General comments

In this study, the authors analyze five different MITgcm general circulation model based ocean reanalyses (3 global and 2 regional) to see how well they simulate the Southern Ocean.  Comparisons are made of the total sea ice extent and volume, planar views of the average temperature and salinity at 550m and February mean SST and SSS, time series of ACC and Weddell and Ross Gyre transport, vertical sections and time series of temperature and salinity that match observations from two WOCE sections that are primarily off the continental shelf, and vertical sections of temperature and salinity across four specific sections of the Antarctic continental shelf.  The authors find that the different reanalyses “agree well” (line 428) with the observed hydrography and circulation in the open ocean, but generally do not represent observed trends in deep water hydrography and do a poor job of representing hydrography on the Antarctic continental shelf.

As a potential user of these products in the Southern Ocean, I’m very happy the authors decided to do this study and I think this will be (at least, selfishly, for me) a fantastic addition to the literature.  I certainly think this study is worth the attention of GMD because of how useful it could be for the community.  The manuscript was mostly (see below) clear and easy to understand.

However, I do have several suggestions that I think will help make this study more suited as a guide for potential future users of these products in the Southern Ocean

1)  There really are not any quantitative comparisons.  I didn’t see any computations of RMSE, correlations, skill metrics, Taylor diagrams, etc.  I realize that quantitative comparisons can be painful to create with different grids and will not be appropriate for every comparison.  However, they would be very useful for trying to judge what particular reanalysis gets closest to observations for a specific measurement that might be of most interest to a potential future user (e.g. RMSE for the S4P line for someone who is interesting in the ocean off West Antarctica) beyond just trying to guess from a visual inspection.

2)  Some potential users of these reanalyses are not just physical oceanographers and I think a very useful comparison for those interested in biogeochemical uses (although physical oceanographers will be interested as well) would be mixed layer depth characteristics.  

3)  The plan views at one depth and the two WOCE cross section are quite useful for studying the open ocean time-mean hydrography, but are they enough to show that the hydrography in the reanalyses “are consistent with observations” (lines 5-6)?  I know it’s impractical to look at the plan views for several different depths (and the authors have already presented two depths), but maybe some integrated measures can be displayed.  For example, a planar view can be compared of the vertically integrated temperature (or heat content) and salinity (or salt content) over the top 1500 m to that from WOA.

I have several other specific comments and suggestions below, but most of these are minor and should be easily dealt with by the authors.  I think this exercise is a great idea, but I do think it needs some work before it will be fully useful for the intended audience.

Specific comments

Line 18:  I think it would be helpful to label the ACC and (especially) the Ross and Weddell Gyres somewhere.  I don’t know if the best way is to add a schematic as an inset somewhere, label them on one of the stream function or MDT plots, or add a new plot (maybe in the Supplementary Information?).

Lines 33-34:  Would it be useful to list some reasons why a regional model (e.g. higher resolution, tuning for a specific region) might be able to match observations better?

Lines 38-48:  I think the previous study of Uotila et al. (Climate Dynamics, 2019) assessing several ocean reanalyses (including GECCO2) in the polar regions should be mentioned somewhere (maybe not necessarily in this paragraph, but somewhere in the manuscript).

Lines 64-67:  I do not agree with this statement as the Uotila et al. study did do a cross-comparison of several reanalyses (although none of the ones here, unless you count the older version of GECCO) and observations for the Southern Ocean.

Lines 115-122:  No details for the ice observations used in section 3.2 are given here (or anywhere else that I could find).

Lines 125-128:  The differences in the bathymetry over the continental shelf are an important point (as pointed out in section 3.5 by the authors), but it is very difficult to see this in Figure 1.  I suggest an additional figure (here or in the Supplementary Information) with a magnified view of the bathymetry that shows more of the continental shelf differences (perhaps a blow-up of the Amundsen/Bellingshausen sector?).

Line 135:  Again, I did not see any information on the sea ice “observations” used in this section.

Lines 140-141:  Have the authors looked at the correlation between the modeled and observed September sea ice extents?

Lines 147-152:  Are there any observed estimates of sea ice volume used here?  Even though there is a black “Satellite-based estimates” legend at the top of Figure 3, I don’t see any black lines plotted in the figure.

Lines 155-156:  I suggest adding a little more information here about exactly where 550 m is a good depth for examining CDW intrusions to the Antarctic ice shelves.  It certainly is a good depth in the Amundsen-Bellingshausen and (I think) Totten, but it is too deep in some areas (i.e. Ross Sea, Prydz Bay) to show much of open ocean originated MCDW heat heading towards the ice shelves.

Lines 166-167:  I disagree that the “cold and relatively salty surface water mass” in the Weddell Sea is a “feature of all” estimates:  The SSS in the Weddell in B-SOSE (139…the one shown in Fig. 5) is pretty fresh compared to WOA (and all the other reanalyses).

Lines 191-194:  The ACC transport is based on the flux through Drake Passage, but exactly where is the difference computed in the stream functions to get at the model Weddell and Ross Gyre strengths (apologies if I missed it)?

Line 202:  Is there a reference for the statement that SSH observations suggest an acceleration of the WG over 2014-2017?

Lines 243-246:  It is hard to see the cold and fresh layer over the slope at the northwest side of the section in the observations.  Is it worth adding a supplemental figure showing a zoomed-in western section?

Lines 287-288:  Do the authors know of any good references on model drift that could be added here?

Lines 299-301:  Similar to the comment above for lines 243-246, it is hard to see the cold and fresh layer over the slope at the western side of the section in the observations.

Lines 334-336:  Do the authors think there might be other reasons, besides the outdated bathymetry, for the shallow thermoclines (i.e. issues with the atmospheric forcing, how glacial ice melt is added, etc.)?

Line 343:  Is it worth actually computing and comparing the sea ice production rates for the different reanalyses?  There are observational based estimates of coastal polynya production and it might be instructive to see how the reanalyses compare.

Lines 392-407:  What about the lack of HSSW and the poorly simulated sea ice production mentioned earlier in lines 342-343?

Technical corrections

Line 139:  Since this sentence is just about the maximum extent, should “2b-c” be “2b”?

Line 140:  I think “in March” should be added after “simulated model extent”, since the sea ice extent model error is greater than 10^12 m^2 in September (Figure 2b).

Lines 143-144:  Since this sentence is just about the minimum extent, should “2b-c” be “2c”?

Line 165:  I did not see any isohalines in Fig. 5.

Line 181 (as well as lines 192, 193, and Table 2):  I think it would be helpful if the authors gave the standard deviation for all the model transport estimates, especially since ranges or standard deviations are given for some of the observation based estimates.

Line 198:  The ACC seasonality is shown in Fig. 8a, not 7a.

Line 200:  The ACC interannual variability is shown in Fig. 7a, not 8a.

Line 249:  I don’t see a row in Table 3 that would be temperature trend in the Weddell “below 2000 m”.  Do the authors mean below 3000 m?

Line 265:  I don’t see a row in Table 3 that would be temperature trend in the Weddell “between 1000-2000 m”.  Do the authors mean between 1500 and 2000 m?

Lines 311-312:  Table 3 shows the Ross Sea temperature change at 1500-2000m for B-SOSE (iteration 139) to be less than observations.

Line 326:  The acronym DSW has not been described up to this point.

Line 373:  Suggesting adding “previous” before “section”.

Line 439:  Typo, “hydrpgrahic”.

Line 450:  First, thanks to the authors for making all these data and scripts available.  I may have missed it, but the link in this line (Nakayama, 2024d) only seemed to include SOSE iteration 105 (didn’t see 139).

Figure 2c, y-axis label:  If there is not enough room to have “March Sea Ice extent”, then I suggest changing “March Sea extent” to “March Ice extent”.

Figure 4 caption:  Several of the letter references are incorrect.  For example, at the beginning “a,g,l” should be “a,f,k”.

Figure 5 caption:  I don’t think the description of the red arrows (“point to the salinity bias”) is sufficient.  I assume they are pointing to where the largest input of freshwater from Hammond and Jones is introduced and ECCO LLC270 and B-SOSE are the two models that use both the ice shelf melt rate and iceberg calving.

Figure 9 caption:  Suggest adding something like “See text for the meaning of the red and blue ellipses in (i)” (or add a description in the figure caption).

Figure 11 caption:  Suggest changing “blue dots” to “blue dot”.

Figure 14 caption:  Suggest changing “cyan dots” to “cyan dot” (or perhaps “blue dot” since both dots in Figure 1 are the same color).

Figure 15 caption:  Suggest changing “compiled available all years” to “compiled all available years”.

Table 3:  The figure caption says these are trends “at 500m and 1000 m”, but none of the depths given in each row fit that description (all are either 1500-2000m or 3000m-bottom).

Good paper that will be heavily cited. I have one somewhat-major suggestion to make it even more useful to the overall Southern Ocean community, and a series of minor ones that are mostly esthetic.

In the following, the ocean reanalyses and state estimates that are the topic of this paper are referred to as “ORAs”.

The one major comment: Show us the mixed layer depth

ORAs are too often used instead of observations for both physics and BGC studies when one wants to look at the time evolution of their favourite process (e.g. carbon cycle, primary production). Sure, I am biased in what I read, but this process often relies on a realistic representation of the mixed layers. Besides, it is surprising that you comment on the Weddell Sea hydrography and gyre strength and lament at the representation of both AABW and shelf processes and yet do not check that the mixed layers there are not-too-spurious.

I would recommend either a table listing for each ORA, for summer and winter separately, ideally for the Weddell and Ross seas separately, the minimum – median – maximum mixed layer depths, or at least a map of the temporal maximum mixed layer depth for the entire Southern Ocean. The time evolution (for each sea) of the spatial maximum in a style similar to Fig 2 would also be very useful.

Minor comments, in order of appearance:

The text contains many typos, starting with the title: ocean reanalysEs, not reanalysIs; and affiliation 5: InsTitute (T missing). I trust a careful reading of the text by the authors and then the copy editors will fix the others.

Table 1: Add which product used which atmosphere (see later comment), and the year of the bathymetry, since you use this year disparity to explain some of the differences later in the text.

In section 2, the methods used to calculate the sea ice extent and volume are not defined. In particular, was a threshold in sea ice concentration used, and if so, which? There is also no information about the sea ice observation source aside from a reference to “satellites” in the figure caption.

 Figure 2 onwards: I appreciate that your colour scheme was consistent across figures. I would however recommend that you modify it to increase readability. Red and magenta are hard to distinguish – swapping red for a yellowish orange should work better, and make it compatible with red/green colour-blindness. Dark blue and black are also hard to distinguish – swapping dark blue for a lighter blue such as rgb(0, 150, 255) should work while still remaining distinct enough from green.

Figure 3: The observational line is missing. You should have data since the early 2000s at least (Envisat)

Line 195: I acknowledge that the main objective of the paper is to describe the performance of the ORAs, not to explain their biases. Yet here and there, you do, relating the biases to bathymetry or shelf processes. This line / entire section presents an opportunity for another one: GECCO3 is systematically weaker than the other ORAs, and it also is the only one that uses NCEP rather than ERA5. Can there be an explanation in the NCEP winds?

Figures 9-10 + 12-13 and Table 3: For panels b,d,f,h,j and the trends, can you please compare them on a similar time period, or at least a similar time span? There could be interannual variability masking or enhancing the trends that you show. Or filter all the data to make sure that you really are showing trends, but that is more effort and not feasible anyway for the short ORAs.

Figures 11 and 14: The observational row is missing. You have the data since you show them on Figs 9-10 and 12-13

Section 3.5: I have no more comment, but wanted to say that I really liked this section.