Dear Authors,

The paper “How can expert knowledge increase the realism of conceptual hydrological models? A case study in the Swiss Pre-Alps” was reviewed by three referees. Referees #1 and #2 have submitted their review in time and the authors have also submitted their responses to the two referees’ comments. Referee #3 sent his evaluation after the deadline, and I decided to join the referee’s comments on the bottom of this report.

All three referees find the paper interesting, well written with valuable results. Referee #1 raised important issues concerning several instances with questionable gaps in the logic and some critical parts of the method description. Both Referees #2 and #3 advise publication after minor revision. The Authors gave convincing responses to the Referees’ comments.

The paper will be reconsidered after “Major revisions”. In the revised version, authors are invited to respond point by point to each comment of each referee, giving convincing proofs to all questions raised by referees. Please highlight clearly what you changed in the revised manuscript, so the referees are able to assess your changes.

Kind regards,

Roger Moussa

____________________________________________________________
Referee #3

Review comments

The article questions the added value of expert knowledge in the simulation of the rainfall-streamflow relationship by analyzing the sensitivity of a process-oriented model to the mapping of dominant runoff processes. The analysis is made on four nested catchments in the Swiss Alps.

I found this is an interesting article, which is clear, easy to follow and well written. I have some requests for a few clarifications and changes in the text. I advise publication after minor modifications.


Specific comments:

1. Introduction: The authors much emphasize the issue of spatial process mapping. They do not mention much the issue of time scale of these processes. Are the time scales of dominant processes well represented in the model? Though this can be a cause of model failure, this is often a neglected problem. I know this is not the main focus of the article, but this issue of time appear several times in the article (e.g. P5,L5). The literature on this may be shortly discussed somewhere in the article, in the introduction or the discussion (see e.g. work by Savenije and others).

2. P4,L22: Maybe show these administrative limits in the first map of Fig. 2 (if it does not make the map too messy) .

3. P5,L11-17: This part may be presented with bullet points, to ease reading.

4. P5,L19: “SF07” is used before it is defined.

5. P6,L3-8: Some parameters do not appear in Fig. 4 and could be added to better follow the model description.

6. P6,L10-13: Is this study made in a forecasting context? If yes, this could be mentioned since it may have an importance in the results, given the role of data assimilation.

7. P8,L29: I was a bit concerned by the fact that only six events are used to make the tests. This is not sufficient to draw general conclusion. Could the authors make an additional test in their discussion section, in which a larger number of events would be tested, to see if this corroborates the detailed results on a few events? I think this would be helpful. If this is not feasible due to data or time constraints, this aspect should be better discussed. What is the feeling of the authors on possible results in other conditions/catchments?

8. P9,L20: A ideal P-factor equal to 1 means that you assume that the confidence intervals built in the modelling process with parameter ensembles are 100% confidence intervals. This may be mentioned since this is not obvious. Predictive uncertainty bounds may be built to contain e.g. 80% or 90% of the observations.

9. P10,L20, P14,L15-20 and P15,L11-14: The Trueb catchment seems to be very special. Precipitation and streamflow are partly blamed for that in the article. Would there be a way to confirm these hypotheses based on the data used?

10. P13,L1-2: My question is: what would be the objective of such a detailed representation if one had the knowledge and computational efficiency? It is as if one would like to make a map of the catchment at the 1:1 scale. Is there any use/objective for that?!

11. P13,L3-10: I missed a clear definition of “realism” in the article. The authors make a parallel between realism and the experimentalist expectation. I would say that the experimentalist has his own model in mind, based on his knowledge, experience, etc. Therefore, confronting the numerical model with experimentalist expectation is no more than confronting two models. Here one expects that the experimentalist better knows how the catchments behaves, but the experimentalist, like any other model, may be wrong also! This issue of defining realism is crucial, but I find that it should be based on objective measures. I think this would need a more in-depth discussion.

12. P15,L10-11: I found this sentence unclear.

13. P16,L4-7: I do agree that this is a strong limitation of this study, which prevents from reaching more general conclusions. I think the authors should try to show tests at least on a larger number of events on their catchments if possible (see comment above). I also think that a short sentence should be added in the abstract to acknowledge this limitation in the conclusion of the study.

14. Table 3: Write “for the bottom-up and top-down” in the caption to keep the same order as the result tables.

15. Table 4: “THY” in the caption and “THI” in the table, to be harmonized. Maybe “No. of ground rain stations…”.

16. Fig. 3: Maybe use the same order of maps as in table 1.

17. Fig. 4: If possible, put “-1” as exponent in the unit.

18. Fig. 8: Even if there are negative values for the Trueb catchment, I think that the median values of distributions not appearing on the graphs could be written in the graphs (with small arrows showing that this is down the bounds).

19. Fig. 9: “the four study catchments”

PS: I didn’t find in the main text references to Figures 2 and 3.
_________________________________________________

Dear Authors,

The revised version gives responses to the main recommendations of both reviewers.

However, some minor recommendations are formulated by Reviewer #2. Please also introduce more discussions in the paper on the points you rejected following the initial recommendations of Reviewer #1.

Kind regards,

Roger Moussa

Dear Authors,

The first version of the revised version gives responses to the main recommendations of both reviewers. Hoever, in the second revised version I didn't find:

- corrections and responses to the recommendations of Reviewer #2 (see below).

- more discussions in the paper on the points you rejected following the initial recommendations of Reviewer #1.

Kind regards,

Roger Moussa


____________________
Reviewer #2: Review of "How can expert knowledge increase the realism of conceptual hydrological models? A case study in the Swiss Pre-Alps" HESS-2017-322 by Antonetti and Zappa

The authors have provided a detailed response to the issues raised in the original review.

The only aspect that remained confusing and in my opinion warrants better clarification is with respect to clarifying "expert knowledge" in hydrology, and contrasting it to other types of knowledge (potentially) available to a hydrologist.

The response says that: "In the revised manuscript we will therefore add a definition of 'expert knowledge' at the beginning of the introductory section and will distinguish in a clearer way between (i) expert knowledge per se and (ii) (top-down or bottom-up) strategies for the use of expert knowledge in conceptual models".

The corresponding change I am seeing is the addition of "(i.e. someone's acquaintance as a result of study and experience)" in the first sentence of the introduction, on Line 5.

To me as a hydrologist this very brief definition is only partially helpful. Under this definition, what other way is there to gain knowledge other than through "study and experience"?

If the manuscript singles out "expert" knowledge as its central study topic, it should provide a clear definition in the context of hydrological sciences and some examples of what IS and what IS NOT expert knowledge. The author's responses that "As acquaintance is uncountable by definition, information to be considered as 'expert knowledge' cannot be listed" and "it [expert knowledge] should not be contrasted to other kinds of knowledge" do not seem convincing to me, at least not in an international research journal in hydrology.

For example, if the authors can contrast "modellers" vs "experimentalists" - despite these not being truly sharp distinctions, what is so special about "expert" knowledge that defies contrasting it to other type of knowledge? Just because it has not been done in previous cited publications is not really a scientific reason to not do it here. Perhaps there is no clear definition and in fact all knowledge is "expert knowledge" - in which case this might be noted in the paper (intro/discussion).

I therefore recommend a revision of the manuscript to clarify this issue.

This request is not intended as a nit-picking or otherwise "tricky" comment - just a request as a reader/reviewer to clarify one of the central concepts in this work.

Dear Authors,

Thank you for the additional modifications and for your response to Reviewer #2. The paper can now be published.

Kind regards,

Roger Moussa