I had a similar problem for western part of US. I used PRISM data which has raster for temperature and precipitation. I assumed that precipitation for the cells which has temperature less than zero is snow. This worked well with real data for me. The raster was about 4 km x 4 km and monthly data was used.