Abstract: Based on the second generation model system data (DERF 2.0) of National Climate Center System, the historical return inspection work in early summer (June) of Heilongjiang Province from 1983 to 2013 was investigated. Through empirical orthogonal function method (EOF) and optimal subset regression (OSR), we analyzed the prediction results of the model downscaling interpretation application, and used short- term climate prediction integrated service score (PS) and anomaly correlation coefficient (ACC) method to evaluate the prediction performance of model downscaling in early summer precipitation of Heilongjiang Province, and compared the results with that of model direct prediction factor. The results showed that the prediction of DERF 2.0 model downscaling was much better than that of the direct prediction of model. In addition, the application of DERF 2.0 model downscaling interpretation had moderate skills for predicting extreme drought and flood years in 2004 and 2009.