Stable Machine‐Learning Parameterization of Subgrid Processes in a Comprehensive Atmospheric Model Learned From Embedded Convection‐Permitting Simulations
Abstract Modern climate projections often suffer from inadequate spatial and temporal resolution due to computational limitations, resulting in inaccurate representations of sub‐grid processes. A promising technique to address this is the multiscale modeling framework (MMF), which embeds a kilometer...
Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
American Geophysical Union (AGU)
2025-07-01
|
| Series: | Journal of Advances in Modeling Earth Systems |
| Subjects: | |
| Online Access: | https://doi.org/10.1029/2024MS004618 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1839611615378407424 |
|---|---|
| author | Zeyuan Hu Akshay Subramaniam Zhiming Kuang Jerry Lin Sungduk Yu Walter M. Hannah Noah D. Brenowitz Josh Romero Michael S. Pritchard |
| author_facet | Zeyuan Hu Akshay Subramaniam Zhiming Kuang Jerry Lin Sungduk Yu Walter M. Hannah Noah D. Brenowitz Josh Romero Michael S. Pritchard |
| author_sort | Zeyuan Hu |
| collection | DOAJ |
| description | Abstract Modern climate projections often suffer from inadequate spatial and temporal resolution due to computational limitations, resulting in inaccurate representations of sub‐grid processes. A promising technique to address this is the multiscale modeling framework (MMF), which embeds a kilometer‐resolution cloud‐resolving model (CRM) within each atmospheric column of a host climate model to replace traditional convection and cloud parameterizations. Machine learning offers a unique opportunity to make MMF more accessible by emulating the embedded CRM and reducing its substantial computational cost. Although many studies have demonstrated proof‐of‐concept success of achieving stable hybrid simulations, it remains a challenge to achieve near operational‐level success with real geography and comprehensive variable emulation that includes, for example, explicit cloud condensate coupling. In this study, we present a stable hybrid model capable of integrating for at least 5 years with near operational‐level complexity, including coarse‐grid geography, seasonality, explicit cloud condensate and wind predictions, and land coupling. Our model demonstrates skillful online performance, achieving a 5‐year zonal mean tropospheric temperature bias within 2 K, water vapor bias within 1 g/kg, and a precipitation root mean square error of 0.96 mm/day. Key factors contributing to our online performance include an expressive U‐Net architecture and physical thermodynamic constraints for microphysics. With microphysical constraints mitigating unrealistic cloud formation, our work is the first to demonstrate realistic multi‐year cloud condensate climatology under the MMF framework. Despite these advances, online diagnostics reveal persistent biases in certain regions, highlighting the need for innovative strategies to further optimize online performance. |
| format | Article |
| id | doaj-art-09345e1b93a943c2bcad4143b55b4ce0 |
| institution | Matheson Library |
| issn | 1942-2466 |
| language | English |
| publishDate | 2025-07-01 |
| publisher | American Geophysical Union (AGU) |
| record_format | Article |
| series | Journal of Advances in Modeling Earth Systems |
| spelling | doaj-art-09345e1b93a943c2bcad4143b55b4ce02025-07-28T14:46:59ZengAmerican Geophysical Union (AGU)Journal of Advances in Modeling Earth Systems1942-24662025-07-01177n/an/a10.1029/2024MS004618Stable Machine‐Learning Parameterization of Subgrid Processes in a Comprehensive Atmospheric Model Learned From Embedded Convection‐Permitting SimulationsZeyuan Hu0Akshay Subramaniam1Zhiming Kuang2Jerry Lin3Sungduk Yu4Walter M. Hannah5Noah D. Brenowitz6Josh Romero7Michael S. Pritchard8NVIDIA Research Santa Clara CA USANVIDIA Research Santa Clara CA USAHarvard University Cambridge MA USAUniversity of California at Irvine Irvine CA USAMultimodal Cognitive AI Intel Labs Santa Clara CA USALawrence Livermore National Laboratory Livermore CA USANVIDIA Research Santa Clara CA USANVIDIA Research Santa Clara CA USANVIDIA Research Santa Clara CA USAAbstract Modern climate projections often suffer from inadequate spatial and temporal resolution due to computational limitations, resulting in inaccurate representations of sub‐grid processes. A promising technique to address this is the multiscale modeling framework (MMF), which embeds a kilometer‐resolution cloud‐resolving model (CRM) within each atmospheric column of a host climate model to replace traditional convection and cloud parameterizations. Machine learning offers a unique opportunity to make MMF more accessible by emulating the embedded CRM and reducing its substantial computational cost. Although many studies have demonstrated proof‐of‐concept success of achieving stable hybrid simulations, it remains a challenge to achieve near operational‐level success with real geography and comprehensive variable emulation that includes, for example, explicit cloud condensate coupling. In this study, we present a stable hybrid model capable of integrating for at least 5 years with near operational‐level complexity, including coarse‐grid geography, seasonality, explicit cloud condensate and wind predictions, and land coupling. Our model demonstrates skillful online performance, achieving a 5‐year zonal mean tropospheric temperature bias within 2 K, water vapor bias within 1 g/kg, and a precipitation root mean square error of 0.96 mm/day. Key factors contributing to our online performance include an expressive U‐Net architecture and physical thermodynamic constraints for microphysics. With microphysical constraints mitigating unrealistic cloud formation, our work is the first to demonstrate realistic multi‐year cloud condensate climatology under the MMF framework. Despite these advances, online diagnostics reveal persistent biases in certain regions, highlighting the need for innovative strategies to further optimize online performance.https://doi.org/10.1029/2024MS004618parameterizationmachine learningconvectionclimatehybrid simulation |
| spellingShingle | Zeyuan Hu Akshay Subramaniam Zhiming Kuang Jerry Lin Sungduk Yu Walter M. Hannah Noah D. Brenowitz Josh Romero Michael S. Pritchard Stable Machine‐Learning Parameterization of Subgrid Processes in a Comprehensive Atmospheric Model Learned From Embedded Convection‐Permitting Simulations Journal of Advances in Modeling Earth Systems parameterization machine learning convection climate hybrid simulation |
| title | Stable Machine‐Learning Parameterization of Subgrid Processes in a Comprehensive Atmospheric Model Learned From Embedded Convection‐Permitting Simulations |
| title_full | Stable Machine‐Learning Parameterization of Subgrid Processes in a Comprehensive Atmospheric Model Learned From Embedded Convection‐Permitting Simulations |
| title_fullStr | Stable Machine‐Learning Parameterization of Subgrid Processes in a Comprehensive Atmospheric Model Learned From Embedded Convection‐Permitting Simulations |
| title_full_unstemmed | Stable Machine‐Learning Parameterization of Subgrid Processes in a Comprehensive Atmospheric Model Learned From Embedded Convection‐Permitting Simulations |
| title_short | Stable Machine‐Learning Parameterization of Subgrid Processes in a Comprehensive Atmospheric Model Learned From Embedded Convection‐Permitting Simulations |
| title_sort | stable machine learning parameterization of subgrid processes in a comprehensive atmospheric model learned from embedded convection permitting simulations |
| topic | parameterization machine learning convection climate hybrid simulation |
| url | https://doi.org/10.1029/2024MS004618 |
| work_keys_str_mv | AT zeyuanhu stablemachinelearningparameterizationofsubgridprocessesinacomprehensiveatmosphericmodellearnedfromembeddedconvectionpermittingsimulations AT akshaysubramaniam stablemachinelearningparameterizationofsubgridprocessesinacomprehensiveatmosphericmodellearnedfromembeddedconvectionpermittingsimulations AT zhimingkuang stablemachinelearningparameterizationofsubgridprocessesinacomprehensiveatmosphericmodellearnedfromembeddedconvectionpermittingsimulations AT jerrylin stablemachinelearningparameterizationofsubgridprocessesinacomprehensiveatmosphericmodellearnedfromembeddedconvectionpermittingsimulations AT sungdukyu stablemachinelearningparameterizationofsubgridprocessesinacomprehensiveatmosphericmodellearnedfromembeddedconvectionpermittingsimulations AT waltermhannah stablemachinelearningparameterizationofsubgridprocessesinacomprehensiveatmosphericmodellearnedfromembeddedconvectionpermittingsimulations AT noahdbrenowitz stablemachinelearningparameterizationofsubgridprocessesinacomprehensiveatmosphericmodellearnedfromembeddedconvectionpermittingsimulations AT joshromero stablemachinelearningparameterizationofsubgridprocessesinacomprehensiveatmosphericmodellearnedfromembeddedconvectionpermittingsimulations AT michaelspritchard stablemachinelearningparameterizationofsubgridprocessesinacomprehensiveatmosphericmodellearnedfromembeddedconvectionpermittingsimulations |