Abstract:Based on the monthly Surface Air Temperature (SAT) of 160 stations in China and the ERA-Interim monthly surface soil moisture reanalysis data from ECMWF, the yearly increment of soil moisture of nine key regions over Eurasia was selected as the predictors through correlation analysis, the Barnett-Preisendorfer Canonical Correlation Analysis (BP-CCA) combining the Ensemble Canonical Correlation analysis (ECC) method were used to develop the ensemble prediction model in order to predict the yearly increment of summer SAT over eastern China, and finally carry out the prediction of summer SAT. The data during 1980-2004 and 2005-2014 are used to perform the historical prediction test and the independent sample test, respectively. Firstly the single factor prediction models of nine predictors were set up with the BP-CCA method, then the ensemble prediction models with the ECC method based on different combinations of the nine predictors were developed, and the predictive skills are also analyzed. Results show that the combinations of different predictors have different predictive skills for summer SAT in China. The soil moisture in the lower reaches of the Lena River, south of the Yellow River, lower reaches of the Yenisei River, West Siberian Plain, and Northwest of the Indian Peninsula, have good predictive effect for summer SAT in North China. The soil moisture in the South of the Yellow River, lower reaches of the Yenisei River, Northwest of the Indian Peninsula, Northeast of the Baikal Lake, and West of the Baikal Lake, have good predictive effect for summer SAT in the Yangtze-Huaihe region. Both of the two ensemble prediction models show good actual predictive skills. The coherence rate between the observed and the predicted SAT anomaly over North China and Yangtze-Huaihe region, is 8/10,7/10, and the average relative root mean square error is 3.4%, 2.7%, respectively. Prediction Scores (PS) of SAT in North China and the Yangtze-Huaihe region by the two models all exceed 80 points, and the generally used anomaly correlation coefficients (ACC) all exceed 0.3, which means that soil moisture implies useful signals for summer SAT in China, and soil moisture can be considered to apply in the summer SAT prediction operations.