Abstract: A study of a relatively long period (12 -24 h) regional thunderstorm forecast is carried out based on the GFS analysis and the forecasting data from NCEP (National Centers for Environmental Prediction). The prediction factors in this study come from the convective indices that are often used to examine the dynamic and thermodynamic environments of thunderstorm occurrence. With the highest CSI (critical success index) as a criterion, the discriminant analysis method and stepwise regression method are used to build the optimal thunderstorm model for each grid in the forecasting region. Therefore, thunderstorm models and the corresponding critical values for the discrimination of the thunderstorm occurrence would change with the location so that the local environment and climate background of the thunderstorm occurrence can be considered. The newly-built thunderstorm forecast method that ensembles the convective parameters (called "ECTF" method) is used for the prediction of the two thunderstorm processes which respectively occur on June and September 2012. The results show that this method can basically predict the area of thunderstorm in East China. Further, to discriminate occurrence of thunderstorm and torrential rain effectively and objectively, the "Ensemble Dynamic Factors precipitation Forecasting method (EDFF)" is introduced. The EDFF method performs well on the diagnosis and tracing of severe precipitation, while the ECTF method is more capable of reflecting the convective weather, such as flash floods, hail, high wind and so on. Extracting the advantages of these two methods, they are integrated and an "integrated dynamic factors-convective parameters severe weather forecasting method" is built, which can be used in the forecast of precipitation and thunderstorm and can also discriminate, to some extent, among the precipitating thunderstorm, the non-precipitating thunderstorm and the non-thunderstorm precipitation.