The non-stationary characteristics of the average runoff in driest days are tested in this paper based on the measured runoff data of Tangnaihai Hydrological Station from 1957 to 2018 using the 5-year moving average and Mann-Kendall abrupt change test to analyze the effect of climate change on the non-stationary of hydrological extreme value. According to the Kendall rank correlation analysis method, the climate index was selected, and the non-stationary generalized extreme value distribution model (generalized extreme value, GEV) was constructed with time and climate index as covariates. Parameter estimation and model selection were carried out, and application effects of the stationary and non-stationary GEV models in the runoff simulation of dry season were compared. The results indicated that the dry season runoff shows obvious non-stationary characteristics. Meanwhile, the simulated value of the stationary GEV model is higher than the measured value, and the GEV model with the western Pacific index as covariate produces better fit for the extreme value, which better explained the volatility of extreme dry events.