Available heavy metals concentrations in agricultural soils: relationship with soil properties and total heavy metals concentrations in different industries

Heavy metal (HM) pollution in agricultural soils has arisen sharply in recent years. However, the impact of main factors on available HMs concentrations in agricultural soils of the three main industries (smelting, chemical and mining industry) is unclear. Herein, soil properties (pH, cation exchange capacity (CEC) and texture (sand, slit, clay)), total and available concentrations were concluded based on the results of 165 research papers from 2000 to 2023 in Web of Science database. In the three industries, the correlation and redundancy analysis were used to study the correlation between main factors and available concentrations, and quantitatively analyzed the contribution of each factor to available concentrations with gradient boosting decision tree model. The results showed that different factors had varying degrees of impact on available metals in the three main industries, and the importance of same factors varied in each industry, as for soil pH, it was most important for available Pb and Zn in the chemical industry, but the total concentrations were most important in the smelting and mining industry. There was no significant correlation between total and available concentrations. Soil properties involved in this paper (especially soil pH) were negatively correlated with available concentrations. This study provides effective guidance for the formulation of soil pollution control and risk assessment standards based on industry classification in the three major industrial impact areas. Available HMs concentrations (Pb, Cd, Zn and Cu) are easily absorbed by plants, this will increase the potential risks of agricultural soils and products. The study used correlation and redundancy analysis to consider the relationship of the main factors on available concentrations from the perspective of industry, and quantitatively analyzed the contribution degree of the factors on available HMs using gradient boosting decision tree model in machine learning. The results showed that the factors in typical industries involving HMs had varying degrees of impact on available concentrations, which provided a theoretical basis for differentiated management in each industry.