Prediction of soil heavy metal concentrations in copper tailings area using hyperspectral reflectance

This study aimed to explore soil spectral characteristics from spectral data obtained in laboratory and field, and establish the hyperspectral prediction models by stepwise multiple regression (SMR) and partial least squares regression (PLSR) for predicting concentrations of six heavy metals (Cd, As, Pb, Cr, Ni and Zn) in tailings area. The performance of SMR and PLSR was also compared. The results show that the preprocessed spectral data can enhance the spectral characteristics of heavy metals in the soil. The maximum correlation coefficients of the six heavy metal elements of Cd, As, Pb, Cr, Ni, and Zn are 0.639, - 0.734, - 0.715, - 0.772, - 0.706, and 0.631, respectively; reciprocal first-order processing and normalized first-order processing have the most significant effect. Compared to field spectral data, the laboratory spectral data have a higher correlation with heavy metal content, the optimal wavebands of Cd, As, Pb, Cr, Ni and Zn are 707 nm, 505 nm, 505 nm, 468 nm, 468 nm and 610 nm, respectively; compared to SMR, the prediction model of heavy metal content established by PLSR method has higher prediction accuracy and better fitting effect. Among them, the model established by PLSR for heavy metal Ni and Cr in soil under reciprocal first-order treatment has the best prediction effect, and the model determination coefficient (R-2) reaches more than 0.7, the verification root mean square error (RMSE) value is only about 10% of the mean value of the heavy metal content in the study area, which shows the better prediction effect.