Dear authors,

both reviewers are highly critical of the manuscript with one suggesting major revisions and one suggesting rejection. The authors' reply proposes substantial changes that may address most of the posed criticism and so I invite you to submit a revised manuscript that carefully addresses all points raised. The revised manuscript will be sent out for review again.

Best regards,
Jakob Zscheischler

Dear Authors,

your revised manuscript has now been seen by the original two reviewers. Both see merit in the revision but still have a number of concerns that need to be addressed before the manuscript can be considered for publication. I therefore invite you to resubmit a point-by-point response to all reviewer comments together with a revised version of the manuscript.

Best regards,
Jakob Zscheischler

Dear authors,

your revised paper has now been seen by the original two reviewers. While they are generally supportive of some of the changes made in the paper, some substantial comments remain, see the attached comments. Please pay particular attention to the newly added part on the CMIP5 models and its integration into the paper.

Kind regards,
Jakob Zscheischler

Dear Authors,
thanks for submitting yet another revised version of this manuscript. There still remain some open issues that need to be addressed before the paper can be considered for publication. Please see the attached report.
Best regards,
Jakob Zscheischler

Dear Authors,
thank you for addressing the remaining reviewer comments. After a careful reading by myself I have a few more comments that I would like to ask you to address before the manuscript can be accepted for publication.
The comments are pasted below this message.
Best regards,
Jakob Zscheischler

L245: why does p need a subscript 1? Since there is no p_2, consider removing the subscript. Currently the subscript leads to some confusion e.g. in the first line of Table 1.
L248: shouldn’t \pi have the subscript i,r instead of 1,r, with i=1,2 for the two clusters?
L249: step 7, the definition of cold spell analogues is unclear. The equation is incorrect when p1 != \pi_{1,r} and does not provide a selection of events. I guess you mean something like analogues should fulfil g_{t,r}^{Z500} >= g_m^{Z500} where g_m^{Z500} is the p_1 th quantile of g_{t,r}^{Z500} in the modelled fields, i.e. P(g_{t,r}^{Z500} >= g_m^{Z500}) = p_1. Please adjust accordingly.
Caption of Table 1: In step 7 in L249 cold spell analogues where defined to always have the same occurrence probability p1. The use of “analogues” for different sets of event sets thus creates ambiguity. Please clearly state for which analysis which set of analogues is used and consider using different terminology. (This point has already led to some confusion in the last round of reviews.)
L274: I assume the analogues used here are those taken from step 7 in L249, which would be fewer than the ones summarised in Table 1? Or where have the analogues that were selected in step 7 been used otherwise? Please clarify.

Dear Authors,

thank you for addressing my remaining comments. I'm now happy to accept your manuscript for publication in ESD.

Best regards,
Jakob Zscheischler