Failure mechanism analysis of the catastrophic Junlian landslide in Southwestern China based on seismic signal analysis and numerical simulation

High-speed long-runout landslides are characterized by abrupt initiation, rapid movement, and extensive impact areas, making them among the most destructive geological hazards in mountainous regions of Southwest China. Because these events often occur suddenly, continuous observations throughout the entire movement process are rarely available, which makes it difficult to accurately characterize their dynamic evolution and failure mechanisms. To address this limitation and improve the understanding of such hazards, this study investigates two successive high-speed long-runout landslides that occurred in Junlian County, Yibin City, Sichuan Province, China, on February 8, 2025, by integrating seismic signal analysis, discrete element simulation, and limit equilibrium analysis. Landslide-induced seismic signals were used to identify movement stages and estimate centroid motion parameters, whereas discrete element simulation and limit equilibrium analysis were used to interpret the initiation failure mode and stability-controlling factors. Results indicate that the two successive landslides can be divided into three stages: initiation and acceleration, collision and redirection, and deceleration and accumulation. The estimated peak centroid velocities of the first and second landslides were approximately 45 m/s and 42 m/s, respectively, subject to significant uncertainty from mass estimation. Discrete element simulation shows that the initiation failure process was characterized by rear tensile cracking, toe shear deformation, progressive connection of the potential sliding surface, and rapid downslope movement after instability. Sensitivity analysis indicates that the rear crack connectivity ratio and internal friction angle were the key factors controlling slope stability, whereas horizontal seismic acceleration and rear-crack water depth had relatively limited influence. Overall, this study provides a seismic-signal-based method for post-event reconstruction of sudden geohazards, supporting mechanism analysis and hazard mitigation of similar landslides.