Modeling the Alps across large spatial and temporal scales using theInstructed Glacier Model (IGM)

The European Alps and its unique features were largely formed from additive processes like
sediment deposition and subtractive processes such as glacial erosion. Spanning millions of
years, these processes that helped shape the Alps are not fully understood due to their complexity
and tenancy to be coupled with each other. Existing landscape evolution models that include
these processes are limited in their computational power - often only allowing a coarser spatial
resolution. A high spatial resolution, and by extension maintaining higher-order physics fidelity,
is also imperative in accurately reconstructing the Alps. We aim to address this limitation by
using the community-led Instructed Glacier Model (IGM) that leverages Graphical Processing
Units (GPUs) and scientific machine learning (SciML) to accelerate computation. Here, we adapt
IGM to be a landscape evolution model (LEM) by including relevant mechanisms for landscape
evolution such as glacial abrasion, quarrying, fluvial erosion, isostatic rebound, and hill-slope
processes.

To demonstrate its capacity, we first benchmark its results against traditional landscape evo-
lution models (i.e. iSOSIA), validating that, though IGM is a physics-informed machine learning
model, it remains a process-based LEM. We furthermore aim to show its efficiency at modeling
across a wide range of scales such as multiple alpine catchments as well as longer temporal
periods such as during the Quaternary. As such, we hope our modeling approach can be used
for various applications such as exploring how glaciers are linked to these underlying processes,
inverse problems to achieve better model-data agreements, and ensembles across long temporal
or spatial scales.