The manuscript from Belgacem et al. is setting out to produce a data product focusing on consistent ocean oxygen values in the western Mediterranean Sea, with some important restrictions, such as only considering data collected on Italian vessels, and only considering the oxygen data from the CTD sonde. It is a well needed attempt to move toward a consistent set of ocean data, fit to determine variability and trends in oxygen in this particular area.

The abstract promises “The quality assurance process involves calibration of CTD measurements against Winkler analyses and the comparison of deep observations with reference datasets, using the crossover analysis”. However, I see only weak connections in the manuscript between Winkler and sonde (CTD) data. I recommend making the

The supplementary information is very short, only 2 tables. These are of interest and I suggest including this information in the main manuscript rather than having a SI. Maybe the information from table S1 could be included in tables 1, for instance. There are a lot of details on how the analysis of Winkler and CTD data is done that needs to be discussed in the manuscript. Table S1 has no information on any post-calibration of the CTD sensor: Was a “post-calibration” needed? If so, how was this done, and how large was the post-calibration?

There needs to be a discussion on the difference of this analysis, that uses CTD sensor data, vs. the Winkler data. It took me a while to understand that the manuscript is about the CTDO data, and not about the discrete bottle data, as is the case for CARIMED and GLODAP. Also, and importantly, if there was an adjustment applied to the CTDO data, was the same adjustment applied to the Winkler data for that cruise? And where is this documented, and where are the bottle data for those cruises? Actually, where are the bottle data for all of these cruises? This connection is particularly important for any future effort to make a data product of oxygen data in the Mediterranean Sea that is more universal, but based on bottle data.

The link to the temporary data set did not work, so I could not inspect the product. This in itself is a disqualifier for publication, that can only happen once the data is on a recognized repository.

Based on my assessment this manuscript is likely worthy of publication, but only after a major revision.

Minor issues:

Abstract: Quality assurance is the process of checking the data while doing the measurements, so that you can react and correct any issues you notice. What you have done here is quality control, not quality assurance.

Line 69: profiles

Line 38. Is that the end of the sentence? Add period.

Line 39: Oxygen is not leading to high productivity, but is the result of high (primary) productivity.

Line 41: OMZs are certainly important, but the OMZ in the Med is very weak, i.e. most areas of the (open) Med is well oxygenated. I suggest to note this in the text. OMZs cannot become “more frequent”, as they the term OMZ refer to an area, not an event. OMZ can become larger/smaller, more, or less, intense, but not more/less frequent. Low oxygen events can, however, become more frequent.

Line 69: Profiles

Table 1: Maybe add a column on which oxygen sensor was used, and if there was a post-calibration needed after comparison with Winkler data.

Lines 102-110: This paragraph probably belongs to the introduction rather than method section.

Line 123: A “cruises” too much?

Line 132: Similarly,

Lines 132-135: As Far as I know, the CARIMED data product is not published, or final. How can you be sure that the oxygen has “reference cruise” quality by simply stating that they were part of a non-ready product? Also, did you make a check on consistency between Winkler data and CTD data for the reference cruises as well, noting that CARIMED is only bottle data`

Would there be another, more objective, way of assigning reference cruises, such as using cruises that used CRMs or certified standards?

Lines 181-219, roughly. Most paragraphs is this section is only one sentence long. It makes it awkward to read. Consider modifying. Much of this information is probably better in a table.

Line 227: Why is this an issue for recent cruises?

Line 232: It seems very strange to average each cruise to one single profile. I cannot see how that can work, except for cruises with limited geographic extent. Why not make an average profile for each of your sub-regions instead, that might work out.

Line 232: Ferer to Figure 7a

Line 242: A 4% offset can be related to conversion between volumetric and gravimetric units, a difference that is close to 4%. Not saying that is the case, but could be worth checking.

Lines 247-249: Is it “between 1000-2000 meters”, or “below 1000 meter”? That is not the same.

Line 248: Capital “T”

Line 270. You might refer to Tanhua 2011 here.

Line 269: Section crossovers. I did not understand if you made a cross-over analysis of all cruises vs. each other, or only the cruises in the product vs. the reference cruises. The text is not clear on that.

Table 4 and Figure 8. The table states correction facto 1,039, whereas the offset in figure 8 is 0,957. That does not match. Why? This is confusing.

Section 4: I would strongly suggest to use coherent definitions, and standardized language for the description of the adjustments to each cruise. I would suggest to make sure the ms is consistent in using “offset” – this is the difference between a cruise and reference cruises: “correction” – this is the reciprocal of the offset and indicate what would need to be done to get consistency based on the determined off-sets – “Adjustment”- this is what you did to the cruise based on different lines of evidence.b Reading the text on each cruise, it is not always clear what you did to the cruise due to fluffy language. I would recommend not to to fancy on language, but be consistent and clear.

Line 427: Very fluffy language, and numbers. Why would an offset of 0.97 lead to an adjustment of 2% - it has to be 3%.

Line 443: Very fluffy – “We think an adjustment of …. “. Tell me what you did, not what you think.

There are many many figures to the section describing the individual cruises. Maybe having one or two cruises with many panels would be much easier to read .

Line 457: I do not understand this, what does this mean, and what is the adjustment in the end?

Line 473: An “%” too much?

Please add a table of applied adjustments! Maybe by adding a column to table 5.

Line 505: ? This paragraph do not really belong to the conclusion.

Line 522: The individual data are not a product, remove the word. The adjusted data product is not the “original”, suggest removing the word.

Belgacem and coll. present a compilation of dissolved oxygen profiles obtained in the western Mediterranean. This dataset covers the period 2004 - 2023, although most of the data were collected between 2004 and 2013. The introduction is brief, as are the methods and results sections, while the article presents the data qualification in great detail. It's true that a data paper should explain the data and quality controls, but in the end this version of the submitted article mainly covers this aspect. What is missing, however, is a contextualization of the topic (oxygenation of Mediterranean waters) based on recent articles, a more detailed description of the data set, and a comparison of the data with other data sets (Argo, other campaigns in the Mediterranean).

Many details are given for quality control. But there are almost no details about how the data was collected. It is only written that (line 39) 1382 CTD oxygen profiles, but there is no information about how they were obtained. This is a real problem because to use a dataset, it is important to know the type of sensors used, their accuracy, how they are calibrated and the calibration frequency, especially for marine waters where the values must be very precise. It would also be useful to explain why there have been far fewer measures since 2014, and to indicate whether this will continue. It is not explained also what are reference cruises (figure caption 2) and how they were defined.

The introduction to Dissolved Oxygen (DO) and the Mediterranean Sea is minimalist, with a few generalisations about this element and some very conventional references, and then very few details about DO dynamics in this sea, which is considered a hot spot for climate change. To justify the interest of the database presented, a more precise state of the art is required. Some work has already been published, in particular on the DO minimum, which is strongly determined by the intermittent convection process. There is an important issue related to the decrease in the intensity of deep convection in the northwestern Mediterranean predicted by the end of the century in recent projections, which may have important consequences on the DO concentration. This needs to be highlighted in the article, as it reinforces the interest in making DO datasets available in the Mediterranean Sea. The Mediterranean Sea is not my usual area of study. However, I would like to suggest some recent articles :

Fourrier & al (2022) Impact of Intermittent Convection in the Northwestern Mediterranean Sea on Oxygen Content, Nutrients, and the Carbonate System JOURNAL OF GEOPHYSICAL RESEARCH-OCEANS DOI10.1029/2022JC018615

Ulses e& al (2021) Oxygen budget of the north-western Mediterranean deep-convection region BIOGEOSCIENCES DOI10.5194/bg-18-937-2021

The description of the quality assurance method needs to be improved, the principle of the method is relevant to be described in the article with an example, but the other cases should be placed as a supplementary section or provided as information at the link where the data can be accessed.

Even though it is a data paper, the data need to be better presented, how they compare with other data sets or Argo profiles. The conclusion also needs to be considered, for example, it is underlined (lines 498) that the accuracy of the measurements, but in fact the article does not introduce what data are available for this region.

The text in the article also needs to be improved. For example, it is strange to start an abstract by mentioning the doi of another dataset (lines 16-18 lines). This sentence can appear in the body of the article, but not in the abstract.

In summary, I have very mixed feelings about this paper submitted by Belgacem and coll. On the one hand, it presents an interesting dataset, but at this stage it needs a thorough rewrite. I would suggest rejecting it and encouraging the submission of a more mature version.