Abstract: Using 24-240 h precipitation ensemble forecast produced by the T213 global ensemble prediction system (EPS) of China Meteorological Administration (CMA) and the precipitation observations between June and August from 2008 to 2011 in the Sichuan basin, we investigate a heavy rainfall calibration method based on ensemble forecast-observation probability matching. The principle of this method is to correct model bias in heavy rainfall values (50 mm) based on a comparison of the probability density distributions of observed vs. ensemble forecasted precipitation amounts. A Corrected Heavy Rainfall forecast value（CALHR）over a model grid point is made by defining an adjustment to the forecast value in such a way that the adjusted cumulative forecast distribution matches the corresponding distribution observed. In particular, the Gamma function is used to simulate the probability density distributions of precipitation and to capture the bias information.This technique is used to perform the calibration of ensemble precipitation forecasts for the Sichuan basin from 28 June to 10 July 10 in 2013.The verified results are analyzed and its limitations are discussed. From this study, it was found that a noticeable systemic bias of precipitation forecasts for the T213 EPS system exists with smaller precipitation amounts than observations. The longer the forecast time is, the smaller the precipitation amounts will be. CALHR value generated by this method is generally smaller than 50 mm, and the longer the forecast time is, the smaller the CALHR values will be. This method is effectively to correct the systemic bias of heavy rain forecast for the T213 EPS system. The heavy rainfall forecasts are improved for yes or no forecast for this method with higher ETs scores and lower missing rate and false alarm rate.