Radar data are assimilated into numerical prediction models to improve precipitation forecasts using three-dimensional (3D) variational (VAR) and hybrid data assimilation (DA) methods. The effect of radar DA (RDA) and prediction accuracy according to DA methods are evaluated by conducting experiments of three heavy rainfall events and analyzing the spatial pattern of accumulated precipitation, profiles of hydrometeors, and cloud microphysical processes. Increments of the initial analysis revealed that the hybrid method simulated the convective band, wind convergence in front of the cold front and a higher amount of water vapor mixing ratio compared to 3DVAR. Based on numerical experiments, the root mean square error of the total cumulative precipitation in 3DVAR was 9.68 mm lower than that of the no-RDA experiment (CTRL), whereas that of the Hybrid was 11.12 mm lower than that of CTRL. RDA improved precipitation forecasts, and Hybrid showed better precipitation forecast accuracy than 3DVAR. Although RDA indirectly changed the water vapor, rain, snow, and graupel mixing ratios, the change in the water vapor amount was the major factor affecting microphysical processes. The increase (decrease) in the water vapor mixing ratio had the greatest impact on precipitation formation (dissipation), whereas the effect of snow, rain, and graupel mixing ratios was relatively small. The results indicate that it is critical to create an environment in which water vapor can transform into precipitating hydrometeors and the hydrometeors can grow through cloud microphysical processes. The accuracy of heavy rain forecast can be improved by hybrid DA method, which can consider the model error in real time, making it more adaptable to the day-to-day changing weather conditions.