Abstract: Storm is one of the major nature disasters during flood season in China. Because of its difficulty to estimate precisely the spatiotemporal distribution of storm and its intensity change，it is impossible to forecast storm accurately. Recently，satellite remote sensing has become a critical approach for storm monitoring. Meanwhile，the Cloud Phase Retrieval as the main topic in the study of rain precipitation formation，is known for facilitating deduction of convective intensity，weather forecasting and weather modification through investigating the distribution of cloud particles. In this study，storm cloud in visible，near-infrared，infrared channels was analyzed and characterized using rain precipitation and FY3A/VIRR data collected from 2009-2011 in the basic weather stations．Cloud particles had significant difference at 0.65 μm and 1.6 μm channels when bright temperature was less than 233 K and these two channels also showed 4 times difference when there are strong air convection and thick cloud. Then using the technique of multispectral cloud phase retrieval，cloud phase was subdivided into ice clouds，cirrus clouds，multi-layers clouds．Using cloud recognition approach，cloud phase characteristics was further analyzed during two strong rainfall periods occurred in the Tianshan Mountains ridges and surrounding areas in Xinjiang. Distribution of cirrus cloud zone and dense ice cloud zone was consistent，indicating that the cloud phase recognition approach is feasible and reasonable. When combining with real time precipitation，the results show that cloud particles in the rainstorm areas are ice cloud phase. The cloud phase retrieval technique in visible，near-infrared，infrared channels facilitates identification of multilayer cloud，precision of cloud phase retrieval and analysis of weather forecasting. Because of its using data detected in visible，near infrared wavelength，it is limited to daytime using. Moreover，based on low temporal resolution data of polar orbiting satellite，it becomes impossible to trace strong convective weather continually. Therefore，further study will be conducted by combining geo-stationary satellite data with previous research results.

Key words: Tianshan Mountains, storm, spectral characteristics, cloud phase identification