I would like to thank the authors for responding to my points. This has led to significant changes to the paper, many of which are positive. However, it also led to many wholly new analyses being presented that didn’t quite address the concerns the last time, and the presentation is still confusing in areas, with either insufficient information or statistical rigor to fully follow the arguments. Therefore, although I still find the ideas being presented here to be highly interesting and exciting, I am left feeling that the paper is not publishable in its current state. In particular, I feel as though it just does not do enough to discuss the uncertainties in the analysis, and has too much certainty in its conclusions. I list my major issues below followed by minor points (which may overlap into the more major ones).
Major points
The discussion on subselecting is still confusing - I think this is because there are two steps that are discussed as one and referred to as “subsampling”? For example, there is the choice to split the record into two, i.e., above and below the value of 0.5 for SNAO, and then again where individual years are removed. I understand the first part of this sub-selection, and it seems warranted given the relationship between SST gradients and summer NAO. But I am still unsure of the second, which seems more focused on getting a higher correlation between the two. I still do not understand the physical basis for this, and am worried that it over emphasizes the potential role of this mechanism. Indeed, the authors say they are not worried about the mechanism, but I’m not sure how they can be so certain. I guess in particular, the logic seems not ideal here, as the authors state they are looking for years where SSTs are being driven by freshwater, but then focus on the relationship with the summer NAO. An obvious question is whether the same years come out if you first look for years where you think the SST anomalies are driven by FW changes? This would go some way to better understand the thinking here.
Although there has been some attempt to explain the physical relationships - I’m still uneasy about the explanation for the two regimes of stronger SST gradients from only the freshwater fluxes related to the summer NAO. In the response the authors say that “The threshold of ~ -0.5 in the summer NAO corresponds to a critical surface freshening above which the shallower, seasonal freshwater anomalies are mixed down”, but supply no evidence. I take it that this is an assumption? I agree that this could be one interpretation, but another is that the SST gradient is not being driven by the same processes (i.e., freshwater) in the other cases?
I don’t fully understand the logic behind the mass balance either, and I worry it is missing key processes (and hence overestimating the role of freshwater). In particular, the authors basically assume that geostrophic currents do not contribute to the SST anomalies (and in the response argue that it is because they are in parallel to density contours). This is obviously the case, but how do you take account of weakened or stronger geostrophic currents and, hence, ocean heat transport convergence anomalies? This may be a bit simple minded, but if the NAC weakens and transports less warm water into the eastern subpolar gyre, but the sea water is being subject to the climatological heat loss, then this will cause a cooling. Another process that isn’t taken account for is whether there is a shift in the currents. Confusingingly, such a mechanism is discussed in terms of the ocean response to the summer NAO, and is argued to make key changes to temperature, but the mass balance model (which is used to argue that freshwater changes are the key process) doesn’t take into account these processes, as far as I can tell.
I still remain confused about the proposed mechanism. Moreover, in order to explain how freshwater affects summer weather, the authors invoke changes in ocean circulation (to explain Fe) and wind-driven changes in heatfluxes and Ekman upwelling. They do this using a very small sample (next point) to produce a range of statistical links, but still appear to conclude that it is freshwater anomalies that drive everything with little discussion of the uncertainties inherent to a coupled system (note that the simulations used still do not test even the specific hypothesis that subpolar SSTs are driving European summer temperatures as opposed to SSTs or external forcing more generally). Therefore, I feel the authors need to do a much better job in reflecting the uncertainties in their conclusions. Ultimately, the way this is written is that the authors seem to think that the freshwater (or really the summer NAO) drives everything, but this seems a bit extreme to me based on the evidence. I would expect a significant amount of discussion time on the limitations of the method and on the exact role of freshwater anomalies (e.g., are the responsible for everything (as presented here), or could they be an important and not fully appreciated feedback?)
I’m still very worried about the robustness of the results in a statistical sense. The regressions for Fw are based on, as far as I can tell, 9 sample years(!). Moreover, even if the authors argue that the auto-correlation is small (it doesn’t look that small over the 1980-2020 period), most of the years in their samples are not wholly independent (e.g., they come in clusters), and occur on a background of significant decadal change in atmospheric circulation and European temperatures, which both may be driven by external forcing). There is no attempt to address the particular uncertainty of small samples in the whole paper, which begs the question, how robust are the results regarding European Weather to further sub-sampling? Given the short time-series, I would at least expect some sort of jackknife resampling (i.e., leaving out 1 year at a time), but I would strongly recommend the authors to also compute figures 6 and 7 using non-consecutive years.
Minor points
Line 166 - should be figure 1c? Relatedly, I would recommend to move figure 1c to 1a (e.g., swap the time-series with the spatial maps). This seems more logical.
Line 190 - I had not realized previously that you use a different spatial area to define your SST gradient. This seems slightly odd in that you may be picking up different processes (and related to my previous worries about not understanding why there is such a threshold in the SNAO SST regression). How sensitive is the SST gradient, and hence the correlations shown in figure 1d, to using fixed boxes for SST?
Line 195 - Better to be specific - you mean for values above the -0.5 threshold?
The process of “optimizing” the regression slope is very unclear - how do you do this? Did you just resample by randomly picking 17 years, and then pick the highest value? What is the physical basis for this? You say that results are not sensitive to this choice, which may be true, but then why “optimize” the correlation to 0.9 (from 0.68)? I think it would be good to show the sensitivity more clearly in the appendix.
Line 280 - its not clear to me that figure 4c shows an intensification of the subpolar gyre - if anything, it is a weakening, in the mean (lower ADT in NAC region means weaker NAC, and higher ADT in Eastern subpolar gyre indicates weaker flow?)
Line 283 - what do you mean by subsampling in this case? When you say ADT on Fw, do you really mean it is the ADT when you only use sNAO value > -0.5. Is the field optimized in the same way as the delta SST?
Paragraph from line 299 - I found it very confusing to understand what anomalies were being explained by what. It starts by talking about sNAO anomalies >-0.5, but then talks about both Fe and Fw events.
This part implies that ocean circulation changes control Fw events, however, this bring up two questions. Firstyly, what sort of circulation anomalies are we talking about? In particular, previous work had talked about shifts in the size and location of the subpolar gyre (e.g., Hatun et al, 2005). Is this the sort of circulation anomaly that the authors are thinking of? Second, the argument that circulation changes are important for Fw seems at odds with the argument that advection doesn’t cause the SST anomalies in the derivation of the freshwater signal, where the authors say geostrophic circulation is not important?
Line 358 - I think this is written with far too much certainty ( the SST fronts destabilize the overlying atmosphere, resulting in an enhanced jet stream), and is presented as though it has been proven within this study. I take the point that this is what might be expected from theory, but a regression onto a short time series is not proof. Indeed, the previous section has just been highlighting how wind anomalies can drive SST gradients. Therefore, I would suggest wording here (and elsewhere) that better reflects the uncertainties. E.g., the enhanced jet along the SST front is consistent with…. Ultimately, if I have understood the analysis correctly - this regression is based on 8 years/events - this is an incredibly small sample.
Figure 6 (and other relevant ones) - its not clear what the time period used for this regression - Previously I would have thought it from 1979 onwards using the 8 events - but, now with figure 9, I’m not sure…
On the subject of figure 6, please can you say more about the significance test, please? What is the test, specifically, and how have you ensured that it is robust to a small sample? Also, how do you take account of the fact that many of your years are clustered together? Maybe a jackknife resampling would be appropriate here to test the sensitivity of your results to omitting years from your analysis.
Section 4.5 - I didn’t really understand what this section was trying to say, other than it took a different analysis approach - e.g., composite. However, I found the analysis a bit confusing in the sense that freshwater anomalies are argued to be important, but the SSS anomalies do not look like the SSTs in this case?
Figure 8 - The climate pattern seen here looks just like the response to the summer NAO… what is the correlation between your summer NAO index and the Tsummer shown in 8a?
Section 4.6 - I found this section very confusing, and it wasn’t clear to me what was being argued - in particular, I thought the authors had argued in their response that this analysis was was based on the interannual variability due to the use of the NAO, but then present an analysis of very long timescale changes. It also raises the question of what is the time periods are being used in all the other analyses in the paper. If this is not a key aspect of the story of the paper, then I suggest removing it - there is already enough in this paper.
Section 4.7 - over what time-period do you compute the relationships?
Line 423 - do you mean figure 10?
Line 443 - I don’t understand what you mean by “projet the SST each summer onto the observed SST pattern - please elaborate
Line 456 - the model simulations still don’t just test the impact of subpolar SST, but include all SST anomalies and external forcings. So I really do not see how you can attribute the signals to the subpolar SST anomalies.
Section 4.8 - Over what time period is the R^2 computed? If this is based on just the 9 points shown in figure 2? How robust is this relationship to jackknife resampling? What happens if you remove all consecutive years? |
