the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
ROMSOC: A regional atmosphere-ocean coupled model for CPU-GPU hybrid system architectures
Abstract. Recent years have seen significant efforts to refine the horizontal resolutions of global and regional climate models to the kilometer scale. This refinement aims to better resolve atmospheric and oceanic mesoscale processes, thereby improving the fidelity of simulations. However, these high-resolution simulations are computationally demanding, often necessitating trade-offs between resolution and simulated timescale. A key challenge is that many existing models are designed to run on central processing units (CPUs) alone, limiting their ability to leverage the full computational power of modern supercomputers, which feature hybrid architectures with both CPUs and graphics processing units (GPUs).
In this study, we introduce ROMSOC, a newly developed regional coupled atmosphere-ocean model. ROMSOC integrates the Regional Oceanic Modeling System (ROMS) in its original CPU-based configuration with the Consortium for Small-Scale Modeling (COSMO) model (v5.12), which can utilize GPU accelerators on heterogeneous system architectures. This combination efficiently exploits the hybrid CPU-GPU architecture of the Piz Daint supercomputer at the Swiss National Supercomputing Centre (CSCS), achieving a speed-up of up to six times compared to a CPU-only version with the same number of nodes.
We evaluated the model using a configuration focused on the northeast Pacific, where ROMS covers the entire Pacific Ocean with a telescopic grid, providing full ocean mesoscale-resolving refinement in the California Current System (CalCS; 4 km resolution). Meanwhile, COSMO covers most of the northeast Pacific at a 7 km resolution. This configuration was run in hindcast mode for the years 2010–2021, examining the roles of different modes of air-sea coupling at the mesoscale, including thermodynamical coupling (associated with heat fluxes) and mechanical coupling (associated with wind stress and surface ocean currents).
Our evaluation indicates that the hindcast generally agrees well with observations and reanalyses. Notably, large-scale sea surface temperature (SST) patterns and coastal upwelling are well-represented, but SSTs show a small cold bias, resulting from too-strong wind forcing. Additionally, the coupled model exhibits a deeper and more realistic simulation of the ocean mixed-layer depth with a more pronounced seasonal cycle, driven by the enhanced wind-driven mixing. On the other hand, our ROMSOC simulations reveal a negative cloud cover bias off the coast of southern California, a common issue in climate models.
- Preprint
(12186 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-2922', Anonymous Referee #1, 27 Nov 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2024-2922/egusphere-2024-2922-RC1-supplement.pdf
-
CEC1: 'Comment on egusphere-2024-2922 - No compliance with the policy of the journal', Juan Antonio Añel, 02 Dec 2024
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.htmlFirst, the server hosted by the ETH that you link to make your data available is not suitable for scientific publication. Therefore, you must publish the data in one of the repositories acceptable according to our policy.
Also, you have archived the ROMS 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, such as Zenodo. Regarding the COSMO model we can not accept that you claim that will be made available upon request. You have to provide evidence of what prevents you of publishing at least the new features of COSMO that you have developed for the submitted work.
Therefore, the current situation with your manuscript is irregular. Please, publish your code 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. Also, please include the relevant primary input/output data.
I have to note that if you do not fix this problem, we will have to reject your manuscript for publication in our journal.
Juan A. Añel
Geosci. Model Dev. Executive Editor
Citation: https://doi.org/10.5194/egusphere-2024-2922-CEC1 -
AC1: 'Reply on CEC1', Gesa Eirund, 18 Dec 2024
Dear Juan A. Añel,
We apologize for not complying with the GMD Code and Data policy, which we fully support. In our impression, the ETH Library is very suitable for scientific publication, as it is certified with the CoreTrustSeal Certification and additionally fulfils all requirements for FAIR. But we provided a copy of all our data/scripts now on zenodo, where we also included the ROMS code as well as the necessary changes to the COSMO code. In order to get the full COSMO license, we refer the reader to contact the CLM community, as we cannot provide the license for the code (https://www.cosmo-model.org/content/consortium/licencing.htm).
The zenodo link for the data/scripts is as follows: https://doi.org/10.5281/zenodo.14275348
We will include this link in a revised version of the manuscript.
Thank you very much for considering our manuscript for publishing in GMD,
Gesa Eirund
Citation: https://doi.org/10.5194/egusphere-2024-2922-AC1 -
CEC2: 'Reply on AC1', Juan Antonio Añel, 18 Dec 2024
Dear authors,
Many thanks for your reply. I would like to ask you regarding two issues:
- First, what is preventing you of sharing the raw model data in an open repository? At least, we would expect that you publish the data for the variables that you study in your submitted work. This can be easily done extracting them from the full output files. Please, clarify the size of the files and if this approach would be feasible.
- Second, thanks for pointing out how to get access to the COSMO model and the problems with its license. However, as COSMO-CLM is not maintained anymore, and the webpage could vanish, it would be good if you store the code of the model in a Zenodo private repository. In this way, the access to the model version and configuration that you have used for your work will be permanently stored, and you keep the control on who can access to the code.
Juan A. Añel
Geosci. Model Dev. Executive Editor
Citation: https://doi.org/10.5194/egusphere-2024-2922-CEC2 -
AC2: 'Reply on CEC2', Gesa Eirund, 10 Jan 2025
Dear Juan A. Añel,
thank you for your reply. Please find below our answers regarding your issues:
1. I apologize for my wording being confusing. With "processed model data" I meant exactly raw model output, but containing only the relevant variables, merged over time (as it would be a lot of individual files otherwise). For some variables, additionally some calculated values (like climatologies) are included. All commands used can be found in the netcdf history of the respective file. I hope this clarifies your issue.
2. We opened a zenodo repository with restricted access as suggested by you which now contains the COSMO and ROMS code for our model to run in coupled mode. The repository can be accessed here: https://doi.org/10.5281/zenodo.14624664.
Best regards,
Gesa Eirund on behalf of all coauthors
Citation: https://doi.org/10.5194/egusphere-2024-2922-AC2
-
AC2: 'Reply on CEC2', Gesa Eirund, 10 Jan 2025
-
CEC2: 'Reply on AC1', Juan Antonio Añel, 18 Dec 2024
-
AC1: 'Reply on CEC1', Gesa Eirund, 18 Dec 2024
-
RC2: 'Comment on egusphere-2024-2922', Anonymous Referee #2, 05 Feb 2025
General Assessment
This manuscript proposes a novel regional coupled atmosphere-ocean model (ROMSOC) that integrates COSMO and ROMS while optimizing performance on a CPU-GPU hybrid architecture. The model is applied for over Northeast Pacific with a focus on the California Current System, evaluating key air-sea interactions over a 12-year hindcast period. The manuscript also provides a detailed discussion of the potential sources of uncertainties for SST simulation bias. In general, the study presents a interesting model development work with a insightful scientific study on SST with solid method. The manuscript is well organized with detailed description of the model and thorough discussions of the result. Therefore, I would recommend it to be accepted if the following comments could be properly addressed.
Comment#1: Not really a comment but more of a curious question: the study demonstrates impressive computational efficiency with effective energy cost on a heterogeneous CPU-GPU system, but for a fixed grid configuration. How does the performance scale with different grid sizes? Similarly, as the author mentioned at line#203 it is constrained by the CPU running ROMS part, how does the performance scale with more CPU nodes? Moreover, although the practice seems very promising, is there any potential limitations, for example, is it applicable for ultra-find grid resolution like 1km?
Comment#2: The manuscript generally presents very detailed discussions regarding the potential causes for simulation bias by the coupled model, but there are limited discussions to quantify the uncertainties induced or reduced by the coupling effort. For example, SST bias of ROMSOC is primarily attributed to wind bias which indicated that the raw SST bias of ROMS is amplified by COSMO. But there is no demonstration of the 10m-wind bias by pure COSMO. I would recommend the revision to consider incorporate such kind of comparison and discussion to help better demonstrate the benefit and “side-effect” of this coupled model.
Comment#3: The study acknowledges a common bias in climate models as there are too few clouds in the southern domain over CalCS region. A more detailed discussion on why ROMSOC exhibits this behavior would be helpful to understand the limit of this specific model. This may not be directly falls into the concern of this study but: some double-momentum mechanisms have been developed for years, does COSMO support any of them, or there is any study ever try it?
Citation: https://doi.org/10.5194/egusphere-2024-2922-RC2
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
253 | 87 | 108 | 448 | 7 | 13 |
- HTML: 253
- PDF: 87
- XML: 108
- Total: 448
- BibTeX: 7
- EndNote: 13
Viewed (geographical distribution)
Country | # | Views | % |
---|---|---|---|
United States of America | 1 | 170 | 37 |
Switzerland | 2 | 28 | 6 |
Germany | 3 | 25 | 5 |
China | 4 | 20 | 4 |
Netherlands | 5 | 18 | 3 |
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1
- 170