For the purpose of applying the surface cloud observation data, satellite images and Doppler radar reflectivity to the GRAPESMeso (Mesoscale of the Global and Regional Assimilation and Prediction System) model， the GRAPES cloud analysis system has been developed for the first time which is based on the LAPS (Local Analysis and Prediction System) cloud analysis program and referring to ARPS Data Analysis System (ADAS). The adjustments of the 3D cloud cover and initial field and cloud water and hydrometeors have been analyzed when the surface cloud observation data, satellite images and Doppler radar reflectivity are assimilated in the GRAPESMeso. A heavy precipitation lasting from June 28 to July 4, 2009 in Hunan is simulated, and the results show： (a) the distribution of 3D cloud cover has been retrieved from the cloud analysis system; (b) the cloud water and hydrometeors have been produced quickly based on the 3D cloud cover and the heights of cloud top and bottom, and also the initial relative humidity field has been adjusted; and (c) the duration of spinup time has been remarkably decreased, and the forecast precipitation has been significantly increased in the early hours, the performance of 24 h precipitation forecast has been also improved obviously.