The weather research and forecasting (WRF) model along with its three-dimensional data assimilation (3D-Var) system was used to improve the initial field by assimilating radar reflectivity and radial velocity from Doppler radars. A rare strong squall line occurring on 4 March 2018 in Jiangxi Province was simulated and studied by using the adjusted initial field. It is found that only assimilation of rainwater, snow, graupel particles retrieved from radar reflectivity and water vapor derived from radar reflectivity could not make a stable improvement in forecasting the composite reflectivity, especially got an opposite effect on forecasting surface gale and precipitation. The assimilation of reflectivity data could significantly improve forecast skill when radar radial velocity was assimilated jointly. The reason is that assimilation of radar reflectivity can efficiently adjust initial hydrometeors and thermal field, but it has little effect on initial dynamic field. As the simulation time went by, the adjustment of dynamic field was unreasonable, a false divergence wind field appeared in the upper troposphere, thus a stratiform cloud area appeared in front of the squall line, which was not captured in reality. So the model could not improve the simulation of vertical wind shear and cold pool as well as rear inflow, and then there was a large gap between forecasted results and observations. Only assimilation of radar radial velocity could improve simulation results, and assimilation of both could efficiently adjust initial hydrometeors and thermal fields as well as dynamic field, and make the physical configuration more compatible with reality.The vertical wind shear and wind field structure were more favorable to occurrence of the strong squall line, and then formed a strong cold pool close to reality, further the forecasting results of radar composite reflectivity, surface wind and precipitation of the squall line agreed much better with observations compared with only assimilation of radar radial velocity.