Coal mining has environmental impacts on surrounding areas, including heavy metal contamination of soil. This study explores the feasibility of using hyperspectral remote sensing to determine the heavy metal (Cr, Ni, Cu, Zn, Cd, Pb) content of soils in a coal‐mining area in the city of Zoucheng, Shandong Province, China. We used a plasma mass spectrometer to measure the heavy metal contents of soils and an ASD Field Spec4 spectrometer to measure soil hyperspectral data. Savitzky–Golay (SG) convolution smoothing and multiplicative scatter correction (MSC) were applied to the data, along with multiple mathematical transformations. Finally, a regression model for estimating heavy metal content of soils was developed using partial least squares regression (PLSR) analysis. Results show that the average heavy metal content of study soils was lower than the national standard value of soil environmental quality. The model's predictive accuracy is extremely high for Ni (R2 = 0.923 and RMSE = 0.831 by modeling; R2 = 0.879 and RMSE = 1.292 by testing); ideal for Cr, Cu, Zn, and Pb; and insufficient for Cd. Preprocessing the reflectance spectra with SG convolution smoothing in combination with MSC and reciprocal logarithm transformation yields the highest model accuracy. Hyperspectral PLSR modeling can effectively predict heavy metal content of soils in coal‐mining areas, and preprocessing spectral data is crucial for achieving high prediction accuracy.
Core Ideas
Quantitative inversion can provide technical support for monitoring of soil heavy metals.
Quantitative inversion can provide theoretical information for further environmental recovery in mining areas.
Hyperspectral remote sensing can quickly evaluate the status of heavy metal contamination in soil.