Based on the rainfall data taken from the Public Meteorological Service Center, China Meteorological Administration (CMA), the calibration of the precipitation forecast is performed by using Frequency Matching Method (FMM) and the threshold method. Then the performance of calibration is evaluated by Mean Absolute Error (MAE), Test Score (TS), etc. The results in the whole 2015 show that: (1) the calibration can significantly reduce the occurrence of false alarms of light rain and improve the qualitative forecast for a "rain or no rain" event. (2) The calibrated precipitation forecasts are much closer to the observation, and the MAE and area bias (dry bias or wet bias) of the ensemble system are also greatly reduced. (3) The improvement by the calibration method is related to the forecasting performance of the ensemble system itself. The greater the forecast errors, the better the calibrated forecast skill.