A meteorological early warning method for mountain torrent disasters is proposed using the fuzzy evaluation method in this paper. The method is based on the dynamic critical rainfall for mountain torrent early warning that considers soil water content saturation, then a correspondence between meteorological warning levels and fuzzy scores for mountain torrent disasters is established based on the fuzzy evaluation method. Equal weight averaging and weighting algorithms constructed based on the coefficient of determination or the relative error of peak flow are respectively adopted. With this method, together with the comprehensive fuzzy scores for meteorological warning calculated by using the precipitation forecasts from CMA-MESO, CMA-SH9, CMA-BJ and intelligent grid forecasting， the meteorological early warning level is determined. The results show that the hit rate of the meteorological early warning results based on the fuzzy evaluation method is comparable to that of the CMA-BJ and higher than other models, the miss rate and false alarm rate are also comparable to those of the CMA-BJ and lower than other models, the TS scores are all higher than those of other models, through the application and verification of the mountain torrent disaster in Hengshui of Anyang River from 17 to 22 July 2021. This method can extend the lead time of mountain torrent prediction and improve the accuracy of early warning.