Frequency Analysis Of Precipitation Extremes Under A Changing Climate

In this paper,the Heihe River Basin is selected as the research area.Based on the daily grid data from 1960 to 2014 in the Heihe River Basin and the future daily precipitation data of 5 GCM models in the CMIP5 data set,the seasonal maximum daily precipitation sequence(SMP)of the Heihe River Basin is extracted.The GEV-CDN model is used to construct the SMP sequence and select a suitable stationary or nonstationary model for the SMPs time series.Finally,according to the optimized GEV model,the frequency characteristics of extreme precipitation of the four seasons in the upper,middle,and lower reaches of the Heihe River Basin are analyzed.The main conclusions are as follows:(1)The test method ADF and MK are used to test the stationarity and trend of 12 SMP sequences in the Heihe River basin from 1960 to 2014.The result shows that under the significance level of 0.1,all SMP sequences in the upper reaches of the Heihe River showed a significant upward trend,and there is no significant trend in other SMPs except winter in the middle and lower reaches.Among all seasons,the upward trend of winter SMP sequences is most significant.The correlation test results show that the midstream-summer SMP sequence is negatively correlated with the East Asian summer monsoon index(EASMI),and 5 of the 12 SMP time series are positively correlated with the Western Pacific subtropical high index(WPI).(2)The optimal results of the GEV-CDN model are consistent with the trend test results.The non-stationary model performs better than the stationary model in the SMP time series with significant trends.For stationary SMP time series,the stationary GEV model is still applicable.For SMP series that are significantly related to the climate index but don't have a significant trend,the non-stationary model with the climate index as a covariate can effectively improve the quality of the fit.(3)For SMP sequences with significant trends,the return levels increase with time,that is,the frequency of extreme precipitation events increases.Uncertainty analysis shows that the higher the return period,the greater the uncertainty of the return level estimation results.Besides,the more complex the model,the lower the reliability of the results when predicting the future extreme precipitation.(4)After the bias correction,GCM's simulation of extreme precipitation is better than before.In the prediction of future extreme precipitation,the deviation between the return levels of the five GCMs models increases as the return period increases,and the deviation between the five GCM models of the RCP8.5 is higher than the RCP2.6 and RCP4.5.The negative change of extreme precipitation relative to the historical period in the future period mainly occurs in the summer and autumn seasons.In winter and summer,there is a positive change,with the largest change in winter.Among three emission scenarios,the change in RCP8.5 scenario is the most obvious in the positive change,and the change in the RCP2.6 scenario is the largest in the negative growth.In addition,the future forecast results based on the trend extension in the historical period are relatively close to the future forecast results under the RCP4.5 emission scenario.