The Advanced Regional Prediction System (ARPS) along with its Data Analysis System (ADAS) were used to analyze the improvement of the initial field by assimilating radar radial velocity and reflectivity data. A local severe convective event that occurred on 21 August 2012 in Jiangxi Province was then simulated and studied by using the adjusted initial field in the Weather Research and Forecasting (WRF) model. The main results are as follows. (1) Both the mesoscale wind fields and cloud, moisture contents of the initial field were improved by assimilation of radar radial velocity and reflectivity data. The diabatic initialization of ADAS could adjust the temperature, moisture and wind fields to balance the dynamic and thermodynamic process of the initial field. (2) Assimilation of radar radial velocity made an important contribution to the improvement of the initial dynamic field, but it had little effect on atmospheric hydrometeors and precipitation forecast. Assimilation of radar reflectivity played a role mainly in adjusting the initial hydrometeors and thermal fields, which efficiently mitigated the spin-up time of the model and significantly improved the quantitative rainfall forecast. Assimilation of both radar radial velocity and reflectivity could result in a rapid development of the mesoscale horizontal wind convergence and vertical motion, adjusted temperature and moisture distributions reasonably in the initial field, and improved quantitative precipitation forecast, especially for 3-hour forecast. (3) The results of simulation show that the three dimensional structure of the convective storm and its evolution have been successfully simulated by assimilating radar radial velocity and reflectivity. The convergence and uplift force of the shear line at the middle and lower levels is one of the major dynamic mechanisms for the organization, development and maintenance of the convective storm. The convective condensation and associated latent heat release played an important role in the development of the convective storm. Therefore, radar data assimilation is of great importance to improve the accuracy of the numerical weather prediction and the model ability for the simulation of severe convective weather.