Deep learning speeds up ice flow modelling by several orders of magnitude
Abstract This paper introduces the Instructed Glacier Model (IGM) – a model that simulates ice dynamics, mass balance and its coupling to predict the evolution of glaciers, icefields or ice sheets. The novelty of IGM is that it models the ice flow by a Convolutional Neural Network, which is trained...
Published in: | Journal of Glaciology |
---|---|
Main Authors: | , , , , , |
Other Authors: | |
Format: | Article in Journal/Newspaper |
Language: | English |
Published: |
Cambridge University Press (CUP)
2021
|
Subjects: | |
Online Access: | http://dx.doi.org/10.1017/jog.2021.120 https://www.cambridge.org/core/services/aop-cambridge-core/content/view/S0022143021001209 |
id |
crcambridgeupr:10.1017/jog.2021.120 |
---|---|
record_format |
openpolar |
spelling |
crcambridgeupr:10.1017/jog.2021.120 2024-09-30T14:37:51+00:00 Deep learning speeds up ice flow modelling by several orders of magnitude Jouvet, Guillaume Cordonnier, Guillaume Kim, Byungsoo Lüthi, Martin Vieli, Andreas Aschwanden, Andy Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung 2021 http://dx.doi.org/10.1017/jog.2021.120 https://www.cambridge.org/core/services/aop-cambridge-core/content/view/S0022143021001209 en eng Cambridge University Press (CUP) https://creativecommons.org/licenses/by-nc-sa/4.0/ Journal of Glaciology volume 68, issue 270, page 651-664 ISSN 0022-1430 1727-5652 journal-article 2021 crcambridgeupr https://doi.org/10.1017/jog.2021.120 2024-09-11T04:04:56Z Abstract This paper introduces the Instructed Glacier Model (IGM) – a model that simulates ice dynamics, mass balance and its coupling to predict the evolution of glaciers, icefields or ice sheets. The novelty of IGM is that it models the ice flow by a Convolutional Neural Network, which is trained from data generated with hybrid SIA + SSA or Stokes ice flow models. By doing so, the most computationally demanding model component is substituted by a cheap emulator. Once trained with representative data, we demonstrate that IGM permits to model mountain glaciers up to 1000 × faster than Stokes ones on Central Processing Units (CPU) with fidelity levels above 90% in terms of ice flow solutions leading to nearly identical transient thickness evolution. Switching to the GPU often permits additional significant speed-ups, especially when emulating Stokes dynamics or/and modelling at high spatial resolution. IGM is an open-source Python code which deals with two-dimensional (2-D) gridded input and output data. Together with a companion library of trained ice flow emulators, IGM permits user-friendly, highly efficient and mechanically state-of-the-art glacier and icefields simulations. Article in Journal/Newspaper Journal of Glaciology Cambridge University Press Journal of Glaciology 1 14 |
institution |
Open Polar |
collection |
Cambridge University Press |
op_collection_id |
crcambridgeupr |
language |
English |
description |
Abstract This paper introduces the Instructed Glacier Model (IGM) – a model that simulates ice dynamics, mass balance and its coupling to predict the evolution of glaciers, icefields or ice sheets. The novelty of IGM is that it models the ice flow by a Convolutional Neural Network, which is trained from data generated with hybrid SIA + SSA or Stokes ice flow models. By doing so, the most computationally demanding model component is substituted by a cheap emulator. Once trained with representative data, we demonstrate that IGM permits to model mountain glaciers up to 1000 × faster than Stokes ones on Central Processing Units (CPU) with fidelity levels above 90% in terms of ice flow solutions leading to nearly identical transient thickness evolution. Switching to the GPU often permits additional significant speed-ups, especially when emulating Stokes dynamics or/and modelling at high spatial resolution. IGM is an open-source Python code which deals with two-dimensional (2-D) gridded input and output data. Together with a companion library of trained ice flow emulators, IGM permits user-friendly, highly efficient and mechanically state-of-the-art glacier and icefields simulations. |
author2 |
Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung |
format |
Article in Journal/Newspaper |
author |
Jouvet, Guillaume Cordonnier, Guillaume Kim, Byungsoo Lüthi, Martin Vieli, Andreas Aschwanden, Andy |
spellingShingle |
Jouvet, Guillaume Cordonnier, Guillaume Kim, Byungsoo Lüthi, Martin Vieli, Andreas Aschwanden, Andy Deep learning speeds up ice flow modelling by several orders of magnitude |
author_facet |
Jouvet, Guillaume Cordonnier, Guillaume Kim, Byungsoo Lüthi, Martin Vieli, Andreas Aschwanden, Andy |
author_sort |
Jouvet, Guillaume |
title |
Deep learning speeds up ice flow modelling by several orders of magnitude |
title_short |
Deep learning speeds up ice flow modelling by several orders of magnitude |
title_full |
Deep learning speeds up ice flow modelling by several orders of magnitude |
title_fullStr |
Deep learning speeds up ice flow modelling by several orders of magnitude |
title_full_unstemmed |
Deep learning speeds up ice flow modelling by several orders of magnitude |
title_sort |
deep learning speeds up ice flow modelling by several orders of magnitude |
publisher |
Cambridge University Press (CUP) |
publishDate |
2021 |
url |
http://dx.doi.org/10.1017/jog.2021.120 https://www.cambridge.org/core/services/aop-cambridge-core/content/view/S0022143021001209 |
genre |
Journal of Glaciology |
genre_facet |
Journal of Glaciology |
op_source |
Journal of Glaciology volume 68, issue 270, page 651-664 ISSN 0022-1430 1727-5652 |
op_rights |
https://creativecommons.org/licenses/by-nc-sa/4.0/ |
op_doi |
https://doi.org/10.1017/jog.2021.120 |
container_title |
Journal of Glaciology |
container_start_page |
1 |
op_container_end_page |
14 |
_version_ |
1811640638291050496 |