Early Warning System For Debris Flows

Debris flows pose significant threats to infrastructure and life. Previous work has established empirical rainfall thresholds for debris flow triggering for single well-studied catchments. However, these approaches suffer from limited transferability to catchments with diverse geology and cannot accommodate varying precipitation patterns driven by a changing climate. In this study we use machine learning (ML) to enhance an existing debris flow early warning system (EWS) for 26 high-impact catchments in the Canton of Valais, Switzerland that triggers alerts when a unique precipitation threshold is exceeded. We extend this methodology to a further 100 catchments in the region with varying geology. Our ML-driven EWS offers a promising avenue to improve debris flow prediction and monitoring, contributing to more effective risk mitigation strategies in a geologically diverse landscape.