A second reason for concern is the distribution of impacts among people and across regions. The impacts of climate change will not be distributed equally. Some individuals, sectors, systems, and regions will be less affected—or may even benefit; other individuals, sectors, systems, and regions may suffer significant losses. This pattern of relative benefits or losses is not likely to remain constant over time. It will be different with different magnitudes of climate change. Some regions may have gains only for certain changes in temperature and precipitation and not for others. As a result, some regions that may first see net benefits eventually may face losses as well as the climate continues to warm.

Research into the distribution of impacts of climate change is in its infancy, in large measure because this research poses several methodological challenges.

A first difficulty is synthesis—the need to reduce the complex pattern of individual impacts to a more tractable set of regional or sectoral indicators. The challenge is to identify a set of indicators that can summarize and make comparable the impacts in different regions, sectors, or systems in a meaningful way. A range of indicators and methods have been put forward. Many models use physical measures such as the number of people affected (e.g., Hoozemans et al., 1993), change in net primary productivity (NPP) (White et al., 1999), or the number of systems undergoing change (e.g., Alcamo et al., 1995).

The most widespread numeraire, however, is economic cost (Nordhaus, 1991, 1994a; Cline, 1992; Hohmeyer and Gaertner, 1992; Titus, 1992; Downing et al., 1995, 1996; Fankhauser, 1995; Tol, 1995; Mendelsohn and Neumann, 1999). This numeraire is particularly well-suited to measure market impacts—that is, impacts that are linked to market transactions and directly affect GDP (i.e., a country's national accounts). The costs of sea-level rise, for example, can be expressed as the capital cost of protection plus the economic value of land and structures at loss or at risk; agricultural impacts can be expressed as costs or benefits to producers and consumers, including the incremental costs of adaptation. Using a monetary numeraire to express nonmarket impacts such as effects on ecosystems or human health is more difficult. It is possible in principle, however. There is a broad and established literature on valuation theory and its application, including studies (mostly in a nonclimate change context) on the monetary value of lower mortality risk, ecosystems, quality of life, and so forth. However, economic valuation can be controversial and requires sophisticated analysis, which still is mostly lacking in a climate change context.

Physical metrics—such as NPP or percentage of systems affected—on the other hand, are best suited for natural systems. When they are applied to systems under human management, they suffer from being poorly linked to human welfare, the ultimate indicator of concern. Some researchers therefore recommend different numeraires for market impacts, mortality, ecosystems, quality of life, and equity (Schneider et al., 2000b). They recognize, however, that final comparisons across different numeraires nonetheless are required; they regard this as the job of policymakers, however.

Persistent knowledge gaps is a second source of difficulty. Distributional analysis depends heavily on the geographical details of climate change, but these details are one of the major uncertainties in the outputs of climate change models. This is particularly true for estimates of precipitation; for example, estimates of water-sector impact can vary widely depending on the choice of GCM.³ Uncertainties continue at the level of impact analysis. Despite a growing number of country-level case studies, our knowledge of local impacts is still too uneven and incomplete for a careful, detailed comparison across regions. Furthermore, differences in assumptions often make it difficult to compare case studies across countries. Only a few studies try to provide a coherent global picture on the basis of a uniform set of assumptions. The basis of most such global impact assessments tends to be studies undertaken in developed countries—often the United States—which are then extrapolated to other regions. Such extrapolation is difficult and will be successful only if regional circumstances are carefully taken into account, including differences in geography, level of development, value systems, and adaptive capacity. Not all analyses are equally careful in undertaking this task.

There are other shortcomings that affect the quality of analysis. Although our understanding of the vulnerability of developed countries is improving—at least with respect to market impacts—information about developing countries is quite limited. Nonmarket damages, indirect effects (e.g., the effect of changed agricultural output on the food-processing industry), the link between market and nonmarket effects (e.g., how the loss of ecosystem functions will affect GDP), and the sociopolitical implications of change also are still poorly understood. Uncertainty, transient effects (the impact of a changing rather than a changed and static climate), and the influence of climate variability are other factors that deserve more attention. Because of these knowledge gaps, distributional analysis has to rely on (difficult) expert judgment and extrapolation if it is to provide a comprehensive picture.

A third problem is adaptation. There has been substantial progress in the treatment of adaptation since the SAR, but adaptation is difficult to capture adequately in an impact assessment. Adaptation will entail complex behavioral, technological, and institutional adjustments at all levels of society, and the capacity to undertake them will vary considerably (see Chapter 18). Various approaches are used to model adaptation (e.g., spatial analogs, microeconomic modeling), but they are prone to systematic errors about its effectiveness. The standard approach used in coastal impact assessment and in many agricultural models is to include in the analysis a limited number of "prominent" but ultimately arbitrary adaptations. This underestimates adaptive capacity because many potentially effective adaptations are excluded (Tol et al., 1998). On the other hand, approaches that are based on analogs—such as the Ricardian approach used by, for example, Mendelsohn et al. (1994), Mendelsohn and Dinar (1998), and Darwin (1999)—probably overestimate adaptive capacity because they neglect the cost of transition and learning. This is especially true for cases in which adaptation in developed countries today is used as a proxy for worldwide adaptation to an uncertain future climate. Only a very few studies model adaptation as an optimization process in which agents trade off the costs and benefits of different adaptation options (Fankhauser, 1995; Yohe et al., 1995, 1996).

The analysis is further complicated by the strong link between adaptation and other socioeconomic trends. The world will change substantially in the future, and this will affect vulnerability to climate change. For example, a successful effort to roll back malaria (as promoted by the development community) could reduce the negative health effects of climate change. On the other hand, growing pressure on natural resources from unsustainable economic development is likely to exacerbate the impacts of climate change on natural systems. Even without explicit adaptation, impact assessments therefore depend on the "type" of socioeconomic development expected in the future. The sensitivity of estimates to such baseline trends can be strong enough in some cases to reverse the sign (i.e., a potentially negative impact can become positive under a suitable development path, or vice versa) (Mendelsohn and Neumann, 1999).

Despite the limits in knowledge, a few general patterns emerge with regard to the distribution of climate change impacts. These patterns are derived from general principles, observations of past vulnerabilities, and limited modeling studies.

IPCC Homepage