The simulation of a heavy rainfall event that occurred in the middle and lower reaches of Yangtze River was conducted to examine the effects of perturbations of land surface variables (soil moisture and soil temperature) and land surface parameter (vegetation fraction) in the ensemble forecast using the Weather Research and Forecasting Model (WRF) Version 3.2.1 and National Centers for Environmental Prediction (NCEP) reanalysis data. The results showed that land surface variable (or parameter) perturbations have a large impact on short term simulation of rainstorm. The time scale that the land surface variable (or parameter) perturbation affects precipitation is lower than 10 h and the smallest time scale is lower than 6 h. From the point of influence mechanism, disturbance of the land surface variables (parameters) changes the surface latent heat flux and sensible heat flux firstly, which has great impact on the local atmospheric temperature, pressure, humidity and wind field by the land atmosphere interaction, and thus affects the intensity and distribution of the heavy rainfall. The ensemble average result is better than the control forecast, which is more stable and credible than the single ensemble members. The analysis of the precipitation probability forecast can provide some useful information about the precipitation forecast especially to heavy rainfall. Overall, the initial perturbation of land surface variables and land surface parameter perturbations are significant to the initial ensemble forecast.