Large eddy simulation of ocean mesoscale eddies
P. Perezhogin, A. Balakrishna, R. Agrawal · 2025
Read Large eddy simulation of ocean mesoscale eddies by P. Perezhogin, A. Balakrishna, R. Agrawal (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-2501-05357.txt.
Introduction
Pavel Perezhogin, Arun Balakrishna, and Rahul Agrawal apply large eddy simulation methods to ocean mesoscale eddies, the swirls that carry much of the ocean's lateral transport. Global ocean models at quarter degree resolution cannot resolve these fluxes, so they must be modeled. The team borrows an explicit filtering approach from three dimensional turbulence work.
Their closure was built into a major ocean model and tested in an idealized double gyre basin. At well resolved grids, simulations matched filtered high resolution truth. As the filter scale neared the energy carrying scales, accuracy dropped and runs stopped converging. Mean flows still came out right at every resolution. A companion model for thickness flux improved the coarsest runs.
Catalog record
| CREATOR | P. Perezhogin, A. Balakrishna, R. Agrawal |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 27 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-05357
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Large eddy simulation of ocean mesoscale eddies P. Perezhogin, A. Balakrishna, R. Agrawal Abstract LES of ocean mesoscale eddies Mesoscale eddies produce lateral (2D) fluxes that need to be parameterized in eddy-permitting ( 1/4^{\circ} ) global ocean models due to insufficient horizontal resolution. Here, we systematically apply methods from the 3D LES community to parameterize lateral vorticity fluxes produced by mesoscale eddies leveraging an explicit filtering approach together with a dynamic procedure. The developed subfilter closure is implemented into the GFDL MOM6 ocean model and is evaluated in an idealized double-gyre configuration, both a-priori and a-posteriori . For sufficiently resolved grids, the LES simulations converge to the filtered high-resolution data. However, limitations in the proposed closure are observed when the filter scale approaches the energy-containing scales: the a-priori performance drops and a-posteriori experiments fail to converge to the filtered high-resolution data. Nevertheless, the proposed closure is accurate in predicting the mean flow in a-posteriori simulations at all resolutions considered ( 1/2^{\circ}-1/8^{\circ} ). Finally, we propose parameterizing the thickness fluxes using a Bardina model which further improves simulations at the coarsest resolutions ( 1/2^{\circ}-1/3^{\circ} ). 1 Introduction Traditional parameterizations of mesoscale eddies attempt to close the energy budget by parameterizing dissipation and backscatter of kinetic energy (KE) ( Jansen & Held, 2014 ) and removal of the potential energy ( Gent & McWilliams, 1990 ) . However, these parameterizations are commonly tuned to maximize a-posteriori performance, while their a-priori accuracy and connection to high-resolution data are unknown. In recent years, there has been a substantial effort to develop mesoscale parameterizations with high a-priori accuracy by applying the large eddy simulation (LES) paradigm ( Fox-Kemper & Menemenlis, 2008 ; Khani & Dawson, 2023 ) or machine learning methods ( Zanna & Bolton, 2020 ) . However, as compared to studies of 3D turbulence, the assessment of these methods has been limited thus far. For instance, the dynamic model of Germano et al. (1991) has been evaluated in ocean models only once ( Bachman et al., 2017 ) . Vorticity fluxes (or momentum fluxes) in quasi-2D fluids are responsible for the forward cascade of enstrophy and inverse cascade of energy. Thus, careful consideration of both dissipation and backscatter is required to produce accurate and numerically stable subfilter closures. Perezhogin & Glazunov (2023) showed that the 3D LES closure of Horiuti (1997) can be adapted for the prediction of vorticity fluxes in 2D decaying turbulence. In particular, the Reynolds stress, which is a part of the Germano (1986) decomposition, is responsible for additional KE backscatter. This finding was further confirmed and exploited by Perezhogin et al. (2024) in an idealized GFDL MOM6 ocean model ( Adcroft et al., 2019 ) . However, subfilter closures in that work were subject to substantial a-posteriori tuning, and thus assessing their a-priori accuracy is challenging. In this light, we propose a methodology for parameterization of mesoscale eddies in the MOM6 ocean model that addresses a problem of a-posteriori tuning without degrading a-priori accuracy. We utilize two approaches adopted by the 3D LES community, namely the explicit filtering approach ( Winckelmans et al., 2001 ; Bose et al., 2010 ) and the dynamic procedure of Germano et al. (1991) . Following the explicit filtering approach, we assume that the LES filter width \overline{\Delta} is larger than the LES grid spacing \Delta x . This improves our predictions of the subfilter flux by allowing a partial inversion of the LES filter ( Carati et al., 2001 ) . Additionally, we suppress the numerical discretization errors by increasing the filter-to-grid width ratio ( \mathrm{FGR}=\overline{\Delta}/\Delta x ). Finally, the dynamic procedure allows us to accurately estimate the free parameters of the parameterization from the resolved flow only. Overall, our approach permits a parameterization with good a-priori accuracy and contains a single tunable parameter FGR, which is adjusted to maximize the a-posteriori performance. The rest of this report is organized as follows. Section 2 describes the MOM6 ocean model, the subfilter closure of vorticity fluxes ( Perezhogin & Glazunov, 2023 ) and a newly proposed closure of thickness fluxes. In Section 3, we show a-priori and a-posteriori results and compare them to high-resolution filtered data. Finally, concluding remarks are made in Section 4. 2 Methods Here, we attempt to improve the dynamical core of the MOM6 ocean model, which solves the stacked shallow water equations, with improved LES subfilter closures. 2.1 Double-Gyre configuration The stacked (i.e., multilayer) shallow water equations in the vector-invariant form are given as \displaystyle\partial_{t}u^{k}-(\omega^{k}+f)v^{k}+\partial_{x}(K^{k}+M^{k})=\mathcal{F}^{k}_{x}, (1) \displaystyle\partial_{t}v^{k}+(\omega^{k}+f)u^{k}+\partial_{y}(K^{k}+M^{k})=\mathcal{F}^{k}_{y}, (2) \displaystyle\partial_{t}h^{k}+\partial_{x}(u^{k}h^{k})+\partial_{y}(v^{k}h^{k})=0, (3) where k is the index of the fluid layer, u^{k} and v^{k} are zonal and meridional velocities, and h^{k} is the thickness of the fluid layer. Further, \omega^{k}=\partial_{x}v^{k}-\partial_{y}u^{k} is the relative vorticity, K^{k}=1/2((u^{k})^{2}+(v^{k})^{2}) is the KE per unit mass, and M^{k} is the Montgomery potential (pressure anomaly). Also, f is the Coriolis parameter and \mathcal{F}^{k} is an additional forcing and dissipation. Note that there is no horizontal molecular viscosity in these equations and thus the small-scale dissipation is provided by the subgrid parameterization only. We consider an idealized GFDL MOM6 ocean model in a classical double-gyre configuration described in Perezhogin et al. (2024) and shown in Figure 1 . Two fluid layers are governed by the stacked shallow water equations (Eqs. ( 1 )-( 3 )), with \mathcal{F}^{k} given by a wind stress applied at the surface and a drag force applied at the ocean floor. The model domain is on a sphere with a spatial extent of (0^{\circ} - 22^{\circ}\mathrm{E}) ~\times~ (30^{\circ} - 50^{\circ}\mathrm{N}) which is approximately 2000~\mathrm{km} ~\times~ 2000~\mathrm{km} . We run experiments for 20 years of model time and use the last 10 years for computation of statistics to isolate from the transient development. The high-resolution simulation at resolution 1/64^{\circ} resolves the mesoscale eddies with 10 grid points, as its grid spacing ( 1.5 km) is 10 times smaller than the first baroclinic Rossby radius ( 15-30 km), the scale at which mesoscale eddies are generated (see blue shading in Figure 1 (b)). In contrast, ocean models with the coarsest resolutions considered ( 1/2^{\circ} and 1/4^{\circ} ) have an insufficient number of grid points per Rossby radius, leading to insufficient eddy energy across all spatial scales (Figure 1 (b)); here eddies are defined as deviations from the 10-years average. In considered baseline experiments, we parameterize the subgrid momentum fluxes with the biharmonic Smagorinsky model having a coefficient of C_{S}=0.06 , which is a default value in the global ocean model ( Adcroft et al., 2019 ) . Smaller values of C_{S} are possible but will lead to the appearance of numerical noise in the vorticity field. The mentioned lack of eddy energy is caused by the too-strong dissipation provided by the biharmonic Smagorinsky parameterization. The remainder of this report focuses on replacing this parameterization with alternative subgrid models to enhance the coarse-resolution ocean models. Figure 1: Idealized GFDL MOM6 ocean model in a double-gyre configuration parameterized with biharmonic Smagorinsky model ( C_{S}=0.06 ): (a) computational domain, (b) spatial spectrum of eddy kinetic energy (EKE) in a box (5^{\circ} - 15^{\circ}\mathrm{E}) ~\times~ (35^{\circ} - 45^{\circ}\mathrm{N}) in the upper fluid layer, (c) snapshot in the high-resolution model ( 1/64^{\circ} ). 2.2 LES of stacked shallow water equations Equations (