Based on the mesoscale numerical forecast system that is operationally implemented in the Chongqing Meteorological Service, the influence of different terrain smoothing schemes in the WRF model on precipitation forecast is studied. The differences between the static terrain height fields generated without smoothing and with two smoothing schemes (sds and 121) in the WRF model are studied and compared in detail. Based on this, one month consistent experiments using different terrain smoothing schemes were carried out, and a typical heavy precipitation case was selected for further comparative analysis. The results show that there are obvious difference between the static terrain height fields generated by different terrain smoothing schemes, especially for the steep plateau regions or mountain regions while the maximum mean absolute error can be more than 462.56 m; the two terrain smoothing schemes (sds and 121) mainly smooth out the topographic features of small scale in the mode terrain, and in general, the smoothing degree of the 121 scheme is greater than that of the sds scheme. The verification results of the one month consistent experiments show that the model terrain smoothing process has a positive impact on the forecasts of heavy or above rainfall, and the prediction results using the 121 smoothing scheme are better than those using the sds smoothing scheme. The comparative analysis results of a typical precipitation case show that different terrain smoothing schemes will cause significant differences in the prediction of vertical velocity and water vapor flux divergence, and such significant differences will further affect the fall area and intensity of heavy precipitation prediction.