A statistical downscaling system for forecasting summer precipitation at stations in China has been established in this study on the basis of real-time prediction of numerical products from the Climate Forecast System (CFS) and observational data. The summer 500-hPa geopotential height in the current year from CFS and the previous autumn-winter sea surface temperature from observations were selected as the two predictors, with corresponding key regions of Pan-East-Asia and the tropical Pacific, respectively. The statistical downscaling hindcast on the 1982-2001 summer precipitation over China improved the performance of the prediction compared with that of the original CFS. The spatial anomaly correlation coefficients increased from 0.03 to 0.31, and the temporal correlation coefficients over most parts of China also increased significantly by the downscaling scheme with a maximum of 0.6. The root mean square error decreased in comparison with the output of the original CFS. Furthermore, we successfully created a hindcast on the 2011 summer precipitation anomaly pattern in China by using this statistical downscaling scheme.