The paper analyzes the sensitivity of a semi-distributed urban drainage model to spatial and temporal resolutions of precipitation input by weather radar data. The application is to an urban catchment in Rotterdam with 4 storms. The authors make an attempt to generalize their results by estimating non-dimensional measures such as the catchment, runoff and sewer sampling numbers for different combinations of rainfall resolution, and studying their effects of metrics of peak flow depth and discharge in the sewer system normalized to a reference spatial resolution.

The paper presents a methodology which aims to identify the uncertainties/errors in predictions which are generated by rainfall input resolution and catchment and urban drainage system discretization. It is clear that the method needs to be tested on more events, and the authors acknowledge this fact.

The paper was reviewed by 4 referees, all of whom are experts in the topic covered by the paper. The reviews were quite critical, 3 of the 4 suggested major revisions or rejection of the paper, 1 called for a minor revision. The comments of the referees were constructive, and I can read between the lines the value they see in the analysis. I recommend therefore to the authors a major revision of their work which will address very clearly the concerns of the referees, in line with the posted author’s responses. Two of the reviewers have requested to review the revised paper, so there will be another fast review round of the final revised submission.

I would especially like to highlight the following key focus points for the revision.

1/ It is clear that 4 events do not constitute a sufficient database to generate statistically sound conclusions, from the start please clearly explain what you aim to achieve by studying only 4 events and exactly why you chose those 4 events (like you do in your response). Please also add more information on the suitability and/or limitations of the semi-distributed rainfall-runoff model rather than a fully-distributed approach in this context (Ref 1), and certainly present details on the catchment delineation you list in your response.

2/ Like practically all referees I also got confused with the notation of the sampling numbers and different resolutions. In your revision please make sure that these are very clearly and consistently used throughout the manuscript.

3/ Ref 2 raised two important points which you should engage with in the introduction. The first is the modelling approach to study sensitivity in circumstances where the validity of the approach itself (rainfall-runoff model) cannot be properly validated. Ref 4 also pointed out the fact that the runoff model has its own uncertainties. The second point is a warning about motivating this study with the issues of climate change leading to a potential increase in extremes. I share this viewpoint and encourage you to make sure that your thinking on the first point (as you outlined in the response) shows up prominently in the introduction and in discussing your results.

4/ Think of re-drawing Fig 5 more than just changing the scales. Ref 1 pointed to the larger variability in event 1 which is lost. I would add that the interesting values, i.e. whether the median is <>1 is not clear at all for events 1-3, only for event 4. I agree with Ref 1 that Fig 7 is a key element of your message and I also did not fully grasp your message there. I suggest that you more clearly explain in the text the increase in variability that comes as you sample precipitation at grid resolutions where precipitation intensity becomes more-and-more independent in space.

5/ There is one aspect of rainfall spatial resolution which I warrants more discussion, going along the lines of Ref 3 on storm location: that is the sampling effect in the aggregation. Starting from a “raw” pixel resolution of say 100 m which is what the instrument measures, there are 25 ways in which a new grid for aggregated precipitation at 500 m may be defined, of which you just chose 1 as far as I understand. This may (or may not) include large uncertainty in the statistics for peak depth and flow. What is the author’s opinion on this?

6/ I appreciate your focus is on spatial effects rather than temporal ones. However you do show some temporal resolution effects as well. If you decide to keep this section, please discuss the causes for the negative delays (time shift) going from 1 min to coarser time steps in Figs 11 and 12. Is this also only a sampling issue? Otherwise it is counter-intuitive with respect to your arguments for smoothing of runoff response by aggregation.

7/ Figure 9 has discharge not water depths.

Peter Molnar

Dear Authors,

Two of the original referees reviewed your revised manuscript and request minor revisions. Please read their comments. Especially Referee 2 raises some points which you should act upon. Please send me an itemized response to those points, so it is clear how they were dealt with in your final revision. If you disagree with the Referee please explain why. Thank you.

Best regards, Peter Molnar

===
Referee 1

The manuscript is much improved, there are a few small errors that would need to be corrected/queries that would need to be explained:
Table 1 - runoff coefficient and storage coefficient: the ',' needs to be replaced by '.' i.e. 0.2 instead of 0,2 etc.
Figure 2 - Event 2, please check if the correct version of this Figure is included? Has the radar location moved in relation to the catchment, or was the black square in the figure accidentally moved somehow? (i.e.,event 2, meridional distance between ca 2 and 5 km, instead of between ca -4 and -7 km for events 1 and 3). Please check and either correct the Figure or explain the Figure in more detail.
Page 15, sentence on lines 8 - 10 (Results...water depths) needs revising as it is currently not clear.
Page 20 line 20: should be 0.05 instead of 0,05m
Please double check references list, i.e. page 24 lines 17, 21 & page 25 lines 5, 11, 14 & 17 - all 'A' seems to have been replaced by 'a'
In the appendix, please check for consistency: is the 'runoff factor' or 'c', the same as the 'runoff coefficient'? And please define 'h'.
===
Referee 2

The revised version exhibits some significant improvements which are satisfactory. However all points were not addressed, and I believe that some minor modifications are still needed before the paper is suitable for publication in HESS.

General comments:
- Some comments of the reviewers or the editor are said to be mentioned in the answer of the authors but are not really addressed in the new manuscript (ex: EC3 - the potential uncertainty of models are not really discussed in the text, neither climate change in the introduction; EC5 – the answer to this is not found in the text). This is quite annoying and should be changed in the revised manuscript (reading again the comments on the first version on the manuscript)
- I still believe it would be relevant to include one hydrograph for at least one event with flow simulated for various resolutions. It would complement well the discussion based on relative values (because as mentioned in the discussion, the highest relative deviations are found for the event 4 which is generating the less runoff).
- Few words on the model validation should be added, to justify the fact that it is legitimate to use it for such a study.

Detailed comments:
1) Introduction
p.4 l.17-21 : given that there is a precise list of questions, we would expect a similar presentation in the conclusion to answer them. l.18 “urban drainage model” is too general in the question, may be you should use “highly detailed semi-distributed model”.
2) Presentation of the case study
p.5 l.19-20 : a map of the catchment with all the runoff areas (the location of the study area within the Rotterdam region could also be indicated in this figure) could be helpful. More quantitative information is given anyway after with the “model lengths” section.
2.2 rainfall data : I am not sure, but I do not think Rotterdam is within the range of the CESAR radar so you should explain how you selected the portion of studied rainfall event and where you implemented it over the catchment (i.e. where and why did you apply the heaviest rainfall).
3) Method
p.8 l.11 : “spatially aggregated”, how was it done? Simply the mean over various pixels or the aggregation is done on radar data before implementing a Z-R relationship. I think it should be explained.
Eq. 1 : is it done on rainfall rate or radar reflectivity (if not the Z is a bit confusing and should be modified).
Figure 4 : indicate what represent the axis
p.8 l.27 : “the magnitude of the distance”; this expression is not clear. Please change.
Dimensionless parameters: include systematically the equation to help the reader
p.11 l.15-16 : use “length” rather than “resolution” because the parameters were introduced as lengths.
3.2.4 : l.28 : “which corresponds to the inverse of a sewer density”. I do not understand this, given definition of the sewer density in 3.1.3. How two quantities with the same physical dimension (a length) can be inverse? Why defining the sewer length LS before and not using it?
4) Results and discussion
p.13 l.7-17 : comments and physical interpretation that would ease the reading should also be added, only stating the figures is not enough.
p.13 l.22-23 : it would help the reader if you just said that it was obtained for various rainfall resolution and that no hydrological modelling are used. Figure 5 : the “vs.” are reversed in the caption and not in the manuscript (other figure captions should also be checked). The legend says mean over the pixels, and the text over the catchment please clarify. From my understanding it is not “standard deviation of normalized rainfall volume” but “normalized standard deviations of rainfall volumes” (as said after by the way). Authors should be much more careful with the words; such imprecise vocabulary make it much harder to understand, and this was already mentioned in the previous reviews !!
p.13 l.28-30 : as said in my previous review (and this has not been addressed), the decrease is due to the artificial transfer of rainfall outside the boundary of the catchment. It could be the other way if the heaviest portion of the storm was outside. So I am not sure that the tendency is really relevant, but the order of magnitude are and this should be stressed. By the way this effect is explained at the sub-catchment scale in sections 4.1.4 and 4.1.5, so it should also be discussed here as well. I think is should also be changed in the abstract.
p.14 l.15-16 : “a more pronounced internal spatial structure” what does it mean and how did you obtained that (there is only a 50m differences in storm decorrelation length). The expression is also used after in the section and the meaning should be explained there as well.
p.14 l.21 : a clear reduction “of” the median
p.15 l.9 : check sentence because not clear
4.1.3 : given that the results are used in the discussion of the previous sub-section (which uses the storm structure), I would suggest to move it before. Figure 8: include the event number in the figure (the “going clockwise” of the caption is not “user friendly” for the reader)
4.1.4 : the “in-sewer” in the title is a bit confusing because it is also the case in section 4.1.2
p.17 l.14-15 : “deviations ... increasing” is not clear. Are you mentioning a monotonic trend in the curves ?
p.17 l.28 : it is not clear to me how the threshold of 0.5 was obtained (I already mentioned this issue in my previous review and this was not addressed).
4.1.5 : again from Figure 10, the threshold of 20 mentioned in the text seems rather arbitrary and should be justified (not obvious from the rather steady trends Fig. 10) or simply not used.
p.18 l.17 : 900m of pipes per 100m2 seems rather high, is not it 100m*100m
===