Use of Neural Networks for Stable, Accurate and Physically Consistent Parameterization of Subgrid Atmospheric Processes With Good Performance at Reduced Precision
Open accessThe DOI matches its Crossref record; 11 of the 11 references it cites were found in a registry or an index.
Abstract
A promising approach to improve climate‐model simulations is to replace traditional subgrid parameterizations based on simplified physical models by machine learning algorithms that are data‐driven. However, neural networks (NNs) often lead to instabilities and climate drift when coupled to an atmospheric model. Here, we learn an NN parameterization from a high‐resolution atmospheric simulation in an idealized domain by accurately calculating subgrid terms through coarse graining. The NN parameterization has a structure that ensures physical constraints are respected, such as by predicting subgrid fluxes instead of tendencies. The NN parameterization leads to stable simulations that replicate the climate of the high‐resolution simulation with similar accuracy to a successful random‐forest parameterization while needing far less memory. We find that the simulations are stable for different horizontal resolutions and a variety of NN architectures, and that an NN with substantially reduced numerical precision could decrease computational costs without affecting the quality of simulations.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Breiman, L. (2001) · Machine Learning · doi:10.1023/a:1010933404324 · Cited by 133,538
Micikevicius, P., Narang, S., Alben, J., Diamos, G., Elsen, E., Garcia, D., Ginsburg, B., Houston, M., Kuchaiev, O., Venkatesh, G., et al. (2017) · arXiv · doi:10.48550/arxiv.1710.03740 · Cited by 1,179 · cited 1×
“Furthermore, previous studies have found that using deep NNs with half-precision multipliers has little effect on accuracy when used for classification (Courbariaux et al., 2014; Miyashita et al., 2016; Micikevicius et al., 2017; Das et al., 2018).”
BackgroundDeep learning to represent subgrid processes in climate models
Rasp, S., Pritchard, M. S., & Gentine, P. (2018) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1810286115 · Cited by 1,044 · cited 5×
“The two main differences in the data processing between YOG20 and the NNs studies mentioned above are that (1) the subgrid corrections were calculated accurately for the instantaneous atmospheric state (YOG20) rather than approximating them based on differences over 3-h periods (Brenowitz & Bretherton, 2018, 2019) or…”
MethodInfluentialCould Machine Learning Break the Convection Parameterization Deadlock?
Gentine, P., Pritchard, M., Rasp, S., Reinaudi, G., & Yacalis, G. (2018) · Geophysical Research Letters · doi:10.1029/2018gl078202 · Cited by 497 · cited 2×
“Previous studies that used NN parameterizations usually predicted the sum of tendencies due to several different processes as a single output (Gentine et al., 2018; Rasp et al., 2018; Brenowitz & Bretherton, 2018, 2019).”
MethodO'Gorman, P. A., & Dwyer, J. G. (2018) · Journal of Advances in Modeling Earth Systems · doi:10.1029/2018ms001351 · Cited by 419 · cited 4×
“…are averages over samples in the training data set, and thus they can automatically satisfy physical constraints such as linear energy conservation (O'Gorman & Dwyer, 2018, YOG20) and make conservative predictions for samples outside of the training data (NNs can also be forced to satisfy analytic…”
BackgroundMethodInfluentialSelf-citationSensitivity of tropical precipitation extremes to climate change
O’Gorman, P. A. (2012) · Nature Geoscience · doi:10.1038/ngeo1568 · Cited by 349
Self-citationClimate goals and computing the future of clouds
Schneider, T., Teixeira, J., Bretherton, C. S., Brient, F., Pressel, K. G., Schär, C., & Siebesma, A. P. (2017) · Nature Climate Change · doi:10.1038/nclimate3190 · Cited by 294
Yuval, J., & O’Gorman, P. A. (2020) · Nature Communications · doi:10.1038/s41467-020-17142-3 · Cited by 283
Self-citationData‐Driven Equation Discovery of Ocean Mesoscale Closures
Zanna, L., & Bolton, T. (2020) · Geophysical Research Letters · doi:10.1029/2020gl088376 · Cited by 259 · cited 11×
“Traditional parameterizations in general circulation models (GCMs) rely on simplified physical models and suffer from inaccuracies which lead to model biases and large uncertainties in climate projections (Bony & Dufresne, 2005; Oueslati & Bellon, 2015; O'Gorman, 2012; Schneider et al., 2017; Wilcox & Donner, 2007).”
MethodInfluentialSpatially Extended Tests of a Neural Network Parametrization Trained by Coarse‐Graining
Brenowitz, N. D., & Bretherton, C. S. (2019) · Journal of Advances in Modeling Earth Systems · doi:10.1029/2019ms001711 · Cited by 231 · cited 11×
“(2018), we find that simulations with the NN parameterization are stable for a variety of configurations, and in contrast to Brenowitz and Bretherton (2018, 2019), they do not exhibit climate drift.”
BackgroundMethodInfluential
Cited by
Physics-informed machine learning: case studies for weather and climate modelling
Kashinath, K., Mustafa, M., Albert, A., Wu, J.-L., Jiang, C., Esmaeilzadeh, S., Azizzadenesheli, K., Wang, R., Chattopadhyay, A., Singh, A., et al. (2021) · Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences · doi:10.1098/rsta.2020.0093 · Cited by 662
Machine Learning in Weather Prediction and Climate Analyses—Applications and Perspectives
Bochenek, B., & Ustrnul, Z. (2022) · Atmosphere · doi:10.3390/atmos13020180 · Cited by 325
The future of coastal monitoring through satellite remote sensing
Vitousek, S., Buscombe, D., Vos, K., Barnard, P. L., Ritchie, A. C., & Warrick, J. A. (2023) · Cambridge Prisms: Coastal Futures · doi:10.1017/cft.2022.4 · Cited by 190
Machine learning for numerical weather and climate modelling: a review
de Burgh-Day, C. O., & Leeuwenburg, T. (2023) · Geoscientific Model Development · doi:10.5194/gmd-16-6433-2023 · Cited by 176
Dueben, P. D., Schultz, M. G., Chantry, M., Gagne, D. J., Hall, D. M., & McGovern, A. (2022) · Artificial Intelligence for the Earth Systems · doi:10.1175/aies-d-21-0002.1 · Cited by 64
Opinion: Optimizing climate models with process knowledge, resolution, and artificial intelligence
Schneider, T., Leung, L. R., & Wills, R. C. J. (2024) · Atmospheric Chemistry and Physics · doi:10.5194/acp-24-7041-2024 · Cited by 53
Differentiable programming for Earth system modeling
Gelbrecht, M., White, A., Bathiany, S., & Boers, N. (2023) · Geoscientific Model Development · doi:10.5194/gmd-16-3123-2023 · Cited by 31
“For example, Yuval et al. (2021) ensure conservation of energy and water in an ANN-based subgrid parameterization by predicting fluxes rather than tendencies of the prognostic model variables.”
BackgroundHan, Y., Zhang, G. J., & Wang, Y. (2023) · Journal of Advances in Modeling Earth Systems · doi:10.1029/2022ms003508 · Cited by 24
Metadata
- Authors
- Yuval, J., O'Gorman, P. A., & Hill, C. N.
- Venue
- Geophysical Research Letters
- Publisher
- American Geophysical Union (AGU)
- Published
- 2021-03-15
- Volume
- 48
- Issue
- 6
- ISSN
- 0094-8276 (ISSN-L)
- 1944-8007
- OpenAlex
- W3094611249
- Cited by
- 98
- References
- 11
- FWCI
- 5.75
- Type
- journal article
- Language
- en
- Open access
- hybrid
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.