Dear Dr. Giacona and co-authors,
thank you for your responses to the reviews. I agree that the dataset you produced is very relevant and its first assessment deserves publication. With respect to the responses to reviewer 3 (statistics performed), your responses do not fully comply with the requests of the reviewer. I understand that you only want to use the Pearson correlation coefficient as an indicator for the relationships, while you basically agree that the reviewer is right about the advantages of using conditional probabilities. As I will send the revised version to the reviewer again, you should extend the statistical analysis section including clear statements about the reasoning and restrictions of your approach in the revision, and give explicit evidence for its basic suitability. With respect to the conditional probabilities, I suggest to compute and provide respective results for some key relationships as an additional reference for the readers.

Please provide a point-by-point response in addition to the revised version, also mentioning relevant comments the reviewers added to the text (not on language issues, which should simply be corrected).

Best regards,
Uwe Ulbrich

Dear Dr Giacona and Co-authors,
as you saw, the reviewer suggests the following alterations before the manuscript is published:
1. conditional means/probabilities instead of biserial point correlation,
2. keep the full categorical variables in the data set published instead of their representation
by a set of binaries (which you might do anyway),
3. rescale the moving averages to the same units as the observations.

I am sure that your paper would benefit from obaying all 3 points, but given your focus on making the dataset available rather than performing a thorough statistical analysis, I leave the decision to you to folllow these recommendations, except that I demand one example of a contigency table (or respective text) as suggested by the reviewer in point 1 ahead of your correlation analysis (p16, line 9). This is, demonstrating the conditional frequencies for the combination of flood events in the rivers mentioned in the subsequent text lines. From the contigency, you can then shift to your correlation approach.

Best regards,
Uwe Ulbrich

Dear Dr. Giacona and co-authors,
I am happy to finally accept your manuscript to publication in NHESS. Surely, there will be a lot of interest in the dataset and your introductory paper. I invite you to distribute the published paper freely once it is available through the NHESS web page after copy-editing and typesetting.
I would be happy to see more results based on the dataset in our journal!
Best regards,
Uwe Ulbrich