Dear Authors,

You have responded to the referees' comments in a satisfactory manner, in most cases you already received a positive feedback from the referees in the discussion forum. Therefore: Please change your manuscript accordingly and submit the revised version (which you basically already provided as supplement to your author comments).

A comment from my side: Reading your paper I had to get used to the structure, which differs from the standard structure (Introduction - Data and Methods - Results - Disscussion and Conclusions). As none of the referees had a problem with it, you may keep it this way.

Yours sincerely,
Uwe Ehret

Dear Authors,

Your revised version from my perspective includes all adjustments requested by the referees.

The manuscript will be forwarded to the referees for approval and/or further comments.

Yours sincerely,

Uwe Ehret

Dear Authors,

All referees recommend publication of your updated manuscript.

Please do one more technical correction according to referee #2:

The paper is now greatly improved and in my opinion, as stated in the public discussion, ready for publication after a technical correction: the sentence in lines 10-11 of the new version is not true.

In fact, the extreme value theorem (Fisher-Tippett-Gnedenko theorem) states that in the case of independent AND IDENTICALLY DISTRIBUTED (not mentioned hypothesis) variables, IF the AM series converge, they can only converge to a GEV distribution - the theorem DOES NOT GRANT THE CONVERGENCE, i.e. the distribution may be not convergent! This aspect is very important in extreme value theory and cannot be neglected in the sentence. Moreover, the convergence to GP of the PoT series requires some additional hypotheses (that are not mentioned in the manuscript).

Given the introductory taste of the sentence, I think this comment can easily be addressed with a slight modification of the sentence to: "The extreme value theory showed that under some hypotheses - including independence - of the random variables, the AM and POT series can asymptotically converge only to the 3-parameters distributions known as GEV and GPD, respectively."