Dear authors,

Thank you for submitting your response to reviewers comments and a revised version of the manuscript.

I appreciate the use of ERA5 and additional discussion on differences between the successful and unsuccessful re-forecasts. Nevertheless, while I understand your limitations in using more datasets, I reckon adding at least one other model could have been valuable following a suggestion by Reviewer#2.

One additional comment that I have on this analysis is that not all observed anomalous polar vortex events are followed by a ‘canonical’ response in cyclone activity. For example, if only 8(10) out of 14 strong polar vortex events are followed by weakening of cyclone activity over Europe (North Atlantic), is it reasonable to expect that if a ‘canonical’ response is found in ERA5 then all reforecasts should also demonstrate the canonical response? Following on from that, I think it might be valuable if Fig. 9 indicated ERA5 cyclone anomalies by week unless ‘canonical’ response held for all 4 weeks (please change ‘SVs’ to strong polar vortex events’ for consistency).

I find it interesting that new Fig 10 shows a stronger polar vortex in forecasts both in SSW and Strong PV events leading to weaker SSW and even stronger (but displaced) Strong PV in models. Is that due to, e.g., a vortex bias in models?

Finally, Fig. 10ab shows that the cyclone mean anomalies in unsuccessful predictions diverge from mean successful forecasts within the first 2-4 days. Perhaps, in addition to 28-day average composites in panels c-j, you can explore what happens during those first days in unsuccessful forecasts? I think this will help further address the question ‘why’ raised by Reviewer#2 and support your claim that the state of the troposphere is important for successful prediction of SSW. Can you expand on this by, e.g., looking at the evolution of tropospheric jet or other variables that represent the tropospheric circulation?

I believe you can further improve the manuscript before it is sent back to reviewers, especially by looking more into stratospheric-tropospheric coupling following anomalous polar vortex events.

Kind regards,

Irina

Minor comments:

Fig.10 caption: I suggest: “Time evolution of cyclone frequency anomaly (in %), zonally averaged over the midlatitude North Atlantic,… Same as (c-f), but for geopotential height anomalies of the 100 hPa surface (Z’100; in gpdm). … indicated buy stippling’

Abstract: regarding the mention of health impacts and infrastructure damage due to cyclone activity in Europe, does it refer to the cyclone activity in general or following anomalous polar vortex events? Has it been demonstrated that anomalous polar vortex can be linked to sever wind events? If not, I would remove this statement from the abstract.

Dear authors,

Thank you for revising the manuscript in response to the reviewers' comments. Both reviewers were pleased with the changes and, importantly, reviewer #2 agreed that using one model would be acceptable.

However, they raised a couple of other major points regarding the sampling issues and dynamical interpretation. I think their suggestion of using bootstrapping is valuable. As for their other point, I can see that in addition to MSLP, you also looked at 850 hPa winds, but I agree that perhaps looking at, e.g., how the upper troposphere (e.g., U300) responds to anomalous polar vortex events may add further insights.

I'll be happy to accept the manuscript once these concerns are addressed.

Kind regards,
Irina

Dear authors,

Thank you for thoroughly addressing the comments made by two Reviewers.
I am pleased to let you know that your paper is now accepted for publication in WCD.

Kind regards,
Irina