Review of “Dynamical drivers of Greenland blocking in climate models” by Clio Michel et al. 

This manuscript investigates the drivers of atmospheric blocking in the Greenland region using several large ensembles of climate models. The representation of blocking and other dynamical features in each of the models is compared to ERA-Interim. Based on this analysis two models (ECHAM6.3-LR and MIROC5) are selected based on their performance in simulating historical blocking in the region of interest. The study continues to investigate several atmospheric features such as Rossby wave breaking and and wind speed in present and future in order to identify the drivers behind the expected changes in winter blocking in the Greenland region. Interestingly, the authors find that different dynamical features seem to be driving blocking in the two models considered. 

This is relevant and timely work on the underlying dynamics driving atmospheric blocking for one of the main blocking regions in the northern hemisphere. I think the study fits well into the scope of Weather and Climate Dynamics and generally I think it should be published and I have no major comments. However, I found the manuscript to contain a large number of smaller issues, inconsistencies, and ambiguities that made it very hard to follow the authors points and sometimes even impossible to be sure what the authors were referring to. I would urge the authors to carefully address these and particularly focus on a consistent description of their methodology as well as their figures. 

Comments

35: CMIP actually stands for Coupled Model Intercomparison Project as far as I know

84: As additional information the author could state how much warming that is approximately compared to the present decade used in the paper. 

Figure S1: I have to admit I’ve trouble interpreting that figure for several reasons:

- The caption is unclear to me: “Difference of blocking frequency where blocking is detected”. Do both panels show blocking frequency or some kind of difference?

- I assume that the figure actually shows absolute blocking frequencies and not any kind of difference. So to show if there are “substantial changes in the results” (line 114 of the main manuscript) the author should plot differences (or even relative difference even though I realize that his might be problematic as well). 

- It is unclear to me what the data-basis for this plot is. ‘All-Hist’ in the title lets me assume that it refers to all 125 members of NorESM1 in the historical period but this does not seem to be the case as the blocking frequency looks too patchy for that – so is it possible that it is only one member?

- The high ‘blocking’ frequency in the Atlantic below 40 degree north is not normally defined as blocking (at least not in the classical sense), right? The authors focus on blocking over Greenland so I would suggest optimizing the colorbar to the blocking frequency there.

Section 2.3: Did the authors calculate the annual blocking frequency and then select Greenland blocking days in DJF from that or was the blocking frequency already calculated only for DJF? In the second case the first and last 4 days of each season will have a lower probability of blocking due to the 5 day persistence criterion I assume?. This should be made clear. 

133: I assume GB stands for Greenland blocking?

134: Why is there a difference between the blocking and the anticyclone definition? Should blocking not just be a stationary anticyclone?

150: Figure S2 does not show that “none of the 30 climatologies of 9 consecutive winters [...] is significantly different from the total 40-year [...] climatology”. as the authors seem to indicate here it just shows the blocking climatologies.

How was this established? Was that done on a grid-point level or somehow averaged? Would we not even expect about 10% of cases to be outside of the 90% range?

figure 1: 

- I personally think that there should be an indication of what is shown as well as the units directly in the figure and not just in the caption.

- “Black lines show the 2, 4 and 6% contours for ERA-Interim (2006-2015).” For me these lines appear grey but that might just be my viewer.

- In panel f the black/grey countours do not seem to match the respective shading. This might just be due to different interpolation but it maybe should be clarified. 

- I assume that the white areas without dots just have zero blocking frequency (rather than significant biases)?

165: Could the authors rephrase this? I did not count but it looks to me as if for all models there are more dotted than un-dotted gird cells even when only focusing on ‘blue’ areas so I don’t think the concluding that “Most of the negative biases are significant” (except for MIROC6) is supported as it is. 

168: This seems to partly contradict the earlier statement that 9 years are enough to establish a robust blocking climatology, right? (even though the argument there was only made for observations). I would argue that for a sufficiently long climatological period the blocking frequency should be a model property and not depend on the realization, right? 

Figure 2: I am again not sure what is shown in this figure. The caption suggest to me that it shows the area average of blocking frequency over the GB region: “mean Greenland blocking frequency” but this is not correct, right? Frequencies are too high for that, so I assume it actually shows Greenland blocking days as defined in line 116? Please clarify. 

Figure S3: The the unit of the shading should be somewhere in this figure. 

Figure 3: I assume that for ERA-Interim not the ensemble mean but the mean over the different periods is shown? 

211: “MIROC5 and ECHAM6.3-LR are the models with the largest variability on the equatorward side of the mean jet and the smallest mean-state biases with respect to wind variability.” Can the authors go into a bit more detail here? Looking at figure 3 I am not sure if I can see what the authors mean with this statement. In particular the standard deviation from MIROC5 does not seem to be higher southwards of its climatological jet than any other models’. 

216: “Climatologies show that AWB is more frequent and located on the equatorward side of the jet

at both low and upper levels (solid contours in Fig. 4 left)”

What is meant by “low and upper levels” and where is this shown in figure 4?

This sentence somehow seems to suggest that the solid contours in figure 4 are an indication of the jet but to know the position relative to the jet the reader actually needs to compare figure 4 to figure 3f, right?

Section 3.3: It would be interestingto discuss these results in the light of this recent study as well: https://onlinelibrary.wiley.com/doi/10.1029/2020JD034082

l251: “ECHAM6.3-LR [...] slightly underestimates the Greenland blocking frequency” I’m not sure if I would call a mean underestimation of about 50% (if I read figure 2c correctly) ‘slightly’. 

Figure S8a: Again it would be interesting to see if the results are consistent with Drouard et al. (2021) when split by AWB and CWB. 

291: CMIP5 was published around 2013 if I’m not mistaken. Can the authors check if the cited studies really all look at CMIP5 data?

292: “we note a decrease in the ensemble mean blocking frequency over Greenland, in particular for ECHAM6.3-LR (up to -0.6% of the time)...” It this referring to Greenland blocking days as defined in line 116 or to blocking frequency directly as I would assume from this statement? 

295: It seems to me that not the change relative to pre-industrial is important but the change relative to the historical period?

302: “The composites over the blocked days are very similar between the present and future experiments (Fig. 7)” Figure 7 does not show the present?

Figure S11: I assume this figure shows change between historical and future?

311: “Over the North Atlantic, CWB do not seem to change much (very weak values and noisy field; not shown)” What does ‘not shown’ mean in this context? CWB is shown in figure 8 and S11, right? There just happen to be not contours in the frame for the case of figure 8b? Or do I misunderstand something?

What does ‘seem to change’ mean? In a given model it either changes or not, right?

336: Except for AWB in MIROC5, right?

Technical comments:

42: ‘have showed’ 

Figure 6: “(g,h,i) Same as (c,d) but for anticyclonic wave breaking frequency”

288: “CWB are”

Review of "Dynamical drivers of Greenland blocking in climate models”

The present manuscript explores the deriver of Greenland blocking events using a set of multi-model large ensembles. This is a nice way to analyze blocking events as they are rare events. The authors showed that two (MIROC5 and ECHAM6.3-LR) out of five models fairly capture the frequency of blocking events compared to the ERA-Interim reanalysis. Then they used these two models to understand the main driver of Greenland blocking events and found an inconsistency in the dynamics of blocking events between the models. More specifically, the author found a link between CWB and Greenland blocking events in ECHAM6.3-LR model in contrast to MIROC5, which CWB is largely underestimated. They further expand their analysis to future climate and report the changes in RWB and eddy-driven jet to investigate the connection with blocking frequency.

I found the topic and contents relevant, and the paper is falling within the scope of Weather and Climate Dynamics. I have no major comments, and I strongly suggest that for publication after some minor revisions, which I believe are needed to improve the quality of the paper and make it clear.

L7: The authors mentioned “four out five models” but then throughout the paper, they focused on the two models which they believe have better performance compared to the other models. Even in conclusion (L324) they mentioned “three out of five” and I suggest they change it in the abstract to be consistent with the analyses in the paper.

L42: “have shown”

L85: I think it sounds better if the authors rephrase the sentence and mentioned 100 and 501 members instead of “times”

L112: Shouldn’t it be “Blocking grid points are identified” instead of “Blocking events are identified” as the index identifies each grid point that satisfies the condition, which does not necessarily mean they are separate events.

L115: I am unsure if I can follow the time fraction (%) definition that the authors explained here. Since it is an important point, they need to explain that clearly. Is the fraction defined as (total blocked gridpoints/(total gridpoints*number of days))?

L116: Have the authors thought about performing clustering analysis to separate Greenland blocks instead of subjective criteria to count the number of blocked days as I think the Greenland region will show up naturally as a region of frequent blocking events base on what the authors have shown in Fig. 1(f).

L133: “GB” spell it out

L166: Over Greenland, negative bias in the frequency of blocking is not significant in the ECHAM6.3-LR model in addition to MIROC5, and it needs to be added. Also, in the following sentence (L168) the author mentioned “strong negative biases in the ensemble mean.” which I think is not true for those two models over Greenland based on what has been shown in Fig. 1.

L211: I believe that ECHAM6.3-LR is the only one that has the smallest mean-state bias in Fig. 3, and biases in MIROC5 are the same as the other models (specially CanAM4 seems to be even better compared to MIROC5).

L247: Fig. 5d-f not Fig. 5c-e 

L269: Can weak absolute vorticity gradient be why the frequency of blocking events is captured with less bias in MIROC5? If yes, I think the authors need to expand this part.

L275: Fig. 6m-o not Fig. 5l-n 

L279: I think the authors mean Fig. S8 instead of Fig. S9.

L293: I understand the numbers are not significant compared to previous studies but are the changes significant compared to each models' current climate?

L297: It is unclear what the authors refer to by saying “large decrease”. Isn’t it a weak decrease as they found here, and it is mentioned in L293?

Last comment: Captions of figures are vague and ambiguous, making it hard for readers to follow. They should be clear as much as possible that readers understand the figures just by reading the captions. Also, if the anomalies are significant (by any t-test) please mentioned that in the caption, otherwise please do statistical significance analysis for those cases that anomalies are shown. 

Review of manuscript wcd-2021-26:

Dynamical drivers of Greenland blocking in climate models

by

Clio Michel et al.

In this study, the HAPPI large ensemble simulations are used to investigate the representation of Greenland blocking (GB) in the HAPPI models and the extent to which GB occurrence is associated with the occurrence of Rossby wave breaking, specifically cyclonic Rossby wave breaking (CWB) over the Labrador Sea / the south of Greenland (70-30W, 50-70N). It is found that most of the 5 HAPPI models considerably underestimate GB frequency; the best agreement with ERA-Interim GB frequency is seen in ECHAM6 and especially the MIROC5 models. Models also tend to underestimate CWB frequency, especially so MIROC5. An analysis of the relationship between CWB and GB occurrence shows that these two phenomena are closely associated in ERA-Interim and ECHAM6, but much less so in MIROC5, leading the authors to conclude that a different dynamical driver of GB must be acting in MIROC5.

Global climate models (GCMs) have long-standing biases in the representation of atmospheric blocking, which are still seen in the latest generation of GCMs. Studies into the nature of blocking (biases) in GCMs therefore remain important, and the present study illustrates an intriguing difference between two GCMs in how they represent GB and CWB. I therefore recommend the study for publication in WCD. I would like to make one suggestion for further analysis, other than that my comments are minor/presentational.

1)

The authors conclude that the dynamical link between CWB and GB is present but not the main ingredient triggering GB in MIROC5. This is, of course, very interesting and nicely illustrated in the paper but the question about the character of GB in MIROC5 is left open. Given the large negative CWB bias in MIROC, many blocking events will occur without CWC in that model.

I would like to suggest the following additional analysis: The authors have derived a set of days (or events) when CWC and GB occurs, respectively. This could be used to derive 2x2 contingency tables of co-occurrence of CWC and GB with frequency counts (No-No, No-Yes, Yes-No, Yes-Yes) for each of ECHAM, MIROC, and ERA-Interim. Furthermore, composites could then be shown with respect to three sets of these events (No-Yes, Yes-No, Yes-Yes), and potentially these could be more illustrative than the current Figures 5 and 6 – it is difficult to be sure, but I would like to encourage the authors to try if they have not already done so.

2) abstract

Make clear early on that this study is about the winter season.

3) page 1, line 23

Surely there are also more wide-ranging impacts of GB due to its association with the NAO and temperature anomalies across much of the Northern Hemisphere? Please add a brief discussion and some references; Chen et al. 2017 looks like a good start.

4) page 3, line 63

What does “enhanced” mean here? With respect to which reference? Maybe it can simply be omitted.

5) page 4, line 121

Typo: occur*s*

6) page 5, line 152

If I understand correctly, these 30 estimates are from overlapping time periods and therefore strongly dependent. Is this considered in the significance testing?

7) Section 3

It is nice to see from this section and the Supplement that the authors have conducted thorough model evaluation. A question that keeps coming up is what constitutes a “good enough” evaluation result to justify pursuing the main aim of the study. Where the authors clear about this before conducting the evaluation? I understand this question is hard and slightly philosophical, but, if possible, a short discussion of this point with respect to this study would be greatly appreciated!

8) page 7, line 208

“Similar standard deviation …” – similar to ERA or to each other?

9) page 7, line 215

“More frequent …” than what? Rephrase this sentence to make clearer that you are talking about a model bias.

10) Table 1

Given the central role of CWB, I suggest adding columns for CWB frequency to this table.

11) page 9, line 237

“zonal wind” – I would add the direction (“westerly”).

12) Section 4.2

Please consider if some of the supplemental material referred to could be promoted to the main manuscript.

13) page 10, line 291

Please add the CMIP6 results, which are now also available (Davini & D’Andrea, 2020).

14) page 11, line 305

Please add appropriate figure cross-reference(s) here (Figures 5 & 7, I think).

15) page 11, line 330

“for the reasons cited above” … please refer to the (sub)section here and/or add a brief reminder of what these reasons are.

16) Section 6

Please add a brief summary/conclusions from Section 5.

References

Chen, X., & Luo, D. (2017). Arctic sea ice decline and continental cold anomalies: Upstream and downstream effects of Greenland blocking. Geophysical Research Letters, 44(7), 3411–3419. https://doi.org/10.1002/2016GL072387

Davini, P., & D’Andrea, F. (2020). From CMIP3 to CMIP6: Northern Hemisphere Atmospheric Blocking Simulation in Present and Future Climate. Journal of Climate, 33(23), 10021–10038. https://doi.org/10.1175/JCLI-D-19-0862.1