Statistical Modelling And Probabilistic Projection Of Extreme Precipitation Over China

Based on the observed daily precipitation,the characteristics and statistical modelling of summer extreme rainfall in China are investigated from their intensity,frequency and duration(IFD)for the period 1961-2012.For the BCC-CSM1.1-m,the multiple bias correction methods are introduced to evaluate their performance in correcting summer daily precipitation,especially extreme rainfall in Jianghuai River Basin(JRB).Furthermore,the probabilistic bias correction approach is presented and assessed in the applicability of the extreme precipitation correcting based on the extreme distribution.Finally,combined with the bias corrected precipitation of best selected global climate models(GCMs)from CMIP5,the above-mentioned statistical model is applied to obtain the future probability changes of extreme rainfall under the RCP4.5 and RCP8.5 scenarios in the late 21st century.The main conclusions of this study are as follows:(1)Based on the newly developed extreme value theory,the extreme wet spell model(WSM)is proposed to jointly characterize the observed IFD distributions of extreme precipitation by the generalized Pareto distribution(GPD),Poisson and geometric distributions.The results indicate that the WSM is able to realistically model extreme rainfall.The WSM probability distribution and non-parametric improved Mann-Kendall trend analysis results show that,extreme precipitation is most likely to occur in the Southern China,especially the Southwest China.The extreme precipitation which lasts one day mainly concentrate in the middle-upper Yangtze River Basin(YRB),the North and Northeast China,while two days and above in middle-lower YRB and its south,especially along the coastal areas.Also,there is a significant increase in both frequency and intensity of extreme rainfall in the middle-lower YRB and its south,but the situation in northeast and southwest China shows opposite changes.Differing from the general decline in number of rainy day,the number of extreme rainfall days and averaged duration over the threshold of 80%shows a same distribution with the extreme precipitation amount.(2)Although BCC-CSM1.1-m has much better performance in climate simulation over China,but still has evident deviation in the daily rainfall especially extreme precipitation in JRB.The simulating ability can be significantly improved,but with different performance in different sides,after the multiple bias correcting,by the Linear scaling(LS),Quantile mapping(QM),Distribution mapping(DM)and Cumulative distribution function transform(CDFT).The QM has the most significant correcting improvement in the probability distributions of daily precipitation and area mean rainfall,and CDFT comes second.For the spatial consistency of extreme precipitation,LS,QM and CDFT have better simulating in the total precipitation and moderate rainy days,and only QM can correct the rainfall days,precipitation intensity and 95%quantile precipitation.(3)The probability bias correction model(XCDFT)is proposed for the extreme rainfall by introducing GPD into the correcting process,and used to assess their ability of BCC-CSM1.1-m simulation correcting for Wuhan,Hefei,Nanjing and Hangzhou in JRB.This demonstrates that XCDFT can preserve the CDF form of the observed calibrated precipitation,and acquire the small changes from the calibration to the verification.The LS,QM,DM and CDFT:main reflect the correction of the mean features,and have poor performance in extreme precipitation.Reversely,the BCC-CSM-1.1-m simulations are significantly improved after XCDFT correction,and shows good agreement between the probability distribution of the verifications and the XCDFT results.(4)Based on the WSM,ACCESS1.3,BCC-CSM1.1-m,CCSM4,HadGEM2-ES and MPI-ESM-MR are evaluated and selected from CMIP5,which can be well performed in the summer extreme precipitation characteristics.Furthermore,the improved quantile mapping is raised to correct the above five selected GCMs by learning from the QM and XCDFT.Combined with the corrected results and the WSM modelling,we can obtain the future probability changes of extreme rainfall.The results show that in the late 21st century,the intensity and frequency of extreme rainfall show substantial increases over most of China,with the largest increase in intensity over the middle-lower YRB and Yellow River Basin,while the highest increase in frequency over southwestern and northeast China.The RCP8.5 scenario was found to produce much larger increases than RCP4.5.Compared to the historical period(1961-2005),the probabilistic projections indicate that 100-year return level will be improved by over 10.9%(20.1%),while the probability of at least five extreme precipitation events per year by 8.2%(12.6%)over most stations of China in RCP4.5(RCP8.5).In contrast,changes in wet spell duration were found to be small.