Thematic Guide to Integrated Assessment Modeling

[HOME] [PREVIOUS] [BOTTOM]

 

How to Select an Integrated Assessment Model

Essay prepared for SEDAC by:

Gary W. Yohe
Department of Economics
Wesleyan University
Middletown, CT. 06459

Contents

1. Introruction
2. A Schematic Portrait of Integrated Assessment Models
3. Choosing IAMs from the SEDAC Menu
4. The Energy Modeling Forum
5. References

Introduction

The short history of integrated assessment in global environmental change research is a perfect example of "letting many flowers grow"; or perhaps more appropriately in the words of Henry Jacoby of MIT, "different horses for different courses". Global change is such a large issue and the underlying uncertainties are so vast that no single model can handle every facet adequately. There were, at last count in the spring of 1996, more than twenty integrated assessment models (IAMs) designed to provide insight into some part of the problem or other. Some sacrifice sectoral and regional detail completely to offer researchers multiple portraits of the global challenge and the ability to explore the fundamental uncertainties. Others sacrifice some flexibility for increased detail in the modeling of various regions or particular sectors. Still others sacrifice even more flexibility to produce enormous amalgomations of detailed interconnected sectoral and regional models. Each approach has its own focus, and each has its own set of strengths and weaknesses. The best advice for a potential user of an IAM is to play into a model's strengths by picking the one whose design best matches your specific needs. To do otherwise is to risk committing errors of judgement and propagating false impressions of the problem.

Chapters in both the Report of Working Group III to the latest Scientific Assessment of the Intergovernmental Panel on Climate Change (Weyant et al. 1996) and Human Choices and Climate Change: A State-of-the-Art Report (Rotmans, Dowlatabadi, and Parson 1996) provide thoughtful reviews of the IAMs that were "up and running" by the beginning of 1995. Readers are referred to either for guidance in making their decisions about which model(s) to consider for their own specific needs. This short essay will offer a schematic portrait of the structure of those chapters even as it pays particular attention both to the models currently available through SEDAC and the Energy Modeling Forum results recorded there.

A Schematic Portrait of Integrated Assessment Models

Peter Wilcoxen, now at the Texas A&M University, presented what he termed the "Modeling Frontier" to a meeting hosted by the United Nations Environment Programme (UNEP) in 1991. The meeting had been called because Dr. Mostafa Tolba, Executive Director of UNEP, wanted to know the answer to the global change issue; and the assembled group needed to explain to the Director not only that they did not know the answer, but also why the answer would not be forthcoming in the near term. Wilcoxen's frontier was an obvious reference to the production possibility frontier that is generally offered as a first model of scarcity in the first week of any introductory economics course. It displayed qualitative measures of geographical scope (from local and national "up" to global) on the vertical axis and modeling detail (from simple to complex) along the horizontal dimension. A convex frontier drawn bowing away from the origin in this quadrant reflected modelers' limited ability to handle both dimensions simultaneously. The point was that only an omniscient deity could create a model capable of support locally and sectorally disaggregated coverage of the entire global change problem. The rest of us are faced with trading modeling detail against geographical coverage.

Choosing IA Models from the SEDAC Menu

Cast in the context of Wilcoxen's diagram, the choice of which model(s) to use can be viewed as a constrained maximization problem--making the best choice subject to the constraint of what is possible and available. Users can, in particular, approach the choice by judging their needs against the modeling frontier as defined by existing models and making their own determination of the relative subjective value of complexity and scope. There are, however, several frontiers to consider. One contrasts scope with complexity in treating the economic and social drivers of change. Another casts scope against complexity in dealing with the economic and social effects of change, on the one hand, and mitigating policy on the other. A third weighs the same tradeoff in the depiction of natural scientific linkages, impacts, and adaptation. And finally, each tradeoff must be viewed in light of its explicit and/or implicit treatment of uncertainty; is the model deterministic, or does it accommodate decision-making under uncertainty.

SEDAC now offers access to results drawn from four separate models, and each can be described in terms that locate it against these four tradeoffs. ICAM2, for example, runs relatively complex greenhouse gas emissions trajectories drawn from relatively simple models of the economies of seven global regions through complex climate and atmospheric chemistry models to produce vast arrays of physical impacts and associated matrices of economic and feedback effects (Dowlatabadi and Morgan 1993; Dowlatabadi, Goulder, and Kopp 1994). The ICAM structure is, in fact, designed to incorporate uncertainty explicitly into each run so that adaptive decisions are made by actors who are not equipped with perfect information. ICAM therefore produces distributions of outcomes that rely to a large degree on "who knows what, and when?"; and no two runs will necessarily give the same results. Researchers who are looking to evaluate mitigation strategies will generally be disappointed with ICAM2 if they try to use it to weigh their relative costs and benefits in an optimization exercise, but researchers who are interested in ecosystem and economic effects and the adaptive decisions that influence them will be pleased with its versatility. Indeed, Dowlatabadi and Morgan would be happy to assist in the incorporation of new impact modules into ICAM's integrated structures.

DICE, by way of contrast, is a global model that pushes greenhouse gas emissions derived from a moderately complex and very aggregate economic model (driven by population and labor productivity and dampened by a simple control cost feedback) through a simple representation of atmospheric and climate interactions and into a single global damage function (Nordhaus 1992, 1994). It is a full-blown dynamic optimization model that has been designed to compute optimal emissions trajectories even as it tracks greenhouse gas concentrations, associated changes in global mean temperature, control costs and damages. DICE can accommodate monte carlo simulation over a small set of key variables to produce distributions of outcomes, but each run assumes perfect information about what the future holds as it makes deterministic calculations of optimal mitigation. Subsequent versions not yet available on SEDAC have, however, included regional disaggregation of emissions and damages (the RICE Model by Nordhaus and Yang 1996) and imperfectly informed optimization (the PRICE Model designed by Kolstad).

Image 2.0 is also available on SEDAC (Alcamo et al. 1994). This is a model whose focus is almost entirely centered on complex atmospheric, climatic and physical effects and feedbacks distributed regionally. The model essentially takes emissions trajectories as given and virtually ignores economic impacts. It does, however, compensate by offering detailed, geographically distributed but nonetheless coordinated physical and ecologic effects across a wide spectrum of variables.

The fourth model available on SEDAC is dubbed MiniCAM (Edmonds et al. 1994; Edmonds, Wise, and MacCracken 1994). It is a simplified, straightforward pass-through representation of its more complicated antecedent - the GCAM model. Even so, it offers moderately complex representations of the economic drivers of emissions, relatively detailed representations of the atmospheric chemistry that drives the climate change, reasonable portraits of complex physical effects, and simple measures of market and nonmarket damages. Policy impacts can be assessed from MiniCAM, and optimal trajectories derived from GCAM can be viewed. It cannot, however, handle the explicit uncertainty analyses supported across 9 regions that can be produced by GCAM.

The choice at SEDAC is limited, at this point; but even casual review of the four offerings suggests that they cover much of Wilcoxen's modeling possibility space. In their collective treatment of economics, the SEDAC offerings include a global model, a model with seven regions, and a model derived from an antecedent with nine (going on 12) regions; and there is a regional variant of the global model available in the literature. Their handling of economic impact spans the same space, but their handling of physical impacts extends to detailed representations of almost micro-grid coverage of the globe. Finally, they include models that can support sensitivity and uncertainty analysis of deterministic outcomes and at least one model that explicitly builds uncertainty into its structure.

The Energy Modeling Forum

The Energy Modeling Forum (EMF), located at Stanford University, has organized 14 different modeling assessment and comparison exercises over more than two decades. The 12th and the 14th exercises of the Forum (EMF-12 and EMF-14) have been directed at topics that span much of energy's sphere of influence in global change. More specifically, EMF-12 investigated differences in modelers' estimates of the cost of reducing carbon emissions along 5 distinct control scenarios (Gaskins and Weyant 1993). EMF-14 broadened the focus to ponder the means by which researchers have looked to conduct integrated assessments of global change, to investigate why different approaches produce different results, to highlight (and try to fill) gaps in knowledge and understanding and to avoid giving even the pretext of offering a consensus view of the future. The motivation behind EMF-14 was simple. The Framework Convention of Climate Change points signatory nations toward stabilizing concentrations of greenhouse gases in the atmosphere at a level that would prevent undesirable anthropogenically induced effects on the climate system. Ultimately, however, each country's action will depend upon its assessment of the relative costs and benefits of policy intervention.

The results of EMF-12 were, therefore, not enough to offer complete insight into those judgments. The purposes of EMF-14 are (1) to compare the various integrated approaches to assess their usefulness in policy development and in setting climate change research priorities and (2) to recommend area for improvement in the assessment process. The standard EMF approach was followed, at least initially, in framing both the EMF-12 and EMF-14 exercises. Representatives of as many integrated assessment modeling teams as possible were invited to participate along with experts on each of the key individual components and linkages (carbon cycle, atmospheric chemistry, climate, energy-economics, physical impacts, valuation, etc...). Working with individuals involved in the development of climate policies and negotiations, participants ran mutually agreed-upon standardized scenarios so that key outputs could be compared. They then confronted a series of policy related problems employing scenarios of their own design. It was impossible for any group to complete every task. Participants tackled, instead, whichever subset of tasks was most tractable for their particular approach; and SEDAC offers access to the results.

Care needs to be taken in their interpretation. EMF-12 and EMF-14 results reflect, most accurately, dispersions of expert opinion about many of the most important indicators of anthropogenic climate change. They do not, however, reflect uncertainty. Indeed, the models that accommodate uncertainty in their analysis report only one set of results to EMF (median or mean or something else); and the dispersion of their results across the full range of uncertainty that they can model generally straddle the EMF ranges. In addition, the EMF results are not all independent. Comparison of modeler's choice and standardized reference cases suggest that individual modelers sometimes adjust their driving variables if they are too far from the "consensus range". It was, in fact, curious to note that early reports displayed wider dispersion across the standardized runs than the modeler's choice trajectories. Speculation about behavior aside, it is undeniable that many of the models are drawn from the same fundamental structures, either in their representations of the science or in their construction of the economic drivers of emissions.

References

Alcamo, J., G.J.J. Kreileman, M.S. Krol, and G. Zuidema. 1994. Modeling the global society-biosphere-climate system: Part 1: Model description and testing. In Image 2.0: Integrated Modeling of Global Climate Change, ed. J. Alcamo, 1-35. Dordrecht, The Netherlands: Kluwer Academic Publishers.

Dowlatabadi, H. and G. Morgan. 1993. A model framework for integrated studies of the climate problem. Energy Policy 21: 209-221.

Dowlatabadi, H., L. Goulder, and R. Kopp. 1994. Integrated Economic and Ecological Modeling for Public Policy Decision Making. RFF Discussion Paper 94-37. Washington, D.C: Resources for the Future.

Gaskins, D. and J. Weyant. 1993. Model comparisons of the costs of reducing carbon dioxide emissions. American Economic Review Papers and Proceedings 83, no. 2: 318-323.

Edmonds, J., H. Pitcher, N. Rosenberg, and T. Wigley. 1994. Design for the global change assessment model. In Proceedings of the International Workshop on Integrative Assessment of Mitigation, Impacts, and Adaptation to Climate Change. Laxenburg, Austria. IIASA.

Edmonds, J., M.A. Wise, and C.N. MacCracken. 1994. Advanced Energy Technologies and Climate Change: An Analysis Using the Global Change Assessment Model (GCAM). Richland, Wash.: Pacific Northwest Laboratory.

Nordhaus, W.D. 1994. Managing the Global Commons: The Economics of the Greenhouse Effect. Cambridge: MIT Press.

Nordhaus, W.D. 1992. An optimal transition path for controlling greenhouse gases. Science 258: 1315-1319.

Nordhaus, W.D., and Z. Yang. 1996. RICE: A Regional Dynamic General Equilibrium Model of Optimal Climate-Change Policy. Department of Economics, Yale University, New Haven, Conn.

Rotmans, Dowlatabadi, and Parson. 1996. Integrated assessment of climate change: Evaluation of methods and strategies. In Human Choices and Climate Change: A State-of-the-Art Report. Forthcoming.

Weyant, J., O. Davidson, H. Dowlatabadi, J. Edmonds, M. Grubb, R. Richels, J. Rotmans, P. Shukla, W. Cline, S. Fankhauser, and R. Tol. 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. Contribution of Working Group III to the Second Assessment Report of the Intergovernmental Panel on Climate Change (IPCC), eds. J. Bruce et al. Cambridge: Cambridge University Press. Forthcoming.

 

 

[SEDAC] [PREVIOUS] [TOP]

 

 

Suggested Citation

Center for International Earth Science Information Network (CIESIN). 1995. Thematic Guide to Integrated Assessment Modeling of Climate Change [online]. University Center, Mich.
CIESIN URL: http://sedac.ciesin.org/mva/iamcc.tg/TGHP.html

 

Our sponsors: