the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 17 Sep 2025
Applying Corrective Machine Learning in the E3SM Atmosphere Model in C++ (EAMxx)
Abstract. The Simplified Cloud-Resolving E3SM Atmosphere Model (SCREAM) is the newest addition to the family of earth system models capable of explicitly resolving convective systems. SCREAM is a kilometer-scale configuration of the advanced E3SM Atmosphere Model (EAMxx), designed for heterogeneous systems. While the enhanced accuracy of kilometer-scale modeling offers significant benefits, it comes with a substantial computational cost, limiting feasible simulation durations to only a few years, even on the fastest supercomputers. Machine learning presents an opportunity for scientists to achieve the high accuracy of storm-resolving models at a significantly reduced cost. Building on the previous success of applying corrective machine learning (ML) to the FV3 model, this study explores the effects of implementing corrective ML in EAMxx-SCREAM. We also address the computational challenges of integrating our implementation of corrective ML, which is written in Python, with the C++/Kokkos EAMxx driver, as well as potential reasons why this approach has not proved as effective for EAMxx-SCREAM as for the FV3 model.
Competing interests: At least one of the (co-)authors is a member of the editorial board of Geoscientific Model Development. The peer-review process was guided by an independent editor, and the authors also have no other competing interests to declare.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(2648 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(2648 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
CEC1: 'No compliance with the policy of the journal', Juan Antonio Añel, 11 Oct 2025
-
CC1: 'Reply on CEC1', Aaron Donahue, 28 Oct 2025
My apologies for submitting a manuscript that did not adhere to the code and data availability requirements of the manuscript. I have taken measures to publish the data and the software used in this manuscript to the acceptable manuscripts. In response to the editor comments:
1) The software is available for download via Zenodo with the DOI of https://doi.org/10.5281/zenodo.17469329
2) The data is available for download via Zenodo with the DOI of https://doi.org/10.5281/zenodo.17469234
3) I no longer reference the Terai et al. paper for instructions on accessing the data from that study. Instead readers are directed to the following zenodo doi's: https://doi.org/10.5281/zenodo.14579433 and https://doi.org/10.5281/zenodo.14578966
This is all reflected in an updated Code and Data Availability section that I can submit as a revision to the paper if requested.
Citation: https://doi.org/10.5194/egusphere-2025-3883-CC1 -
CEC2: 'Reply on CC1', Juan Antonio Añel, 30 Oct 2025
Dear authors,
Many thanks for your reply, and the willingness to address the outstanding issues in it. However, some issues remain. Regarding the Terai et al. data, the submitted version of your manuscript states that is the source for high resolution data used in the study. Now you refer to readers to a Zenodo repository, which does not contain the data, but some text with external links to sites that are not permanent repositories. Therefore, we can not consider your manuscript in compliance with the policy of the journal until you have addressed this issue and published in a suitable repository all the data that you have used for your work.
In this regard, the situation of non-compliance of your manuscript remains, and we can not consider it in its current form for publication or review in our journal.
Please, address the outstanding issues and reply to this comment with the relevant information.
Juan A. Añel
Geosci. Model Dev. Executive Editor
Citation: https://doi.org/10.5194/egusphere-2025-3883-CEC2 -
AC1: 'Reply on CEC2', Aaron Donahue, 19 Nov 2025
Thank you again to the editor for helping us understand how to make sure our data availability section is in line with GMD policy. After some discussion offline I believe a solution has been found. I will break the ~300GB of data we used for this study into a set of Zenodo repositories and add the doi and reference for them to the manuscript. While I establish these repositories would it be possible to move forward with the review process? I will post another public comment with links to each one once they are complete.
Citation: https://doi.org/10.5194/egusphere-2025-3883-AC1
-
AC1: 'Reply on CEC2', Aaron Donahue, 19 Nov 2025
-
CEC2: 'Reply on CC1', Juan Antonio Añel, 30 Oct 2025
-
CC1: 'Reply on CEC1', Aaron Donahue, 28 Oct 2025
-
RC1: 'Comment on egusphere-2025-3883', Anonymous Referee #1, 04 Dec 2025
The paper is in general well written and well structured. It provides interesting findings about transferring a concept of a machine learning based correction scheme of subgrid source terms from the FV3GFS general circulation model to the EAMxx general circulation model. The content of this paper is helpful for the modeling community as it clearly states encountered problems of such a migration. Despite this I recommend the authors to expand the interpretation of the results of their hybrid simulations. Moreover all figures require an improved description what is shown in the respective captions. Likewise I strongly recommend to the authors to adjust their way how they cite data or software used to fulfill the standards of Copernicus journals.
As a result I recommend major revision of this submitted manuscript. Attached you will find some specific comments and a detailed list of technical comments with respective line numbers in the submitted manuscript.
-
AC2: 'Reply on RC1', Aaron Donahue, 21 Apr 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-3883/egusphere-2025-3883-AC2-supplement.pdf
-
AC2: 'Reply on RC1', Aaron Donahue, 21 Apr 2026
-
RC2: 'Comment on egusphere-2025-3883', Anonymous Referee #2, 03 Feb 2026
The article presents work with the aim to use machine learning to nudge a coarse version of the EAMxx-SCREAM model towards the results of a high resolution run with the goal to obtain results corresponding to high-res runs at significantly lower computational costs. For this, the authors follow previous work with another Earth system model, FV3, and use the openly available algorithms that were successfully implemented for FV3. However, in the present article, the results of the nudging were less promising than in the previous work. The authors speculate as to why this is the case. They also argue that, despite the less effective nudging, there are additional benefits stemming from the model developments made, such as Python to C++ bridging added to the model which will also be beneficial for other developments.
In general, this is an interesting and highly timely work that should be published. I also feel very favorable to publishing work that has not proven entirely as successful as initially hoped. However, in the present form, the article feels incomplete and lacks motivation and discussion and thorough investigation as to why the work was not as effective as for the FV3 model, which would make the article useful for readers attempting similar endeavors. I therefore recommend major revisions as detailed in the attachment.Â
-
AC3: 'Reply on RC2', Aaron Donahue, 21 Apr 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-3883/egusphere-2025-3883-AC3-supplement.pdf
-
AC3: 'Reply on RC2', Aaron Donahue, 21 Apr 2026
Interactive discussion
Status: closed
-
CEC1: 'No compliance with the policy of the journal', Juan Antonio Añel, 11 Oct 2025
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, you have archived part of your code on GitHub. However, GitHub is not a suitable repository for scientific publication. GitHub itself instructs authors to use other long-term archival and publishing alternatives. Additionally, you have not made the data used in your manuscript (all input and output data) publicly available, instead linking to sites such as nersc.gov or previous publications to access some of it. Given all these shortcomings, your manuscript should not have been accepted for Discussions and peer review.
Therefore, the current situation with your manuscript is irregular. Please, publish all your code and data in one of the appropriate repositories and reply to this comment with the relevant information (link and a permanent identifier for it (e.g., DOI)) as soon as possible, as we can not accept manuscripts in Discussions that do not comply with our policy.Â
Also, you must include a modified 'Code and Data Availability' section in a potentially reviewed manuscript, containing the information of the new repositories.Given that no reviewer has been secured for your manuscript yet, I advise the Topical Editor to refrain from inviting reviewers or taking any action on your submission until the issues with the code and data have been adequately addressed and resolved.
Juan A. Añel
Geosci. Model Dev. Executive EditorCitation: https://doi.org/10.5194/egusphere-2025-3883-CEC1 -
CC1: 'Reply on CEC1', Aaron Donahue, 28 Oct 2025
My apologies for submitting a manuscript that did not adhere to the code and data availability requirements of the manuscript. I have taken measures to publish the data and the software used in this manuscript to the acceptable manuscripts. In response to the editor comments:
1) The software is available for download via Zenodo with the DOI of https://doi.org/10.5281/zenodo.17469329
2) The data is available for download via Zenodo with the DOI of https://doi.org/10.5281/zenodo.17469234
3) I no longer reference the Terai et al. paper for instructions on accessing the data from that study. Instead readers are directed to the following zenodo doi's: https://doi.org/10.5281/zenodo.14579433 and https://doi.org/10.5281/zenodo.14578966
This is all reflected in an updated Code and Data Availability section that I can submit as a revision to the paper if requested.
Citation: https://doi.org/10.5194/egusphere-2025-3883-CC1 -
CEC2: 'Reply on CC1', Juan Antonio Añel, 30 Oct 2025
Dear authors,
Many thanks for your reply, and the willingness to address the outstanding issues in it. However, some issues remain. Regarding the Terai et al. data, the submitted version of your manuscript states that is the source for high resolution data used in the study. Now you refer to readers to a Zenodo repository, which does not contain the data, but some text with external links to sites that are not permanent repositories. Therefore, we can not consider your manuscript in compliance with the policy of the journal until you have addressed this issue and published in a suitable repository all the data that you have used for your work.
In this regard, the situation of non-compliance of your manuscript remains, and we can not consider it in its current form for publication or review in our journal.
Please, address the outstanding issues and reply to this comment with the relevant information.
Juan A. Añel
Geosci. Model Dev. Executive Editor
Citation: https://doi.org/10.5194/egusphere-2025-3883-CEC2 -
AC1: 'Reply on CEC2', Aaron Donahue, 19 Nov 2025
Thank you again to the editor for helping us understand how to make sure our data availability section is in line with GMD policy. After some discussion offline I believe a solution has been found. I will break the ~300GB of data we used for this study into a set of Zenodo repositories and add the doi and reference for them to the manuscript. While I establish these repositories would it be possible to move forward with the review process? I will post another public comment with links to each one once they are complete.
Citation: https://doi.org/10.5194/egusphere-2025-3883-AC1
-
AC1: 'Reply on CEC2', Aaron Donahue, 19 Nov 2025
-
CEC2: 'Reply on CC1', Juan Antonio Añel, 30 Oct 2025
-
CC1: 'Reply on CEC1', Aaron Donahue, 28 Oct 2025
-
RC1: 'Comment on egusphere-2025-3883', Anonymous Referee #1, 04 Dec 2025
The paper is in general well written and well structured. It provides interesting findings about transferring a concept of a machine learning based correction scheme of subgrid source terms from the FV3GFS general circulation model to the EAMxx general circulation model. The content of this paper is helpful for the modeling community as it clearly states encountered problems of such a migration. Despite this I recommend the authors to expand the interpretation of the results of their hybrid simulations. Moreover all figures require an improved description what is shown in the respective captions. Likewise I strongly recommend to the authors to adjust their way how they cite data or software used to fulfill the standards of Copernicus journals.
As a result I recommend major revision of this submitted manuscript. Attached you will find some specific comments and a detailed list of technical comments with respective line numbers in the submitted manuscript.
-
AC2: 'Reply on RC1', Aaron Donahue, 21 Apr 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-3883/egusphere-2025-3883-AC2-supplement.pdf
-
AC2: 'Reply on RC1', Aaron Donahue, 21 Apr 2026
-
RC2: 'Comment on egusphere-2025-3883', Anonymous Referee #2, 03 Feb 2026
The article presents work with the aim to use machine learning to nudge a coarse version of the EAMxx-SCREAM model towards the results of a high resolution run with the goal to obtain results corresponding to high-res runs at significantly lower computational costs. For this, the authors follow previous work with another Earth system model, FV3, and use the openly available algorithms that were successfully implemented for FV3. However, in the present article, the results of the nudging were less promising than in the previous work. The authors speculate as to why this is the case. They also argue that, despite the less effective nudging, there are additional benefits stemming from the model developments made, such as Python to C++ bridging added to the model which will also be beneficial for other developments.
In general, this is an interesting and highly timely work that should be published. I also feel very favorable to publishing work that has not proven entirely as successful as initially hoped. However, in the present form, the article feels incomplete and lacks motivation and discussion and thorough investigation as to why the work was not as effective as for the FV3 model, which would make the article useful for readers attempting similar endeavors. I therefore recommend major revisions as detailed in the attachment.Â
-
AC3: 'Reply on RC2', Aaron Donahue, 21 Apr 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-3883/egusphere-2025-3883-AC3-supplement.pdf
-
AC3: 'Reply on RC2', Aaron Donahue, 21 Apr 2026
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 6,894 | 1,113 | 220 | 8,227 | 144 | 140 |
- HTML: 6,894
- PDF: 1,113
- XML: 220
- Total: 8,227
- BibTeX: 144
- EndNote: 140
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Aaron S. Donahue
Elynn Wu
W. Andre Perkins
Peter M. Caldwell
Christopher S. Bretherton
Finn O. Rebassoo
Jean-Christophe Golaz
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(2648 KB) - Metadata XML
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, you have archived part of your code on GitHub. However, GitHub is not a suitable repository for scientific publication. GitHub itself instructs authors to use other long-term archival and publishing alternatives. Additionally, you have not made the data used in your manuscript (all input and output data) publicly available, instead linking to sites such as nersc.gov or previous publications to access some of it. Given all these shortcomings, your manuscript should not have been accepted for Discussions and peer review.
Therefore, the current situation with your manuscript is irregular. Please, publish all your code and data in one of the appropriate repositories and reply to this comment with the relevant information (link and a permanent identifier for it (e.g., DOI)) as soon as possible, as we can not accept manuscripts in Discussions that do not comply with our policy.Â
Also, you must include a modified 'Code and Data Availability' section in a potentially reviewed manuscript, containing the information of the new repositories.
Given that no reviewer has been secured for your manuscript yet, I advise the Topical Editor to refrain from inviting reviewers or taking any action on your submission until the issues with the code and data have been adequately addressed and resolved.
Juan A. Añel
Geosci. Model Dev. Executive Editor