Accuracy and uncertainty analysis of soil Bbf spatial distribution estimation at a coking plant-contaminated site based on normalization geostatistical technologies

作者：Liu, Geng; Niu, Junjie; Zhang, Chao; Guo, Guanlin^*

来源：Environmental Science and Pollution Research, 2015, 22(24): 20121-20130.

DOI：10.1007/s11356-015-5122-2

• Geostatistics
• Normalization methods
• Risk estimation
• Contaminated site
• Polynuclear aromatic hydrocarbons

Data distribution is usually skewed severely by the presence of hot spots in contaminated sites. This causes difficulties for accurate geostatistical data transformation. Three types of typical normal distribution transformation methods termed the normal score, Johnson, and Box-Cox transformations were applied to compare the effects of spatial interpolation with normal distribution transformation data of benzo(b)fluoranthene in a large-scale coking plant-contaminated site in north China. Three normal transformation methods decreased the skewness and kurtosis of the benzo(b)fluoranthene, and all the transformed data passed the Kolmogorov-Smirnov test threshold. Cross validation showed that Johnson ordinary kriging has a minimum root-mean-square error of 1.17 and a mean error of 0.19, which was more accurate than the other two models. The area with fewer sampling points and that with high levels of contamination showed the largest prediction standard errors based on the Johnson ordinary kriging prediction map. We introduce an ideal normal transformation method prior to geostatistical estimation for severely skewed data, which enhances the reliability of risk estimation and improves the accuracy for determination of remediation boundaries.

• 出版日期2015-12
• 单位环境基准与风险评估国家重点实验室; 太原师范学院; 中国环境科学研究院