Topographic effects substantially distort surface reflectance and vegetation indices in mountainous regions, leading to systematic biases in ecological and biophysical parameter retrievals. This study proposes an illumination-stratified topographic correction method (SCSCTS), which extends the conventional Sun-Canopy-Sensor plus C (SCS + C) model by introducing an illumination-based zonal radiative compensation scheme and explicitly accounting for terrain occlusion and adjacent terrain reflection. Using Landsat 8 surface reflectance data from a karst mountainous region in Guizhou, China, the performance of SCSCTS was evaluated and compared with SCS + C and the path length correction (PLC) model under spring and winter conditions. Quantitative results show that, relative to SCS + C, SCSCTS reduced the dependence of near-infrared reflectance on topographic illumination by up to 99.4% under low solar elevation conditions and by 53.8%-67.2% in spring. The proportion of anomalous reflectance pixels decreased by 72.0% in spring and 93.9% in winter, while shaded-sunlit reflectance differences were reduced by 39.8% and 80.8%, respectively. Land-cover-weighted radiometric deviations decreased by 10.3% under high and 24.9% under low solar elevation conditions. At the vegetation index level, SCSCTS reduced terrain-related correlations of EVI, NIRv, and RDVI by 78.6% in spring and 86.1% in winter compared with SCS + C, with regression slopes approaching zero. Compared with PLC, SCSCTS exhibited lower seasonal sensitivity and more stable correction performance under strong shadow conditions. The illumination-stratified correction framework improves radiometric consistency for both reflectance and vegetation indices, suppresses overcorrection effects, and enhances the robustness of remote sensing parameter retrievals in complex mountainous environments. This approach provides a physically based and operationally feasible solution for topographic correction in mountainous remote sensing applications.
Shen et al. (Thu,) studied this question.