the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
ISSM-ML: a modular framework for integrating machine learning models into the Ice-sheet and Sea-level System Model (ISSM v2026.4)
Abstract. Machine learning (ML) models are increasingly used in glaciology to replace computationally expensive model components or emulate specific processes. These approaches, however, are commonly developed as standalone tools or within application specific software and cannot be readily deployed within established ice-sheet modeling frameworks. We present ISSM-ML, a modular framework for integrating externally trained, Python-based ML models into the Ice-sheet and Sea-level System Model (ISSM). Model development and training remain within the Python ecosystem, while ISSM invokes selected ML modules at runtime and retains control of model initialization, numerical solvers, physical analyses, time stepping, and other procedures in its original workflow. We demonstrate the framework using a multilayer perceptron to replace the nonlinear velocity-dependent term in the Weertman basal-friction law. Two training strategies are evaluated, one using conventional scaling and the other using distribution-aware preprocessing. When the scaling-only model is coupled to ISSM, the stress balance solver fails to converge. The distribution-aware model, in contrast, allows ISSM to converge reliably throughout the transient simulation, with a good agreement with the analytical-friction reference. These results demonstrate that offline prediction accuracy alone is not sufficient for evaluating an ML component intended for numerical-model coupling. Their behavior must also be assessed within the simulation environment in which they operate, including convergence, stability, and feedback with evolving model states.
- Preprint
(2445 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 21 Dec 2026)
-
CEC1: 'Comment on egusphere-2026-6193 - No compliance with the policy of the journal', Juan Antonio Añel, 02 Nov 2026
reply
-
AC1: 'Reply on CEC1', Gong Cheng, 03 Nov 2026
reply
Dear Dr. Añel,
Thank you for your comments and for highlighting these issues.
We have now updated our repositories to include the necessary materials, including the training dataset and trained machine learning model weights. We have also clarified the software dependencies and specified the versions of the libraries used in this work.
We would like to clarify that ISSM is an open-source ice-sheet model written in C++, with MATLAB and Python interfaces serving as front ends. Our ISSM–ML implementation supports both interfaces, allowing users to run the coupled model without relying on MATLAB. GNU Octave can also be used as an alternative to MATLAB. We have clarified these points in the revised "Code and Data Availability" section.
Regarding the Helheim Glacier simulations, the data used in this study were generated using ISSM as part of previously published work (Cheng et al., 2022; Koo et al., 2025). The corresponding simulation code and datasets are publicly archived on Zenodo. The dataset used to train the friction emulator is also included in the updated FrictionEmulatorTrain repository.
We have revised the Code and Data Availability section for the next version of the manuscript as follows:
Code and data availability. The coupled ISSM–ML model is available in the ISSM 2026.4 release https://doi.org/10.5281/zenodo.18166184 (Morlighem et al., 2026). ISSM is an open-source C++ ice-sheet model with MATLAB and Python interfaces, both of which support the coupled ISSM–ML implementation. GNU Octave can also be used as an alternative to MATLAB. The friction emulator training code is available in the FrictionEmulatorTrain v1.0.1 release https://doi.org/10.5281/zenodo.18166184 (Liu, 2026), which includes the training code, training dataset, and trained model weights saved as checkpoint files. The training data were generated using ISSM in previous studies (Cheng, 2022, Koo, 2025), with the corresponding Helheim Glacier simulation code and data archived at https://doi.org/10.5281/zenodo.5879143 (Cheng, 2021) and  https://doi.org/10.5281/zenodo.11392220 (Koo, 2024). The software versions used in this work were Python 3.14.4, NumPy 2.4.3, PyTorch 2.11.0, SciPy 1.17.1, and pybind11 3.1.0.
We hope these updates address the concerns regarding code and data availability and reproducibility. Thank you again for your guidance and consideration.
Best regards,
Gong Cheng, on behalf of all co-authorsCitation: https://doi.org/10.5194/egusphere-2026-6193-AC1 -
CEC2: 'Reply on AC1', Juan Antonio Añel, 03 Nov 2026
reply
Dear authors,
Thanks for addressing this issue so quickly. I have checked the repositories and we can consider now the current version of your manuscript in compliance with the code policy of the journal.
Juan A. Añel
Geosci. Model Dev. Executive Editor  Â
Citation: https://doi.org/10.5194/egusphere-2026-6193-CEC2
-
CEC2: 'Reply on AC1', Juan Antonio Añel, 03 Nov 2026
reply
-
AC1: 'Reply on CEC1', Gong Cheng, 03 Nov 2026
reply
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 285 | 126 | 31 | 442 | 32 | 37 |
- HTML: 285
- PDF: 126
- XML: 31
- Total: 442
- BibTeX: 32
- EndNote: 37
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Dear authors,
Unfortunately, after checking your manuscript, it has come to our attention that it does not comply with our "Code and Data Policy".
https://www.geoscientific-model-development.net/policies/code_and_data_policy.html
First, in the repositories that you cite in the Code and Data Availability section of your manuscript you have not included output data or model weights, which we require to be published in the case of machine learning models. Secondly, your work relies on a set of libraries and third-party software, both private and free, for which you do not provide version numbers. When the presented work relies on third-party software this is of the greater importance, as different versions of software can have different implementations of algorithms, functions, etc. and therefore produce different results. Therefore, please, provide the versions of all the software used in your work. Moreover, as you use the privative software Matlab, which code is unknown, it would be good if you can clarify if your code is compatible or not with GNU Octave, to avoid such dependence on a non-FLOSS software.
Finally, you must provide repositories for all the data used and produced in your work. At least you seem to use data simulations from a glacier model for the Helheim glacier (subsection 5.1). However, in your manuscript no model is cited to produce such data and no repository is listed for the code and data. Please, clarify this, and if you have used (as it seems to be the case) other model to produce data for validation, provide a repository acceptable according to the policy of the journal that contains both of them.
The GMD review and publication process depends on reviewers and community commentators being able to access, during the discussion phase, the code and data on which a manuscript depends, and on ensuring the provenance of replicability of the published papers for years after their publication. Therefore, please, solve the mentioned issues as soon as possible. We cannot have manuscripts under discussion that do not comply with our policy.Â
Later, if the Topical Editor decides to continue with the review or publication process of your manuscript and you are requested to upload a new version of it, then The 'Code and Data Availability’ section of your manuscript must also be modified to cite the new repository locations, and corresponding references added to the bibliography.Â
I must note that if you do not fix this problem, we cannot continue with the peer-review process or accept your manuscript for publication in GMD.
Juan A. Añel
Geosci. Model Dev. Executive Editor