Based on the meteorological data of three different models in CMIP5, the mutual information method was used to selectthe forecasting factors and the RBF neural network model was used to predict the changes of snowfall runoff in the western mountainousareas of Tianshan Mountains under different climate scenarios (RCP2.6, RCP4.5, RCP8.5). The results show that: (1) Thefuture runoff will continue to rise during 2020-2030, and be stable in 2060. In additon, the future runoff increases significantly innon-flood season and decreases in flood season; (2) The main factors affecting the runoff in different scenarios under differentmodes were analyzed by the grey correlation method. Besides, the future changes of the relevant factors were analyzed. It is foundthat the main reasons for the increase of runoff in non-flood season and the decrease of runoff in flood season are that the precipitationincrease while the evaporation decrease or increase slightly in the non-flood season, and the precipitation decrease while theevaporation increase with the temperature in the flood season.