The paper outlines an easily accessible and quality-controlled product from multiple sources. It is explained well where data comes from and why there is a need for this data product in addition to the national database of the Canadian Integrated Ocean Observing System.

With the impact of decreasing ice sheets of Greenland getting more and more important, it is crucial to compile such datasets to build a strong community effort around it. 

One of the purposes described are using it to review the changes in the ocean climate of Atlantic Canada. The paper conducts an initial analysis and finds that there are strong changes present in the Northwest Atlantic Ocean including decadal fluctuations. The authors make it clear that this is not an in-depth study of the area. The analysis done here is well suited for preliminary results. The analysis demonstrates that the dataset has potential.

This dataset also incorporates an older database, Climate, which was put offline in 2010 – through this dataset, they make the data accessible again. Not all data were quality controlled beforehand and the authors performed a separate quality controlled which is well described. This adds further value even if some of the data is available elsewhere. Other quality control algorithms are correctly cited and described where possible. 

The manuscript is also transparent about known issues, like lack of metadata or no negative sign for temperature measurements. The format of yearly NetCDF files binned by pressure is well chosen as a easily digestable format.

I only have minor comments listed below.

L39 Are there any considerations when comparing century old data to more recent one? I am aware that for each individual datasets you discuss pre-screening and data issues, but perhaps it would be interesting to mention changing data quality here.

L120 What an annoying issue to encounter! Was this an issue with these specific instruments or measuring protocols? Perhaps a little information would reassure someone working with the data that all of these issues have been caught and it’s not a problem for further datasets.

L145 The way the automatic flagging sounds good - but I’m just wondering if there were any additional manual checks after flagging? Perhaps there were too many flagged profiles, but in that case it would also be interesting to hear how much was flagged.

L299 Are the headers of the resulting product the same as in this step?

L303 Great to provide this example command and also the example header above. Perhaps most users don’t need aid with opening data, but I think this is useful for demonstrating how easy to access the dataset is.

L314 You said you expect that the Climate quality control is better – that is probably right, but it would be great to hear more about why you think it’s better or how much worse the other data is

L329 Why exceeding the depth by 200 m - and not strictly that depth? Is this to account for some uncertainties in the GEBCO dataset?

L343 As before, it would be good to hear if manual checks were at some point performed

L360 This overview is useful, but perhaps it can be moved to the beginning of the section, after which the steps are described?

L414 It’s ARGO floats, not ARGOS

L414 Can you clarify if ARGO floats are included in the dataset? I know that you mentioned “profiling floats” before in section 3, but perhaps it would be good to address the ARGO program separately

L575 I mentioned this before, but for data stretching back to 1920 I would like to hear a bit more about differences in data quality or biases

L600 You have several recommendations before you say you’re updating the dataset in the coming years. Are these recommendations simply recommendations for further datasets or are they future plans for CASTS?

Figure 5 You previously mentioned in the text e.g. how many months on average are available in a given dataset that you included. Perhaps it would be interesting to add information about seasonal bias. This could be a map similar to this one, which shows total number of profiles, but instead e.g. average number of months covered in a year or a similar indicator. Maybe this would look too similar to the total number of profiles, however.

Figure 8 I think for the purpose of this plot it is still okay, but note that these rainbow-style colourmaps are not perceptually uniform (see e.g. Thyng et al. 2016 at http://dx.doi.org/10.5670/oceanog.2016.66)

This paper describes a chronologically assembled quality-controled data set of hydrographic (T-S) profile data mostly located along the shelves and continental slopes off maritime Canada (but extending wider). 

The title also mentions the Eastern Arctic. This is a very Canadian-oriented designation for that part of the Canadian Arctic (including Baffin Bay), and I am not sure how much it is used elsewhere. Here, the description/investigation stops at Davies Strait: what is referred to as Canadian Atlantic shelf… (even though the data coverage is 35-80°N and 42-100°W; thus also includes for example Flemish Cap, but not the slope region east of it). Considering the focus and the examples provided to illustrate the products, I am not so sure that 'Eatsern Arctic' should be part of the title.

A large compilation of data is assembled from different sources, with a rather logical selection and validation process. There are however no correction of potential biases  (for example, for MBTs or XBTs), and some data types are skipped (mooring, gliders, and possibly seal-borne probes, but that is not clear for the latest with subsets are included). THe effort is very valuable and worth been published in ESSD. However, I still have some concerns:

- the presentation for the data is as time series mostly from a few best sampled stations. This is restrictive, with thus a focus onthe temporal sampling, but not the spatial sampling. Although the finally-poduced data set is indeed chronologically sorted, what this brings for other parts of the whole domain is not really discussed. However, ranges of variability (or standard deviations) are used in a spatial way to select/vlidate the data, which are not presented.They should at least be included in supplementary material (I could not open the Supplementary mterial, which seems only to accept Winzip as a tool, and not other tools)

- Furthermore, the different times series presented  (with a clear focus on the annual (or seasonal) means (due to the sampling and natural higher frequency variability, even after removing an average seasonal cycle)) are not associated with uncertainty estimates. There is one attempt to create an index based on spatially-distributed data (the CIL on the GB), but it is not clear from the presentation how it is estimated (how the spatial information is taken into account, for instance and at which level of the averaging...). The huge area over which it is estimated (extending right over the Flemish cap to the east and to Newfoundland and station 27 to the west) is not clear. I wondered reading it whether station 27 would not include a strong percentage of the winter data. Is there an assumption that it has no spatial structure (at least statistically). But how true is it over such a large area (and what about CIL area indices?)

Finally, the choice of the time series is at times a little bit peculiar, and it is not completely clear what the criteria were for selecting these illustrating examples (in section 6) (see specific comments in the detailed comments).

Detailed comments:

Introduction: l. 21, connection of sea ice loss in the Arctic and Greenland Ice sheet loss to the freshening in the NASPG rather controversial…  (accumulation for example recently of liquid freshwater in the Arctic after sea ice melt suggesting that net budget of the export of fresh water (either liquid or in sea ice) is not that straightforward…)

The later mention on that page of the confluence of water masses for the large decadal or longer time scale variability in this region right, but on coastal areas (and for T) also with large continental influences.

l. 85: the collection from WOD is only for station data and MBT. Interestingly, not XBTs or pinnipeds. Is it that they are kown to be better validated elsewhere? What about gliders (surveys known off Halifax or south-east Newfoundland) (I see it mentioned in 3.3, with the mention that they were removed; data not calibrated?)?

l. 88: add year of the occupation of the Challenger Halifax sections (I guess 1873). L. 89 ‘missing years between 1873 and 1910’ is it ‘all years’ or ‘some of the years’

l. 95: for the NAFC database, among other countries mentioned, would the UK have been doing surveys in the area of Newfoundland-Labrador been a dominion (1909-1949)

l.. 279: strange to use keyword CTD for data in 1912-2023 (and title of 3.8 is supposed to be 2002-2023, as also explained l. 282)

In tests, wondering how many profiles removed by the 4.2 test. The test on T-S profiles with 0 standard deviation if more than 3 measurements is a bit extreme. Depending on reported resolution, season, range of depth, this could be physical

4.4 the depth issue. There was a published paper using seal data off south-east Greenland showing that the maximum depths reached by the seals were some times larger than the bottom depth reported in GEBCO, indicating small scale errors in GEBCO (but I don’t recall whether this was exceeding 200m, and possibly bathymetry much better on Labrador – Newfoundland shelves (and certainly better in the southern part of the domain!). Are these errors of depth of cast mostly associated with XBTs, or also station/CTD casts (in which case a position error is likely, such as sometimes happens if data reported initially in degrees, minutes and not decimal degrees. With XBTs, it is unfortunately common that cut of profile when reaching the bottom is not correctly applied (I am sorry I am not able to find the reference)

Here too, wondering what is the number of profiles removed by the check and which type.

l. 414: ‘such as Argo’ (not ‘ARGOS’)

For the illustration of time series, l. 436: why use the CIL Minimum temperature at Rimouski station, instead of CIL thickness layer (as is used for the Grand Banks surveys?). I am also wondering about the choice of showing the vertically averaged T and S. Maybe this should be argued a little. Is there a little bit more time coverage if the focus was on specific layers and not the vertically averaged (for example, near surface layer maybe a little more sampled at some of the sites?). Is the variability in the different layers well correlated? (well, this is discussed a little through the upper layer stratification (0-50m), but there does it make much senses to present its annual average, as it presents a very large seasonal cycle; should the focus be more seasonal for that parameter?)

Also, in the explanation of Fig. 8 on lines 441-445, I am not exactly sure of what is done, as the figure 8 suggests that this is the spatial average, corresponding to the box on Fig. 1. However, there is some spatial variability through the region, and how is it is taken into account. Is the time series analysis done first at each grid point and then averaged spatially, or is the assumption that there is no spatial structure of the CIL depth through this large region. If so, what justifies it?

A bit difficult in section 6 to follow the figures. They are all cited at the beginning of section-, but after discussed sequentially.

l. 450: station 27 representative of the NW Atlantic as a whole. To me rather far-fetched. Maybe OK for the shelf conditions, but not so close with NW Atlantic off-shelf, even rather close by…

l. 468: ‘instead of melting ice from the Arctic…’, could be more ‘intensified export of freshwater’(freshwater can be transported either in liquid water or sea ice; with possible contribution of increased melt of continental ice’, as the station is very close to shore, and could receive melt from Baffin Island or further north, as well as Hudson Bay; I am less sure for Greenland)

l. 470: the pre-requisite of that statement is that there is near-surface intensification of these T and S signals. At that point, we do not know whether this is the case. I suggest to use ‘could’ instead of ‘should’. Indeed, the results presented are not that coherent (see the weak stratification in the late 2010s).

l. 505: I was also wondering whether there was a tendency to have a correlation between the Prince 5 and Halifax2 salinity anomaly time series. At least, it is the impression given by the figures (for the most recent period).

l. 521: ‘correlated with seasonally averaged St. Lawrence runoff’… (this is not shown, but I wondered whether the results cited from other papers are with the time series plotted: is it correct?)

l. 533: how is it distinctly different from other local water masses, such as the continental slope water (I assume east of the Banks?). To which extent is it shelf winter water, maybe formed a little upstream and advected to the area sampled on the Grand Banks and its vicinity (the box plotted on Fig. 1 is rather large) (to some extent, this is commented after on l. 536, and even after on l. 540-541… and I suggest that all these comments and explanations on the CIL be grouped together.

l. 538: I assume that it should be ‘… a strong seasonal signal, with temperature in the upper 50 m climatologically larger than 7°C…’

l. 558: the end of the sentence ’especially during… at shallower depths’. What does it refer to in the sentence? Maybe a separate sentence, indicating that the shallower depths have been more systematically sampled in recent years than earlier.

l. 566: I would skip ‘, whether through ice melt or river runoff’, which I find restrictive for the conclusions of Florindo-Lopez et al (2020) paper. Off course, this paper deals with interesting surveys of the Seal Island section, further north, and in particular the Coastal current near southern Labrador. It is not clear to me that this is the water mass most sampled at station 27, which I suspect also has a large inflow of the Baffin Bay branch.

On Fig. 1 box ‘Grand Banks’, goes quite far east on Flemish Pass, but does not cover the Grand Banks further south… I guess the choice is due to data cover.