Characterizing glacial and paraglacial flood processes across scales using environmental seismology

Glacial and paraglacial floods are among the most destructive natural hazards in high-mountain regions. These events result from cascades of processes, which rapidly transfer large amounts of water, sediment and energy across entire catchments. Their initiation typically occurs in remote, poorly instrumented areas, while impacts propagate far downstream, strongly limiting process-based understanding at the regional scale. Here, we present preliminary results from an ongoing analysis of glacial and paraglacial hazards in the Bhote Koshi catchment (Nepal), one of the best instrumented glacierized basins at the regional scale, with continuous seismic monitoring since 2016. This study is conducted within the framework of the French PEPR IRIMA program (project IRIMONT), which aims to improve the assessment and mitigation of natural hazards in mountain regions through integrated and interdisciplinary approaches.

First, we focus on the July 2016 glacial lake outburst flood (GLOF). Seismic records of this event provide a unique opportunity to investigate its mechanics from initiation to far-field propagation. Preliminary analyses reveal distinct seismic signatures associated with different phases of the flood, characterized by systematic variations in amplitude, frequency content and phase coherence as a function of time and distance. These signatures indicate an exceptional capacity of the GLOF to mobilize large boulders, leading to seismic energy levels and inferred sediment transport that far exceed those observed during seasonal hydrological events. In parallel, we investigate the temporal evolution of slope instabilities in the Bhote Koshi catchment following the 2015 Gorkha earthquake. We apply unsupervised machine learning approaches to cluster seismic signals, identify recurrent signal families, and establish a baseline of background hydrological and geomorphic activity at the catchment scale. The seismic observations reveal sustained post-seismic landslide activity, with evolving signal characteristics reflecting the progressive relaxation of hillslopes modulated by hydrometeorological forcing. 

Overall, these preliminary results demonstrate the potential of environmental seismology, combined with data-driven approaches, to bridge the gap between local process understanding and regional-scale hazard assessment.