Abstract
Mapping spatial distribution of soil contaminants at contaminated sites is the basis of risk assessment. Hotspots can cause strongly skewed distribution of the raw contaminant concentrations in soil, and consequently can require suitable normalization prior to interpolation. In this study, three normalization methods including normal score, Johnson, and Box-Cox transformation were performed on the concentrations of two low-molecular weight (LMW) PAHs (i.e., acenaphthene (Ace) and naphthalene (Nap)) and two high-molecular weight (HMW) PAHs (i.e., benzo(a)pyrene (BaP) and benzo(b)fluoranthene (BbF)) in soils of a typical coking plant in North China. The estimating accuracy of soil LMW and HMW PAHs distribution using ordinary kriging with different normalization methods was compared. The results showed that all transformed data passed the Kolmogorov-Smirnov test, indicating that all three data transformation methods achieved normality of raw data. Compared to Box-Cox-ordinary kriging, normal score-, and Johnson-ordinary kriging had higher estimating accuracy of the four soil PAHs distribution. In cross-validation, smaller root-mean-square error (RMSE) and mean error (ME) values were observed for normal score-ordinary kriging for both LMW and HMW PAHs compared to Johnson- and Box-Cox-ordinary kriging. Thus, normal score transformation is suitable for alleviating the impact of hotspots on estimating accuracy of the four selected soil PAHs distribution at this coking plant. The findings can provide insights into reducing uncertainty in spatial interpolation at PAHs-contaminated sites.