Abstract
A precise estimation of the heavy metal concentrations in soils using multispectral remote sensing technology is challenging. Herein, Landsat8 imagery, a digital elevation model, and geochemical data derived from soil samples are integrated to improve the accuracy of estimating the Cu, Pb, and As concentrations in topsoil, using the Daxigou mining area in Shaanxi Province, China, as a case study. The relationships between the three heavy metals and soil environmental factors were investigated. The optimal combination of factors associated with the elevated concentrations of each heavy metal was determined combining correlation analysis with collinearity tests. A back propagation network optimised using a genetic algorithm was trained with 80% of the data for samples and subsequently employed to estimate the heavy metal concentrations in the area. The validation results show that the RMSE of the proposed model is lower than those of the existing linear model and rule-based M5 model tree. From the spatial distribution map of the three metals concentrations using the proposed method, there are findings that high concentrations of the heavy metals studied occur in the mining area, across the slag storage area, on the sides of the road used for transporting ore materials, and along the base of slopes in the area. These findings are consistent with the survey results in the field. The validation and findings validate the effectiveness of the proposed method.