CMIP5全球气候模式对中国上空空气静稳日数模拟能力评估

空气静稳日数变化与污染物浓度变化密切相关,评估气候模式对空气静稳日数的模拟能力是进行未来预估的基础.本文利用15个CMIP5(Coupled Model Intercomparison Project phase 5)全球模式的模拟结果与观测数据,分别计算了1961-2005年逐年中国上空空气静稳日数,并利用统计方法分析了中国上空空气静稳日数的标准差、相对均方根误差、区域平均的时间序列、趋势分布和EOF(Empirical Orthogonal Function)主要模态变化特征,评估了CMIP5模式对中国上空空气静稳日数的模拟能力.结果表明:多模式集合平均结果可以模拟出空气静稳日数由沿海向内陆逐渐增加的分布特征,单个模式对空气静稳日数空间分布的模拟能力相差较大.多模式集合平均可以较好地再现夏、冬季的空气静稳日数.15个模式中,CanESM2和:IPSL-CM5B-LR对中国大部分区域的模拟效果较好,多模式集合平均的模拟能力优于单个模式.与观测相比,多模式集合平均的1961-2005年空气静稳日数年际变化波动较小,多数区域的多模式集合平均的空气静稳日数高于观测值.对于逐年的冬季空气静稳日数,大多数区域的多模式集合平均存在高估.在中国东部和新疆大部,多模式集合平均可以较好的模拟出空气静稳日数变化趋势的空间分布特征,但是数值偏小.多模式集合平均也能较好的模拟出空气静稳日数的EOF1和EOF2特征向量分布型,但对前三个EOF的时间系数序列模拟能力差.

Evaluation of air stagnation in China by CMIP5 models

The change of air stagnation days is closely related with the variation of air pollutant concentration, while evaluating the performance of simulating air stagnation days will provide a basis for projection. In this paper, we calculated the number of air stagnation days during 1961-2005 by employing the outputs from 15 Coupled Model Intercomparison Project phase 5 (CMIP5) models and observation data. Then the models' ability to simulate air stagnation days was assessed. The statistical approach we used is listed as follows: standard deviation, root-mean-square error, time evolution of each sectors, trend pattern and EOF (Empirical Orthogonal Function). We found that the performance of multi-model ensemble (MME) is high, but that of individual models varies widely. The CanESM2, IPSL-CM5B-LR and MME work well over the most parts of China. And MME can reproduce the spatial pattern of observed air stagnation days in summer and winter well. While the observed time evolutions of spatial averaging fluctuate greater than the simulated ones over China and its eight sub-regions, and MME values of most sectors in China are higher than the corresponding observed values. In winter, MME values are higher than the observed ones over most sectors. There is a good agreement in air stagnation trends pattern between the observational and model data over the Eastern part of China and most part of Xinjiang, but the spread of simulated trends is smaller than the one of observed trends. MME has the ability to capture the spatial distribution of EOF1 and EOF2; however, it fails to reproduce the observed time coefficient series of EOFs.