Improving image-based tracking of slow-moving landslides with optical PlanetScope data: A case study from the Central Andes

Optical image offset tracking is a commonly used technique for mapping surface displacements caused by landslides, glaciers, and earthquakes in planar direction. With its daily temporal and high spatial resolution of 3 m, the PlanetScope cubesat constellation provides an excellent data set for studying dynamic surface changes. However, the limited relative geolocation accuracy among PlanetScope scenes with ~10 m RMSE for the PlanetScope SuperDove constellation poses a problem for the identification of slow-moving targets whose annual displacement rates remain well below this value. In this study, we have used PlanetScope data to measure surface displacement over a slow-moving landslide with velocities between 1 and 6 m/yr in the NW Argentine Andes through image cross-correlation techniques. In this steep, rugged environment, not only the misalignment from scene to scene, but also topography-related artifacts and the changing terrain over time, are sources of error. We present several correction steps to improve coregistration accuracy between PlanetScope scenes that lower the relative geolocation accuracy between selected image pairs into the subpixel range. Through an improved scene-to-scene alignment we can better distinguish displacement signal from noise and thus obtain a better understanding of the dynamics of this slow-moving landslide and its driving factors.