Key points are not available for this paper at this time.
We present the concept and capabilities of IGM (https://github.com/jouvetg/igm), a Python-based modeling tool designed for efficiently simulating glacier evolution across various scales. IGM integrates ice thermomechanics, climate-driven surface mass balance, mass conservation, and other processes. Within IGM, the update of all physical model components involves a series of mathematical operations on horizontal raster grids, performed by the Tensorflow library. This design choice results in high parallelization capabilities, particularly beneficial when executed on GPU hardware. The most challenging aspect of parallelization within IGM is the ice flow model, which leverages a physics-informed convolutional neural network trained from high-order ice flow physics. Conversely, components like the positive-degree day surface mass balance or the enthalpy thermal scheme take advantage of pixel-wise parallelization. Beyond its computational efficiency, IGM offers a user-friendly coding structure and modularity to promote community development, an OGGM-based module for accessing the data, data assimilation through underlying automatic differentiation tools, and post-processing vizualization routines. We present a comprehensive workflow, which includes data preprocessing, inverse and forward modeling, and rendering of results, enabling the rapid modeling of any mountain glacier globally by providing its RGI ID within a few minutes.
Jouvet et al. (Fri,) studied this question.