| 08 Sep 2025
Impact of mesoscale eddy parameterization on Arctic Atlantic Water circulation and heat transport in the eddy-permitting grey zone
Abstract. The Arctic Ocean is undergoing rapid change, yet many CMIP-type climate models struggle to accurately represent its circulation and water masses. A key feature of the system is the topographically controlled boundary currents that transport warm, saline Atlantic Water northward at intermediate depths into the Atlantic Water layer. An important process affecting these boundary currents is the lateral flux of heat and salt driven by mesoscale eddies. Because the deformation radius is relatively small in the Arctic Ocean, numerical simulations require kilometer-scale resolution to fully capture eddy dynamics. Most of the current climate models, however, operate in a non-eddying regime, relying on mesoscale eddy parameterizations – typically combining isopycnal diffusion (Redi) with eddy-induced advection (Gent & McWilliams, GM). As horizontal resolution increases, future models will shift from an eddy-parameterized to eddy-permitting regime, entering a grey zone where eddies are only partially resolved and the role of GM parameterization becomes less straightforward. This study investigates the use of GM parameterization in eddy-permitting models, focusing on its effect on the northward transport of Atlantic Water in the Nordic Seas and Arctic Ocean. We conduct realistic simulations where we vary GM diffusivity strength and test two different GM scalings. These experiments are compared with a high-resolution model-truth simulation and observational data. Consistent with earlier idealistic studies, we find that omitting the GM parameterization excludes important eddy–mean flow interactions, which in our case results in stronger meridional overturning, barotropic circulation, and northward transport. Conversely, applying GM results in weaker circulation and dampens resolved eddy fluxes, while the parameterized fluxes introduce biases, particularly in the temperature distribution in the Greenland Sea.
Review of “Impact of mesoscale eddy parameterization on Arctic Atlantic Water circulation and heat transport in the eddy-permitting grey zone” by Pemberton et al., egusphere-2025-4241
The manuscript describes numerical experiments with a regional model of the Arctic Ocean with two different resolutions and different ways of parameterising unresolved eddies with the GM/Redi scheme.
The introduction is well written and easy to read. The manuscript sets out to address important questions of eddy parameterisation in resolutions where the eddy field is partly resolved (eddy-permitting), impact on notoriously poor model skill in Arctic simulations (with often too warm ocean temperatures), but these questions are never clearly formulated and hence the manuscript turns out to be a long and sometimes confusing description of a comparison between the different model simulations and some observations.
To be fair, the presented results are not straightforward to interpret. For example, sometimes higher resolution reduces meridional overturning, in other places it increases it. Using small GM parameters increases meridional overturning over no GM parameterisation in the coarser setup, but using large GM parameters reduces the overturning. In a different place the model with the lowest overturning is the one without GM parameterisation (but not consistently). These phenomena are described but not explained.
Another example: in the Norwegian and Greenland Seas the “worst” model (low resolution, no GM) gives the closest fit to observations (Fig11), better than the “model truth”, but in the Fram Strait and West-Spitzbergen Current it is among the worst (as expected). In the interior Arctic (Fig12), the high resolution “model truth” has the worst fit to observations.
I can only speculate, that the choice of experiments makes it very difficult to extract information specifically about eddy fluxes (no eddy fluxes are explicitly computed, neither for the high resolution model nor the parameterised eddy fluxes which could be diagnosed as well) and eddy activity and GM/Redi, as the “model truth” is eddy resolving only in part of the domain, and the change in resolution also introduces confounding factors such as better resolution of topography (coastlines and bathymetry), that will most likely have a profound effect on the simulation. Different spurious mixing in high and low resolution may also be a problem that is not addressed, also the GM/Redi parameterisation introduced a strong vertical mixing contribution.
As a consequence the results section of the manuscript is very difficult to read (and also less carefully written than the very good introduction) and is very difficult to extract any conclusions. The authors leave the reader with little to no explanation for their sometimes unintuitive results (and they acknowledge the unintuitive results in l494 without further explanation). The final section is not much more than a summary. Many important subjects are touched or implied but not properly addressed (e.g. why does higher resolution lead to an additional warming and salinity bias, Fig12, also observed in other models of the Arctic Ocean), and the modelling reader is left with the conclusion that things are difficult with and without GM/Redi and that we need more work (but that’s something that was also so before reading the manuscript).
I recommend that the authors focus on fewer “metrics” and not only describe but also interpret the results (where do eddy fluxes, parameterized or not, but diagnosed from the simulations lead to the differences seen in the diagnostics?), also in the light of benefits and shortcomings of the GM-parameterisation in the Arctic. Is GM an appropriate parameterisation for Arctic eddies? What does one need to do to “get it right”? Maybe diagnose GM coefficients from the high-resolution run. I also recommend additional experiments, e.g. a high-resolution simulation with the coarse resolution topography to exclude additional confounders, a GEOMETRIC experiment with smaller values of alpha to get a better agreement with the observational estimates.
Additional comments (sometimes repetitive of the main points):
page 2
l38: AW reach -> reaches? (Since AW abbreviates “Atlantic Water” and not “waters”)
page 4
l62: attration -> attraction (maybe better attention?)
page 5
Caption of Fig2: I would use: $L_d/\min(dx,dy)$ (so min => \mathrm{min}
Caption of Fig2:
The abbreviations ARCTIC12, ARCTIC025 only become clear later an maybe should be explained also here in the caption
page 6
l127: the -> a?
The way this description is written implicitly assumes that the reader is familiar with the different options of NEMO (in my opinion, ‘the’ vector-invariant formulation, should be ‘a’ vector-invariant formulation, etc.)
l130: a diffusivity coefficient that is scaled by a defined velocity scale Udiff = 0.0193 and L.
Not clear to me how this works exactly (again a little bit o NEMO insider knowledge is required here). Is the coefficient Udifff*L? Or Udiff*L**2 (which would make more sense to me)? Same for viscosity, where I would argue that one needs to include the timestep as well to guarantee stability for large viscosities.
page 7
l140: The background values for vertical diffusivity and viscosity are tuned down to 5.4·10−6 and 5.4·10−7 m2s−1
Probably not important for this paper (although GM + Redi can have a strong vertical contribution to vertical mixing), but I thought that the TKE scheme of NEMO (Gaspar et al, Blance+Delecluse) uses a be minimum TKE value that accounts for the background mixing (due to internal tides) and no explicit vertical background diffusivities are required. If you use them, then the background mixing is introduced twice (according to my understanding, see also Brueggemann et al, 2024, doi:10.1029/2023MS003768)
l150: a landfast ice parameterization is utilized
Here and in general, I would appreciate references for the various schemes (also Flather, Orlanski, etc.)
l154: ARCTIC12 without GM scheme and this serves as a model truth calculation
Model thruth without full eddy field? How so?
l161: higher kappa_GM
Than the “standard GM scheme scaling coefficients”?
l166: while the GEOMETRIC scaling mostly yields values that are too high
As far as I can see the GEOMETRIC values could be tuned down by alpha, why didn’t you use a smaller value of alpha to avoid the overestimation of kappa_GM?
page 9
l180: Dai and Trenberth dataset -> the Dai and Trenberth dataset ?
page 10
l210: However, to facilitate comparison to observational estimates and other modelling studies we use this common computation of "heat" transport.
I would still call this a temperature transport to avoid confusion.
page 10
L214 analysis -> analyse/analyze
l214: by looking at
Too colloquial to my taste (you can look at a photo or painting etc). I would use something more specific like inspect, investiage, analyse, etc.
page 11
l237: the contribution of EKE drops off significantly in the high resolution experiments
Here it becomes clear that even the 12th deg simulation does not resolve any eddies (in the interior Arctic) and all that EKE shows is kinetic energy of variability on monthly timescales as the EKE is not much higher for Arctic12 than for the other simulations.
page 12
l240: I would rewrite as:
The drop off in eddy activity between the experiments and the dampening effect of the GM scheme on the resolved eddy field can also be clearly detected in snapshots of relative vorticity and potential temperature (Figures A1 and A2).
l256: start -> starts
l253 Nordic Seas.
The results are not quite as expected as the high resolution stream function is weaker than the coarse one without GM. Clearly the eddy activity in this region contributes to the overall hydrographic gradients and hence the mean circulation. Something that would be nice to discuss here (but not done)
l262: For the subpolar gyre, it’s even different with Arctic12 being the strongest and ARCTIC025-noGM being relatively weak.
page 14
l268: Section 3.3 Meridional overturning circulation
Results are not conclusive and require explanation
l288: At the Iceland–
page 16
l298: In our analysis this contribution is instead included in the Faroe–Scotland gateway making our Scotland–Norway inflow lower.
Not clear why the analysis was adjusted according to the observations before comparison.
page 16
l302: possibly due to better resolving the channels of the gateways.
It’s important to differentiate between the effects of resolution internal dynamics (eddies) and topography.
l306: “The northward heat transport that AW …”
Although important for the temperature (bias) of AW (in models) and the entire Arctic, the “heat” flux comparison is difficult and potentially inconclusive, because it does not only show the “heat” flux differences but also differences in mean state, if the volume flux is not balanced. It would be possible to have closed volume budget (for fluxes through all straits of the Arctic) in the model, so I strongly recommend to use that property of the model and have “real” heat fluxes.
page 18
l324: Although the in- and outflows in the model experiments are higher than estimates reported Tsubouchi et al. (2012).
Main clause is missing
page 19
l348: leads
page 20
l374: is a well-defined
l379: The GM scheme impacts
page 21
Fig9: no observations? There are multiple observations of temperature alone this section, even current velocity (e.g. Beszczynska-Möller et al., 2012). These direct measurements would be easy to compare to (much easier than, say, inferred “heat” or volume flux)
l389: equals 233 meter
Earlier (l221) a different depth (623m) was chosen to identify the AW layer for the MKE and EKE plots in Fig4.
page 23
l425: observed values.
Not really “observed” but inferred from observations and an inverse model, so maybe “observation-based estimates”?
page 25
show more uniform differences
l454: Also in the entire Eurasian basin is the AW layer thickness over estimated with model experiments showing 980−1370 compared to 630 in PHC3.0.
Grammar (the AWI layer thickness is overestimated); units (meters) are missing
page 27
l472: The meridional overturning is weaker in the northern North Atlantic using GM, however, no sensitivity is seen in the Nordic Seas overturning
But not in a consistent way!
l483: omitting the GM scheme yields too wide AW core
Counterintuitive as GM should flatten isopycnal, here is seems that omitting GM flattens the isopycnal and stretches the AW across the entire strait. Not explained.
Also “yields a too wide AW code” or yields an AW core that is too wide”
page 28
l503: It should be noted that we did not spend a great deal of effort to tune the new GEOMETRIC scheme.
Maybe this would have been necessary to obtain a more realistic GM-parameter field?
l505: observational estimates by Kusters et al. (2025) show that the GM diffusivities have a vertical structure with reduction of diffusivities with depth in some regions.
I am not sure if that’s the best conclusion from their paper. It rather shows that the assumption of an eddy parameterisation with a uniform GM coefficient may not be appropriate. The fact that the coefficient is not uniform says that the parameterisation is not complete.
page 29
l530: Data availability.
No available data from the simulation results
page 31
l602: Kusters, N., Balwada, D., and Groeskamp, S.
Incomplete reference, no journal.