Global Landslide Susceptibility Mapping Using Multi-Model Machine Learning Approaches on Geospatial Satellite Data

This study introduces a high-resolution, global landslide susceptibility model employing a multi-model machine learning framework. Aiming to surpass the 2016 NASA model, the approach leverages an enhanced global landslide catalog (UGLC) for training, a global higher-accuracy 90m MERIT DEM, and over 100 global predictive variables encompassing topographic, geological, and environmental factors. Developed on high-performance computing for operational efficiency and scalability, the model prioritizes reliability and interpretability to support future dynamic early warning systems, linking baseline susceptibility with real-time monitoring for improved disaster response in global landslide risk assessment.