Integrating VNIR Spectroscopy and Environmental Data into the RUSLE Model for Accurate Soil Erodibility and Erosion Prediction at Watershed Scale

ABSTRACT This study aimed to predict the soil erodibility factor (K factor) using visible and near-infrared (VNIR) spectroscopy coupled with a partial least-squares regression (PLSR) model and to integrate the predicted K factor into the RUSLE model for improved soil erosion assessment. The K factor was calculated based on the Universal Soil Loss Equation (USLE) model and also predicted using VNIR spectroscopy coupled with the PLSR model. The combination of these factors generated a GIS map illustrating soil loss. Our results revealed a significant negative correlation between soil permeability and the USLE K factor (r = −0.75, p < 0.05). Conversely, percent silt showed the strongest positive correlation with the K factor (r = 0.55, p < 0.05). Results also showed that the K factor had significant correlations with several spectral bands, including 450 nm (r = 0.36, p < 0.05), 641 nm (r = −0.21, p < 0.05), 990 nm (r = 0.26, p < 0.05), and 1,910 nm (r = 0.22, p < 0.05). Notably, the mean VNIR-PLSR K factor (0.28 t hr MJ−1 mm−1) was twice that of the USLE K factor (0.13 t hr−1 MJ−1 mm−1). The average soil loss predicted by the RUSLE model using the VNIR-PLSR K factor was 39.2 t hr−1 yr−1, i.e., more than double the average soil loss estimated using the USLE K factor (18.6 t hr−1 yr−1). Soil loss ranged from negligible in flat areas to over 60 t ha−1 yr−1 in mountainous regions. Approximately 20 percent of the study area experienced very low to low soil loss, while another 20 percent faced very severe soil erosion.