times, and to lay the foundation for converting them into a long-term operational observing systems.
Future observations will require more attention to data quality, homogeneity, and continuity if we are to understand the nature of decade-to-century-scale climate variability. Currently under consideration is a Global Climate Observing System (GCOS)1 that includes observations of terrestrial, atmospheric, and oceanic aspects of climate. Since GCOS would be built around existing observing networks and environmental problems, it is essential that scientists effectively convey long-term climate-monitoring needs to the agencies sponsoring space-based and conventional observing systems. Among the climate-monitoring issues that should be addressed in the near future are stability of network sites, intercomparability of instruments, and increased sampling in data-sparse regions. Two concerns for both present and future observations are the better collection and documentation of metadata on observing instruments and practices as well as processing algorithms, and improved data-archiving practices and data-management systems. All these will increase the value of climate monitoring for decade-to-century-scale research; most essential, of course, is assurance of a long-term commitment to observational objectives.
The climate record provides the key to developing, refining, and verifying our hypotheses regarding the forcing agents responsible for climate fluctuations, be they anthropogenic or natural, internal or external to the climate system, global or regional, or persisting one or many decades. Diaz and Bradley (1995) suggest that the many decade-to-century climate fluctuations evident in both observations and proxy records may indeed have natural origins, since similar fluctuations seem to occur in the climate record both before and after humans became capable of modifying climate.
Often physically based models are the best means of testing our hypotheses about the cause of the fluctuations, as described in the modeling sections of this volume, but in addition they can be effectively used to help discern the physical consistency of apparent climate fluctuations and change. Cayan et al. (1995) demonstrate this approach; they use an ocean general-circulation model to help explain the climate jump over the North Pacific, as documented by Trenberth (1990; Trenberth and Hurrell, 1995, in this volume). Cayan et al. show how the atmosphere and ocean can act together to maintain decadal-scale climate fluctuations on a large spatial scale.
Potentially important factors in explaining climate fluctuations on decade-to-century time scales are land-surface and atmospheric feedback effects. Walsh (1995) provides ample evidence that snow cover has important feedback effects on the climate system on short (daily and interannual) and long (thousands of years) time scales, but on the time scales of interest to us the impact of snow is still not well understood. However, Nicholson (1995) and Shukla (1995) present evidence that the feedback between land-surface characteristics and the atmosphere has led to the prolonged drought in the Sahel. Karl et al. (1995) document an asymmetric increase of the mean maximum and minimum temperatures over many portions of the global land mass. Although they cite a number of potential causes of this multi-decadal trend, such as increases of anthropogenic atmospheric sulfate aerosols (Charlson et al., 1992) and greenhouse gases, empirical evidence suggests that, at least in some regions, observed increases in cloud cover play an important role in modulating the surface temperature. The forcing responsible for the increase in cloud cover remains unknown.
There are many important aspects of climate forcings and associated responses that cannot be covered here. Of particular relevance are the known changes in solar irradiance associated with the sunspot cycle. Recently Friis-Christensen and Lassen (1991) have used the length of the sunspot cycle to explain the decadal fluctuations and trends of land temperatures. Although a linear response of the surface temperature to the sunspot cycle length implies some unlikely responses of the temperature record in the early part of the time series (Kelly and Wigley, 1992), it is clear that changes in solar irradiance must continue to be monitored as a potential source of global climate fluctuations or change.
Recently, Elliott et al. (1991) have found evidence for an increase of tropospheric water vapor since 1973, leading to an enhanced greenhouse effect. However, the many types of observing and data-processing inhomogeneities in the upper-air moisture record make it immensely difficult to separate spurious trends and discontinuities from the true climate signal, even when the signal may be as large as a 10 percent increase in specific humidity.
A critical atmospheric quantity affecting surface temperature is the variability of cloud cover—its type, height, and spatial distribution. Using the ISCCP data base, Hartmann et al. (1992) showed the net forcing of various cloud types as a function of season and latitude. A recent study by Rossow (1995), however, reveals that this data set has serious biases that make it inappropriate for decadal-scale climate assessments. Meanwhile, conventional in situ data, analyzed on a national basis by a number of researchers, show a widespread general increase in total cloud cover