel), or models ticular abrupt-change mechanism by giving it a (because they would represent a change in the scenario of its own. causal structure represented in models). 4.6.4. Uncertainty in multivariate Abrupt changes can pose particular challenges or qualitative scenarios for deciding the number and range of scenarios to include in an assessment or decision-support As the characterization of future conditions exercise, either because their consequences are within scenarios grows more complex, so does so extreme or because they would fundamen- the process of representing uncertainty within tally change our understanding of how the sys- them. While many of the issues discussed above tem operates. The decision of whether and how in the simplified context of scenarios on a single to consider these uncertainties consequently variable also apply to multi-dimensional scenar- turns on the balance between their probability ios, several additional issues arise. – which is believed to be low but not well characterized – and their high consequences, which The most basic of these is that with multiple must be evaluated relative to the scenarios’ in- dimensions of variation in scenarios, it is nec- tended use. This will be a particularly difficult essary to decide which uncertainties are repre- choice when only a few scenarios are being gen- sented. Even when scenarios include only erated. For example, in a coastal impacts as- multiple quantitative variables, it is no longer sessment the enormous consequences of the possible for a few scenarios to span all corners of the joint distribution of these variables. Rather, they must combine variations in ways 189 NRC 2002.

76 Global-Change Scenarios - Their Development and Use that are most illuminating and important for the teristics associated with each realization purpose at hand, massively reducing the dimen- through both qualitative reasoning to fill in a sionality of the problem to make it intelligible narrative, and assembly of data and model re- for users. In addition, increasingly detailed and sults to build a parallel quantitative description realistic scenarios often specify characteristics to the extent this is judged useful. Repeated, that are qualitative, or described less precisely critical iteration between the qualitative and than cardinal variables. For example, alterna- quantitative elements is conducted, to bring ad- tive scenarios might specify that current trends ditional relevant knowledge and expertise to of globalization increase, stagnate, or reverse, bear and to check for consistency. or that decision-making capacity on climate change increases or decreases. Such charac- Even rich narrative multivariate scenarios must teristics may be judged crucial to include be- imply certain claims of likelihood. Every sce- cause they may be among the most important nario included must be deemed likely enough The approach most drivers of preferred choices or consequences to merit the resources and attention spent on de- widely proposed to of concern. veloping and analyzing it. This applies even to represent key extreme-event scenarios that are intentionally uncertainties in Scenarios of this kind pose substantial further constructed to capture the low-probability tail challenges in representing uncertainty and in- of the distribution, since even they must be per- scenarios is to seek terpreting its meaning. Relative to the simple ceived likely enough to merit time and attention underlying structural quantitative scenarios we have considered up to given their severity. Since users would reject uncertainties that this point, these lie in a much higher dimen- any scenario that they persistently judged too satisfy two conditions: sionality space of future possibilities; they may implausible to consider, when decision-makers they appear to be not lie in any ordinal relationship to each other; find a scenario exercise useful, it validates de- most important in and they may include characteristics whose def- velopers’ judgment that each scenario was influencing outcomes initional boundaries are not precisely specified. likely enough to consider. of concern, and they Defining a small set of scenarios to reasonably are linked with span the most important uncertainties is conse- In a purely mathematical sense, any one specific variation in many quently even more difficult than for simple rich multivariate scenario must be arbitrary and other factors. quantitative scenarios. of vanishingly small probability. There are, however, ways in which it may be reasonable to The approach most widely proposed to repre- assign non-zero probabilities to multivariate sent key uncertainties in such scenarios is to scenarios. First, if scenario designers in fact seek underlying structural uncertainties that sat- succeed at identifying a few deep structural un- isfy two conditions: they appear to be most im- certainties that strongly condition outcomes on portant in influencing outcomes of concern or many other characteristics in a scenario, then relevant decisions; and they are linked with vari- the richness of a scenario description need not ation in many other factors. These underlying imply that it is vanishingly unlikely. Whether uncertainties can be simple discrete states such this is true or not is a judgment to be made by as peace or war, prosperity or stagnation; or, as scenario developers and users in each applica- in several major global environmental scenar- tion. If they are sufficiently careful in their de- ios, they can be deeper societal trends, such as velopment and critical examination of more or less globalization or shifts in societal scenarios, their judgment may well be correct. values toward greater environmental concern, But there will often be no way to further test from which variation in many factors is as- these judgments, so it is of course possible that sumed to follow. the proliferation of additional detail in scenarios – even detail that developers and users recog- This approach, formalized in the Shell scenarios nize is crucial for determining valued outcomes method,190 involves two steps: first identifying a and preferred choices – is arbitrary or erroneous. small number of fundamental uncertainties and a small set of alternative realizations of each; A second way in which rich, detailed scenarios and then elaborating additional future charac- may be judged sufficiently likely to consider concerns the precision with which scenario characteristics are specified. In rich multivari- 190 Shell International 2003.

77 The U.S. Climate Change Science Program Section 4 - Challenges and Controversies in Scenarios for Climate Change

ate scenarios, many characteristics are often nent long after the decision had been made by a specified diffusely: economic growth may be writing team no longer in operation, it would merely “high” or “low,” rather than being stated have been virtually impossible for the group to as a particular value. Even when a characteris- retrospectively assign such probabilities. But tic is stated quantitatively, its particular value rather than rely on this argument of managerial may be treated as merely illustrative of a range infeasibility alone, SRES organizers offered a of similar values; e.g., annual GDP growth vigorous substantive defense of their initial de- might be set at 4 percent because a user needs a cision. This defense relied in part on the state- numerical model input, but it is understood to ment that the six marker scenarios were all represent a broad range of similar values that all “equally sound,” without providing any guid- count as “high” growth. Interpreted in this way, ance regarding what this meant other than ex- a multivariate description may remain likely plicitly denying that it meant “equally likely.” The simpler the enough to merit examination – and indeed, a Describing each of the six marker scenarios as modest number of scenarios may exhaust the set “equally sound” represents the entirely reason- contents of of potential futures that matter for the issue at able case that in the developers’ judgment these scenarios, the hand. Here one is not assigning likelihood to all needed to be considered seriously – without more readily they the precise numerical assumptions used to flesh making any further judgment as to their likeli- lend themselves to out the details of a scenario, but rather to a thick hood. While clearly frustrating to those wanting explicit slice of future conditions that resemble that sce- to use the scenarios as a basis for policy, the re- quantification of nario more than the other scenarios in the set. sult is entirely consistent with the IPCC man- probabilities. date to do assessment, but not to reach policy 4.6.5 The debate conclusions. over quantifying probabilities However, this debate will continue; it rests in A major debate in the use of global-change sce- part on different conceptions of the meaning narios has concerned whether or not to specify and typical contents of a scenario. The simpler quantitative probabilities associated with sce- the contents of scenarios, the more readily they narios. This debate is central to the meaning lend themselves to explicit quantification of and use of scenarios, and has been sharpest over probabilities. When scenarios consist only of the IPCC’s SRES scenarios. Developers of the alternative time-paths of a single quantitative SRES scenarios decided at the outset of their variable, or one such variable is of predominant process that they would make no attempt to as- importance, it is straightforward and sensible to sign probabilities to scenarios, in part because understand the intervals between those time- they were adopting the Shell approach of de- paths to have probabilities associated with them veloping scenarios from storylines, in which and there are several strong arguments for being quantitative probabilities are usually avoided. explicit about these probabilities. First, stating After the scenarios were published, several crit- probabilities allows comparative risk assess- ics argued that since the most prominent and ment between scenarios and explicit exploration important outputs were the projections of emis- of risk-reducing strategies.191 Second, sophisti- sions under the six marker scenarios, it was nat- cated decision-makers whose choices depend ural – and essential for development of rational on uncertainty in these variables need probabil- climate-change policy – to describe the distri- ity information about possible values, not just a bution of emissions in probabilistic terms. For set of alternative values, to evaluate choices – example, how likely are 2100 emissions to lie whether their approach to decision-making is above the 30 GtC of scenario A2 or below the based on expected values, risk-aversion, seek- 5.2 GtC of B1? Should the range spanned by ing robust strategies, or some other approach. all 40 SRES scenarios be understood to com- Finally, when such scenarios are presented with- prise 90 percent of all probability? 99 percent? out probability judgments, users may attach All of it? their own, often via simple heuristic devices that may misrepresent the developers’ understand- Developers of the SRES scenarios stood by ing. For example, many subsequent users of the their initial decision not to quantify probabilities. Since the controversy only became promi- 191 Webster 2003.

78 Global-Change Scenarios - Their Development and Use

SRES emissions scenarios have simply assumed ture all the uncertainty of concern to the deci- the probabilities they needed to conduct further sion-maker – i.e., scenarios are intended to be assessments, using such simple devices as mutually exclusive and collectively exhaustive – counting scenarios or assuming a uniform dis- there may even be a reasonable basis for nu- tribution over the entire range. merical probability.

Opponents of explicit quantification of proba- The second argument for rejecting probabilistic bilities do not dispute that such probabilities can description of socio-economic conditions is coherently be assigned to simple scenarios in based on “reflexivity” – the proposition that one or two quantitative variables. Rather, they scenarios may influence the behavior or deci- raise principled objections to the appropriate- sions driving the scenarios, so probability judg- ness of attempting to quantify probabilities for ments about scenarios could reflect back on more complex scenarios, particularly those in- themselves, becoming either “self-fulfilling” or Concern about volving socio-economic conditions, as well as (more plausibly) “self-denying” prophecies. reflexivity appears practical objections to the use of probabilities Section 4.1 addresses this issue in some detail, more serious for even in the case of simple quantitative scenarios. in particular in the distinction between how to scenarios prepared in treat mitigation decisions in scenarios to inform Many researchers are less comfortable using mitigation decisions and impacts or adaptation- close consultation probabilities for complex scenarios that include related decisions. We might only add here that with national explicit socio-economic elements than for un- for scenarios of global emissions, reflexivity mitigation policy- certainties that are purely bio-physical, such as could only operate if both the influence of sce- makers, and it is for probabilities of different rates of climate change, nario judgments on their users’ behavior and the this reason (among conditional on a particular emissions scenario. influence of their users’ behavior on global others) that we judge Four main arguments are advanced against the emissions were extremely strong. Moreover, it explicit attempts to use of probabilities for such scenarios. is not evident why scenarios with explicit like- assign probabilities lihood judgments should raise this concern, less valuable for First, some argue that the large multivariate while scenarios presented without such judg- scenarios prepared in space of possibilities from which such scenarios ments – which also presume some claims of such settings. are drawn, and the vague and qualitative way plausibility or likelihood – should not. Concern that some scenario characteristics are specified, about reflexivity appears more serious for sce- make it impossible to coherently define the narios prepared in close consultation with na- boundaries of the outcome space to which prob- tional mitigation policy-makers, and it is for this abilities are being assigned. There is no way to reason (among others) that we judge explicit at- clearly define the interval “between” one sce- tempts to assign probabilities less valuable for nario and another; and if probability is attrib- scenarios prepared in such settings. uted to a slice of possibilities around each scenario rather than to the intervals between Third, some argue that it should not be scenario them, is it not possible to define clearly the developers or experts who make judgments boundaries of the slice to which the probability about likelihood of alternative scenarios, but is assigned. To the extent that scenarios de- users – particularly when scenarios are used to scribe different types of worlds, which are dis- inform high-stakes public decisions. But this tinguished from each other by alternative depends on the details of the content and use of resolution of a few key uncertainties – e.g., high scenarios. For some scenario elements in some or low growth, high or low globalization – settings, particularly use of scenarios to advise where the location of the boundary is not pre- specific policy decisions, the scenario users cisely specified, it may be difficult to create a may be as expert as the developers in associated shared understanding of these boundaries be- uncertainties and risks, or more so. But in such tween users and creators. But if assigning a pre- settings, the use of scenarios normally high- cise numerical probability is judged too difficult lights critical examination of these assumptions, in these cases, less precise descriptions such as and users have the knowledge and assertiveness “highly likely,” “more likely,” “less likely,” or to probe, critique, modify, or reject scenario el- “roughly equal” could be assigned. In some ap- ements that they find weak, including probabil- plications where scenarios are intended to cap- ity judgments. When scenarios are produced to

79 The U.S. Climate Change Science Program Section 4 - Challenges and Controversies in Scenarios for Climate Change

serve many diverse users and consequently can- An additional practical argument against quan- not rely on intensive interplay with representa- tifying probabilities is that attempting to do so tive, well-informed, and challenging users, may represent a distraction that uses time, gen- scenario developers frequently have the best ac- erates conflicts, and is of little value to scenario cess to available knowledge relevant to forming users. Whether this is indeed the case, however, probability judgments. Not making these judg- is in part a judgment to be made by scenario ments explicit is withholding information that users, not developers. Opponents of quantified users may need to understand and interpret the probability argue that users typically only need scenarios. If scenarios and their underlying rea- scenarios to pass some probability threshold. soning and assumptions are presented clearly Beyond this threshold, they will seek robust enough, users can make informed choices choices that yield acceptable outcomes under all whether or not to use probability judgments that possibilities, so further refinement of probabil- Overall, we find the are provided. ity serves no purpose. This argument has merit, but only to the extent that it accurately describes arguments in favor of Finally, some argue that probabilities cannot how these scenarios will be used. Quantitative quantifying be known, or even sensibly estimated, for assignment of probabilities to scenarios when probabilities to be socio-economic futures – perhaps because high-stakes decisions are implicated is clearly strongest for socio-economic processes and mechanisms are difficult and contentious, as the SRES contro- scenarios whose intrinsically less knowable than biophysical versy illustrates. Even if this argument cor- major outputs are ones, perhaps due to the unpredictable effects rectly characterizes how scenarios are used, projections of one of human creativity and leadership, and perhaps users might still be able to profitably exploit quantitative variable because causation does not operate in the more detailed probability information if it were (or very few), and human domain as it does in the bio-physical do- available – although one must also consider the weakest for complex main. Although these arguments raise deep risk that non-technical users might somehow be multivariate scenarios philosophical questions, as a practical matter more likely to misunderstand scenarios with ex- with substantial probabilistic projections are routinely done in plicit probability judgments attached (perhaps qualitative or some socio-economic domains, including pro- by taking a stated probability distribution as the jections of population and economic growth, “true” distribution) than to misunderstand a narrative elements. but not, or not well, in others, such as projecting simple collection of scenarios presented with no technological innovation. Provided the basic such probability information (perhaps by taking concept of subjective probability is accepted, the range presented to embrace the totality of weaker knowledge and deeper uncertainties can all possibilities). It is also possible that engag- be accommodated by broadening the relevant ing scenario users in an attempt to assign prob- uncertainties rather than declining to make abilities, even only illustratively, could both probabilistic judgments, but the question remains draw on relevant knowledge of uncertainties of whether the resultant broad uncertainty ranges that they possess more than scenario develop- are meaningful or operationally useful. ers, and provide a valuable device to probe and sharpen their understanding of the situation. Several practical objections have also been Any argument based on the information needs raised to associating explicit likelihood judg- of specific users becomes less persuasive as the ments with scenarios. These include the diffi- set of potential uses and users, and the likely di- culty of developing probability estimates from versity of their information needs, grow larger. multiple information sources that can achieve sufficient agreement from diverse experts, and Overall, we find the arguments in favor of quan- the non-intuitive nature of probability distribu- tifying probabilities to be strongest for scenar- tions in using scenarios to communicate with ios whose major outputs are projections of one non-expert users. These are both valid con- quantitative variable (or very few), and weakest cerns, although active areas of research and de- for complex multivariate scenarios with sub- velopment in expert elicitation techniques and stantial qualitative or narrative elements. The in simple intuitive devices to communicate un- controversy over probabilities in SRES reflected certainty are making some progress in mitigat- in part different perceptions of what type of sce- ing them. narios these were. SRES initially followed a storyline-based process and rejected quantifi-

80 Global-Change Scenarios - Their Development and Use cation of probabilities on that basis. Subsequent Moreover, even for rich narrative scenarios, the efforts, however, consisted predominantly of de- arguments against rendering probability judg- veloping quantitative emissions projections and ments are strongest when the exercise is pro- neglected further development of the storylines. duced for a small number of users with similar Moreover, with a few significant exceptions, responsibilities and concerns. In such a setting, subsequent applications of the scenarios have intensive interaction between scenario develop- principally used their emissions figures, some- ers and users can provide whatever additional times together with population and GDP, and detail about, or confidence in, the scenarios that made little or no use of the underdeveloped sto- users may require to benefit from the scenarios. rylines that lay behind them. The controversy When scenarios serve potential users who are over quantitative probability in this case sug- more numerous and diverse, however, such in- gests that when quantitative projections are a tensive interaction is not possible. As a result, major output of a scenario exercise, developers the value of explicit likelihood judgments in- may have a responsibility to go further in char- creases. To the extent that future global-change acterizing the likelihood of the resultant emis- exercises continue to strengthen their qualita- sions intervals than would be appropriate for the tive aspects and the integration between quali- more complex underlying storylines. tative and quantitative – valuable directions for future efforts – they should still seek to move further toward explicit characterization of likelihood than has been done thus far.

BOX 4.8. The Global Business Network Abrupt Climate Change Exercise

In 2002, the Office of Net Assessments (ONA), a small strategic planning office in the Office of the US Secretary of Defense, asked the Global Business Network (GBN), a consulting firm expert in scenario methods, to develop a scenario of potential national-security implications of abrupt climate change. This request was stimulated by widespread scientific interest at the time in abrupt climate change, particularly shifts in North Atlantic circulation, including a 2002 report by the National Academy of Sciences.192 In addition, several scientific papers had reported changes in Atlantic circulation and salinity that some scientists thought might indicate impending larger disruption, as well as new evidence of rapid climate shifts in the past.193 GBN staff developed the scenario by reviewing scientific literature and informally consulting with climate and ocean scientists.194 They reviewed three past climate events of diverse severity and decided to base their scenario on the one in the middle, the century-long period of strong cooling 8,200 years ago. Coming after an extended warm period, this event brought cooling of about 5 °F over Greenland, with cold and dry conditions extending around the North Atlantic basin and substantial drying in mid-continental regions of North America, Eurasia, and Africa.195 For their future abrupt-change scenario, the authors constructed a path of climate change to reach these conditions by 2020. The pathway involved rapid warming through 2010, as high as 4 – 5 °F per decade in some regions,196 followed by a rapid turn to cooling around 2010, as melting in Green- land freshens the North Atlantic and substantially shuts down the thermohaline circulation. By 2020, hypothesized conditions have approached those of the 8,200-year event – cooling of 5 °F in Asia and North America and 6 °F in Europe, with widespread drying in major agricultural regions and intensification of winter storm winds. The authors acknowledge that the scenario pushes the boundaries of what is plausible, both in the rapidity of changes and in the simultaneous occurrence

continued on next page

192 NRC 2002. 193 See, e.g., Dickson et al. 2002, Hansen et al. 2001, Gagosian 2003, Curry and Mauritzen 2005, Fairbanks 195 Alley et al. 1997. 1989. 196 Note: these regional projections are 5-10 times faster 194 Global Business Network 2004. than the IPCC’s projected global 21st-century warming.

81 The U.S. Climate Change Science Program Section 4 - Challenges and Controversies in Scenarios for Climate Change

BOX 4.8., continued from previous page.

of extreme changes in multiple world regions. They contend that this is defensible and useful, however, for an exercise focused on sketching the nature of challenges posed by a plausible worst case.197 The socio-economic and security implications of the climate scenario were developed judgmentally, in consultation with ONA. Incremental changes are projected for the first 10 years, with general increase in environmental stresses and approximate maintenance of present disparities between rich and poor countries. After 2010, catastrophic cooling in Europe and drying of major agricultural regions worldwide brings widespread shortages of food, due to decreased agricultural production; of water, due to shifted precipitation patterns; and of energy, due to shipping disruptions from increased sea ice and storminess. These shortages produce 400 million migrants over the period 2010-2020, as desperate scarcity generates violent conflict in Europe, Asia, and the Americas. Ex- tending their speculation on security implications into the 2020s, the authors hypothesize widespread southward migration of Europeans and near-collapse of the European Union, sustained conflict in East and Southeast Asia including struggles between China and Japan over access to Russ- ian energy supplies, and increasing political integration of a fortress North America to manage security risks and refugee flows. Controversy and criticism The project was completed in October 2003, its report published in February 2004 and reported in Fortune Magazine the same month.198 A few weeks later, the London Observer claimed to have ob- tained the report secretly and used the scenario to criticize US refusal to join the Kyoto Protocol.199 Subsequent news coverage took up the theme that the report was secret or suppressed, and suggested the reason was that the scenario called for more urgent action on climate change.200 In the resultant controversy, ONA stated – correctly – that the report did not represent US policy, but was merely a speculative consultant’s study. Although the controversy subsided after a few weeks, interest and concern about the possibility of abrupt climate change, although not of this precise character, have continued to grow.201 This scenario is a sketch of an abrupt climate-change event, with little fine-scale detail about the hypothesized changes or underlying reasoning and no attempt to suggest how likely or unlikely such an event is. Rather, it seeks a preliminary answer to the question, what might the worst case look like? Such questions are more often posed in security studies than other fields, because of the unique nature of responsibilities of military organizations – responding to diverse, novel, unknown threats with extremely high cost of failure. Many climate-change decision-makers could likely benefit from such upper-bound scenarios too, but this exercise is the only example of a worst-case scenario produced for climate change. Major official assessments have focused overwhelmingly on average or best-guess projections. But the response to this report vividly illustrates the risks of worst-case or extreme scenarios. Produced in consultation with a sophisticated user – and in this case, one closely connected to senior decision-makers – who thoroughly understands the outer-bound nature of the underlying assumptions, they can be valuable devices for preliminary risk assessment and threat identification. But in a wider and polarized policy debate they are hard to explain and may be misunderstood or mis- represented. Attempting to manage the process through secrecy appears counterproductive, fore- going the potential value such analyses could provide to multiple decision-makers. More promising might be to integrate extreme-case scenarios explicitly into analyses that also present multiple mid- range scenarios.

197 GBN 2004a:7, 2004b; Schwartz interview. 198 Stipp 2004 (released, January 26, 2004). 199 London Observer 2004. 200 San Francisco Chronicle 2004, Providence Journal 2004. 201 E.g., Alley 2004 cites it as a useful worst-case assessment.

82 Global-Change Scenarios - Their Development and Use

Conclusions and 5 Recommendations SECTION

This section presents our conclusions regarding the present state of development and use of scenarios for climate-change applications, and some recommendations for specific changes or initiatives to advance current practice to make scenarios more useful.

Before doing so, we briefly reprise some key definitional points, because uses of the term scenarios are so divergent. We have defined scenarios as descriptions of future conditions produced to inform decision-making under uncertainty. This definition distinguishes scenarios from assessments, models, decision analyses, and other decision-support activities. Scenarios may be developed and used in conjunction with these – for example, scenarios can provide descriptions of potential future conditions used as inputs to such activities – but are not identical to these, and not alternatives to them.

We have also distinguished scenarios from other types of future statements intended to inform decisions, such as projections, predictions, and forecasts. Relative to these, scenarios tend to be more multivariate (but still schematic), tend to be developed in groups, and tend to presume lower predictive confidence. The last condition is the case in part because scenarios tend to be used in situations where the basis for forecasting is less established because of deeper uncertainties, or for situations that pertain to further in the future beyond the range for which there is high confidence in specific projections, even contingent ones.

Having distinguished scenarios from these related activities, we consider a broad set of scenarios of diverse characteristics and uses, including simple and complex scenarios, quantitative and qualitative scenarios as well as various combinations of the two, and scenarios whose primary use and interpretation is positive or normative. Where we intend our conclusions and recommendations to apply to only certain types or uses of scenarios, we state this explicitly. Unless stated otherwise, they pertain to all types of global-change scenarios we are considering.

83 The U.S. Climate Change Science Program Section 5 - Conclusions and Recommendations

5.1 USE OF SCENARIOS IN of background scientific knowledge, we ex- CLIMATE-CHANGE DECISIONS pect this trend to continue, and to broaden to other types of climate-related decisions. Scenarios can make valuable contributions to climate-change decision-making. Scenarios of global emissions and result- Many of the decisions that will comprise ant changes in atmospheric trace-gas society’s response to climate change – concentrations and climate are a core whether mitigation, adaptation, or other re- requirement shared by many diverse sponses – involve high stakes, deep uncer- climate-related decisions. Although cli- tainties, and long time horizons. Scenarios mate-change decision-makers and their can help inform these decisions by structur- particular information needs are highly di- ing present knowledge and uncertainty, verse, many will need scenarios of global prompting critical examination of present emissions and resultant climate change, assumptions and practices, stimulating new and many more will need information that insights, identifying key pitfalls and oppor- depends on these. Consistent scenarios of tunities, or providing a framework for the as- global emissions and climate change, pro- sessment of particular decisions. For some vided centrally at the national or interna- decisions, which involve irreversible near- tional level, can serve these diverse needs – term commitment to choices whose conse- if they are presented with enough trans- quences extend over a horizon involving parency and documentation of their under- substantial uncertainties, some form of sce- lying reasoning and assumptions. nario-based reasoning may be essential. Beyond global emissions and resultant cli- There is a big gap between the use of sce- mate change, decision-makers’ needs narios in current practice and their po- from scenarios are highly diverse. Differ- tential contributions. Despite their evident ent climate-change decision-makers will value and capability, many climate-related have highly variable needs from scenarios, decisions that could benefit from scenarios in the factors and variables included, the (e.g., many decisions regarding long-term time and spatial scale at which they are pro- management and investments in climate- vided, and the nature of uncertainties rep- sensitive areas such as freshwater systems resented. The means for meeting these or coastal zones) are not using them. In- additional needs will likely be diverse, too. deed, many such decisions are still being Some will call for separate, specialized sce- made without considering climate change nario production capabilities. A major dis- at all. Conversely, many climate-change tinction in scenario-related needs can be scenarios have only weak and indirect con- drawn between impacts and adaptation nections to practical decisions related to managers, mitigation policy-makers, and climate-change mitigation or adaptation. energy resource and technology managers.

Interest in considering and using climate- Impacts and adaptation managers need change scenarios is sharply increasing. scenarios that project impacts relevant to There is increasing interest in considering their specific responsibilities, and the climate-change scenarios in diverse deci- major determinants of vulnerability and sion and planning processes. This trend is adaptive capacity. Impacts and adaptation strongest for planning and decisions con- managers include both national officials cerned with climate-change impacts and and others responsible for more specific adaptation. The trend reflects advances in domains of impact. These decision-mak- scientific understanding of climate change, ers need climate-change scenarios, driven gradual maturation of models and analytic by specified global emissions scenarios, to tools, and increasing recognition by deci- provide information about potential cli- sion-makers of the potential importance of mate-related stresses on their areas of re- climate change. Given the high general con- sponsibility. In addition, they need other cern about climate change and the advance environmental and socio-economic infor-

84 Global-Change Scenarios - Their Development and Use mation specific to their areas of responsi- when they share mitigation responsibilities bility, at appropriate spatial and temporal or undertake mitigation initiatives. Serious scales. Meeting these needs will require pursuit of greenhouse-gas mitigation will both easy access to centrally produced cli- require major policy innovations that carry mate scenario information with associated significant risks of many kinds, including tools and support, and development of de- the effectiveness and cost of the policies centralized capabilities for developing and but also their effects on government budg- applying additional scenario-related infor- ets, competitiveness of particular industries, mation. Many of these specific informa- opportunities for national technological cation needs are likely to be similar in pabilities, etc. Decision-makers consider- character for many particular locations and ing such policies will need scenarios of types of impact. global and national emissions trends, resultant climate change, and aggregate im- Meeting information needs for impacts pacts. In addition, they will need to and adaptation requires a cross-scale or- consider many factors specific to their ju- ganizational structure. These decisions’ risdictions – e.g., national policies, institu- combination of centralized and decentral- tions, economic structure, technological ized information needs suggests the need capabilities, and the detailed structure of na- for a linked network of institutions at national emissions – and information about the tional and sub-national levels to develop policy environment for their choices, in- scenarios. Such a structure would combine cluding alternative scenarios of other na- central provision of globally consistent tions’ mitigation strategies, international emissions and climate change scenarios; mitigation decisions, and implementation decentralized elaboration of these scenar- and compliance. ios with additional variables required for regional impacts and adaptation analyses; Scenarios for mitigation decisions should and provision of tools and resources to sup- include a wide range of baseline emis- port development and use of scenarios. sions assumptions and should not pre- judge the likely level of mitigation effort. Scenarios for impact and adaptation man- Scenarios used to inform mitigation deci- agers should be based on emissions as- sions should consider the full range of po- sumptions that include a likely range of tential mitigation choices on the agenda, mitigation interventions, now and in the defined relative to baseline assumptions future. The emissions assumptions under- that, as much as possible, reflect only ef- lying scenarios for impacts managers forts already enacted or committed, includ- should be based on the likely range of fu- ing a range of reasonable assumptions ture global emissions trajectories, includ- about implementation and compliance. ing explicit assumptions about what This assumption typically implies a wider degrees of further mitigation effort are range of emissions futures than is consid- likely over time. This will typically imply ered in scenarios to support impacts and a narrower range of emission futures than is adaptation decisions. considered in scenarios to support mitigation decisions. Mitigation decision-makers can use target-driven scenarios for backcasting. Mit- Mitigation policy-makers need scenarios igation decision-making may also benefit that project alternative emissions trends in from scenarios that impose explicit future their own jurisdiction and others, and the environmental targets such as limits on major factors that will influence mitiga- emissions or atmospheric concentrations, tion opportunities, constraints, and costs. together with assumptions about policy and Mitigation policy-makers are usually offi- implementation elsewhere, and reason cials who make national policy and partic- backwards to explore alternative paths to, ipate in international negotiations, but this and implications and requirements of attain- group also includes sub-national officials ing that goal, including feasibility, costs, and

85 The U.S. Climate Change Science Program Section 5 - Conclusions and Recommendations

tradeoffs. These must be defined in ways rel- based on new knowledge, experience, and evant to the level of decision-making being priorities, as well as further developments informed, i.e., alternative national targets in scenario-related methods. Such updates to inform national policy-making, in the are needed much more frequently than the broader context of alternative global base- decision time horizons. lines or global targets. 5.2 USE OF SCENARIOS Informing mitigation decisions requires IN CLIMATE-CHANGE capacity for scenario development at the ASSESSMENTS national level. While core scenarios of global emissions and climate-change can Large-scale, official assessments are the provide some of the required input into mit- major use for scenarios at present and are igation decisions, these decisions require likely to remain an important use. Large- additional information that must be pro- scale, official assessments represent the vided at the national or sub-national level most prominent demand for climate-related where the decisions are being considered, scenarios at present, and are likely to re- generated in consultation with relevant de- main major users, particularly for coordi- cision-makers or their surrogates. nated scenarios of global emissions and resultant climate-change. Energy resource and technology managers need scenarios that represent the Within assessments, scenarios mainly political and economic environment for serve to support further analysis, model- energy investments, including mitigation ing, and assessment. When scenarios are policies. Energy resource and technology used in assessments, certain users are managers concerned with private responses clearly identified: e.g., climate modelers to mitigation policy primarily need scenar- are major users of emissions scenarios, ios that represent alternative policy while impacts assessors and modelers are regimes. Emissions and climate change major users of climate-change scenarios. underlie these as influences on policy de- These users have specific scenario needs, cisions, but do not capture the most impor- and close consultation is possible between tant uncertainties for these decision- scenario producers and users to meet these makers. While many actors may wish to needs. Substantial progress has been made generate these scenarios privately to keep in providing useful scenarios for these their assumptions and analyses confiden- groups, at both the national and interna- tial, there may also be value in multi-party tional level. These efforts should be con- collaborative scenario-building exercises in tinued and expanded. which today’s policy-makers and corporate planners jointly examine what range of pol- The presentation of scenarios in assess- icy, economic, and energy regimes is plau- ments leads to many additional, unforeseen sible or likely over the 30- to 50-year time uses. Scenarios presented in large-scale as- horizons relevant for investment and tech- sessments gain prominent dissemination nology-development decisions. that results in their being put to many uses their developers did not foresee. Scenarios Scenarios must be periodically revised and should pursue clarity of documentation and updated. For all types of decisions and de- transparency about underlying reasoning cision-makers, developing scenarios, ap- and assumptions, to improve the ease of use plying them to inform decisions, and and reduce the risk of misunderstanding in refining scenario methods, are iterative such derivative uses, although they cannot processes. Limitations to present scenar- anticipate all information needs of an open- ios or methods do not in general justify de- ended set of diverse potential uses. laying consideration of such decisions, any more than scientific uncertainties do. Still, In assessments, scenarios can strongly in- scenarios must be periodically updated, fluence issue framing. Also because of

86 Global-Change Scenarios - Their Development and Use

their prominent dissemination, scenarios information about trends in world emis- presented in major assessments can exer- sions, underlying socio-economic condi- cise substantial influence over the framing tions at the scale of major world regions, of policy discussions or provide simple, and the large-scale pattern of global policy widely used metrics of the seriousness of response. They can also provide access to the issue. They may consequently exercise climate-model scenario output, plus tools, broad influence over many decisions that data, and support for producing finer-scale depend upon such aggregate perceptions of scenario information needed for particular seriousness. The prospect of such influ- impact and adaptation applications. Miti- ence further heightens the responsibility for gation and adaptation decisions and asso- transparency in production of scenarios. ciated assessments at national or smaller spatial scale will need more detailed and Scenarios contain unavoidable elements finer-scale climate and socio-economic in- of judgment in both their production and formation than can be provided by central- use. Although they draw on relevant data, ized scenarios, so these must be extended knowledge, and analysis, scenarios in- and/or modified by national and sub-na- evitably contain elements of judgment. In tional scenario processes. addition to putting serious responsibilities onto scenario developers, this implies that Scenarios of emissions and resultant cli- there is no authoritative way to resolve ar- mate change should be global in scope guments over whether a scenario is plausi- and century-scale in time horizon. Core ble or not. When a wide enough range of emissions and climate-change scenarios potential futures is considered, some sce- should specify major climate-relevant narios are likely to draw criticism, in part emissions and other perturbations, globally motivated by opposition to their foresee- and for major world regions. They should able implications for action. Any scenario extend over a time horizon of at least 100 can be attacked as unreasonable, specula- years (including some that extend 200-300 tive or unlikely, and close enough scrutiny years to support assessments of sea level of any scenario can usually reveal incon- rise), with interim results at roughly sistencies, but these do not provide suffi- decadal resolution. cient basis for excluding a scenario from consideration. Indeed, scenarios designed Emissions scenarios of several distinct to represent extreme events, or to lie near logical types will be needed to serve di- one end of a distribution of potential out- verse purposes. These will include some comes, should by definition appear un- combination of alternative baselines, alterna- likely. The most productive response to tive levels of incremental stringency of miti- such criticisms lies in transparency about gation effort, and specified future targets to the process, reasoning, and assumptions support backcasting and feasibility analysis. used to produce scenarios. Such transparency can shift arguments to underlying For some uses, emissions scenarios should uncertainties, and help limit biases in the be coupled to explicit scenarios of alter- production of scenarios. native socio-economic futures. For these scenarios, the range of potential socio-eco- 5.3 CHARACTERISTICS nomic and policy futures considered should OF “CORE” EMISSIONS AND be wider than has been considered to date, CLIMATE SCENARIOS including scenarios of policy failure and conflict, and a wide range of stringency and Centrally provided scenarios of emissions timing of mitigation effort. For example, and climate change can help inform miti- what if development stagnates in major gation and adaptation decisions at na- world regions? What if world emissions tional and sub-national scale, but these grow sharply for several decades with lit- will usually require additional informa- tle control effort, followed by a subsequent tion as well. Central scenarios can provide shift to stringent mitigation efforts? What

87 The U.S. Climate Change Science Program Section 5 - Conclusions and Recommendations

if part of the world makes a lot of effort and been made thus far. These efforts should part makes very little? Considering such seek to connect alternative qualitative and varied future histories is crucial for consid- narrative scenarios not just to alternative ering long-term risks and opportunities parameter values in quantitative models, from major mitigation choices. but also to alternative forms of causal relations and model structures. Generating Scenarios should reflect various explicit multiple alternative model quantifications degrees of coordination, depending on based on the same narrative and associated their intended uses. Some uses will re- causal logic is one promising route to en- quire groups of simple coordinated scenar- riching understanding of uncertainties in ios to provide standardized inputs for key quantitative variables such as future downstream modeling and analysis – e.g., economic output and emissions. standard emissions scenarios as inputs to climate models and standard climate sce- 5.4 SCENARIO PROCESS: narios as inputs to impact assessments – for DEVELOPER-USER exploring present uncertainties and track- INTERACTIONS ing developments of knowledge over time. Other scenarios should be based on multiple There is value in collaboration between models using common input assumptions. scenario developers and users, particu- Non-standardized scenarios produced at the larly at the beginning and ending stages initiative of researchers and modelers should of a scenario exercise. The appropriate de- also be produced, which explore alternative gree and means of this collaboration vary assumptions or meet specific user needs, substantially among scenario exercises. provided these meet basic standards of User engagement is most important in the quality control, transparency, and docu- initial scoping and design of a scenario ex- mentation. ercise, and in the evaluation and application of the scenarios generated. The value Some scenarios should seek to link quali- of user engagement in details of scenario tative and quantitative elements. Some development, quantification, elaboration, scenarios of socio-economic conditions, and checking, depends on the specific case. whether produced to support global emissions scenarios or impacts assessments, The ease of achieving such collaboration should include both qualitative and quanti- and its value are likely to be greater when tative elements and sustained analytic ef- scenario users are clearly identified, few forts to link the two. Qualitative or narrative in number, and similar in their interests scenario elements can provide a vehicle to and perspectives. When potential scenario explore major historical uncertainties with users are identified, relatively few, and rel- large implications for global emissions, cli- atively homogenous, close and intensive mate change, and vulnerability to climate collaboration between users and develop- impacts; provide a coherent rationale and ers is likely to be most productive. When logical structure to connect assumed tra- potential users are numerous and diverse, jectories for multiple variables, including intensive engagement may be infeasible both quantitative and qualitative ones; and and more structured processes for consul- provide guidance to other analysts or users tation, representation, and information ex- who may wish to extend the scenarios by change are needed. While progress has elaborating additional detail. Achieving been made in new methods to allow larger these benefits will require more sustained numbers to participate in scenario exer- effort to integrate model-based projections cises, further development of such methods of quantitative variables with qualitative is needed. and narrative scenario elements, to iterate between these, and to critically examine each element in light of the other, than has

88 Global-Change Scenarios - Their Development and Use

5.5 COMMUNICATION consistency within individual scenarios OF SCENARIOS can be an intensive and time-consuming process, but is crucial to avoid problems Effective communication of scenarios is that can discredit a scenario exercise. essential, in forms useful to audiences of diverse interests and technical skills. Sce- In scenario exercises that use multiple narios must be communicated effectively to models to explore potential uncertainties their potential users, including both techni- in future conditions, consistency between cal and non-technical audiences. In addi- models should be pursued primarily tion to the contents or outputs of scenarios, through coordination of inputs, not out- communication should include associated puts. Use of multiple models in parallel to documentation, tools, and support for their produce alternative descriptions of future use. Various methods should be used to conditions can improve understanding of promote broad dissemination of scenario uncertainties, if models are run under information; for instance, presentations, re- consistent assumptions about exogenous ports, websites, and centralized data distri- inputs. Forcing models to generate consis- bution centers. To facilitate user tent trajectories for endogenous outputs understanding of results, various methods poses several risks, including suppressing should be used to communicate numerical variation from alternative causal structures and technical information, including multi- that could provide valuable insights into ple tabular, summary, and graphical formats, uncertainties, and encouraging over-confi- ideally with user-interactive capabilities. dence from spurious precision. For quantities that are exogenous in some models and Transparency of underlying reasoning endogenous in others, the appropriate treat- and assumptions is crucial. Scenario comment varies case by case, but it is not gen- munication should include transparent dis- erally desirable to force multiple models to closure of underlying assumptions, models, convergent values of such variables with- and reasoning used to produce the scenar- out more detailed examination of the un- ios, to support the credibility of scenarios, derlying uncertainties. to alert potential users to conditions under which they might wish to use or modify Imposing consistent outputs in multi- them, and to inform criticism and improve- model exercises can be useful, however, ment of scenarios. This should include ex- when these outputs represent common plicit identification of the major goals for policy evaluation. For example, uncertainties represented in each scenario consistent constraints on some environ- and the sources of underlying information, mentally relevant target such as emissions, whether drawn from the scientific litera- atmospheric concentrations, or radiative ture, formal expert-elicitation exercises, or forcings, can be used to examine inter- informal judgments of the scenario team. model uncertainties in the technological, It is possible in virtually all cases to for- economic, and resource conditions associ- mulate simple, accessible, honest descrip- ated with meeting the specified targets. tions of why a scenario was undertaken, why it was necessary, what was done, how Transparency in reporting scenario and and why, and why it merits respect as a rea- model differences as well as underlying sonable judgment. assumptions and reasoning can help mitigate the effects of inconsistencies among 5.6 CONSISTENCY AND scenarios. Ideally, multiple scenarios in an INTEGRATION IN SCENARIOS exercise should differ only on those elements intentionally chosen to distinguish Any scenario should be internally consis- them, and be consistent in all other factors. tent in its assumptions and reasoning, to However, this is not always possible, par- the extent this can be established given ticularly when scenarios are produced present knowledge. Carefully pursuing using different models. Pursuing maximal

89 The U.S. Climate Change Science Program Section 5 - Conclusions and Recommendations

transparency about the models, assump- global or regional climate, or of prominent tions, and reasoning underlying each quantitative impact measures, such as scenario – perhaps by issuing detailed di- changes in river flows or sea level, al- agnostic reports that include explicit dis- though they neither capture all relevant cussion of points of weakness, uncertainty, uncertainty nor avoid the inevitably sub- and disagreements, and the means used to jective nature of such probability judg- resolve them – can mitigate any resultant ments. Explicit probability judgments are confusion. more difficult and controversial when they involve socio-economic factors for which 5.7 TREATMENT OF quantitative models are not available, and UNCERTAINTY IN SCENARIOS that do not depend in well understood ways on identified quantitative parameters. Such Some scenario exercises should include factors include major technological inno- more explicit characterization of likeli- vations, large-scale changes in attitudes or hood judgments than has been practiced norms, or policy response. so far. The advantages of being more explicit about the probability judgments that Attempting to include explicit probability underlie scenario exercises are likely to judgments is likely to be most useful and outweigh the disadvantages. Such specifi- successful when key variables are few, cation should be pursued further than has quantitative outcomes are needed, and po- been done in major global-change scenario tential users are numerous and diverse. exercises to date, although not necessarily The case for assigning explicit likelihood in all scenario exercises. The means avail- judgments is strongest when scenarios’ able to express these judgments are of most salient components are quantitative widely varying specificity, ranging from projections of a few key variables, such as agreed terminology202 to explicitly quanti- emissions or average temperature change fied probability distributions. All such over the globe or some region, because the judgments should include explicit ac- technical barriers to assigning probabilities knowledgement of their inevitably subjec- are least severe in this case. The case is tive elements and appropriate caveats to strongest when a primary purpose of the help users avoid mistaking them as objec- scenario exercise is to provide inputs to tively true. other quantitative assessment activities, or to inform decisions that primarily depend Explicit probability judgments are easiest on one or a few key quantitative variables, to produce and least controversial in sce- because such uses are most likely to require narios generated using quantitative mod- probability judgments. The case is els of climate change or specific impact strongest when the set of potential scenario domains. Scenarios generated using such users and uses is large and heterogeneous, models can be conditioned on specific as- because this situation provides the least op- sumed values for socio-economic inputs portunity for informal communication of such as emissions, and can represent ex- implicit judgments of likelihood or priority plicitly and quantitatively the effects of through intense, sustained collaboration be- specified variation in initial conditions or tween scenario developers and users. unknown parameter values. These devices are also available, although in less wide- Attempting to include explicit probability spread use, in economic models used to judgments is likely to be least useful and project emissions. These devices aid in successful when scenarios specify multi- constructing distributions of key quantita- ple characteristics, including prominent tive characteristics, such as measures of narrative or qualitative components; when the purpose of a scenario exercise is sensitivity analysis or heuristic exploration;

202 See, e.g., the consistent uncertainty language proposed for IPCC reports by Moss and Schneider 2000.

90 Global-Change Scenarios - Their Development and Use and when potential users are few, similar, Providing explicit probability judgments al- and known. When scenarios are primarily lows scenario users to choose whether to construed as rich, qualitative narratives that use them or not. Some users may choose present major alternative historical and to use these explicitly in their subsequent socio-economic trajectories, the technical analysis or decision support, while others obstacles to explicit probability assignment may use them only to help decide which sce- are greatest and the likely confidence in narios to use, and still others may disregard scenario developers’ subjective probability them entirely. Users may select a different judgments lowest. When the main purpose group of scenarios or a different subset of of a scenario exercise is to stimulate critical the uncertainty range for various reasons, in- or creative thought, to probe the limits of a cluding differences in risk aversion, differ- model or decision strategy through sensi- ences in the scope of their decision author- tivity analysis, or to explore ways of meet- ity, or differences in their assumptions ing a specified target, explicit probability about decisions by other actors (present or assignment provides little or no benefit. future). Presenting complete descriptions of When users are few, similar, and specifi- scenarios together with underlying as- cally identified, they can be intensively in- sumptions and reasoning, including proba- volved in scenario production, allowing bility judgments, preserves all these options effective informal communication of like- for users. lihood judgments without stating them explicitly. Under these conditions, scenario Scenario exercises should give more at- exercises can also be structured to engage tention to extreme cases. Some uses of sce- users in the potentially instructive activity narios require consideration of low-probability, of assigning and discussing their own prob- high-consequence extreme cases, such as loss ability judgments, rather than putting that of a major continental ice sheet or major responsibility exclusively on the researchers changes in meridional ocean circulation. or analysts developing scenarios. Consequently, such scenarios should be included in large, general-purpose scenario ex- Centrally provided scenarios of global ercises producing emissions or climate- emissions and climate change should at- change scenarios, together with more likely tempt to include explicit probability judg- middle-case scenarios. When extreme sce- ments. Because of the large, diverse set of narios are included in an exercise, it is espe- users for these scenarios, explicit likelihood cially critical to be explicit and transparent judgments should be provided for ranges of about the reasoning and assumptions under- key quantitative outputs, including global lying each scenario, and scenario developers’ emissions and global-average temperature judgments of relative likelihoods. change. Scenarios should typically include several paths that span a wide range of In addition to enhancing the utility of sce- judged uncertainty, e.g., 95 percent to 99 nario outputs, probabilistic methods can percent. The associated probability judg- contribute throughout the scenario devel- ments may include some that are uncondi- opment process. Developing scenarios re- tional and some that are conditioned on quires making many judgments about un- specific assumed prior conditions. Such es- known characteristics and developing many timates should be provided by multiple arguments and pathways to link these. Sce- groups using diverse methods. As for all narios based on quantitative models typically such probability judgments, their unavoid- require specifying many exogenous inputs ably subjective nature and the specific as- and parameters. Even narrative scenarios re- sumptions on which they are conditioned quire specifying values of multiple charac- should be stated explicitly and prominently teristics, both qualitative and quantitative. Explicit discussion of uncertainties and as- . sociated probabilities can help structure

91 The U.S. Climate Change Science Program Section 5 - Conclusions and Recommendations

and facilitate many aspects of the scenario To help fulfill these presently unmet development process, including deciding ap- needs, CCSP should establish a propriate ranges of variables to consider, program to: defining boundaries of what outcomes are • Commission scenarios for use in as- considered plausible, elaborating associat- sessments and decision-support ed causal mechanisms and linkages, dis- activities. This task includes facil- cussing and integrating knowledge and itating agreement among relevant judgments from multiple participants, and producers and users on standard sce- clarifying disagreements. Explicit conver- narios in cases where multiple as- sation about probabilities can support in- sessment activities require stan- sights throughout these processes, in addi- dard scenarios for comparability, and tion to supporting communication of scenario convening and supporting a di- judgments to users. verse collection of more extensive and detailed scenario-related exer- 5.8 EXPANDING AND SUSTAIN- cises, by multiple groups using a ING CAPACITY FOR PRODUC- wide range of models and ap- TION AND USE OF SCENARIOS203 proaches.

Present scenario capacity is inadequate. Al- • Disseminate scenarios with associated though scenario-based reasoning is required documentation, tools, and guidance for many aspects of global change assessment materials. This task includes mul- and decision support, the present capacity to tiple forms of support and pro- produce, disseminate, apply, evaluate, and gram-building for diverse groups adapt scenarios is inadequate. There has not seeking to apply, modify, and extend been enough continuity to enable effective existing scenario-based informa- learning, because scenarios are typically tion at various regional and sectoral produced de novo for each major application. scales, through providing data, There has not been enough transparency about models, tools, resources, and asso- methods, reasoning, and assumptions. Con- ciated documentation and technical structing integrated scenarios and exploring support, in multiple forms and alternative methods has been difficult, in part through multiple media. because scenario exercises have tended to be • Commission various groups to eval- dominated by use of quantitative models, sep- uate scenarios and their applica- arated along disciplinary lines. Inadequate re- tions and to develop improved sources have been devoted to methods de- methods. This task includes defin- velopment, for scenarios and related decision- ing and promulgating standards for support tools Finally, there has been no sys- quality control – which, given the tematic evaluation and critique of scenarios need for diverse approaches, would or their application. principally concern matters of process such as transparency, documentation, evaluation, and dissemination of results and supporting information. The task also includes broad efforts to develop scenario-related skills, tools, and methods, e.g., by providing resources for methods development and evaluation projects; conducting and es- tablishing procedures for evaluation 203 Recommendations made in this report regarding pro- of particular activities and pro- grammatic and organizational changes, and the ade- grams; and convening workshops, quacy of current budgets, reflect the judgment of the report’s authors and the CPDAC and are not necessarily the views of the U.S. Government.

92 Global-Change Scenarios - Their Development and Use

conferences, etc., to evaluate lysts who are developing and using progress overall, or in particular scenarios and related methods. Al- methodological areas. though established as a US national program, it must also support, col- • Archive results and documentation re- laborate, and coordinate with par- lated to all these tasks, to provide allel activities in other nations and historical perspective and institu- internationally, and with relevant tional memory for future scenario- sub-national activities. related activities. This task includes preserving for retrospective • The program should integrate and bal- scrutiny a wide range of materials: ance goals and criteria related to not just the methods, contents, and scientific and technical quality, results of scenario exercises, but also and those related to utility and rel- the progressive evaluations of par- evance to users. This balance is ticular activities and approaches, and needed for the program to support of the entire program. In its ongo- promising but speculative activities, ing development and evaluation of to encourage creativity and diversity methods, the program should not of approach, to avoid being captured draw rigid boundaries between sce- by any particular discipline or mod- nario development and application eling approach, and to be willing to and other related methods of as- make and explain judgments about sessment and decision support for quality and promise that reflect long-term global change issues. both scientific and practical considerations. To achieve this, the pro- Several conditions in the design and gram needs broad discretion over the management of this new program type of projects supported, includ- would be required to ensure its ef- ing sponsoring fairly sharply tar- fectiveness. geted activities, supporting speculative activities, and investing to de- • The program should build and main- velop and assess capabilities that do tain strong connections with out- not yet exist. side relevant expertise, and analytic and modeling capability. While • The program should be insulated from the program must develop enough political control. For the scenarios internal expertise to be a full par- and analyses based on them to be ticipant in debates over scenarios perceived as credible by their diverse and assessment methods, it cannot users, the program needs enough in- and should not attempt to impose a sulation from political control, at unilateral vision of preferred sce- both the national or international narios, methods, or approaches. level, to prevent scenarios from Rather, it must build and maintain becoming proxies for conflict over close collegial connections with near-term policies, and to allow outside networks of researchers exploration of the implications of al- and analysts in multiple fields of ex- ternative futures that represent plau- pertise, including emissions mod- sible risks but that some political ac- elers, climate scientists and mod- tors would find objectionable. elers, impacts researchers, and resource managers. These relation- • The program should strive for max- ships would be facilitated by es- imum transparency in its own ac- tablishing governing mechanisms, tivities, in addition to demanding such as a senior advisory board, it from activities it supports. The drawn from the broad communities program should strive for maximal of researchers, modelers, and ana- transparency regarding inputs, mod-

93 The U.S. Climate Change Science Program Section 5 - Conclusions and Recommendations

els, assumptions, and reasoning cess to particular dissemination employed in developing scenarios, outlets, and access to resources. A as well as any significant disagree- weak “clearinghouse” that solicits, ments that arose and how they supports, or publicizes scenarios but were resolved and any remaining cannot exercise quality control, weaknesses recognized by the de- propose and stimulate new direc- velopers. The broader and more ditions, or convene critical reviews, of verse the collection of intended the whole enterprise and of partic- uses and users, the more crucial is ular exercises, is not an adequate transparency of the scenario-pro- model for the program. duction process – because different users may require scenarios pro- • The program will require an adequate duced using different underlying as- sustained resources. The program sumptions, and they must be able to must build and maintain a sophis- track the underlying logic to exer- ticated analytic capability, and de- cise this choice. This would enhance velop skills and institutional mem- credibility in the scenario-develop- ory regarding prior experiences, ment process. While calls for such successes, and failures. This re- transparency are widely made, ex- quirement precludes the program perience suggests it is difficult to being a series of ad hoc one-time ac- achieve, particularly for such mat- tivities or a part-time, unfunded bur- ters as disagreements or recognized den imposed on people and organ- weaknesses that may risk profes- izations with other full-time re- sional embarrassment. Still, achiev- sponsibilities. ing more transparency and more widely informed debate on such matters is essential for advancing scenario methods.

• The program will require the authority and resources necessary to articulate and promulgate standards for transparency, consistency (e.g., of units and formats), and quality control. This task involves facilitating discussions among the community and formulating persuasive guidelines and supporting arguments. It also requires use of incentives such as seals of approval, access to participation in particular processes, ac-

94 ACRONYMS A1, A2, B1, B2, A1FI LRTAP IS92 IPCC IIASA HadCM2 GtC GFDL GDP GCM GBN EPA ENSO EMF DWR DOD DDC CFCs CCTP CCSP names ofspecif Transboundary Air PollutionTransboundary Air Convention onLong-Range scenarios namedIS92athrough IS92f the IPCCin1992,includingspecific series ofemissionsscenariosproducedby Climate Change Inter-governmental Panel on Laxenburg,Systems Analysis, Austria Institutefor International Applied Version 2 UK Hadley Centreclimatemodel, gigatonnes (billionmetrictons)ofcarbon laboratory climate modelproducedby this pheric Administration (NOAA), and US NationalOceanographic and Atmos Geophysical FluidDynamicsLaboratory, domesticproduct gross general circulationmodel Global BusinessNetwork Environmental Protection Agency Oscillation o/Southern El Ni Energy ModelingForum (California) of Department Water Resources ofDefense(U.S.)Department Center(IPCC) Data Distribution chlorofluorocarbons Climate Change Technology Program Climate ChangeScienceProgram Emissions Scenarios published on intheIPCCSpecialReport ic emissionsscenarios Global-Change Scenarios-TheirDe VOC VEMAP UKCIP TGICA SRES PPP PDO OTA ONA OECD NYCHP NYCDEP NRC NAST NAPAP MER MEC MEA olatile organic compounds Analysis Project Vegetation Ecosystem Mappingand Program United KingdomClimateImpacts Analysis (IPCC) forImpactandClimate Support Task GrouponDataandScenario (IPCC) onEmissionsScenarios Special Report purchasing-power parity DecadalOscillation Pacific United StatesCongress of Office Technology Assessment, ofDefense) Secretary ofNet ofUS Office Assessments (Office and Development Organization forEconomicCo-operation New York ClimateandHealthProject Environmental Protection New York of CityDepartment National ResearchCouncil(U.S.) TeamSynthesis National Assessment Assessment Program Precipitation National Acid market exchange rates US National Assessment Metropolitan EastCoastassessmentof Millennium Ecosystem Assessment velopment andUse 95 The U.S. Climate Change Science Program Acronyms

96 REFERENCES Alcamo, J., A. Bouwman,J. Edmonds, A. Gruebler, T. Morita,and Barrow Barney Barras, J., S.Beville, D. Britsch,S.Hartley, S.Hawes, J. Johnston, N.W.,Arnell, M.J.L. Livermore, S.Kovats, P.E. Levy, R.Nicholls, Alley, R.B., 2004: climatechange. Abrupt Alley Allen Berkhout Berkhout sions Scenarios.” In: A. Sugandhy, 1995: An Evaluation oftheIPCCIS92Emis- cil onEnvironmental Quality, Washington, DC,USA. 03-334,USGeologicalReport Survey, Washington, DC,USA. jected CoastalLouisianaLandChanges: 1978-2050 Roy, S.Sapkota, andJ. Suhayda, 2003: P. Kemp, Q. Kinler, J.A. Martucci, Porthouse, D. Reed, K. Environmental Change sessments: characterizingtheSRESstorylines. nomic scenariosforglobal-scaleclimatechangeimpactas- M.L. Parry, 2004:Climateandsocio-eco- andS.R.Gaffin, November 2004,62-69. nent, widespreadevent 8,200years ago. and P.U. Clark,1997:Holoceneclimateinstability:apromi- the IPCC’s Third Assessment Report. (eds.)], CambridgeUniversity Press,Cambridge,UK. H. Lee,B.A. Callander, andK.Maskell E.Haites,N. Harris, Emission Scenarios ing ofClimateChange andanEvaluationoftheIPCCIS92 ability inCanada: Past, Present andFuture change scenariosforCanada.In: Revise Environmental Futuresstudy, SPRU, Brighton,UK. Scenario andUserGuidance.Final oftheReview Report and 165-168. mate impactassessment.GlobalEnvironmental Change Canada, Ottawa. , R.B., T. Sowers, P.A. Mayewski, M.Stuiver, K.C. Taylor, , M.,S.Raper, in etal.,2001:ClimateChange-Uncertainty , G.O., 1981: , E.,P. Gachon,andD. Caya, 2004:Creatingclimate , F. and J. 2001:Foresight Hertin, Futures2001:Revised , F. andJ. 2000:Socio-economicfuturesforcli- Hertin, Global 2000ReporttothePresident. Coun- [Houghton, J.T., L.G.MeiraFilho, J. Bruce, Climate Change 1994:RadiativeForc- 14, 3-20. Climate Change and Vari- Science Scientific American Scientific Historical andPro- Geology 293, 5529,430. . Environment 26, 6. pnFile . Open Global 10, , Global-Change Scenarios-TheirDe B Brewer Bretherton Bracken, P., 1990:Gaminginhierarchicaldefenseorganizations. Bracken, P., 1977:Unintendedconsequencesofstrategic gam- Boer, G.J., G.M.Flato,andD. Ramsden,1999b: A transientcli- Boer, G.J., G.M.Flato,M.C.Reader, andD. Ramsden,1999a: A Berkhout Berkhout Brewer rewer Press, Cambridge,UK. Munn (eds.)].CambridgeUniversity Press,Cambridge,UK. Sustainable Development oftheBiosphere [Clark, W.C. andR.E. Jenkins, andJ.J. Ephraums(eds.)].CambridgeUniversity Change: The IPCCScientific Assessment climatechange.In: dent greenhouse-gas-induced (eds.)]. Brassey’s, London. agement In: ing. SimulationandGames century. ClimateDynamics16,427-450. house gasandaerosolforcing:projectedclimateforthe21st mate changesimulationwithhistoricalandprojectedgreen- century. recordforthe20th sign andcomparisonwiththeinstrumental gasandaerosolforcing:experimental de- jected greenhouse transient climatechangesimulationwithhistoricalandpro- ing machines. tures inclimateimpactassessment:usingscenariosaslearn- resources/publications/pub_dets.asp?ID=34. sessment: A Guideto Their UseintheUK UKCIP, economic scenariosforclimateimpactassessment.In: Hulme, andJ. Skea, 2001:PresentationoftheUKCIPsocio- O’Riordan, D. Cobb,L.Ledoux,R. Tinch, J. Palutikof, M. don. A.C., D. Van Opstal,andJ.H. Barkley (eds.)].Brassey’s, Lon- Brink: andPracticeinCrisisManagement [Goldberg,Theory Cambridge, MA. Problem-Solving.of Military Harvard University Press, , G.D., 1986:Methodsfor synthesis:policy exercises. In: , G.D., 1990:Discovery isnotprediction. In: Avoiding the , G.D. andM.Shubik,1979: The War Game:aCritique Avoiding theBrink: Theory andPractice inCrisisMan- , F., J. and Hertin, A. Jordan,2002:Socio-economicfu- , F., J. I.Lorenzoni, Hertin, A. Jordan,K. Turner, T. , F.P., J.D. K.Bryan, Woods, etal.,1990: Time-depen- Climate Dynamics Socio-Economic Scenariosfor ClimateImpact As- [Goldberg, A.C., D. Van Opstal,andJ.H. Barkley Global Environmental Change 16, 405-426. 8 , 283-318(3September). [Houghton, J.T., G.J. , www.ukcip.org.uk/ velopment andUse 12, 83-95. Climate 97 The U.S. Climate Change Science Program References

Burkett, V.R., D.A. Wilcox, R. Stottlemeyer, W. Barrow, D. Fagre, J. Dickson, B., I. Yashayaev, J. Meincke, B. Turrell, S. Dye, and J. Hoffort. Baron, J. Price, J. Nielsen, C.D. Allen, D.L. Peterson, G. Ruggerone, 2002. Rapid freshening of the deep North Atlantic Ocean over the and T. Doyle, 2005: Nonlinear dynamics in ecosystem response to past four decades. Nature 416, 832. climate change: case studies and policy implications. Ecological Complexity 2, 357-394. Doyle, T.W., R.H. Day, and J.M. Biagas, 2003: Predicting coastal retreat in the Florida Big Bend Region of the Gulf Coast under climate California DWR, (Dept of Water Resources), 2005: California Water change induced sea-level rise. In: Integrated Assessment of the Cli- Plan Update 2005. Bulletin 160-05. Available at http://www.wa- mate Change Impacts on the Gulf Coast Region - Foundation Doc- terplan.water.ca.gov. ument [Ning, A.H., R.E. Turner, T. Doyle, and K. Abdollahi (eds.)]. LSU Press, Baton Rouge, USA. Castles, I. and D. Henderson, 2003a: The IPCC emissions scenarios: an economic-statistical critique. Energy and Environment 14(2-3), 159-185. Duncan, N. and P. Wack, 1994: Scenarios designed to improve decision making. Planning Review 22(4), 18-25. Castles, I. and D. Henderson, 2003b: Economics, emissions scenarios, and the work of the IPCC. Energy and Environment 14(4), 415-435. The Economist, 2003a: Hot potato. The Intergovernmental Panel on Cli- mate Change had better check its calculations. February 13. CCSP, 2007: Scenarios of Greenhouse Gas Emissions and Atmospheric Concentrations. A report by the U.S. Climate Change Science Pro- The Economist, 2003a: Hot potato revisited: a lack-of-progress report on gram and the subcommittee on Global Change Science Research the Intergovernmental Panel on Climate Change. November 6. [Clarke, L., J. Edmonds, J. Jacoby, H. Pitcher, J. Reilly, and R. Richels (Eds.)]. Department of Energy, Office of Biological & En- EMF (Energy Modeling Forum), 2006: EMF 22: Greenhouse Gas vironmental Research, Washington, DC, USA, 102 pp. Stabilization. Prospectus. www.stanford.edu/group/EMF/projects/ EMF22/emf22prospectus.pdf. Chess, C. and K. Purcell, 1999: Public P and the environment: do we know what works? Environmental Science & Technology 33, 2685- Envision Sustainability Tools, 1999: Lower Fraser Basin QUEST Model 2691. Structure, Ver 1.0. Vancouver BC. www.sdri.ubc.ca/downloads/ Models.pdf Cowling, E.B., 1992: The performance and legacy of NAPAP. Ecologi- cal Applications 2(2), 111-116. Fairbanks, R.G., 1989: A 17,000-year glacio-eustatic sea level record: influence of glacial melting rates on the Younger Dryas and deep Cook, C.W., 1939: Scenery of Florida Interpreted by a Geologist. The ocean circulation. Nature 342, 7, 637-642. State Geological Survey, Tallahassee, FL, USA. Gagosian, R.B., 2003: Abrupt climate change: should we be worried? Curry, R. and C. Mauritzen, 2005: Dilution of the northern North At- Presented to World Economic Forum, Davos, Switzerland, January 27. lantic Ocean in recent decades. Science 308, 5729, 1772-1774. GBN (Global Business Network), 2004a: Executive Summary, An Abrupt Cubasch, U., G.A. Meehl, G.J. Boer, R.J. Stouffer, M. Dix, A. Noda, Climate Change Scenario and Its Implications for US National Se- C.A. Senior, S.C.B. Raper, and K.S. Yap, 2001: Projections of future curity, Oct. 2003. www.ems.org/climate/pentagon_climatechange.pdf, climate change. In: Climate Change 2001: The Scientific Basis [Houghton, J.T., G.J. Jenkins, and J.J. Ephraums (eds.)]. Cambridge GBN (Global Business Network), 2004b: Abrupt Climate Change. Press University Press, Cambridge, UK. Release, February. www.gbn.com/ArticleDisplayServlet.srv?aid=26231.

Davis, G., 2003: Meeting future energy needs: choices and possibilities. Giorgi, F., B. Hewitson, J. Christensen, M. Hulme, H. von Storch, P. The Bridge 33(2), 16-21. Whetton, R. Jones, L. Mearns, and C. Fu, 2001: Regional climate information: evaluation and projections. In: Climate Change 2001: de la Chesnaye, F. and J. Weyant (eds.), 2006: Multi-greenhouse gas mit- The Scientific Basis [Houghton, J.T., G.J. Jenkins, and J.J. Ephraums igation and climate policy. The Energy Journal, special issue. (eds.)]. Cambridge University Press, Cambridge, UK.

DeWeerd, H., 1967: Political Military Scenarios, P-3535. Rand Corpo- Godet, M. and F. Roubelat. 1996. Creating the future: the use and mis- ration, Santa Monica, CA. use of scenarios. Long Range Planning 29, 2, 164-171.

DeWeerd, H., 1975. A contextual approach to scenario construction. Simulation and Games 5, 403-414.

98 Global-Change Scenarios - Their Development and Use

Gosselink, J.G., 1984: The Ecology of Delta Marshes of Coastal IPCC (Intergovernmental Panel on Climate Change), 1990: Climate Louisiana: a Community Profile. FWS/OBS-84/09, U.S. Fish and Change: The IPCC Scientific Assessment [Houghton, J.T., G.J. Jenk- Wildlife Service, Washington, DC, USA. ins, and J.J. Ephraums (eds.)]. Cambridge University Press, Cam- bridge, UK. Gough, C., N. Castells, and S. Funtowicz, 1998: Integrated assessment: an emerging methodology for complex issues. Environmental Mod- IPCC (Intergovernmental Panel on Climate Change), 1996a: Climate eling and Assessment 3, 19-29. Change 1995: The Science of Climate Change [Houghton, J.T., L.G. Meira Filho, B.A. Calander, N. Harris, A. Kattenberg, and K. Greenberger, M., G.D. Brewer, W.W. Hogan, and M. Russell. 1983: Maskell (eds.)]. Cambridge University Press, Cambridge, UK. Caught Unawares: The Energy Decade in Retrospect. Ballinger, Cambridge, MA, USA. IPCC (Intergovernmental Panel on Climate Change), 1996b: Climate Change 1995: Economic and Social Dimensions of Climate Change Gregory, R. and T. McDaniels, 2005: Improving environmental decision [Bruce, J.P., H. Lee, E.F. Haites (eds.)]. Cambridge University Press, processes. Appendix B in: NRC 2005, Decision Making for the En- Cambridge, UK. vironment. National Academy Press, Washington DC, USA. IPCC (Intergovernmental Panel on Climate Change), 2001a: Climate Grübler, A. and N. Nakicenovic, 2001: Identifying dangers in an un- Change 2001, The Scientific Basis [Houghton, J.T., Y. Ding, D.J. certain climate. Nature 412, 5-15. Griggs, M. Noguer, P.J. van der Linden, and D. Xiaosu. Cambridge University Press, Cambridge, UK. Grübler, A., N. Naki enovi , J. Alcamo, G. Davis, J. Fenhann, B. Hare, S. Mori, B. Pepper, H. Pitcher, K. Riahi, H. Rogner, E.L. La Rovere, IPCC (Intergovernmental Panel on Climate Change), 2001b: Climate A. Sankovski, M. Schlesinger, R.P. Shukla, R. Swart, N. Victor, and Change 2001: Impacts, Adaptations, and Vulnerability [McCarthy, T.Y. Jung, 2004: Emission scenarios: a final response. Energy & En- J.J., O.F. Canziani, N.A. Leary, D.J. Dokken, and K.S. White. Cam- vironment 15, 1, 11-24. bridge University Press, Cambridge, UK.

Hansen, B., W. Turrell, and S. Østerhus, 2001: Decreasing overflow from IPCC (Intergovernmental Panel on Climate Change), 2001c: Climate the Nordic Seas into the Atlantic Ocean through the Faroe Bank Change 2001: Mitigation, [Metz, B., O. Davidson, R. Swart, and J. Channel since 1950. Nature 411, 927-930. Pan (eds.)]. Cambridge University Press, Cambridge, UK.

Hausrath, A.H., 1971: Venture Simulation in War, Business, and Poli- IPCC-TGCIA (Intergovernmental Panel on Climate Change-Task tics. McGraw-Hill, New York, USA. Group on Scenarios for Climate Impact Assessment), 1999: Guide- lines on the Use of Scenario Data for Climate Impact and Adapta- Herrick, C., 2002: Atmospheric Science and the Constitution of Public tion Assessment, Version 1, [Carter, T.R., M. Hulme, and M. Lal Policy: The Case of the National Acid Precipitation Assessment Pro- (eds.)]. gram (NAPAP). Case study prepared for the American Meteoro- logical Society, 2002 Summer Policy Colloquium. Johns, T.C., R. Carnell, J. Crossly et al., 1997: The second Hadley Cen- http://www.ametsoc.org/atmospolicy/PolicyCaseStudies.html. tre Coupled Ocean-Atmosphere GCM: model description, spinup, and validation. Climate Dynamics 13, 225-237. Holling, C.S. (ed.), 1978: Adaptive Environmental Assessment and Man- agement. Wiley, New York, USA. Jones, W.M., 1985: On Free-Form Gaming. Rand Note N-2322-RC, Rand Corporation, Santa Monica, CA, USA. Holman, I.P., P.J. Loveland, R.J. Nicholls, S. Shackley, P.M. Berry, M.D.A. Rounsevell, E. Audsley, P.A. Harrison, and R. Wood, 2002: Kahn, H. and A.J. Wiener, 1967: The Year 2000: a Framework for Spec- REGIS - Regional Climate Change Impact Response Studies in East ulation on the Next Thirty-Three Years. Macmillan, New York. Anglia and North West England. UK Climate Impacts Programme, UK. Kalkstein, L.S. and J.S. Greene, 1997: An evaluation of climate/mortal- Holtsmark, B.J. and K.H. Alfsen, 2005: PPP Correction of the IPCC ity relationships in large U.S. cities and the possible impacts of a cli- emissions scenarios: does it matter? Climatic Change 68, 1, 11-19. mate change. Environmental Health Perspectives 105, 84–93

Huss, W.R., 1988: A move toward scenario analysis. International Jour- Keith, D.W., 2000: Geoengineering the climate: history and prospect. nal of Forecasting 4, 377-388. Annual Review of Energy and Environment 25, 245-284.

99 The U.S. Climate Change Science Program References

Keith, D.W., M. Ha-Duong, and J.K. Stolaroff, 2006: Climate strategy Lomborg, B. 2001: The Skeptical Environmentalist. Cambridge Uni- with CO2 capture from the air. Climatic Change 74, 1-3. versity Press, Cambridge, UK.

Kemp-Benedict, E., C. Heaps, and P. Raskin, 2002: Global Scenario London Observer, 2004: Now the Pentagon tells Bush: climate change Group Futures: Technical Notes. Stockholm Environment Institute will destroy us. February 22:1. observer.guardian.co.uk/interna- PoleStar Series Report no. 9. tional/story/0,6903,1153513,00.html,

Kinney, P, et al., 2005: Assessing Potential Public Health and Air Qual- Lorenzoni, I., A. Jordan, M. Hulme, R.K. Turner, and T. O’Riordan, ity Impacts of Changing Climate and Land Use in Metropolitan New 2000: A co-evolutionary approach to climate change impact assess- York. Columbia Earth Institute, New York, USA. ment: part i. integrating socio-economic and climate change scenarios. Global Environmental Change 10, 1, 57-68. Kinney, P.L., J.E. Rosenthal, C. Rosenzweig, C. Hogrefe, W. Solecki, K. Knowlton, C. Small, B. Lynn, K. Civerolo, J.Y. Ku, R. Goldberg, C. MacCracken, M.C., E.J. Barron, D.R. Easterling, B.S. Felzer, and T.R. Oliveri, 2006: Assessing potential public health impacts of changing Karl, 2003: Climate change scenarios for the U.S. National Assess- climate and land uses: the New York Climate & Health Project. In: ment. Bulletin of the American Meteorological Society 84, 12, 1711- Climate Change and Variability: Impacts and Responses [Ruth, M., 1723. K. Donaghy, and P. Kirshen (eds.)]. New Horizons in Regional Sci- ence, Edward Elgar, Cheltenham, UK. MacCracken, M.C., E. Barron, D. Easterling, B. Felzer, and T. Karl, 2001: Scenarios for climate variability and change. See NAST 2001. Kittel, T.G.F., N.A. Rosenbloom, T.H. Painter, D.S. Schimel, and VEMAP Participants, 1995: the VEMAP Integrated Database for McLean, R.F., A. Tsyban, V. Burkett, J. Codignotto, D. Forbes, V. Ittekkot, Modeling United States Ecosystem/Vegetation Sensitivity to Climate N. Mimura, and R.J. Beamish, 2001: Coastal zones and Marine Change. Journal of Biogeography 22, 4-5, 857-862. ecosystems. In: Climate Change: Impacts, Adaptation, and Vulner- ability [McCarthy, J.J., O.F. Canziani, N.A. Leary, D.J. Dokken, and Leggett, J., W. Pepper, R.J. Swart, J.A. Edmonds, L.G. Meira Filho, I. K.S. White (eds.)]. Cambridge University Press, Cambridge, UK. Mintzer, M.X. Wang, and J. Watson, 1992: Emissions scenarios for the IPCC, an update: assumptions, methodology, and results. In: Cli- Manabe, S. and R.J. Stouffer, 1979: A CO2-climate sensitivity study mate Change 1992: The Supplementary Report to the IPCC Scien- with a mathematical model of the global climate. Nature 282, 491-493. tific Assessment [Houghton, J.T., B.A. Callander, and S.K. Varney (eds.)]. Cambridge University Press, Cambridge, UK. Manabe, S., R.J. Stouffer, M.J. Spelman, and K. Bryan, 1991: Transient responses of a coupled ocean-atmosphere model to gradual changes Lempert, R.J., D.G. Groves, S.W. Popper, and S.C. Bankes, 2006: A gen- of atmospheric CO2. part i: annual mean response. Journal of Cli- eral, analytic method for generating robust strategies and narrative mate 4, 8, 785-818. scenarios. Management Science 52, 4, 514-528. Manabe, S. and R.T. Wetherald, 1967: Thermal equilibrium of the at- Levine, R.A., 1964a: Crisis games for adults. In: Crisis Games 27 Years mosphere with a given distribution of relative humidity. Journal of Later: Plus C’est Déjà vu. P-7719, RAND Corporation, Santa Mon- the Atmospheric Sciences 24, 3, 241-259. ica, CA, USA. Manne, A., R. Richels, and J.A. Edmonds, 2005: Market exchange rates Levine, R.A., 1964b: Crisis games: a rejoinder to Tom Schelling and to or purchasing power parity: does the choice make any difference to some extent to Bill Jones. Crisis Games 27 Years Later: Plus C’est the climate debate? Climatic Change, 71, 1-2, 1-8. Déjà vu. P-7719, RAND Corporation, Santa Monica, CA, USA. MEA (Millennium Ecosystem Assessment), 2005: Ecosystems and Human Levy, M.A., 1995: International co-operation to combat acid rain. In: Well-Being: Synthesis Report. Island Press, Washington, DC, USA. Green Globe Yearbook of International Co-operation on Environ- ment and Development 1995 [Bergesen, H., G. Parmann, and Ø. MEA (Millennium Ecosystem Assessment), 2006: Ecosystems and Thommessen (eds.)]. Oxford University Press, Oxford, UK. Human Well-Being: Scenarios. Island Press, Washington, DC, USA.

Lofgren, B.M., F.H. Quinn, A.H. Clites, R.A. Assel, and A.J. Eberhardt, Meadows, D.H., D.L. Meadows, J. Randers, W.W. Behrens III, 1972: 2000: Water resources. In: Preparing for a Changing Climate: Great Limits to Growth: A Report to the Club of Rome. Universe Books, Lakes Overview. New York, USA.

100 Global-Change Scenarios - Their Development and Use

Mearns, L.O., C. Rosenzweig, and R. Goldberg, 1992: Effect of changes Morgan, M.G., R. Cantor, W.C. Clark, A. Fisher, H.D. Jacoby, A.C. Jane- in interannual climatic variability on CERES-wheat yields: sensitiv- tos, A.P. Kinzig, J. Melillo, R.B. Street, and T.J. Wilbanks, 2005: ity and 2 x CO2 general circulation model studies. Agricultural and Learning from the U.S. National Assessment of Climate Change. Forest Meteorology 62, 159-189. Environmental Science & Technology 39, 9023-9032.

Mearns, L.O., C. Rosenzweig, and R. Goldberg, 1996: The effect of Morton, R.A., N.A. Buster, M.D. Krohn, 2002: Subsurface controls on changes in daily and interannual climatic variability on CERES- historical subsidence rates and associated wetland loss in southcen- wheat: a sensitivity study. Climatic Change 32, 257-292. tral Louisiana. Transactions - Gulf Coast Association of Geological Societies 52, 767-778. Mearns, L.O., M. Hulme, T.R. Carter, R. Leemans, M. Lal, and P. Whet- ton, 2001: Climate scenario development. In: Climate Change 2001: Moss, R.H. and S.H. Schneider, 2000: Uncertainties in the IPCC TAR: The Scientific Basis [Houghton, J.T., Y. Ding, D.J. Griggs, M. recommendations to lead authors for more consistent assessment and Noguer, P.J. van der Linden, X. Dai, K. Maskell, and C.A. Johnson reporting. In: Cross Cutting Issues Guidance Papers, Intergovern- (eds.)]. Cambridge University Press, Cambridge, UK. mental Panel on Climate Change, Geneva, Switzerland.

Michaels, P.J., 2003a: Science or political science: an assessment of the Mote, P.W., E.A. Parson, et al., 2003: Climate and the water, forests, and US National Assessment of the potential consequences of climate salmon of the Pacific Northwest. Climatic Change 61, 1-2, 45-88. variability and change. In: Politicizing Science: the Alchemy of Pol- icymaking [Gough, M. (ed.)]. Publication 517, Hoover Institution Naki enovi , N. and R. Swart (eds.), 2000: Special Report on Emis-sions Press, Stanford, CA, USA. Scenarios. Cambridge University Press, Cambridge, UK. http://www.grida.no/climate/ipcc/emission/ Michaels, P. (ed.), 2003b: Bad math. World Climate Report 13, 10. Naki enovi , N., A. Grübler, S. Gaffin, T.T. Jung, T. Kram, T. Morita, H. Miller, S.S., 1990: NAPAP: a unique experience. Environmental Science Pitcher, K. Riahi, M. Schlesinger, P.R. Shukla, D. van Vuuren, G. and Technology 24, 12, 1781-1782. Davis, L. Michaelis, R. Swart, and N. Victor, 2003: IPCC SRES revisited: a response. Energy & Environment 14, 2-3, 187-214. Mitchell, J.F.B., S. Manabe, V. Meleshko, and V. Tokioka, 1990: Equi- librium climate change and its implications for the future. In: Cli- NAPAP (National Acid Precipitation Assessment Program, 1982: An- mate Change: The IPCC Scientific Assessment [Houghton, J.T., G.J. nual Report: National Acid Precipitation Assessment Program. Na- Jenkins, and J.J. Ephraums (eds.)]. Cambridge University Press, tional Acid Precipitation Assessment Program, Washington, DC, USA. Cambridge, UK. Nash, J.F., 1950: Equilibrium points in n-person games. Proceedings of Mitchell, R.B., W.C. Clark, D.W. Cash, and N.M. Dickson (eds.), 2006: the National Academy of Sciences 36 (1), 48-49. Global Environmental Assessments: Information and Influence. MIT Press, Cambridge, MA, USA. NAST (National Assessment Synthesis Team), 2000: Climate Change Impacts on the United States: The Potential Consequences of Cli- Morita, T. and H-C. Lee, 1998: IPCC SRES Database. Database pre- mate Variability and Change, Overview. Cambridge University pared for IPCC Special Report on Emissions Scenarios. http:www- Press, Cambridge, UK. cger.nies.go.jp/cger-e/db/ipcc.html. NAST (National Assessment Synthesis Team), 2001: Climate Change Morita, T., J. Robinson, A. Adegbulugbe, J. Alcamo, D. Herbert, E.L. La Impacts in the United States: Potential Consequences of Climate Rovere, N. Nakicenovic, H. Pitcher, P. Raskin, K. Riahi, A. Change and Variability and Change. Cambridge University Press, Sankovski, V. Sokolov, B. de Vries, and Z. Dadi, 2001: Greenhouse Cambridge, UK. gas emissions: mitigation scenarios and implications. In: Climate Change 2001: Mitigation [Metz, B., O. Davidson, R. Swart, and J. NRC (US National Research Council), 1996: Understanding Risk: In- Pan (eds.)]. Cambridge University Press, Cambridge, UK. forming Decisions in a Democratic Society. [Stern, P.C. and H.V. Fineberg (eds.)]. Committee on Risk Characterization. National Morgan, M.G. and D.W. Keith, 1995: Subjective judgments by climate Academy Press, Washington, DC, USA. experts. Environmental Science & Technology 29, 468-476. NRC (US National Research Council), 1999: Our Common Journey: A Morgan, M.G., M. Kandlikar, J. Risbey, and H. Dowlatabadi, 1998: Why Transition toward Sustainability. Board on Sustainable Develop- conventional tools of policy analysis are often inadequate for prob- ment. National Academy Press, Washington, DC, USA. lems of global change. Climatic Change 41, 271-281.

101 The U.S. Climate Change Science Program References

NRC (US National Research Council), 2002: Abrupt Climate Change: Pittock, A.B., R.N. Jones, and C.F. Mitchell, 2001: Probabilities will help Inevitable Surprises. Committee on Abrupt Climate Change. Na- us plan for climate change. Nature 413, 249. tional Academy Press, Washington, DC, USA. Providence Journal, 2004. Pentagon report plans for climate catastro- NRC (US National Research Council), 2005: Decision Making for the phe. March 3, A1. Environment: Social and Behavioral Science Research Priorities. Committee on Human Dimensions of Global Change. National Raisanen, J. and T. Palmer, 2001: A probability and decision-model Academy Press, Washington, DC, USA. analysis of a multi-model ensemble of climate change simulations. Journal of Climate 14, 3212-3226. Parikh, J.K., 1992: IPCC strategies unfair to the South. Nature 360, 507-508. Raskin, P., T. Banuri, G. Gallopin, P. Gutman, A. Hammone, R. Kates, and R. Swart, 2002: Great Transition: the Promise and Lure of the Parikh, J.K., 1998: The emperor’s new clothes: long-range energy-use Times Ahead. Global Scenario Group, Stockholm Environment In- scenarios by IIASA-WEC and IPCC. Energy 23(1), 69-70. stitute, Boston.

Parson, E.A., 1996: What can you learn from a game? In: Wise Choices: Reilly, J., P.H. Stone, et al., 2001: Climate change – uncertainty and cli- Games, Decisions, Negotiations [Zeckhauser, R., R.L. Keeney, and mate change assessments. Science 293, 430. J.K. Sebenius (eds.)]. Harvard Business School Press, Boston, USA. Renn, O., T. Webler, and P. Wiedemann, 1995: Fairness and Compensa- Parson, E.A., 1997: Informing global environmental policy-making: a tion in Citizen Participation. Kluwer, Dordrecht, The Netherlands. plea for new methods of assessment and synthesis. Environmental Modeling and Assessment 2(4), 267-279. Roberts, L., 1991: Learning from an acid rain program. Science 251, 1302-1305. Parson, E.A., 2003: Protecting the Ozone Layer: Science and Strategy. Oxford University Press, New York, USA. Robinson, J.B., 1982: Energy backcasting: a proposed method of policy analysis. Energy Policy 10(4), 337-45. Parson, E.A., 2006: Reflections on air capture: the political economy of active intervention in the global environment. Climatic Change 74, 1-3. Robinson, J.B., 2003: Future subjunctive: backcasting as social learning. Futures 35, 839-856. Parson, E.A. and K. Fisher-Vanden, 1997: Integrated assessment models of global climate change. Annual Review of Energy and the En- Rosenberg, N., P. Crosson, K. Frederick, W. Easterling, M. McKenney, vironment 22, 589-628. M. Bowes, R. Sedjo, J. Darmstadler, L. Katz, and K. Lemon, 1993: The MINK methodology: background and baseline. Climatic Parson, E.A. and M.G. Morgan, with A. Janetos, L. Joyce, B. Miller, R. Change 24, 7-22. Richels, and T. Wilbanks, 2001: Socioeconomic context for climate impact assessment. In: Climate Change Impacts on the United Rosenzweig, C. and W.D. Solecki (eds.), 2001: Climate change and a States, NAST. Cambridge University Press, Cambridge, UK. global city: the potential consequences of climate variability and change - metro east coast. Columbia Earth Institute, New York, USA. Parson, E.A., R.W. Corell, et al., 2003: Understanding climate impacts, vulnerabilities, and adaptation in the United States: building a ca- Rosenzweig, C. and D.C. Major, 2006: Climate Impact Assessment of pacity for assessment. Climatic Change 57(1), 9-42. Environmental Infrastructure Systems: Phase I Final Report: Scop- ing for Phase II. Center for Climate Systems Research, Columbia Payne, J.T., A.W. Wood, A.F. Hamlet, R.N. Palmer, and D.P. Lettenmaier, University, New York, USA. 2004: Mitigating the effects of climate change on the water resources of the Columbia River Basin. Climatic Change 62(1-3), 233-256. Rothman, D., J. Robinson, and D. Biggs, 2003: Signs of life: linking indicators and models in the context of QUEST. In: Implementing Sus- Perhac, R.M., 1991: Usable science: lessons from acid rain legislation, tainable Development - Integrated Assessment and Participatory NAPAP. Power Engineering 95, 10, 26-29. Decision-Making Processes [Abaza, H. and A. Baranzini (eds.)]. Ed- ward Elgar, Cheltenham, UK. Pitcher, H.M., 2005. Downscaling: something for nothing? Presentation to workshop on global-change scenarios, Snowmass Colorado, Rubin, E.S., 1991: Benefit-cost implications of acid rain controls: an July 26. evaluation of the NAPAP integrated assessment. Journal of the Air and Waste Management Association 41(7), 914-921.

102 Global-Change Scenarios - Their Development and Use

Ruosteenoja, K., T.R. Carter, K. Jylha, and H. Tuomenvirta, 2003: Fu- Shubik, M., 1975: The Uses and Methods of Gaming. Elsevier, New ture Climate in World Regions: An Intercomparison of Model-Based York, USA. Projections for the New IPCC Emissions Scenarios. Finnish Envi- ronment Institute, Helsinki. Shinkle, K.D. and R.K. Dokka, 2004: Rates of Vertical Displacement at Benchmarks in the Lower Mississippi Valley and the Northern Gulf Russell, M., 1992: Lessons from NAPAP. Ecological Applications 2(2), Coast. NOAA Technical Report NOS/NGS 50, US Department of 107-110. Commerce, National Oceanic and Atmospheric Administration, Washington, DC, USA. San Francisco Chronicle, 2004. Pentagon-sponsored climate report sparks hullabaloo in Europe. February 25, A5. Smil, V., 2005: Energy at the Cross Roads. MIT Press, Cambridge, MA, USA. Schelling, T.C., 1964: An uninhibited pitch for crisis games. In: Crisis Games 27 Years Later: Plus C’est Deja VuI. P-7719, RAND Cor- Southeast Regional Assessment Team, 2002: Preparing for a Chang- poration, Santa Monica, CA, USA. ing Climate: Potential Consequences of Climate Variability and Change, Southeast. Publication 8-40002, Global Hydrology and Cli- Schelling, T.C., 1994: Nuclear history program oral history transcript. mate Center, University of Alabama in Huntsville, USA. NHP Berlin Crisis Oral History Project, Interview with Thomas Schelling. University of Maryland Center for International Security Stipp, D., 2004: The Pentagon’s weather nightmare: the climate could Studies, Nuclear History Program, College Park, MD, USA. change radically, and fast. Fortune February 9, 100.

Schelling, T.C., 1983: Climate change: implications for welfare and pol- Stockholm Environment Institute, 1999: POLESTAR: System Manual icy. In: Changing Climate, Report of the Carbon Dioxide Assess- for Version 2000. POLESTAR Series Report No. 2., Stockholm En- ment Committee, US National Research Council. National Academy vironment Institute, Sweden. Press, Washington DC, USA. Stouffer, R.J., S. Manabe, and K. Bryan, 1989: Interhemispheric asym- Schneider, S.H., 2001: What is “dangerous” climate change? Nature metry in climate response to a gradual increase of CO2. Nature 342, 411, 17-19. 660-662.

Schneider, S.H., 2002: Can we estimate the likelihood of climatic Svedin, U. and B. Aniansson (eds.), 1987: Surprising Futures. Notes changes at 2100? Climatic Change 52(4), 441-451. from an International Workshop on Long-term World Development, Friibergh Manor, Sweden, January 1986. Report 87:1, Swedish Schoemaker, P.J.H., 1995: Scenario planning: a tool for strategic think- Council for Planning and Coordination of Research, Stockholm. ing. Sloan Management Review, Winter. Swart, R.J., P. Raskin, and J. Ribonson, 2004: The problem of the future: Schultz, R.L. and E.M. Sullivan, 1972: Developments in simulation in sustainability science and scenario analysis. Global Environmental social and administrative science. In: Simulation in the Social and Change 14, 137-146. Administrative Sciences [Guetzkow, H., P. Kotler, and R.L. Schultz (eds.)]. Prentice-Hall, Englewood Cliffs, NJ, USA. Tebaldi, C., L.O. Mearns, R.L. Smith, and D. Nychka, 2004: Regional probabilities of precipitation change: a Bayesian analysis of multi- Schwartz, P., 1991: The Art of the Long View: Planning for the Future in model simulations. Geophysical Research Letters 31, L24213. an Uncertain World. Currency Doubleday, New York, USA. Tebaldi, C., R.L. Smith, D. Nychka, and L.O. Mearns, 2005: Quantify- Semenov, M.A. and J.R. Porter, 1995: Climatic variability and the mod- ing uncertainty in projections of regional climate change: a Bayesian eling of crop yields. Agricultural and Forest Meteorology 73, 265-283. approach to the analysis of multimodel ensembles. Journal of Cli- mate 18(10), 1524-1540. Shell International, 2001: Energy Needs, Choices, and Possibilities: Scenarios to 2050. Global Business Environment. Terleckyj, N.E., 1999a: Analytic Documentation of Three Alternate Socioeconomic Projections, 1997-2050. NPA Data Services, Shell International, 2003: Scenarios: an Explorer’s Guide. Global Busi- Washington, DC, USA. ness Environment. www-static.shell.com/static/royal-en/downloads/ scenarios_explorersguide.pdf. Terleckyj, N.E., 1999b: Development of Three Alternate National Projection Scenarios,1997-2050. NPA Data Services, Washington, DC, USA.

103 The U.S. Climate Change Science Program References

Toth, F. and T.W. Wilbanks, 2004: Considering the Technical and So- Van der Heijden, K., 1996: Scenarios: The Art of Strategic Conversation. cioeconomic Assumptions Embedded in the SRES Scenario Fami- John Wiley and Sons, Chichester, UK. lies. Guidance Paper, Fourth Assessment Report, IPCC Working Group 2. September 2004. van Notten, P.W.F., J. Rotmans, M.B.A. van Asselt, and D.S. Rothman, 2003: An updated scenario typology. Futures 35, 423–443. Tversky, A. and D. Kahneman, 1974: Judgment under uncertainty: heuristics and biases. Science 185, 1124-1131. van’t Klooster, S.A. and M.B.A. van Asselt, 2006: Practicing the scenario-axes technique. Futures 38, 15-30. UKCIP (UK Climate Impacts Programme), 1998: Climate Change Sce- narios for the United Kingdom. UKCIP Technical Report No. 1, Vaughan, D.G. and J.R. Spouge, 2002: Risk estimation of collapse of Oxford, UK. http://www.ukcip.org.uk/resources/publications/ the West Antarctic Ice Sheet. Climatic Change 52, 65-91. pub_ dets.asp?ID=11. VEMAP Members, 1995: Vegetation/Ecosystem Modeling and Analy- UKCIP (UK Climate Impacts Programme), 2000: Climate Change: As- sis Project (VEMAP): comparing biogeography and biogeochem- sessing the Impacts, Identifying the Responses [McKenzie Hedger, istry models in a continental-scale study of terrestrial ecosystem M., M. Gawith, I. Brown, R. Connell, and T.E. Downing (eds.)]. responses to climate change and CO2 doubling. Global Biogeo- UKCIP Short Report, Oxford, UK. http://www.ukcip.org.uk/re- chemical Cycles 9(4), 407-437. sources/publications/pub_dets.asp?ID=16. von Neumann, J. and O. Morgenstern, 1944: Theory of Games and Eco- UKCIP (UK Climate Impacts Programme), 2001: Socio-Economic Sce- nomic Behavior. Princeton University Press, Princeton, NJ, USA. narios for Climate Change Impact Assessment: A Guide to Their Use in the UKCIP. UKCIP, Oxford, UK. ukcip.org.uk/resources/publi- Wack, P., 1985a: Scenarios: uncharted waters ahead. Harvard Business cations/documents/34.pdf. Review 63(5), 73-89.

UKCIP (UK Climate Impacts Programme), 2005: Measuring Progress: Wack, P., 1985b: Scenarios: shooting the rapids. Harvard Business Re- Preparing for Climate Change through the UK Climate Impacts Pro- view 63(6), 139-150. gramme [West, C. and M. Gawith (eds.)]. UKCIP, Oxford, UK. http://www.ukcip.org.uk/resources/publications/pub_dets.asp?ID=68. Wallsten, T.S. and R.G. Whitfield, 1986: Assessing the Risks to Young Children of Three Effects Associated with Elevated Blood Lead Lev- UK House of Lords, 2005: The Economics of Climate Change. Select els. ANL/AA-32, Argonne National Laboratory, Argonne, IL, USA. Committee on Economic Affairs. 6 July. The Stationery Office Lim- ited, London, UK. Washington, W.M. and G.A. Meehl, 1989: Climate sensitivity due to increased CO2: experiments with a coupled atmosphere-ocean gen- UK Office of Science and Technology, 2002: Foresight Futures 2020: eral circulation model. Climate Dynamics 4(1),1-38. Revised Scenarios and Guidance. Department of Trade and Industry, London, UK. WBCSD (World Business Council on Sustainable Development), 1997: Exploring Sustainable Development: WBCSD Global Sce- US Army Corps of Engineers, 2004: Louisiana Coastal Area (LCA), narios, 2000-2050. www.wbcsd.org/DocRoot/FFiAJwjBGGNjla- Louisiana, Ecosystem Restoration Study. New Orleans District, wOAipD/exploringscenarios.pdf. Louisiana, USA. Webster, M.D., 2003: Communicating climate change uncertainty to US Climate Change Science Program and Subcommittee on Global policy-makers and the public. Climatic Change 61, 1-2, 1-8. Change Research, 2003: Strategic Plan for the U.S. Climate Change Science Program. Washington, DC, USA. www.climate- WEC/IIASA (World Energy Council, International Institute for Applied science.gov/Library/stratplan2003/final/default.htm. Systems Analysis), 1995: Global Energy Perspectives to 2050 and Beyond. World Energy Council, London, UK. US EPA (US Environmental Protection Agency), 1989: The Potential Effects of Global Climate Change on the United States [Smith, J. and D. Tirpak (eds.)]. EPA-230-05-89-050, Washington, DC, USA.

US OTA (US Congress, Office of Technology Assessment), 1993: Preparing for an Uncertain Climate (2 vols.). OTA-O-567 and -568, US Government Printing Office, Washington, DC, USA.

104 Global-Change Scenarios - Their Development and Use

Weyant, J., O. Davidson, H. Dowlatabadi, J.A. Edmonds, M. Grubb, E.A. Parson, R. Richels, J. Rotmans, P.R. Shukla, R.S.J. Tol, W. Cline, and S. Fankhauser, 1996: Integrated assessment of climate change: an overview and comparison of approaches and results. In: Climate Change 1995, Economic and Social Dimensions of Climate Change [Bruce, J.P., H. Lee, and E.F. Haites (eds)]. Cambridge University Press, Cambridge, UK.

Weyant, J.P. and J.N. Hill, 1999: Introduction and overview, the costs of the Kyoto Protocol: a multi-model evaluation. The Energy Journal, Special Issue, vii-xliv.

Weyant, J.P. (ed.), 2004: EMF 19: alternative technology strategies for climate change policy. Energy Economics 26, 4.

Wigley, T.M.L., R. Richels, and J.A. Edmonds, 1996: Economic and environmental choices in the stabilization of atmospheric CO2 concentrations. Nature 379, 240-243.

Wilby, R.L. and T.M.L. Wigley, 1997: Downscaling general circulation model output: a review of methods and limitations. Progress in Phys- ical Geography 21, 530-548.

Williams, K.L., K.C. Ewel, R.P. Stumpf, F.E. Putz, and T.W. Workman, 1999: Sea-level rise and coastal forest retreat on the west coast of Florida. Ecology 80(6), 2045-2063.

105 The U.S. Climate Change Science Program References

106 CONTACT INFORMATION

Global Change Research Information Office The Climate Change Science Program incorporates c/o Climate Change Science Program Office the U.S. Global Change Research Program and the 1717 Pennsylvania Avenue, NW Climate Change Research Initiative. Suite 250 Washington, DC 20006 To obtain a copy of this document, place an order at 202-223-6262 (voice) the Global Change Research Information Office 202-223-3065 (fax) (GCRIO) web site: http://www.gcrio.org/orders

CLIMATE CHANGE SCIENCE PROGRAM AND THE SUBCOMMITTEE ON GLOBAL CHANGE RESEARCH

William J. Brennan, Chair Jacqueline Schafer Department of Commerce U.S. Agency for International Development National Oceanic and Atmospheric Administration Acting Director, Climate Change Science Program Joel Scheraga Environmental Protection Agency Jack Kaye, Vice Chair National Aeronautics and Space Administration Harlan Watson Department of State Allen Dearry Department of Health and Human Services

Jerry Elwood EXECUTIVE OFFICE Department of Energy AND OTHER LIAISONS

Mary Glackin Melissa Brandt National Oceanic and Atmospheric Administration Office of Management and Budget

Patricia Gruber Stephen Eule Department of Defense Department of Energy Director, Climate Change Technology Program William Hohenstein Department of Agriculture Katharine Gebbie National Institute of Standards & Technology Linda Lawson Department of Transportation Margaret McCalla Office of the Federal Coordinator for Meteorology Mark Myers U.S. Geological Survey George Banks Council on Environmental Quality Jarvis Moyers National Science Foundation Gene Whitney Office of Science and Technology Policy Patrick Neale Smithsonian Institution U.S. Climate Change Science Program 1717 Pennsylvania Avenue, NW • Suite 250 • Washington, D.C. 20006 USA 1-202-223-6262 (voice) • 1-202-223-3065 (fax) http://www.climatescience.gov