Thunderstorm clouds were identified and tracked by K-means clustering algorithm and forecasted through the kind of linear extrapolation prediction. Based on the statistical relationship of Z=AIb between radar reflectivity factor Z and precipitation intensity I, Z values measured by radar had been compared with those data of the corresponding ground actual rainfall intensity. By regression statistical analysis, the coefficients of A and b under different rainfull can be deduced respectively and the estimation equation was obtained. The identifying and tracking method about thunderstorm clouds was described in detail by selecting Nanjing radar data about the precipitation process on September 18, 2007 and the forecast result was also verified. The result indicated that thunderstorm clouds were able to be identified, tracked and estimated with extrapolation by this method and the short-time precipitation was accurate.