Leveraging artificial intelligence and digital terrain models for active and inactive landslide detection

This study presents an innovative method for landslide detection using a boosted cascade of simple features within a machine learning framework to efficiently and accurately identify areas susceptible to landslides. By leveraging AdaBoost algorithm-enhanced decision trees, this method analyzes terrain features from hillshade extracted from Digital Terrain Models (DTMs). The southwestern part of Cyprus, characterized by diverse geological-geotechnical conditions and rugged geomorphology, serves as the test area. This study evaluated the effectiveness of various DTM resolutions and cell sizes in detecting different types of landslides: active, dormant, and relict. The findings indicate that high-resolution DTMs and smaller cell sizes are most effective for identifying relict landslides with subtle features, albeit at high computational costs. For active landslides with steeper geomorphologies, a coarser cell size combined with moderate-resolution DTMs achieves an 80% success rate, significantly reducing computational time.