that significant change has taken place (Figure 1d). Many of the papers in this volume contribute insights and data on the form of such climate variations.
The identification of the characteristics of climate variability involves several issues. Differentiating climate ''change" from climate "variability" is a matter of the time scale. What appears to be a trend, in a single decade's recording, may reveal itself as fluctuating variability over a period of a century. The Dust Bowl period in the central United States in the 1930s represents a short, decadal-scale natural variation in climate for a specific region. In contrast, the "little ice age", which lasted from the 1400s to the 1800s, represents a variation on a time scale of centuries. (During the little ice age the decadal variability was typical of that observed during modern conditions.) An additional complication arises because variation that occurs on one time scale may influence changes on other time scales. The well-documented El Niño / Southern Oscillation phenomenon, which is characterized by variation over 4- to 7-year periods, seems to be affected by longer-time-scale variations and, in turn, to modulate the variations themselves.
Even after the time scale of interest has been defined, it may be difficult to recognize and classify natural variability as evidenced in modern instrumental observations, because anthropogenic change may already have contaminated them significantly. Limitations on data continuity (loss of records, changes in instrumentation, gaps, lack of data quality assurance) are also a serious handicap. Longer records, especially those based on proxy indicators of climate (i.e., indirect measures of climate such as the width of annual growth cycles in trees and corals), represent an excellent potential additional data set for estimating or extracting natural variability.
An additional difficulty in assessing variability is that change and variation are often characterized by strong spatial dependency. That is, while some region or regions of the earth may be experiencing significant climate variation over a given period, other regions may show virtually no change. Similarly, a specific variable (e.g., global temperature) may show a marked change between decades, while another (e.g., global pressure fields) may not show any significant difference. Where and how climate is measured can influence the findings and conclusions.
Our understanding of the causes of decade-to-century-scale variability is limited, in part because there are many possible causes, typically disguised by complex interactions. They include inherent variability (deterministic or random within individual components of the earth system), internal variability associated with the coupled ocean-atmosphere-cryosphere system, and forced variability such as solar variation and volcanic aerosol loading of the atmosphere. One key challenge is to isolate (if possible) the signatures of the various possible mechanisms so that we can discriminate between cause and effect as we examine the climate record. Also, a greater understanding of the magnitude of the forcing required to produce observed variations will allow us to focus on the mechanisms most likely to contribute to climate variability. It seems likely that several mechanisms, operating in concert, are responsible for the variations apparent in the observed climate record.
Understanding the mechanisms that produce natural variability will require a hierarchy of climate models, including coupled models that are capable of addressing the interactions between the components of the earth system. The development of models that will yield simulations or predictions that can be verified and validated against the observational record, is another major challenge. Models that can describe the nature of decade-to-century-scale variations will serve not only as a measure of our understanding, but as a tool to increase this understanding. They provide our primary opportunity to predict climate variability, although ultimately prediction may be achieved through a variety of means. Given specified forcing scenarios, models may provide viable climate-response scenarios. But statistical characterization is another possible tool. For example, it would be useful to know that a particular extreme climate event (e.g., major regional flooding) tends to occur in clusters over a several-decade period rather than irregularly and infrequently.
Making the distinction between natural variations and human-induced changes—and ultimately predicting future changes—will require more complete characterization of the climate system's variability in both space and time, a greater understanding of the causes of decade-to-century-scale climate variability and the mechanisms by which it is produced, and the development of more advanced predictive models. The papers included in this volume reveal significant progress in identifying the behavior of many key components of the climate system, their climate signatures, their internal modes of variability, and their interactions with different components of the earth system. The coupling of observations and models, together with the availability of both long-term, consistent, and high-quality observations and proxy data records, will be critical to this assessment of climate and its change.
The importance of climate variability to society motivated the organization of a workshop by the Climate Research Committee (CRC) of the National Research Council's Board on Atmospheric Sciences and Climate. Not only have climate variations on decade-to-century time scales received significantly less attention than seasonal and interannual climate variations, or even the glacial/interglacial periods, but for nearly two decades natural variations have been overshadowed by much-publicized concerns about the possibility of long-term changes caused by increases in