Summary

This manuscript discusses several indicators of the position of the Gulf Stream and finds that those based on following a specific isotherm at a specific depth (i.e., 15ºC at 200 m or 12ºC at 400) become biased northward relative to the maximum velocity core in models under warming scenarios. This result is perhaps a bit obvious, but it is worthwhile to make this point explicitly as well as quantify the size of the bias. The manuscript is clearly written and thoroughly documented. I am not convinced, however, that the location of the maximum velocity core is the best indicator for the “true” latitude of the Gulf Stream (see comment 1) and the statistical tests are not necessarily appropriate to the hypotheses being tested (see comment 2). The second comment should be straightforward to address and it is possible the first can be addressed by providing additional motivation in the introduction.

Specific Comments

1. The authors motivate interest in knowing the latitude of the Gulf Stream by invoking the Gulf Stream’s connection to the AMOC and the Gulf Stream’s impact on ecosystems. They then adopt the location of the maximum surface velocity (i.e., the SSH front at the surface) as the “true” location of the Gulf Stream, but it is not clear that the location of the SSH front is relevant to the AMOC or ecology. Biology, in particular, is much more sensitive to temperature than SSH or current speed, so the location of the Gulf Stream temperature front (at various depths) is likely more relevant to biology and regional climate than the location of the SSH front. Chi et al. (2019) showed that the two fronts have distinct spatiotemporal variability—especially downstream of the New England Seamounts—such that motion of one is not necessarily correlated with motion of the other. It is likely that the 12 & 15ºC isotherms also become biased north of the location of the temperature front under warming scenarios, but the size of the bias is likely to be different than the bias between these isotherms and the location of the SSH front. In light of these facts, I suggest that the authors provide stronger motivation for their interest in the location of the maximum velocity at the surface.
2. While both the changes and slopes of the various indicators are all different from each other, it is not clear that these differences are statistically significant. Tables S1 and S2 give p-values for the slopes, but these only indicate that the slopes are significantly different from zero. Since the trends are being compared to each other, rather than zero, it would be more appropriate to give confidence intervals for the slopes so that the reader can assess whether or not the confidence intervals overlap.

General comments

In the manuscript by Garcia-Suarez and collaborators, the authors test the hypothesis of whether isotherm base criteria for the location of the Gulf Stream are appropriate in the context of a warming ocean. To do this they use two simulations, using two different model architectures, forced by two different scenarios. Their results suggest that widespread warming in the slope associated with the climate change signal can be erroneously interpreted as a northward shift in the current axis. The manuscript is well written and the language is clear. The overall framing of what current and future changes in the world’s western boundary current systems tell us about the overturning circulation and continental shelf environment is a very relevant one and it would be of interest to a large part of the community. However, I’m afraid that the study lacks both the novelty and rigour that are expected for a peer reviewed scientific publication. The question that is posed by this study, presented as their hypothesis in lines 50-51, is already addressed in Todd and Ren 2023, which the authors already cite in the manuscript. Todd and Ren clearly articulate how to decompose temperature signals to isolate warming from axis shift based in the observations. A similar approach could have been taken here to isolate warming from shift. With regards to experimental design, while I value the authors’ effort to demonstrate that their results hold across multiple models, the different model architecture and different forcings lead to very different GS behaviours, and the manuscript presents no evidence that the either model is able to reproduce variability. Unfortunately, I believe the work required to bring this manuscript into acceptable form is beyond the scope of the review process, and as such I have to recommend against publication. 

Specific comments

(1) The paper goes back and forth between the concept of the GS as a driver of the changes in the Slope (both introduction and discussion) and changes in the slope explaining the warming in the northern flank of the GS (the explanation for the mismatch between velocity and thermal proxies of the GS). I think there’s a real opportunity for the authors to address this question, but a more systematic approach to decomposing changes in temperature and circulation in the western North Atlantic would be needed. 

As a starting point considering the actual spatial patterns in warming may already reveal some interesting information. At the moment because the Slope temperature is calculated as an average over the region relative to the mean GS position in 2015, and figure 4 shows that at least in CM2.6 there is a non negligible trend in the actual GS path (both blue and black lines in panels c and d confirm, and the magnitude of the temperature trends over the entire period in FOCI are much less clear due to the nature of the forcing post 2040 I suspect), some of the warming attributed to the Slope is likely a real shift in the GS axis. 

In the discussion of the warming in the slope as an explanation for the apparent shift in the GS inferred by T15, the authors show evidence that the slope is warming at an accelerated rate compared to the subtropical ocean (panels c/d versus e/f). If this is the case, the thermal gradient across the GS must also be increasing. For the historical period, because a big portion of the observed temperature changes are salinity compensated, as discussed in Todd and Ren (2023), they have no impact in the current structure. Using the heave/spice framework as presented could help distinguish which part of the warming is a driver and which is an impact. This framework along with the vertical structure of the temperature anomalies shown in Figure 5 offers some likely candidates for what are the actual drivers of the warming in the continental slope and shelf (lines 156-161) in these models. 

(2) Amongst the key differences between models that make it hard to assess whether these models capture the real behaviour of the Western North Atlantic are the following. 

Firstly, in the assessment of the mean GS path (Figure 1) in CM2-6 the SSH and maximum velocity criteria show important differences (with the sign of the difference varying across longitudes), this is compared to both observations and FOCI for which the two criteria tend to agree to the west of the Grand Banks. Secondly in the discussion about the Slope supported by figure 5, in CM2.6 the slope continuously cools down for most of the control run, while it oscillates around 0 in FOCI. The magnitude of the warming for different layers is also quite different between models, with both the 200 and 450m levels warming faster than the surface in FOCI, but only the 450m layer warming faster than the surface in CM2.6 according to Figure 5. 

(3) Lastly, as I final note, the North Wall is not expected to match the maximum velocity core (as stated in lines 145-146), but instead represents the northern flank of the current. As such it is expected to be located to the north of the GS core even in equilibrium. The question is whether the two metrics vary in a coherent way. Figure 4 suggests that within the eastern part of the domain, where the Gulf Stream is more coherent (A1 domain according to the labelling), a deeper isotherm could be used to track the evolution of the current core quite accurately in both models. A shallower isotherm, such as T15 does overestimate the evolution based on a simple linear fit. However, in close inspection, for FOCI (A1), most of the difference between T15 and T12 occurs at the beginning of the record, prior to 2024, while they approximately track each other from that point onwards. As I said, I think this work presents some very interesting preliminary results that the authors could build on in a future study.