cover. However, Lamb (1955) showed that there is a detectable decrease in the 1000 to 500 mb thickness as a layer of air passes over a large area of snow cover, implying that snow cover can contribute to the maintenance of a trough of cold air, which in turn helps to maintain the snow cover.
Some of the earliest work on the role of snow cover in seasonal atmospheric variability was directed at the Asian monsoon. Blanford (1884), Walker (1910), and others postulated a link between Himalayan winter snow cover and the strength of the Indian summer monsoon by reasoning that extensive snow cover could retard the heating of the Asian landmass. This link has been substantiated in recent years with the aid of satellite-derived measurements of snow cover (Hahn and Shukla, 1976; Dey et al., 1984; Dickson, 1984; Bhanu Kumar, 1988). On the basis of data for the 1967-1980 period, linear correlation coefficients between December-to-March snow extent in the Himalayas and June-to-September monsoon rainfall were found to be approximately -0.6 (Dey et al., 1984; Dickson, 1984).
While the monsoon-snow correlations seem to provide the basis for useful monsoon predictions at ranges of several months, there are several important caveats. First, snow cover over the Himalayan region was not mapped consistently during the first part (1967-1972) of the satellite record (Ropelewski et al., 1984). Second, as shown in Figure 2 (from Bhanu Kumar, 1987), the agreement between the interannual fluctuations of snowfall and monsoon variables has been noticeably poorer since 1980. Inclusion of data from 1981 to 1985 lowered the snow-rainfall correlations from -0.60 to -0.38 (r2˜0.14); the 95 percent significance threshold for a sample of N = 19 years is approximately 0.45. Third, Indian monsoon rainfall appears to correlate as highly with snow cover over the remainder of Eurasia as it does with Himalayan snow cover (Dickson, 1984). The physical linkage between snowfall and monsoon rainfall may therefore be more complicated than implied by the proposed effect of snow on the timing of the heating of the Himalayan and Tibetan region. For this reason, the modeling studies (e.g., Barnett et al., 1989) described below provide potentially important diagnostic information.
Other observationally derived linkages between snow cover and the atmospheric circulation over seasonal time scales have been addressed by Afanas'eva et al. (1979), who examined the position of the Planetary Upper-air Frontal Zone (PUFZ) over Eurasia during the autumn and spring. Time variations of the PUFZ and the snow boundary correspond closely (r = 0.78). However, as noted above, such correlations are at least partially attributable to the fact that the position of major upper-air features (e.g., the jet stream) is a primary determinant of the position of the snow boundary. Interannual variations of snow cover over Eurasia were also examined by Toomig (1981), who found that the annual value of the absorption of solar radiation at Soviet stations is a strong function of the springtime absorption, which in turn depends on the albedo during spring. Toomig also reported modest correlations (r ˜ 0.45) between the surface albedo of early spring and the surface air temperatures through July at several stations. A similar lag relationship