Following the “frequency matching” concept, two approaches have been tested to improve ensemblebased precipitation forecasts for the case of Beijing “7.21” (21 July 2012) severe rainstorm and the period of rainy season from 1 May to 31 July 2010 in China. One approach is “probabilitymatched ensemble mean (PM)”, which uses ensemble memberbased precipitation frequency to calibrate the frequency of simple ensemble mean to correct smoothing effect caused by ensemble averaging; another is “bias correction”, which uses observed precipitation frequency to calibrate the precipitation frequency of each ensemble member to remove systematic forecast bias caused by model deficiency. The results show that (1) PM can effectively correct the problem of “toowidespread light precipitation and toomuchreduced heavy precipitation” in a simple ensemble mean, especially so for extremely heavy rains. The larger ensemble spread is, the more improvement will be achieved. However, PM can barely improve total precipitation amount forecasts and has no ability to correct the systematic bias of the model. (2) Although bias correction is not effective in correcting precipitation position errors, it can effectively remove systematic bias in precipitation amount forecasts and, consequently, improve ensemble spread and probabilistic forecasts significantly. Bias correction of ensemble mean forecast needs to be done independently rather than through biascorrected ensemble members via simple averaging. (3) PM approach also works well even after each ensemble member’s bias has been removed in advance. Therefore, both PM and bias correction need to be done jointly to maximize benefit through model forecast postprocessing in an operational environment.