Dear Authors,

I have read the reviewer comments (plus my own) and the author responses in detail. I think we are nearly there and I appreciate everyone’s time and effort on this manuscript. Both reviewers were positive about the manuscripts publication (as am I) but we need to make sure the paper has the right context and discussion and is technically sound. Here are some final comments (I will want to check these but the paper will not go out for review again) that I think should be better defined to the reader for clarity. Please do think of these of how interested I am in the paper than just sending the authors more to do!:

1) In the abstract, you end with: “Setting up a global catchment model has to be a long-term commitment of continuous model refinements to achieve more useful results for water management”. - I’m not sure refinement is the right word at this stage, the model has really been more crudely applied to the GRDC data with limited discussion or evaluation of reasons for model failures or refinement or the quality of the GRDC data themselves. So I think this sentence needs to be improved. I’m also not sure there is any proof that these results are ‘more useful’ even to water managers, who would normally not be interested in a global model but one that they understand and is tailored for their local area or have resources to go with it. Perhaps, eventually, such global models may have utility to those users who have limited or no local resources but that is not this model or close to it. Other models, such as WaterWorld are far more focused (including a network of training resources and funded workshops) to help such limited resource managers, see https://iwaponline.com/hr/article/44/5/748/1145/WaterWorld-a-self-parameterising-physically-based. If the authors wish to persue this idea of being relevant to water managers then I think they should cite and discuss such other initiatives tht have been around a long time.

2) I probably agree with one reviewer that certainly the intro is not ideal, in particular the first paragraph is very general and lacks references to prove such statements as “they have high credibility among practitioners and water managers”. The start could be much better developed to discuss where the community is with large sample hydrology, which ultimately this paper sits under (again simply the model is basically calibrated to a global sub-sample of river flow gauges). There are also more recent papers that have made simulations at continental scales and/or involved in large sample hydrology which should be improved.

3) When addressing the novelty of the current paper the authors should state what they mean by “relatively high resolution” in terms of the first global application on line 125. They should also note here for other ‘lower resolutions’ what has therefore been achieved and thus where this study fits in terms of a resolution development that is novel. I am certainly not clear on what the authors are specifically trying to claim here….

4) I think the authors miss-understood my comment about preserving ‘mass balance’, I just want to clarify this. My understanding is that HYPE allows for a flux in the GW stores that (from the WIKI manual) allows “Regional groundwater flow to outside model system”. What I take this to be effectively a ‘loss function’ meaning that this water is never seen again in the model domain. It’s important to understand for such models that such ‘mass balance losses’ can occur as not every model works on that principle (and really it should be understood for different environments how much water is needing to be lost to gain ‘good’ predictions). Of course I expect the model to maintain ‘mass balance’ in terms of the numeric’s being correct in each store or function, else the model would have coding errors.

5) Please better explain the reasons for the choices of catchments that were involved in calibration and those for independent model evaluation (even if purely random) in lines 242-244

6) The section on model setup needs to better explain how many parameters are being estimated, or it can be included in the parameter estimation section. It seems that globally if I have this correct there are separate parameter estimations for 36 different classes (plus some others for lakes), but for each of those the text does not state the number of parameters being tuned in each. I would also note that nowhere is it clear that although a 50km grid is being used for the forcing data I don’t believe the paper explains how that is applied to each HRU or catchment size. So is this then ‘averaged’ over the HRU if the HRU is larger and is the forcing data fractionally weighted to properly reflect the catchment boundary delineation compared to the actual gridded data. Lines 277-278 sort of point to an explanation but it isn’t 100% clear.

7) In catchment setup please give more statistics beyond the mean as to what catchment sizes are being run (as in say the 5th and 95th catchment size that each HRU is being run, and so to understand better what is meant by a certain resolution and that variation). Lines 320-321 do not resolve the statistics of HRU sizes at all well!

8) With regards to the approach to model calibration and evaluation there is an inconsistency with the stated number of catchments used in the model calibration and evaluation earlier in the paper and then those reported in lines 421-423. At least this is confusing as more catchments are brought into the calibration pool. This needs to be better explained please and this inconsistency removed.

9) To me an important part of the paper is to understand why we might treat different ‘physiographic’ areas differently (it basically is core to the whole results) and if this is justified in any hydrologically meaningful way (and of course there will always be multiple ways to do this, none perfect). However the authors only state they made 36 classes using ESA CCI 1.6 data, but no explanation is given as to why/how. I think the reader needs to understand ‘why 36 classes’. For example I appreciate this is a paper which I am a co-author on (sorry) but the authors might be interested in some of the implications of what has recently been shown in Knoben, W.J.M., Woods, R.A. and Freer, J.E. (2018) A Quantitative Hydrological Climate Classification Evaluated With Independent Streamflow Data. Water Resources Research 54(7), 5088-5109 and why Köppen-Geiger may not be the best separator hydrologically - for example.

10) Referring to point 5) above the reader has no way of knowing if the model evaluation was truly ‘independent’, in that if sub basins were involved in the calibration then they would not be independent, hence why a clarification of the choice of catchments and how this was separated is needed.

11) It’s great to have flow signatures, please explain why this whole table 5 is being reported and better explain the authors comment that help the modeler to examine whether the process description and model structure are valid across 485 the landscape or if the regionalization of parameter values must be reconsidered for some parts of a 486 large domain”. This is critical to inform the reader how this is approached.

12) I am not clear on what evidence presented the authors can so categorically state “WWH version 1.3 describes major hydrological features globally and important spatial variability in factors controlling the runoff mechanisms”. That seems a little bold for me on the rest of the evidence submitted and the call from one reviewer to be ‘more modest’. In my previous editor comments I stated: I also feel that there needs to be a better evaluation as to why certain regions behave well or not in terms of increasing a scientific interpretation in the paper to be novel for publication… But really the authors have not changed their approach to add any form of evaluation (only a small change to the discussion) but yet still they have this statement in the results. They either need to make some form of analyses that really reflects this concept that they maintain that the flow signatures effectively help diagnostics to understand if the model has the right processes or they change the way that they comment on what the results can (and do) say. We all know the GRDC gauges used have major impacts beyond those that are natural and this is bound to cause significant problems in interpreting any results. The authors make very little mention of this throughout the results and I do think there needs to be improvements to self reflection as to where the model fails (or reasons why it might fail). Figure 7, for example, is not usefully discussed in any way that helps the reader interpret any form of ‘process’ reasons why the model is failing. This does not require some analyses to earth observation, there are more immediate and tangible things that could be explored in the results just to see reasons for main model failures purely on stream flow that would provide some useful insights to the reader! (else this is a numerical fitting exercise only, but that is not what the authors are trying to state).

13) Figure 6 shows some lack of understanding of what the KGE metric scales to and needs to be adjusted so that it is technically correct. The current figure title suggests that they believe KGE at zero scales in the same way as Nash-Sutcliffe (all values above zero are better than the long term observed mean). But that is not the case. I’m sorry again to highlight another of our recent papers but it is very pertinent to this comment. See https://www.hydrol-earth-syst-sci-discuss.net/hess-2019-327/ where Knoben et al. show that it is actually -0.41 that is equivalent to the long term observed mean and other matters with interpreting KGE scores.

14) I am not sure the authors can state that “the model provides a first platform for catchment modelling to be further refined and experimented with at the global, regional and local scales”. Again I have noted the work of Mark Mulligan and this may well not be the only such platform. Please ensure the correct context is given here….

Thanks very much to all, Jim

The authors have done a very good job of their final responses and both the reviewers were enthusiastic as am I of the paper for publication. Therefore the paper can move to publication. Thank you to the authors and the reviewers for all their time and expertise on this paper.... Jim