Abstract: Based on the daily 12-36 h, 36-60 h, 60-84 h, 84-108 h, 108-132 h and 132-156 h ensemble precipitation probability forecasts over China (0 -60 N, 70 -140 E) from May 25 to August 31 during 2007-2011 from the global ensemble models of CMA, ECMWF, JMA,NCEP and UKMO taken from the TIGGE archives, we assembled the precipitation probability forecasts from each of the above 5 centers in 2011 by the bias-removed ensemble mean (MBRE) and the traditional PoorMan techniques (POOL), and then estimated their forecast skills by calculating the Rank Probability Score (RPS) and the Brier Score (BS). The results show that the multi-model ensemble precipitation probability forecast technique has a higher forecast skill than any single model, and the MBRE technique is better than the POOL technique for the 12-156h precipitation forecasts. It is also found that both POOL and MBRE improved considerably the probabilistic forecasts compared toa single model for all light, moderate and heavy rain events. Furthermore, MBRE has a higher forecast skill than POOL.