This study has investigated the likelihood of a re-occurrence of the extremely cold European winter of 1963 under present climate conditions and what such a winter would look like. To investigate this, the authors employ a range of techniques, some involve accounting for the dynamical features which gave rise to the extremely cold winter, while others are standard statistical methods, e.g. extreme value analysis. All the methods gave approximately similar answers, that such an event could occur today but is less likely, and if it did occur, the temperatures would be approximately 1.5 degrees warmer than the original 1963 event.

The paper provides a clear and thorough assessment of the theoretical occurrence of the 1963 winter under today’s climate. It is an interesting study, and would have been more interesting if they had found that the winter could not occur today. However, since the conclusion is that an extremely cold winter from the past is less likely to occur and would be warmer if it did occur under today’s warmer climate is not especially groundbreaking. The paper does raise the issue of mal-adaption by society towards warmer winters, and I think this is very valuable. I also think the comparison of the methods will be of interest to the community. Overall, I would recommend the paper be published with minor revisions. 

General Issues

1. While the result that the cold winter can still occur and would be warmer is not wholly surprising, the paper does makes a good point that society may be adapting under the assumption of warmer winters while extremely cold winter are still very possible. I think this could be raised up in the paper. Perhaps introducing the idea as part of the motivation for the work so it is in the reader’s minds as they go through.
2. It is a potentially provocative finding that the unconditional statistical method of using a fitted GEV gave as good a result as methods which incorporated knowledge of the dynamics. Many statistical based studies are criticized for not including knowledge of the dynamics of a situation, and here we have a case where this knowledge provided no additional benefit, and for an extreme event no less. Obviously, this is a single case and it could be random chance that the statistical method did so well. Given that the authors have used so many different methods, I think it would be a nice addition to the paper for the authors to briefly comment on how the methods compare, especially the unconditional statistical method with the methods incorporating knowledge of the dynamics.
3. The title is a little sensationalist and I’d suggest changing it to match the style of the WCD journal.
4. The second paragraph in the Results section discusses the failure of models to show pronounced forced changes in atmospheric circulation. This is not a topic that is really investigated by this study. CESM2 simulations were run to perform the model boosting analysis, but there is no assessment of how or why or to what extent CESM2 fails to show forced changes. The discussion reads more like a commentary on the failure of models in this particular aspect, and does feel connected to the rest of the study. If this is a main motivation or theme in this study, then this failure of models should be introduced in the introduction and its implications on the use of model data in this study needs to be assessed, not merely commented on. Without expanding on the issue raised in this paragraph and incorporating it more fully into the study, I’d recommend removing the paragraph, more specifically, sentences from line 204 to 212.
5. Figure captions for figures 2, 3, and6 need to be expanded to better explain what is being shown in the figures. This is especially true of figures 3 which is important for the paper.  

Minor Issues

24: You use the expression “led to”. This suggests some precession in time, that the pressure anomalies occurred and then afterwards, there was a negative NAO anomaly. Perhaps change to “resulted in” or “comprised”.

65-90: The dynamical adjustment method. Is it reasonable to separate the dynamic and thermodynamic influences on surface temperature in such a linear way? Consider adding a sentence or two discussing the caveats or limitations of this assumption/approach.

78: “The first dynamical adjustment approach (dark blue line in Fig. 1) uses ERA5 to train the regression model, and the spatial pattern of sea level pressure (SLP) over a circulation domain over Europe and the North Atlantic.” Possibly it is just my reading of the sentence, but it feels a little awkward. The ‘and’ feels like it is part of the regression statement. Consider changing this to “along with”.

82: Start a new paragraph at “We use a second method…” It provides a cleaning break when reading. Possibly change to “We also use a second…”

169-183: Please revise the structure of this paragraph. It jumps straight into what the method leads to and only describes the method itself towards the end of the paragraph. I suspect this is one of those cases where the author is so familiar with the method, he forgets that the reader may not understand what he is talking about from the beginning.

174-181: Please explain in the paragraph why you use the single coldest winter in December for the first boosting, but two coldest Januarys for the second boosting.

176: Please write the dates out in full. Using “01.12 to 15.12” could be written as “1st to 15th December”.

180: Again, write the dates out in full.

219: Remove the double brackets.

245: The return period of the 1963 event was 119 years. For such an event to occur today, the return period would be 371 years. However, the uncertainty for the occurrence today is 97 to 7680 years. That is to say, the return period of the event today may still be 119 years at the 95% level. This should be commented on.

251-252: “This indicates, incidentally, that the storyline approach is not providing larger effects of climate change compared to the probabilistic approach, or possibly exaggerating these effect.” This sentence is confusing. Does this mean the storyline approach is ‘not providing larger effects’ or is exaggerating effects (i.e. to make larger)? These seems to suggest storyline approach either does’t make the effects larger or it does make them larger. Please rephrase the sentence.

259: This is a good point about mal-adaptation. I think it is a shame that this only now appears in the paper and wasn’t raised in the introduction.

270: Change “similarly” to “similar”.

Figure 2: Subplot a needs a grey line in the legend to explain what the grey lines are, and caption could explain how they relate to the blue line. You said (d) shows difference between (c) and (d), think you meant (b) and (c).

Figure 3: This figure needs a lot more explanation in the caption. Possibly, this could be done in part in the paragraph on lines 239-260 where the figure is discussed. Instead of simply making a statement and referencing (Fig.3), instead reference (Fig.3a red line). This would make it easier for the reader to connect the point your are discussing with the specific feature in Figure 3.

Figure 6. This needs more explanation in the caption.  

Review of “An extreme cold Central European winter such as 1963 is unlikely but still possible despite climate change”

The manuscript asks the question of how would the cold 1963 European winter look like if similar atmospheric conditions were to develop in the recent, warmer, climate. Furthermore, it asks whether events as cold as winter 1963 might still be possible today. For this purpose, a number of different methodologies recently develop in the context of extreme event attribution are adopted and put to this test case. They conclude that climate change has warmed an event like winter 1963 by about 1.5 degrees and an event as cold as winter 1963 could still happen today, though it this is unlikely. 

I have very much appreciated the multi-method approach which is adopted in the manuscript. The overall finding  that the different methods provide very consistent answers is an important result, since it helps confirming the validity of the proposed approaches in this, and other, applications. I think this is an important result, which might benefit being stressed a bit more.  The overall conclusion that such cold winter conditions would be warmer today is certainly not unexpected, but a quantification is still useful. The manuscript is well written, and the figures, though  packed with perhaps too much information, are nice and clear. Overall, I have very few remarks for this manuscript, which I fully recommend for publication. 

General (minor) comments:

- There is some repetition of the description of the approaches between the “method” section, and the result section. This would be fine if the methods section were at the end of the paper, but given the structure of this journal, I think repetitions could be reduced.

- The authors could consider summarising the results from the different methods in a table.

Specific minor comments

Line 31: It is not clear there is a long cold tail in the temperature distribution from Fig 1a, though I agree the tail is clearly evident in Fig 3. Please clarify. 

Line 95: remove parenthesis in the citation of Shepherd 2016

Line 106: “with the same spacing” is not clear. Please clarify.

Line 115: please specify the estimated percentile

Line 117-125: this text should be removed from here, since it discusses results and not a method.

Line 126: it’s not clear to me why 2.5 is called “unconditional”. Wouldn’t 2.4 be unconditional too? The exact methods differ, but it seems to be me that the main difference between 2.4 and 2.5 is that the first is based on climate model data, and the latter on reanalysis data. 

Line 131: please add mean after DJF.

Line 137: do you mean a linear function of GSAT? If not, what function?

Line 155: “utilising importance sampling techniques” would require some more discussion on how this is implemented. It received considerable less space than the boosting methodology. For example: is the reshuffling performed daily, or over blocks of a given length to better preserve autocorrelation?

Line 163: please remove continuous.

Line 175: what is the CESM2-ETH ensemble? Please clarify

Line 175-185: this text could be reduced, since discussed in the results section.

Line 185: If the discussion of the method is repeated in the results section, such as here, then please add a reference to the section in which that method was discussed. 

Line 186: If I get it right, the winter cold extremes such as 1963 have warmed less than the DJF mean temperatures. But don’t we expect cold extremes to warm faster than the mean, due to, e.g. plank feedback and polar amplification?  Could you provide some discussion of this unexpected result? Can it be a consequence of the impact of the mean circulation trend on the mean temperature trend? Or other processes are at play. 

Line 261: Please note that method 2.5 had already answered this question. To the extent that return levels of such amplitude anomaly are still associated to a return period in the present day climate, events colder than 1963 are still possible. Please discuss. 

Line 239: Following on the previous point, I don’t understand why the CESM2-LE approach is “conditional on the 1963 atmospheric circulation”. That approach is just looking at the trend in a low percentile. It includes both changes in the magnitude and frequency of cold events. If the we assume that looking at a low percentile implies conditioning on “tail events”, then the GEV approach from WWA should be equally considered a conditional analysis. 

Line l 268-273: some redundant text, since the boosting is already discussed in the methods section.

L 300: The only information about the ability of CESM2 comes from Fig 4a, which is qualitative. Could you please quantify the mean temperature bias, and the bias in a low temperature quantile? That would be useful to add some confidence on the model results.

Line 310: why not showing a map for the temperature anomaly associated to the empirical importance sampling? It could be showed in place of the second boosted CESM simulation which does not add much information. 

Fig 1 caption: add (blue dashed) after “atmospheric circulation” and (black dashed) after “global mean temperature”. 

Fig 2 caption: Please specify the method used to estimate the blu line in Fig 2a. b): replace over Germany with over Europe. The last (d) should be a (b).

Fig 3 caption: The main message seems that the different methods give consistent answer, more than there is uncertainty.

Fig 4: what is the unit of the Circulation (SLP) axis? Please specify.

This is a clear and concise paper, using a range of appropriate methodologies, to examine an important and useful question. I appreciate the discussion of the potential role of natural variability in the dynamical trends, and find the results and conclusions to be clear and well communicated. I recommend publication, subject to some minor comments below.

Line 26. I suspect the c in O’conner should be capitalized? (also line 37 and other references to O’Conner (1963)

Line 31. The long cold tail isn’t immediately apparent from Fig. 1, I realized after looking at Fig. 3 that you do have the distribution shown on the far right, but this isn’t referenced in the caption or labelled in any way, and so is easy to miss.

Line 80. Why SLP for the ERA5 model and z500 for the CESM LENS trained model? Is daily SLP not available in the CESM LENS? Do you think this influences the effectiveness of explaining surface temperature extremes?

Line 95. Citation shouldn’t be in parentheses

Line 130. I assume you mean 90-day running means?

Line 174. CESM2-ETH is only defined later, on line 275.

Line 180. I think you mean Fig. 6 here? Also, it’s a little confusing to reference Fig. 5 (or 6) before Fig. 4, although I understand that it’s just to give some more information on the methodology.

Section 3.1: You show that average winter temperatures over Germany have increased by 2.5C, and explain that some of this is due to dynamical impacts, but it would be useful to have a quantitative assessment of how much thermodynamical warming has occurred over Germany over this time period. In section 3.2 you show that the coldest winter (1963) would be only 1.4C warmer in the present climate, but also make the argument that the coldest winters occur when there is advection from regions that are experiencing greater thermodynamical warming. This would suggest to me that we might expect the coldest extremes to warm faster than the average temperature, but it is hard to make this comparison without know the thermodynamical contribution to average Germany winter temperatures.

Sections 2 and 3: There could be clearer separation of the results between different sections. For example, I think section 2.4 reports results that might be (more) useful in section 3. Similarly, in line 245 you state that the winter 1963 event has a return period of 371 years in 2021, but then in the next section, 3.3, ask whether such an event as the 1963 winter would be possible in today’s climate – it seems like you already answered this in the previous section if it has a return period of 371 years. I understand that you look at other methodologies in Section 3.3, but I suggest some re-arrangement to help this flow a little better.

Line 280. I don’t think the word ‘bias’ is quite correct here. I would recommend ‘residual’ or similar.

Fig. 4. The x-axis labels of circulation and albedo for the top row are closer to the plots below, and so look more like titles for those plots than x-axis labels for the upper plots.

Section 4. I would consider renaming this section as Discussion and Conclusions, as the first paragraph seems more discussion than conclusion.