Optimal weighted model for ensemble forecast of the surface air temperature in mainland China and its regional applications

Based on Coupled Model Intercomparison Project,Phase5(CMIP5) muti-models, an optimal weighted model for ensemble forecast (Op-SE) of the surface air temperature in mainland China was presented. In order to assess the capability of Op-SE, it was compared with equally-weighted ensemble (EE) and superensemble (SE), and anomaly correlation coefficient (ACC) and root-mean-square-error (RMSE) were chosen to evaluate their forecasting skills. As shown from the results, ACC between Op-SE and observation is optimal in most of China, especially eastern China, which indicates that ACC has passed the significance test at 0.5 level. However EE has poor performance in ACC. As for RMSE, EE is also relatively weak. And Op-SE is better than SE in eastern China, while SE is better in a few other areas like Sichuan Basin. The comprehensive assessments of ACC and RMSE show that the forecast skill of Op-SE is best in northeast, north, eastern, southwest and northwest of China, but is not good enough in Sichuan Basin and southern Gansu. In conclusion, Op-SE provides a new method on selecting outstanding models into ensemble climate forecast, which can improve the forecast skill in regions to some extent.