Dear Authors,
the doubts on the lack of novelty raised by Ref. #1 are similar to those that prompted me to ask you to revise the abstract and the Introduction of your first submission (prior to the one published in the discussion).
Your reply at the time was convincing and I believe that the two novel points you mentioned:
“Note that while previous studies have provided comparisons of ensemble methods, none have explicitly studied their effects on high flow prediction accuracy – which is the focus of our research. In addition, previous studies have not systematically evaluated the effects of using a combination of resampling strategies with ensemble methods.”
indeed provide enough novel content to support the publication of your work on HESS, as suggested also by Ref#2 and Ref #3, who recommend a major revision but highlight the worth of the analyses.
In fact, additional guidelines on the use of resampling and/or ensemble techniques may be useful to many colleagues using data-driven models. Also the issue of the ability to reproduce high/extreme values (and of extrapolation) and of the heteroskedasticity in the errors are important for any kind of model.
The Introduction is improved in clarifying the novelty issues, but as highlighted by Ref#1’s comments, the stress on such points may be even stronger (especially in the Abstract, where there is too much focus on the algorithms, whose main features are familiar only to a niche of readers). Your reply in the discussion elaborate further on the novel content of the work, and I think you may revise the manuscript accordingly, also trying to attempt to infer some general interpretations that go further than the specific algorithms, in order to make your analysis and discussion more robust (see RC1-17).
All the referees have carefully read the manuscript and constructively suggested a list of necessary clarifications and some minor additional analyses, that you have already addressed and added in your comprehensive replies and supplements (even if unfortunately, too late in the discussion to allow them to provide their opinion on the additional work you presented).
I therefore invite you to submit the revised version (that is probably practically ready), that I will ask to one or two referees to revise.
Best regards,
Elena Toth

Dear Authors,
like Ref#2 and Ref#3, I thank you for having revised the manuscript, addressing also the suggestions on the need for better explaining the novelty of the work and adding additional discussion of the results (focussing on the concept of diversity-in-learning and the advantages of combined resampling - ensemble methods), as required by Referee #1.
Congratulations and thanks for having submitted your work to HESS,
Elena Toth