Based on the six outputs of models estimated by the multimodel downscaling ensemble prediction system (MODES), temperature and precipitation were evaluated in the prediction performance and analyzed by using the symbol consistency method (Pc). According to the results of the evaluation, the equalweight averaging scheme and the optimal scheme are used to statistically integrate the six model products for local application from 2012 to 2015The results indicate that the prediction performance of the model products is different, and the prediction performance of ECMWF is higher than that of NCC and NCEP. By comparing the two schemes, the scores of the optimal scheme are higher than the average score of six outputs and the equalweight averaging scheme no matter temperature or precipitation. Meanwhile, the scores of the optimal scheme are higher than those of the released prediction products. Therefore, using the optimal scheme can effectively improve the effect of MODES on predicted performance.