over the past several decades (Dutton et al., 1991; Henderson-Sellers, 1990; Karl and Steurer, 1990; London et al., 1991; McGuffie and Henderson-Sellers, 1989; Parungo et al., 1994). Clearly, a much more focused effort is required to better understand decadal and multi-decadal fluctuations of cloud cover, since they often have very important feedback effects on other climate quantities. For instance, cloud feedbacks work differently during the day from the way they do at night. They therefore differ at high and low latitudes, making generalizations speculative.
Also, there is no discussion in this volume of the world's glaciers, in spite of their known tendency to fluctuate markedly on decade-to-century time scales. Long-term trends of climate change are integrated by mountain glaciers, and during the past century mountain glaciers all over the world have been declining (IPCC, 1990). The World Glacier Monitoring Service (1993) summarized these changes over the past several decades, and the USGS Satellite Image Atlas of the World discusses observed variations of hundreds of glaciers over the past several centuries.
Identification and explanation of the forcings and feedbacks responsible for decade-to-century-scale climate fluctuations are essential to distinguishing between natural and anthropogenic impacts on the climate system, as well as to developing any predictive skill with respect to these phenomena. Further progress in understanding this gray area of climate change will depend upon our ability to address three topics. First, we must be able to document climate fluctuations without spurious discontinuities and trends. Second, we must ensure that we are adequately monitoring potential forcings and feedbacks internal and external to the climate system, so that we will have the data necessary for testing hypotheses about their operation. Last, data analysts must work closely with modelers (and vice versa) to test the hypotheses we formulate on a variety of climate models, ranging from simple one-dimensional climate models to complex coupled ocean-atmosphere general-circulation models. The re-analysis modeling projects of the United States (Kalnay and Jenne, 1991) and the European community are likely to enhance our confidence in both how and why climate has varied on decade-to-century time scales. (It must be emphasized, however, that the value of any re-analysis effort can be jeopardized by the use of data that are biased or, most troublesome, inhomogeneous in time.) Only by advancing in all three of these areas will we be able to overcome our ignorance about multi-decadal climate fluctuations and make predictions with any degree of assurance.