基于长程相关性特征的北京气候中心气候系统模式对中国气温的模拟性能评估

传统对数值模式模拟效果的评估主要是基于均值、趋势、概率密度分布、极值等方面的特征比较模拟要素与观测值的差异或是通过分析模拟值与观测值之间的相关性来定量判断.这类评估方法主要考虑了模式模拟结果与观测资料在统计上的差异,缺乏对两者在动力学特征上的比较.鉴于此,本文基于气象观测资料演变过程中所展现的长程相关性特征,利用去趋势波动分析方法对观测资料和模式模拟数据进行标度分析,研究模式模拟结果是否具有类似于观测资料中的长程相关性特征,进而通过比较观测资料和模式模拟结果的标度指数,从大气演变的内在动力学特征上对模式模拟性能进行评估.本文对北京气候中心气候系统模式对中国地表气温的模拟性能进行了评估,结果表明,该模式能够较好地反映出中国区域气温要素的长程相关性特征,对东北、西北中东部、江淮、江南东部等地模拟效果较好,但对于青藏高原地区及西北大部、华北、黄淮等地的模拟效果相对较差,其中对青藏高原地区和西北西部的模拟效果最差.

Performance evaluation of Chinese air temperature simulated by Beijing Climate Center Climate System Model on the basis of the long-range correlation

The traditional evaluation of simulation perfomance for numerical model is based on the differences in average, trend, probability density distribution, extreme value, variance and self-correlation between observational data and the simulated data by numerical model. These evaluation methods mainly depend on the comparison of statistical difference between the simulation and observation, but ignore the comparison in the dynamical characteristics of climate system. In view of this, in the present paper we use detrended fluctuation analysis (DFA) method to analyze the scaling characteristics of daily temperature from observational data and the corresponding simulated data by climate model, and investigate whether the long-range correlation of meteorological elements can exist in the simulated data of climate model. If so, we can assess the performance of climate model based on the comparison of scaling index between the observational data and the simulated data. As an example, the performances of Beijing Climate Center Climate System Model (BCC_CSM1.1(m)) model, including simulated daily average temperature, daily maximum temperature and daily minimum temperature, are assessed by using DFA. The results indicate that the long-range correlation characteristics of temperature in China can be simulated by BCC_CSM1.1(m). But the simulation errors are relatively large in Qinghai-Tibet Plateau, most West China, North China, and Huanghuai district. Thereinto, and the performance of BCC_CSM1.1(m) in Qinghai-Tibet Plateau and western West China is worst in China.
Keywords：detrended fluctuation analysisscaling exponentlong-range correlationmodel performance evaluation