This paper examines the winter atmosphere-autumn SST relationship in ERA5 and CMIP6 models to assess the representation of atmospheric response to SST over the North Atlantic region. It starts with a MCA analysis between SLP and SST in simultaneous and lagged analysis, finding largely consistent concurrent MCA but not for lagged MCA in CMIP6 against ERA5. This underestimation of lagged response in MCA is then assessed, and seems to be confirmed, in an independent approach of atmospheric circulation analogue  on the autumn surface heat flux. This is a comprehensive analysis of the CMIP6 models against observations and provides a reference that can be compared against using other methods. The results are generally interesting. Nevertheless, the complex analysis, with multiple data sets and multiple methods (MCA, EOF, concurrent and lead-lag analysis, and analogue method) makes it a great challenge to follow.  I recommend a minor revision before the publication of the paper.

Major comments:

My major comments concerns with the presentation of the results, instead of the result.

The paper uses a comprehensive analysis strategy that employes multiple methods, but didn’t give a guideline of this complex strategy of analysis. For example, the lead-lag MCA analysis has been used to isolate atmospheric driving of and response to SST in the first half of the paper. Then, it uses the analogue method on surface heat flux to re-examine the feedback. This is my understanding after some long-time contemplation, because the paper didn’t lay out this strategy clearly. Similarly, the paper also uses MCA analysis and then compares with the EOFs of single variables to check if the coupled mode explains the dominant variability in each field…

It will be useful to readers to state clearly the strategy of the paper. For example, why using the lead-lag MCA is not sufficient and the analogue method need to be used? I suggest the authors to add a subsection explicitly in 2.3 Method to describe the overall strategy, and the reason why so many different methods have to be used.

Also, it will be very helpful for readers, after the application of a method, specify what has been learned and what are the remaining problems, and how this problem can be resolved using another method. This naturally introduces another method, making the paper flow more smoothly. For example, after the lead-lag MCA analysis, what has been learned on the response and feedback, and what is the remaining question. How this remaining question is to be addressed, using another method, such as the analogue method.

It is perhaps also useful to highlight that this paper concerns more on the atmospheric response to SST (or the feedback of SST on the atmosphere in models against observations), since the atmospheric driving of SST seems to be a robust feature in models as in observations. It is the feedback part that show diverse results in the models, consistent with GCM simulations (Kushnir et al 2002) and the signal-noise paradox.  Accordingly, the discussion on the feedback part should be given higher priority than the driving part, than in the current version of the paper.

Review of Anderson et al “Boreal autumn to winter North Atlantic atmosphere-ocean coupling in large ensemble climate simulations”

This manuscript investigates the coupling between boreal Autumn ocean variability and Winter atmospheric circulation in ERA5 reanalysis and CMIP6 models. They use a variety of techniques including MCA, EOFs and a circulation analogue decomposition. They find that while models generally accurately represent contemporaneous modes of ocean-atmosphere co-variability, they do not show a lagged relationship between Autumn sea surface temperature and the Winter North Atlantic Oscillation (NAO) as strong as that seen in reanalysis. They conclude from the circulation analogue decomposition that the Autumn SST-winter NAO relationship is not the result of SST variability independent of atmospheric-driving. The paper raises questions about the mechanisms by which Autumn SSTs relate to winter circulation.

This paper addresses an important question which fits well within the scope of WCD and is generally well written. I have a few relatively minor comments and recommend publication once these points have been addressed.

Comments:

1. Statistical significance. It would be useful to have stippling (or similar) showing statistical significance in many of the figures (e.g. figures 1 and 4). This would help to indicate in which regions the coefficients are particularly unusually anomalous and less likely the result of random correlations.
2. The authors show large differences in the MCA coefficients for Autumn SST and winter SLP between ERA5 and models (figure 5) - is there a way to show that this isn’t consistent with sampling uncertainty given the relative length of the observational record?
3. On the question of the causes of the relationship of Autumn SST and Winter SLP, it would be useful to know to what degree SSTs persist from Autumn to Winter – does an MCA analysis or regression analysis Autumn SST and winter SST show strong co-variability?
4. In the discussion, the authors mention a few potential causes of the Autumn SST – winter SLP relationship including SST anomalies generated in autumn persisting to winter and remote driving via Rossby waves from the tropical Pacific. Firstly, does figure 7e, showing a relationship between autumn circulation-driven THF and weakly positive NAO-like SLP, give some evidence for the former? Secondly, could the stratosphere be an additional player in creating a lagged relationship (e.g. Collingwood et al 2024)

Refs

Collingwood, E., Scaife, A.A., Lu, H., Sinha, B., King, J., Marsh, R. and Marshall, G., 2024. Meridional wind in the upper stratosphere: A source of winter NAO predictability. Geophysical Research Letters, 51(24), p.e2024GL111717.