In this study,based on the European Centre for Medium-Range Weather Forecasts (ECMWF) of 2 m surface air temperature and automatic weather observatory data of China,the correlation among the initial field bias,historical bias,Kalman filter predicted bias the real-time bias is analyzed.Four daily maximum and minimum temperature forecast regression schemes are designed.A comparison among ECMWF,CMA and provincial forecasting is investigated.The results show that the improved scheme's predicted temperature,historical bias,initial field bias and Kalman filter inversion bias as the predictor are all optimal.In addition,the scheme's forecast quality for the maximum and minimum temperatures in 2017 is significantly better than that of both ECMWF and CMA,particularly in areas featuring complex terrain.Compared with the best-performing provincial forecasting,the maximum temperature MAE is 8.24%~13.97% lower than the provincial forecast,while the forecast accuracy is increased by 1.24%~3.57%,and the daily minimum temperature MAE is 9.43%~17.69% lower than the provincial ones.The forecast accuracy rate increased by 1.77%~2.72%.Additionally,it shows the greatest improvement within 1 day,thus indicating that the correction effect decreases considerably with the forecast lead time.