This study provides an extremely detailed evaluation of the representation of blocked Euro-Atlantic regimes in a large sample of numerical climate models. The analysis is thorough and robust, yet I am doubtful of some of the interpretations provided, as well as on the overall framing of the results.

Major Comments

1. There is a broad literature on the nature of blocked versus zonal flows in the North Atlantic, and more broadly the midlatitudes, from the seminal works of Charney and DeVore in the late 70s to more recent theoretical perspectives (e.g. Faranda et al., 2016). Here the authors define three blocked regimes, and overlook entirely the rich discussion around the existence or lack thereof of zonal versus blocked equilibria in the mid-latitude flow, and the work on zonal flow regimes. While I still believe that the analysis of the regimes defined by the authors may be useful, it is less general than the “Euro-Atlantic Weather Regimes” they claim in the title, and may more aptly be called an analysis of blocked flows in the North Atlantic.
2. Related to the above, a general criticism that could be moved to the rationale of this study is the idea put forth by the authors that time invariance of regimes is always a desireable feature of regime definition. This then leads the authors to only identify blocked regimes. One could well imagine that information provided by the change in the regimes in the historical period may contain useful dynamical information and be used to evaluate model behaviour (perhaps with the support of AMIP simulation), while the future changes in regimes may be useful to compare different models between themselves. In this study, the authors wish to obtain time-invariant regimes, but that comes with the hidden assumption that the system the regimes are computed on is sufficiently stationary over the analysed timeperiod for regime stability to be a positive and not a detrimental feature of the regimes.
3. 451-460 I find this section a bit chaotic. First you introduce the figures in block, and then you discuss some of these in detail (but not others) in the various sub-sections. As the paper is quite long, if this part is not essential I would suggest that the authors remove it and introduce the figures when they are being described in each subsection, perhaps moving the figures that are not commented on in detail to the Supplement. Also, throughout Sect. 5, the authors could state explicitly “(not shown)” when in the text they comment on some aspect of their regression model that is not explicitly illustrated in the figures. There are also several statements in Sect. 5 that are supported by the figures but that are not followed by a reference to any figure.

Other Comments

These are not necessarily “minor” comments per se, as some pertain to the interpretation of the results, but they are more specific than the above comments.

1. 18 Did the authors actually intend to highlight the North Pacific here?
2. 68-69 I am not sure I follow the logic whereby these regimes being well-represented in climate models make them well-suited for model evaluation. The logic risks being somewhat circular.
3. 86 How do you obtain 1 degree data from the reanalyses that have a lower native resolution?
4. 93 It is odd to refer to table S3 before table S2. Perhaps swap the order of these?
5. 119 This is not true, and one could easily devise a distance metric where this does not hold.
6. 130 Is 4 EOFs not a very restrictive number? What fraction of the variance do they explain? Many studies, including some the authors cite, use more EOFs as basis for the clustering. I appreciate that this classifications has been previously used by the authors, but some further details on how it reflects intraseasonal variability should be provided.
7. 144-145 Could you explain better what role the maximisation plays here? One could imagine simply computing an area-weighted average pattern correlation between the matching regimes in the two datasets as a comparison metric. Or do you mean that you identify the matching regimes as the pairs with the maximum pattern correlation? In the latter case, I would recommend rephrasing the sentence to state this more clearly.
8. 176 This is perhaps a detail, but is the blocking in the AR regime for a threshold of 0.4 significantly above average? The difference shown in Fig. S3 relative to climatology looks to be minimal.
9. 228-229 It is odd to have this bit about the sign of the EOFs here. Should you not mention this somewhere in Sect. 2.3, perhaps when you introduce pattern correlation?
10. 259 Do the authors intended 1 degree here?
11. Starting l. 265 Could you not state this in simpler terms, i.e. that using adjusted R^2 and Ridge Regression you find that 6 predictors is the optimal number, and that the 6 predictors you list above are in turn the optimal ones? Also, does the optimality of the chosen predictors hold for all regime behaviours considered in Fig. S6 or only for a subset?
12. Methods: You should specify how you identify blocking. The description in the caption of Fig. S3 does not allow for reproducibility.
13. 315-319 Here you could note that Grahms and colleagues have indeed recently adopted a 7-regime definiton, and have also argued that fewer regimes may not sufficiently describe the intraseasonal weather variability in the region.
14. 322-323 Is it really the case? If you plot the difference between CMIP5 and CMIP6 in the lower RHS panel of fig. 4, does 3 actually stand out?
15. 324-326 Does this claim not contrast with the fact that CMIP6 models have higher fidelity for essentially all regime numbers?
16. 346 But by this same metric the location of the Northern peak seems to have become worse?
17. 408 I am not sure I understand how Fig. 11 shows something about “whether the individual models are able to represent the correct levels”. The authors still show bulk statistics, much like in Figs. 9 and 10.
18. 412 Analysus --> analysis
19. 447 Please introduce the variance ratio in the methods, especially as you also mention this in the context of the “classical” k-means clustering regimes, which may cause confusion for some readers.
20. 507 “persistence (the jet wobbling more within distinct regimes)” I am not sure I follow this logic. A priori, one could expect some of the regimes to be more persistent when characterised by a low latitudinal jet variance. If all regimes, when persistent, are associated with large latitudinal jet variability, then clearly the identified regimes reflect only a specific aspect of “the longitudinal and latitudinal variability of the jet” that the authors mention in the introduction.
21. 27 The last point in the list, which supports the choice of analysis framework, should perhaps come first (or at least closer to the top of the list)?
22. 568-570 “The latitudinal variability of the jet (the ‘wobbling’) interpreted as regime variability.” Does this not contradict what you stated in Sect. 5.3 (“our result therefore adds additional clarity by demonstrating that the extra variability primarily projects onto regime persistence”)?
23. 583 “too many extreme rainfall events”. Usually the models have the opposite problem, namely that they do not produce enough extreme rainfall compared to obsservations.

Fig. S3: There is an “?” in the caption.

Fig. S3: “For a threshold of 0.4 the neutral regime features less blocking than in the overall climatology at all longitudes”. Is this really the case? It is sort of hard to tell from the figure, but it does look as though this does not hold around 10E?

Figs. 9 and 10 Please label the axes.

Figs. 9 and 10: The number of models included in the different years will affect the spread shown in the plots. Can the authors indicate, either in the figures themselves or in an SI figure/table how many models were included for every year?

Fig. 11 Caption: “mulit” --> “multi”

Fig. S8 and following: please check the captions to see whether they refer to the correct figures in the main text.

Faranda, D., Masato, G., Moloney, N. et al. The switching between zonal and blocked mid-latitude atmospheric circulation: a dynamical system perspective. Clim Dyn 47, 1587–1599 (2016).

Grams, C. M., Beerli, R., Pfenninger, S., Staffell, I., & Wernli, H. (2017). Balancing Europe’s wind-power output through spatial deployment informed by weather regimes. Nature climate change, 7(8), 557-562.

The authors present a comprehensive and detailed assessment of the phenomenology of Atlantic weather regimes in reanalysis and historical coupled simulations.

I recommend that the manuscript is accepted after minor revisions.

I wonder whether the title can be improved. I see that the study focuses on the phenomenology of regimes and not on their dynamics.w

In the abstract you refer to ‘wintertime circulation’ and then mention that this is dominated in the Atlantic by non linear flows. Please be more specific here, for instance mention you look at low frequency atmospheric variability.

Line 15, complex orography and external forcing are not a feature of the wintertime atmospheric circulation, please refer to an interaction with them.

Table 1 please clarify what is meant with ‘experimental’.

Line 84 why use one ensemble member? Is the ensemble member generated with a lower resolution compared to the reanaysls ( I think for ERA5 this is the case)?

Lines 111-119 this is unclear, please provide more details about the methodology and its interpretation.

Line 130 I was surprised by the use of 4 EOFs, form literature I would expect a higher number of EOFs. Is this common to other studies? Do the authors have an idea of how this choice can influence their results?

Line 136-143 Again please provide more details on the methodology.

Line 160 consider giving an explicit definition of stability and fidelity.

You introduce a set of predictors that mix Eddy forcing and atmospheric horizontal resolution with

remote drivers. I disagree that eddies and resolution can be labeled as predictors and put on the same level of other drivers.

Line 314 please report in the text the conclusion of DS

You should mention that the 7 regime approach has been used elsewhere in scientific literature.

Line 329 the ‘general convergence in ensemble behaviour’ is rather vague, please explain.

Line 345  but the northern regime latitude is actually worse in CMIP6 (?)

The construction and discussion of figure 7 is a bit hard to follow.

Caption of figure 9, what is the ‘full ensemble spread’?

I understand that the study is based on the assumption that historical simulations should reproduce the observed behaviour of fixed regimes in the reanalysis. While this is certainly a reasonable exercise, the hypotheses introduced by the authors should be clearly discussed.

Line  543 this point is rather obscure, consider rewriting it.

Line 550 I am not sure I can but the point of the suitability for intermodal comparison. Please pur forward the argument with a broader discussion or consider dropping it.