Normal non-stationary hydrological frequency analysis methods are prone to subjectivity when selecting covariates and establishing the functional relationship between statistical parameters and covariates. Besides, these methods can only produce design flood estimations, lacking the ability of conducting simultaneous uncertainty analysis. To address the above shortcomings, the ARIMA-MCMC model was established by introducing the time-varying statistical parameters of the fitting period into the MCMC sampling process, then the parameter sampling for the non-stationary design flood frequency distribution model was conducted under future climate change conditions. Based on the posterior distribution of the parameters, the design flood frequency was calculated, and the corresponding confidence intervals were derived. Taking the Xiaodeshi Station in the Yalong River Basin as an example, the variation of design flood in the station under future climate change conditions was quantatively analyzed by ARIMA-MCMC model. The results show that the convergence effect of the parameter sampling based on ARIMA-MCMC method is excellent, and the values of D for all three scenarios are smaller than the critical value of 5% significance level. Except for the design values with P=0.1% and P=0.05% under scenario SSP2-4.5, the design daily maximum flow in other scenarios increased significantly compared to historical periods, with the increases of 7.1%-10.5% and 13.9%-27.2% under the scenarios of SSP1-2.6 and SSP5-8.5, respectively. The ARIMA-MCMC method established in this study can effectively conduct the non-stationary design flood frequency analysis under changing conditions.