Inversion of Total Copper Content in Mining Soils with Different Spectral Pretreatment Techniques Using AHSI/ZY1-02D Data

Satellite hyperspectral remote sensing is a useful technology to extract soil heavy metal content. This paper introduces an inversion method of soil copper (Cu) content using AHSI/ZY1-02D data and assesses the inversion accuracy with field investigation data. Different spectral pretreatments are compared and then an inversion model of soil Cu content based on random frog leaping and regression tree ensemble is established. Compared with the original spectrum, the utilized spectral transformations can significantly increase the number of sensitive bands for soil Cu inversion, and the correlation between Cu content and spectrum can be improved as well. We also find the reciprocal spectrum and its differential transformations have the highest correlation with the soil Cu content. Moreover, the R2 and RMSE of the inversion model are 0.826 and 596.79 mg/kg, respectively, indicating that AHSI/ZY1-02D hyperspectral data can be used to estimate the total Cu content in the soil at mining area.