In order to study the effect of MM5 Adjoint-model Assimilation System on the improvement of rainfall prediction and that of cloud drift wind data on improvement of heavy rain prediction,the MM5 adjoint-model assimilation system is used to assimilate conventional observation data and CDW (cloud drift wind) data in a case of heavy rainfall taking place from 14th to 15th June 2002 in the south of the Yangtze River.Simulation experiments of different schemes are performed.The results show that the MM5 Adjoint-model Assimilation System had the better effect on improving rainfall prediction;and GMS-5 CDW data improved the quality of the initially upper wind and meso-and micro-scale weather systems,thus the quantity of precipitation forecast,especially heavy rainfall forecast.