Soil spectral simulation and soil parameter retrieval with an optimized four-flux MLG radiative transfer model

Soil moisture (SM), with its strong water absorption characteristics, profoundly alters the soil reflectance spectra and reduces the reliability of predicting other soil parameters. To achieve high-precision spectral simulation and remove SM effects efficiently, we propose a novel method based the optimized four-flux Maheu-Letoulouzan-Gouesbet (MLG) radiative transfer model. The model incorporates key soil properties (e.g., SM, organic matter, iron oxides, texture fractions) to better characterize soil spectral formation mechanisms. Additionally, the genetic algorithm, with a spectral continuity constraint, is developed for global parameter optimization, thereby simplifying the calculation process and improving the simulation reliability. Based on the optimal parameters, the SM factor is independently excluded and then the spectra are reconstructed to remove SM effects. Results demonstrated that the optimized MLG model achieved the highest accuracy in soil spectral simulation under both dry and wet conditions. Compared with the Kubelka-Munk (KM) model, the coefficient of determination (R 2 ) improved from 0.87 to 0.94 in the ICRAF dry soil dataset and from 0.87 to 0.94 in the Nong'an wet soil dataset. The model maintained high accuracy (R 2 = 0.81) even when the SM content exceeded 30 %, with strong effects in the water absorption bands around 1400 and 1900 nm. After removing SM effects from in-situ field spectra, corrected spectra exhibited stronger correlations with organic matter, iron oxides, and texture fractions, significantly enhancing prediction accuracy (R 2 > 0.90, the relative percent difference (RPD) increase by 0.88–3.00). This physically-grounded method underscores the potential for accurate soil spectral simulation and SM effects removal, thereby strengthening the foundation for large-scale soil monitoring via satellite remote sensing.