model experiments provide a potentially useful vehicle for diagnosing the effects of snow cover. However, the mixed nature of the model-derived conclusions pertaining to the impacts of snow cover will become apparent as the survey proceeds.
In a study of snow-induced effects in non-winter months, Yeh et al. (1983) used a simplified version of the Geophysical Fluid Dynamics Laboratory's general-circulation model (GCM), containing idealized geography and a limited computational domain. The model contained no diurnal cycle, and its cloudiness was prescribed to be zonally uniform and seasonally invariant. The complete removal of the snow cover in mid-March was found to reduce the water available to the soil through snow melt, thus decreasing the soil moisture during the spring and summer in the region of snow removal. The drying of the soil resulted in an increase of surface temperature at high latitudes by 2°C to 8°C for the subsequent 3 to 4 months. The temperature increase extended into the upper troposphere, thereby reducing the meridional temperature gradient and the zonal wind in high latitudes.
A similar conclusion about the snow-hydrology-temperature linkage was obtained from a more realistic GCM by Yasunari et al. (1989). In this experiment, the 5° × 4° version of the Japanese Meteorological Institute GCM was run for 6 months beginning March 1. The experimental runs were identical to the control runs except for the addition of 5 cm (water equivalent) of snow in the snow-covered portion of the 30° to 60°N zone of the Eurasian continent. The results, shown in Figure 3, contained evidence of both (1) an albedo feedback, which suppressed temperatures over lower latitudes (e.g., Tibetan Plateau) by 2°C to 3°C during spring, and (2) a snow-hydrology-temperature linkage, which suppressed temperatures over middle latitudes by 2°C to 3°C during the summer months of June to August. During the summer, the anomalous Eurasian heat sink also appeared to induce a stationary Rossby wave
pattern extending from eastern Asia to northern North America.
The most thorough investigation of spring-summer feedbacks involving snow cover is Barnett et al.'s 1989 study of Eurasian snow impacts on a low-resolution (T2 1) version of the European Centre for Medium-Range Weather Forecasting's model. In Barnett et al.'s first experiment, snow extent corresponding to observed extremes was prescribed and interactions between snow and the surface hydrology were suppressed in order to isolate the albedo effect. The atmospheric response to the snow anomalies was local and confined primarily to air temperature and upper-air geopotential (but not sea-level pressure). All significant signals vanished when the snow disappeared in the spring, and the albedo effect had no sustained impact on the development of the Asian monsoon.
In Barnett et al.'s second experiment, rates of snowfall over Eurasia were doubled and halved so that the subsequent melt and evaporation could induce changes in the regional hydrology. The two sets of simulations showed statistically significant differences extending through the subsequent two seasons. The results derived from the doubled snowfall were characterized by significantly lower surface and tropospheric temperatures from May through July, higher pressures over Asia and lower pressures over North America, weaker zonal winds over the Arabian Sea, weaker surface convergence over southern Asia, and a weaker monsoon over southeast Asia. The Indian monsoon, however, was not substantially weaker in the "heavy snow" simulations, although this result may be partially attributable to the model resolution. The sea-level pressure signal over Asia and North America is stronger in the model than in the real world; the exchange of mass between the two continents may have been exaggerated in the model because the sea surface temperature (SST) distribution was prescribed climatologically in all the model runs.
In general, the physical mechanisms underlying Barnett et al.'s model response are similar to those of Yeh et al. (1983). The doubling of snowfall by Barnett et al. also produced a general weakening of the wind in the Southern Hemisphere and along the equator. While this response is similar to that which occurs prior to the warm phase of an ENSO event, the model with prescribed SST cannot sustain an ENSO event. In order to address the snow-ENSO link in more detail, Barnett et al. performed additional experiments with a coupled ocean-atmosphere model. The conclusion was that "the snow/monsoon signal has all the characteristics necessary to trigger the Pacific portion of an ENSO event, but the signal is too small by a factor of at least 2. In balance, it appears that snow-induced monsoon perturbations may be one of the (multiple) triggers that can initiate an ENSO cycle" (Barnett et al., 1989, p. 683-684). The elimination of major model biases and the improvement