水环境模拟参数随机场特异值及稳健变异函数的研究  

Research on the outlier of parameters random fields in water environmental simulation and the robust variation function

作　　者：牟向玉[1] 张征[1] 尚晓颖[1] 徐永利[1] 

机构地区：[1]北京林业大学环境科学与工程学院,北京100083

出　　处：《安全与环境学报》2008年第4期88-92,共5页Journal of Safety and Environment

摘　　要：根据变异函数不稳健性的产生机理,对原始监测数据进行正态变换并对特异值的识别及处理作初步研究。以廊坊市地下水采样中Cl-质量浓度为区域化变量,采用影响系数法(ICM)和估计邻域法(ENM)识别并处理其特异值;采用Cressie-Hawkins和中位调节法两种稳健统计学方法对采样数据进行系统处理和分析。结果表明,对空间变异性较大的变量(Cl-质量浓度),其特异值经ICM和ENM处理后的估计效果均优于未处理特异值的效果,其中ICM对特异值的处理效果略好于ENM。采用稳健变异函数计算方法时能得到结构特征较强的变异函数,且处理特异值后再进行稳健计算比单纯使用稳健公式的效果好。可见,获得稳健变异函数及其估计量对改善变异函数的结构性和理论模型的主要参数有明显效果。This paper is concerned about the authors＇ research on the robustness of the spatial variogram and the treatment of the outlier in water environmental simulation. For our research purpose, we have chosen sampling data from the underground water in Langfang City, Hebei. Based on the analysis of the mechanism of the unrobustness of the variation function, we have made the first attempt to make a series of experiments by means of normal variation and the outlier recognition and rejection. Testing the outliers of sampling data by using Influencing Coefficient Method and Estimating Neighborhood Method, we have gained more satisfactory results, which indicate that the estimation effect after treating the samples collected with the two methods proves somewhat advantageous over those unprocessed. More exactly speaking, the processed data of Cl^- mass concentrations are endowed with higher spatial ability, which may imply the Influencing Coefficient Method is advantageous over the Estimating Neighborhood Method. Based on what has been tested, the sampling data were systematically treated and analyzed with the two robust statistical methods, that is, the Cressie-Hawkins Method and the Median Regulation Method. The results of the carefully planned analyses show that the Median Regulation Method proves better in its efficiency than Cressie- Hawkins Method. Since Cl^- mass concentration is characteristic of anisotropy, the best choice of the treatment of the outlier and robust statistical methods should be with the Median Regulation Method. The results also show that we shouldn＇t over-simplify the general purpose of the evaluation methods but analyze problems in accordance with the different situations. It can also be concluded that the variation function of higher spatial constitutive property found through the computational methods of robust proves to be more remarkable than not through the treatment of the outlier. The results thus indicate that the robust variogram and its estimation can be done more effectively on th

关 键 词：环境地学 水环境模拟参数随机场 地质统计学 稳健统计学 稳健变异函数 特异值 

分 类 号：X523[环境科学与工程—环境工程] P641.2[天文地球—地质矿产勘探]