The T639L60 model data of 2010 are used as the forecast fields, and the NCEP data from 1980 to 2009 are used as the historical sample database. The analogue examples of T639L60 model forecast fields are picked out from the sample database, based on statistical forecast with dynamic analog, and the correction scheme of T639L60 model data is designed with the multiple regression method. The correction effectiveness of the scheme is tested using the T639L60 model data from January 2011 to June 2011 The results show that the correction effectiveness of the forecast fields over three days in China is obvious, based on statistical forecast with dynamic analog, and the correction effectiveness of synoptic situation fields is better than element fields.