Based on the summer (from June to August) precipitation data from 29 m eteorological stations distributed over Northern Xinjiang, the precipitation ind ex is calculated. Adopting assemble regression method constructed by two stepwis e regressions, including the initial factors constructed by slidingcorrelation method on account of the correlation coefficient stability, it is attempted to p redict the summer precipitation index along with the factors of the previous mon thly atmospheric circulation indexes. While the sample number is 47, the correla tion coefficient between the original data and the simulated data is as high as 0.7248. There are 10 prediction models to be set up with different sample size r espectively, and the independent test is analyzed. The summer precipitation is t o be lower in 2008, of which the index is 106.9. The results show that the stati stical predict model based on the integration of sliding correlation, stepwise r egression and assemble prediction has a certain prediction ability. The analys is method synthesizing regression and time series could be applied widely to sho rtterm climatic prediction operation.