Satellite-based tracking of slow-moving landslides: challenges and perspectives

Slow-moving landslides represent a significant hazard to local communities and infrastructure in mountainous regions worldwide. Given their challenging and often inaccessible terrain, satellite imagery holds great potential for monitoring landslides from space. In this study we use optical data from Sentinel-2 and PlanetScope satellites for tracing surface displacement across slow-moving landslides through image cross-correlation. Our work particularly focuses on the variables affecting measurement precision, including orthorectification errors and mismatches due to variable shading or seasonal snow cover. Erroneous measurements can be reduced when image pairs are carefully selected based on their view angles and sun positions. This practice, however, severely limits the number of potential image pairs, resulting in disconnected networks of displacement maps. This in turn poses problems when solving for a displacement time series using an inversion technique. Here, we evaluate the effect of network connectivity and measurement noise on inversion results using both synthetic and real-world data. Our findings support the extraction of accurate displacement estimates from remotely sensed data, advancing the detection potential of landslides and their dynamic behaviors.