Review of "The ACCESS-CM2 climate model with a higher resolution ocean-sea ice component (1/4°)" by Huneke et al.

General Comments

This manuscript introduces and evaluates a new configuration of the Australian Community Climate and Earth System Simulator (ACCESS-CM2) coupled model, based on a ¼-degree ocean-sea ice component (ACCESS-CM2-025). 

The primary objective of developing this new model was to enhance the representation of ocean mesoscale features and broaden the scope of climate modeling research applications. The manuscript comprehensively assesses the simulated climate of ACCESS-CM2-025 over a 500-year present-day simulation by comparing it against various observation-based datasets, the existing lower-resolution version of the same coupled model (ACCESS-CM2-1), and two forced ocean-sea ice models (ACCESS-OM2), OM2-025 (0.25° resolution) and OM2-1 (1° resolution), which share the same ocean-sea ice components as the coupled models but are driven by JRA-55-do.

Overall, the study concludes that the higher grid resolution of the ocean-sea ice component in CM2-025 has improved many aspects of the simulated climate, especially in dynamically active regions. Still, some biases persist, pointing to remaining deficiencies in the atmospheric or oceanic model components, or a need for even higher resolution. The manuscript is well-written, and the authors effectively present the evaluation of this new coupled system using key metrics. My main suggestion, detailed below, is to include additional information about the ¼-degree ocean setup and to clarify the relative contributions of increased horizontal resolution versus mesoscale parameterizations to some of the observed improvements. The manuscript is well-suited to Geoscientific Model Development, and I strongly recommend publication pending minor clarifications as outlined below.

Specific Comments

Lines 46-48: Please briefly list what are the different approaches that the modeling centers take to the issues of eddy-permiting resolutions. Which of these approaches does ACCESS-CM2-025 follow?

Table 1: It would be helpful to include the total number of processors used for each configuration, along with the breakdown across model components. Additionally, the manuscript should clarify why performance differs so substantially between OM2-1 and OM2-025 compared to CM2-1 and CM2-025.

Lines 124-125: Please briefly describe the sub-grid scale parameterizations used in this study, including both mesoscale and submesoscale schemes, as well as the vertical mixing scheme employed. In particular, given that modeling centers differ in whether Gent–McWilliams/Redi is applied at this resolution, it would be helpful to clarify the approach taken here. Additionally, please explain how liquid and frozen freshwater inputs were represented or distributed in the model.. 

Lines 138-140: Based on the slope of the ocean temperature time series, it appears that the top-of-atmosphere energy imbalance differs substantially between CM2-1 and CM2-025, with CM2-025 showing a significantly reduced imbalance. Since the only change between these configurations is the ocean and sea ice horizontal resolution (with no additional atmospheric tuning), this is a notable result that the authors may wish to emphasize more clearly.

Lines 215–217: The statement, “The fact that both higher-resolution models exhibit the same behaviour confirm the requirement of a higher grid resolution to adequately resolve the spatial structure of the ACC, even if the integrated transport deviates more from the observational estimate,” may overstate the attribution to resolution alone. Given that the GM parameterization is also applied, how can the authors be certain that the improved ACC structure is solely due to increased resolution? Clarifying the role of GM in this context would strengthen the argument.

Table 2 and Figures 2 and 3: Is the thickness of the first model layer, used for extracting SST and SSS, the same as that of the corresponding level in the WOA13 vertical grid? If not, were any interpolations applied to facilitate the comparison? Additionally, please include a reference in Table 2 to the observation-based MLD dataset used.

Lines 256–263: This paragraph mixes observational and model definitions of MLD, making it difficult to identify the key point. The observational dataset referenced (https://www.seanoe.org/data/00806/91774/) uses an updated MLD definition based on a 0.03 kg/m³ density difference from 10 m depth, as described in Treguier et al. (2023). Therefore, the appropriate citation should be de Boyer Montégut et al. (2023) rather than the original 2004 paper. In contrast, the model uses a buoyancy threshold of 0.0003 m/s² relative to the surface, following Griffies et al. (2016). It is unclear how these two criteria are “nearly identical,” as claimed. A clearer justification or quantification of their similarity would be helpful.

Lines 263–269 and Figure 5: The authors compare the maximum monthly climatological mixed layer depth rather than analyzing a specific month or season. While this approach highlights the deepest annual MLD, it may obscure important aspects of the seasonal cycle, such as the timing and rate of MLD deepening and shoaling. Including an assessment or brief discussion of how well the seasonal evolution of MLD is represented in the models, particularly in regions with strong seasonal variability, would strengthen the analysis. Given the presence of open-ocean polynyas in the models, a focus on a specific month or season would also be especially informative. Finally, a bias map (model minus observations) would add value, as Table 2 only covers a limited number of regions.

Lines 270–277; lines 295-296: It is surprising that open-ocean polynyas occur in the ¼° model. Were these events persistent, and how long did they typically last? It would be helpful to clarify whether the authors attempted to enhance restratification processes, either through mesoscale or submesoscale parameterizations, during model development to reduce or eliminate the occurrence of such features.

Figure 6: The thick tick marks on the x-axis are misaligned between the upper and lower panels. Which one is correct? Based on the latitude labels in the lower plots, the transitions between ocean basins do not appear to occur at 30°S as indicated.

Line 296: In addition to highlighting improvements to the vertical mixing scheme, the authors might also consider discussing improvements to re-stratification processes via mesoscale and submesoscale parameterizations.

Lines 319–321: The authors might consider including a comparison of the AMOC vertical structure in depth space at 26°N, where long-term observational data are available (e.g., from the RAPID array), to provide a more direct evaluation against observations.

Lines 429–431 and Figure 10: I do not clearly see the “smaller but negative bias dominating in the higher latitudes” as stated. Please double-check this claim for accuracy. Additionally, consider highlighting the zero-contour line in Figure 10 to make the sign and structure of the bias more apparent.

Editorial / Typographical Comments

• Line 216: “confirm” => confirms.
• Figure 8: Please consider increasing the size of the two plots to improve readability and better highlight the differences between the model configurations.
• Lines 338-339: “(fresh bias in the Southern Hemisphere for CM2-025)”, this is misleading as CM2-1 shows a near-surface fresh bias in the Southern Hemisphere. Please clarify.
• Lines 368-369: Please improve this sentence; “... CM2-1 rather poor” perhaps should be “... CM2-1, which is rather poor”.
• Line 434: ".. and Aghulas.." -> "The CM2-025 biases improve in the Kuroshio and Agulhas region when compared with observations." (Spelling of "Agulhas").
• Line 519: "...onto the finder ocean grid." -> "...onto the finer ocean grid."

References:

Treguier, Anne Marie, Clement de Boyer Montégut, Alexandra Bozec, Eric P. Chassignet, Baylor Fox-Kemper, Andy McC. Hogg, Doroteaciro Iovino et al. "The mixed-layer depth in the Ocean Model Intercomparison Project (OMIP): impact of resolving mesoscale eddies." Geoscientific Model Development 16, no. 13 (2023): 3849-3872.

Review of “The ACCESS-CM2 climate model with a higher resolution ocean-sea ice component (1/4 degree)” by Huneke et al.

This manuscript provides a model documentation of a new ACCESS CM2-025 coupled climate model, which has an increased ice-ocean resolution of ¼ degree and uses the same atmosphere land components as ACCESS CM2-1 (which has a 1 degree ice-ocean resolution). The manuscript compares 4 simulations each run for 500 years: Year-2000 control experiments performed with the two coupled models and ice-ocean simulations forced with repeat year forcing performed with the ice-ocean components of each coupled model. The authors compare these four simulations to assess the impact of increasing the ocean and sea ice resolution from 1 to ¼ degree. They find some simulation benefits associated with moving to ¼ degree resolution, including increased eddy kinetic energy, improved ENSO dynamics, and reduced model drift. However, a number of climate biases are unchanged, and others are degraded. They also find that the CM2-025 model exhibits a large centennial-scale variability associated with deep convection in the North Atlantic.

The paper is clearly written, easy to follow, well referenced, and has a number of results that will be of interest to the broader modelling community. The subject matter is certainly appropriate for GMD. I appreciated that the paper was well scoped, which made it very readable for a model documentation paper. I have a number of minor comments for the authors to consider, and also one major comment related to the sea ice analysis in the manuscript. I am recommending major revisions, since I consider the sea ice comment essential to address, but I expect that most of my other comments should be relatively straightforward to address.

Also, note that I only read the comments of Reviewer 1 after doing this review, in an effort to provide an independent assessment. I agree with their assessment, and there are some commonalities between our reviews.

Major comments:

Sea ice analysis:

Section 3.2: The observed sea ice extent values look strange, both for the Arctic and Antarctic. Which NSIDC observational product is being used here? Was the polar hole in the Arctic taken into account? If I compute sea ice extent climatologies using the NSIDC SIC CDR dataset (https://nsidc.org/data/g02202/versions/5), I get the following approximate values, which are notably different from those shown in Fig. 8 and Table 2:

Arctic (Max, Min): 15, 6.5

Antarctic (Max, Min): 19, 4.5

Please double check the sea ice observational values, as these large differences have important implications for the model results. This will affect many of the statements throughout section 3.2, for example Lines 367-369, 372-375, and 385-390.

Minor comments:

L37-38: This is a bit confusing since the resolution is the inverse of grid spacing, thus has different units than the Rossby radius. Suggest changing to something like: “A lateral grid spacing of less than half of the Rossby deformation radius is needed…”

L71: “…from the atmosphere to the ocean on larger scales…”

L70-81: Also suggest mentioning that sufficiently high atmospheric resolution is required to capture these small-scale coupled air-sea interactions.

Table 1: It is useful to know the performance in simulated years per day of each configuration, but it would also be helpful to know the number of CPU cores required for these simulations. Please add another column providing this.

L115: I assume that the sea ice model uses an ice-thickness distribution? Please add the number of ice thickness categories here.

L121-124: Since the impacts of these bathymetry changes come up a number of times in the manuscript, it would be helpful to provide some additional details on the exact changes that were made (i.e. the number of grid cells changed, the change in seafloor height, etc).

L125-126: Given the strong focus on eddy impacts in this manuscript, it would be helpful to provide more details here on the GM implementation. For example, what diffusivity coefficients are used in the various simulations? Is a lateral resolution function employed? Is there a vertical structure function used? Where is the GM parameterization active in the 0.25 degree runs?

Section 2.1: Do the ice-ocean simulations include a surface salinity restoring? If so, please add the restoring data set used and the restoring strength.

L128-136: L128 states that the models are initialized from a prior PI control simulation, then L132 says that WOA is used. Which one is it? This should be clarified.

L139: The ocean heat uptake and interior ocean warming also results from TOA energy imbalance. This should also be mentioned here.

L137-148: Given that this is a year 2000 control, we expect SST and interior ocean warming due to the TOA energy imbalance. It would be helpful to have a sense of how much of the model drifts shown in Figure 1 are due to the Year-2000 forcing and how much are due to erroneous drift associated with model physics and numerics errors. Were any 1850 pre-industrial control runs performed with these configurations? Comparing these to the Year-2000 control would be a way to estimate the contributions from Year-2000 forcing in Figure 1. If any pre-industrial control runs exist, I suggest looking at them and adding some comments on this in the text.

Table 2: I don’t see a superscript “b” in this table.

L164-168: It seems that the ocean component also plays a role in the Southern Ocean SST biases. In particular, the ocean-only OM2-025 run displays a number of Southern Ocean SST bias features that are also present in the CM2-025 run, albeit with smaller magnitude, suggesting that ocean physics errors are also a contributor to these biases. I suggest mentioning this point here and softening the current text, which makes a too-definitive attribution to the atmospheric component.

L187-190: Suggest citing Khosravi et al. (2022) here (https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2021EF002282), who show that similar positive salinity biases in the Canadian Basin are ubiquitous across CMIP6 models.

Line 190: Change to: “Other challenging regions…”

Table 2: I suggest adding RMSE values for SST and SSS to Table 2 (i.e. the RMSE based on local SST/SSS errors). This would help with, for example, the point made on L195-196.

L206-208: I’m confused by this argument. The Southern Ocean SST bias is positive, which should reduce the equator-to-pole temperature gradient, reduce the meridional density gradient, and weaken the thermal wind contribution. Please clarify this line of argumentation.

L210: “this type of model”

L232: What temporal resolution data is used to compute this standard deviation, for both the models and observations?

Figure 6: Suggest adding an annotation to the figure (maybe a vertical line at 30S) to make it clear that two different zonal regions are being plotted on the same figure.

L330: What is the observational target for AABW density? Or it this statement derived from the T/S plots in Fig. 7?

Figure 7: Suggest adding text annotations to panels 7a-h indicating the RMSE of these plots. This would provide a useful summary statistic for quickly comparing the models.

Figure 8: Suggest reformatting this figure to be one row and two columns to improve the aspect ratio of the panels.   

Lines 376-384: Does the sea ice concentration field in the Ross Sea show an imprint of the deep oceanic convection that occurs in this region? It is not visible in the sea ice extent, but I wonder if a patch of reduced sea ice concentration is visible. If so, it would be worth adding a comment on this.

Lines 397-398: Is this difference really being driven by ice export rather than differences in thermodynamic growth? This statement needs to be backed up by some analysis of the sea ice mass budget terms or the velocity fields. Or the term “export” could simply be removed from this sentence.

Line 412: “Sea of Okhotsk”

Figure 10: The strip of data in the Arctic (north of roughly 80N) looks strange. Are these missing data values?

Figure 11: It would be helpful to also see the ENSO signal along the west coast of South America. Suggest including more data, such that the white ocean regions in Figs. 11a-c also show regression values.

L448: “…too far East…”

L449: “…extend too far…”

L458: The text annotated in the plot say 0.09 rather than 0.11. Which one is correct?

L460-465: It would also be useful to see the power spectrum of the Nino 3.4 SSTA. Is this also markedly improved in the CM2-025 model? I leave it up to the authors whether they want to add this to Figure 11.

L503-507: Please provide more details on this. What aspects of the CM2-1 climate mean state would make it less prone to multidecadal variability?

Section 4.2: Does this model exhibit any multidecadal variability in the Southern Ocean associated with deep convection in the Ross and/or Weddell Seas? Given the mixed layer depth biases in these regions, they could be a candidate for multidecadal polynya variability arising from a charge/discharge mechanism of subsurface temperatures. This arises in many models, for example the GFDL CM4.0 model (see Fig. 3 of Held et al. (2019), 10.1029/2019MS001829). I am curious whether you see this in any of the ACCESS configurations.

L 525: Change to: “For this metric, CM2-025 compares best of all four models…” to make it clear that this statement is referring to Figure 4.

L 547: Suggest changing to: “affects the AMOC and global mean surface temperatures.”

L551-554: Are there any planned future uses of CM2-025 (e.g. CMIP simulations, initialized predictions, etc.)? If so, it would be interesting to mention those in this final paragraph.

Mitch Bushuk

NOAA GFDL

Review of “The ACCESS-CM2 climate model with a higher resolution ocean-sea ice component (1/4◦)” by Huneke et al.

Review by Brandon Reichl

This article summarizes a higher-resolution ocean counterpart to the ACCESS CM2 climate model.  The original model simulation utilizes a 1 degree ocean grid spacing.  The new simulation uses a refined grid with 0.25 degree grid spacing.  This is reported as the only difference between the two models, such that the comparison demonstrates the impact of resolving part of the mesoscale eddy field with the finer grid-spacing of the 0.25 degree model.  The results are presented for the coupled “CM” version alongside a forced ocean and sea-ice experiment “OM” for context of the role of coupling.  Overall the impact of refined resolution is mixed, with several aspects of the simulation being improved as expected, but other important model biases either being unaffected (or even emerging) in the 0.25 degree grid.  The results are valuable and yield new insight into the role of ocean resolution in coupled climate models, so I support publishing this manuscript.

The study is well written and the analysis highlights many relevant features of a finer resolution ocean and is clearly presented.  While there are previous studies to compare eddy-permitting and eddy-parameterizing models in the literature, I agree with the authors’ statement that it is still useful to document this model and report the impact of high-resolution in different models since it can be very model specific.  The approach to present results from the CM experiment alongside the OM experiment also helps clarify the role of horizontal resolution in coupled vs forced ocean models.  I have several comments that could be taken by the authors’ to clarify and improve certain aspects of the analysis and discussion in a revision.

General comments

1. I appreciate the brevity of this article to keep the message succinct, since the CM2-1 and OM2 series models are already published resources. However, I do think some additional explanation of the eddy closure framework adopted in this work would be incredibly useful for a reader and the messages here.  Perhaps it can be briefly summarized in the text, but more detail in an appendix could be warranted?  It would be most useful to summarize the scale-aware eddy parameterization, especially since it is not retuned in the CM2-025 experiments.  This could aid some potential reflection on scientific benefits vs computational cost for resolving vs parameterizing eddies that would bolster the discussion section.  It would also benefit the paper to mention all employed parameterizations that may be relevant for the results and biases highlighted in this work, including for surface ocean vertical mixing (KPP?), any submesoscale parameterization, and any other relevant included parameterizations that can affect the mesoscale eddy field (lateral viscosity, diffusivity, etc).  
2. Somewhat related to Point 1, a comment on the lack of “retuning” of CM2-025 compared to CM2-1 may aid some of the discussion.  E.g., CM2-1 probably included some parameter optimization at some point in its development that targeted specific aspects of the simulation with a 1 degree ocean (Bi et al. 2020 documented albedo settings that were considered for improving ice and SST climatologies). If I understood it that the mesoscale eddy closure employed here utilizes a resolution function, then perhaps the hypothesis is that the model should not need any retuning.  But in principle resolving some eddies is likely superior to parameterizing all eddies, so that this isn’t so straightforward to me.  The clear advantage for not retuning the model here is that it allows for directly comparing the effect of only changing ocean resolution. But likely the mean bias comparisons (e.g., SST, etc.) could be quite sensitive to the lack of retuning, since CM2-025 is not likely optimized in this configuration.  To clarify this comment, I don’t mean to imply that it is necessary to retune the model for this work, it just may warrant some discussion.

Specific comments

L26: “with only few assumptions on the climate forcing”

There are many other assumptions in these models, so I’m not quite sure I understood what was intended by the statement.  Perhaps clarify.

L30: “Increasing the lateral grid resolution of the ocean model has a high priority as the ocean is a key component that links different parts of the climate system.”

This statement doesn’t obviously justify increasing the ocean resolution instead of the atmosphere or adding complexity elsewhere/ensembles/etc.  This could be worded differently to better emphasize why this work focused on refining the ocean resolution.  Was there a specific application or result in mind for a 0.25 degree model?

L32: “coastline, passages, and narrow straits”

Also representing narrow coastal shelves.

L35: “as detailed further below”

Suggest to be more specific. In the intro?  In this study?

Section 2:  Please confirm that the ocean time steps are also independent of resolution, or give their values (or state if dynamic).

L120:  “deepening of individual grid cells on the Antarctic continental shelf to avoid model crashes” 

Do you understand why this is only an issue in the coupled model?  The atmospheric coupled model resolution is much coarser than JRA, so I would have expected the OM model to experience stronger forcing.  Is there some coupled feedback mechanism?

L123:”as deepening of the gateways of the Mediterranean and Baltic Seas which otherwise salinify or freshen in the coupled model due to limited exchange between the respective marginal sea and the open ocean.”

Please clarify if this makes the gateway depths or cross-sectional areas closer to the measured values?

L130: “Using present-day forcing allows for comparison of the model output to observations which are lacking the required spatial coverage for the pre-industrial climate which is often applied for climate model control simulations.”

An alternative approach would have been to include a PI control spin-up for each model, followed by historical simulations with observed 1850-present forcing.  Any comments on the reason for using a present condition to fully spin-up the model instead?  Is there any consequence for this choice to the comparison to obs (or do you expect considering the transience of the present day forcing to matter?).

L135: “that is representative of the mean state”

Any comments on what is missed by excluding transience in these runs? Such as the various interannual atmospheric variabilities that may rectify on the mean conditions?  Perhaps it makes it somewhat more consistent with the coupled runs in the sense of the fixed climate forcings, but probably this is complicated since the coupled runs can simulate interannual modes of variability.

L151: “The analysis is guided by known CM2-1 biases”

This “known CM2-1 biases” qualifier is used a few times to this point.  It could help to include a paragraph earlier discussing these known biases that will be explored (and maybe also why this resolution is expected to improve the specific bias).

Table 2:  The mixed layer depth box regions could be shown in Figure 5 if it doesn’t make it too busy, e.g., to help a reader see what region encompasses the ACC.

Figure 2:  Some integrated metrics should be included here in the figure or caption to aid the comparisons, e.g., mean and RMS bias, spatial correlations.

L180: “The simulated global mean sea surface salinity (SSS) in CM2-1 matches well the observed estimates (0.02, Table 2), while CM2-025 is biased fresh (-0.1). Similar to SST, however, the global mean hides large spatial patterns of sizable magnitude revealing problems of both models to adequately simulate various dynamical features (Fig. 3).”

Would it be more appropriate to discuss the RMS biases rather than absolute given these caveats?

Figures 2&3:  The Caspian Sea is presumably masked out of the models?  I suggest to mask them out in the bias map rather than showing it as zero bias.  Similarly, I wonder if CM2-1 Black Sea is also masked out?  It seems in general the land mask in the figures is probably not consistent with the land mask from the models, which results in several “white” missing values regions in all figures which may be confusing due to many colormaps also containing white.  It may be cleaner to use the model’s land mask for all figures and eliminate any correspondence between the colormap and the color of any remaining missing values.

L192:  “in CM2-025 is improved in the Baltic and Mediterranean Seas”

It has improved compared to what?  The Baltic appears to have significant fresh bias in Figure 3a compared to 3b.  I presume that is due to the salt restoring in 3b, but it seems awkward to say it is improved without some context.  I wonder if there are similar strait transport issues in other seas with narrow exchanges that were not adjusted in this new grid, like the Black Sea and Persian Gulf?

L194:  “modest sea surface salinity restoring”

You could give the restoring piston velocity.

L261:  “nearly identical”

Please clarify the differences.  Treguier et al. (2023, https://doi.org/10.5194/gmd-16-3849-2023) noted many important differences that can arise from inconsistencies, so that it can be difficult to compare MLD metrics if not 100% like-to-like model:obs definitions.

L291:  “All configurations overestimate the maximum MLD which excludes the atmospheric forcing to be primary source for the biases.”

It isn’t obvious to me that you can rule out the possibility that both the JRA repeat year forcing and coupled model are not good representations of the important features of the atmospheric state.  I tend to agree that a significant fraction of the bias is likely originating in some part of the ocean component, but it may not be the whole story and JRA can certainly have its own issues.

L296: The deep MLD bias source could easily be caused by much more than just the vertical mixing scheme, it is likely to also be impacted by other processes and parameterizations that set the interior stratification.  Notably, the chosen eddy parameterizations for meso and submesoscale eddies strongly impacts the wintertime stratification/restratification.  Perhaps mention this here as well.

L296:  Any comments on the summertime mixed layer depths?  Are they not sensitive to the resolution?

L314:  Some more insight for these AMOC calculations would help here.  E.g., do these overturning strengths include any parameterized transports from the eddy closures?  Is the model the maximum overturning in density space (as per the figures) and can you please clarify that it is like-to-like with this observational estimate (it wasn’t obvious to me from the reference that the quoted obs overturning value was computed in density space).  Do the obs provide an error estimate?  Are you also able to provide the density and/or depth of the peak overturning to compare with the obs?

You might include in this discussion - there is a potential role for specifics of the vertical coordinate along with the horizontal resolution for improving deep convection and overturning watermass pathways (e.g., as discussed by Wang et al., 2015 in the context of a z* vs density based ocean component in an otherwise identical GFDL CM2 models, https://doi.org/10.1016/j.ocemod.2014.12.005; CM2M used MOM4, but I think shares several similarities with the MOM5 configuration here).

Figure 10:  What is the red/orange stripe at the northern edge?  Maybe a plotting artifact?

L359:  “The JRA55-do reanalysis product of the ACCESS-OM2 models”

This wording confused me at first, it may be clearer to say “The JRA55-do reanalysis product used to force the ACCESS-OM2 models.”

S3.3:  Can you also provide any comments on the impact on precipitation, if there is any impact?

L459: “This result suggests that CM2-025 produces ENSO variability with a life cycle much closer to the observations”

Figure 11 clearly demonstrates the reduced biennial variability.  A power spectrum of SST variance, e.g. in the NINO3 region, might aid this discussion if there is more variability to show in the longer ENSO periods.  (e.g., as in Figure 2 of https://doi.org/10.1029/2009GL038710)

Section 4.2:  Is there any connection to the Southern Ocean polynas, besides simply being in regions of bottom water formation?  I understand the mechanisms as hypothesized here are (mostly) local, but it is at least curious that these strong multi-decadal oscillations in deep convection regions occur in both high latitude regions in this model.  You may at least elaborate on the Southern Ocean polynya timescales in this discussion and contrast the two high-latitude variabilities.

Grammar

L448: “displays the maximum ENSO SSTA to far east”

To -> too