The Simulation of Extreme Precipitation over Hunan Province based on the Statistical Downscaling Method of Transform Cumulative Distribution Function (CDF-t)
Intensify,frequency and duration of Extreme Precipitation would increase in the future under a warming climate. Especially for Hunan where is sensitive to the climate change. Based on the CMIP5 historical simulations datasets,the ability of the CMIP5 models in simulating the spatial pattern andinterannual variability of extreme precipitation over Hunan province are evaluated using the statistical downscaling method of transform cumulative distribution function (CDF-t) combined with four extreme precipitation indices. The results show that due to the low resolution of GCM models,characteristics of extreme precipitation related to the terrain and atmospheric circulation over Hunan were not exactly reproduced. There are great differences between patterns,and the model sets have relatively poor simulation results. CDF-t statistical downscaling can improve CMIP5 simulation of extreme precipitation in Hunan by establishing the functional relationship between large-scale variable CDF and the same regional variable CDF. As far as the spatial structure is concerned,this method can greatly improve the spatial structure ability of the model to simulate the heavy rain days (R10) and the continuous five-day maximum precipitation (R5d) ,and shows a high consistency between the models,especially the effect of R10 improvement is the most remarkable. Compared with the observation,the spatial average absolute error in Hunan area reaches 2. 18 d,which is lower. The absolute error before scale is reduced by 45. 46%. As far as time variability is concerned,this method can greatly improve the time variability ability of model simulation R90P and R5d. After scaling down,the IVS value decreases from 2. 2 and 1. 5 to 0. 3 and 0. 6 respectively.