The present manuscript presents a high level overview of the differences and similarities between three configurations of the CMCC ocean model in forced simulations. The comparisons are limited to  the simulated state - no of process level explanations of the differences are provided. Further, while the emphasis is on the impact of horizontal resolution, there are a large number of other differences in the models being compared: vertical resolution, topography, the sea ice component model and the sea ice initialization, the salinity restoring timescale, the way that runoff forcing is applied, along with the more typical adjustments to viscous and diffusive parameterizations. So, while the authors have put in a considerable amount of effort in compiling model metrics, the resulting manuscript is rather unsatisfying. We have little new insight into how and why the explicit representation versus parameterization of the mesoscale impacts the simulated ocean state. As a documentation of what was done the manuscript may be adequate with some additional work to correct imprecise descriptions, but I do not envision the manuscript having much of an impact beyond those who might wish to use the CMCC models and it is unlikely to advance the field. Additionally, the manuscript suffers from being poorly prepared with many missing references and poor language constructions.

Detailed Comments:

Line 26 “enforces our ability in understanding ”: poor wording, perhaps is a prerequisite to develop our understanding

Line 29-30: ensemble size is a mother strong trade-off in both prediction and climate modeling.

Line 31 “start grid spacing”: rather, typical grid spacing. No standard (a specification) exists.

Line 33 “miss key processes”: it is not necessarily the case that key processes are missed. They may be parameterized.

Line 36 “Despite”: While simulations at this resolution …

Line 38: “access to which” : to assess to what extent

Line 60 “resolution dependent”: as discussed at top this is not a convergence study in the sense of numerical analysis. Many other aspects of the simulation besides horizontal resolution are change.

Line 68: Manral et al 2013 ref missing

Line 104: suggest starting new paragraph with sentence beginning “While the best …”

Line 154 “in coupled runs”: all the runs described herein are forced not coupled?

Line 189: do not correspond to what _was_ found in 

Line 203-204: I do not understand this distinction. The difference from the initial state taken from observations is the model bias?

Line 210: Lellouche et al 2021 ref missing

Table 2: is the standard deviation stated the inter annual standard deviation of the global mean SST or the mean standard deviation of the global spatial deviation of SST?

Line 254: I don’t understand the difference between “model physics”, which I generally take to be parameterizations and “unresolved processes” which require parameterization?

Line 260: Most of the SST biases are reduced”: This is not visually that obvious. State the rms error of annual mean SST at each resolution.

Section 3.3.2 and Figures 5-6: This discussion would be much improved by showing panels with the summer MLD (JJA for NH and JFM for SH) as a single plot and winter MLD (vice versa) as a single plot with the color scale appropriate to each season. The discussion of similarities in summer or lower latitude features is completely obscured by the full annual range color scale. This is also a very long section with little insight beyond that they are different. To what extent can we attribute differences to changes in preconditioning of water masses due to differences in large scale flow versus changes in the restartification power of mixed layer eddies in each case? Does the ORCA1 model include a parameterization (e.g. Fox-Kemper e al) of submesoscale mixed laser restratification included in its GM parameterization?

Line 338-342 discussion of higher order statistics in Johnson and Lyman 2022: What is the relevance to this study none of these statistics are evaluated in the paper

Line 349: Johnson and Lyman 2022 ref missing

Line 350: again, the use of the full dynamic range in the axis scale makes it difficult to compare the quality of the simulation of shallower mixed layers. The relative error could be just as large as for deeper M.

Line 374 “representations is underrepresented”: nonsensical phrase

Line 377: “unable to represent flow instabilities …”: yes, but the figure is showing mean low speed, not EKE

Line 385: “dependent on model numerics”: how so? Don’t all of the models use the same numerical methods to solve the equations of motion?

Line 386: “impact of mesoscale dynamics”: This has not been shown. It could simply be the impact of viscous boundary layer dynamics or topography which also differ across configurations

Line 404 “passed”: past

Line 421 “Figure 10 shows role of mesoscale eddy field …” : again, the figure shows the mean flow and no analysis of eddy-mean flow interaction is provided.

Section 3.3.2 (sigma overturning): A more precise definition of how the stream function was calculated at each resolution is required. Was it computed from Eulerian mean (monthly, annual, climatological?) velocity and density fields? Was the GM eddy-induced velocity included in the ORCA1 result? The authors should write down the integral with averaging operators in the appropriate places for clarity. This is important in trying to understand whether differences seen are related to “bolus velocity”, diapycnal processes or surface forcing. The reference to Andrews and McIntyre suggest that we should be interpreting something about eddy induced transport, but the discussion is unclear. One of the major differences is the structure of the strong clockwise cell in the ACC region which I presume is related to the degree of compensation of the Deacon cell. No discussion of this feature is provided, nor are we sure how to interpret the result given the uncertainty about exactly what is being shown.

Line 466: “This suggests that longer integrations are required for GLOB16 …”: It was previously stated that the analysis of all cases was for the first cycle (Line 150). Why is that only GLOB16 requires longer integration to be compared?

Figures 13-14 and accompanying discussion: This is a long discussion of the GLOB16 results alone with no explicit comparison across resolutions. It seems unnecessary if the purpose of the paper is investigating resolution dependence rather than an assessment of GLOB16 alone.

Line 560: Treguier 2012 and Robert 2016 refs missing

This manuscript provides a high-level assessment of the performance of the CMCC GLOB16 1/16° global ocean and sea ice simulation relative to two related simulations at 1° and 0.25° (ORCA1 and ORCA025) and range of observationally estimated values. The text is clearly written and logically organised, and the figures are appropriate and well-presented. While the manuscript provides few new insights, it gives a good overview of the main strengths and weaknesses of GLOB16 which will be of use to users of that model and as a point of comparison for other modelling efforts internationally. I would like to see it published after major revision to include more details of interest to the ocean modelling community, and I've provided detailed suggestions in the hope that this will help.

While the comparisons to observation are well-chosen and cover many of the most important assessment criteria for models of this sort, there are some significant omissions:

- Several references are made to sea ice processes and differences between the two sea ice models as explanations for differing ocean features, but no sea ice results are presented to back up these speculations. Comparing sea ice results (such as climatological maps of concentration and time series of ice volume, area and extent) with observations and between LIM2 and CICE would be very helpful to aid interpretation of the ocean biases and understand the differences between these models.

- There could also be more comparison of the variability at different resolutions relative to observations, e.g. maps of sea level standard deviation, to highlight the impact of increased resolution.

- No information on model computational performance is provided, e.g. the relative core-hours per simulated year at each resolution. This would be useful for practitioners in deciding whether the improved solution fidelity at high resolution is worth trading off against a shorter simulation duration within finite computational and walltime budgets. Other technical details (e.g. number of computational cores, parallelisation efficiency, etc) would also be helpful.

The depth of analysis could also be improved. Many parts of the text simply give a verbal description of what the plots show. This does not provide much value to the reader, and in some places the descriptions are also incorrect. Additional physical insights and interpretations would be very helpful.

An assessment of the impact of increased resolution in GLOB16 is hampered by many important differences between this model and ORCA1 and ORCA025 (e.g. a different sea ice model, twice as many vertical levels, many different parameterisations). This is unfortunate but perhaps  beyond the scope of this paper to rectify.

Re. Code and Data Availability - can the configuration parameter files for GLOB16, ORCA1 and ORCA025 be provided? And links to source code for LIM2 and CICE4?

There are many small typos and errors, which I have listed in the last section below.

Numbers in my detailed comments below are line numbers.

Comments on content:

56: Mention that this multi-resolution approach is similar to that taken at other modelling centers, eg Storkey et al 2018 https://doi.org/10.5194/gmd-11-3187-2018, Adcroft et al 2019 http://dx.doi.org/10.1029/2019MS001726, Kiss et al 2020 https://www.geosci-model-dev.net/13/401/2020/

84: if the model has been "extensively upgraded", then Iovino et al. 2016 is not very useful as a reference. Please provide a list of the important changes that have been made since Iovino et al. 2016

93: give the equation of state, and say what the prognostic variables are (e.g. conservative or potential temperature)

105: "While the best approach to identify the impact of grid resolution should be to change only resolution and associated physics in the suit of models, this was not the case in similar previous studies (Chassignet et al., 2020, Kiss et al., 2020, Li et al., 2020)." - while there were unavoidable differences in parameterisation due to the differing ability to resolve processes at different resolutions, Kiss et al. 2020 made significant efforts to harmonise the configurations across resolution. It sounds like an effort in that direction would be possible here and would more cleanly highlight the effects of improving resolution. Is that beyond the scope of this project?

Table 1: as also pointed out by Alexander Shchepetkin, "barotropic sub-step [sec]" should be "barotropic sub-steps". It would also be worth noting that this was a configuration error (the barotropic timestep is an order of magnitude smaller than needed for CFL stability) which would adversely affect the computational cost, but not the numerical solutions.

186: does the differing surface vertical resolution play a role in this SST difference?

190: "Kiss et al. (2020) where the 1/10° ocean surface ... with the largest bias from observations" no, obs in Kiss et al 2020 fig 3b is an anomaly offset by 18°, so this plot only compares trend, not bias

157: specify whether the salt restoring is constrained to have zero total flux. If not, how significant is the salt flux for the drift in total ocean salt mass?

228: "The SSS drift is offset by the surface salinity restoring that is incorporated into the codes to enforce salt conservation in the model ocean (in Sect. 2)." - do you mean "constrain SSS drift" rather than "enforce salt conservation"? Conserving total ocean salt mass would require balancing any net flux into sea ice and nonzero net restoring salt flux.

229: "The restoring of SSS drives its quasi-stationary evolution, the salt exchange between ocean and sea-ice due to ice formation and melting, is the only source of salt for the ocean." - so is there zero net salt restoring flux?

254: drift may also be due to biased forcing (as mentioned later in this paragraph), or adjustment from an unbalanced initial condition

257: or initial condition... though less likely for surface bias

Fig 4: if this bias was plotted relative to the WOA initial condition it would remove any effect from initial condition bias relative to ERSSTv5

Figs 5,6: Summer differences are very hard to discern. It would be better to plot winter in both hemispheres in fig 5 and summer in both hemispheres in fig 6, with different color scale.

Figs 5,6: extend latitude range further south to show Ross and Weddell convection

Fig 5,6: add citation for (a) to caption

Fig 7 adds little beyond what is in 5&6, other than an additional obs dataset - could be supplementary material.

Fig 7: cite obs datasets in caption

279: "The observed MLD is diagnosed through a density threshold criterion as the depth over which the potential density increases by 0.03 kg m-3 from the reference value of surface potential density taken at 10m depth; resulting values are mapped on a monthly basis at 1°x1° spatial resolution (de Boyer Montégut et al., 2004). The same density threshold method is applied to model output as recommended by Griffies et al. (2016) to compute the MLD in OMIP models." - This is not the same method because Griffies et al 2016 specify using the top model level as a reference, not 10m. Treguier et al 2023 state CMCC use the top model level, not 10m. Or was MLD recalculated to use 10m (ie different from the Griffies et al method)? If not, this difference may contribute to the bias, as Treguier et al discuss. Please clarify what was done and its bearing on comparison with obs.

297: also mention MLD bias at 0°E in mid-north Greenland Sea: several hundred metres too shallow

367: could also comment on the differences in the Zapiola anticyclone at different resolution - seem similar to what was seen by Kiss et al 2020.

384: "In an eddy-rich regime, the ocean model is less diffusive/viscous" is back to front: At high res, diffusivity/viscosity can be reduced while maintaining numerical stability; this allows WBCs to be realistically narrow and inertial, and the resolution of the internal Rossby radius allows baroclinic instability and an eddying flow

384: "less diffusive/viscous leading to an improvement in the strength and position of WBCs" - could relate this to narrowing the Munk (Laplacian) or Haidvogel et al 1992 (biharmonic) viscous WBC scales (A_lap/beta)^(1/3) and (A_bih/beta)^(1/5). Perhaps these scale values could also be useful in table 1:

ORCA1: Munk: (1e4/2e-11)^1/3~80km

ORCA025: Haidvogel: (1.8e11/2e-11)^1/5~25km

GLOB16: Haidvogel: (0.5e10/2e-11)^1/5~12km

Haidvogel et al 1992 ref: http://dx.doi.org/10.1175/1520-0485(1992)022%3C0882:BCSIAQ%3E2.0.CO;2

434: "The Antarctic coastal current is also clearly represented, it flows westward along the Antarctic coast and meets the eastward-flowing ACC at the Drake Passage, emerging as the Malvinas current." - isn't the Malvinas fed by the Antarctic circumpolar current, not the Antarctic coastal current? 

465: More discussion of the overturning south of 30°S would be good, e.g. in comparison with Farneti et al 2015. What depth does 1036.8kg/m^3 correspond to in GLOB16 in this region? Why is this subpolar cell mostly disconnected from the abyssal export cell in GLOB16 and ORCA1? It seems more connected to the surface cell at ~50S, 1036kg/m^3 than the abyss. The circulation below 1037kg/m^3 looks quite different from Kiss et al 2020 fig 7 and many of the models in Farneti et al 2015 fig 17 (and ORCA1 looks very different from the 1° CMCC model Farneti et al show) - can you point this out and speculate as to why? Also in ORCA1 a much larger fraction of the southern ocean upwelling is recirculated south of 35S in an intense clockwise cell rather than joining the NADW cell, and this overturning cell is stronger than at higher resolution - why?

589: is it feasible to do a RAPID-like calculation with your model data to verify this?

Further details needed:

86: specify the latitude at which the tripolar cap starts

87: is this a Mercator grid? If so, say so. e.g. replace "increases poleward as cosine" -> "meridional spacing decreases poleward as the cosine"

105, 121: state the ice grid used - is it the same as the ocean model? If not, what interpolation methods are used?

105, 121: state what fields are coupled between the ocean and sea ice in each direction

111: specify what is meant by the "in-house sea ice module"? is it CICEv4.1 (line 116)? But that's not in-house.

118: specify maximum ocean depth in the 3 models

120: "such as" - be specific. What else was used? e.g. Redi?

121: give GM parameter values

122: "multi-category" - specify how many

Table 1: suggest to also specify

- ice model dynamic and thermodynamic timesteps

- ocean-ice coupling timestep

- whether downslope transport or mixing schemes were used at low res

- whether Rayleigh was drag used in any straits at low resolution

133: specify JRA55-do version (1.4?)

136: specify the interpolation methods that were used from JRA55-do to the model grid in each model

141: is relative wind also used for stress calculation on ice?

145: "at the ocean surface in GLOB16" - is this literally in the top 0.8m, with no distribution to depth at all?

160: "the sea ice models used in our two systems employ different bulk salinity affecting the salt release from the sea ice to the ocean" - specify these values - how different are they? can the effect on stratification and circulation be quantified or at least estimated?

Fig 1,2 captions: specify averaging period (are annual means plotted?)

Small errors, typos, suggestions for clearer phrasing, etc:

7: "This paper describes the GLOB16 global ..." so GLOB16 is defined prior to first use on line 13

13: access -> assess

13: mesoscale activities -> resolving mesoscale processes

21: effort -> efforts

29: trade-off among -> trade-offs between

33: Both model configurations do not resolve -> Neither resolution resolves

L6: require -> requiring

38: access to which -> assess to what

51: is -> are

58: (low-, nominally 1° horizontal grid spacing), eddy-permitting (medium-, nominally 0.25°) to eddy-rich (high-, 0.0625°) resolutions -> (low-resolution, nominally 1° horizontal grid spacing), eddy-permitting (medium-resolution, nominally 0.25°) to eddy-rich (high-resolution, 0.0625°)

73: activities -> processes

84: , -> , and

88: - the -> . The

90: Outline -> An outline

93: momentums -> momentum

103: C grid -> a C grid

105: suit -> suite

107: computation cost of GLOB16 -> computational cost of the GLOB16

117: refinement of meridional grid to 1/3° -> meridional refinement to 1/3°

123: delete "Note that the two sea ice models LIM and CICE employ different bulk salinity, affecting the salt release from the sea ice into the ocean." - repeated in line 160

141: JRA55 -> JRA55-do

148: GLOB16 grid -> The GLOB16 grid

154: "in coupled runs" is unclear. I suggest "in ORCA1 and ORCA025" (assuming this is what was meant)

167: that warms -> warming

169: staying anyway -> but staying

170: specify whether the 0.1°C cooling is over 1 cycle or 6

175: over -> out of

187: 2c -> 1b

189: what found -> what was found

fig 1: swap b & c in caption

205: largely -> greatly

212: resolutions -> resolution

229: evolution, the -> evolution, and the

244: in 2018 -> by 2018 ?

258: activities -> activity

259: GLOB16 bias -> GLOB16

261: define "WBC" acronym on first use

262: differences are also in -> improvements are also seen in

265: regions -> regions than ORCA1 and ORCA025

268: experiments with a -> experiments due to a

275: insert last sentence of this paragraph here as the 2nd sentence

288: The mixed -> The winter mixed

290: Ross and Weddell convection is cut off in figs 5,6

298: remove "slightly" - this is quite a big difference

301 high-latitude -> northern high-latitude

305: horizontal -> zonal?

306: remove "caveats"

314: NH Sept MLD not interpretable with this colormap

329, table 1: were downslope transport or mixing schemes used at low res?

350: the observation one -> the spread of observations

360: (Fig. 8), zoomed in the key dynamical regions (Fig. 9 and 10), -> (Fig. 8), and zoomed in to the key dynamical regions (Fig. 9 and 10). The

404: passed -> past

404: follows closely -> closely follows

407: to OSCAR -> to the OSCAR

408: remove "in amplitude"

408: The decaying eastward along the Kuroshio extension and magnitude match the observed one -> The GLOB16 Kuroshio extension magnitude and its eastward decay match observations

409: the 170°E longitude -> 170°E

410: toward 145°E, to rapidly decay westwards. -> until 145°E, but decays too rapidly further east.

Fig 9: fix panel numbering (a-d happens twice)

464, 474, 475, fig 11: stated transports are double what is shown in fig 11 - is the text incorrect or is the contour interval 2 Sv, not 1 Sv as stated in the caption for fig 11?

480: "~2 Sv in density space (~ 6 Sv below 3000 m in depth space)." Are these back to front? Looks like about 6 Sv in fig 11 (or 12 Sv if the contour interval is 2 Sv)

467: "A portion" - quantify this

468: we do present -> we present

473: 55°N -> 65°N?

Fig 11: plot contours beyond current range (both positive and negative) - seem to be missing some in the far south, unless the extrema are very flat. Also state in caption how colour is related to overturning direction

497: till -> until

Fig 12 caption: Swap "dashed" and "solid" and state that ORCA1 is also plotted.

512: two decades -> decade ?

513: can't see low in 2009 in either obs record in Fig 13. "observed in 2009 and 2010" -> "observed at 26.5°N in 2005, 2010, 2011 and 2013"?

518, 528, fig 13: why not plot the other 2 models as well?

Fig 13 caption: state that the time scale is compressed prior to 2000.

539: 15°N -> 18°N?

540, 541: swap 10 - 20°N and 20-30°N (or is fig14a key wrong?)

542: MHT peaks around 24°N -> MHT peaks around 24°N but is not well constrained given the error bars

542-589: MHT -> AMHT

548: remove "The strongest heat transport is found in the eddy-rich ocean." - repeated below

550: GLOB16 tracks the ECMWF estimates and compares well with TF08 -> GLOB16 tracks the ECMWF estimates and compares well with TF08 except north of 40°N

552: Hirschi et al. 2019 -> Hirschi et al. 2020 ?

556: 18°N -> 15°N ?

565: slightly exceeds the total MHT between the equator and 15°N -> slightly exceeds the total MHT between the equator and 15°N and south of 20°S

568: gyre one increases to level off -> gyre component increases to level off at total AMHT

574: remove "not shown for ORCA025 575 and ORCA1" - this difference is visible in fig 14a.

578: misrepresent -> misrepresents

579: 2010, 2011, 2013 and 2017 minima -> 2010, 2011, 2013, 2017 and 2019 minima ?

Fig 14b: why not include the other 2 models?

608: negative -> positive

617, table 1: was Rayleigh drag used in the straits at low resolution?

629: plot Donohue et al value in fig 15c

631: below the most recent estimates (Xu et al., 2020) -> below the most recent estimates (Xu et al., 2020) and some eddy-rich models (Kiss et al., 2020)

Fig 15 caption: (b) the Indonesian Throughflow -> (b) the southward Indonesian Throughflow

Fig 15 caption: Cite sources for obs in a & b

671: provide -> provides

679: Indian to the Pacific -> Pacific to Indian

683: (weaker than observed values) -> (weaker than observed values and some other eddy-rich models)