and investments in infrastructure. Changes in frequency and intensity of extreme weather events may accompany such changes in climate (Karl et al., 1996), such as the devastating Midwestern floods that struck the United States in 1993 and again in 1997. The remarkable change in the flood frequency of the American River above Sacramento, California, is the subject of a current NRC study in the Water Science and Technology Board. The Folsom Dam was built in 1945 to provide flood protection for Sacramento. Eight floods greater than the largest flood in the 1905-1945 period have occurred since 1945. A similar situation exists for several of the other Sierra Nevada rivers in California. These high floods have led people to question the level of flood protection actually provided by the dam, and, more important, how flood risk should be analyzed.
Better information on likely climate change and the associated regional patternsfor example, the probability that such floods may occur in clusters, say six or seven times over a 20-year periodwould not only permit the mitigation of negative impacts but afford the opportunity to exploit positive impacts. Governments and individuals alike would benefit from advance knowledge of any climate changes that would have a major impact on agriculture, energy production and utilization, water resources and quality, air quality, health, fisheries, forestry, insurance, recreation, and transportationall fundamental to society's well-being, all vulnerable to any prolonged change or abrupt shift in our climate system. Not only would society benefit from increased climate-prediction skill by being better prepared to ward off adverse climatic consequences, but advance knowledge of climate variations would also enable society to capitalize on opportunities, such as increased geographical ranges for certain crops.
Unfortunately, the subtlety of slow changes over long time scales (relative to diurnal, seasonal, and interannual variations) tends to disguise their potential long-term severity, and thus limits society's willingness to address them in advance; this lack of urgency is exacerbated by the uncertainty in scientists' ability to forecast such change. Given the requisite understanding of climate variability, we hope to ultimately forecast and detect alterations in climate change (distinguishing natural variability from anthropogenic change), providing a rational basis for future policy and infrastructure-management decisions.
The limitations of the instrumental data on which our current state of understanding is based are readily exposed by evaluating their ability to help answer some of our most fundamental questions involving decadal or centennial change. For example, questions such as "Is the planet getting warmer? Is the hydrologic cycle changing? Are the atmospheric and oceanic circulations changing? Are the weather and climate becoming more extreme or variable? Is the radiative forcing of climate changing?" cannot yet be answered definitively. Each one of these apparently simple questions is actually quite complex, both because of its multivariate aspects and because global spatial and temporal sampling is required to address it adequately. The global observing systems needed to provide the answers are either inadequate or non-existent. For science to provide society with the information it needs, better data are essential. The models that will yield predictions require these data to improve our understanding of decade-to-century-scale climate change, its rate and range of variability, its likelihood and distribution of occurrence, and the sensitivity of the climate to changes in the forcing (natural and anthropogenic).
The fundamental need to develop a good scientific understanding of climate variability and change over decade-to-century time scales, the inadequacy of our current understanding, and the limited resources available to increase this understanding all point to the need for a nationally recognized dec-cen science plan. The present report articulates the primary scientific issues that must be addressed in order to advance most efficiently toward the necessary understanding. In developing this plan, the members of the Dec-Cen panel have taken special care to recognize that research directed toward decade-to-century-scale change and variability will differ in two remarkable respects from research directed at shorter-time-scale variability.
First, research on these intermediate time scales is relatively new. As noted above, only recently have we obtained sufficiently long high-resolution paleoclimate records to allow the examination of past change on dec-cen time scales, and acquired faster computers and improved models that can perform the long simulations needed for studying such change. Consequently, we are on the steep slope of the learning curve, with new results and dramatic insights arising at an impressive rate. The fundamental scientific issues requiring our primary attention are evolving rapidly. Flexibility and adaptability in response to new opportunities and promising directions will be imperative if we are to optimally advance our understanding of medium- and long-range climate change and variability.
Second, the paradigm developed for the study of climate change on seasonal-to-interannual time scales cannot be applied to the study of climate problems on longer time scales. We have recently achieved considerable success in studying short-time-scale climate problems by generating hypotheses and models that are quickly evaluated and improved through analysis of the existing and rapidly expanding instrumental records. For longer-time-scale problems, the existing paleoclimate records are still too sparse and the historical records too short; as for future records, multiple decades will be required before even a nominal comparison with model predictions becomes possible. Furthermore, the change in atmospheric composition as a consequence of human actions represents a forcing whose future trends can be estimated