Chapter 6 Methodology for Developing the MA Scenarios · 2011. 6. 12. · Chapter 6 Methodology for Developing the MA Scenarios Coordinating Lead Authors: Joseph Alcamo, Detlef van

Chapter 6

Methodology for Developing the MA Scenarios

Coordinating Lead Authors: Joseph Alcamo, Detlef van Vuuren, Claudia RinglerLead Authors: Jacqueline Alder, Elena Bennett, David Lodge, Toshihiko Masui, Tsuneyuki Morita,* Mark

Rosegrant, Osvaldo Sala, Kerstin Schulze, Monika ZurekContributing Authors: Bas Eickhout, Michael Maerker, Kasper KokReview Editors: Antonio Alonso Concheiro, Yuzuru Matsuoka, Allen Hammond

Main Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

6.2 Background to the MA Scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

6.3 Overview of Procedure for Developing the MA Scenarios . . . . . . . . . 1486.3.1 Organizational Steps6.3.2 Scenario Storyline Development and Quantification6.3.3 Synthesis, Review, and Dissemination6.3.4 Linkages between Different Spatial and Temporal Scales

6.4 Building the Qualitative Scenarios: Developing Storylines . . . . . . . . . 151

6.5 Building the Quantitative Scenarios: The Global Modeling Exercise . . 1526.5.1 Organization of the Global Modeling Exercise6.5.2 Specifying a Consistent Set of Model Inputs6.5.3 Soft-linking the Models6.5.4 Modeling Changes in Biodiversity6.5.5 Comments on the Modeling Approach

6.6 Discussion of Uncertainty and Scenarios . . . . . . . . . . . . . . . . . . . . . 1556.6.1 Using the Scenario Approach to Explore Uncertainty6.6.2 Communicating Uncertainties of the Scenarios6.6.3 Sensitivity Analysis

6.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156

APPENDIXES: DESCRIPTIONS OF MODELS

6.1 The IMAGE 2.2 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157

6.2 The IMPACT Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158

6.3 The WaterGAP Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160

6.4 The Asia-Pacific Integrated Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 161

*Deceased.

PAGE 145

145

................. 11411$ $CH6 10-27-05 08:42:01 PS

146 Ecosystems and Human Well-being: Scenarios

6.5 Ecopath with Ecosim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163

6.6 Terrestrial Biodiversity Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166

6.7 Freshwater Biodiversity Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169

6.8 Multiscale Scenario Development in the Sub-global Assessments . . . 170

REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171

BOXES

6.1 MA Procedure for Developing Scenarios

6.2 A Comparison of Global and Sub-global ScenarioDevelopment

6.3 Models Used in MA Global Modeling ExerciseAppendix 6.1 The Three Ecopath with Ecosim Models Used

FIGURES

6.1 Status of Information, from Low to High, versus Degree ofControllability

6.2 Overall Methodology of MA Scenario Development

6.3 Reporting Regions for the Global Modeling Results of theMA*

6.4 Linkages between Models

6.5 Scale for Assessing State of Knowledge and StatementConfidence

6.6 Scheme for Describing ‘‘State of Knowledge’’ or Uncertaintyof Statements from Models or Theories

Appendix 6.1 Structure of IMAGE 2.2 ModelAppendix 6.2 Structure of IMPACT ModelAppendix 6.3 Structure of WaterGAP ModelAppendix 6.4 Structure of AIM/Water ModelAppendix 6.5 Structure of AIM/Ecosystem ModelAppendix 6.6 Structure of Ecopath with Ecosim (EwE) ModelAppendix 6.7 Structure of Ecopath Model

*This appears in Appendix A at the end of this volume.

PAGE 146

Appendix 6.8 Multiscale Design of the Sub-global Assessments inSouthern Africa and Portugal

TABLES

6.1 Global Modeling OutputAppendix 6.1 Overview of Major Uncertainties in the IMAGE

ModelAppendix 6.2 Level of Confidence in Different Types of Scenario

Calculations from IMAGEAppendix 6.3 Overview of Major Uncertainties in the IMPACT


Calculations from IMPACTAppendix 6.5 Overview of Major Uncertainties in WaterGAP


Calculations from WaterGAPAppendix 6.7 Overview of Major Uncertainties in AIM ModelAppendix 6.8 Level of Confidence in Different Types of Scenario

Calculations from AIMAppendix 6.9 Overview of Major Uncertainties in Ecopath with

Ecosim (EwE) ModelAppendix 6.10 Level of Confidence in Different Types of Scenario

Calculations from Ecopath with EcosimAppendix 6.11 Overview of Major Uncertainties in the Terrestrial

Biodiversity ModelAppendix 6.12 Overview of Major Uncertainties in the Freshwater

Biodiversity Model

................. 11411$ $CH6 10-27-05 08:42:03 PS

147Methodology for Developing the MA Scenarios

Main Messages

The Millennium Ecosystem Assessment scenarios break new ground inglobal environmental scenarios by explicitly incorporating both ecosys-tem dynamics and feedbacks. The goal of the MA is to provide decision-makers and stakeholders with scientific information on the links between eco-system change and human well-being. Scenarios are used in this context toexplore alternative futures on the basis of coherent and internally consistentsets of assumptions. The scenarios are novel in that they incorporate feed-backs between social and ecological systems and consider the connectionsbetween global and local socioecological processes.

The approach to scenario development used in the MA combines qualita-tive storyline development and quantitative modeling. In this way, thescenarios capture the aspects of ecosystem services that are possible toquantify, but also those that are difficult or even impossible to expressin quantitative terms. The MA developed scenarios of ecosystem servicesand human well-being to 2050, with selected results up to 2100. Scenarioswere developed in an iterative process of storyline development and modeling.The storylines covered many complex aspects of society and ecosystems thatare difficult to quantify, while the models helped ensure the consistency of thestorylines and provided important numerical information where quantificationwas possible.

In the MA, scenarios were partly quantified by using linked global modelsto ensure integration across future changes in ecosystem services. Avail-able global models do not allow for a comprehensive assessment of the link-ages among ecosystem change, ecosystem services, human well-being, andsocial responses to ecosystem change. To assess ecosystem change for alarger set of services, several global models were linked and run based on aconsistent set of scenario drivers to ensure integration across future changesin ecosystem services.

While advances have been made by the MA in scenario development toexplore possible futures of the linkages between ecosystem change andsociety, still further progress is possible. Using quantitative and qualitativetools, the MA scenarios cover a large number of ecological services and driv-ers of ecosystem change. In the course of developing these scenarios, we alsoidentified areas where analytical tools are relatively weak. For quantification ofecosystem service scenarios, we particularly need models that further disag-gregate services to local scales, address cultural and supporting ecosystemservices, and consider feedbacks between ecosystem change and human de-velopment.

6.1 IntroductionThe goal of the Millennium Ecosystem Assessment is toprovide decision-makers and stakeholders with scientific in-formation on the links between ecosystem change andhuman well-being. The MA focuses on ecosystem services(such as food, water, and biodiversity) and on the conse-quences of changes in ecosystems for human well-being andfor other life on Earth. Ecosystem change, on the otherhand, is significantly affected by human decisions, oftenover long time horizons (Carpenter 2002). For example,changes in soils or biodiversity of long-lived organisms canhave legacy effects that last for decades or longer. Thus itis crucially important to consider the future when makingdecisions about the current management of ecosystem ser-vices.

PAGE 147

The MA developed scenarios to provide decision-makersand stakeholders with scientific information on the linksbetween ecosystem change and human well-being. Thischapter describes the methodology used to develop the sce-narios. It first provides background information on scenar-ios in general, followed by an overview of the methodologyused in the MA scenario development. The developmentof the qualitative storylines and the global modeling exer-cise are then described in detail. Finally, we briefly describehow uncertainty and scale issues were handled in the sce-narios. Eight Appendixes provide detailed descriptions ofthe models we relied on.

6.2 Background to the MA ScenariosApplied (natural and social) sciences have used many meth-ods for devising an understanding of the future, includingpredictions, projections, and scenario development. (For anoverview of methods, see, e.g., Glenn and Gordon 2005.)Each approach has its own methodology, levels of uncer-tainty, and tools for estimating probabilities. It should benoted that these terms are often not strictly separated in theliterature. The conventional difference, however, is that aprediction is an attempt to produce a most likely descriptionor estimate of the actual evolution of a variable or system inthe future. (The term ‘‘forecasting’’ is also often used; it isused interchangeably with prediction in this chapter.) Pro-jections differ from predictions in that they involve assump-tions concerning, for example, future socioeconomic andtechnological developments that may not be realized. Theyare therefore subject to substantial uncertainty. Scenarios areneither predictions nor projections and may be based on anarrative storyline. Scenarios may be derived from projec-tions but often include additional information from othersources.

Over the years, experience with global assessment proj-ects in the ecological and environmental realm has shownthat prediction over large time periods is difficult if not im-possible, given the complexity of the systems examined andthe large uncertainties associated with them—particularlyfor time horizons beyond 10–20 years. In the case of eco-systems, heterogeneity, non-linear dynamics, and cross-scaleinteractions of ecosystems contribute to system complexity(Holling 1978; Levin 2000). Furthermore, ecological pre-dictions are contingent on drivers that may be even moredifficult to predict, such as human behavior. As a result,people rarely have enough information to produce reliablepredictions of ecosystem behavior or environmental change(Sarewitz et al. 2000; Funtowicz and Ravetz 1993). Despitethese problems with predicting the future, people need totake decisions with implications for the future. Scenario de-velopment offers one approach to dealing with uncertainty.

The MA uses the IPCC definition of scenarios as ‘‘plau-sible descriptions of how the future may develop, based ona coherent and internally consistent set of assumptions aboutkey relationships and driving forces’’ (such as rate of tech-nology changes and prices) (IPCC 2000b). As such, scenar-ios are used as a systematic method for thinking creatively

................. 11411$ $CH6 10-27-05 08:42:03 PS


about complex, uncertain futures for both qualitative andquantitative aspects. Figure 6.1 presents an overview of theadditional value created by scenario analysis compared withmore deterministic approaches such as predictions.

Scenarios can serve different purposes (see, e.g., Alcamo2001; van der Heijden 1997). They can be used in an ex-plorative manner or for scientific assessment in order to un-derstand the functioning of an investigated system. For theMA, researchers are interested in exploring hypothesizedinteractions and linkages between key variables related toecosystems and human well-being. Scenario outcomes canthen form part of planning and decision-making processesand help bridge the gap between the scientific and the policy-making communities. The MA scenarios can also be usedin an informative or educational way. Depending on theprocess used, scenarios can also challenge the assumptionsthat people have about the future and can illustrate the dif-ferent views on their outcomes held by participants of thescenario-building exercise.

In general, scenarios contain a description of step-wisechanges or a storyline, driving forces, base year, and timesteps and horizon (Alcamo 2001). They are often classifiedby the method used to develop them, their goals and objec-tives, or their output. One classification of scenarios distin-guishes between ‘‘exploratory’’ and ‘‘anticipatory’’ scenarios.Exploratory scenarios are descriptive and explore trends intothe future. Anticipatory scenarios start with a vision of thefuture that could be optimistic, pessimistic, or neutral andwork backwards in time to discern how that particular futuremight be reached. This type of scenario is sometimes alsoreferred to as a ‘‘normative’’ scenario. The MA scenarioswere developed using mostly exploratory approaches.

Finally, scenarios can consist of qualitative information,quantitative information, or both. Chapter 2 contains someexamples of each of these types of scenarios. Qualitativescenarios, using a narrative text to convey the main scenariomessages, can be very helpful when presenting information

Figure 6.1. Status of Information, from Low to High, versusDegree of Controllability. PDF�probability distribution function.The figure shows the domain of traditional decision tools such asutility optimization and the domain where scenarios may be helpful.(Adapted from Peterson et al. 2003)

PAGE 148

to a nonscientific audience. Quantitative scenarios usuallyrely on modeling tools incorporating quantified informa-tion to calculate future developments and changes and arepresented in the form of graphs and tables.

Both scenario types can be combined to develop inter-nally consistent storylines assessed through quantificationand models, which are then disseminated in a narrativeform. (A ‘‘storyline’’ is a scenario in written form, and usu-ally takes the form of a story with a very definite messageor ‘‘line’’ running through it.) This approach was used todevelop the MA scenarios. The qualitative scenarios (story-lines) provide an understandable way to communicate com-plex information, have considerable depth, describecomprehensive feedback effects, and incorporate a widerange of views about the future. The quantitative scenariosare used to check the consistency of the qualitative scenar-ios, to provide relevant numerical information, and to ‘‘en-rich’’ the qualitative scenarios by showing trends anddynamics not anticipated by the storylines. By ‘‘consis-tency’’ we mean that the storylines do not contain elementsthat are contradictory according to current knowledge. Onthe other hand, the goal of developing consistent scenariosshould not lead us to omit elements that may look contra-dictory but in reality are only surprising new connectionsor results. Often, uncovering these unanticipated connec-tions that challenge current beliefs and assumptions is oneof the most powerful results of the scenarios analysis. Forexample, is a scenario about climate impacts only consistentwhen it assumes that climate change will lead to globalwarming? Indeed, there are ‘‘surprising’’ yet plausible andconsistent scenarios that postulate that climate change willlead to cooling of parts of the lower atmosphere.

Together, the qualitative and quantitative scenarios pro-vide a powerful combination that compensates for some ofthe deficits of either one on its own. The combination ofqualitative with quantitative scenarios has been used inmany recent global environmental assessments, such asIPCC’s Special Report on Emissions Scenarios (IPCC2000b), UNEP’s Global Environment Outlook (UNEP2002), the scenarios of the Global Scenario Group (Raskinet al. 1998), and the World Water Vision scenarios (Cos-grove and Rijsberman 2000; Alcamo et al. 2000).

The distinction between qualitative and quantitativescenarios is sometimes blurred, however. Qualitative sce-narios can be derived by formalized, almost quantitativemethods, while quantitative scenarios can be developed bysoliciting numerical estimates from experts or by usingsemi-quantitative techniques such as fuzzy set theory. Sto-rylines can also be interspersed with numerical data andthereby be viewed as both qualitative and quantitative.

As noted in earlier chapters, the main objective of theMA scenarios is to explore links between future changes inworld ecosystems and their services and human well-being.The scenario analysis focuses on the period up to 2050,with selected prospects for 2100.

6.3 Overview of Procedure for Developing the MAScenariosThis section describes the process used to develop the MAscenarios. The procedure consists of 14 steps organized into

................. 11411$ $CH6 10-27-05 08:42:05 PS


three phases (see Box 6.1); the details of the storyline devel-opment and the modeling exercise are explained in latersections. In the first phase, the scenario exercise was orga-nized and the main questions and focus of the alternativescenarios were identified. In the second phase, the storylineswere written and the scenarios were quantified using aniterative procedure. During the third phase, the results ofthe scenario analysis were synthesized, and scenarios andtheir outcomes were reviewed by the stakeholders of theMA, revised, and disseminated. These elements are also in-dicated in Figure 6.2. While Figure 6.2 suggests that activi-ties were completed once processed, in reality earlieractivities were often revised during an iterative process.

Two essential activities within the overall scenario de-velopment framework were the formulation of alternativescenario storylines and their quantification. These two ele-ments were designed to be mutually reinforcing. The de-velopment of scenario storylines facilitates internalconsistency of different assumptions and takes into accounta broad range of elements and feedback effects that are ei-ther difficult to quantify or for which no modeling capabilityexists, or both. Based on initial storylines, the quantificationprocess helps to provide insights into those processes wheresufficient knowledge exists to allow modeling, and to takeinto account the interactions among the various drivers andservices. During scenario development, several interactionswere organized between the storyline development and themodeling exercise in order to increase the consistency ofthe two approaches.

6.3.1 Organizational Steps

The first phase in the MA scenario development consistedof establishing a scenario guidance team, composed of the

BOX 6.1

MA Procedure for Developing Scenarios

Phase I: Organizational steps

1. Establish a scenario guidance team.2. Establish a scenario panel.3. Conduct interviews with scenario end users.4. Determine the objectives and focus of the scenarios.5. Devise the focal questions of the scenarios.

Phase II: Scenario storyline development and quantification

6. Construct a zero-order draft of scenario storylines.7. Organize modeling analyses and begin quantification.8. Revise zero-order storylines and construct first-order storylines.9. Quantify scenario elements.

10. Revise storylines based on results of quantifications.11. Revise model inputs for drivers and re-run the models.

Phase III: Synthesis, review, and dissemination

12. Distribute draft scenarios for general review.13. Develop final version of the scenarios by incorporating user feed-

back.14. Publish and disseminate the scenarios.

PAGE 149

Figure 6.2. Overall Methodology of MA Scenario Development

chairpersons and secretariat of the Scenarios WorkingGroup, to lead and coordinate the scenario-building proc-ess. In addition, a larger panel, composed mainly of scien-tific experts, was assembled to build the scenarios.

The scenario guidance team conducted a series of inter-views with potential users of the scenarios to obtain theirinput for developing the goals and focus of the scenarios.This was especially important for the MA because the num-ber of potential users is very large and diverse. These inter-views also ensured input from stakeholders and users earlyon in the study. Understanding the needs and desires ofusers and their outlook on future development helped theteam to devise the main focal questions of the scenarios.

Based on the results of the user interviews and discus-sions with the scenario panel, the objectives, focus, leadingthemes, and hypotheses of the scenarios were derived bythe scenario guidance team and panel (and later confirmedby the MA Assessment Panel). For the MA, the main objec-tive of scenario development was to explore alternative de-velopment paths for world ecosystems and their servicesover the next 50 years and the consequences of these pathsfor human well-being. Based on these results, the scenarioteam clarified the focal questions to be addressed by thescenarios (for the rationale behind the choice of questions,see Chapter 5). The main question was:

What are the consequences of plausible changes in developmentpaths for ecosystems and their services over the next 50 yearsand what will be the consequences of those changes for humanwell-being?

The key focal question was then defined through a series ofmore specific questions—that is, what are the consequences

................. 11411$ $CH6 10-27-05 08:42:06 PS


for ecosystem services and human well-being of strategiesthat emphasize:• economic and human development (e.g., poverty eradi-

cation, market liberalization) as the primary means ofmanagement?

• local and regional safety and protection, giving far lessemphasis to cross-border and global issues?

• development and use of technologies, allowing greatereco-efficiency and adaptive control?

• adaptive management and local learning about the con-sequences of management interventions for ecosystemservices?Ecosystem services are defined as ‘‘the conditions and

processes supported by biodiversity through which ecosys-tems sustain and fulfill human life, including the provisionof goods’’ (MA 2003; see also Chapter 1). Ecosystem proc-esses are seldom traded in markets, typically have no marketprice, and therefore usually do not enter in economicdecision-making or cost-benefit analyses even though theyare essential for human well-being. The MA considered thefollowing interlinked categories of ecosystem services: pro-visioning services (food, fresh water, and other biologicalproducts), supporting and regulating services (including soilformation, nutrient cycling, waste treatment, and climateregulation), and cultural services.

6.3.2 Scenario Storyline Development andQuantification

Following a review and evaluation of current and past sce-nario efforts, scenario building blocks for driving forces,ecological management dilemmas, branch points, and so onwere mapped out. Storyline outlines were then developedaround these building blocks. Additional details on the sto-ryline development are provided later in this chapter.

While the initial storylines were being developed, ateam of modelers representing several global models wasorganized to quantify the scenarios. Five global models cov-ering global change processes or ecosystem provisioningservices and two global models describing changes in bio-diversity were chosen. Criteria that were used to selectthese models included global coverage, publications ofmodel structure and/or model application in peer-reviewedliterature, relevance in describing the future of ecosystemservices, and ability to be adapted to the storylines of theMA. (The Ecopath with Ecosim models used to describemarine ecosystems and their service to global fisheries formsan exception to the rule of global coverage, as no globalmodel for this issue was available.) Although all the modelshad been developed previously, linkages among models andprojections out to 2050 and 2100 did require adjustmentsfor several of them. Test calculations were carried out usingpreliminary driving force assumptions. These test calcula-tions were helpful in identifying the potential contributionof different models to the analysis and in clarifying the pro-cedures of linking the different models.

After a series of iterations, the zero-order storylines wererevised and cross-checked for internal consistency. Onemeasure used to accomplish this was the development oftimelines and milestones for the various scenarios.

PAGE 150

In the next step, the modeling team, in consultationwith the storyline team, developed quantitative drivingforces that were considered to be consistent with the story-lines. Obviously, there is room for interpretation regardingthe consistency of driving forces and the storylines, andmultiple sets of driving forces are possible. The drivingforces and quantified drivers for the modeling exercise cho-sen for the MA scenarios are discussed in detail in Chapters7, 8, and 9. Based on the model outcomes of the quantita-tive scenarios, the scenario team further elaborated oradapted the storylines. A number of feedback workshopswith the MA Board and stakeholder groups were held toimprove the focus and details of the storylines.

Based on the results of the first round of quantified sce-narios, small adjustments in the specification of drivers andlinkages among models were made, new model calculationswere carried out with the modeling framework, and thestorylines were revised (in other words, there was one itera-tion between storyline development and quantification).Ideally, a series of iterations between storyline improve-ment, quantification, and stakeholder feedback sessionswould have helped to better harmonize the quantitative andqualitative scenarios, but time constraints limited the num-ber of iterations for the MA. The quantified scenario resultsare described in detail in Chapter 9, while the scenario sto-rylines can be found in Chapter 8.

6.3.3 Synthesis, Review, and Dissemination

The scenario outcomes were assessed in the context of thefocal questions and user needs of the various MA usergroups. These results are described in Chapters 11, 12, 13,and 14, based on an analysis of both qualitative and quanti-tative scenario outcomes. Feedback from the assessmentcomponent of the scenario team led to further refinementof the storylines and the provision of additional model de-tails.

The scenarios, consisting of the qualitative storylines andquantitative model calculations, were disseminated for re-view by interested user groups. This was accomplishedthrough presentations, workshops, the MA review process,and Internet communications. Reviewer comments werethen incorporated into the scenarios. Both review and dis-semination are considered important elements for the suc-cess of the scenario exercise.

6.3.4 Linkages between Different Spatial andTemporal Scales

In order to deal with the multiscale aspects of the relation-ships between ecosystem services and human well-being,the MA called for a large number of sub-global assessmentsin addition to the global assessment. Several of the sub-global assessments also developed scenarios. As these wereoften targeted at specific user groups or addressed very spe-cific questions, it was not always possible to directly link thesub-global assessments to the global assessment. (See alsoBox 6.2.) Nevertheless, to harmonize the global and sub-global scenario exercises as much as possible, the followingsteps were taken:

................. 11411$ $CH6 10-27-05 08:42:08 PS


BOX 6.2

A Comparison of Global and Sub-global ScenarioDevelopment

One of the goals of both the global and sub-global scenarios was toprovide foresight about potential futures for ecosystems, including theprovision of ecosystem services and human well-being.

Despite similar goals, sub-global scenario development was some-what different from global scenario development. The key differenceswere the extensive use of quantitative models in the global scenarios,greater involvement of decision-makers in the sub-global scenarios,and use of sub-global scenarios directly as a tool for decision-making(versus a broader learning-focused global scenario set).

Because the group of decision-makers at the global scale is morediffuse, involvement of decision-makers in the global scenario develop-ment was less intense than it was for the sub-global scenarios. Repre-sentatives from the business community, the public sector, and theinternational conventions were periodically informed of progress in thedevelopment of the global scenarios and asked for feedback. The sub-global assessments, because they often focused on issues for whichkey decision-makers could be identified, had closer contact with theirprimary intended users. The ultimate result of having decision-makersmore involved in scenario development was that the scenarios them-selves were built more as a direct tool for engaging people in decision-finding processes. Thus, in most sub-global assessments, scenario de-velopment focused on futures over which local decision-makers haveat least some direct control.

The global scenarios provide four global storylines from where alook down enriches these stories with regional and local details. Thesub-global scenarios provide a large number of local stories fromwhere a look up enriches the stories with regional and global ‘‘details.’’A more complete description of the sub-global scenarios can be foundin Chapter 9 of the MA Multiscale Assessments volume.

• Representatives of some sub-global assessments partici-pated in the global scenario team and contributed to thescenario development.

• Members of the global scenario guidance team partici-pated at various occasions in meetings of the sub-globalscenario assessments, explaining both the preliminaryglobal scenario results and the procedure followed in de-veloping the global scenarios.

• Some of the sub-global assessments used the storylinesof the global assessment as background for their work orotherwise linked their scenarios to the global assessment.

• After the storylines and the model runs of the globalscenarios were finalized, results and findings of the sub-global assessment were used to illustrate how the scenarioscould play out at the local scale. (See Boxes in Chapter 8.)As well as addressing changes in ecosystems and their

services at several spatial scales, the MA also considered dif-ferent temporal scales. For a more detailed discussion on thegeneral issue of scales in the MA, see the MA conceptualframework report (MA 2003). (See also Chapter 7 of thisvolume and Chapter 4 in the MA Multiscale Assessmentsvolume).

The question of temporal scale was important for theconstruction of the MA scenarios. Although the global sce-

PAGE 151

narios were primarily developed to the year 2050, scenarioresults of the quantitative scenario elaboration were also re-ported for 2020, 2050, and 2100. The 2020 report provideda link between the scenarios and medium-term policy ob-jectives, such as the 2015 Millennium Development Goals.It also linked the global and sub-global scenarios, many ofwhich extend only to approximately 2025. Meanwhile, theresults for 2100 impart insight into longer-term trends inecosystem services. Results for 2100 were only reported forparameters that are determined by strong inertia within thenatural system, such as climate change and sea level rise.

While several of the models used within the modelingexercise perform their calculations for 10–40 global regionsor countries, a much lower resolution was chosen for re-porting. Quantitative results (Chapter 9) are mainly pre-sented for six reporting regions: sub-Saharan Africa, MiddleEast and North Africa, the Organisation of Economic Co-operation and Development, the former Soviet Union,Latin America, and Asia. (See Figure 6.3 in Appendix A.)These are sometimes aggregated into ‘‘rich’’ or ‘‘wealthy’’countries and ‘‘poorer’’ countries.

The reasons for using this lower resolution include theamount of information that could be presented within thisvolume and checked for internal consistency. In addition,some models use a global grid of half-degree latitude andlongitude to calculate changes in environmental and eco-logical parameters. The latter are presented in case they arerelevant. Grid-level results should be interpreted as broad-brush visualizations of the geographic patterns underlyingthe scenarios, not as specific predictions for small regions oreven grid cells.

6.4 Building the Qualitative Scenarios:Developing StorylinesSignificant emphasis was placed on storyline development.Storylines can be provocative because they challenge thetendency of people to extrapolate from the present into thefuture. They can be used to highlight key uncertainties andsurprises about the future. They can consider nonlinearitiesand complicated causal links more easily than global modelscan. Moreover, they can incorporate important ecologicalprocesses, which so far have not been satisfactorily consid-ered in existing global models. (See Chapter 3.) Since theMA’s goal for scenarios development was to specificallyconsider the future of ecosystems and their services, story-line development was used to incorporate processes that themodels could not fully address. Moreover, the qualitativestories provided the input variables for the global models.

The qualitative storylines were developed through a se-ries of discussions among the scenario development panelalternating with feedback from MA user groups and outsideexperts. The storyline development followed six steps:• identification of what the MA user groups wanted to

learn from the scenarios,• development of a set of scenario building blocks,• determination of a set of basic storylines that reflected

the MA goal and responded to user needs,

................. 11411$ $CH6 10-27-05 08:42:08 PS


• development of rich details for the storylines,• harmonization of the storylines with modeling results,

and• feedback from experts and user groups and its incorpora-

tion into the final storylines.Although these steps are presented in order, the process

in reality cycled through some of the steps many times untilit was felt that consensus was reached on the storylines.

Key questions about the future and main uncertainties ofMA user groups were identified through a series of feedbacktechniques. Approximately 70 leaders and decision-makersfrom around the world and in many different decision-making positions were interviewed about their hopes andfears for the future. A formal User Needs survey developedby the MA at the beginning of the assessment was used asadditional input. This survey was sent to representatives ofthe MA user community and contained questions on ex-pected outcomes of the MA process. Synthesis of these sur-veys led to the formulation of the key questions listedearlier.

In the second step, the scenario team developed a num-ber of scenario building blocks, including the factors differ-entiating the scenarios. In addition, the scenario teamidentified possible driving forces of socioecological systemsinto the future, as well as the main uncertainties of thesedriving forces and the prospects for being able to steer them.Other scenario building blocks included discussions on eco-logical dilemmas that decision-makers are likely to face inthe near future, possible branching points of scenarios, theoccurrence of cross-scale ecological feedback loops, and as-sumptions that decision-makers hold about the functioningof ecological systems (such as whether ecological systemsare fragile or resilient).

A first set of scenario storylines was developed using anumber of different development paths to distinguishamong them. This was done through a combination ofwriting, presentations, and discussions within the scenarioteam and feedback from other working groups within theMA. Once these storylines were developed, they were pre-sented to a wider group of experts, including the MABoard, members of the World Business Council for Sustain-able Development, scenario experts from other scenario ex-ercises, and several decision-maker communities. Feedbackfrom this exercise led to further refinement of the storylines.

As the results of the quantified scenarios became avail-able, they were compared with the qualitative storylines.This led to further discussions about the logical pathways tothe final sequence of events in the scenarios. These discus-sions were encouraged by the structure of the scenarioteam, which included both storyline-writers and membersof the modeling teams. As a result of these discussions, sto-rylines and model driving forces were adjusted. These dis-cussions also led to new interpretations of the storylines intomodel parameters.

6.5 Building the Quantitative Scenarios: TheGlobal Modeling Exercise6.5.1 Organization of the Global Modeling Exercise

As noted in previous sections, the storyline developmentwas complemented by building quantitative scenarios using

PAGE 152

a linked set of global models. The purpose of the modelingexercise was both to test the consistency of the storylines asdeveloped in the first round and to elaborate and illustratethe scenarios in numerical form. This ‘‘quantification of thescenarios’’ had five main steps:• Assembling several global models to assess possible fu-

ture changes in the world’s ecosystems and their ser-vices. These models are briefly described in Box 6.3 andin the Appendixes. In addition, several models wereused to describe certain aspects of changes in biodiversity.

• Specifying a consistent set of model inputs based on thescenario storylines.

• Running the models with the specified model inputs.• Soft-linking the models by using the output from one

model as input to another (we use the term soft-link asthe models were not run simultaneously).

• Compiling and analyzing model outputs about changesin future ecosystem services and implications for humanwell-being. The models were used to analyze the futurestate of indicators for ‘‘provisioning,’’ ‘‘regulating,’’ and‘‘supporting’’ ecosystem services. These indicators arelisted in Table 6.1. The analysis of modeling results ispresented in Chapter 9.

6.5.2 Specifying a Consistent Set of Model Inputs

The first version of the storylines of the MA scenarios (andin particular, tables containing their main characteristics)formed the basis of the main model assumptions for thequantitative exercise. Over several workshops, the story-lines were translated into a consistent set of model assump-tions that closely corresponded to the ‘‘indirect drivers’’ ofecosystem services. These included:• population development, including total population and

age distribution in different regions;• economic development as represented by assumed growth

in per capita GDP per region and changes in economicstructure;

• technology development, covering many model inputs suchas the rate of improvement in the efficiency of domesticwater use or the rate of increase in crop yields;

• human behavior, covering model parameters such as thewillingness of people to invest time or money in energyconservation or water conservation; and

• institutional factors, such as the existence and strength ofinstitutions to promote education, international trade,and international technology transfer. The latter are rep-resented directly (trade barriers, for instance, and importtariffs) or indirectly (income elasticity for education) inthe models, based on the storylines.For each of these factors, trends were developed for

model inputs that corresponded to the qualitative state-ments of the storylines. For example, statements in the sto-rylines about ‘‘high’’ or ‘‘low’’ mortality were interpretedsuch that the trend in mortality would be in the upper orlower 20% of the probabilistic demographic projections.(See also Chapter 9.) Another example is that the scenariowith the highest level of agricultural intensification was as-sumed to have the fastest rate of improvement of crop yield

................. 11411$ $CH6 10-27-05 08:42:09 PS


BOX 6.3

Models Used in MA Global Modeling Exercise

The models in the global modeling exercise include:

• The IMPACT model of the International Food Policy Research Insti-tute in the United States, which computes food supply, demand,trade, and international food prices for countries and regions (Rose-grant et al. 2002).

• The WaterGAP model of the University of Kassel in Germany, whichcomputes global water use and availability on a watershed scale(Alcamo et al. 2003a, 2003b).

• The AIM global change integrated model of the National Institute forEnvironment Studies in Japan, which computes land cover and otherindicators of global change worldwide, with an emphasis on Asia(Kainuma et al. 2002).

• The IMAGE 2.2 global change model of the National Institute ofPublic Health and the Environment in the Netherlands, which com-putes climate and global land cover on a grid scale and severalother indicators of global change (IMAGE-team 2001).

• The Ecopath with Ecosim model of the University of British Columbiain Canada, which computes dynamic changes in selected marineecosystems as a function of fishing efforts (Pauly et al. 2000).

and largest expansion of irrigation development. An over-view of model inputs is provided in Chapter 9.

6.5.3 Soft-linking the Models

To achieve greater consistency between the calculations ofthe different models, they were ‘‘soft-linked’’ in the sensethat output files from one model were used as inputs toother models. (See Figure 6.4.) The time interval of datathat were exchanged between the models was usually oneyear. The following model linkages were included:• Computations of regional food supply, demand, and

trade from the IMPACT model were aggregated to the17 IMAGE world regions and the 12 IMAGE animaland crop types. These data were then used as input tothe IMAGE land cover model that computed on a globalgrid the changes in agricultural land that are consistentwith the agricultural production computed in IMPACT.In addition, IMAGE was used to calculate the amountof grassland needed for the livestock production com-puted in IMPACT. Two iterations were done betweenIMPACT and IMAGE to increase the consistency of theinformation on agricultural production, availability ofland, and climate change. (First, preliminary runs wereused as a basis for discussion between the two groups onthe consistency of trends under each of the four scenar-ios for agricultural production, yields changes, and im-pacts on total land use in each region; for the final runs,an additional iteration was done, providing informationfrom IMAGE to IMPACT on yield changes resultingfrom climate change and use of marginal lands.) Thelinkage between IMPACT and IMAGE was done for2000, 2020, 2050, and 2100.

PAGE 153

Table 6.1. Global Modeling Output

Model Used toEcosystem CalculateService Indicator Indicator

Direct drivers of ecosystem changeClimate change temperature change, precipitation IMAGE

change

Changes in land areas per land cover and land IMAGEuse and land use typecover

Technology water use efficiency, energy IMPACT, IMAGE,adaptation and efficiency, area and numbers WaterGAPuse growth, crop yield growth, and

changes in livestock carcassweight

Exogenous fertilizer use, irrigation, wage IMPACT, IMAGEinputs rates

Air pollution sulfur and NOx emissions AIM, IMAGEemissions

Provisioning servicesFood total meat, fish, and crop IMPACT

production; consumption; trade,food prices

Food potential food production, crop IMAGE, AIMarea, pasture area, and area forbiofuels

Fish stock Ecopath/Ecosim

Fuelwood biofuel supply IMAGE, AIM

Fresh water annual renewable water WaterGAP, AIMresources, water withdrawals andconsumption, return flows

Regulating servicesClimate net carbon flux IMAGEregulation

Erosion erosion risk IMAGE

Supporting servicesPrimary primary production IMAGE, AIMproduction

Food security calorie availability, food prices, IMPACTshare of malnourished pre-schoolchildren in developing countries

Water security water stress WaterGAP, AIM

Input to biodiversity calculationsTerrestrial land use area, climate change, IMAGEbiodiversity nitrogen and sulfur deposition

Aquatic river discharge WaterGAPbiodiversity

climate change IMAGE

• Changes in irrigated areas computed in IMPACT wereallocated to a global grid in the WaterGAP model andthen used to compute regional irrigation water require-ments. These irrigation water requirements were thenadded to water withdrawals from the domestic and in-

................. 11411$ $CH6 10-27-05 08:42:10 PS


Figure 6.4. Linkages between Models

dustrial sectors (changes for these sectors were calculatedby WaterGAP) and compared with local water availabil-ity. From this comparison, WaterGAP estimated waterstress on a global grid.

• Changes in temperature and precipitation were calcu-lated in IMAGE based on trends in greenhouse gas emis-sions from energy and land use. These climate changecalculations were used in WaterGAP to computechanges in water availability and in IMPACT to estimatechanges in crop yield.

• The IMAGE model was used to compute changes inelectricity use and livestock production, the latter ob-tained from IMPACT, which were used by WaterGAPto estimate future water requirements in the electricityand livestock sectors.

• The model for Freshwater Biodiversity used inputs ofriver discharge from WaterGAP and climate fromIMAGE.

• The calculations of changes in terrestrial biodiversity re-lied on calculations of land cover changes, nitrogen de-position, and climate from IMAGE. Calculations weredone on the basis of annual IMAGE output data.

• The AIM modeling team used the same drivers in termsof population, economic growth, and technology devel-opment, but no linkages were made between the AIMmodel and outputs of other models.Linkages between models are seldom straightforward,

and usually require the upscaling (aggregation) or downsca-

PAGE 154

ling of various data. The IMPACT model, for instance, usesa more detailed regional disaggregation level than theIMAGE model. In almost all cases, however, regional scal-ing was possible by combining regions (upscaling) or byassuming proportional changes (downscaling). The modelswere not recalibrated on the basis of the new input parame-ters provided by the other models, but in most cases themodels had been calibrated using comparable internationaldatabases. The new linkages therefore did not lead to majorinconsistencies in assumptions between the models.

6.5.4 Modeling Changes in Biodiversity

Currently, there are no global models that describe changesin biodiversity on the basis of global scenarios. For the MAanalysis, several smaller models or algorithms were developedon the basis of linkages between global change parametersand species diversity to describe elements of biodiversitychange. Outcomes from these tools were then used for amore elaborate discussion of the possible impacts of the dif-ferent scenarios for global biodiversity. The methods usedare discussed in detail in Chapter 10.

6.5.4.1 Terrestrial Biodiversity

The possible future changes in terrestrial biodiversity underthe four MA scenarios were explored on the basis of theassessment of changes in native habitat cover over time, cli-mate change, and changes in nitrogen deposition. A review

................. 11411$ $CH6 10-27-05 08:42:13 PS


of global threats to biodiversity identified land use change,climate change, the introduction of alien species, nitrogendeposition, and carbon fertilization as major driving forcesfor extinction (Sala et al. 2000). Simple and well-establishedexisting relationships between these threats and species di-versity were used to explore the possible impacts under theMA scenarios. For land use change, for instance, the spe-cies-area relationship was used to describe potential loss ofplant diversity.

A second important cause of potential change in futureglobal biodiversity is climate change. The potential impactsof climate change were explored using several tools thatprovide insight into changes in biomes and plant diversityas a result of climate change. More qualitative assessmentswere made for the impacts of nitrogen and invasive species.Finally, changes in local biodiversity were directly estimatedon the basis of the land cover change scenarios.

6.5.4.2 Aquatic and Marine Biodiversity

Oberdorff et al. (1995) developed a global model to de-scribe the diversity of fresh waters as a function of riverdischarge, net primary production per area of the water-shed, watershed area, and fish species richness of the conti-nent. As river discharge can be used as a measure of the sizeof the freshwater habitat, Oberdorff ’s model is, in fact, anapproximation or expression of the species-area curve forfreshwater biodiversity. In our exercise, we updated the fishspecies numbers and discharge data in Oberdorff ’s modeland used it to describe changes in fish biodiversity as a func-tion of changes in drivers that affect river discharge.

For marine biodiversity, several studies are available onthe impacts of fisheries on the marine diversity for differenttrophic levels. However, to date no global model has beendeveloped. Instead, we applied several regional models todifferent seas around the world using the MA storylines tospecify their main assumptions. The results of these local-ized case studies served to indicate possible changes inglobal biodiversity. In addition, a qualitative discussionbased on expert knowledge of the impacts of other pres-sures, including climate change, on marine biodiversityadded to the interpretation of model results.

6.5.5 Comments on the Modeling Approach

The approach of using global models to quantify the scenar-ios has certain disadvantages that should be made explicit.The outcomes of global modeling exercises are highly un-certain, and their assumptions, drivers, and equation systemsare often difficult to explain to a nontechnical audience.Furthermore, the models brought together in the MAglobal modeling exercise were developed independentlyand were therefore not fully compatible. For example, theyhave different spatial and temporal resolutions.

The advantages of using global models as part of the sce-nario building process outweigh the disadvantages, how-ever. Although the model results are uncertain, the modelsthemselves have been published in the scientific literatureand have undergone peer review. They have also beenfound useful for linking science with policy issues in earlier

PAGE 155

international policy-relevant applications. The models usedin the MA exercise provide insights into the trends of manydifferent types of ecosystem services, including global foodproduction, the status of global freshwater resources, andglobal land cover. In combination, they provide a uniqueopportunity to generate globally comprehensive, rich, anddetailed information for enriching the MA storylines.

While more than 20 indicators were computed in thelinked modeling system, coverage of global ecosystem ser-vices and feedback effects remained limited. As noted, wetried to make up for this deficit by developing qualitativestorylines, which in text form can describe additional indi-cators and aspects of ecosystem services. At the same time,the modeling exercise addressed some of the deficits of thestorylines. For example, model calculations can be used tointerlink outcomes for various ecosystem services and to ex-plore the consistency of the storyline assumptions. As partof an overall approach, the storyline-writing and modelingexercise complement each other.

6.6 Discussion of Uncertainty and Scenarios

6.6.1 Using the Scenario Approach to ExploreUncertainty

As explained in the introduction, the main reason to usea scenario approach to explore the future development ofecosystem services is that the systems under study are toocomplex and the uncertainties too large to use alternativeapproaches, such as prediction. (See also Chapter 3.) There-fore scenario analysis is used as a tool to address the uncer-tainty of the future. The MA scenario analysis providesconcrete information about plausible future developmentpaths of ecosystem services and their relation to humanwell-being. The range of scenarios exemplifies the rangeof possible futures and in so doing helps stakeholders anddecision-makers to design robust strategies to preserve eco-system services for human well-being.

The high level of uncertainty about the future of ecosys-tem services also implies that is not possible to distinguishbetween the probability of one scenario versus another. Inscenario analysis we sometimes have an intuitive sense thatone scenario is more probable than another, but for the MAand most scenario exercises it is not fruitful to dwell ontheir relative probabilities. With regards to the MA scenar-ios, other scenarios are also possible, and it is highly unlikelythat any of the four scenarios developed for the MA wouldmaterialize as described. In other words, the four scenariosare only a small subset of limitless plausible futures. Theywere selected because they sampled broadly over the spaceof possible futures, they illustrated points about ecosystemservices and human well-being that the MA was charged toaddress, and they enabled us to answer the focal questionsposed by the MA Scenarios Working Group.

6.6.2 Communicating Uncertainties of theScenarios

Despite the uncertainty of the future, the scenarios containstatements that we intuitively judge as more likely than oth-

................. 11411$ $CH6 10-27-05 08:42:14 PS


ers. To communicate this certainty/uncertainty we use theexpressions shown in Figure 6.5. This scheme was developedfor handling uncertainty in assessments of the Intergovern-mental Panel on Climate Change (Moss and Schneider2000).

Associated with each statement of confidence is a quan-titative confidence level or range of probability. Accordingto this scale, a confidence level of 1.0 implies that we areabsolutely certain that a statement is true, whereas a level of0.0 implies that we are absolutely certain that the statementis false. It should be noted, however, that in this volumeconfidence levels are typically not estimated numerically.Instead, they are based on the subjective judgments of thescientists. Also, it is unusual to make statements that do nothave at least medium certainty (unless they are high-riskevents).

Another way to communicate uncertainty is shown inFigure 6.6, which describes a set of expressions for describ-ing the state of knowledge about models and parametersused for constructing the MA scenarios. These expressionscan be used to supplement the five-point scale of Figure 6.5in order to explain why a model outcome is associated withhigh, medium, or low confidence. These expressions areused, for example, in the Appendix of this chapter to de-scribe the uncertainties of different aspects of the modelsused in the global modeling exercise. They are also usedextensively in Chapter 9 to explain estimates of futurechanges in ecosystem services.

6.6.3 Sensitivity Analysis

In some cases, formal sensitivity analysis was used as part ofthe global modeling exercise to estimate the uncertainty ofcalculations. For example, the MA population scenarioswere selected from a stochastic calculation of populationprojections. (See Chapter 7.) Another example is the assess-ment of the uncertainty of climate change on water avail-ability. (See Chapter 9.) A third example is the use of

Figure 6.5. Scale for Assessing State of Knowledge andStatement Confidence (Moss and Schneider 2000)

PAGE 156

Figure 6.6. Scheme for Describing ‘‘State of Knowledge’’ orUncertainty of Statements from Models or Theories (Moss andSchneider 2003)

Monte-Carlo analysis as part of the calculation of changesin terrestrial biodiversity. (See Chapter 10.)

6.7 SummaryThe goal of the MA is to provide decision-makers andstakeholders with scientific information about linkages be-tween ecosystem change and human well-being. SeveralMA scenarios were developed to explore alternative futureson the basis of coherent and internally consistent sets ofassumptions. Scenario development was chosen instead ofother approaches, such as predictions, as scenarios are bettersuited to deal with the large inherent uncertainties of thecomplex relationships between ecological and human sys-tems and within each of these systems.

An important aspect of the MA scenarios is that theyneed to take into account ecosystem dynamics and ecosys-tem feedbacks. As earlier global scenarios have been gener-ated for other purposes, incorporation of realistic ecosystemdynamics is a novel aspect of the MA scenarios.

The MA developed scenarios of ecosystems services andhuman well-being from 2000–50 with selected outlooks to2100. The MA scenarios were developed by first definingqualitative storylines, followed by quantification of selectedstoryline drivers and parameters in an iterative process. Thedevelopment of scenario storylines allowed the process tofocus on internal consistency of different assumptions andalso to take into account a broad range of elements andfeedback effects that often cannot be quantified. Based oninitial storylines, the quantification process helped to pro-vide insights into processes where sufficient knowledge ex-ists to allow for modeling and to take into account the

................. 11411$ $CH6 10-27-05 08:42:18 PS


interactions among the various drivers and ecosystem ser-vices.

APPENDIXES: Descriptions of Models

Appendix 6.1 The IMAGE 2.2 ModelThe IMAGE modeling framework—the Integrated Modelto Assess the Global Environment—was originally devel-oped to study the causes and impacts of climate changewithin an integrated context. At the moment, however,IMAGE 2.2 is used to study a whole range of environmentaland global change problems, in particular in the realm ofland use change, atmospheric pollution, and climatechange. The model and its sub-models have been describedin detail in several publications (see, in particular, Alcamoet al. 1998; IMAGE-team 2001).

Model Structure and Data

In general terms, the IMAGE 2.2 framework describesglobal environmental change in terms of its cause-responsechain. Appendix Figure 6.1 provides an overview of thedifferent parts of the model.

The cause-response chains start with the main drivingforces—population and macroeconomic changes—thatdetermine energy and food consumption and production.Cooperation with a macroeconomic modeling team (CPB1999) working on a general equilibrium model ensures inseveral cases an economic underpinning of assumptionsmade. Next, a detailed description of the energy and foodconsumption and production are developed using theTIMER Global Energy Model and the AEM Food De-mand and Trade model (for the MA, the latter was replacedby a link to the IMPACT model). Both models accountfor various substitution processes, technology development,and trade.

The changes in production and demand for food andbiofuels (the latter are calculated in the energy model) haveimplications for land use, which is modeled in IMAGE ona 0.5 by 0.5 degree grid. Changes in both energy consump-tion and land use patterns give rise to emissions that areused to calculate changes in atmospheric concentration ofgreenhouse gases and some atmospheric pollutants such asnitrogen and sulfur oxides. Changes in concentration ofgreenhouse gases, ozone precursors, and species involved inaerosol formation comprise the basis for calculating climaticchange. Next, changes in climate are calculated as globalmean changes that are downscaled to the 0.5 by 0.5 degreegrids using patterns generated by general circulationmodels.

The Land-Cover Model of IMAGE simulates thechange in land use and land cover in each region driven bydemands for food (including crops, feed, and grass for ani-mal agriculture), timber, and biofuels, in addition tochanges in climate. The model distinguishes 14 natural andforest land cover types and 5 humanmade types. A cropmodule based on the FAO agro-ecological zones approachcomputes the spatially explicit yields of the different crop

PAGE 157

groups and grass and the areas used for their production, asdetermined by climate and soil quality (Alcamo et al. 1998).In case expansion of agricultural land is required, a rule-based ‘‘suitability map’’ determines which grid cells are se-lected. Conditions that enhance the suitability of a grid cellfor agricultural expansion are its potential crop yield (whichchanges over time as a result of climate change and technol-ogy development), its proximity to other agricultural areas,and its proximity to water bodies.

The Land-Cover Model also includes a modified versionof the BIOME model (Prentice et al. 1992) to computechanges in potential vegetation (the equilibrium vegetationthat should eventually develop under a given climate). Theshifts in vegetation zones, however, do not occur instanta-neously. In IMAGE 2.2, such dynamic adaptation is mod-eled explicitly according to the algorithms developed byVan Minnen et al. (2000). An important aspect of IMAGEis that it accounts for significant feedbacks within the sys-tem, such as temperature, precipitation, and atmosphericCO2 feedbacks on the selection of crop types and the mi-gration of ecosystems. This allows for calculating changes incrop and grass yields and, as a consequence, the location ofdifferent types of agriculture, changes in net primary pro-ductivity, and migration of natural ecosystems.

Application

The IMAGE model has been applied in several assessmentstudies worldwide, including work for IPCC and analysesfor UNEP’s Global Environment Outlook (UNEP 2002).For instance, the IMAGE team was one of the six modelsthat took part in the development of the scenarios ofIPCC’s Special Report on Emission Scenarios (IPCC2000a; de Vries et al. 2000; Kram et al. 2000). The modelhas also been used for a large number of studies that aim toidentify strategies that could mitigate climate change,mostly focusing on the role of technology or relevant tim-ing of action (e.g., van Vuuren and de Vries 2001). IMAGEalso contributed to European projects, including the regu-larly published State of the Environment report on Europe ofthe EU/European Environment Agency and work for theDirectorate-General for the Environment of the EuropeanCommission. Recently, the geographic scale of IMAGE wasfurther disaggregated to the country level in Europe, usingthe model in a large project for land use change scenarios inEurope. In addition, on a project basis the capabilities of themodel to describe the nitrogen cycle are improved. In re-cent years, the links to biodiversity modeling have also beenimproving.

Uncertainty

As a global Integrated Assessment Model, the focus ofIMAGE is on large-scale, mostly first-order drivers of globalenvironmental change. This obviously introduces some im-portant limitations to its results, and in particular how tointerpret their accuracy and uncertainty. An importantmethod to handle some of the uncertainties is by using ascenario approach. A large number of relationships andmodel drivers that are currently not known or that dependon human decisions are varied in these scenarios to explore

................. 11411$ $CH6 10-27-05 08:42:19 PS


Appendix Figure 6.1. Structure of IMAGE 2.2 Model. Phoenix, WorldScan, and TIMER are submodels of the IMAGE 2.2 model.

the uncertainties involved in them (see IMAGE-team2001). For the energy model, in 2001 a separate project wasperformed to evaluate the uncertainties in the energy modelusing both quantitative and qualitative techniques. Throughthis analysis we identified that the model’s most importantuncertainties had to do with assumptions for technologicalimprovement in the energy system and how human activi-ties are translated into a demand for energy (includinghuman lifestyles, economic sector change, and energy effi-ciency).

The carbon cycle model has also been used in a sensitiv-ity analysis to assess uncertainties in carbon cycle modelingin general (Leemans et al. 2002). Finally, a main uncertaintyin IMAGE’s climate model has to do with the ‘‘climate sen-sitivity’’ (that is, the response of global temperature com-puted by the model to changes in atmospheric greenhousegas concentrations) and the regional patterns of changedtemperature and precipitation. IMAGE 2.2 has actuallybeen set up in such a way that these variables can be easilyvaried on the basis of more scientifically detailed models,

PAGE 158

and a separate CD-ROM has been published indicating theuncertainties in these relationships in detail. To summarize,in terms of the scheme discussed in this chapter on the cer-tainty of different theories, most of IMAGE would need togo into the category of established but incomplete knowl-edge.

Appendix Tables 6.1 and 6.2 give an overview of themain sources of uncertainty in the IMAGE model.

Appendix 6.2 The IMPACT ModelIMPACT—the International Model for Policy Analysis ofAgricultural Commodities and Trade—was developed inthe early 1990s as a response to concerns about a lack ofvision and consensus regarding the actions required to feedthe world in the future, reduce poverty, and protect thenatural resource base.


IMPACT is a representation of a competitive world agricul-tural market for 32 crop and livestock commodities, includ-

................. 11411$ $CH6 10-27-05 08:42:22 PS


Appendix Table 6.1. Overview of Major Uncertainties in theIMAGE Model

Model Component

EnvironmentalEnergy and Related Land Use and System and

Uncertainty Emissions Land Cover Climate Change

Model integration in larger rule-based scheme for allo-structure economy, feedbacks algorithm for cating carbon

allocating land pools in the car-dynamic formulationuse bon cycle modelin energy model

(learning by doing,multinomial logit)

Parameter resource assump- biome model climate sensitivitytions parameter climate change

settinglearning parameters patternsCO2 fertilization multipliers in car-

bon model (im-pact of climateand carbon cycle)

Driving population assumptionsforce economic assumptions

assumptions ontechnology change

lifestyle, materialintensity, diets

environmental policies

agricultural productionlevels (from IMPACT)

Initial emissions in base initial land use / climate in basecondition year (1995) land cover year (average

maps global values andhistoric energy usemaps)historic land

use data (FAO)

Model downscalingoperation method

ing all cereals, soybeans, roots and tubers, meats, milk, eggs,oils, oilcakes and meals, sugar and sweeteners, fruits andvegetables, and fish. It is specified as a set of 43 country orregional sub-models, within each of which supply, demand,and prices for agricultural commodities are determined.The country and regional agricultural sub-models are linkedthrough trade, a specification that highlights the interde-pendence of countries and commodities in global agricul-tural markets.

The model uses a system of supply and demand elasticit-ies incorporated into a series of linear and nonlinear equa-tions to approximate the underlying production anddemand functions. World agricultural commodity prices aredetermined annually at levels that clear international mar-kets. Demand is a function of prices, income, and popula-tion growth. Growth in crop production in each country isdetermined by crop prices and the rate of productivitygrowth. (See Appendix Figure 6.2.) The model is writtenin the General Algebraic Modeling System programming

PAGE 159

Appendix Table 6.2. Level of Confidence in Different Types ofScenario Calculations from IMAGE

High Established but Well-establishedincomplete energy modeling andclimate impacts on scenariosagriculture and biomes

carbon cycle

Level of Low Speculative CompetingAgreement explanationsgrid-level changes in

driving forces global climate change—including estimates ofimpacts of land degra-uncertaintydationlocal climate change

land use change

Low High

Amount of Evidence(theory, observation, model outputs)

language and makes use of the Gauss-Seidel algorithm. Thisprocedure minimizes the sum of net trade at the interna-tional level and seeks a world market price for a commoditythat satisfies market-clearing conditions.

IMPACT generates annual projections for crop area,yield, and production; demand for food, feed, and otheruses; and prices and trade. It also generates projections forlivestock numbers, yield, production, demand, prices, andtrade. The current base year is 1997 (three-year average of1996–98) and the model incorporates data from FAOSTAT(FAO 2000); commodity, income, and population data andprojections from the World Bank (World Bank 1998,2000a, 2000b) and the UN (UN 1998); a system of supplyand demand elasticities from literature reviews and expertestimates; and rates for malnutrition from ACC/SCN(1996)/WHO (1997) and calorie-child malnutrition rela-tionships developed by Smith and Haddad (2000). For MApurposes, the projections period was updated from 1997–2025 to 2100. Additional updates on drivers and parametersare described in Chapter 9.

Application

IMPACT has been applied to a wide variety of contextsfor medium- and long-term policy analysis of global foodmarkets. Applications include commodity-specific analyses(for example, for roots and tubers (Scott et al. 2000), forlivestock (Delgado et al. 1999), and for fisheries (Delgadoet al. 2003)) and regional analyses (for example, on the con-sequences of the Asian financial crisis (Rosegrant and Rin-gler 2000)). In 2002 a separate IMPACT-WATER modelwas developed that incorporates the implications of wateravailability and nonagricultural water demands on food se-curity and global food markets (Rosegrant et al. 2002).

Uncertainty

As IMPACT does not contain equations with known statis-tical properties, formal uncertainty tests cannot be carried

................. 11411$ $CH6 10-27-05 08:42:22 PS


Appendix Figure 6.2. Structure of IMPACT Model

out. However, the robustness of results has been tested andthe sensitivity of results with respect to the various driversin the model have been carried out through numerous sce-nario analyses, as described in Rosegrant et al. (2001) andthe other IMPACT publications cited in this assessment.The drivers associated with uncertainty that have importantimplications for model outcomes include population andincome growth, as well as drivers affecting area and yieldand livestock numbers and slaughtered weight growth.

Appendix Tables 6.3 and 6.4 summarize the points re-lated to uncertainty in the model, based on the level ofagreement and amount of evidence.

Appendix 6.3 The WaterGAP ModelThe principal instrument used for the global analysis ofwater withdrawals, availability, and stress is the WaterGAPmodel—Water Global Assessment and Prognosis—developedat the Center for Environmental Systems Research of theUniversity of Kassel in cooperation with the National Insti-tute of Public Health and the Environment of the Nether-lands. WaterGAP is currently the only model with globalcoverage that computes both water use and availability onthe river basin scale.


The aim of WaterGAP is to provide a basis both for anassessment of current water resources and water use and foran integrated perspective of the impacts of global change on

PAGE 160

the water sector. WaterGAP has two main components, aglobal hydrology model and a global water use model. (SeeAppendix Figure 6.3.)

The global hydrology model simulates the characteristicmacroscale behavior of the terrestrial water cycle to estimatewater availability; in this context, we define ‘‘water avail-ability’’ as the total river discharge, which is the combinedsurface runoff and groundwater recharge. The model calcu-lates discharge based on the computation of daily water bal-ances of the soil and canopy. A water balance is alsoperformed for open waters, and river flow is routed via aglobal flow routing scheme. In a standard global run, thedischarge of approximately 10,500 rivers is computed. Theglobal hydrology model provides a testable method for tak-ing into account the effects of climate and land cover onrunoff.

The global water use model consists of three main sub-models that compute water use for households, industry,and irrigation in 150 countries. Beside the water withdrawal(total volume of water that is abstracted from surface orgroundwater sources), the consumptive water use (waterthat is actually used and not returned to the water cycle asreturn flow) is quantified. Both water availability and wateruse computations cover the entire land surface of the globeexcept Antarctica (spatial resolution 0.5�—that is, 66,896grid cells). A global drainage direction map with a 0.5� spa-tial resolution allows for drainage basins to be chosen flexi-bly; this permits the analysis of the water resources situationin all large drainage basins worldwide. For a more detailed

................. 11411$ $CH6 10-27-05 08:42:25 PS


Appendix Table 6.3. Overview of Major Uncertainties in theIMPACT Model

ModelComponent Uncertainty

Model based on partial equilibrium theory (equilibrium between de-structure mand and supply of all commodities and production factors)

underlying sources of growth in area/numbers and produc-tivity

structure of supply and demand functions and underlyingelasticities, complementary and substitution of factor inputs.

Parameters Input parametersbase year (three-year centered moving averages for) area,yield, production, numbers for 32 agricultural commoditiesand 43 countries and regions

elasticities underlying the country and regional demand andsupply functions

commodity prices

driving forces

Output parametersannual levels of food supply, demand, trade, internationalfood prices, calorie availability, and share and number ofmalnourished children

Driving force Economic and demographic driversincome growth (GDP)

population growth

Technological, management, and infrastructural driversproductivity growth (including management research, con-ventional plant breeding)

area and irrigated area growth

livestock feed ratios

Policy driversincluding commodity price policy as defined by taxes andsubsidies on commodities, drivers affecting child malnutri-tion, and food demand preferences

Initial baseline—three-year average centered on 1997 of all inputcondition parameters and assumptions for driving forces

Model —operation

description of the model, see Alcamo et al. (2000, 2003a,2003b), Alcamo (2001), and Döll et al. (2003).

Application

Results from the model have been used in many nationaland international studies, including the World Water As-sessment, the International Dialogue on Climate and Water,UNEP’s Global Environmental Outlook, and the WorldWater Vision scenarios disseminated by the World WaterCommission.

Uncertainty

Appendix Tables 6.5 and 6.6 summarize some of the mostimportant sources of uncertainty in WaterGAP calculations.

PAGE 161

Appendix Table 6.4. Level of Confidence in Different Types ofScenario Calculations from IMPACT

High Established but Well-establishedincomplete changes in consumptionprojections of area patterns and food

demandprojections of irrigatedarea, yield

projections of livestocknumbers, production

number ofmalnourished children

calorie availability

Level of

Low Speculative Competing

Agreement/

explanations

Assessment

projections of commod-ity prices

commodity trade

Low High

Amount of Evidence(theory, observations, model outputs)

Alcamo et al. (2003b) found that the magnitude of uncer-tainty of model calculations was very spatially dependent.

Appendix 6.4 The Asia-Pacific Integrated ModelAIM—the Asia-Pacific Integrated Model—is a large-scalecomputer simulation model developed by the National In-stitute for Environmental Studies in collaboration withKyoto University and several research institutes in the Asia-Pacific region. AIM assesses policy options for stabilizingglobal climate, especially in the Asia-Pacific region, withobjectives of reducing greenhouse gas emissions and avoid-ing the impacts of climate change. Modelers and policy-makers have recognized that climate change problems haveto be solved in conjunction with other policy objectives,such as economic development and environmental conser-vation. The AIM model has thus been extended to takeinto account a range of environmental problems, such asecosystem degradation and waste disposal, in a comprehen-sive way.


AIM/Water estimates country-wise water use (withdrawaland consumption in agricultural, industrial, and domesticsector), country-wise renewable water resource, spatial dis-tribution of water use and renewable water resources withresolution of 2.5� � 2.5�, and basin-wise water stress index.Appendix Figure 6.4 presents an overview over the maincomponents of the AIM/Water sub-model. Future scenar-ios of population, GDP, technological improvements, andhistorical trends of population with access to water supplyare the basic inputs used in the estimation of water use. Thecountry-wise water use is then disaggregated to grid cells inproportion to the spatial densities of population and crop-

................. 11411$ $CH6 10-27-05 08:42:26 PS


Appendix Figure 6.3. Structure of WaterGAP Model. Top: overview of main components. Bottom: WaterGAP hydrological model. (Döll etal. 2003)

PAGE 162................. 11411$ $CH6 10-27-05 08:42:29 PS


Appendix Table 6.5. Overview of Major Uncertainties inWaterGAP Model

ModelComponent Uncertainty

Model evapotranspirationstructure river transport time

snowmelt mechanism

Parameters watershed calibration parameter

parameter for allocating total discharge to surface and sub-surface flow

Driving force local precipitation inputs, frequency of rain days

Initial current direction of flows in flat and wetland areascondition grid resolution of current water withdrawals

Model downscaling method for country-scale domestic and indus-operation trial withdrawals; interpolation of climate data

Appendix Table 6.6. Level of Confidence in Different Types ofScenario Calculations from WaterGAP

High Established but Well-establishedincomplete annual withdrawals inwater stress industrialized countries

annual withdrawals in annual water availabilitydeveloping countries where there are long-

term hydrologic gaugesannual water(about 50% of the areaavailability in areasof Earth)without long-term

hydrologic gaugesLevel of

Low Speculative CompetingAgreement

explanationsreturn flowswater quality

freshwater biodiversity(WaterGAP contributesto these calculations)

Low High

Amount of Evidence(theory, observations, model outputs)

land. The change in renewable water resource is estimatedby considering future climate change as input data. In orderto obtain a water stress index, water withdrawal and renew-able water resources are compared in each river basin. Seealso Harasawa et al. (2002) for a more detailed description.

AIM/Agriculture estimates potential crop productivityof rice, wheat, and maize with the spatial resolution of 0.5�� 0.5�. Climatic factors are then taken as inputs to simulatenet accumulation of biomass through photosynthesis andrespiration. They include monthly temperature, cloudiness,precipitation, vapor pressure, and wind speed. The physicaland chemical properties of soil such as soil texture and soilslope are also considered in estimating suitability for agricul-ture (Takahashi et al. 1997).

PAGE 163

AIM/Ecosystem is a global computable general equilib-rium model. The model structure is shown in AppendixFigure 6.5. It is an economic model with 15 regions and15 sectors. The model has been developed for the period1997–2100 with recursive dynamics. Prices and activitiesare calculated in order to balance demand and supply for allcommodities and production factors. AIM/Ecosystemmodel is linked to AIM/Agriculture model in terms of landproductivity changes resulting from climate change. Themain drivers of these dynamics are population, productioninvestment, and technology improvement. In this model,various environmental issues such as deforestation and airpollution are included. These interact with the economythrough provision of resources and maintenance and degra-dation of the environment. This model thus consistentlyestimates economic activities such as GDP and primary en-ergy supply, the related environmental load such as air pol-lution, and environmental protection activities such asinvestments in desulfurization technologies (Masui et al.forthcoming).

Application

The AIM model has been used in the development of oneof the marker scenarios for IPCC/SRES. The extendedversion was used for UNEP’s GEO3 report. Long-termscenarios of environmental factors quantified using AIM/Water, AIM/Agriculture, and AIM/Ecosystem have beenused for the MA.

Uncertainty

The AIM models are based on a deterministic framework.The time scale of each model is more than 100 years. Thislong-term framework stresses theoretical consistency. It ispreferable to models that use past trends because those arenot suitable for studying long-term dynamics. Our ap-proach to uncertainty is not to evaluate each parameter orfunction individually but to assess the robust options or pol-icies derived from the various simulation results.

Appendix Tables 6.7 and 6.8 summarize the uncertaint-ies in each model.

An option or sets of options related to the elements inthis table are introduced to the models and simulated underthe different scenarios. When the options always producesimilar results even in different scenarios, they are regardedas robust.

Appendix 6.5 Ecopath with EcosimEwE—Ecopath with Ecosim—is an ecological modelingsoftware suite for personal computers; some components ofEwE have been under development for nearly two decades.The approach is thoroughly documented in the scientificliterature, with over 100 ecosystems models developed todate (see www.ecopath.org). EwE uses two main compo-nents: Ecopath, a static, mass-balanced snapshot of the sys-tem, and Ecosim, a time dynamic simulation module forpolicy exploration that is based on an Ecopath model.

................. 11411$ $CH6 10-27-05 08:42:30 PS


Appendix Figure 6.4. Structure of AIM/Water Model

Appendix Figure 6.5. Structure of AIM/Ecosystem Model


The foundation of the marine fisheries calculations is anEcopath model (Christensen and Pauly 1992; Pauly et al.2000). The model creates a static, mass-balanced snapshotof the resources in an ecosystem and their trophic interac-tions, represented by trophically linked biomass ‘‘pools.’’The biomass pools consist of a single species or of speciesgroups representing ecological guilds. Pools may be furthersplit into ontogenetic (juvenile/adult) groups that can thenbe linked together in Ecosim.

Ecopath data requirements of biomass estimates, totalmortality estimates, consumption estimates, diet composi-tions, and fishery catches are relatively simple and are gen-erally available from stock assessment, ecological studies, or

PAGE 164

the literature. The parameterization of Ecopath is based onsatisfying two key equations: the production of fish and theconservation of matter. In general, Ecopath requires inputof three of the following four parameters: biomass, pro-duction/biomass ratio (or total mortality), consumption/biomass ratio, and ecotrophic efficiency for each of thefunctional groups in a model. Christensen and Walters(2004) detail the methods used and the capabilities and pit-falls of this approach.

Ecosim has a dynamic simulation capability at the eco-system level, with key initial parameters from the base Eco-path model. (See Appendix Figures 6.6 and 6.7.) The keycomputational aspects are:• use of mass-balance results (from Ecopath) for parameter

estimation;

................. 11411$ $CH6 10-27-05 08:42:38 PS


Appendix Table 6.7. Overview of Major Uncertainties in AIM Model

Uncertainty

AIM/Water

Model Component Water Withdrawal Model Renewable Water Resource Model

Model structure assumption that the spatial pattern of population and land use choice of method for estimating potential evapotranspirationwill not change in future

Parameter assumption regarding water use efficiency improvement in assumption of the model parameter for relating actual andeach sector potential evapotranspiration

assumption of the model parameter for estimating urbaniza-tion ratio

Driving force population climate projected by GCM

increasing trend in population with access to water

degree of economic activity

Initial condition error in the estimated sectoral water withdrawal in the base error in the estimated renewable water resource in the baseyear year

error in the observed climate data

Model operation procedure to develop climate scenario from GCM result

AIM/Agriculture

Model structure choice of method for estimating photosynthesis ratio

Parameter assumption of the model parameter which describes crop growth characteristics

Driving force future climate projected by GCM

Initial condition error in the observed climate data

error in the soil data

Model operation procedure to develop climate scenario from GCM result

AIM/Ecosystem

Model structure based on the general equilibrium theory (equilibrium between demand and supply of all commodities and production factors)

investment function in each period

structures of production function and demand function: especially elasticity of substitution among the inputs

Parameter change of preference

relationship between cost and performance in pollution reduction

Driving for

Documents

Chapter 6 Methodology for Developing the MA Scenarios · 2011. 6. 12. · Chapter 6 Methodology for Developing the MA Scenarios Coordinating Lead Authors: Joseph Alcamo, Detlef van