Two case studies involving extensive winter storms over the US Great Plains were investigated. First, precipitation in the form of snow vs. non precipitation events were classified for large areas affected by these storms using NEXRAD radar estimates and ground-based synoptic "first order" weather stations of the US National Weather Service as truth. Next, AMSU measurements in the 23-183 GHz region were matched up with the classified events. Results show that snowfall can be reliably captured from AMSU measurements for a range of precipitation events using mainly the 54, 150, 176 and 180 GHz higher frequency channels. Based on these findings, a preliminary snowfall retrieval algorithm was developed and successfully tested over the US territory, capturing a major portion of snowfall and complex storm dynamics surprisingly well. However, this study was done for a limited geographical region, and work is continuing to develop a global snowfall retrieval scheme over land.