24个CMIP5模式对长江流域模拟能力评估

根据1961~2005年长江流域气象站点的实测月降水量和气温数据,采用第5期全球耦合模式比较计划CMIP5(the Fifth Phase of Coupled Model Intercomparison Project)中24个全球气候模式(GCM)的模拟结果,通过计算模拟变量和观测变量平均值的相对误差、归一化的均方根误差、时间和空间相关系数,采用M-K趋势分析方法,分别选用在长江流域模拟气温和降水较好的5个模式进行集合平均,从时间的演变规律和空间的分布特征两方面,检验该模式集合对长江流域模拟气温和降水的能力。研究结果表明:各个模式模拟气温的能力要明显好于模拟降水的能力,但模拟气温较好的模式模拟降水的能力并不一定突出;模式集合的结果表明:在时间尺度上,模式集合平均结果与观测值拟合程度较好,且模式集合的结果振荡幅度较观测值小;在空间尺度上,模式集合的空间分布趋势与观测值大致相同,说明采用的模式集合结果用于预估未来长江流域降水的时空分布特征和演变规律是可行的。

ASSESSMENT ON 24 GLOBAL CLIMATE MODELS IN THE CMIP5 OVER THE YANGTZE RIVER

Global climate models (GCMs) have been widely used for projections of future climate change and have provided unprecedented opportunities to analyze the potential effects at regional and continental scales. However, the simulation capacity of each GCM varies among different regions. Therefore, assessing the performances of GCMs in specific regions is essential for further study of adapting and mitigating the effects of climate change at regional scales. Yangtze River Basin, as one of the most important political and economic areas in China, has increasingly suffered from flood and drought disasters in recent years because of climate change. In this study, the performances of 24 Global Climate Models (GCMs) in the fifth phase of Coupled Model Intercomparison Project (CMIP5) were assessed by comparing the model outputs with the observations from 127 meteorological stations for the period of 1961-2005 over the Yangtze River. Top 5 GCMs for annual air temperature and precipitation were selected by the calculation of the Mean Absolute/Relative Error (AE/RE), Normalized Root of Mean Square Error (NRMSE), Time and Spatial correlation coefficient as well as two statistical indices of M K trend analysis methods. However, no GCM fell in the best five for both air temperature and precipitation, and likewise no GCM fell in the worst five for them over the Yangtze River. According to recent studies, the ensemble mean of all the models usually shows better agreement with the observations than any single model. Thus, the ensemble outputs of top 5 models for air temperature and precipitations were evaluated in terms of time and spatial distribution characteristics, respectively. The results suggest that the simulated annual temperature of GCMs is superior to the precipitation outputs, and this may result from the uncertainty and variability of precipitation. The averaged ensemble results show that the set outputs well fit the observations during 1961-2005, whereas oscillation amplitude of the simulations is smaller than that of observations. Furthermore, the ensemble results could capture the spatial distribution characteristics for both air temperature and precipitation over large areas except for some areas which show great divergence with less observations and high altitude. The ensemble outputs can consequently be used to project future temperature and precipitation over the Yangtze River. In order to assist policymakers and water managers in adopting strategies based on scientific understanding, further study should focus on the following aspects: (1) whether the density of meteorological stations and the interpolation methods have significant influences on the assessments; (2) whether the simulation capacities of GCMs can be improved by using finer resolution data and reducing the time scale data; and (3) how to improve the performances of GCMs in CMIP5 over the areas with large deviation of terrain