Simulation And Projection Of Temperature Extremes Over China Based On Multi-model Dynamical Downscaling And Bias Correction

A variable resolution atmospheric general circulation model LMDZ4 which is one-way nested into five global climate models (GCMs) (BCC-csml-1-m,CNRM-CM5, FGOALS-g2, IPSL-CM5A-MRand IPSL-CM5A-MR) is used to conduct an ensemble dynamical downscaling data sets for China during 1961-2005.The simulation capability of extreme temperature of each model has been evaluated comprehensively. The equidistant CDF matching method (EDCDFm) and the CDF-Transform Method (CDFt) are used for the bias correction of daily dynamical downscaling outputs. And EDCDFm method with good performance is selected to project the future changes of climate for the early (2016-2035), middle (2046-2065)and end (2080-2099) of the twenty-first century. The main results are as follows:(1) Compared with GCMs, LMDZ4 shows its superiority in depicting a good regional terrain including the Tibetan Plateau and Sichuan Basin, and can better show the spatial distribution of extreme temperatures in China. However, the improvement of dynamical downscaling shows significant regional differences. For mean minimum temperature, mean maximum temperature and frost days, the simulation of downscaling is mainly improved in Northeast, Northwest, the Tibetan Plateau and Southwest China, with the correlation coefficients being increased to 0.95, and the normalized root mean square differences being below 0.5 ( 0.5d) , and the improvement of the correlation coefficient for mean minimum temperature and mean maximum temperature is increased with the terrain height. The improvement of heat wave duration index in Northeast, North and Southwest China is significant, but there are large differences between models. Furthermore, compared with GCMs, the downscaling models are able to reproduce, to a certain extent, the spatial distribution of the trend of mean maximum temperature and mean minimum temperature in China, and decrease the trend errors of mean minimum temperature and frost days in Northeast, the North, the Tibetan Plateau and Southwest China. The downscaling models ensemble also performs well in reproducing the observational spatial field in climate state and trend of temperature extremes in China.(2) The EDCDFm method and CDFt method can better correct the spatial distribution of daily mean temperature and extreme temperature, and significantly reduce the biases of the model simulation. The average bias of the daily mean temperature in China is reduced from -1.15? to -0.15? and -0.14?. The spatial correlation coefficient of each index is increased to above 0.99 (except for the heat wave index is 0.75) , and the root mean square error is close to zero. However, the cold bias of CDF-t method in winter is obviously higher than that of EDCDFm method, and the temperature change curve of EDCDFm method is closer to the observation than CDF-t method. Therefore, it is more appropriate to use the parameterized EDCDFm method to project future changes of temperature.(3) In the RCP4.5 emission scenario, the temperature over China shows a warming trend. The mean temperature is projected to increase by 0.76?, 1.84? and 2.10?with respect to the 1986-2005 during 2017-2036, 2046-2065 and 2080-2099 respectively. The spatial of future change for the mean temperature, maximum temperature and minimum temperature in the three periods have a good consistency,warming in the northern China is higher than that in the south, and the temperature change in the autumn and winter is larger than that in the spring and summer.Uncertainties of temperature change projection are large in Tibetan Plateau and the Sichuan Basin. The frost days decrease significantly, especially in the Tibetan Plateau, the frost days in the three periods decrease more than 15 days, 30 days and 40 days, respectively.The variation of the heat wave index is the smallest, and the increase of the heat wave is mainly in the eastern China, and the heat wave in the south China is more than 2 days.