Abstract: Using satellite remote sensing information, the distribution of cloud-top particle can be retrieved and the physical mechanism of certain weather phenomenon can be explained. Nowadays, cloud phase inversion method based on passive remote sensing instruments on 8 μm, 11 μm and 12 μm spectral bands (corresponding to three MODIS infrared spectral channels: Ch29, Ch31 and Ch32) are relatively practical, which is called tri-spectral cloud phase inversion method. EOS/MODIS satellite data in Akzo Prefecture of Xinjiang and observations from relevant weather station are analyzed to study a snowstorm. The change of cloud phase is revealed by tri-spectral brightness and temperature difference scatter diagram, which clearly shows the whole process of cloud clustering from water cloud to hybrid cloud, and finally to ice cloud. It also reveals the physical mechanism how this weather system occurs and finally develops to blizzard. It can be seen that the result from tri-spectral cloud phase inversion method is completely according with actual circumstance. Therefore, it can be concluded that tri-spectral cloud phase inversion method has a theoretical advantage and a wide application foreground in practice. Meanwhile, this method of MODIS data cloud phase recognition can be applied to analysis of blizzard clouds by FY-3 satellite.