普通及指示克里格法在水稻禁产区筛选中的应用

在对农产品产地进行监测基础上,以川东南水稻生产区为研究对象,探讨了普通克里格法及指示克里格法在农产品产地禁产区筛选中应用的可行性.研究区土壤监测统计结果表明,川东南水稻生产区以重金属Cd污染为主,区域Cd最高值为0.53mg·kg-1;监测点超过<土壤环境质量标准>(GB 15618-1995)2级标准值的比例达34.98%.探索性分析结果显示,研究区土壤Cd异常值较少,土壤Cd监测值及指示值实验半变异函数均为各向同性,采用人工干预条件下的基台值自动拟合法拟合普通及指示克里格实验半变异函数,并根据交叉验证结果确定分别采用带块金值的指数模型及带块金值的球形模型套合结构拟合普通及指示克里格实验半变异函数.通过普通和指示克里格插值结果筛选水稻禁产区后,应用误判率法评估了两种方法应用于水稻禁产区筛选的优缺点.结果表明,普通克里格法训练数据集的误判率为3.89%,验证数据集误判率为26.09%;指示克里格法训练集的误判率为0.78%,验证数据集误判率为21.73%,指示克里格法的估计精度要优于普通克里格法,更适合于研究区水稻禁产区的筛选.

Application of ordinary and indicator Kriging methods for screening non-rice-cropping areas

Based on the monitoring data of heavy metal concentrations at the rice producing district located in southeastern of Sichuan Province,P. R. China,possibility of applying ordinary and indicator Kriging to screen non-rice-cropping areas was investigated. Statistical analysis of monitoring data obtained from the study district showed that the rice producing district in the southeastern of Sichuan Province,P. R. China was mainly contaminated with cadmium,with the highest concentrations of Cd up to 0.53 mg·kg^-1. About 34.98% of the samples had a Cd concentration exceeding the second class of the National Soil Environment Guidelines (GB 15618-1995). The exploring spatial analysis showed that there were few outliers in the raw data and the experimental semivariograms of raw data and indicator data were isotropic. The method of automatic sill fitting under human intervention was used to fit the experimental semivariograms of ordinary and indicator Kriging. Then,on the basis of the cross-validation results,an exponential model with nugget effect and a nested spherical model with nugget effect were adopted to match the ordinary and indicator experimental semivariograms,respectively. The advantages and disadvantages of screening non-rice-cropping areas using the two models were assessed in terms of error rate(ER).The results showed that the ER of training data set and validation data set for ordinary Kriging was 3.89% and 26.09%,respectively,while for indicator Kriging they were 0.78% and 21.73%,respectively,which indicated that the indicator Kriging showed more advantages than ordinary Kriging when used for screening non-rice-cropping areas.