Estimation of soil organic matter content based on spectral indices constructed by improved Hapke model

Soil organic matter (SOM) content is an important indicator to measure the degradation degree and fertility of soil. However, most current SOM prediction methods are based on statistical learning theory, overlooking the transmission process and physical mechanism of reflectance spectra, and lacking the physical basis of soil remote sensing. In this study, a method for estimating SOM content based on spectral indices constructed by the improved Hapke model was proposed, which started from the radiative transfer process of soil reflectance spectra and used the converted reflectance r and single scattering albedo ω as means to construct spectral indices. The prediction accuracy of these spectral indices with sensitive bands selected from laboratory-measured data (Data1) was validated using field high-spectral data (Data2), and the potential application in remote sensing of spectral indices was validated using multispectral data (Data3). As expected, these spectral indices exhibit good prediction accuracy for both field hyper-spectral data (TBI37: R2P is 73.88; RPD is 2.02) and field multispectral data (TBI17: R2P, is 67.19; RPD is 1.78). The comparative results indicate that, in terms of both accuracy and stability, spectral indices constructed by the improved Hapke model outperform those based on spectral reflectance. This study reduces the complexity of model calibration effectively, and the constructed spectral indices have clear physical meaning and good potential for fast and high accuracy prediction of SOM content at large scales.