Forthcoming in
Proceedings of the 1994 A&WMA
Global Climate Change Conference: Phoenix April 5-8,
Air & Waste Management Association, Pittsburgh, 1994.
The Policy Evaluation Framework (PEF) is a decision analysis tool that enables decision makers to continuously formulate policies that take into account the existing uncertainties, and to refine policies as new scientific information is developed. It is designed to provide a framework for integrating and evaluating the best available information from the diverse elements that influence climate policy. PEF encourages exploration of the policy implications of alternative technological, economic, physical, and biological assumptions and scenarios .
PEF integrates deterministic parametric models of physical, biological, and economic systems with a flexible decision tree system. The deterministic models represent greenhouse gas emissions, atmospheric accumulation of these gases, global and regional climate changes, ecosystem impacts, economic impacts, and mitigation and adaptation options. The decision tree system captures the key scientific and economic uncertainties, and reflects the wide range of possible outcomes of alternative policy actions. The framework contains considerable flexibility to allow a wide range of scientific and economic assumptions or scenarios to be represented and explored
A key feature of PEF is its capability to address both mitigation policies and investments in anticipatory adaptation to protect ecological and economic systems, as well as interactions among such options. PEF's time structure allows issues related to the timing and flexibility of alternatives to be evaluated, while the decision tree structure facilitates examining questions involving the value of information, contingent actions, and probabilistic representations .
This paper is intended to introduce PEF to the global climate policy community. The paper provides an overview of the structure, modules, and capabilities of PEF, and discusses selected results from an initial set of illustrative applications. Fuller descriptions of the PEF methodology and results can be found in the EPA's forthcoming Integrated Assessment of Global Climate Change [1] and The Global Climate Policy Evaluation Framework [2].
PEF's structure combines two components: (1) a deterministic model that describes the physical, biological, and economic impacts of greenhouse gas emissions; and (2) a decision tree system that organizes relevant information about the decisions and uncertainties. These components support the analytical tools that make it possible to evaluate policy alternatives under uncertainty. Figure 1 shows the relationship between PEF's deterministic model and its decision tree. For a given set of assumptions, a specific physical and economic scenario, and a particular policy decision, the deterministic model calculates the resulting physical and economic impacts. The decision tree executes the deterministic model multiple times to evaluate the effect of various uncertainties and decisions.
The scope of the framework includes both economic and ecosystem impacts, and both near-term and long-term adaptation and mitigation decisions. The near-term decisions are the primary focus of the model; the long-term decisions are included to provide a more realistic model of the consequences of near-term decisions, and to investigate the implications of the timing of actions.
Designing the deterministic model required tradeoffs between the detail of physical and economic representations and the simplicity needed for a practical analytical tool that can be used to rapidly evaluate an array of policy options and scenarios. This type of structure provides transparency, which also helps policy makers understand and interpret the results. For most processes, the model represents the physical relationships with parametric equations that can be calibrated to the results of larger, more detailed models. This approach provides the flexibility to represent many differing assumptions, opinions, or results. As PEF evolves, various relationships in the deterministic model will be updated to reflect new information.
The deterministic model uses reduced form models where it is appropriate to do so--that is, where the key aspects of the underlying phenomena are relatively well understood. The atmospheric process and global climate relationships, for example, use this approach. Where there is not yet a good understanding of the underlying processes, or where it is not reasonable to represent these processes with a reduced form model, the deterministic model uses structured functional forms. The economic impacts and ecosystem impacts relationships, for example, use this approach. All modules include enough parametric flexibility to represent, or to bound, any reasonable scenario. This capability allows sensitivity analyses to help prioritize refinements of the model. Each module is described below.
An individual mitigation alternative includes the target reductions in U. S. and rest-of-world emissions over time for each greenhouse gas. The realized reductions, as a percentage of the target reductions, are specified separately for U. S. and rest-of-world emissions. Each alternative may also include the cost of the alternative, or the cost can be estimated using an endogenous model in PEF. Although PEF can estimate reductions in any greenhouse gas, the cost model currently estimates only the cost of carbon dioxide reductions. If an alternative scenario includes reductions in other gases, the cost of those reductions must be specified. This structure makes it possible to incorporate the cost and efficacy of mitigation from other models or studies.
A particular adaptation alternative includes four components: the dollar investment in economic adaptation by sector over time; the dollar investment in protecting ecosystems (ecosystem adaptation) by ecosystem type over time; the allocation of each sector's economic adaptation investment across the regions; and the allocation of each ecosystem type's adaptation investment across the regions. Adaptation alternatives may differ both in amount and timing of spending. In PEF, the cumulative investment, less depreciation, can reduce the impact of climate change on the economy and on ecosystems. The spending is expressed in annual rates of spending in billions of constant dollars. The parameters that describe the effect of adaptation on the impacts and the rate at which the adaptation investment depreciates are specified separately.
Given a particular mitigation alternative, with its associated emissions reductions, the emission modules estimate the annual emissions from U. S. and rest-of-world sources of each gas. Emissions are built up from the baseline emissions scenario, the target reduction due to mitigation, and a parameter representing the efficacy of mitigation.
PEF provides several alternatives in selecting the gases to include in an analysis. The model includes specialized atmospheric process relationships for carbon dioxide, methane, and nitrous oxide. Carbon dioxide can be modeled alone or with these other gases. Beyond these, PEF can include any other gases as well, such as CFCs or HCFCs, that fit into the standard relationships described below. The mitigation alternatives may address any of these gases.
The module uses the Maier-Reimer and Hasselmann model, as described by T. M. L. Wigley [3], for future carbon dioxide emissions, while it uses a single-sink decay model for past carbon dioxide emissions and for all emissions of other gases. The module uses the IPCC radiative forcing relationships [4] to calculate radiative forcing directly from the concentrations of greenhouse gases. Following the IPCC [4], it assumes a linear relationship between changes in radiative forcing and the equilibrium temperature. It uses a parametric model to describe the relationship between equilibrium and realized temperature, which treats each increase in equilibrium temperature as a pulse that becomes realized over time. A parametric model, based on STUGE [6], is used to calculate the change in global sea level.
The choice of regional climate variables is driven by the requirements of the damage functions. The model can be easily expanded to meet the requirements of the economic damage functions and the ecosystem failure relationships (described below) as they evolve.
General circulation models still offer little insight into the nature of regional changes in temperature and precipitation, runoff, and soil moisture. PEF therefore provides flexible general relationships for regional changes, so that the sensitivity of analytical results to different assumptions about regional climate effects can be explored. The existing structure includes separate variables for each of the regional climate indicators to allow analyses of the influence of an individual indicator and the value of resolving its uncertainty.
Economic impacts are those that would typically appear in GDP calculations or other monetized economic measures. In general, the economic impacts include effects on activities that take place within economic markets, or on goods that are exchanged on economic markets. The model also places a monetary value on the ecosystem impacts, wherever possible. These generally are effects that are not manifested in markets.
PEF allows spending on both ecosystem and economic adaptation to influence the economic impacts. Spending is accumulated into a stock of investment, which depreciates at a rate determined by an input variable. Adaptation spending targeted at ecosystems are also allowed to affect the economic sector through cross-effect parameters. The cross effects of ecosystem adaptation may increase or decrease the economic impacts. The savings due to economic adaptation are a function of both the adjusted stock of adaptation investment and the impacts before adaptation. Both the depreciation rates and the efficacy of adaptation may vary across sectors.
PEF assumes that the total impacts to a sector can be estimated by adding the impacts due to temperature change, precipitation change, and the other climate indicators. It could, however, be modified to include interaction terms between climate indicators. Damages may be estimated using any combination of linear functions, power functions of arbitrary power, logistic functions, and step functions. PEF uses a separate set of parameters for each combination of sector, region, and climate indicator.
The model does not address reactive adaptation explicitly. Instead, the damages in each sector are meant to represent the impacts of climate change, net of the impacts of reactive adaptation. Anticipatory adaptation may reduce damages either by reducing the sensitivity of an economic sector to climate change or by reducing the need for reactive adaptation with more effective anticipatory adaptation.
In estimating the fractional loss of ecosystems, the model assumes that changes in climate make some fraction of the ecosystem valueless. This could represent total failure of some fraction of land that had been populated by a particular type of ecosystem, or it could represent a partial loss of all of the land of that ecosystem type within the region.
Once the model has estimated a fractional loss, it then uses a valuation function to estimate the value per unit area by type and region. As with the economic impact module, any combination of linear, arbitrary power, logistic, and step functions can be used to estimate the ecosystem damages. This would allow, for example, the value per acre of a particular type in a given region to increase sharply as the ecosystem becomes scarce. Similar to the economic impacts module, adaptation investments may be made to reduce the amount of ecosystem damage. Ecosystem adaptation is modeled in the same way as economic adaptation, with a stock of investment, depreciation, efficacy of adaptation, and cross effects from economic sector investments.
In the final step, the model combines the area lost, the valuation per acre, and the level of economic and ecosystem adaptation into an annual impact, measured in dollars, of ecosystem or nonmonetary damages. This version of the model, while crude, can perform "what-ifs" analyses regarding non-market impacts to provide useful policy insights. Despite its simplistic nature, this model represents an improvement over those that ignore the potential non-market impacts of climate change.
Relating this to the levels described in the overview, the use of two decision periods is part of level two and is embedded in PEF's relationships. As such, it could be changed, but not on a regular basis. The beginning and end of the horizon, and the length of the time step, are part of level three and thus can be changed from run to run. Table 1 describes the specific timing assumptions embedded in the current version of PEF.
PEF provides a range of built-in analyses, which can be subdivided into deterministic and probabilistic analyses. Deterministic analyses investigate the preferred policy options under various climate and impact scenarios. Probabilistic analyses use the likelihoods of different outcomes to evaluate appropriate options under conditions of uncertainty. Concise descriptions of each of the analysis types are provided in Table 2.
The preferred alternatives are also strongly sensitive to the impacts scenarios. Figure 6 shows PEF results for appropriate mitigation and adaptation policies under combinations of economic sensitivity scenarios. (The scenarios are defined by the impacts, as a percentage of gross domestic product, that result from a 2. 5°C increase in realized temperature.) Although these results are preliminary, they suggest that uncertainties in the impacts resulting from a given level of climate change may be as important to policy decisions as the uncertainty in the extent of climate change itself.
Because mitigation and adaptation work through different mechanisms, they interact asymmetrically. Whereas investment in adaptation are directly targeted at the impacts, investments in mitigation reduce impacts indirectly through changes in the climate. Therefore, while potentially more expensive, adaptation could have larger effects on impacts more quickly. Thus, mitigation becomes less cost-effective when high levels of adaptation investment reduce the impacts, while adaptation is still effective in the presence of high levels of mitigation,
Figure 9 shows the preferred long-term emission reductions for different threshold scenarios. The results to date suggest that the preferred reduction in emissions is more sensitive to the time at which the threshold occurs than to the magnitude of the impacts.
PEF has been used to investigate the relationships between the key uncertainties and the available policy alternatives. Nothing hinders the adaptability of the model to new information in science or economics. Further efforts will focus on refining the assumptions and scenarios used, especially in developing sectorally disaggregated impact functions.
2. Chan, Nathan Y. et al.; The Global Climate Policy Evaluation Framework: A technical report prepared by Decision Focus Incorporated for the United States Environmental Protection Agency, forthcoming.
3. Wigley, T. M. L. "A Simple Inverse Carbon Cycle Model," Global Biogeochemical Cycles, December 1991.
4. Climate Change: The IPCC Scientificc Assessment; Houghton, J. T.; Jenkins, G. J.; and Ephraums, J. J. eds.; Cambridge University Press, Cambridge: 1990.
5. Climate Change 1992: The Supplementary Report to the IPCC Scientific Assessment; Houghton, J. T., Callander, B. A., and Vainey, S. K., eds.; Cambridge University Press, 1992.
6. Wigley, T.M.L.; Holt, T.; and Raper, S. C. B. STUGE: an Interactive Greenhouse Model. Climatic Research Unit, University of East Anglia, Norwich, U.K., Oct. 1991.
