The authors used the plot data and T-logP data of Micaps (Meteorological Information Comprehensive Analysis and Processioning System) to analyze thunderstorm precipitation and non-thunderstorm precipitation in Hangzhou, Quzhou, and Taizhou. The prediction factors in this study that can tell the differences between the environments of thunderstorm precipitation and non-thunderstorm precipitation are convective available potential energy, temperature differences between 850 hPa and 500 hPa, K index,and 2-m height temperature. An index has been formulated from analysis of parameters derived from proximity soundings and decision trees. The quality of this index is examined with precipitation activities for the period from 2004 to 2013 and 2016 using ECMWF fine grid forecast data and observation data. The results show that the threat score (TS) is more than 0.53, the hit rate is 71% and the false rate of thunderstorm forecast (43%) is higher than that of non-thunderstorm precipitation (10%) for most areas and seasons. This method basically can also predict the areas of different types of precipitation from examining two thunderstorm processes which occurred in the spring and summer of 2016, respectively. In summary, the advantage of this method is reflected in the fact that it not only can make fine precipitation classification forecast but also can create expressive prediction for medium and long-term forecast.