One main concern in climate science is to reduce uncertainties on sea level predictions. In particular, these uncertainties depend on the ice sheet mass loss including Greenland. Iceberg calving at Greenland glaciers accounts for up to one half of ice loss at marine-terminating glacier termini. Under certain circumstances, near-grounded glacier termini release ice through buoyancy-driven calving of unstable icebergs which capsize against the terminus. Slow rotations of such icebergs exert a force on the ice cliff, responsible for the generation of magnitude 5 glacial earthquakes. Such seismic signals have been recorded for the last twenty-five years by the global seismological networks including broadband stations GLISN deployed in Greenland. Since the early 2000s, occurrences of glacial earthquakes have increased and their spatial distribution has also evolved on the continent. One concern is the evolution of the volume of capsizing icebergs.

Field observations on capsizing icebergs are lacking, making then seismic signals remarkable tools to quantify and monitor calving events and investigate physical processes at stake. We aim to extract information from continuous seismograms. Indeed, the characteristics of generated seismic waves depend on the volume of the iceberg and the whole dynamic of the capsizing iceberg (Sergeant et al. GRL 2016, JGR 2018). The global aim of this work is to compare recorded and synthetic seismic signals calculated with a model of iceberg capsize by solving an inverse problems on the model components. 

Iceberg capsize dynamics involves fluid dynamics coupled to solid mechanics and then depends on: iceberg-water interactions, iceberg-glacier friction, glacier-sea floor friction, elasto-viscoplastic deformation of ice; and only little field data is available. Solving directly fluid flow, solid motion, and contact equations even in two dimensions is very costly and can hardly be used to generate catalogs of iceberg capsizes and generated forces. Therefore, a simplified mechanical model of icebergs capsizing in water has been developed based on few assumptions. The proposed model, named SAFIM (semi-analytical floating iceberg model) accounts for hydrostatic pressure, pressure drag, and added mass in order to mimic the effects of hydrodynamic flow on the iceberg motion. The effects of these fluid components have been validated based on a separate state-of-the-art Computational Fluid Dynamics code which can handle free surface and arbitrary iceberg configurations. The comparison between the two models helps on refining the assumptions made on capsizing-iceberg-water interactions in SAFIM model, especially pressure drag. Indeed, pressure drag presents a very important effect to be accounted for to accurately model iceberg capsize dynamics and generated seismic signals. Finally, the SAFIM model is used to reproduce seismic signals and then estimate the dimensions of capsizing icebergs for few well-documented events.