Ice-flow model emulator based on physics-informed deep learning

Abstract Convolutional neural networks (CNN) trained from high-order ice-flow model realisations have proven to be outstanding emulators in terms of fidelity and computational performance. However, the dependence on an ensemble of realisations of an instructor model renders this strategy difficult t...

Full description

Bibliographic Details
Published in:Journal of Glaciology
Main Authors: Jouvet, Guillaume, Cordonnier, Guillaume
Format: Article in Journal/Newspaper
Language:English
Published: Cambridge University Press (CUP) 2023
Subjects:
Online Access:http://dx.doi.org/10.1017/jog.2023.73
https://www.cambridge.org/core/services/aop-cambridge-core/content/view/S0022143023000734
id crcambridgeupr:10.1017/jog.2023.73
record_format openpolar
spelling crcambridgeupr:10.1017/jog.2023.73 2024-10-06T13:49:38+00:00 Ice-flow model emulator based on physics-informed deep learning Jouvet, Guillaume Cordonnier, Guillaume 2023 http://dx.doi.org/10.1017/jog.2023.73 https://www.cambridge.org/core/services/aop-cambridge-core/content/view/S0022143023000734 en eng Cambridge University Press (CUP) http://creativecommons.org/licenses/by/4.0/ Journal of Glaciology page 1-15 ISSN 0022-1430 1727-5652 journal-article 2023 crcambridgeupr https://doi.org/10.1017/jog.2023.73 2024-09-11T04:04:56Z Abstract Convolutional neural networks (CNN) trained from high-order ice-flow model realisations have proven to be outstanding emulators in terms of fidelity and computational performance. However, the dependence on an ensemble of realisations of an instructor model renders this strategy difficult to generalise to a variety of ice-flow regimes found in the nature. To overcome this issue, we adopt the approach of physics-informed deep learning, which fuses traditional numerical solutions by finite differences/elements and deep-learning approaches. Here, we train a CNN to minimise the energy associated with high-order ice-flow equations within the time iterations of a glacier evolution model. As a result, our emulator is a promising alternative to traditional solvers thanks to its high computational efficiency (especially on GPU), its high fidelity to the original model, its simplified training (without requiring any data), its capability to handle a variety of ice-flow regimes and memorise previous solutions, and its relatively simple implementation. Embedded into the ‘Instructed Glacier Model’ (IGM) framework, the potential of the emulator is illustrated with three applications including a large-scale high-resolution (2400x4000) forward glacier evolution model, an inverse modelling case for data assimilation, and an ice shelf. Article in Journal/Newspaper Ice Shelf Journal of Glaciology Cambridge University Press Journal of Glaciology 1 15
institution Open Polar
collection Cambridge University Press
op_collection_id crcambridgeupr
language English
description Abstract Convolutional neural networks (CNN) trained from high-order ice-flow model realisations have proven to be outstanding emulators in terms of fidelity and computational performance. However, the dependence on an ensemble of realisations of an instructor model renders this strategy difficult to generalise to a variety of ice-flow regimes found in the nature. To overcome this issue, we adopt the approach of physics-informed deep learning, which fuses traditional numerical solutions by finite differences/elements and deep-learning approaches. Here, we train a CNN to minimise the energy associated with high-order ice-flow equations within the time iterations of a glacier evolution model. As a result, our emulator is a promising alternative to traditional solvers thanks to its high computational efficiency (especially on GPU), its high fidelity to the original model, its simplified training (without requiring any data), its capability to handle a variety of ice-flow regimes and memorise previous solutions, and its relatively simple implementation. Embedded into the ‘Instructed Glacier Model’ (IGM) framework, the potential of the emulator is illustrated with three applications including a large-scale high-resolution (2400x4000) forward glacier evolution model, an inverse modelling case for data assimilation, and an ice shelf.
format Article in Journal/Newspaper
author Jouvet, Guillaume
Cordonnier, Guillaume
spellingShingle Jouvet, Guillaume
Cordonnier, Guillaume
Ice-flow model emulator based on physics-informed deep learning
author_facet Jouvet, Guillaume
Cordonnier, Guillaume
author_sort Jouvet, Guillaume
title Ice-flow model emulator based on physics-informed deep learning
title_short Ice-flow model emulator based on physics-informed deep learning
title_full Ice-flow model emulator based on physics-informed deep learning
title_fullStr Ice-flow model emulator based on physics-informed deep learning
title_full_unstemmed Ice-flow model emulator based on physics-informed deep learning
title_sort ice-flow model emulator based on physics-informed deep learning
publisher Cambridge University Press (CUP)
publishDate 2023
url http://dx.doi.org/10.1017/jog.2023.73
https://www.cambridge.org/core/services/aop-cambridge-core/content/view/S0022143023000734
genre Ice Shelf
Journal of Glaciology
genre_facet Ice Shelf
Journal of Glaciology
op_source Journal of Glaciology
page 1-15
ISSN 0022-1430 1727-5652
op_rights http://creativecommons.org/licenses/by/4.0/
op_doi https://doi.org/10.1017/jog.2023.73
container_title Journal of Glaciology
container_start_page 1
op_container_end_page 15
_version_ 1812177695899910144