Skip to Main Content
Monitoring soil salinization has been a difficult process in arid lands due their large spatial and temporal variability. Hyperspectral remote sensing has offered a potential for faster detection of salinization process but mostly from empirical approaches. In this paper, an integrated approach combining model inversion and empirical regressions has been proposed for soil salt content (SSC) estimation from hyperspectral information obtained from controlled laboratory experiments. All soil samples were artificially salinized using Na2SO4, NaCl, and Na2CO3 (99% purity) salts to different levels and to different soil-moisture conditions, since soil moisture often jointly affects reflectance spectra with SSC. Hapke model was calibrated and validated for its simulation on soil reflectance and showed good agreements with measured data. The optimal values of single scattering albedo that was inversely retrieved from the Hapke model had good relationships with SSC at 2000-2200 nm for each treatment even under various soil-moisture conditions. Taking usage of these findings, the integrated approach obtained high accuracies on SSC estimations with R2's of 0.90, 0.86, and 0.72 and slightly dropped R2 's of 0.89, 0.81, and 0.67 for NaCl-, Na2SO4-, and Na2CO3-type saline soils under respective dry and wet conditions. The R2 decreased to 0.55 and 0.53 for dry and wet soils when salt types were ignored. The integrated approach provides a novel as well as an efficient way for SSC estimation from reflected spectra, and hence, we foresee its potential applications for large-scale SSC mapping from reflectance measurements.