Abstract: To study the causes of mountain flood geological disasters in Nyingchi, Tibet, the authors analyzed several major impact factors such as precipitation, land use, soil type, slope and vegetation index. Combined with the environmental geology and the spatial and temporal distribution characteristics of precipitation, Logistic regression method was used to establish the probability zoning of the disasters. Based on the daily precipitation data of the disaster day and six days before the disaster day, an early warning model of mountain flood geological disasters was obtained by the analysis of the Logistic regression. The results showed that the probability of disasters in most areas of Nyingchi was less than 40% , while the remaining area covered approximately 14.8% (18000 km2). We applied this early warning model to test the previous statistical data of mountain flood geological disasters, and found that the prediction accuracy of the occurrence and nonoccurrence of the disasters was about 87.2% and 97.7%, respectively.