Abstract: On the basis of the integrated ecological risk assessment of China, this research chooses the risk sources data as the main source of uncertainties to analyze their impact to the risk value calculation and the sensitivities of the risk value to them. Monte Carlo smulation method is used to perform these proprieties for each pixel throughout the country with the special scale of 10 km×10 km. In the simulation progress for each pixel, the uncertainty is set firstly, and then the probability distribution of the risk source levels are computed according to their frequency statistics for at least last 50 years. At last, the simulation of risk value is carried out reference to the equitation of integrated ecological risk assessment. Meanwhile, the sensitivities of the risk value to all risk sources are got in term of Spearman rank correlation coefficients, the variance contribution of which can be used to figure out the sensitivity structure of all risk sources. The uncertainties are set as 50% and 25% to indicate high uncertainty and low uncertainty respectively, and the simulation times is set as 10000 for each pixel. The results show that 1) the simulated mean risk values are close to the calculated risk values, the ratio of them for all pixels are concentrated in 0.80?1.20 with either high or low uncertainty, and 0.95?1.05 for low uncertainty, although some differences exist between different hazard regions; 2) the simulated maximum risk values are also close to the calculated risk values, which are mostly less than one and a half of the calculated, except for the Qinghai-Tibet region; 3) for the whole country, three main risk sources that the risk value most sensitive to are droughts, floods and landslides/debris flows.

Key words: integrated ecological risk assessment, uncertainty analysis, sensitivities analysis, Monte Carlo simulation

摘要： 以全国综合生态风险评价为基础, 选择风险源数据为主要分析对象, 分析冰雹、大风等10种风险源的不确定性对风险值计算的影响, 以及风险值对风险源分级的敏感性。采用蒙特卡洛模拟法, 对全国范围内10 km×10 km的栅格进行逐一分析。在每个栅格的模拟过程中, 首先设定数据分析的不确定性, 然后按照风险源等级划分的频数统计设定风险源分级模拟值的概率分布, 利用全国综合生态风险评价的基本公式, 对栅格的风险值进行模拟; 同时, 利用Spearman秩相关系数求解风险值对风险源分级的敏感性, 通过方差贡献求得模拟风险值对所有风险源的敏感性结构。设定50%和25%两个水平分别表征高不确定性和低不确定性, 模拟次数设置为10⁴次, 结果显示: 1) 模拟风险的平均值与风险计算值接近, 两者之比多数介于0.95~1.05之间, 但不同风险分区的性质会存在差异; 2) 风险最大模拟值普遍高于风险计算值, 但除青藏区外多数风险最大模拟值与风险计算值的比例仍集中在1.0~1.5之间; 3) 全国范围内, 风险最敏感的3个主要风险源为干旱、洪涝以及滑坡和泥石流。

关键词: 综合生态风险评价, 不确定性分析, 敏感性分析, 蒙特卡洛模拟