Based on the monthly precipitation data of 52 observational stations in June, July and August from 1961 to 2000, the features of summer precipitation in Guizhou Province were analyzed by means of abrupt change analysis and wavelet analysis techniques. A prediction model was developed and the prediction experiment of summer precipitation in Guizhou Province was conducted with the mean generating function. The results indicate that summer precipitation in Guizhou increased obviously in recent 40 years and its inter-annual variation was remarkable with varying periods: 8-10, 4-6, 4, and 2 years. Abrupt change mainly occurred in 1971 and 1994, and precipitation began to decrease from l994. The result shows that the mean generating function has good prediction capability for short-range climate changes, especially for extreme precipitation events.