Coupled model for susceptibility mapping of hydro-fluctuation belt rock degradation in the Jinsha River reservoirs

Hydro-fluctuation belt (HFB) rock mass degradation poses significant risks to bank slope stability. While existing studies primarily focus on laboratory experiments and localized investigations, systematic regional-scale susceptibility mapping of HFB rock masses remains unexplored. This study focuses on the four cascade reservoirs along the Jinsha River in China. Systematically assesses 17 factors across four categories: topography, geology, environment, and reservoir operations. By combining the information value (IV) model with machine learning algorithms, an evaluation framework integrating geostatistical and machine learning methods is constructed, achieving the first regional-scale susceptibility mapping of HFB rock mass degradation. The results indicate that compared with the single IV model, the coupled models exhibit better applicability, among which the IV + RF (random forest) model performs the best (area under curve = 0.928, kappa = 0.897). The key driving factors of degradation include slope, bank slope shape, lithology, proximity to faults and rainfall intensity. Spatial analysis reveals that high-sensitivity zones within the reservoir areas predominantly occur in Triassic, Jurassic, Ordovician-Silurian, Cambrian, and Proterozoic strata, characterized by interbedded fragile-competent sequences. The findings of this study can deepen the understanding of susceptibility evaluation methods for HFB rock mass degradation and provide new insights for reservoir disaster prevention, mitigation, and long-term safety monitoring.