Review of Hamitouche et al.

The authors offer a regionally-focused comparison of two hydrological routing models, CaMa-Flood and WRF-Hydro, using default parameters, as well as a calibrated WRF-Hydro. The paper is a nice contribution to the Mediterranean region hydro modeling scientific literature. I suggest Major Revision, with three Major comments and several Minor comments.

Major comments.

1. Compare results with a baseline. To show improvement in river discharge simulations, it would be helpful to compare the results with an existing baseline/benchmark. For instance, in the introduction, part of the motivation was that the HD model was underestimating discharge in the Med. Similarly, CaMa and WRF-Hydro also tended to underestimate discharge. Is there any way to see if either model actually offered an improvement? Even if it is qualitative? If not, this point should still be mentioned as a limitation of the work.
2. Integrate the calibration results. As written and presented, the calibrated WRF-Hydro results seem like a separate add-on, rather than an integrated component. Some integration occurred starting on line 466, but this was hard to follow.
3. Improve clarity, organization, and flow. The authors include a lot of good content, but sometimes it was hard to follow and assumed prior familiarity. Specific suggestions for this are offered in the Minor Comments listed line-by-line, below.

Minor comments

Line 25. Suggest: “Not only does it provide essential freshwater input…”

Line 30: Suggest tightening up this paragraph to better link with previous paragraph, which was only about the importance of the regional discharges in the Med Sea. Suggest revising this first sentence (also not sure if the Nearing reference fits, since that ref is more about AI-forecasting). Maybe: “Timely and accurate river discharge estimates into the Mediterranean are critical for managing water resources and risks in the region.” Then second sentence could merge with the 3^rd to be, “In a broader context, understanding the interplay between hydrological processes and regional climate is essential, and underscores the need to study the coupling of ….“

Line 56. Suggest replacing “Nowadays”. Maybe something about Recent advances?

Line 60-64. This has important information but is hard to follow for someone not very familiar with these models; suggest revising for clarity. What is HD? How do the different horizonal resolutions factor in? What is HydroPy LSM?  

Line 75. About this sentence -> “This highlights the importance of conducting detailed sensitivity analyses on standalone routing models, before adding them into coupled climate or Earth system models.” <- I think this is too general, and what is the sensitivity analysis here? Seems more like a comparative evaluation. Also, can you link this to a specific regional model or modelling experiment, such as the Med-CORDEX experiment, rather than the general coupled ESMs? Or later when you say “Euro-Mediterranean regional coupled models”?

Line 84. This is the first time you mention Med-CORDEX. Even though you define CORDEX earlier, you assume that the reader is familiar with this particular experiment. Suggest adding more context for the reader. Maybe also point to Figure 1 here.   

Line 91-93. The calibration seems like an add-on as written. Can you better integrate this into your experimental design? Why don’t you do calibration experiments on CaMa-Flood?

Line 95-96. Going forward, to show improvement, it’s best to compare to an existing baseline (such as the Med-CORDEX simulations, or the underestimating HD model you talk about earlier).

Line 116: This paragraph doesn’t fit as-is under the 2.1 heading (Study area and river discharge observations). If it does stay, then this is the first time the ENEA-REG model is mentioned, and the reader needs it to be defined, and more information about what it is. Looks like you don’t say it’s the atmosphere-land-ocean model until line 135 – should this information be moved to that section?  Line 119: You mention the Med-CORDEX protocol – should some of the protocol info be mentioned in the introduction to help the reader with background?

Line 188: Why is the CaMa model not calibrated and compared during this study? Do you only calibrate wrf-hydro since the PyWrfHydroCalib package is available?

Figure 2, Figure 3, Table 2 etc: Is there a reason to not present all the results together? So include the three (i) default CaMa, (ii) default WRF-Hydro, and (iii) calib WRF-Hydro?

Line 310. You mentioned earlier that the HD model also tends to underestimate discharge. Is there a way to compare how much HD underestimates vs these models? It can help if there is a prior baseline to compare to, even if only qualitatively.

Line 434. This seems better to either go earlier (after Figure 5 and the KGE results) or in a discussion section, otherwise it seems out of place and hard to follow.

Line 464. This is a new topic and I recommend adding a new subsection here, and some introductory sentences to guide the reader on what was done and why. Also, I’m not following what is being shown in Table S5. Line 468 – I suggest reminding the reader that the ENEA-REG model is downscaled with WRF. The Table S5 caption needs to also remind the reader that WRF is downscaling the ENEA-REG. I had trouble crosswalking what was in the text and in the supplemental Table S5, suggest revising. Is there a better/clearer way to show the results in Table S5, and/or should it be brought into the main manuscript rather than the Supplemental?

Line 493. I don’t see a Fig S7.

Line 495. Would be good to recognize that including a baseline for Euro-Med models would be helpful to see if there was an improvement.    

The paper by Hamitouche et al. presents the performance of two hydrological models, namely CaMa-Flood and WRF-Hydro, driven by the ENEA-REG atmosphere-land-ocean coupled model run at a 12km resolution over the Med-Cordex domain. The performances analyzed concern default (uncalibrated) versions of both models, as well as a calibrated version of WRF-Hydro. While the authors highlight the higher performance of the calibrated version of WRF-Hydro, they present this work (to my understanding) as a preliminary step for further application “for offline atmosphere-hydrological simulations to close the water cycle at the land-ocean interface” (L526).

In brief, the key findings presented in this paper are the performances of two uncalibrated models and one calibrated model, which utilize data from an offline climate model as input, on a series of Mediterranean basins. These findings can be, of course, useful, but presented alone, rather than well inserted in a workflow providing much more advanced results, are no more than calibration exercises or, in the case of the uncalibrated models, simple applications of already existing hydrological models, not providing any novelty or relevant contribution to the field. Indeed, model calibration is a preliminary and unavoidable necessity for any hydrological model. The two research questions presented (LL95-97) have quite obvious answers, especially the second.

I strongly suggest that the research presented be strengthened by associating this initial evaluation/calibration step with a more meaningful analysis (e.g., long-range hydrological reanalysis, sensitivity analysis, or exploring issues related to hydrological extremes). The results achieved so far seem too preliminary. Furthermore, in reviewing the study, I suggest considering the following comments.

Introduction: It is unclear why these two hydrological models were chosen and not (also) others. In addition, especially concerning WRF-Hydro, there is a vast amount of references describing significantly more advanced research, in which the calibration issue has been overcome with interesting strategies to be taken into consideration, simulating even multiple basins simultaneously for a large number of years, both in one-way and fully-coupled modes. A basic search on Scopus reveals approximately 300 documents containing “WRF-Hydro” in the Article Title, Abstract, or keywords.

One model requires daily runoff as input, while the other uses 6-hour meteorological data. One model is calibrated, the other not. Any comparison between these two models looks quite unbalanced. I would propose that the results of the analysis be presented as a sequence of steps to further improve the hydrological output (even though I don’t consider the uncalibrated analysis interesting).

The calibration and validation methodology is rather unclear. Reading LL231-235, the reader cannot understand for each basin analysed how many years were used for calibration and how many for validation. The sentence “The optimized parameters from the calibration were then used to evaluate the model over the entire 1990–2014 period” suggests that the years used for calibration were also used for validation. A simple Table would have helped. No hydrographs are shown in the entire manuscript, nor in the Supplementary Material. In addition, a detailed comparison between ENEA-REG and WRF-Hydro runoff is lacking (I can only see Fig. S4, which shows the mean seasonal cycle for only one river).

In addition, concerning the calibration strategy, the values of the calibrated parameters should be duly analysed and discussed. For example, the parameter smcmax represents the maximum soil moisture content for each soil type. Why is it higher than 1 in most basins? Please consider the issue of equifinality seriously while dealing with a multi-parameter calibration.

L146: None of the simulations considers reservoir operations and lakes. That’s probably one reason why some of the rivers are simulated very poorly. The rivers’ flow should be preliminary naturalized.

L178: Hydrological routing model components of WRF-Hydro were run on the same 6 km spatial resolution grid. Such a resolution is very low for most of the Mediterranean catchments. Additionally, this approach does not utilize one of the main features of WRF-Hydro, namely subgrid disaggregation and aggregation.

L200-202: “snowmelt parameters were left at their default settings due to the limited relevance of snowmelt in most of the study region and for potential future regionalization over other snow-free basins.” This sentence sounds very strange for basins for which the Alps or the Pyrenees are fundamental. Probably, this is another reason why some of the rivers are simulated very poorly.

Conclusions: Regarding calibration for fully coupled atmosphere-hydrological models, please note that many papers demonstrate that it is based on different concepts compared to offline calibration, and it should not be directly based on observed meteorological inputs.