Abstract: The conditional probability density functions of daily precipitation in 16 stations of Guangxi in Summer are estimated and then the probability distributions of maximum amount of daily precipitation in 1~20 days are recurrented by using daily amount of precipitation data set. The results show that the Gamma distribution can be accurately approximated to sample frequency distribution of daily precipitation given wet preceding day or dry preceding day. For the most stations the probabilities of maximum amount of daily precipitation in 10 days achieve their largest value (about 0.2) at 20~30 mm, while in 20 days at 30~50 mm. The probabilities at which the maximum amount of daily precipitation in 10, 20 days in the stations of Beihai and Qinzhou are more than 100 mm are evidently more than those in other stations.