Abstract: Based on the Desert Oasis Gobi Regional Assimilation and Forecasting System (DOGRAFS), four experiments are designed for a heavy precipitation occurred in the Yili vally of Xinjiang from 31 July to 1 August 2016, and a continuous test in July 2016, to investigate the impacts of microwave radiance data assimilation on rainstorm forecasts in Xinjiang. Four experiments were conducted, including (1) conventional synoptic and sounding data only (Ctrl), (2) a combination of conventional and AMSU-A data (Exp1), (3) a combination of conventional and MHS data (Exp2), (4) a combination of conventional, AMSU-A and MHS data (Exp3). The results showed that the adjustments to the initial temperature, geopotential height and humidity were notably larger by assimilating AMSU-A data (Exp1 and Exp3) than MHS data (Exp2). The adjustments to temperature and geopotential height were positive in low level and negative in middle-high level, while there were several negative centers of humidity adjustments focusing between 800-600 hPa. However, assimilating MHS data had a very weak impact on the temperature and geopotential height and a fine-tuning to the humidity. From the precipitation forecasts and objective verifications, assimilating MHS data had an advantage on the rainfall area and large threshold rainfall while the AMSU-A data had a negative effect.