Landslide Activity in the European Alps ‐ Part 1: A Machine Learning Based Instrumental Catalog From the Analysis of Seismological Data

Abstract Landslides in the European Alps are a growing concern in the context of climate change. However, landslide catalogs for the European Alps remain incomplete, often lacking precise timing and containing few, if any, entries for remote areas. Seismology offers a complementary approach at the regional scale by detecting large landslides (e.g., volume > ) that generate seismic signals. We present an automated workflow to detect and locate seismogenic landslides using continuous seismic data from 440 stations across the European Alps (2017–2018). The workflow combines automatic seismic signal detection, supervised machine learning classification, and NonLinLoc relocation. Starting from a training data set in which landslide seismic signals represent only 11%, we achieve true positive rates of up to 88% through a data augmentation strategy. We also automatically locate 6/7 documented landslides with errors below 15 km. Finally, we identified 144 previously undocumented landslides, 41 of which form clusters in specific Alpine massifs. The presented method provides a foundation for long‐term monitoring of landslide activity across the Alps and for investigating how meteorological conditions influence landslide occurrence in different mountain regions.