glacial changes have been compiled for about 500 glaciers worldwide (e.g., Wood, 1988). Figure 6 (from Wood, 1988) shows the trends in glacier mass balance over approximately 20 years, expressed as the percentage of observed glaciers having a positive mass balance. While this percentage for all glaciers is close to 50 percent, the percentage for an individual region such as the European Alps can undergo multi-year excursions well above or below 50 percent.
Aside from CO2-doubling experiments, there have been very few three-dimensional model simulations of the land-atmosphere system over decadal time scales. In this respect, one may argue that less attention has been given to the simulation of land-atmosphere interactions than to the simulation of large-scale ocean variations over decade-to-century time scales. Examples of the latter are the experiments of Bryan (1986), Weaver et al. (1991), and Yang and Neelin (1993). Moreover, most models containing an interactive land surface treat the surface physics quite crudely: the "bucket" method is typically used to handle soil moisture, multi-level soil treatments are generally not included, and runoff often "disappears" from the system.
Most global climate models do include a thermodynamically and hydrologically interactive snow cover. In its assessment of climate models, the IPCC (1990) found that the snow cover simulated by several models was "broadly realistic," leading to the statement that "the simulated snow extent should thus not distort the simulated global radiative feedbacks" (IPCC, 1990, p. 113). Even if one gives the benefit of the doubt to that assertion, the snow cover simulated by all models appears to contain significant errors, particularly over eastern Asia. More comprehensive diagnostic assessments of the simulated snow cover (extent and water equivalent) in the context of both the present climate and projections of climate change are needed.
A more sobering finding concerning model simulations of the feedback between snow and radiation is contained in the IPCC's (1992) update, which cites Cess et al.'s (1991) comparison of the snow feedback (under perpetual April forcing) in 17 global climate models. The temperature sensitivity or feedback parameter, λ, associated with snow cover was found to vary from negative values in some models to a wide range of positive values (Figure 7). The sensitivity parameter in clear-sky regions ranges from values corres-