Towards a Unified Data-Driven Boundary Layer Momentum Flux Parameterization for Ocean and Atmosphere
2025
Read Towards a Unified Data-Driven Boundary Layer Momentum Flux Parameterization for Ocean and Atmosphere (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-01766.txt.
Introduction
Renaud Falga, Sara Shamekh, and Laure Zanna train one neural network to describe turbulent momentum mixing in both the ocean and the atmosphere. Boundary layer turbulence sets how winds and currents evolve. Climate models need formulas, parameterizations, for the vertical fluxes it produces. Traditional formulas miss effects like upgradient fluxes entirely.
The network learns offline from coarse-grained large-eddy simulations spanning many turbulent regimes in both fluids. Normalizing flux profiles by their surface values exposes a self-similar shape shared across regimes, enabling joint training. Given a mean wind or current profile, the network predicts the flux profile. Plugged into a single-column atmospheric model, it beat the model's built-in parameterization. It cut errors by a factor of two to three. It stayed robust even when the surface flux was biased by up to 30 percent.
Catalog record
| YEAR | 2025 |
|---|---|
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 57 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-01766
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Towards a Unified Data-Driven Boundary Layer Momentum Flux Parameterization for Ocean and Atmosphere Abstract Boundary layer turbulence, particularly the vertical fluxes of momentum, shapes the evolution of winds and currents and plays a critical role in weather, climate, and biogeochemical processes. In this work, a unified, data-driven parameterization of turbulent momentum fluxes is introduced for both the oceanic and atmospheric convective boundary layers. An artificial neural network (ANN) is trained offline on coarse-grained large-eddy simulation (LES) data representing a wide range of turbulent regimes in both fluids. By normalizing momentum flux profiles with their surface values, we exploit a self-similar structure across regimes and fluids, enabling joint training. The ANN learns to predict vertical profiles of subgrid momentum fluxes from mean wind or current profiles, capturing key physical features such as upgradient fluxes that are inaccessible to traditional first-order closure schemes. When implemented online in the Single Column Atmospheric Model (SCAM), the ANN parameterization consistently outperforms the SCAM baseline parameterization in replicating the evolution of the boundary layer wind profiles from the LES, especially under convective conditions, with errors reduced by a factor of 2–3 across regimes. ANN performance remains robust even when the surface momentum flux is biased by ±30%, and generalization is confirmed by testing on LES cases excluded from the training dataset. This work demonstrates the potential of machine learning to create unified and physically consistent parameterizations across boundary layer systems in climate models. † † journal: Journal of Advances in Modeling Earth Systems (JAMES) † † authors: Renaud Falga, Sara Shamekh, Laure Zanna † † corresponding: Renaud Falga, [email protected] Courant Institute of Mathematical Sciences, New York University, New York, NY, USA keypoints A data-driven parameterization for both atmospheric and oceanic boundary layer momentum flux is developed and implemented in SCAM. The neural network is able to predict upgradient fluxes, highlighting the potential to outperform traditional eddy diffusivity schemes. The ANN parameterization outperforms the CLUBB scheme in replicating the LES wind profiles, especially in convective regimes. Plain Language Summary In both oceanic and atmospheric boundary layers, the transfer of momentum that is induced by turbulence has a significant impact on winds, currents, and many large-scale climatic phenomena. It is therefore essential for coarse-resolution models to accurately parameterize the subgrid effects of turbulence on circulation patterns, yet current methods still struggle to represent the fluxes of momentum. In this study, we introduce a novel machine learning approach that utilizes a neural network trained on high-resolution simulations to predict the subgrid fluxes of momentum in both the atmosphere and the ocean. By uncovering a common structure in turbulent behavior, the same model can be applied to both systems. When tested in a single-column climate model framework, the neural network outperforms traditional methods, especially during convective conditions. This unified approach offers a promising path to improve weather and climate predictions. 1 Introduction The oceanic and atmospheric boundary layers exhibit important physical similarities. In both, the transport of heat, momentum, and tracers is dominated by turbulence, which can be generated either by shear or by surface radiative forcing that generates buoyancy. Boundary layers are typically classified as stable, neutral, or convective, depending on their stratification. Convective boundary layers (CBLs), specifically, share particular similarities across fluids and can form well-mixed layers with nearly uniform temperature, winds, or currents. Although the nature of the surface forcing differs — with convection driven by daytime heating in the atmosphere and by nighttime cooling or precipitation in the ocean — the resulting turbulent dynamics and vertical structure are comparable. In both fluids, vertical turbulent fluxes such as momentum and heat peak at the surface and decrease throughout the boundary layer (BL), reflecting analogous energy transfer processes. Because of poor representation of boundary layer processes, numerical weather forecasts and climate models struggle to accurately simulate near-surface atmospheric variables like temperature and wind Edwards et al. (2011) ; Beljaars (2011) , or oceanic mixed layer depth and sea-surface temperature Belcher et al. (2012) . In the atmosphere, different biases in temperature and moisture can arise from the choice of BL parameterization, both in stable and convective conditions X-M. Hu et al. (2010) . The biases in wind, temperature and clouds also induce important biases in near-surface oceanic variables including sea surface temperature (SST) in coupled climate runs I. Richter (2015) . Specifically, these positive SST biases are partly caused by unrealistically low stratocumulus cloud cover Z. Hu et al. (2008) and an underestimated near-surface wind speed that reduces upwelling Vannière et al. (2014) , which highlights the complex interactions between the atmospheric and oceanic boundary layers. The biases mentioned above could be reduced by improving the representation of boundary layer turbulence in coarse resolution models. Among the turbulent fluxes, the fluxes of momentum play a central role in shaping the vertical structure of wind and current profiles. In global climate models (GCMs), BL turbulence needs to be parameterized, and accurately representing momentum fluxes is critical for reliable weather and climate projections. The misrepresentation of atmospheric momentum fluxes has been shown to lead to biases in surface wind patterns and resulting Hadley circulation in GCM simulations J.H. Richter & Rasch (2008) , which can in turn affect the cloud cover. Boundary layer clouds are commonly poorly simulated in GCMs Teixeira et al. (2011) , and the impact of clouds on future climate is one of the largest sources of uncertainty within climate model projections Caldwell et al. (2016) . Accurate parameterization of atmospheric boundary layer (ABL) momentum flux has been shown to improve the representation of clouds in GCMs Suselj et al. (2021) . Momentum fluxes are key critical processes in the oceanic boundary layer (OBL) as well. It is estimated that 90% of the excess heat induced by anthropogenic greenhouse gases is absorbed by the ocean Resplandy et al. (2019) , and the strength of convectively-driven mixing partly determines the amount of heat and carbon uptake. Improving representation of oceanic boundary layer momentum fluxes actually leads to an improvement in the simulation of global ocean heat uptake Shimura et al. (2020) . Given their impact on surface exchanges as well as large-scale atmospheric and oceanic processes, it is essential to develop accurate parameterizations of momentum fluxes for GCMs. Among the most common schemes that have been adopted in ABL and OBL parameterizations are the first-order eddy-diffusivity schemes. In such models, the momentum flux is expressed as the opposite of the vertical gradient of mean wind (or current) multiplied by an eddy-diffusivity coefficient; the flux is thus said to be downgradient. However, these schemes usually fail to represent convective boundary layers, because of the well-mixed structure of the wind and current profiles. Brown et al. (2006) . Additionally, in reality, upgradient fluxes can occur because of non-local transport or mesoscale dynamics, and therefore cannot be represented by these approaches Dixit et al. (2021) ; Helfer et al. (2021) . Some schemes added a mass-flux term to represent the effect of non-local mixing on turbulent transport. However, it has been shown that these so-called mass-flux schemes fail to represent the vertical momentum transport for both shallow and deep convection Zhu (2015) . The representation of wind and boundary layer clouds can be improved by higher order schemes that add prognostic