Detecting landslides with satellites

In a climate with more rain, the extent of rain-triggered landslides is expected to increase. NVE's landslide database RegObs provides an overview of registered landslides in Norway as a basis for assessment, notification and mapping of landslide hazard. The Norwegian Public Roads Administration and Bane Nor contribute to updating the database after landslide events affecting their infrastructure. Unfortunately, the database is incomplete since few landslides are registered where there are no roads, railway lines or other infrastructure. In order to improve prediction models, there is a need for new methods to detect and record landslides.

Landslides can be detected by comparing images of an area taken before and after a heavy rain event. Mapping landslides in this way requires good images for comparison, which is not always easy. Parts of the terrain will always be in shadow, and there may be clouds in the way. Radar images can contain abundant "noise" that varies from image to image. PhD candidate Erin Lindsay's new method involves removing the "noise" by combining several images taken in the time before and after the landslide into two composite "before" and "after" images which are then compared. The image processing is done quickly with the web platform Google Earth Engine. The technique enables the use of satellite data to discover many new landslide events, also in areas without infrastructure, and thus improve the prediction models for landslides.

Read more:
Lindsay E, Frauenfelder R, Rüther D, Nava L, Rubensdotter L, Strout J & Nordal S: Multi-Temporal Satellite Image Composites in Google Earth Engine for Improved Landslide Visibility: A Case Study of a Glacial Landscape. Remote Sensing 2022, Vol 14(10), 2301; doi.org/10.3390/rs14102301, ISSN 2072-4292 (Published online 10 May 2022)