Climate variability on seasonal-to-interannual time scales has a discernible effect on ecological systems and human welfare in different parts of the world. Existing and new methods of prediction offer the possibility of real, although imperfect, skill in predicting aspects of the climate several months to a year or more in advance in a number of geographical regions. As Chapter 3 shows, humanity has developed a variety of techniques for coping with climatic variations. The techniques actually employed vary across regions and types of activity and with the economic, social, cultural, institutional, and technological characteristics of sectors, regions, and users, and their availability and usefulness changes over time. These coping techniques shape the human consequences of climatic variations and the potential usefulness of climate forecast information.
The introduction of skillful predictive information into a social system adapted to unpredictable climate variability introduces new problems and opportunities. Chapter 4 shows that the way actors respond to new predictive information depends on their perceptions of its value, importance, and accuracy; their prior expectations about climate; the institutional structures and constraints in which they operate; and other factors. Their perceptions can be influenced by education and by the mode and manner in which forecast information is conveyed to them. New information does not benefit all recipients equally, and some may even be worse off as a result of improved forecasts, depending on their situations and the specifics of the forecast and its dissemination, not all of
which are understood. As Chapters 4 and 5 show, there are techniques by which the utility of forecasts can be improved and methods by which the utility of the forecasts and their effects on recipients can be evaluated.
Useful applications of climate information to societal problems are beginning to be made, mostly in a haphazard and disorganized manner. Nevertheless, the practical potential of seasonal-to-interannual climate forecasts to produce socially beneficial information is beginning to become apparent. To realize fully this potential, it is necessary to conduct systematic investigations in pursuit of two goals:
This chapter summarizes the panel's findings and outlines a series of scientific questions, the investigation of which will help society approach these goals.
Climate Forecasting and Its Uses
Coping with Seasonal-to-Interannual Climate Variation
The effects of climatic events and climate forecasts on human populations are shaped by the coping strategies that people have developed over long periods of time for living in variable climates. To make forecasts optimally useful, it is necessary to understand these coping systems. Specifically:
The Potential of Climate Forecasts
Although climate forecast information has great potential social value, its actual value may fall short of the potential for many reasons. In addition, information that is valuable for some purposes or for some recipients may not be for others. Much depends on how well forecast information is matched to users' needs, structured in accord with their modes of understanding, and delivered through systems that are effective for particular types of recipients. Responses to past forecasts also affect the use of new ones. Specifically:
Measuring the Consequences of Climate Variability and Forecasts
Scientific capability to measure and model the effects of seasonal-to-interannual climatic variability is well developed in some sectors (e.g., agriculture, water resources) and only beginning to be developed in others (e.g., human health, environmental amenities). Scientific capability to judge the value of climate forecasts is in its infancy.
We propose a program of research addressed to the ultimate goals of understanding and increasing the value of seasonal-to-interannual climate forecasts. Because this field of research is so new, it makes little sense to be highly prescriptive. Rather, we have identified a series of scientific questions that can provide programmatic guidance. The questions fall into three broad categories: research on the potential benefits of climate forecast information, on improved dissemination of forecast information, and on estimating the consequences of climatic variations and of climate forecasts. Research that clearly addresses these questions will yield progress toward the ultimate goals. We also emphasize the potential value of studying past climate fluctuations and forecasts, such as those of the 1997-1998 period, as an important approach to addressing all three categories of questions. Although in our view a rather open-ended research program makes the most sense at present, it will be reasonable
for program managers to reassess progress from time to time and to reconsider whether a more prescriptive science plan is advisable at some later date.
Potential Benefits of Climate Forecast Information
1. Which regions, sectors, and actors would benefit from improved forecast information, and which forecast information would potentially be of the greatest benefit to them?
Research on this question would aim to set an agenda for climate science from the point of view of the consumers of forecast information. Climate forecasts can be improved in multiple ways (e.g., in different regions, for different parameters, over different lead times), but these different kinds of improvements are probably not equal in terms of the social benefits they could bring. Research on this question would provide a voice of consumer demand to the climate science community.
This research should proceed from the recognition that the usefulness of forecast information is typically specific to culture, region, sector, institutional context, and other factors that influence the strategies actors use for coping with climatic variations. Thus, the benefit a forecast can bring depends not only on its accuracy but also on the nature of the human activities that occur in the region covered by the forecast and on the ability of particular actors in that region to change their behavior beneficially on the basis of information in it. For example, a storm will have dramatically different effects depending on whether or not it hits a populated area and, if it does, on how the population is organized to cope with events: how it has constructed its buildings, how it is insured against losses, how effective its warning systems are, how exposed its essential services such as electric power and food supply are to storm damage, and various other social factors that vary across locations and change over time. Thus, the value of a forecast concerning the probability of serious storms will depend on these variables as well as on the skill and accuracy of the forecast.
Because the value of a forecast is specific in these ways, it cannot be reliably determined without considering the affected activities in the region it covers and how the information in the forecast relates to the realistic options recipients have for benefiting from it. Research should therefore be directed separately at regions, groups, sectors, and institutions believed to be important in terms of the costs of climatic variability or the value of forecasts. Research should also be directed toward developing a more complete taxonomy of coping strategies as a step toward a theory of
coping. As theory improves, it is likely to become clearer which studies of decision makers and institutions are likely to be most informative.
The research effort should recognize that value judgments are inherent in any attempt to quantify the potential social benefit of improved climate forecasts. Thus, it is not necessarily advisable to seek a single metric, such as money, for evaluating all possible improvements in forecast skill. The research effort should take note of the fact that particular improvements in forecast skill may benefit some users more than, or at the expense of, others, and that the benefits and costs may be of different kinds, not easily comparable. Decisions on research priorities should also take into consideration the fact that some of the sectors, groups, or countries that may have a great potential to benefit from improved climate forecasts may have very limited capability to gain that benefit without special efforts by government or international agencies. Part of the research should concern the access of groups and sectors to useful coping strategies because providing additional coping strategies may make forecast information considerably more valuable.
Research on this question should include efforts to identify the effective responses available to actors in climate-sensitive sectors and the major constraints on their action. It should recognize that the usefulness of information may depend on the coping strategies that affected actors use, the level of skill in the forecast, its spatial or temporal resolution, the identity of the climatic parameters that can be skillfully forecast, or other attributes. The research would aim to match improvements in forecast information to the informational needs of potential forecast users. It would include efforts to estimate the benefit that might be obtained from optimal responses to improved forecasts with particular attributes. It would also consider constraints on actors' ability to take advantage of information in climate forecasts (e.g., the time sensitivity of decisions to renew insurance contracts), which may be loosened either by improving forecast skill or by changing the context of the decision. Research on this question can help establish the likely social benefit of improved forecasts in the current institutional environment and help determine how much responses to forecasts might be improved by policy interventions.
Various research methods are available for estimating the potential benefits of improved climate forecasts. One general approach is to build models that estimate the benefits that would come from optimal response to forecasts. This approach estimates the sensitivity of outcomes for particular sectors and groups to particular climate parameters and the extent to which actors could improve their outcomes given improved forecasts with given lead times. It quantifies the benefit that forecast improvements might bring to each sector and group. By using techniques such as value-of-information analysis, sector-specific studies can estimate the po-
tential benefit to particular sectors of particular improvements in forecast information.
Another general approach is based on querying the potential users of climate forecast information to learn what they want to know, when they need to know it to take advantage of the information, what information sources they would use, how they use currently available forecasts, and related questions. This strategy can be pursued by using well-established survey methodologies with representative samples of particular user groups. Skilled survey research operations exist in many countries that are capable of gathering high-quality data. However, careful attention must be paid to the design of questions because of the limited experience to date constructing reliable items on climate and climate forecasting.
The user-focused approach can also be implemented by using methods of structured discussion (e.g., workshops, conferences, and ongoing advisory bodies) involving the producers and consumers of forecasts: climate scientists, economic and social scientists, government officials, and representatives from climate-sensitive sectors and groups. This approach, although not quantitative like surveys or modeling, provides a valuable supplement to them for at least two reasons. First, the users of forecasts often have knowledge about their information needs that may not readily occur to modelers or the designers of survey instruments. Thus, dialogue can facilitate the other methods. Second, ongoing interaction between scientists and information users is likely to lead rather quickly and directly to improvements in the ways climate information is delivered. It may be a low-cost and efficient way of conveying information to climate scientists.
Ideally, the modeling and user-based approaches should be conducted in parallel. Models can identify kinds of forecast information that users may never have anticipated getting that would, in fact, be valuable to them. Models, interpreted in light of their limitations, can also help set priorities for developing forecast information on the basis of the potential benefit it can provide. User-based approaches offer greater certainty that the consumer's perspective is being conveyed to the producers of forecasts. They have the added advantage of focusing attention on information that is likely to be beneficial under actual, not only optimal, conditions of use. Also, discussion methods are likely to set in motion communication processes between producers and consumers that will enhance mutual understanding and actual use of forecast information.
Structured discussion methods are not yet very well developed, however. Because of this, and because it is reasonable to presume that discussions between the producers and consumers of forecasts will continue to be useful for some time, it is important for efforts that use the discussion
approach to have an explicit evaluative component, to learn how to conduct such discussions most effectively. This research should collect data on how structured discussion methods function in order to learn how best to organize them. Past research suggests that the involvement of user representatives in defining research questions and designing the messages that explain the researcha participatory research approachmay greatly increase the usefulness of scientific information. Research should be directed toward learning how best to apply this principle in particular types of situations.
2. Which regions, sectors, and actors can benefit most from current forecast skill?
This research would proceed from the viewpoint of climate science. It would begin with existing forecast capabilities and explore how to get the most social benefit from the dissemination of this forecast information. It would use the same modeling approaches used to address the previous question in order to estimate the value of forecast information if optimally used and would also investigate differences between optimal and actual response to forecasts among particular groups of users to identify those that might gain significant additional benefit from available forecast information under appropriate conditions.
Improved Dissemination of Forecast Information
3. How do individuals conceptualize climate variability and react to climate forecasts? What roles do their expectations of climate variability play in their acceptance and use of forecasts?
To improve the dissemination of forecast information, it is necessary to develop a basic understanding of the perceptions, beliefs, and mental models that individuals in different cultures and climatic regions use to understand climatic variability and interpret forecast information. Mental models should be investigated for their accuracy, for how they respond to new information, for how they incorporate information about uncertainty, and for how they vary by geographical region, cultural circumstances, education, and so forth. An issue of particular importance is how people will interpret a probabilistic forecast as a result of the perceived accuracy or inaccuracy of previous forecasts.
This research should investigate the roles of cognitive and affective mechanisms in the actual responses of individuals and organizations to climate forecasts. In particular, the research should examine the roles played by prior expectations of climate variability, by interpretations of
past forecasts, and by knowledge of adaptive coping strategies and beliefs in their effectiveness. To the extent that such expectations, interpretations, and knowledge are found to affect the use of climate forecasts, research should address how accurate expectations can be created and how an appropriate behavioral repertoire can be established by educational and informational interventions. To the extent that perceptions, expectations, and beliefs are identified that act as barriers to the effective use of climate forecasts, research should address how to alter those by appropriately organized information or education. To the extent that surprising (i.e., unexpected) outcomes are found to be required to motivate individuals or organizations to modify their beliefs and behavior, research should examine how to provide such educational surprises at small costs. Comparative studies of these questions across cultures and sectors may be particularly informative, as they have the advantage of distinguishing between components of those processes that are universally shared and those that are culture- or situation-specific.
One promising approach to these questions is through case studies of responses to short-range forecasts and to forecasts of the 1997-1998 El Niño/Southern Oscillation (ENSO) warm phase. Such research might examine how forecast information was delivered by scientists, the mass media, private information vendors, and others; who had access to the information; and how the information was received, understood, and used. It might test hypotheses developed from analogous situations to draw tentative conclusions about which characteristics of forecast information and its delivery increase its use by particular classes of recipients. Other kinds of studies, including experimental ones, can refine such tentative hypotheses and conclusions.
Research on these questions can be helpful in designing messages that convey climate forecast information in ways that are compatible with recipients' mental models, that accurately represent uncertainty and probability, and that do not mislead them about the level of skill the forecasts contain. Involvement of forecast user groups in such research is likely to increase the practical value of the findings.
4. How do organizations interpret climatic information and react to climate forecasts? What are the roles of organizational routines, cultures, structures, and responsibilities in the use and acceptance of forecasts?
These questions parallel those under question 3. The research would address the same questions, but it would focus on organizational behavior. Among the important organizations for study are firms in climate-sensitive sectors, organizations that provide coping mechanisms (e.g., in-
surance, reinsurance, disaster preparedness), and organizations that might interpret forecast information for large numbers of recipients (e.g., extension organizations, trade groups, mass media organizations).
5. How do recipients of forecasts deal with forecast uncertainty, the risk of forecast failure, and actual forecast failure? What are the implications of these reactions for the design of forecast information?
Research on this question should address the factors discussed in Chapter 4 that are known to limit the usefulness of probabilistic information in the context of climate forecasts. Given the difficulty individuals and organizations typically have in interpreting and acting on probabilistic information, research should address how to translate the uncertainty of climate forecasts into a format more compatible with the deterministic nature of users' reactions to such forecasts. Research should also examine the impact of different ways of communicating the inherent uncertainty of forecasts and the risk of forecast failure on users' willingness to use forecast information and on their reactions to actual forecast failures. If insufficient confidence in communicating climate forecasts hinders users' acceptance, and overconfident communication aggravates the negative impact of forecast failures, research should identify presentation styles and information formats that maximize ex ante acceptance while minimizing ex post disappointment. Comparative studies of these questions should address the influence of individual and cultural differences in interpreting uncertainty and forecast failure.
6. How are the effects of forecasts shaped by aspects of the systems that disseminate information (e.g., weather forecasting agencies, mass media) and of the forecast messages? How do these effects interact with attributes of the forecast users?
Research on these questions is key to understanding what makes for effective delivery of climate forecast information. A major focus should be on systems of information delivery. As Chapter 4 shows, research on responses to past climate forecasts and to analogous kinds of information has generated several promising hypotheses about how to deliver climate forecast information most effectively. Further research on responses to recent climate forecasts is likely to generate additional ones. However, such hypotheses require testing and modification for future applications. Research on information delivery might include experiments with aspects of information delivery systems, such as with participatory development of information and with the use of communication channels specifically selected or designed to reach particular sectors or types of actors
within sectors, especially the low-income, poorly educated, and cultural minority sectors that are often poorly served by informational campaigns. It might investigate the potential roles of trade associations, professional societies, and other groups that might take on intermediary roles, interpreting forecast information for particular user groups. It might examine how mass media organizations process climate forecast information and how information from mass media sources affects recipients' understandings. It might experiment with organized interaction between intermediaries and their audiences aimed at making information more useful. It might also include experiments with different ways of presenting particular climate forecasts and measurement of their effects on users' levels of understanding and their willingness to act. The evidence reviewed in Chapter 4 strongly suggests that involvement of forecast users in the design of this kind of research can greatly increase the practical value of the findings.
7. What are the ethical and legal issues created by the dissemination of skillful, but uncertain, climate forecasts?
Because seasonal forecasts can have significant benefits and costs and because these may be distributed unevenly across human populations and ecosystems, scientific climate forecasting raises ethical and legal issues. Already, scientists and public officials have been held responsible in court for costs associated with actions taken using their forecasts. Ethical research questions address when and how to issue forecasts, how to deal appropriately with uncertainty, how forecast skill should be developed to achieve an appropriate distribution of the benefits, and how ethical beliefs (e.g., concerning the rights of nonhuman species or equity among human populations) do and should affect the development, presentation, and dissemination of forecast information. Legal research questions include assessing case law regarding responsibility for climate, weather, and analogous forecasts; the treatment of scientific uncertainty in the legal system; the relationship between impacts and liability settlements; and the role of legal institutions (e.g., water and property rights) in coping with climatic variability and climate forecasts.
Consequences of Climatic Variability and of Forecasts
8. How are the human consequences of climatic variability shaped by the conjunctions and dynamic relationships between climatic events and social and other nonclimatic factors? How
do seasonal forecasts interact with other factors and types of information in ways that affect the value of forecasts?
Climatic variations influence resource productivity, economic development, and human and ecosystem health, but these impacts are typically mediated or exacerbated by coping strategies and by other trends and events such as demographic and technological changes, economic and political conditions, and resource management strategies. Improving the capability to estimate the human consequences of climatic variations requires improved understanding of the social and other non-climatic phenomena that combine with climatic ones to produce these consequences. For example, the many studies of climatic impacts on agriculture show clearly that yields depend not only on climate but also on available technology, soils, prices, agricultural policies, and individual farm management. In the general case, the effects of climatic variations may depend on population growth in and migration to areas that experience large climate variations; economic and infrastructural development in such areas; the level of dependence of human populations on food and other essential goods and services delivered from outside their immediate vicinity; technologies and practices affecting land use and water demand; the distribution of economic resources; the levels of income and education among affected actors; the availability of insurance and insurance-like institutions; the structure of markets for the supplies and outputs of affected actors; and the condition of formal emergency warning and response systems. Research on the effects of climatic variability should distinguish the effects of climate from those of such variables as these and clarify the dependency of climatic effects on these other variables. A variety of methods is available for these tasks, as shown in Chapter 5.
Some of the nonclimatic events that influence climatic impacts change systematically over time. For example, future hurricane damage is highly dependent on trends in population migration and the rate of building construction and in certain coping systems, such as the adoption of hurricane-resistant building codes and practices, as well as on storm frequency and intensity. The specific effects of these and other nonclimatic factors on the sensitivity of particular sectors are not well understood, however. Future research should examine how changes in social conditions affect the sensitivity of particular groups and activities to particular kinds of climate variation.
This research should especially emphasize climatic variations about which predictive skill is improving, human activities and groups that are believed to be highly sensitive or vulnerable to these variations, and human activities and groups that may become increasingly vulnerable as a
result of social changes. The research would identify opportunities for policy changes that might reduce the likelihood of catastrophic outcomes or increase the ability of human activities and groups to benefit from expected climate variations and from forecasts of them. In conjunction with research on coping strategies, this research would make it possible to estimate the future benefits of climate forecasts in the context of expected future social conditions.
9. How are the effects of forecasts shaped by the coping systems available to affected groups and sectors? How might improved forecasts change coping mechanisms and how might changes in coping systems make forecasts more valuable?
To estimate and increase the value of climate forecasts to society, it is necessary to understand the current coping systems available to groups and sectors of society. Coping mechanisms for dealing with climate variability are both formal and informal and range from individual behaviors to national policy. Some coping systems will enhance the benefits from forecasts, whereas the limited flexibility of others may constrain the ability to take advantage of the information. Research should be conducted on how improved forecasts may alter currently effective coping strategies (e.g., how better forecasts might change the products of the plant breeding industry, the use of insurance by farmers and other vulnerable actors, and the operation of the insurance and industry and government relief programs). Research should also be conducted on the public policies and institutional mechanisms that affect coping strategies (e.g., government farm subsidies) to gain understanding of how well they serve to mitigate the negative effects of climate variations and how they might serve best in an environment of improved forecast skill. Research should also explore issues of access to coping strategies that might benefit particular groups. It should address the ways the usefulness and value of forecasts may depend on changes in coping systems as a result of such forces as population changes, migration, economic development, and political changes, as well as the potential for modifying coping systems so as to make forecasts more valuable.
10. Which methods should be used to estimate the effects of climate variation and climate forecasts?
A variety of modeling strategies and discussion-based qualitative methods is available for estimating these effects, and there is a place for many of them given the current state of knowledge. Some methods may be more accurate or more useful for certain purposes, and other methods for other purposes. Some of the research on this question should examine
and compare the outputs of different modeling methods to shed light on the usefulness of each as well as to increase understanding of the underlying phenomena. Some of the research should use discussions, surveys, and other techniques to estimate the effects of climatic events and of forecasts. A special focus of nonmodeling research should be on outcomes for which good quantitative data do not exist or for which the value of quantitative data is uncertain.
11. How will the gains and losses from improved forecasts be distributed among those affected? To what extent might improved forecasting skill exacerbate socioeconomic inequalities among individuals, sectors, and countries? How might the distribution of gains and losses be affected by policies specifically aimed at bringing the benefits of forecasts to marginalized and vulnerable groups?
Research on this question should examine who might gain and lose from improved forecasts and the factors affecting the distribution of gains and losses. This research is of more than academic interest, as various public policy decisions may affect the distribution of benefits and costs from forecasts. For example, the benefits of agricultural innovations and many kinds of information have flowed first to large-scale, well-educated actors. It is important to anticipate whether climate forecast information is likely to be distributed in ways that follow that model and to examine the effects of targeted efforts to deliver the information to groups otherwise unlikely to benefit. Similarly, the response of the insurance industry (including firms, reinsurers, and regulatory authorities) to forecasts will lead to differential effects as a function of who is insured and how. Policies could alter these effects. Institutional changes, such as in water rights laws, may also affect the distribution of the benefits of forecasts. In developing countries, access to world markets or government regulation of agricultural prices might interact with the availability of forecasts in determining the distribution of benefits.
Decisions about who distributes forecast information (e.g., national weather services, extension organizations, private-sector vendors) also have implications for the distribution of the benefits and costs of forecasts. In general, public policies adopted for a variety of purposes may affect access to forecast information and the ability to make adjustments in response to climate forecasts. Thus, research illuminating the distribution of the benefits of climate forecasts and the effects of policy interventions on this distribution is likely to be relevant to a wide range of public policy choices. In addition, research might examine the possibility that, even if everyone made optimal responses to better forecasts, these re-
sponses would leave some groups worse off than they would have been without the forecasts. This might happen, for example, to farmers who cannot increase their production while farmers in another region use forecasts to produce a bumper crop that drives down prices.
12. How adequate are existing data for addressing questions about the consequences of climate variability and the value and consequences of climate forecasts? To what extent are existing data sources underexploited?
For example, can existing data sets that have emerged from socioeconomic panel surveys in many countries of the world be fruitfully merged with appropriately geocoded information on climate and weather over time to better quantify the effects of climate variability? What is the value of surveys that elicit speculative information on what respondents would do differently with forecast information or what they would have done differently if they did not have such information (such as in the 1997-1998 El Niño event) for estimating the true impact of improved forecasts? Can existing demonstration projects provide adequate information on the value of existing forecast data? Are long-term ecological research sites areas in which adding a human dimension to the data collected would provide improved information on the relationship between human behavior and climate variation?
The Value of Studying Past Climate Fluctuations and Forecasts
Past climate fluctuations (e.g., droughts, heat waves, flooding) provide natural experiments to examine ex post responses of human systems and the environment. They are the only situations that permit direct observation of human and institutional behavior in response to dynamic climate. The same is true with past climate forecasts. Such natural experiments potentially reveal important details about the sensitivities of human activities to climate variability and about responses to forecasts. They can be used to help identify and quantify biophysical responses to climate fluctuation, social costs and benefits of both the climatic events and any available forecasts, and coping mechanisms. Careful analysis of the effects of past forecasts would allow the forecasting community to benefit from seeing how the forecasts were used (or not used), how forecast use was shaped by the forecast information and its delivery, and what can be done to improve the usability of the forecasts. There have been a number of natural experiments in recent years, including, in the United States, forecasts in relation to the drought of 1988, the Mississippi flood of 1993, and various forecasts of hurricane tracks. Of course, the 1997-1998 El
Niño and its forecasts have presented a highly relevant natural experiment that deserves detailed study as a way of investigating most of the scientific questions listed here. In interpreting such natural experiments, it is important to consider such variables as the length of lead time forecasts offer and the degree of confidence scientists have in their forecasts.