The study uses the VIKING20X-JRA-short ocean hindcast simulation to explore mechanisms explaining the anomalously fresh and cold conditions in the subpolar North Atlantic during 2012-2016. Utilizing Lagrangian tracking, the authors show that the freshening/cooling seen at OSNAP-East to a large extent is caused by a higher fraction of Labrador Sea source water due to increased recirculation in the subpolar gyre. The study addresses a timely question, and makes for an interesting contribution in the debate regarding subpolar variability. That being said, I have several questions and comments related to the mechanism proposed and how it relates to previous studies. Moreover, the manuscript is quite dense with results, and being precise with definitions and language will be crucial for getting the main messages across to a reader.

Mechanisms and relationship to related work

1. The introduction highlights particularly the ‘warming hole’. However, as the warming hole is a multi-decadal signal thought to be caused by effects that are not the main focus here (Keli et al. 2020; cloud feedback, low-latitude AMOC), the introduction reads a little misleading. I recommend restructuring so that the 2012-2016 freshening and cooling is the focus, as well as previously proposed mechanisms for interannual to decadal variability (i.e. expand on mechanisms mentioned in l.34-42 and in Section 5).
2. Throughout, it’s unclear to me exactly how the proposed mechanism agree/disagree/link to previous proposed mechanisms in literature. Formulations such as l.575 ‘we find neither of these to be fully convincing’ make it sound like you discard all previous work, but surely your mechanism is not entirely independent of what’s been proposed earlier?  
3. l.314-316, l.322-324 and Fig. 8: You do see signs of an eastward shifted subpolar front, but you seem quite dismissive of horizontal redistribution in the conclusion (l.600-601). In addition to Holliday et al. 2020 and Desbruyres et al. 2021, Kenigson et al. 2020 and Asbjørnsen et al. 2021 are relevant for such horizontal shifts. Asbjørnsen et al. 2021 is also quite similar in terms of the particle tracking approach.
4. Are you seeing the subpolar trend reversal in VIKING20X as documented in Desbruyres et al. 2021?
5. Fig. 8a-b: Eastern and western parts look quite different. Interesting that OSNAP-E-37W-500m total (Fig. 5a) is dominated by what’s happening at 37W-21W. A point that should be noted? The fractional evolution seen for the eastern part is more consistent with what is seen for the Nordic Seas inflow in Asbjørnsen et al. 2021.     
6. l.358-360 and onwards: In any sort of AMOC change discussion, it needs to be clear that there is no consensus on whether the AMOC actually has systematically weakened over the 20^th century until today (e.g., debate summarized in Jackson et al. 2022, Latif et al. 2022).
7. l.445-446: Labrador Coastal Current and ‘the other’ – are both these branches what’s typically called the Labrador Current? I also don’t get the LC-Arctic and LC-Atlantic distinction (l.411-412).
8. You make a convincing case for the freshening/cooling seen at OSNAP-E is due to a higher fraction of Lab. Sea water (Fig. 5, Fig. 7, Table 1) related to longer residence times in the SPG (Fig. 10a). I don’t fully get why more SPG recirculation necessarily must be due to reduced heat loss and deepening isopycnals (l.435-441). Do you reference studies showing this mechanism for the SPG anywhere? Perhaps this point needs to be repeated in the conclusions.

Lagrangian analysis:

1. Are particles released over the full OSNAP-East line as said in l.75-76 or only over the part of OSNAP-East shown in Figure 3? Are you analysing releases over the upper 1000m as said in l. 76 or the upper 500m as stated in l.111. I think the information in l.111-116 needs to come earlier to avoid conflicting messages.
2. l.84-86: A bit unclear. How about explaining it as: ‘Each particle represents a volume transport of 0.001797 Sv. The number of particles released along the 2-dimensional OSNAP-E section is scaled with the model velocity normal to the section at the release time.’
3. l.87-89: What is the reasoning for choosing Parcels and not a tool like Ariane where streamlines are computed analytically?
4. l.89-95: I don’t quite get the type of errors discussed here (Errors induced by temporal resolution? Errors related to lacking diffusion? Errors related to particle release number? Errors related to assumption of stationarity?). What are the 32 subsets?
5. Gulf Stream and Lab. Sea are defined as the two ‘upstream origins’ for OSNAP-E water (l.96). Then you subdivide Lab. Sea ‘origin’ by ‘upstream origin’ again (l.102): Hudson Bay, Davis Strait. Greenland Sea, Denmark Strait, SPG. Later in the manuscript the word ‘source’ is frequently used. I would be very clear with the definitions and use them consistently throughout. Perhaps Lab. Sea and Gulf Stream is your ‘upstream sources’ while the Labrador source region is subdivided by four ‘origin regions’?
6. Figure 3: I’m confused by the lines on the map. Lines over land should be gone? Are particles crossing the orange line cutting across the Lab. Sea defined as having Lab. Sea origin? Is the north-south orange line the Loop path/Slope Sea pathway definition? I get the east-west green line is the Gulf Stream definition, but what about the north-south thin green line? What about showing definitions for Davis, Hudson, Greenland Sea origins in panels f-h, respectively?
7. l.209-212: Such fractions will depend on the depths evaluated over and the source definition used. In Koul et al. 2020, particles are initialised in the upper 100m only. In Asbjørnsen et al. 2021, 26% of the Iceland-Scotland Ridge inflow (upper 1000m or so) has a subpolar or Arctic origin (Davis, Hudson, Denmark straits, or circulation in the SPG), with 42% of the surface inflow being water from Davis Strait and Hudson Bay.

Minor comments / technical corrections:

1. l.80: reference in parenthesis or write ‘shown in Biastoch et al. 2021’.
2. Fig. 1 and Fig. 2 is not referenced in the text until page 9. Either reference earlier or rethink figure-order.
3. l.107-110: Would shorten this point. Doesn’t need to be its own paragraph.
4. Section 2.4: Unless you want to expand on OSNAP measurements and how the EN4 product is produced, I would delete this section and move the sentence somewhere suitable.
5. Section 2.5: Not really necessary information – I recommend deleting the section and put the information about code availability in the data statement at the end.
6. l.216-217: Stated already in section 2.2, not necessary to repeat.
7. Starting sentences with ‘So’ or ’But’ (sometimes) making them not ‘full sentences’ with a subject, verb, and an object: e.g., l.220, l.247, l.255, l.269, l.427, l.465, l.519, l.546.
8. l.220-223: No need to repeat the details here.
9. l.294-305: I would cut this section and save such summarizing statements for the conclusion. If you are worried some of the interesting results might be ‘lost’ because the manuscript is quite dense, you could even do a bullet point list summarizing the main results in the conclusion.
10. l.337-340: check punctuation and parentheses.
11. l.341-344, l.411:413: check parentheses.
12. Section 5.3 title: title sounds like you again will discuss what was addressed in 5.1. I would find an alternative pointing to the Labrador Sea focus.
13. l.414-416: ref. Fig. 10?
14. Fig. 10 and l.414-419: What depth is the ‘light upper layers’ - are you still looking at 0-500m?
15. l.528-532: Too long sentence – difficult to decipher.
16. Legends Fig. 7, Fig. 8, Fig. 10 looks quite messy. I would display text as a structured column. Fig. 8 could have one common legend on the side.

Mentioned literature:

Kenigson et al. 2020 – 10.1175/JPO-D-20-0071.1

Asbjørnsen et al. 2021 –  10.1175/JCLI-D-20-0917.1

Desbruyres et al. 2021 - 10.1038/s43247-021-00120-y

Jackson et al. 2022 -  10.1038/s43017-022-00263-2

Latif et al. 2022 - 10.1038/s41558-022-01342-4

This article sheds new light on the main drivers behind the cooling and freshening recently experienced in the upper eastern subpolar North Atlantic. To this end, the authors analyze the outputs of a historical hindcast (years 1980 to 2019) performed with the eddy-rich ocean–sea-ice model VIKING20X, using a variety of techniques and diagnostics, from lagrangian particle tracking, to water mass transformation analyses. 

The article is well-timed and well written, the methodology applied is sound, figures are clear, and many of the results will be of high interest to the climate community at large, and to anyone with specific interests in the recent changes experienced in the North Atlantic and its surroundings. 

I recommend a minor revision of the manuscript and enclose a list of comments that the authors would need to address to render the article suitable for publication in Ocean Science. 

General Comments:

• Agreement of the VIKING20X-JRA hincast with observations is mentioned several times throughout the text, but it is mostly derived from visual comparisons, which can be misleading. In some cases the agreement is clear, but in other cases is much less evident (see comments #4, #16 and #18 further down). Supporting these statements with some specific metrics, like linear correlations between the hincast and the observations, will help to determine more precisely to what extent they agree with each other.
• Given the comprehensive list of processes and mechanisms that are analyzed in the paper, which in many cases are interconnected, it would be very useful to include at the end of the article a schematic figure summarizing the chain of events that give rise to the freshening and cooling of the eastern subpolar North Atlantic, as supported by the model analysis.
• Several figures from other articles are cited throughout the text, urging the readers to keep jumping from one article to another, and thus hindering the overall readability of the article. Some of those figures include indices that could be easily incorporated in others figures of this manuscript (see comments #6, #10 and #11), making intercomparison between those and the VIKING20X indices much more straightforward. I strongly recommend the authors to include them. 

Specific Comments:

1. [Lines 18-19] This sentence seems incomplete and inaccurate. It should (1) mention that there is a cooling on the eastern subpolar North Atlantic, not a surface warming anomaly, and (2) it should also specify that this is simulated in response to a “weakening” of the ocean circulation. Also, I recommend the authors to avoid the use of the term “predict” in the sentence, as “predictions” generally refer to historical simulations initialized from observations to phase the model with the observed internal variability. However, the warming hole is a feature that consistently appears in uninitialized historical simulations, and is therefore deemed to be mostly externally forced. 
2. [Line 108] It is unclear which “common technique” it refers to.
3. [Figure 4b, caption] Is it “number of days” or “number of years”?
4. [Lines 177-181, Figure 2] While the simulated upper temperature variability in the hindcast shows a good agreement with the observed timeseries from both OSNAP and EN4 products, this is not the case for the simulated salinity. This is particularly clear for EN4, which can be compared for a substantially longer period, showing large discrepancies in terms of both the high and the low-frequency variability. Indeed, a close inspection to Figure 2 reveals that the differences between EN4 and VIKING20X-JRA are not stationary in time, and reflect more than the systematic mean state bias stated in the text. It is true that salinity observations are quite scarce and therefore objective analyses like EN4 are subject to large uncertainties, but it remains to be checked whether the simulated variability is within the range of the observed uncertainty.
5. [Lines 311-313] What do you mean by “contrasting evolution of transit times”? Are their associated histograms (like in Figure 4) substantially different? And in which way?
6. [Lines 334-340] Instead of referring the reader to Figure 10 in Biastoch et al 2021, it would be preferable if the authors included the SPG index in a subpanel of Figure 9, where it can be directly compared with the AMOC indices.
7. [Lines 339-340] The two parentheses referring to Figures 5 and 7 need to be closed.
8. [Lines 363-364, Figure 9] Why did you choose to plot the AMOC at 29°N and not at the same latitude of the RAPID array (i.e. 26°N)? It would be indeed very interesting to include a direct comparison of the hindcast with the RAPID data, to learn more about the model realism in representing dynamical aspects, like the North Atlantic ocean circulation.
9. [Lines 366-368] I would not say that the AMOC is responsible for a reduction in Gulfstream transport. The Gulfstream does not respond to the AMOC, it is a component of the AMOC and as such contributes to its variability, not the other way around.
10. [Lines 371-372] This can be directly shown to the reader if, as suggested in Comment #6, the SPG index is included in Figure 9.
11. [Lines 378-380] Here you compare again with an observed index from another article that could be easily included in this one (in Figure 9). 
12. [Line 386] I do not find this summary statement fully justified. There has been no specific analysis in section 5.2 to rule out the AMOC weakening as a main contributor to the freshening of the eastern subpolar North Atlantic.
13. [Line 394] The parenthesis needs to be closed.
14. [Lines 520-521] It is important to specify that the doubling of the annual mean depth of those isopycnals only occurs in VIKING20X-JRA. In EN4, the increase in the depth of the isopycnals is very subtle and only discernible for σo > 27.65 kg/m3.
15. [Lines 526-528] Figure 14 a-c → Figure 14c (as this is the only panel showing the isopycnals)
16. [Lines 537-539] From Figure 14 it is not possible to say if the model realistically represents the variability in salinity and density. Indeed, as mentioned in Comment #14, the strong deepening of the isopycnals in VIKING20X-JRA (one of the major reasons behind the freshening in the eastern subpolar North Atlantic identified in the paper) is not seen in EN4, and in particular in the upper layers.
17. [Lines 544-545] Can you explain why it would be counter-intuitive?
18. [Lines 589-593] Not all the model results are fully supported by observational data, in particular the deepening of the isopycnals in the Labrador Sea. Furthermore, the evolution of the Labrador Sea vertical structure only compares well with the EN4 data (Figure 14) for temperature. The agreement with the observed salinity and density is much more limited. It would be worth discussing whether, and if yes, in which sense, this poor agreement affects the main findings derived from the VIKING20X-JRA analysis.

In this study, the authors investigated mechanisms of the exceptional freshening event in the eastern subpolar North Atlantic using a high-resolution VIKING20X model run. The authors conducted a thorough study starting with a Lagrangian tracking analysis that leads step by step to the conclusion that the freshening event in the eastern subpolar gyre is due to reduced heat loss in the western subpolar gyre. Overall, I find the manuscript logically organized with convincing conclusions. However, I believe several general concerns need to be addressed before the manuscript can be accepted for publication.

 General comments:

1. This study is based on the high-resolution VIKING20X-JRA-Short hindcast simulation, which is able to capture the great freshening and cooling event in 2014-2016. The pattern and timing of the freshening and cooling from the model is very consistent with those derived from EN4. However, the magnitude of the freshening and cooling in the model is substantially larger than in EN4. Since the hindcast is a free run without data simulation and bias correction, the authors argue that the model serves as a dynamically consistent tool to examine the freshening event.

This leads to my first concern: given the large simulated bias in the magnitude of the freshening/cooling event, the model could be overly sensitive to a certain mechanism (e.g., heat loss in the Labrador Sea) that contribute to the freshening, while it may underestimate other mechanisms (e.g., AMOC). The authors point out that warm and salty bias prior to the 2014-2016 event is a common feature of hindcast simulations. However, this is not a valid argument that we shall expect a stronger freshening in the model than in observation. Can the authors explicitly explain the reason for this much stronger freshening in the model? Do other realizations of the VIKING20X hindcast simulations show a similar freshening event? If the answer is yes, can the same mechanism explain the freshening? In any case, a clearly stated disclaimer is needed in the discussion to remind readers that compared to observations, model bias both prior to and during the freshening event is strong. The proposed mechanism may be subject to model bias.

2. In section 5.2, the role of the AMOC in driving the freshening is discussed. It is concluded that the weakening Gulf Stream source, determined via particle tracking, is associated with the AMOC in the subtropics and is not related to the subtropical gyre circulation. What I find missing here is the AMOC in the subpolar north Atlantic. What is the role of the subpolar AMOC in the great freshening event? Does the subpolar AMOC also weakens around a similar timing in VIKING20X? How is the magnitude and time of the weakening (or maybe strengthening) of the subpolar AMOC in the model compared to the subtropical AMOC? The authors have studied the great freshening event in the subpolar North Atlantic, used OSNAP East section as the termination of the Lagrangian tracking method, and determined that the Labrador Sea heat loss plays a key role in driving the freshening event. However, the authors have avoided investigating the AMOC in the subpolar North Atlantic.

3. I do not find Section 6.3 convincing. First, I do not see how the modeled isopycnal depths “agree closely” with observation. The model has substantially shallower isopycnals, particularly at large depths. This means that the model has a substantially higher density in the Labrador Sea. However, I cannot understand how the model can have fresher and warmer water and at the same time higher density throughout the water column.

Minor comments:

1. I do not think it is necessary to start the introduction with the “warming hole”. It might be more straightforward if you directly start with text describing the recent “freshening and cooling” event.
2. In Figure 2d, why is there direct-path water crossing 60°W.
3. Section 5 need to be reorganized. Both section 5.1 and 5.3 are subpolar-gyre-related mechanisms. And what does it mean by basin-scale in section 5.2? Gyre circulation is also basin-scale.
4. Line 364, how is “subtropical gyre recirculation” defined?
5. Line 377, I do not think there is a consensus on whether the AMOC has declined since the 1990s. Models and proxies suggest that AMOC has declined (e.g., Rahmstorf et al., 2015; Ceasar et al., 2018, 2021), while observation-based reconstructions have not found a significant AMOC decline (Fu et al., 2020; Worthington et al., 2021; Caínzos et al., 2022).
6. Lines 459, please specify the density of the “lighter” waters.
7. Lines 461, it is confusing to call waters lighter than 27.50 kg m^-3 as the “lightest” waters.
8. In Section 6.2, it is concluded that due to reduced heat loss over the Labrador basin, transformation from lighter to denser water mass is weakened. Therefore, the steady inflow in the upper layer (<27.65 kg m^-3) must be balanced by an enhanced outflow also in the upper layer. Does this indicate that the overturning in the Labrador Sea weakens, while gyre circulation in the Labrador sea strengthens? This leads back to Section 5.1, lines 335-340, where it is found that the SPG is not responsible for the freshening. How would the authors reconcile the discrepancy here?
9. 14 needs to be reorganized. Fig. 14(d,e,f) is cited before Fig. 14(a,b,c).
10. Line 536, salinity issue does not make temperature comparison reliable, the sentence needs rephrasing.

Reference:

Rahmstorf, S., et al. (2015). Exceptional twentieth-century slowdown in Atlantic Ocean overturning circulation. Nature Climate Change, 5(5), 475–480

Caesar, L., et al. (2021). Current Atlantic meridional overturning circulation weakest in last millennium. Nature Geoscience, 14(3), 1–120

Caesar, L., et al. (2018). Observed fingerprint of a weakening Atlantic Ocean overturning circulation. Nature, 556(7700), 191–196

Fu, Y., et al. (2020). A stable Atlantic meridional overturning circulation in a changing North Atlantic ocean since the 1990s. Science Advances, 6(48), eabc7836

Worthington, E. L., et al. (2021). A 30-year reconstruction of the Atlantic meridional overturning circulation shows no decline. Ocean Science, 17(1), 285–299

Caínzos, V., et al. (2022). Thirty years of GOSHIP and WOCE data: Atlantic overturning of mass, heat, and freshwater transport. Geophysical Research Letters, 49, e2021GL096527