Summary

This study aims to investigate frequency, duration, and size of blocking under climate-change scenarios. This is an important topic. Weather type decomposition (WTD) is used in this study, which makes it original.

However, I have more than major reservations as outlined below.

Major comments

1. WTD

1a. Discussion on advantages and limitations of WTD

I find no discussion in the paper about advantages and limitations of WTD. (except description of how it is done and that it agrees with blocking indices). One of the limitations of WTD is that WTD is less flexible and it does not come with a native measure of blocking center and size.

1b. Fitness of WTD

I wonder how well WTD can summarize the Z500 variability. In the process of k-mean, is variance between cluster much larger than variance within the same cluster? Or do the 4 clusters explain a very high percentage of variance? Or is the clustering very clear cut?

1c. Interpretation of WTD

If the WTD clustering is not that clear cut, it is hard for me to interpret the results. When “cluster centroid” is found different in some models, what does it mean? Is it because reanalysis-blocking-like pattern occurs less frequent? Or is it because some boundary cases (under reanalysis clustering) occur more frequent and that the cluster boundary needs to be put elsewhere?

Also the “blocking frequency”. Do changes come from the 5-day requirement, or the overall frequency of weather type? If it has to do with the overall frequency of weather type, the question again is whether it comes from changes in frequency of centroid-like patterns or boundary-case-like patterns?

Same question for the “blocking center”. Does it come from the weather type shifting in location? If so, is it because of centroid-like patterns or boundary-case-like patterns?

1d. Insignificant results

If the WTD clustering is not that clear cut, I wonder if this will cause extra variability that stops you from drawing significant results. Most results using DeltaZ500_SSP are not statistically significant.

2. DeltaZ500_HIST may be irrelevant

Many results are based on DeltaZ500_HIST, in which the overall higher geopotential height in warmer climate is not removed. I cannot see how this overall higher geopotential height would link to weather impact or air pollution, which is likely why authors are interested in blocking. Measures of blocking based on DeltaZ500_HIST go too far, and become irrelevant to weather impacts.

Minor comments

3. Line 103: How do authors determine which (out of 4) weather regime is the European blocking weather regime?

4. I prefer to say “size” or “area” of blocking, in place of “extension”. The latter is not clear to me and let me think of temporal extension (duration), or some kind of extension of concept.

5a. Line 131: I suggest change “center method” to “composite method”. Because “composite” is really the step that differs from the “DG method”.

5b. But actually, the authors made a few modifications to the DG method that makes it very similar to the center method. For example, authors require DG blocking day to be a subset of WTD blocking day. Perhaps authors need to say what are the major remaining differences (if any) in the two method.

6. Line 7: I think your methodology to quantify size of blocking does not “rely on the WTD”. I don’t think it is WTD-native or WTD-specific. I don’t think this is sufficiently different from other studies (like Nabizadeh et al. 2019) that I would claim “new”.

7. Line 9: “Geopotential height increase” might be more accurate than “pressure increase”.

8. Explanations of methods are disordered. For example, the 3 paragraphs in section 3.1 go by talking first about WTD, then an overview of all steps, and lastly the calculation of Z500 anomaly (which is done before WTD).

9. Line 105: I would suggest to add a bracket “(including the mean)” after “annual cycle”. Because “annual cycle” can sometimes only refer to the seasonal variation from the mean.

10. Line 119: I am not sure the description on treatment of “hole” is complete that others can reproduce. Let 0 be non-blocking and 1 be blocking. What would the code say about 001010100, and 001110101011100?

11. Line 125: Is “hole” included in “blocking days”?

12. Line 136: I prefer a simpler phrase “non-zero” in place of “non vanishing”.

13. Line 136: I would suggest to mention the 75m/100m threshold here in main text, rather than having to find it in supplement.

14. Line 154: With Fig. 1, what is being evaluated is not ability to reproduce the “blocking weather regime” but “composites” (as defined in line 124-127).

15. Fig. 1: I assume this figure is based on DeltaZ500_HIST, so the overall higher geopotential height is included? From Fig. 3, I guess the overall Z500 increase is more than 25m in SSP2. Why don’t I see an increase of RMSD because of this?

16. Table 1: The resolution of GFDL is said to be 1 degree on https://wcrp-cmip.github.io/CMIP6_CVs/docs/CMIP6_source_id.html. Could the authors please check? I assume the argument made in line 179 is based on the resolution when the model is run, not the resolution of the output.

17. Line 211/227: I am not sure the similarity between DeltaZ500_HIST and DeltaZ500_SSP is entirely interesting. The overall increase in geopotential height is a shift of all clusters in a hyper space. So by construction, it has no effect on the clustering result. The only difference is the seasonal variation around the mean. The similarity in results can only suggest the seasonal cycle does not alter enough from HIST to SSP to alter the clustering result.

18. Fig. 5: The peak of ERA5 at 27-28 days look suspicious.

19. Line 247: Perhaps you can clarify “variability”. Do you mean inter-model variability, or inter-event variability?

20. Line 267: I don’t think the 0.1% is a significant digit if the area only has two significant digits.

21. Line 270: I think Nabizadeh et al. 2019 is based on DeltaZ500_SSP. And the increase you are talking is drastically larger than 17% in Nabizadeh et al. 2019. I don’t know if I would call this agreement.

22. Line 304-305: The sentence looks contradicting to me. (“may not match”, “agrees well”)

23. Supplement Step A: I assume this step applies both to the center method and the DG method. But the step uses “blocking center”. What is the “blocking center” for the DG method? Also, for the DG method, is there at most one such center/blob on each day, such that step 8 in the DG method only does temporal mean but not event mean?

I find this manuscript to be interesting and I enjoyed reading it and thinking about the results. The authors have certainly done a substantial amount of work. However, I have some suggestions. I think these have to be addressed before the manuscript can be considered for publication.

Major suggestion #1

You need to provide a strong rationale for why you use the weather typing approach for identifying blocks as opposed to a feature-based method that allows simultaneous tracking in space and time. What are its advantages? I don’t think that simply stating that others have used it before is sufficient.

There are places in the manuscript where you say things that really make me wonder, why did they choose such an approach, such as:

Line 261-262:

“This is due to the fact that the events have been identified via a partitioning algorithm (k-means) and not via blocking indexes designed for geopotential fields that are typical during atmospheric blocking.”

You don’t need to remove these sentences, but earlier, in the methods section, you need to make a convincing argument as to why this method useful. Given that there are multiple blocking indices that are designed for geopotential field, and that you felt compelled to compare your method against one of these, I want to know: why use the weather type approach?

Related to this, at line 340 – 341, you write:

“Our results are in agreement with previous findings where blocking events are defined with blocking indexes. This confirms that the application of the WTD is also a good strategy to analyse blocking event characteristics.”

For me, this is a bit confusing. Why do we need another strategy that gives us the same information that we already have?

Major #2: Suggestion affecting multiple locations:

Your choice to use shorthand names for the anomalies based on the climatology used is a bit confusing. For instance, when I first examined Figure 3, I thought there was a typo. Is there any chance that you might consider creating shorthand names that refer to both the removed and the dataset used, then, for instance:

Z500_HIST composites during the winter HIST period (1980-2009) could be named: Z500_{HIST_HIST}

Whereas,

Z500HIST composites during the winter SSP2 period (2070-2099)  could have the shorthand name:

Z500_{HIST_SSP2}

This change would also help in clarifying what the x- and y-axes are referring to in Figure 4, are all of these showing results using the historical climatology, or is the x-axis the future climate with the historical climate removed? I eventually sorted out the answer to these questions, but with your existing format, it was more difficult than it needs to be.

Minor Suggestions:

Line 99: possible typo, I think the word “so” should be “to”

Line 130 and multiple places elsewhere: when you use the word extension, do you mean the same thing as extent? For me, extension suggests an action, such as expansion or shifting in the location of the block, whereas, extent suggests the instantaneous location of the block. I am curious to see if the other reviewers or the editor agree with me on this. If they do not, you can leave it as is.

Lines 131 – 135: This explanation did not make sense to me, mainly because you sometime mention composites and other times you do not, i.e., on line 132, you write:

“The center method starts from the detection of the center of each blocking event.”

Okay, for me, because you say “for each blocking event”, I think to myself: this does not involve composites. But then the next sentence talks about defining the centers based on the anomaly for the composite. Please try to re-write this description to remove any confusion.

Line 186: I think Figure S3 should be included in manuscript as a result figure not a supplemental figure. In the current form of the manuscript, a comparison of Figures 2 and 3 gives a strong suggestion that blocking changes significantly with climate change. But that is not the result that you are presenting. Instead, the differences in Figures 2 and 3 is mainly a due to a difference in the mean state for the 20 and 21^st centuries. Right? If you see it differently, then please explain.

Lines 252 – 257: This discussion of block intensity when considering delta-Z500HIST vs delta-Z500SSP is a bit awkward for me. Given that in the present climate we define blocks as anomalies with respect to the climatology, your discussion of future block “intensity” relative to the historical climatology, seems a bit arbitrary. For context, if we were discussing heat waves, we know that humans feel uncomfortable when the temperature is above a certain threshold, e.g., 32 ˚C. So, when someone studies future warm events, there is a reason to look at anomalies with respect to our current climatology. But for block intensity, does that same thing hold? i.e., when surface pressure is above some threshold are there specific impacts on humans? If so, please explain. If not, then I wonder if you might consider streamlining this section and only discussing the intensity changes when comparing delta-Z500HIST for the historical runs with delta-Z500SSP for the future runs. Especially since your discussion of Figure 3 already makes the point about the impact of the change in the climatology on the Z500 anomalies.

Figure 7, one result you have found, that I don’t think it mentioned (pardon me if I missed it), is that the model-to-model differences in intensity are larger that the intensity differences for SSP2 vs SSP5. This suggests to me that even though the models are getting some aspects of the physics of blocking correct, there is room for improvement.

Lind 331, you write:

“Climate change will significantly increase the extension of blocking events in the future especially in the worst-case scenario.”

I don’t think your results agree with this statement. As you discuss, the difference is related to a change in the climatology. For me, this implies that the blocks themselves, in terms of their size and impact on the circulation, will be similar to what we observe in the current climate. Perhaps a method for testing this would be to address the question: are the pressure gradients associated with the blocks strengthening?

Philosophical Suggestion:

Whenever a “trends” manuscript is written, I think the authors should take time to ponder the question: Are we (or somebody else perhaps) going to have to write this same paper in N years when the next generation of climate models are released? If the answer is yes, then why do we need this paper now? If the answer is no, or not exactly, then why not – what have we established here that is robust to potential changes in models? And maybe some elements will be updated, but are there are least some elements that are novel in this study that are not going to be superseded by results from an improved set of climate models, and/or are the hypotheses or theories developed in this work that provide a simpler guide to how we can interpret the trends?

Obviously, I have not reviewed the majority of trends papers written, so most of the time this question is not addressed. So, I won’t hold you to a higher standard than what exists in the literature. However, I encourage you to consider engaging in this activity and perhaps including a paragraph at the end of the paper addressing this issue. No problem if not.