Abstract: Satellite radiation data assimilation can be easily influenced by clouds, and huge amount of satellite data cannot be directly assimilated to numerical model in cloudy cases. One reason is that the initial cloud structure cannot be easily obtained, which is an important input parameter for an observation operator invariational system. Without the data of cloud structure, the observation operator cannot correctly simulate the brightness temperature in cloud condition. Because of the high cloud fractions in the tropics and the close links between clouds and extreme synoptic processes, such as typhoons and heavy rains, highaccuracy threedimensional cloud structures are needed for the assimilating satellite radiation data, especially infrared channels. Based on the GRAPES3DVAR（Global and Regional Assimilation and Prediction Enhanced System）and the simulated cloud structures of CHANCHU typhoon by WRF（Weather Research and Forecasting Model）, which are added to the RTTOV（fast radiative transfer model for TOVS）as input variables to improve the observation operator, the infrared channel brightness temperatures are assimilated into the numerical model and its impact to typhoon forecast are analyzed. Results show that while providing the initial cloud structures, some of NOAA16 (National Oceanic and Atmospheric Administration) HIRS/3（High Resolution Infrared Radiation Sounder) channels, such as channels 4-8,10, and 13-19, can be assimilated properly, the temperature and moisture structure of typhoon can be improved, and the HIRS/3 infrared channels assimilation have positive impact on shortterm precipitation forecast.