Abstract: CINRAD/SA Doppler radar data and ARPS data analysis system (ADAS) are used to adjust the initial field, then the adjusted initial field is inputted into the mesoscale model WRF (The Weather Research and Forecasting Model) to simulate two rainstorm processes. The two rainstorm processes produce a flash flooding over Huaihe River Basin during Meiyu of 2003. According to the difference of initial data to be used in the numerical simulation, four experiments are designed and conducted in the study. The initial field obtained without use of radar data in control run (Exp CNTL), with use of radial velocity in Exp RADAR-V, with use of radar reflectivity in Exp RADAR-V, with both radar reflectivity and radial velocity in Exp RADAR-RV. The simulated outputs of these four experiments are compared and verified. The results show that after the adjustment of initial field using radar data through the ARPS complex cloud analysis scheme, both wind field and moisture field are adjusted at the region of radar observation, and the initial fields carries more information about moisture, cloud, latent heat, meso-scale circulation, and so on. At the same time, the simulation results also show that the spin-up time of meso-scale model is shortened and accuracy of rainstorm simulation is improved. Through the quantitative verification of the simulation results, both the ETS and the HSS of experiments simulating the radar data in initial fields are proved higher than those of experiments not simulating the radar data in initial fields, and the BS verification score for the former experiments is more closer to 1.Based on the comparing of the results of these four experiments, it can concluded that the impacts of Doppler radar data assimilation on initial field are different between Exp RADAR-V and RADAR-R. In ADAS system, radar radial velocity observations are used to adjust the wind field, but radar reflectivity observations are used to adjust the thermal and moisture fields. The significant differences exist between simulation results of four experiments. On the basis of the quantitative verification, the Exp RADAR-V, which uses radar radial velocity, gets the highest score and is able to simulate meso-scale convective system and its rainfall more accurately. However, it is noticed that the other experiment, such as Exp RADAR-R and Exp RADAR-RV, also can improve the accuracy of rainfall simulation at a certain extent, though not as good as the results of Exp RADAR-V. In addition, the improvements of precipitation prediction are notable in the first six-hour simulation.