Global warming has intensified hydrological extreme events and resulted in disasters around the world. For disaster management and adaption of extreme events,it is essential to improve the accuracy of extreme value statistical models. In this study,BayesTheorem is introduced to estimate parameters in the Generalized Pareto Distribution (GPD) model which is applied to simulate the distribution of monthly average runoff minima during dry periods in mountain areas of Urumqi River. BayesTheorem treats parameters as random variables and provides machinery way to convert the prior distribution of parameters into a posterior distribution. Statistical inferences based on posterior distribution can provide a more comprehensive representation of the parameters. An improvedMarkov Chain Monte Carlo (MCMC) method,which can solve high-dimensional integral computation in the Bayes equation,is used to generate parameter simulations from the posterior distribution. Model diagnosis plots are made to guarantee the fitted GPD model is appropriate. Then based on the GPD model with Bayesian parameter estimates, monthly average minima corresponding to different return periods can be calculated. The results show that the improved MCMC method is able to make Markov chains converge at a high speed. Compared with the GPD model based on maximum likelihood parameter estimates,the GPD model based on Bayesian parameter estimates obtain more accurate estimations of minimum monthly average runoff. Moreover,the monthly average runoff minima in dry periods corresponding to 10 a,25 a,50 a and 100 a return periods are 0.60 m~3 /s,0.44 m~3 /s,0.32 m~3 / s and 0.20 m~3 /s respectively. The lower boundary of 95% confidence interval of 100a return level is-0.238 m~3 / s,which implies that Urumqi River is likely to cease when 100 a return level occurs in dry periods.