Abstract
In the past, different approaches for automated landslide detection based on multispectral satellite image analysis were developed to focus on distinct landslide triggering events by analyzing bi-temporal image acquisitions. However, many regions, including the Longnan administrative region in Northwestern China, experience ongoing landslide activity requiring a continual multi-temporal analysis. Previous works have shown promising results using automated landslide mapping approaches incorporating multi-temporal NDVI image analysis completed with topographic analysis. With this study we test different multi-temporal change indices derived from Sentinel-2 images in high spatial and temporal resolution. The identified objects are assigned likelihood classes, indicating a certainty to which they can be regarded as landslides. As a result, the detection maps can be further analyzed in relation to triggering and predisposing factors and improve the regional process understanding.