作 者：
李凯;赵华甫;吴克宁;江红蕾;刘霈珈;许伟伟
单 位：
北京师范大学减灾与应急管理研究院;中国地质大学土地科学技术学院;江苏省地质调查研究院
关键词：
插值方法;采样数量;重金属Cd;预测精度
摘 要：
对局部存在重金属污染地区采用适宜的插值方法和布设合理的采样点对重金属污染状况监测具有重要的意义。运用单因子指数法得到土壤重金属Cd污染指数,并在镇域内布设2033个采样点的基础上,通过随机抽样法抽取1830、1423、1017、610、203五个采样点样本子集。首先,运用普通克里金法(OK)、径向基函数法(RBF)和反距离权重法(IDW)对该地区土壤重金属Cd污染指数进行插值预测,并通过交叉验证法进行精度检验。然后,在反距离权重法的基础上,对五个采样子集进行插值精度分析,得到大致合理采样数量。结果表明:(1)利用全集2033个采样点对3种插值方法进行交叉验证分析可知,RMSE表现为IDW(3.018)>RBF(2.942)>OK(2.837),ME表现为OK(-0.0736)>IDW(0.0214)>RBF(0.0096),MAE表现为IDW(0.5668)>RBF(0.5575)>OK(0.5227),3种插值方法在整体预测精度上差异不明显。而对于污染区域的识别,IDW在轻度污染区、中度污染区和重度污染区预测上表现出较大的优势,能较好的反应污染区域的空间变异特征。因此,认为IDW为较适宜的空间插值方法。(2)对不同采样数量的样本进行交叉验证分析可知,RMSE、ME和MAE在1017个采样点到610个采样点误差变化幅度分别为29.84%、71.31%和36.99%,误差增加幅度较前三个子样本间明显增大。在空间特征识别方面,2033、1830、1423和1017个采样点反映的污染区的空间分布特征非常相似,610和203个采样点预测的污染区域面积明显扩大,对各级污染区域的空间特征细节表现能力较差。因此,对于该镇域内的土壤重金属Cd污染指数的研究,1017个左右采样点是比较合理的采样数量。
译 名：
Suitable Interpolation Method and Reasonable Sampling Quantity of Cd Pollution Index in Soil
作 者：
LI Kai;ZHAO Hua-fu;WU Ke-ning;JIANG Hong-lei;LIU Pei-jia;XU Wei-wei;School of Land Science and Technology, China University of Geosciences;Key Laboratory of Land Regulation Ministry of Land and Resources;Institute of Disaster Reduction and Emergency Management, Beijing Normal University;Geological Survey of Jiangsu Province;
关键词：
Interpolation method;;Sampling quantity;;Heavy metal Cd;;Prediction accuracy
摘 要：
Using appropriate interpolation method and setting reasonable sampling point is of great importance for monitoring condition of heavy metal pollution. In this paper, we used the single factor index method to obtain soil pollution index of Cd and extract random samples of 1830, 1423, 1017, 610 and 203 from a total of 2033 samples.First of all, we predicted the spatial distribution of Cd pollution index using ordinary Kriging(OK), radial basis function(RBF) and Inverse distance weighting(IDW) methods and checked the accuracy by the cross validation method. Secondly, we analyzed the interpolation accuracy for the five subsets using inverse distance weighting method to get a general reasonable number of samples. The results showed that(1) The complete 2033 sampling points were used to make cross validation analysis for three interpolation methods. The RMSE result decreased as follows: IDW(3.018) > RBF(2.942) > OK(2.837), the ME result OK(-0.0736) > IDW(0.0214) > RBF(0.0096), and the MAE result IDW(0.5668) > RBF(0.5575) > OK(0.5227). The prediction accuracy showed no obvious difference among three interpolation methods. Taking the identification of polluted area as another standard, IDW method showed a greater advantage to better reflect spatial variability characteristics of regional pollution in districts of slight pollution,moderate pollution and severe pollution.(2) The cross validation about different sampling sizes showed that the error variation amplitude of RMSE, ME and MAE from 1017 to 610 sampling points was 29.84%, 71.31% and 36.99%,respectively. And the amplitude of ME was obvious bigger compared with the first three sub samples. From the spatial feature recognition, 2033, 1830, 1423 and 1017 sampling points reflected the very similar spatial distribution characteristics in contaminated area, but the area of regional pollution was significantly expanded and the detail of spatial characteristics at all pollution levels was considerably decreased from 610 and 203 sampling points. Therefore,the reasonable sampling quantity was about 1017 sampling points in this town for the study of Cd pollution index in soil.
