Based on the GSI assimilation system and WRFARW model, a severe rainfall event which occurred in southern China for the period in 17-18 April 2016 was simulated by the method of rapid update cycle. The experiments used radar radial wind data and other conventional data in assimilation cycle. Several kinds of forecast variables were analyzed to find how the assimilation cycles and difference data would influence the forecast result. The results showed that despite the limitation of the single type of observation, there is a certain improvement effect on the false precipitation prediction by the use of rapid update cycle. Radar radial wind data and conventional data mixing assimilation could improve the humidity and precipitation prediction skills, especially at heavy rainfall levels ETS score. Although the simulated rainfall peak value is less than real observation, the assimilation of radar radial wind data could effectively improve the vertical velocity of the squall line, so the occurrence time and intensity of heavy rainfall are much closer to the real observation.