Based on temperature forecasts of regional ensemble forecast models from six dif ferent operational centers and scientific research institutions, a study of bias correction and multimodel consensus has been done to reduce the system error and to improve the forecast precision. The objective and quantitative verificat ion results show that the forecast precision could be improved observably with t he biascorrection method. Both of the two methods used to be build the co nsensus forecast equations show a great improvement on forecast ability too. Cons ensus forecasts are more precise than any single model. And between the two kind s of consensus forecasts, the multiple regression analysis shows a better effect than the arithmetic average method.