Abstract: Using the hourly rain gauge data from 157 national automatic weather stations (AWS) over Sichuan province between May and September from 2010 to 2016, by applying the Pearson-Ⅲ distribution function, we have conducted a fitting of the hourly rainfall in Sichuan to find out the spatial probability distribution that exceeds different thresholds. Based on the results, we calculated the probability distribution pattern and their extreme values under different return period scenarios of hourly rainfall. The results indicate that rainfall with low frequency but with high grade of hourly rainfall occurs the most often along mountains in the western Sichuan Basin. The rainfall that occurs frequently but with the high grade of hourly rainfall seldom happens in the east part of Panxi region. High value areas of hourly rainfall within a recurrence period of 50 years are located in the northern part of Leshan, in the junction between northwestern Suining and Mianyang, and in the northern part of Dazhou, whose the maximum value can be over 60 mm. The extreme hourly rainfall within a recurrence period of 100 years is up to 70 mm, with a distribution similar to that with a recurrence period of 50 years. The deviation coefficient (C_v) of the Pearson-Ⅲ distribution function to hourly rainfall has a logarithmic decreasing correlation with the altitude of AWSs, the statistic coefficient (R2) between them being up to 0.6545, indicating that terrain has an effect to the hourly rainfall patterns significantly. In addition, the k-means clustering method can be used to group hourly rainfall over Sichuan province.