CMIP5模式对中国极端气温及其变化趋势的模拟评估

Evaluation of the Extreme Temperature and Its Trend in China Simulated by CMIP5 Models

本文基于中国区域逐日气温资料和CMIP5中30个全球气候模式资料,计算了平均日最高气温(TXAV)、平均日最低气温(TNAV)、热浪指数(HWDI)、霜冻日数(FD)、和暖夜指数(TNF90)5个极端气温指数,评估各模式对中国区域极端气温的气候平均场和趋势的模拟能力。研究结果表明,大部分模式能够较好地模拟出极端指数的气候平均场,其中对TNAV、TXAV和FD平均场模拟能力较强,大部分模式平均场与观测场的相关系数超过0.90,但对TNF90和HWDI的模拟能力相对较低,相关系数均低于0.70,且各模式的模拟能力存在较大的差异。对极端指数的趋势模拟来说,模式模拟的中国区域平均各极端气温指数的线性变化趋势与观测相同,但大多数模式模拟趋势的强度偏弱。相比于气候平均场,模式对极端气温指数趋势空间场模拟较差,除TNAV有1/3的模式平均场与观测场的相关系数超过0.60外,模式模拟其余指数的相关系数均低于0.60。模拟极端气温气候平均场的能力最优的5个模式为:IPSL-CM5A-MR、CMCC-CM、IPSL-CM5A-LR、MPI-ESM-MR和MPI-ESM-P。趋势空间场模拟最好的5个模式为:MPI-ESM-P、CANESM2、ACCESS1-3、BCC-CSM1-1和Nor ESM1-M。模式对极端气温指数的时空模拟能力一致性较差,但基于气候平均场或趋势空间场的优选模式,相比于所有集合模式平均,模拟能力均有一定程度的改善。

Based on the observation of daily temperature data and 30 models data provided by Coupled Model Intercomparison Project (CMIP5), five extreme temperature indices including mean maximum temperature (TXAV), mean minimum temperature (TNAV), heat wave duration index (HWDI), frost days (FD) and warm nights (TNF90) were calculated to evaluate simulation capability of each model in terms of spatial field in climate state and the trend of extreme temperature in China. Most models can reflect the spatial pattern of extreme temperature indices in climate state. The spatial pattern of TNAV, TXAV and FD can be better simulated, as the correlation coefficient of most models are more than 0.90, and there are good consistency between models. TNF90 and HWDI are poorly captured by models, as the correlation coefficients are lower than 0.70 and there are large differences between models. Models also simulate the same trend change of average extreme temperature index at national scale as the observation, but most simulation results are weaker than the observation. Compared with the spatial pattern in climate state, the models simulation of trend spatial pattern for extreme temperature are not good. Thereinto, TNAV is relatively well simulated, the correlation coefficients of a third of models are larger than 0.60. The correlation coefficients for other indices are lower than 0.60. Based on space simulation ability, the top 5 models are IPSL-CM5A-MR, CMCC-CM, IPSL-CM5A-LR, MPI-ESM-MR and MPI-ESM-P. Based on time simulation ability, the top 5 models are MPI-ESM-P, CANESM2, ACCESS1-3, BCC-CSM1-1 and NorESM1-M.The models simulation ability ranks of space and time are in poor consistency. Nonetheless, compared with the average multi-model ensemble (AMME), simulation ability of the best multi-models ensemble (BMME) based on the space or time simulation are improved.