Rapid generation of event-based landslide inventory and susceptibility assessment using state of the art approaches

Event-based landslide inventories are crucial for understanding the impact of triggering events such as earthquakes and extreme rainfall on landslides and improving hazard and risk analysis. These inventories are scarce and often generated through visual interpretation, making them labor-intensive and dependent on mapper skills. This thesis explores different deep learning and machine learning approaches for automated landslide detection in the Himalayas and Western Ghats (India), and evaluates the quality and completeness of the resulting inventories. The thesis also compares pixel- and object-based approaches for landslide susceptibility assessment in Austria. The study focuses on the proper evaluation of the completeness and quality of landslide inventories, the use of GIS and remote sensing technology for mapping objects automatically, and the comparison of per-pixel and object-based approaches for landslide susceptibility mapping. The dissertation leads to a deeper understanding of the importance of optical and topographical factors for the generation of event-based landslide inventories and susceptibility assessments.