A Method for Predicting Unsaturated Loess Landslides Based on Rainfall Intensity and Validation

Landslide prediction remains a pivotal yet enduring challenge in disaster risk reduction. This study presents a novel physical model tailored for unsaturated loess slopes in China’s Loess Plateau, uniquely integrating dynamic rainfall infiltration processes with depth-resolved shear strength degradation. Unlike conventional approaches reliant on static thresholds, the model quantifies stability through moisture-dependent shear strength and stability coefficients across soil depths, capturing the transient interplay between rainfall intensity and stratigraphic heterogeneity. Validation was achieved via meticulously designed laboratory experiments, where a 0.5 m high, 40° inclined slope with controlled drainage (10 cm gravel base) was subjected to simulated rainfall. Pore water pressure and volumetric moisture content were monitored at 5 cm depth intervals (L1–L5), revealing three distinct infiltration phases: rapid percolation, stabilization, and saturation. Key findings include: (1) Pore water pressure surges exponentially as saturation initiates, particularly in deeper layers; (2) Shear strength deteriorates progressively, with surface layers losing 70% of strength within 150 minutes, while deeper strata exhibit delayed but complete strength loss upon saturation; (3) Stability coefficients ( Fs ) decline nonlinearly with depth, transitioning from shallow sliding (residual strength insufficient to resist sliding forces) to deep-seated failure (total strength loss). The model’s reliability was further corroborated by numerical simulations replicating field-scale conditions, demonstrating its capacity to predict both timing and depth of slope instability. By bridging theoretical rigor with practical validation, this work advances landslide prediction frameworks, offering actionable insights for region-specific risk assessment and mitigation strategies in loess terrains.