Deep Learning of Systematic Ocean Model Errors in a Coupled GCM From Data Assimilation Increments.
Saved in:
| Title: | Deep Learning of Systematic Ocean Model Errors in a Coupled GCM From Data Assimilation Increments. |
|---|---|
| Authors: | Verma, Tarun1 (AUTHOR) tarunverma.geos@gmail.com, Lu, F.2 (AUTHOR), Adcroft, A.2 (AUTHOR), Zanna, L.3 (AUTHOR), Gnanadesikan, A.1 (AUTHOR) |
| Source: | Journal of Advances in Modeling Earth Systems. Apr2026, Vol. 18 Issue 4, p1-22. 22p. |
| Subject Terms: | *Circulation models, Artificial neural networks, Data assimilation, Deep learning |
| Abstract: | We present a novel, data‐driven approach to predict systematic model errors in the ocean component of a coupled general circulation model leveraging deep learning and data assimilation. We examine the skill of the proposed scheme in learning systematic model errors, including their spatial patterns, variance, scales, and test its sensitivity to different predictors and neural network architecture. The scheme utilizes local state variables such as ocean temperature, salinity, velocities, and surface fluxes to predict corrections to temperature tendency for the upper 1,000 m in the ocean on daily timescales. The performance is evaluated on the withheld test data set and compared against the empirical climatological temperature corrections that are geographically dependent. The performance is depth‐dependent, with significant improvements over the benchmark in the upper 20 m in the ocean. It degrades rapidly with depth but remains comparable to the climatology benchmark. Neural networks can capture up to 40−50% $40-50\%$ of the daily variance in temperature increments in the upper 20 m relative to the benchmark's 20% $20\%$. The improvements are associated with networks predicting finer spatiotemporal scales than the benchmark. They are expected to perform better in reducing surface ocean mixed layer bias than previously used techniques. Despite being column‐local without geographical inputs, networks can sufficiently reproduce spatial patterns on daily and longer timescales. The patterns consist of corrections to regional dynamical features such as western boundary currents, equatorial undercurrents, bathymetry‐related corrections in the Southern Ocean, and warm surface increments over subtropical and midlatitude belts. Plain Language Summary: The ocean is a complex system, and we use ocean general circulation models to study it. However, these models are imperfect and have errors in representing the subgrid‐scale processes. We present a new way to correct these errors using deep learning and data assimilation. This method uses information about the ocean thermodynamic state, such as temperature, velocity, and surface fluxes, to predict and correct errors. We found that the new method performs better in the upper 20 m of the ocean and captures a significant fraction of daily corrections to the temperature equation. This new method can help us reduce bias in the upper ocean mixed layer. Key Points: Neural networks can predict spatiotemporal patterns of data assimilation increments using only local state variablesThe skill of neural networks exceeds that of the state‐independent climatological benchmark in the upper oceanThe predictability of upper ocean increments from fluxes and vertical gradients highlights biases in ocean mixed layer representation [ABSTRACT FROM AUTHOR] |
| Copyright of Journal of Advances in Modeling Earth Systems is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | GreenFILE |
|
Full text is not displayed to guests.
Login for full access.
|
|
| FullText | Links: – Type: pdflink Text: Availability: 1 |
|---|---|
| Header | DbId: 8gh DbLabel: GreenFILE An: 193321851 AccessLevel: 6 PubType: Academic Journal PubTypeId: academicJournal PreciseRelevancyScore: 0 |
| IllustrationInfo | |
| Items | – Name: Title Label: Title Group: Ti Data: Deep Learning of Systematic Ocean Model Errors in a Coupled GCM From Data Assimilation Increments. – Name: Author Label: Authors Group: Au Data: <searchLink fieldCode="AR" term="%22Verma%2C+Tarun%22">Verma, Tarun</searchLink><relatesTo>1</relatesTo> (AUTHOR)<i> tarunverma.geos@gmail.com</i><br /><searchLink fieldCode="AR" term="%22Lu%2C+F%2E%22">Lu, F.</searchLink><relatesTo>2</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Adcroft%2C+A%2E%22">Adcroft, A.</searchLink><relatesTo>2</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Zanna%2C+L%2E%22">Zanna, L.</searchLink><relatesTo>3</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Gnanadesikan%2C+A%2E%22">Gnanadesikan, A.</searchLink><relatesTo>1</relatesTo> (AUTHOR) – Name: TitleSource Label: Source Group: Src Data: <searchLink fieldCode="JN" term="%22Journal+of+Advances+in+Modeling+Earth+Systems%22">Journal of Advances in Modeling Earth Systems</searchLink>. Apr2026, Vol. 18 Issue 4, p1-22. 22p. – Name: Subject Label: Subject Terms Group: Su Data: *<searchLink fieldCode="DE" term="%22Circulation+models%22">Circulation models</searchLink><br /><searchLink fieldCode="DE" term="%22Artificial+neural+networks%22">Artificial neural networks</searchLink><br /><searchLink fieldCode="DE" term="%22Data+assimilation%22">Data assimilation</searchLink><br /><searchLink fieldCode="DE" term="%22Deep+learning%22">Deep learning</searchLink> – Name: Abstract Label: Abstract Group: Ab Data: We present a novel, data‐driven approach to predict systematic model errors in the ocean component of a coupled general circulation model leveraging deep learning and data assimilation. We examine the skill of the proposed scheme in learning systematic model errors, including their spatial patterns, variance, scales, and test its sensitivity to different predictors and neural network architecture. The scheme utilizes local state variables such as ocean temperature, salinity, velocities, and surface fluxes to predict corrections to temperature tendency for the upper 1,000 m in the ocean on daily timescales. The performance is evaluated on the withheld test data set and compared against the empirical climatological temperature corrections that are geographically dependent. The performance is depth‐dependent, with significant improvements over the benchmark in the upper 20 m in the ocean. It degrades rapidly with depth but remains comparable to the climatology benchmark. Neural networks can capture up to 40−50% $40-50\%$ of the daily variance in temperature increments in the upper 20 m relative to the benchmark's 20% $20\%$. The improvements are associated with networks predicting finer spatiotemporal scales than the benchmark. They are expected to perform better in reducing surface ocean mixed layer bias than previously used techniques. Despite being column‐local without geographical inputs, networks can sufficiently reproduce spatial patterns on daily and longer timescales. The patterns consist of corrections to regional dynamical features such as western boundary currents, equatorial undercurrents, bathymetry‐related corrections in the Southern Ocean, and warm surface increments over subtropical and midlatitude belts. Plain Language Summary: The ocean is a complex system, and we use ocean general circulation models to study it. However, these models are imperfect and have errors in representing the subgrid‐scale processes. We present a new way to correct these errors using deep learning and data assimilation. This method uses information about the ocean thermodynamic state, such as temperature, velocity, and surface fluxes, to predict and correct errors. We found that the new method performs better in the upper 20 m of the ocean and captures a significant fraction of daily corrections to the temperature equation. This new method can help us reduce bias in the upper ocean mixed layer. Key Points: Neural networks can predict spatiotemporal patterns of data assimilation increments using only local state variablesThe skill of neural networks exceeds that of the state‐independent climatological benchmark in the upper oceanThe predictability of upper ocean increments from fluxes and vertical gradients highlights biases in ocean mixed layer representation [ABSTRACT FROM AUTHOR] – Name: AbstractSuppliedCopyright Label: Group: Ab Data: <i>Copyright of Journal of Advances in Modeling Earth Systems is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract.</i> (Copyright applies to all Abstracts.) |
| PLink | https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=8gh&AN=193321851 |
| RecordInfo | BibRecord: BibEntity: Identifiers: – Type: doi Value: 10.1029/2025MS005155 Languages: – Code: eng Text: English PhysicalDescription: Pagination: PageCount: 22 StartPage: 1 Subjects: – SubjectFull: Circulation models Type: general – SubjectFull: Artificial neural networks Type: general – SubjectFull: Data assimilation Type: general – SubjectFull: Deep learning Type: general Titles: – TitleFull: Deep Learning of Systematic Ocean Model Errors in a Coupled GCM From Data Assimilation Increments. Type: main BibRelationships: HasContributorRelationships: – PersonEntity: Name: NameFull: Verma, Tarun – PersonEntity: Name: NameFull: Lu, F. – PersonEntity: Name: NameFull: Adcroft, A. – PersonEntity: Name: NameFull: Zanna, L. – PersonEntity: Name: NameFull: Gnanadesikan, A. IsPartOfRelationships: – BibEntity: Dates: – D: 01 M: 04 Text: Apr2026 Type: published Y: 2026 Identifiers: – Type: issn-print Value: 19422466 Numbering: – Type: volume Value: 18 – Type: issue Value: 4 Titles: – TitleFull: Journal of Advances in Modeling Earth Systems Type: main |
| ResultId | 1 |
