Detecting large-scale landslides in high-altitude permafrost region along the Qinghai-Tibet engineering corridor using InSAR and slope unit-based clustering

Permafrost degradation on the Qinghai–Tibet Plateau, driven by climate warming, has increased slope instability and landslide hazards, threatening key infrastructure such as the Qinghai–Tibet Railway and Highway. This study addresses landslide detection in the challenging Qinghai–Tibet Plateau environment, introduces a landslide identification method integrates InSAR with slope unit-based clustering. Ascending and descending Sentinel-1A SAR images were processed using time-series InSAR to generate LOS deformation velocity maps over an 8.4-year period. A 10 km-wide buffer zone along both sides of the railway and highway was designated as the study area. By integrating optical imagery, field investigation, and DBSCAN clustering with slope unit segmentation, 104 landslides including retrogressive thaw slumps were identified. Spatial distribution patterns were analyzed, and a SHAP-based interpretation of dominant factors and their interactions was conducted. Results show that elevation is the decisive factor in landslide occurrence, while precipitation is the dominant controlling variable. Lithology and fault proximity also have significant impacts. Retrogressive thaw slump development is primarily controlled by fault activity and active-layer thickness in permafrost regions. This study provides a robust technical framework and essential data to support landslide early detection and infrastructure protection assessment in high-altitude permafrost environments.