Non‐Local Parameterization of Atmospheric Subgrid Processes With Neural Networks

Wang, Peidong; Yuval, Janni; O’Gorman, Paul A.

doi:10.1029/2022ms002984

Cited by 19 publications

(18 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In a related study, Wang et al. (2022) show that including non‐local inputs improves offline performance for the NN momentum parameterization developed here, and future work should investigate its online performance. Overall, our results show that using high‐resolution simulations to evaluate subgrid fluxes provides useful information for the design of parameterizations, and that NN parameterization for momentum is a promising alternative to existing parameterizations.…”

Section: Discussionmentioning

confidence: 77%

“…The staggering of momentum variables on the model grid poses challenging for learning a momentum parameterization and future work could investigate how best to deal with this issue which may improve online performance at all resolutions. In a related study, Wang et al (2022) show that including non-local inputs improves offline performance for the NN momentum parameterization developed here, and future work should investigate its online performance. Overall, our results show that using high-resolution simulations to evaluate subgrid fluxes provides useful information for the design of parameterizations, and that NN parameterization for momentum is a promising alternative to existing parameterizations.…”

Section: Discussionmentioning

confidence: 83%

See 1 more Smart Citation

Neural‐Network Parameterization of Subgrid Momentum Transport in the Atmosphere

Yuval

O’Gorman

2023

J Adv Model Earth Syst

Self Cite

View full text Add to dashboard Cite

Attempts to use machine learning to develop atmospheric parameterizations have mainly focused on subgrid effects on temperature and moisture, but subgrid momentum transport is also important in simulations of the atmospheric circulation. Here, we use neural networks to develop a subgrid momentum transport parameterization that learns from coarse‐grained output of a high‐resolution atmospheric simulation in an idealized aquaplanet domain. We show that substantial subgrid momentum transport occurs due to convection. The neural‐network parameterization has skill in predicting momentum fluxes associated with convection, although its skill for subgrid momentum fluxes is lower compared to subgrid energy and moisture fluxes. The parameterization conserves momentum, and when implemented in the same atmospheric model at coarse resolution it leads to stable simulations and tends to reduce wind biases, although it over‐corrects for one configuration tested. Overall, our results show that it is challenging to predict subgrid momentum fluxes and that machine‐learning momentum parameterization gives promising results.

show abstract

Section: Discussionmentioning

confidence: 77%

Section: Discussionmentioning

confidence: 83%

Neural‐Network Parameterization of Subgrid Momentum Transport in the Atmosphere

Yuval

O’Gorman

2023

J Adv Model Earth Syst

Self Cite

View full text Add to dashboard Cite

show abstract

“…Therefore, further tests, both offline and online (coupled), are needed to see if 3D GWP schemes improve the circulation variability in GCMs. That said, there is existing evidence for SGS modeling of other physical processes that would benefit from including neighboring columns, providing further incentive for considering horizontally non‐local parameterizations (e.g., Fatkullin & Vanden‐Eijnden, 2004; Guan et al., 2022; Wang et al., 2022). Adding to the complexity, we have found that the GWD due to SGS lateral momentum fluxes could be sensitive to the methods used to extract them.…”

Section: Summary and Discussionmentioning

confidence: 99%

Quantifying 3D Gravity Wave Drag in a Library of Tropical Convection‐Permitting Simulations for Data‐Driven Parameterizations

Sun

Hassanzadeh

Alexander

et al. 2023

J Adv Model Earth Syst

View full text Add to dashboard Cite

Atmospheric gravity waves (GWs) span a broad range of length scales. As a result, the un‐resolved and under‐resolved GWs have to be represented using a sub‐grid scale (SGS) parameterization in general circulation models (GCMs). In recent years, machine learning (ML) techniques have emerged as novel methods for SGS modeling of climate processes. In the widely used approach of supervised (offline) learning, the true representation of the SGS terms have to be properly extracted from high‐fidelity data (e.g., GW‐resolving simulations). However, this is a non‐trivial task, and the quality of the ML‐based parameterization significantly hinges on the quality of these SGS terms. Here, we compare three methods to extract 3D GW fluxes and the resulting drag (Gravity Wave Drag [GWD]) from high‐resolution simulations: Helmholtz decomposition, and spatial filtering to compute the Reynolds stress and the full SGS stress. In addition to previous studies that focused only on vertical fluxes by GWs, we also quantify the SGS GWD due to lateral momentum fluxes. We build and utilize a library of tropical high‐resolution (Δx = 3 km) simulations using weather research and forecasting model. Results show that the SGS lateral momentum fluxes could have a significant contribution to the total GWD. Moreover, when estimating GWD due to lateral effects, interactions between the SGS and the resolved large‐scale flow need to be considered. The sensitivity of the results to different filter type and length scale (dependent on GCM resolution) is also explored to inform the scale‐awareness in the development of data‐driven parameterizations.

show abstract

“…The GCM’s column structure reflects the importance of separating the vertical in the model topology, and the importance of convection and its timescales in atmosphere and ocean. The structural independence of columns has been seen as a limitation, but new methods can use nonlocal predictors ( 102 ). The stochastic parameterizations mentioned in Section 3 also impose nonlocal (in space and time) coherence to the stochasticity ( 103 ).…”

Section: What Might a Future Modeling Landscape Look Like?mentioning

confidence: 99%

Are general circulation models obsolete?

Balaji

Couvreux

Deshayes

et al. 2022

Proc. Natl. Acad. Sci. U.S.A.

View full text Add to dashboard Cite

Traditional general circulation models, or GCMs—that is, three-dimensional dynamical models with unresolved terms represented in equations with tunable parameters—have been a mainstay of climate research for several decades, and some of the pioneering studies have recently been recognized by a Nobel prize in Physics. Yet, there is considerable debate around their continuing role in the future. Frequently mentioned as limitations of GCMs are the structural error and uncertainty across models with different representations of unresolved scales and the fact that the models are tuned to reproduce certain aspects of the observed Earth. We consider these shortcomings in the context of a future generation of models that may address these issues through substantially higher resolution and detail, or through the use of machine learning techniques to match them better to observations, theory, and process models. It is our contention that calibration, far from being a weakness of models, is an essential element in the simulation of complex systems, and contributes to our understanding of their inner workings. Models can be calibrated to reveal both fine-scale detail and the global response to external perturbations. New methods enable us to articulate and improve the connections between the different levels of abstract representation of climate processes, and our understanding resides in an entire hierarchy of models where GCMs will continue to play a central role for the foreseeable future.

show abstract

Non‐Local Parameterization of Atmospheric Subgrid Processes With Neural Networks

Cited by 19 publications

References 38 publications

Neural‐Network Parameterization of Subgrid Momentum Transport in the Atmosphere

Neural‐Network Parameterization of Subgrid Momentum Transport in the Atmosphere

Quantifying 3D Gravity Wave Drag in a Library of Tropical Convection‐Permitting Simulations for Data‐Driven Parameterizations

Are general circulation models obsolete?

Contact Info

Product

Resources

About