Soil toxic metals have strong spatial heterogeneity, and their sources vary among regions. Thus, this study integrated the Catreg and geographically weighted regression (GWR) models to quantitatively extract the main source proxies (numerical and categorical variables were analyzed simultaneously) for different toxic metals and analyze the spatial heterogeneity of the distributions of these sources. Pb, Cd and Hg were the predominant toxic metals in soil. Of the samples with Pb, Cd and Hg, 84.12 %, 68.03 % and 41.57 % exceeded the background values, and 5.36 %, 6.42 % and 5.43 % were moderately contaminated according to the geoaccmulation index, respectively. Industrial activities, with relative importance values of 17.82 %, 31.54 % and 26.51 % for Cd, Hg and Pb, respectively, were the predominant source of these metals especially, in their high-content cluster areas (central urban areas). Soil available phosphorus was another important factor (relative importance values of 13.03 %, 13.41 % and 25.55 % for Cd, Hg and Pb, respectively), and agricultural activities (especially the overuse of phosphoric fertilizers) were identified as an anthropogenic source of these toxic metals. Soil parent material had the greatest influence on As and Cr, with relative importance values of 19.88 % and 19.09 %, respectively, especially in their high-content accumulation area (the eastern coastal area), indicating that these toxic metals mainly come from natural sources. Slope had important impacts on toxic metal accumulation (relative importance values of 17.48 %, 21.22 %, 12.40 % and 16.13 % for Cd, Hg, Cr and As, respectively) by influencing industrial distribution and pollutant migration. By changing the soil adsorption capacity, soil organic matter (explaining 13.01 % of Pb) and soil pH (explaining 14.58 % of As and 12.40 % of Cr) strongly influenced toxic metal accumulation. This study highlights the benefits of the integrated Catreg-GWR model for analyzing multiple spatially heterogeneous environmental data types (numerical and categorical variables), providing a potential foundation for local pollution prevention.
Read full abstract