Abstract: A simple analysis scheme proposed by Brewster K in 1996 is used to retrieve cloud microphysical variables such as water mixing ratio and rainwater mixing ratio of cloud, as well as humility variable of the model air from Doppler weather radar's reflectivity. The initial fields of meso-scale numerical model derived from model's data assimilation system can include cloud microphysical messages and show where the air is saturated within model space. A numerical experiment for a rainfall process near Maanshan City of Anhui Province in China on June 19 of 2002 is performed by the Advanced Regional Prediction System (ARPS) based on this scheme. The background fields and boundary conditions come from the globe model's output of AVN in NCEP, and a 600 km×600km domain is setup with 2km×2km horizontal resolution, and 21 layers in vertical. Results show that 1 hour's forecast of the atmospheric composite reflectivity with ARPS matches the detailed echo picture observed by the radar at target time, and the non-uniform distribution of precipitation with some rainfall clusters matches the non-uniform distribution of echoes with several strong clusters. The clouds represented by cloud microphysical variables can be tracked both in horizontal and vertical directions from initial time to the time of forecast, and their changes are consistent with the observed radar reflectivity. Also some of the meso-scale wind structures with convergence and divergence appear near the strong echoes, which result from atmosphere thermodynamic adjustment to the clouds. Comparatively, the forecast doesn't match the observation if not considering such scheme including radar reflectivity within data assimilation system for initial field of the model. So it seems that this analysis scheme used here would be helpful for numerical nowcasting of the precipitation.