Projections Of Future Climate Change And Uncertainty Over China Based On Bayesian Model Averaging

Based on 35 climate models in the Coupled Model Intercomparison Project phase five (CMIP5) historical simulation experiment and future projection under RCP4.5 scenario, Climatic Research Unit (CRU) and Chen data are used as observation, then we apply the Bayesian Model Averaging (BMA) to project the change and uncertainty of climate over China in the 21st century. The main results are as follows:(1) BMA can simulation the temperation well, and the best model whose average weight is the highest is CESM1-CAM5, NorESM1-M and MIROC5 are also better. While the performance in simulating precipitation is ordinary but better then the ensemble average mean, and the best three models are CNRM-CM5, FGOALS-g2 and HadGEM2-AO.(2) There is a consistently increase in temperature change, and the change during winter are bigger than summer, decreasing from north to south. The maximum value centers are northeast and northwest of China, while southeast coast and southern of the Tibet are the minimum centers. It is very likely (>80%) that surface air temperature increases by 0.7℃ in Northern China, and the probability of Southern China exceeds 50% at the beginning of 21st century. With the same probability of Northern (>80%) and Southern China (>50%), the warming range are 1.5 and 2.0℃, when it comes to the middle and end of 21st century; It’s obviously that the precipitation will incrase in most parts of China, and the change range during winter is higher than summer. At the end of 21st century, the range of winter and summer rainfall is found to be 11.2% and 6.6%, while the Tibet and inner Mongolia are the maximum value centers.(3) Based on the SNR, the reliability of estimated temperature results by BMA is very high, and the SNR is generally higher than 3. Compared with summer, the reliability of winter is lower in most regions except the eastern of Tibet Plateau, and the reliability at the middle of 21st century is found to be higher than the beginning and the end but with the same spatial distribution. Besides, Tarim Basin is the maximum center with the value of 7, while the southeastern of Tibet Plateau and northern of Daxingan Mountain are the area where SNR are the lowest but higher than 1; However, the reliability of precipitation estimated results is so poor that the SNR is almost under 1 across the country, and south China has a minimum value center, where the SNR even under 0.2. Besides, the reliability of estimated results in winter is slightly higher than summer.