In a warming world, debris-covered glaciers are of increasing relevance in high mountain areas worldwide. Still, the complex process interactions between ice flow, debris input, transport and its effect surface mass balance that govern the advance and retreat of these glaciers are not well understood. To gain a more complete understanding of the entire system, we include and couple all these processes in a numerical flow model and aim to investigate the relationship between erosion rates debris fluxes, and debris cover thickness patterns in a changing climate.

Our approach exploits recent advances in deep learning that drastically reduce computational cost for ice flow modelling in the Instructed Glacier Model (IGM), enabling model runs at higher spatial resolution and over longer timescales. Issues of mass conservation and diffusion in englacial debris transport are circumvented by utilizing Lagrangian particle tracking approach for modelling debris transport. We assign a representative debris volume to each particle and seed particles based on user-defined rules, setting thresholds for variables related to steep rock slopes. Once particles melt out at the ice surface down-glacier, their representative debris volumes are evaluated to compute debris cover thicknesses. Mass balance is then modified accordingly using an Østrem curve.

First applications to the two contrasting cases of Zmuttgletscher (Switzerland) and Satopanth Glacier (India) look promising and are reproducing general debris cover patterns well without modifying seeding locations to fit observations. We also examine model sensitivity to key parameters, including model grid size, seeding frequency, debris input quantity, and various seeding conditions.