By using the detailed data of 1464 daily heat stroke cases and daily meteorological factors from 1994 to 2006 in Wuhan , the linear and nonlinear correlation coefficients between the daily number of heat stroke cases and meteorological factors are calculated, the key factors are selected, the nonlinear models are established, and the meteorological grades for heat stroke are divided. The results show: 1) The air temperature is the key factor which leads to the heat stroke events, and the accumulation of meteorological factors in or over three days will be more effective for numerous heat stroke cases; 2) nonlinear correlation will be more suitable for the number of heat stroke cases and meteorological factors. The nonlinear models are established between the daily average number of heat stroke cases and the mean temperature, the minimum relative humidity in the preceding three days (including the present day). The standards of five grades are worked out according to the daily average number of heat stroke cases, the test indicates a good result through historical and independent samples.