Scale-Aware Parameterization of Cloud Fraction and Condensate for a Global Atmospheric Model Machine-Learned From Coarse-Grained Kilometer-Scale Simulations

Cyril Morcrette; Tobias Cave; Helena Reid; Joana da Silva Rodrigues; Teo Deveney; Lisa Kreusser; Kwinten Van Weverberg; Chris Budd

doi:10.1029/2024MS004651

Scale-Aware Parameterization of Cloud Fraction and Condensate for a Global Atmospheric Model Machine-Learned From Coarse-Grained Kilometer-Scale Simulations

Cyril Morcrette, Tobias Cave, Helena Reid, Joana da Silva Rodrigues, Teo Deveney, Lisa Kreusser, Kwinten Van Weverberg, Chris Budd

Research output: Contribution to journal › Article › peer-review

Abstract

Kilometer grid-length simulations over a variety of different locations worldwide are used as training data for a deep-learning model designed to predict clouds in a global climate model. The inputs to the neural network are profiles of temperature, humidity and pressure from the high-resolution model, averaged to the scale of the climate model. The outputs are profiles of cloud fraction and in-cloud liquid and ice water contents. The high-resolution data is coarse-grained to a range of sizes, allowing the model to learn how the cloud formation depends on the size of the area being considered. The machine-learned cloud fraction and cloud condensate scheme is coupled to a global climate model and used to run multi-year simulations where the clouds predicted by the neural-network are fully interacting with the rest of the model.

Original language	English
Article number	e2024MS004651
Journal	Journal of Advances in Modeling Earth Systems (JAMES)
Volume	17
Issue number	4
Early online date	18 Apr 2025
DOIs	https://doi.org/10.1029/2024MS004651
Publication status	Published - 30 Apr 2025

Data Availability Statement

Code for defining the nesting suite and the diagnostic output, for coarse‐graining the high‐resolution model data,for rebalancing it and training the neural network is available from Morcrette (2024a). A portion of the trainingdata is also provided there. Code for converting the Python‐trained neural network to Fortran is available fromMorcrette (2024b).

Acknowledgements

Thanks are due to Omar Jamil for discussions on a precursor to this project. Thanks are due to Keith Williams for his encouragement and support and to Paul Field for his comments on an earlier version of this paper. We wish to thank Pavel Perezhogin and two anonymous reviewers for their comments which helped improve the manuscript.

Funding

CM, HR and JdSR acknowledge support from the AI4PEX programme. This work began as a Masters project carried out by TC, as part of a Mathematics with Data Science for Industry MSc at the University of Bath. TD, LK and CB acknowledge support from the EPSRC programme grant in “The Mathematics of Deep Learning,” under the project EP/L015684/1.

Funders	Funder number
Engineering and Physical Sciences Research Council	EP/L015684/1.

Keywords

climate
cloud
hybrid model
machine learning
parameterization

ASJC Scopus subject areas

Global and Planetary Change
Environmental Chemistry
General Earth and Planetary Sciences

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1029/2024MS004651Licence: CC BY

Cite this

Morcrette, C., Cave, T., Reid, H., da Silva Rodrigues, J., Deveney, T., Kreusser, L., Van Weverberg, K., & Budd, C. (2025). Scale-Aware Parameterization of Cloud Fraction and Condensate for a Global Atmospheric Model Machine-Learned From Coarse-Grained Kilometer-Scale Simulations. Journal of Advances in Modeling Earth Systems (JAMES), 17(4), Article e2024MS004651. https://doi.org/10.1029/2024MS004651

Scale-Aware Parameterization of Cloud Fraction and Condensate for a Global Atmospheric Model Machine-Learned From Coarse-Grained Kilometer-Scale Simulations. / Morcrette, Cyril; Cave, Tobias; Reid, Helena et al.
In: Journal of Advances in Modeling Earth Systems (JAMES), Vol. 17, No. 4, e2024MS004651, 30.04.2025.

Research output: Contribution to journal › Article › peer-review

Morcrette, C, Cave, T, Reid, H, da Silva Rodrigues, J, Deveney, T , Kreusser, L, Van Weverberg, K & Budd, C 2025, 'Scale-Aware Parameterization of Cloud Fraction and Condensate for a Global Atmospheric Model Machine-Learned From Coarse-Grained Kilometer-Scale Simulations', Journal of Advances in Modeling Earth Systems (JAMES), vol. 17, no. 4, e2024MS004651. https://doi.org/10.1029/2024MS004651

@article{f5ff3e4f5a8544b1b387c47ce757905e,

title = "Scale-Aware Parameterization of Cloud Fraction and Condensate for a Global Atmospheric Model Machine-Learned From Coarse-Grained Kilometer-Scale Simulations",

abstract = "Kilometer grid-length simulations over a variety of different locations worldwide are used as training data for a deep-learning model designed to predict clouds in a global climate model. The inputs to the neural network are profiles of temperature, humidity and pressure from the high-resolution model, averaged to the scale of the climate model. The outputs are profiles of cloud fraction and in-cloud liquid and ice water contents. The high-resolution data is coarse-grained to a range of sizes, allowing the model to learn how the cloud formation depends on the size of the area being considered. The machine-learned cloud fraction and cloud condensate scheme is coupled to a global climate model and used to run multi-year simulations where the clouds predicted by the neural-network are fully interacting with the rest of the model.",

keywords = "climate, cloud, hybrid model, machine learning, parameterization",

author = "Cyril Morcrette and Tobias Cave and Helena Reid and {da Silva Rodrigues}, Joana and Teo Deveney and Lisa Kreusser and {Van Weverberg}, Kwinten and Chris Budd",

year = "2025",

month = apr,

day = "30",

doi = "10.1029/2024MS004651",

language = "English",

volume = "17",

journal = "Journal of Advances in Modeling Earth Systems (JAMES)",

issn = "1942-2466",

publisher = "John Wiley and Sons Inc.",

number = "4",

}

TY - JOUR

T1 - Scale-Aware Parameterization of Cloud Fraction and Condensate for a Global Atmospheric Model Machine-Learned From Coarse-Grained Kilometer-Scale Simulations

AU - Morcrette, Cyril

AU - Cave, Tobias

AU - Reid, Helena

AU - da Silva Rodrigues, Joana

AU - Deveney, Teo

AU - Kreusser, Lisa

AU - Van Weverberg, Kwinten

AU - Budd, Chris

PY - 2025/4/30

Y1 - 2025/4/30

N2 - Kilometer grid-length simulations over a variety of different locations worldwide are used as training data for a deep-learning model designed to predict clouds in a global climate model. The inputs to the neural network are profiles of temperature, humidity and pressure from the high-resolution model, averaged to the scale of the climate model. The outputs are profiles of cloud fraction and in-cloud liquid and ice water contents. The high-resolution data is coarse-grained to a range of sizes, allowing the model to learn how the cloud formation depends on the size of the area being considered. The machine-learned cloud fraction and cloud condensate scheme is coupled to a global climate model and used to run multi-year simulations where the clouds predicted by the neural-network are fully interacting with the rest of the model.

AB - Kilometer grid-length simulations over a variety of different locations worldwide are used as training data for a deep-learning model designed to predict clouds in a global climate model. The inputs to the neural network are profiles of temperature, humidity and pressure from the high-resolution model, averaged to the scale of the climate model. The outputs are profiles of cloud fraction and in-cloud liquid and ice water contents. The high-resolution data is coarse-grained to a range of sizes, allowing the model to learn how the cloud formation depends on the size of the area being considered. The machine-learned cloud fraction and cloud condensate scheme is coupled to a global climate model and used to run multi-year simulations where the clouds predicted by the neural-network are fully interacting with the rest of the model.

KW - climate

KW - cloud

KW - hybrid model

KW - machine learning

KW - parameterization

UR - http://www.scopus.com/inward/record.url?scp=105005152438&partnerID=8YFLogxK

U2 - 10.1029/2024MS004651

DO - 10.1029/2024MS004651

M3 - Article

AN - SCOPUS:105005152438

SN - 1942-2466

VL - 17

JO - Journal of Advances in Modeling Earth Systems (JAMES)

JF - Journal of Advances in Modeling Earth Systems (JAMES)

IS - 4

M1 - e2024MS004651

ER -

Scale-Aware Parameterization of Cloud Fraction and Condensate for a Global Atmospheric Model Machine-Learned From Coarse-Grained Kilometer-Scale Simulations

Abstract

Data Availability Statement

Acknowledgements

Funding

Keywords

ASJC Scopus subject areas

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this