tively sophisticated (e.g., nonlinear) models are required for the diagnosis of snow-atmosphere interactions.
With regard to (4) and (5), the validity of parameterizations involving snow cover in global climate models may be a major constraint on further progress toward an understanding of snow-atmosphere feedbacks. Priority should be given to the parameterization of quantities and processes such as snow surface albedo, the influence of snow on the boundary layer structure, and the disposition of the liquid water produced by snow melt.
The preparation of this paper was supported by the National Science Foundation through Grants DPP-9214793 and ATM-9319952. We thank Norene McGhiey for typing the manuscript.
DAVID A. ROBINSON
Rather than commenting on Dr. Walsh's paper, I am going to show you a little about my work on hemispheric snow cover over the last several decades and also on the century time scale I am beginning to put together. This is at Dr. Walsh's suggestion, I hasten to add.
Figure 1 gives you an idea of the annual cycle of snow cover over Northern Hemisphere lands. The area covered by snow ranges from 40 to 50 million km2 during the winter over the Eurasian and North American continents to several million km2 in summer, primarily on top of the Greenland ice sheet.
In recent years there has been a dearth of snow cover over Northern Hemisphere lands (cf. Figure 3 in my paper in this volume). It has been true for the last 5 to 6 years over both Eurasia and North America. So it is seen hemisphere-wide, with many record lows set in the last couple of years, particularly 1990. I remind you that this is based on 20 years of observation. Nonetheless, this record, produced weekly by NOAA, happens to be the most consistent long-term satellite-derived data set available for any climatic variable. Charts go back to 1966, but most of us who have looked at the data feel that it is relatively homogeneous and credible only from 1972 on.
Looking at seasonal variations in snow cover over the past 20 years (cf. my Figure 4 in this volume), you see that there has been very little variation over the winter seasons in Eurasia, and even less in North America. This agrees with what Dr. Walsh showed earlier for sea ice. Fall cover has varied a bit more, but spring and summer show the greatest year-to-year variations in cover. In recent years snow cover has decreased in March through June over both North America and Eurasia. These spring decreases in recent years have been most pronounced over northeastern Asia and down the leeward side of the Canadian Rockies into the northern portions of the United States.
So that is what we have seen hemispherically over the past 20 years. Unfortunately, we have no data sets going back longer than that on a hemispheric scale, but there have been some recent efforts to try to extend local and regional data sets back to the turn of the century. For instance, at Denison, Iowa, over the course of this century there appears to be an increase in the number of snow-covered days with =1 inch (2.5 cm) of snow on the ground during the fall, winter, and spring.
As for other data sets to examine, we are beginning to sort out daily data from a set containing several hundred Russian stations. In some cases, the records go back to the turn of the century. We have also been examining station data in western China extending back into the 1950s.