Review on “Analyzing the informative value of alternative hazard indicators for monitoring drought risk for human water supply and river ecosystems at the global scale”

In this paper the authors present an approach for the selection and calculation of streamfloe drought hazard indicators for monitoring human surface water supply and for river ecosystems. In doing so, the authors discuss and propose to consider the habituation of people and ecosystems to the streamflow regime. For this purpose, eight existing drought indicatoss are compared and quantified and three new ones are proposed based on the global hydrological model WaterGAP2.2d.

This article has positive aspects, in particular the effort in modelling that the authors have made is remarkable, however it needs to be reviewed thoroughly as there are several points that need to be clarified and addressed by the authors. In short, the article has potential, but the authors need to make an effort to focus the analysis and make a readable manuscript. I hope that the comments below will help in this direction.

Structure of the article

First, the structure of the paper could be improved. It is very long, repetitive in some aspects, lacking clarity on the objectives and results, that should be better highlighted both in the description and conclusions sections.

Similarly, the order of the sections does not follow a story, some sections could be shortened, removed, or moved to increase the readability. This structure surely makes it very difficult to read. For example, the introduction should be significantly shortened to focus on the description relevant to the objectives of this work. Section 3.4 should be moved into the methodologies section. In addition, there is an annex that includes only one figure that is quite relevant. I suggest including it in the main text, if the number of figures is an editorial requirement I suggest removing some other figures (e.g. Figure 9).

The methodology can also be reduced, focusing on the description of the proposed new indicators and model validation. In addition, some aspects should be clarified, e.g., eq(1) for SPI, X is noted as a generic variable, if this variable is precipitation this representation could be misleading.

Focus of the manuscript and sectoral risk representation

The authors indicate that the focus of the article is on analysing indicators for monitoring drought risk in very specific sectors (human water supply and river ecosystems). I agree that the selection of the hazard indicator is key to determine this dimension of drought risk and the discussion on that direction is more than welcome and needed. However, apart from referring to these sectors as the focus of the article and several speculative and unsupported assertions, there is no information in this article on how drought specific indicators affect these sectors. The dynamics of how these sectors or systems are affected is surely complex, depending on various factors that determine their exposure and vulnerability beyond whether they rely on upstream reservoirs or the systems are seasonally dependent. However, this is not enough to characterize the vulnerability of these sectors. Further discussion and analysis in this regard is needed.

On the one hand, considering the way the article is structured, orienting the analysis in the description and comparison between indicators, derived metrics, etc. in a concise and targeted manner can improve the focus and structure of the article.  On the other hand, to strictly evaluate whether the proposed indicators are valid to represent risk, a more detailed analysis of the proposed sectors is needed, with a description of their vulnerabilities and how their impacts are produced as a consequence of the combination of the different dimensions of risk.

Grid cells – case study selection and description

The selection of the two grid cells seems to be motivated by characteristics derived from some of the modelled variables. However, no description of these sites exists in the manuscript. A quick search turns up that one gridcell is in central eastern Paraguay (perhaps including a portion of the Paraguay river) and the second near Firenze (Arno). Both points with very different realities regarding how they might be exposed and how they are vulnerable to droughts and surely each will have a very specific risk profile. Here I see a missed opportunity, as one of the objectives of this paper is to find out how the exposed systems can be used. Surely, a discussion along these lines would greatly improve the discussion on the usefulness of the various indicators. Indeed, the comparison between indicators is merely informative, which does not enable identification or validation between them (Please refer to my final comment)

Similarly, March 2002 has been used to describe Figures 3 and 4. But it is not clear why this period was chosen or how this comparison can be extrapolated to the whole period.

Comparison not validation

As proposed in this article, the fundamental purpose of any drought indicator is to represent the sectoral impacts in the best possible way. Indeed, the validation of the best indicator should be consistent in how it represent sectoral impacts. In this sense, it is perfectly legitimate to compare indicators, but it is not possible to validate or rank one over another without looking at independent variables that represent potential impacts. Please elaborate further on this issue

Review “Analyzing the informative value of alternative hazards indicators for monitoring drought risks for human water supply and river ecosystems at the global scale”

This paper analyses which drought hazard indicators are suitable for assessing and monitoring streamflow drought risks for human surface water supply and for river ecosystems. The authors recommend considering the habituation of people and ecosystems to the streamflow regime when selecting indicators. Eight existing indicators and three new indicators are proposed and evaluated using the model results of WaterGap.

The paper has potential but needs a proper revision before publication. The paper is very long and thereby is not always that clear/focussed and repetitive in some aspects. Overall sections can be shortened and may be ordered in a more logical way. For example, the introduction is very lengthy and lacks focus relevant to the objective of the work. Also, the methodology is at some aspects lengthy (too lengthy) whereas other aspects are not discussed at all. Altogether, from the current version of the manuscript, it is hard to judge the full potential of the research and research paper.   

Methods and data:

Section 2.1: First, a minor comment, the model names WaterGAP2.2d, WGHM, WaterGAP are used for the same model I assume. However, this is a bit confusing. Furthermore, previous model results are used (namely from Müller Schmied et al 2021) and all details on model description and evaluation are not discussed in this paper. However, to better understand the results it would be useful to at least read a summarized description of the, for this study, most important parameters, assumptions made, uncertainties, and sensitivities and how this has, or has not, an impact on the evaluation of your drought indicators.

Section 2.2: this section is very long and could be focused more on the new indicators and evaluation of the results.

In the method section, there is no description of how the results will be presented or evaluated/compared.  

Figures:

It is not clear why the two grid cells were chosen. Also, it is not clear why March 2002 was chosen.

In the current manuscript legends and labels are off for e.g. Fig 3 (what are the lower and upper labels in the left legend, and the upper label in the right); Fig 4, placement of ‘0’ and what is meant with ‘0 >0’; what comes after ‘1’; Fig 5, placement of ‘0’; etc.

This study proposes a new classification of streamflow drought hazard indicators (SDHI) to guide researchers or practitioners in selecting an appropriate indicator(s) for their objective(s). In order to demonstrate the importance of the proper selection of SDHIs in a drought assessment, the authors compared eleven SDHIs, including three new indicators, and quantified their similarities and discrepancies. Because human society and ecosystems adapt to a local streamflow regime, this study underscores their habituation implicitly assumed in each SDHI. This includes how one can define a normal condition and a condition under a lack of water. In this context, the results show that drought severity, i.e., accumulated drought magnitude, rather than mere magnitude at a time step, may need to be evaluated to investigate prolonged drought events. The dry season in the arid region, which is not suitable for a conventional threshold method, and the water deficit in comparison to water demand are also paid particular attention in this study. Classifying SDHIs and organizing their features, as a result, the authors stress that one needs to select an appropriate SDHI(s) in view of their assumption(s) and operational drought early warning systems need to cover a wide range of DHIs to support a widespread users’ demand.

The paper provides several important insights regarding SDHIs. Because a better understanding of the application and interpretation of DHIs is a crucial challenge, the results are valuable for the drought research community and practitioners. However, I have some major concerns that the authors need to address during this discussion phase.

1. Structure of this manuscript

I assume the authors have a certain intention, but I would say that the current structure of the manuscript is not so reasonable to me. Due to the structure, it was difficult for me to read the manuscript and grasp this study's point and novelty. Therefore, the author needs to reconsider the design of the storyline. The followings are my suggestions:

• I recommend the authors move Section 3.1 to Method (or partly, Introduction) because this section describes (i) debatable points, difficulties, or caveats in selecting/applying SDHIs, (ii) characteristics of each SDHI that serve as premises for subsequent analysis/discussion, and (iii) the definitions of key terms (e.g., magnitude/severity, conceptual/operational, etc.). Moreover, this section includes relatively general contents, some of which were mentioned in the Introduction, and refers to many preceding studies. Probably, the authors aim to present the classification as a new result, but I would say that it should be reasonable to present the classification as a new classification “method” in Chapter2 and demonstrate the needs and validity of the classification in the Result Chapter.
• In conjunction with the previous comment, I would recommend the authors modify the structure of Chapter4 and rename Chapters 3 and 4 as Results and Discussion. This is because the paragraphs in P. 37 and 38 in Chapter4 give a good summary of the results. Although the authors describe the essence of a comparison at the end of each comparison in Chapter 3, these key lines are fragmented in the Chapter. Thus, the summary and interpretation of the results fit the best right after Chapter3 (= at the beginning of Chapter4) with Table 3. I would title the summary section “systematic approach for selecting streamflow drought hazard indicators”, and the Recommendation could be section 4.2.
• The position of the description in Model validation is not reasonable (section 3.4) #1; the 1^st paragraph of the section has to be in section 2.1 because these are not results of this study and are justifications for the use of WaterGAP.
• The position of the description in Model validation is not reasonable (section 3.4) #2; the validation results (the 2^nd paragraph of section 3.4) have to be presented before the model-based analyses (i.e., section 3.3).

2. Model validation

I require the authors to present a more description on the model validation concerning stream drought reproducibility. Currently, no figure has been presented for section 3.4. Also, I expect an additional analysis to evaluate how well the simulated stream flow data reproduce hydrological drought events detected by the observation data. I would say that, at least, the comparison of observation- and simulation-based drought detections at the selected two gauge stations should be presented similarly to Figure 2.

3. The selected two gauge stations

The basin names and their characteristics (lines 507-510, lines 516-517) should be described right after the first sentence of section 3.2 in order to explain the phrase “at two GRDC gauge stations with different streamflow regimes”.

4. March 2002
   • I am curious about why March 2002 was selected as an example in this manuscript. An additional short line is expected in this regard.
   • The authors need to discuss the generality of the results in Section3.3 for March 2002, although I assume that similarities and discrepancies among SDHIs are similar for other months and years too.

5. Other

The writing often lacks sharpness. It includes many restated phrases (for example, 21 “i.e.” in total) and repetition (e.g., the authors repeatedly stress that severity is a cumulated value.). This manuscript can be more concise to deliver the authors’ messages to readers.