Projection And Evaluation Of Statistical Downscaling Of Regional Daily Precipitation Over Yangtze-Huai River Basin Based On Self-Organizing Maps

Based on the ERA-40 daily reanalysis data from 1961-2002,and the observed daily precipitation data at 56 meteorological stations which locate in the Yangtze-Huai River basin,this study uses a new downscaling method based on Self-Organizing Maps(SOMs) to produce downscaled summer precipitation estimates at each station.The simulation capabilities of the statistical downscaling approach for monsoon precipitation and extreme precipitation over east China have been assessed by independent samples test.Subsequently the downscaling model is applied to simulate daily precipitation at the same 56 stations for the period 1986-2005 by using predictor sets simulated by BCC-CSM1.1(m),MPI-ESM-MR and IPSL-CM5A-MR.Furthermore,the downscaling model is applied to construct scenarios of the daily precipitation during the 21th century (2016-2035,2046-2065, 2081-2100) by using predictor sets simulated by three GCMs (i.e. BCC-CSM1.1(m),MPI-ESM-MR and IPSL-CM5A-MR) forced by Representative Concentration Pathways 4.5.The main results are as follows:(1)The SOM derives a transfer function between large scale weather patterns and local daily precipitation.The downscaling approach provides a faithful reproduction of observed probability distributions and temporal characteristics of precipitation.The Brier Score for all stations are almost 0,and the Significance Score are above 0.8. The average bias of downscalede number of days with precipitation greater than lmm and 10 mm,summer total precipitation,simple daily intensity index,extreme daily precipitation threshold and fraction of total precipitation due to events exceeding the 95th percentile of the climatological distribution for wet day amounts are below 11%.Furthermore,the downscaling approach is able to reproduce,to a certain extent,the temporal variability of precipitation characteristics.(2)The downscaling approach improved GCMs’ability of daily precipitation probability distribution simulation.Compared with the raw GCMs,the biases of precipitation indices for downscaled results are reduced by 40% to 60%.The biases of all precipitation indices are almost below 20%, the spatial correlation coefficients are increased to 0.9,and the root mean square errors are below 0.5,which improved consistency of multi-mode precipitation simulation. Furthermore,the downscaling approach improved the ability of temporal variability of extreme precipitation characteristics simulation.(3)The current downscaling model is broadly valid and stationary for the projection of future changes.The regional average of precipitation indices will gradually increase in the magnitude of change over time;in 2016-2035.The percentage change of precipitation indices of all stations are below 10% with the precipitation reduction of west stations and increase of east stations.Precipitation indices will increase by 20% in 2046-2065 and 30% in 2081-2100.The projection change of downscaling results can be explained by changes in the frequency of weather patterns.The frequency of wet weather patterns at Nanjing station increase significantly and the frequency of dry weather patterns decrease during 2046-2065 and 2081-2100.Therefore,precipitation increase significantly at those two periods.