In the present study,based on the maximum and minimum surface air temperatures of 587 stations in China,ERA-interim reanalysis soil moisture(SM) data and the extended reconstructed sea surface temperature(ERSST) data,the possible linear connection between the previous(spring and winter) sea surface temperature(SST),soil moisture and summer extreme temperature in China during the period of 1979-2009 is investigated.Statistical summer extreme temperature prediction models were developed using Barnett-Presisendorfer Canonical Correlation Analysis(BP-CCA),the Ensemble Canonical Correlation(ECC) method,and newly defined summer extreme temperature indices.In addition,the scores for the prediction models with predictive skill in the independent sample tests were calculated.The results showed that the summer extreme temperature in China has a close relationship with the previous SST and SM anomaly,while the spatial distribution of the previous SST anomaly was similar to the PDO pattern,and the previous SM anomaly in southern China,the Tibetan Plateau,northeastern China and the western part of northwestern China.The cross-validation tests showed that the BP-CCA models based on the previous winter predictors had higher predictive skill than those based on the previous spring predictors.Meanwhile,the BP-CCA models based on SM had higher predictive skill than those based on SST.Additionally,the independent sample tests showed that the ECC models based on the previous SM and SST had high predictive skill for summer extreme temperature in China.The study results revealed that the previous soil moisture and sea surface temperature contain valuable signals for summer extreme temperature in eastern China,and can be considered for inclusion in summer extreme temperature prediction operations.