Abstract: The moist physical processes (mutual transformation of water substance), cloud and rain formation, and corresponding dynamic and thermal effect on precipitation forecast is important for the meso-scale and micro-scale numerical weather prediction model. In order to improve the cloud microphysical parameterization of GRAPES_Meso model, the surface cloud observation data, satellite image and Doppler radar reflectivity are analyzed, and the GRAPES Cloud Analysis System is developed referring to ARPS model and ARPS Data Analysis System (ADAS).The progress and principle of cloud analysis in meso-scale numerical weather prediction model is described at first, and then the technical feasibility of leading cloud analysis into GRAPES_Meso model is analyzed, based on which the cloud analysis module is designed. There are several monomial sub-modules in the GCAS (GRAPES Cloud Analysis System) such as the observation pre-treatment sub-module, 3D cloud cover analysis sub-module, humidity and moisture contents filed analysis sub-module. The special pre-treatment process has been designed for domestic observation data. The GRAPES Cloud Analysis System is developed for the first time which is based on LAPS (Local Analysis and Prediction System) cloud analysis program and referring to ARPS model and ADAS. The adjustment of the 3D cloud cover and initial field and cloud water and hydrometeors is analyzed when surface cloud observation data, satellite image and Doppler radar reflectivity is assimilated in the GRAPES_Meso. The performance of the GRAPES Cloud Analysis System is inspected by the simulation of the landing Typhoon Molave (0906). First, the distribution of 3D cloud cover is retrieved from the cloud analysis system. Second, the cloud water and hydrometeors is produced quickly based on the 3D cloud cover and the height of cloud top and bottom, also, the initial field is adjusted. Third, the forecast of typhoon track in the simulation has a little error, however, the cloud analysis system has displayed positive effects on typhoon landing sites and typhoon track and intensity forecasting compared to the controlled experiment. Finally, the feasibility and correctness of cloud analysis is verified by analysing the radar reflectivity derived from cloud analysis system, indicating the performance of total precipitation forecast is improved obviously.The feasibility of Cloud Analysis System and the assimilating capability are discussed. The System is suitable for operational run for its low computation cost and hot-swap support. Further more, there are some questions which deserve to be more widely debated, for example, Cloud Analysis System should match with the model, region and resolution. The quality control of Doppler radar reflectivity factor data, coordination between initial moist thermal and dynamic field are also very important.