Aiming at the poor performance of the TS verification method, objectbased verification method is introduced, by which the objects in the two datasets are matched in area, position, morphology and intensity to get potential forecast information of spatial field. Taking EC_THIN, T639, ensemble forecast of WRF and BCSH for example, for the purpose of integrating the characteristics of multimodel, the optimal times about the forecast of each severe precipitation are selected and the boxandwhisker plots are used. The circulation patterns and influence systems of severe precipitation are typed. The results show that the score is the best about the forecast of heavy rainfall caused by tropical cyclone with midlow latitude circulation. The score of lowpressure vortex and shear line of heavy rainfall processes is better, of which the best forecast of model is BCSH and EnMAX. The forecasts of 12 models for pressure trough are ordinary. When the system is extratropical cyclone with heavy rainfall, the score is the worst.