Statistical Downscaling Of Temperature And Precipitation Based On The Multimodel Ensemble Forecast Using CMIP5 Data

Based on the CMIP5 runs of 17 models for climate system hindcast of air temperature and precipitation:BCC-CSML1.1 CanCM4, CCSM4,CFSV2-2011,CMCC-CM, EC-EARTH et al., and ERA reanalysis data and JRA-55 data are used as observations of temperature and precipitation. We use multi-model ensemble mean and Superensemble Prediction methods to integrate the hindcast of seventeen models, then we conduct the Statistical downscaling research, and evaluate the results of the ensemble and downscaling. On the basis, we use model projection data to estimate the temperature and precipitation for the period from 2006 to 2025.Studies have shown that the multi-model integrate results is better than single model, which rewards the best SUP effect, and its time series is divided into the training period and the forecasting period. Since the length of the training period and the model numbers have some influence on the results of SUP, the RMSE, ACC and MAE are used to evaluate the hindcast results of multi-model ensemble mean and SUP results.To improve the efficiency and accuracy of calculation, we choose the length of training period for temperature and precipitation is 34 and 35 year, the optimal model numbers both are fifteen. The regional average RMSE of temperature is about 0.43℃ smaller than EMN, which precipitation is about 0.45mm/d smaller than EMN.Then, the statistical downscaling based on linear regression was subsequently used to improve the interpolated the surface air temperature and precipitation hindcast results.The results show that the multi-model ensemble hindcasts are superior to those of individual models in terms of the root-mean-square errors (RMSE) and the anomaly correlation coefficients (ACC) of the surface air temperature and precipitation after statistical downscaling. The improvement of individual models is better than multi-mode integrated. In addition, the statistical downscaling can also get better results of time changes and spatial distribution over the East Asian for temperature and precipitation. For the hindcast of the temperature, the uncertainty over the oceans is less than over the continents,and precipitation is vice. Overall Super-ensemble Prediction and statistical downscaling temperature and precipitation for return better results.In the RCP4.5 scenario, the projection of 2005 is used to estimate the temperature and precipitation for the period from 2006 to 2025, and the results show that the surface air temperature over East Asia would increase, and the increment of the surface air temperature over the oceans is smaller than that over the continents;while precipitation in most areas will also increase, but the rate of increase is smaller.