While it is realized that the fraction of precipitation as snow should be eventually taken care of by the cloud/precipitation parameterization scheme in atmospheric general circulation models (GCMs), most (if not all) of the current GCMs still use a simple relation and/or temperature threshold to provide input of snow and rain to the Earth's surface. Historically, the National Center for Atmospheric Research (NCAR) GCMs have used the surface air temperature at 2.2C to determine whether precipitation is snowfall or rainfall following Auer (1974). Recent phases of the Project for Intercomparison of Land-surface Parameterization Schemes (PILPS) concerning the cold season processes have employed a threshold at 0C. The U.S. Army Corps of Engineers (1956) developed a formula relating snowfall to temperature for a wide range, so the mixed rain and snow is accounted for temperatures ranging from -2C to 4C. We have done a series of GCM runs to sepcifically address the impacts of the partitioning of precipitation into snowfall versus rainfall on snowpack hydrology, general circulation, and other hydroclimatological variables using the NCAR CCM3 coupled with our latest physcilcally-based multi-layer snowpack model. The preliminary results indicate that using 2.2C threshold produces more snow in most area of north hemisphere land than using 0C, however, it produces less snow in Central Asia due to decreased total precipitation. The feedbacks of snow surface and general circulation will be presented at the meeting.