A Novel Statistical Method of Defining Geochemical Baselines and Source Identification for Trace Metals in Soil in Zhangjiagang County, China

Abstract

Establishing accurate geochemical baselines by normalization procedures (NP) is important for distinguishing contaminated soils from the uncontaminated ones. In the areas with intensive anthropogenic influences, geochemical baselines are easily affected by the contaminated soils. The commonly used least squares method is not robust enough because it is sensitive to outliers, resulting in distorted regression relationship. A novel robust regression method, which can diminish the influences of outliers proposed, is proposed for defining more accurate geochemical baselines and the source of trace metals. Zhangjiagang County, which is under with intensive anthropogenic influences and high pollution risks, was selected as the study area. Trace metal data obtained from subsoils were used as normalizers, and SMDM-estimators were used in the robust regression for defining the geochemical baselines and the sources of trace metals in this area. The robust statistical method can successfully be applied in normalization procedures for defining soil geochemical baselines. The SMDM robust regressions reduced the influence of outliers, building a more reasonable regression fitted to data points for the selected trace metals in topsoil and subsoil. Subsoil trace metals are less affected than topsoil, so they can be used as normalizers to replace general elemental normalizers. It is found that the geochemical baselines defined by the novel robust regression method proposed in this study can fully reflect the natural variability of soil trace metals. Moreover, this method can also apparently reduce the effects of outliers on statistical analysis, improving the accuracy of geochemical baselines in the areas with intensive anthropogenic influences and high pollution risks.