Dear Authors,
I would like to warmly thank the two Referees, who have uploaded their detailed and very constructive comments very timely and are both waiting for your revised version, to be written accordingly to the intentions you have listed in your replies.
Their comments were, as I expected, mainly positive on the content/value of presenting the proposed algorithm, but asking a major revision of the applications and of the presentation.
I keep thinking that you need a comparison not only with NSGA but also with MEAS, which is more similar in its concepts than the NSGA: perhaps our Referee Andreas Esfradiatis may help you in this.
And I look forward to see the application of both the open-source hydrological model and MORDOR over more than one catchment: since you chose option c) among the alternatives given by A. Esfradiatis, the paper needs an interpretation of the results you will obtain both changing the model and changing the catchment.
I agree with the referees on the need to improve/explain better:
- The 5 rules (section 2.1.1)
- The relationship between Caramel and the two ‘inspiring’ algorithms
- The metrics for assessing optimization algorithm (section 3)
- The meaning of the algorithm internal parameters/weights and the reasons for their choice (inside the Kursawe test function section, where their impact may be examined). Show impact of using different values for such parameters (not only ‘feelings’)
- The hydrological models (once you will have two): while GR is well-known and plenty of documentation is accessible, whereas for MORDOR you need to better explain how it works and especially its differences from GR.
- The catchments (once you will have two): their climate, dominant hydrological processes and differences (possibly chose catchment with different hydro-climatic regimes)
On A.Esfradiatis’ comment on p. 1, ll 24-25: it is not a matter of adding the references but of explaining the main reasons for applying a multi-objective calibration, that is not solely due to addressing the statistical properties of the errors.
On hydrological modelling performances (A.Esfradiatis’ comment on p. 7): I would suggest, rather than adding Nash-Sutcliff, to analyse the three separate components of the KGE.
I look forward to reading the revised version, that I will ask both referees to comment on.
Since we are asking a considerable amount of work, please let the Editorial Office know if you need more time for the revision phase.
Best regards,
Elena Toth

Dear Authors,
you did indeed a considerable amount of work in this revision, in particular adding, as suggested, the MEAS algorithm (very useful to assess the differences with Caramel) and also an additional case study, so that now we have two real-world basins and two models, which is needed for generalisation purposes.
I do approve that you have given to the paper a more clear-cut identity, focussing on the algorithm, even if personally find that now the algorithm description is too long (but let's wait for the referees' opinions on that: they may suggest you where/how to shorten it).
Since both referees kindly offered to help us again, reviewing the revised version (that is indeed subtantially improved), I am asking for thier opinion.
All the best,
Elena Toth

Editor Decision: Publish subject to revisions (further review by editor and referees) (31 Jan 2020) by Elena Toth
Comments to the Author:
Dear Authors,
you did indeed a considerable amount of work in this revision, in particular adding, as suggested, the MEAS algorithm (very useful to assess the differences with Caramel) and also an additional case study, so that now we have two real-world basins and two models, which is needed for generalisation purposes.
I do approve that you have given to the paper a more clear-cut identity, focussing on the algorithm, even if personally find that now the algorithm description is too long (but let's wait for the referees' opinions on that: they may suggest you where/how to shorten it).
Since both referees kindly offered to help us again, reviewing the revised version (that is indeed subtantially improved), I am asking for thier opinion.
All the best,
Elena Toth

Dear Authors,
as I wrote already, this revised work is indeed a significant step forward, and it is almost ready for publication,
Nonetheless the referees have some constructive suggestions for improving it even more: in addition to make some technical minor changes suggested by both referees, Dr. Andreas Efstratiadis asks to provide a wider review of the state-of-the-art and to add a short conclusive section, and I am sure you may add such parts. Andreas also believes that adding an analysis on a different benchmark would help the readers to better assess the value of your algorithm: I would invite to consider if it is possible for you to provide such different benchmark, unless it takes too much time.
I do invite you, therefore, to submit a last revision trying to address such comments, in order to make your work even more appreciated by the HESS readers.
Best regards,
Elena Toth

Dear Authors,
many thanks for having clarified/integrated the last few points as suggested by the Authors.
We fully understand that, especially due to this difficult period, it is not possible for you to perform additional computations, that may be carried out ion a future study.
The phrase at l. 26 p. 2 is still not clear, due to the English phrasing: i would suggest:
"Evolutionary algorithms are advantageous not only because...."?
Many thanks for having shared your algorithm with the HESS community: I hope it will be widely used.
Best regards,
Elena