Towards an automated regional landslide mapping system using remote sensing images 

Regional-scale landslide mapping is a crucial activity that contributes to estimate triggering thresholds to be used in early warning systems, emergency reliefs, hazard maps, and spatial planning. However, mapping hundreds or thousands of individual landslides is often a tedious and time-consuming preparatory step for those who require a quality landslide inventory in the other end. There is an increasing availability of free, relatively high-resolution satellite images, along with great developments in image classification techniques using statistical models and machine learning, and future prospects of drone-based land use monitoring.

The innovation potential of Erin Lindsay’s PhD project for the partners lies in development of an accurate, user-friendly GIS-based tool for landslide mapping from remote sensing images. The tool may be based on a large global landslide database, which utilizes a machine-learning approach to image classification. This will supplement existing landslide identification efforts in Norway. It could also be used in other countries to develop a landslide catalogue, or early warning systems of their own, or to provide rapid landslide mapping following extreme triggering events, for the purposes of emergency response.

The supervisor team is professor Steinar Nordal (NTNU), Dr.scient Jose Cepeda (NGI), PhD Graziella Devoli (NVE) and PhD Lena Rubensdotter (NGU).