R E V I E W

of the manuscript (egusphere-2023-106) entitled

“Technical note: Determining Arctic Ocean cold halocline and cold halostad layer depths based on vertical stability”  by E. P. Metzner and M. Salzmann

The manuscript proposes a new method for estimating the depth of the Cold Halocline Layer (CHL). The topic is important and warrants a lot of attention in the published paper. Thus, I am very positive that the authors have the potential to produce a nice publication.

However, at the current stage, the manuscript suffers from several major shortcomings:

• I did not find a convincing and satisfactory comparison of the three methods defining the halocline. The majority of the materials presented in the manuscript are about the illustration of the application of the methods and not their comparison. This comparison should include an evaluation of each method's performance and an explanation of the benefits of using each method. Right now, the way it is done is not satisfactory. I would like to see, for example, individual temperature (T) and salinity (S) profiles where the authors show what each method provides and explain the physical reasons for that. I found Fig. 3, which is devoted to method comparison, to be hard to read and not informative.  
• It looks like the authors misinterpret the water structure of the Arctic Ocean, which has direct implications for their comparative analysis of the three methods used for the definition of the halocline base. In the Eurasian Basin, CHL is the layer where T is close to the freezing point (that is why it is called cold) and S rapidly increases. However, below the CHL, there is the second portion of halocline—the lower halocline water—in which T and S increase with depth. The authors should define what they investigate. In the Canadian Basin, this structure is further complicated by the presence of two other halocline water varieties: Pacific summer and winter waters. Showing the Nordic Seas and Siberian shelves should be excluded from their analysis, for example.
• By the way, it looks to me like the authors miss two golden opportunities to give a beautiful explanation to their methods. a) Fig. 2 from a single ITP record shows a mismatch of results between the authors’ stability method (ST) and the density ratio (DR) method after winter convection. I may be wrong, but it is worth checking whether the ST method captures the beginning of the new halocline formation, which is not captured by the two other methods. The halocline formation is a very important topic of Arctic oceanography. b) When the authors analyzed halostad, a possible interpretation may be that they analyzed the boundaries of a variety (winter or summer) of Pacific water. It is worth checking, and if it is true, this may be a nice touch to "sell" the method.
• The way materials presented is often not good enough – see my detailed comments provided to the authors.

Despite these deficiencies, I believe that the authors can improve the manuscript to the level suitable for publications. That is why I give a “major revision” to this manuscript.

Below are my specific comments.

Comments:

1. Line 16: As defined, it is not CHL, but CHL+Lower Halocline Water in the Eurasian Basin.
2. Line 17: 300m may be correct for the Canadian Basin but not for the Eurasian Basin.
3. Line 43: The halocline has a complex structure indeed, but this was first described long time ago – see papers by e.g. Rudels, Aagaard, Carmack, Steele.
4. Line 57: Same as for line 43, plus add here Pacific waters.
5. Data description.

• Please provide vertical resolution of original data, time covered by observations, show distribution of data coverage in time and space for annual and seasonal coverage, show separately spatiotemporal coverage provided by ITP and other sources. Provide accuracy of observations.
• I do not understand the need of this complex vertical data interpolation (lines 89-90). I suspect that the original (raw) data have 1 or 2m resolution which should be sufficient for the purposes of the study.
• Exclude all points from the shelves where there is no typically halocline found. The same is true for the Nordic Seas.
• Regions are shown in Fig 1 but not used. What is the purpose for that?

1. Line 101: Why T is given in K, not in C?
2. Line 113: Shallower, not smaller.
3. Section 2.2.4 and further discussion: All these methods should be illustrated using individual profiles where everything is clearly marked and visible. Same section: You may use definition of the upper Atlantic Water layer by using 0^oC isotherm.
4. Section 2.2.5: The authors may want to clearly define halostad by giving some physical interpretation for this layer.
5. Line 144: Taking 0.001m threshold seems misleading since the original data have a 1m vertical resolution (or even coarser).
6. Section 2.3. I found the use of kriging more misleading than helpful in this study.Actually there is no need in that at all since the authors used another much simpler method which serves for the purposes.
7. Line 168: Filtration of outliers using such a severe threshold of 25/75% (which is less than 1.5 standard deviation) needs an explanation. Sensitivity study where the authors show how sensitive their estimates to different thresholds may be helpful.
8. Line 198: An example where the authors misinterpret water masses: I think they found that the CHL disappeared but the lower halocline water (not AW) is below the SML.
9. Line 202: Please provide specific for the cases when the criteria were not met: what criteria, why, etc.
10. Line 204: I did not understand what is written there.
11. Fig 2: I think this, plus Fig 7, is a good plot to be used for the method interpretation. I suggest to move line definition from the panel - the authors have space below the map. This case would be nice to illustrate further by using individual profiles of T & S for different regimes.

Similar plot for the Canadian Basin, plus individual profiles from there, would be a good illustration for regional halocline differences.

1. 3. Not a good figure. I would completely eliminate it since I found no information in the current version.  
2. Section 3.2. What does “occurrence” mean in this context?
3. Fig 4: Please eliminate point in the Nordic Seas and over the shelves. I suggest to eliminate panels with kriging. The point here is not to show the differences between spatial interpolation methods, but between halocline definition. So, I suggest to show additional panels for differences between methods: e.g. ST-DR and ST-TD. Again, please explain “occurrence”. I did not understand the ”nearest-neighbor” method: If the nearest data point to the grid point is used, why there was averaging then?
4. 5: The same as for Fig 4: Please eliminate estimates from kriging and compare methods of halocline definition and not the methods of spatial interpolation. I.e. show CHL-DR minus CHL-ST and CHL-TD minus CHL_ST.
5. Lines 245-247: Often, some analysis of SML is given (like in this paragraph). This is not the topic of the study, first. SML parameters may be given, if they serve the purpose of illustrating the methods. Otherwise, please skip these places. As they are now, they raise questions about newness of these results (e.g. asking for comparison with the previous papers on the subject).
6. Fig 6: The caption is not clear – the difference is not defined. “(e) to (f)” is not clear. Cos fit is not illustrated.
7. Fig 7. Is good, but a) add T &S (like in Fig 2) b) move line definition from the panel.
8. Fig 8: What is the point of giving fraction of grid points? What is halostad depth? Give a physical interpretation (Pacific water?).
9. Line 314: What is “semi-saline”?
10. Line 321: How can we see the point made for the Barents Sea?
11. Line 322: This discussion and comparison with Steele work needs much more thourough analysis – not just a single line.
12. Line 325 – What does this “lack: mean?

Review on the manuscript EGUSPHERE-2023-106

Summary:

The present study proposes a new method for the detection of the cold halocline layer base depth in the Arctic Ocean. The authors define a new criterion based on vertical stability and compare the results to those obtained with two previously used methods: one based on the density ratio, and one based on temperature differences. Using the ITP and UDASH databases, they derive pan-Arctic maps of the cold halocline layer base depth using all three methods.

General assessment:

The topic is well within the scope of Ocean Science and is of particular importance given the complex and varying structure of the Arctic halocline throughout basins. However, several crucial points would need to be addressed, regarding the manuscript organization as well as the clarity of the method and completeness of the results. Below are listed my major and minor comments. For these reasons, I recommend that the manuscript undergo major revisions before being potentially suitable for publication.

Major comments:

• The manuscript lacks a clear, physical explanation of what the CHL and cold halostad are, and how the proposed method detects their boundaries. I would also suggest the authors emphasize the goal aimed to be achieved by defining a new criterion, e.g., enabling the robust detection of the CHL base depth across basins and seasons using only one criterion.
• It is my understanding that here, the authors equate the CHL base to the point of maximum stability. Wouldn’t the point of maximum stability rather be within the “cline” you are considering, and not necessarily at its base? If I misunderstood, I would suggest the authors clarify their method and the physical reasoning behind it, as stated above.
• The manuscript still lacks an actual evaluation of the performance of each of the detection methods presented here. I like the large scales comparisons, but it lacks some quantitative estimates (which ideally would take into account the varying seasons and basins) and some idea of which method performs best. Because of the diversity of situations, different tests may lead to different rankings of the presented methods. And that is fine, as long as these various results are explicitly presented and discussed.
• The organization of the manuscript lacks fluid connections, both in the introduction and in the presentation of the results. Subsections seem to be organized thematically but without a clear logical order. I would suggest the authors consider reorganizing the overall manuscript (introduction and results) and the abstract as follows:

1) Clear introduction of what SML, CHL, and cold halostad are, and why they matter (as already done in part for the CHL).

2) Presenting the previous methods that have been used to define these layers, and underlining eventual (knowledge) gaps in these methods.

3) Introducing the new method, the physical reasoning behind its development, and the goal it aims to achieve.

4) Demonstrating how the results of the new method compare to results from previous methods (qualitatively and quantitatively, by adding some basin-wise statistics for example), and what are the gains of the new method.

Minor comments:

In the introduction in general:

It would be good to improve logical connections between paragraphs. For example, L26-27: what is the connection between Atlantification discussions and CHL characteristics? You could finish the previous paragraph by commenting on possible changes in the strength of stratification within the CHL that could either boost or hinder further Atlantification.

In general, I would suggest reshaping your introduction as follows (as stated above):

- Defining the broad concepts of cold halocline layer and cold halostad

- Underlining their importance for our understanding of the present and future Arctic ocean characteristics

- Listing criteria for their definition used so far, and eventually pros and cons

- Explaining clearly the motivation behind creating a new set of criteria, and the problem you aim to address

- Finally, introduce the organization of the manuscript (as you already do)

L22-26: Remind the readers in 1-2 sentences what Atlantification is. In fact, you partly do so in the following sentence, but this should come earlier. And isn’t ice loss a symptom/characteristic of Atlantification, too?

L38-39: Here, and generally throughout the manuscript, please define concepts as early as they appear. This sentence should come as you mention the density ratio, 2 lines above. In short, put the sentence line 37-38 at the end of this paragraph.

L34-52: It would be best to reshape this series of 3 paragraphs into one, listing all existing used criteria to define the base of the CHL and eventually their pro and cons.

L57: You already introduced this concept above when citing Bertosio (2020) and (2022). Would be best to merge the descriptions of upper and lower CHL into one paragraph here, keeping the relevant citations.

L75: “Level III data”: This processing level naming convention is rather opaque for unfamiliar readers. State more clearly what this entails (visual inspection, vertical interpolation, salinity spikes or bias corrections… etc)

L91: “Gaussian filter”: as above, please introduce each processing method explicitly.

L89: Is this reasonable for all profiles? Could some profiles, especially the oldest ones, have a vertical sampling of 5 to 10 m? If that is the case, please state so and briefly discuss why you think such a high-resolution interpolation is appropriate and reasonable. I would also suggest the authors consider a vertical resolution that is less fine and closer to the native profiles’ vertical resolutions.

L139: it seems this sentence has grammar issues.

L141: …“only the lowest of these layers is identified as a cold halostad.“ Why is that, physically?

L169: 25^th/75^th percentile: This is an extremely stringent test. Can you explain why you took such a high threshold? As the other reviewer stated, it would be good to know how sensitive your results are to the “outlier” threshold you chose.

L254: In general, I would try to limit references to previous specific figures from previous papers. It is easier if you directly remind the readers what were the findings shown that figure through text, citing the source paper, in discussing your results in light of it.

L314: If you mean “comparatively low-salinity” then use “low salinity”.

Fig. 2 and 7: I like these figures, and it would be great to have one of such plots for -when possible- each subregion and season. Even on one ITP, you could for example indicate when the buoys are in Canada vs. Makarov vs. Amundsen / Nansen basins.

Fig. 3: This is an interesting visualization, but it is rather under-used in the manuscript. I would suggest the author consider replacing it with plots presenting the vertical profiles, either in the introduction to present your various concepts (SML, upper and lower CHL, cold halostad), or in your results by grouping profiles in similar regions or seasons.

Fig. 4 and 5: I still do not quite get the goal of using 2 interpolation methods (NN and kriging). If there is no other objective that showing the pan-Arctic results in 2 different ways, I would suggest the authors pick one of these methods and eliminate the other, in order to make the manuscript more fluid.