InSAR-Derived Localized Deformation and Slow-Moving Landslide Inventory in Northern Pakistan

The distribution of slow-moving landslides has significance in landscape modification and hazard assessment. Monitoring such landslides is more challenging because of their spatial and temporal variability, particularly in tectonically active regions. In these regions, structural discontinuities exert significant control, requiring risk assessment at both local and regional aspects. The Karakoram-Hindu Kush-Himalaya (KHH), i.e., the orogenic belt in northern Pakistan, is a natural laboratory for studying the geological hazards due to its neotectonism, high seismicity, and diverse precipitation patterns. This region encompasses a complex geological framework, including the Nanga Parbat Syntaxis, Hazara-Kashmir Syntaxis, Hunza fault system, Tirich Mir Suture Zone, Shyok Suture, and the Indus-Kohistan Suture zones. These structures prompt various land sliding activities, yet inventories of slow-moving landslides remain scarce in northern Pakistan. The region is also traversed by the China-Pakistan Economic Corridor (CPEC) through the Karakoram Highway (KKH), an old Silk Road, where landslides severely threaten infrastructure and transportation. Here, we report our recent work regarding the actively slow-moving landslide distribution in this 19350 km² region. We combine the InSAR phase-gradient stacking technique with a deep learning-based YOLOv3 network to detect localized deformation from thousands of wrapped interferograms. Analyzing eight years of Sentinel-1 data (2016-2024), we detected and mapped 1,066 active slow-moving landslides in the Hazara Kashmir region. Further, we extended this analysis to the Khunjerab-Chitral alternate route of the CPEC, detecting 859 active landslides along this corridor. Several large, rapidly moving landslides were also recognized, posing significant risks to underlying villages and the route’s stability. These results are validated using optical imageries and field observations to create the first comprehensive inventory of slow-moving landslides in northern Pakistan. Validation against geomorphological features, published landslides, and field observation confirmed an overall precision of 87%, with detected targets corresponding to landslide features, while 13% were classified as false detections. This study underscores the critical need for monitoring and managing geological hazards in this rapidly uplifting tectonic region.