Referees provided really useful and constructive comments. The authors' replies are detailed. The paper, and the authors replies, will be sent back to referees.

I found all referees's comments useful, and the auhotrs response shows they are preapred to revise the paper. They are invited to do so, answering to all referees' comments. After it is done, the revised paper, with the authors replies, will be sent back to referees.

Dear authors,
Today I have received finally the comments from Referee #2. The referees provide comments which mean "revision". In your replies please exaplain how have you addressed the comments (esp. this concerns Referee #1 where there may have been misunderstanding, e.g. w.r.t. answering the comment #2 - it was mentioned in you email of 8 Sept.)
W.r.t. comments of the Referee #2 please try to explain your position in a convincing way. I am not sure (re)doing so much experiments is possible at this stage, but his/her commens have reasonable grounds and have to be addressed. To some extent this can be done by clearly (re)formulating limitations and correspnding recommendations.

Please address the referees' comments. Referee 2 raises three points. The third is about the need to reflect the performance at extremes. Please provide the reasoning why do you think your metirc does that (or not), and why. Please also address the other two points. Please ensure that your explanations to the referees are also reflected in the text.
I am inviting the auhtors to review English again - please read the paper attentively... there are still quite a number of missing words and obvious grammatical mistakes. Please see referee 1's comments in the text.

Dear authors,
The comments of the reviewers are attached.

Referee #2:
The comment on using CDF-based criteria instead of (removing) deterministic error-based (RMSE): yes, I understand the reasons for this comment (and I think you as well). However, RMSE provides improtant information about the models performance even if the models are based on uncertain data or scenarios. I see no problem of using RMSE if its use is justified.

To answer the referee's comment, I suggest to include a bit stronger justification for using the model skills (7-10) against the CDF-based, with the references to literature on this point. It would be also good to include the use of CDF-based metrics into the "recommendation" part in section 6 (if this does not contradict your understanding).

Indeed RMSE formula (8) is wrong - division by n is forgotten.
Bias (Eq 9): typically PBIAS is used which not a difference which you use, but a ratio
PBIAS = 100 * [ sum( sim - obs ) / sum( obs ) ]
would be useful to refer to a source that suggests and justifies use of formula (9).

In sec. 6 I would also clearly state the limitations of this work, and strengthen the "outlook" (future work, and there is enough to say here) which is now just one sentence.

Referee #1:
Indeed, it never hurts to scan the paper again to polish English.


inter model --> inter-model
Inter Tropical Convergence Zone --> Intertropical Convergence Zone [this is a more widely sprad variant of formulation] (or inter-tropical)
down scaling --> downscaling

I hope you can quickly address these (minor) issues. After your revision, the paper will -not- be sent to referees, but will be reviewed by the editor only.

Thank you.

Thank you very much for addressing all the comments! A very good paper.