Abstract: The main purpose of this paper was to presenta new method for interpretation of the products of ECMWF Ensem ble Prediction System(EPS). The basic assumption was that the EPS member srepresent, in a certain degree, the future weather situations. In order to resolve the problem that the synoptic scales are not predictable in the medium range, the conception of weather type was introduced. The approach for automatic clas-sification was Dynamic Fuzzy proposed by Diday, and a new distance called Displacement and Maximum Correlation was adapted. During the it eration, the initial gravity was defined by the weather types. In China, the weather types in summer and winter were determined by the automatic classification. The premier applications showed that EPS may predict the at mospheric evolution in winter, but the predict ability in summer was relatively poor.