基于序贯高斯条件模拟的土壤重金属含量预测与不确定性评价——以宜兴市土壤Hg为例  

Prediction of Soil Heavy Metals Content Based on Sequential Gaussian Simulation and Evaluation of Its Uncertainties:A Case Study of Soil Hg Content in Yixing

作　　者：徐辰星 濮励杰[1,2] 朱明[1,2] 徐彩瑶[1] 张濛[1] 许艳[2,3] XU Chenxing;PU Lijie;ZHU Ming;XU Caiyao;ZHANG Meng;XU Yan(School of Geographic and Oceanographic Sciences,Nanjing University,Nanjing 210023,China;The Key Laboratory of the Coastal Zone Exploitation and Protection,Ministry of Land and Resources,Nanjing 210023,China;School of Environmental Sciences and Engineering,Suzhou University of Science and Technology,Suzhou,Jiangsu 215009,China)

机构地区：[1]南京大学地理与海洋科学学院,南京210023 [2]国土资源部海岸带开发与保护重点实验室,南京210023 [3]苏州科技大学环境科学与工程学院,江苏苏州215009

出　　处：《土壤学报》2018年第4期999-1006,共8页Acta Pedologica Sinica

基　　金：国家自然科学基金项目(41230751;41201180)资助~~

摘　　要：土壤重金属含量的定量预测和预测结果的不确定性评价对生态风险评估具有现实意义。将1 155个表层土壤Hg样本分成两个无交集的数据集,309个样本用于模拟,另外864个样本用于验证,运用序贯高斯条件模拟方法对宜兴市土壤Hg的空间分布进行估值和模拟,并通过与传统地统计方法中的简单克里格估值进行对照,同时利用高斯序贯指示模拟探索单点和多点联合的不确定性。结果表明:基于同一个半方差模型,100次序贯高斯条件模拟方法得到的E-type型与简单克里格估值在空间分布上基本一致,两种方法的精度基本相同;随机抽取了第1、25、50、75、100次序贯高斯条件模拟的单次实现,将模拟结果和插值结果与846个点进行验证,5个单次实现的平均预测误差和均方根预测误差略高;以0.15 mg kg^(-1)作为临界阈值的单点超阈值概率虽然较高,但其置信度不足以划定污染区域,在划定污染区域时应采用多点联合概率评估其可靠性。Quantitative prediction of soil heavy metal content and evaluation of its uncertainties is of great practical significance to ecological risk assessment.In this study,a total of 1155 soil samples for Hg content analysis were sorted into two disjoint datasets,a simulation dataset of 309 samples and a validation dataset of 846 samples.In order to validate accuracy and reliability of the results of the usage of the Sequential Gaussian Conditional Simulation（SGCS）method,comparison was performed of this method with the Simple Kriging Interpolation method based on the same Semi-variance model parameters.Furthermore,Sequential Gaussian Indicator Simulation（SGIS）was used to delineate areas with soil Hg content beyond the threshold value,and to explore for uncertainties with single points and multi-point combination.Results show that the E-type of 100 rounds of SGCS and the SK prediction are quite similar,the difference between the two is small in Average Prediction Error and Root Mean Square Error.This paper sampled the 1^（st）,25^（th）,50^（th）,75^（th）,and 100^（th） round of single implementation of the SGCS method for comparison with the 846 points in the verification dataset in simulated value and interpolated value.The Mean Prediction Error and the Mean Square Prediction Error of the five rounds of single implementation are coincidentally and relatively higher,（within the range of 0~0.01）.When threshold was set to be 0.15 mg kg^（-1）,the single-point critical probability is relatively higher.However,its confidence level is by a certain degree not enough to delineate polluted areas.Multi-point joint probability should be used to evaluate the reliability of the contaminated area.

关 键 词：土壤重金属 序贯高斯条件模拟 不确定性评价 土壤制图 地统计 

分 类 号：P285.1[天文地球—地图制图学与地理信息工程] S159.9[天文地球—测绘科学与技术]