Abstract: In this paper, a general Gamma probability model is built for the precipitation amount in various statistical periods based on the data from 5 long-record stations selected. An improved maximum likelihood method to compute the models parameters is also presented. By means of some empirical curve fitting, it is found that the shape and scale parameters of the model are related to the statistical periods (number of days) and both the parameters and number of statistical days possess the relationship of the quadratic curves. In addition, it is also shown that there are significant seasonal variations in these parameters, which reflect the characteristic of the monsoon climate in China.