Landslide clusters pose significant threats to mountainous regions worldwide, with their complex failure mechanisms and dynamic behaviors requiring comprehensive investigation. This study focuses on the Hongyacun landslide cluster in Qinghai Province, China, a representative example of multi-stage slope failures triggered by hydrological and geological interactions. By integrating field observations with advanced numerical modeling, we aim to reconstruct the full kinematic process of four sub-landslides within the cluster and elucidate the critical factors governing their initiation, motion, and deposition. The research provides insights into flow-slide dynamics under rainfall conditions, addressing a key gap in hazard assessment methodologies for analogous landslide-prone regions. A multidisciplinary approach was employed combining detailed field surveys and three-dimensional numerical simulations. Field investigations mapped the spatial distribution and geometric characteristics of the four sub-landslides (total volume: 2.46 x 10(6) m(3)), while geotechnical analyses identified moisture-induced strength reduction as a primary destabilization factor. The Landslides Post-Failure 3D (LPF3D) simulator was implemented to reconstruct landslide kinematics under two scenarios: natural (pre-rainfall) and rainfall-saturated states. The simulations incorporated soil rheological parameters, hydrological conditions, and terrain data, with particular attention to fluid-solid interactions during motion. Numerical simulations revealed distinct motion patterns between dry and saturated conditions. Continuous rainfall infiltration increased soil saturation by 18-25%, reducing shear strength while enhancing material fluidity. This hydrological transformation generated significant hydrodynamic effects during sliding, with fluid drag forces amplifying mobility rather than providing resistance. Sub-landslide H2-2 exhibited the most hazardous behavior, achieving peak velocity of 32.5 m/s within 70 s and traveling 700 m-42% farther than dry-state simulations predicted. Deposit patterns from all sub-landslides showed >85% spatial consistency with field observations, validating the model's predictive capability. This study demonstrates that rainfall-induced pore pressure development creates dual destabilization effects: reducing shear resistance while enabling fluid-mediated lubrication. The paradoxical role of hydrodynamic forces-enhancing mobility through drag-induced momentum transfer rather than damping-explains the exceptional runout distances observed. The H2-2 sub-landslide's predominance in damage potential correlates with its unique geometric positioning and hydrological connectivity within the cluster. These findings advance understanding of multi-landslide interaction mechanisms and provide a validated framework for assessing rainfall-triggered landslide cascades. The LPF3D methodology proves particularly effective for hazard zonation in complex terrain, offering critical inputs for early warning systems and mitigation planning in comparable geological settings.

来源平台：FRONTIERS IN EARTH SCIENCE