Abstract
To elucidate the hydro-mechanical evolution of rainfall-triggered loess-mudstone interface landslides and improve monitoring and early warning, we conducted a large-scale indoor physical model test under artificial rainfall conditions. The model was instrumented with pore-water pressure and earth pressure sensors, as well as terrestrial laser scanning (TLS) for spatially continuous, hourly displacement mapping. XRF/XRD and mechanical tests were adopted to quantify the water-induced softening characteristics of mudstone.The results show a power-law increase in water absorption and rapid strength degradation, with the unconfined compressive strength decreasing from 4.90 MPa to 0.82 MPa within 3 h. Rainfall promotes the formation of an interfacial water film and argillation of mudstone, which weakens inter-particle bonding and significantly reduces the interface shear strength, representing the key trigger for sliding. Pore-water pressure evolves quasi-synchronously with rainfall but with a slight lag, exhibiting a three-stage pattern: stable - accelerated rise - rapid decline. TLS captured deformation precursors at the crest and slope surface at approximately 1560 min, providing a 140-min lead time over sensor-detected anomalies (approximately 1700 min).TLS-derived displacement fields cross-validate with pressure-based indicators to characterize progressive destabilization, which culminates in failure under continuous rainfall. This study clarifies the water-film-controlled softening mechanism and demonstrates the superior early-warning sensitivity of TLS for interface-type landslides, providing a scientific basis for multi-index fusion monitoring and the formulation of refined early-warning thresholds.