Dear Editor, dear Reviewer 1,

Please find our point-by-point reply to the reviewer comments below. For easier reading, the reviewer comments are in italics and our replies are in standard font.

The manuscript by Ho and Ehret investigates the potential of small, medium and large reservoirs designed for flood control (flood protection or multipurpose reservoirs) to mitigate drought. They select 30 examples from more than 600 existing reservoirs in Baden-Württenberg (Southwest Germany) and apply a “perfect runoff prediction” into these remaining reservoirs by using the conceptual hydrological model LARSIM. These inflows to the reservoirs are labelled “semi-natural”. Based on this inflow over a 24 year daily time series, the authors tried different operating rules for flood retention and drought release. The prerequisite for this was that flood protection should not be impaired by the new operating rules. The operating rules are based on two-point hedging rules (where hedging storage begins at one point and ends at another).

The idea using flood retention basins to mitigate droughts is worth exploring. However, the manuscript has significant flaws. I would recommend rejecting the paper and resubmitting it after a complete revison.

We thank the reviewer for their time in writing these comments. However, we feel that the reviewer’s rejection of the manuscript is based in a fundamental misunderstanding of our work. It seems that the reviewer is under the impression that our work aims to provide a comprehensive plan to mitigate drought in the study area using a network of flood retention basins. This is not the case. Rather, we seek to demonstrate that pre-existing individual flood basins can be repurposed for alleviation of drought conditions downstream. We will make this point clear in a revised version of the manuscript. Please see our related detail reply to referee comment 2 further below, as well as our point-by-point responses to the reviewer’s comments.

My main concerns are the following (see more details below):

1. The connection between drought mitigation and water release is solely based on Q70. This is too simple to draw conclusions about drought defense.While we agree that drought is an inherently complex and multivariable phenomena that, in general, cannot be defined by a single observed variable, we would like to maintain that Q70 is a simple yet comprehensive method to define drought. We can use Q70 to draw conclusions about drought defense for two reasons: First, we do not claim to have a holistic drought protection strategy; rather, in the introduction, we state that we are looking specifically at streamflow drought. Second, use of streamflow thresholds such as the Q70 are a long-standing and commonly accepted way to evaluate a reservoir’s impact on drought conditions (Shih and Revelle, 1995; Wu et al., 2022; Brunner, 2021; Padiyedath Gopalan et al., 2020; Chang et al., 2019) and in general for streamflow drought, though often at higher percentiles (Hisdal et al., 2004; Knight et al., 2011; Knight et al., 2013; Vigiak et al., 2018; Yarnell et al., 2020; Lubw, 2024; Van Loon et al., 2010; Van Loon et al., 2012). Moreover, in our analysis the Q70 is a stand-in for a demand curve, where any failure to meet the demand is a shortage. Because we do not explicitly know the demands downstream of the river, we assume that if there is a general shortage of water—indicated here by the Q70—there is a user downstream who is lacking water. In a revised version of the manuscript, we will clarify these points more explicitly.
2. The spatial and hydrological context of the (arbitrary chosen) reservoirs is missing. Drought protection requires a model for river basin management.

   
   
   Based on this and the previous comments, we are under the impression that there is a misunderstanding about the purpose of these reservoirs and of our work: this study is meant to assess, from a water supply perspective, how much drought mitigation benefit each existing (flood retention) reservoir, working independently, can produce for its local surroundings. It is not to present a plan for complete water resources management in the area. We use the Q70 as a reference for drought conditions downstream and as a stand-in for water demand so that we can assess a reservoir’s ability to supplement low flows; further water resources management plans are beyond the scope of this paper. We will seek to stress this better in the abstract of a revised version of our manuscript.

   The main spatial and hydrological context of our work is the German state of Baden-Württemberg, which provides a common hydroclimatic background that is discussed in the paper. The spatial contexts of these reservoirs are, for the purposes of downstream discharge regulation, largely irrelevant because they operate independently. Therefore, their spatial context only matters insofar as their impacts (due to geography, etc.) on the hydrology, which—for the purposes of comparing some several hundred reservoirs—can be summarized by their inflows. The hydrological context is expressed via the AF (formerly called SF, which we have adapted due to the reviewer's suggestion)—because the AF is based on the mean annual flow and the average low flow (i.e. the Q70), the AF aims to summarize the local hydrological context and relate it to the reservoir’s capacity.

   Also, we would like to emphasize that the reservoirs were chosen non-arbitrarily. We aimed to choose reservoirs with a range of physical (volume) and hydrological (summarized by the AF) characteristics that would represent the existing reservoirs in Baden-Württemberg and documented this in the manuscript in section 2.2.

3. The parameter SF is not helpful (as the authors admit). Why didn’t they choose a different parameter?
   
   
   
   

   The AF is a combined indicator representing the number of times the reservoir can be filled throughout a typical year, based on the difference between the average flow volume and the average low flow volume. This is a slightly modified version of a storage ratio (volume of inflow to capacity) used in the state of Baden-Württemberg to categorize flood retention basins. For this purpose, it has proven very useful, and therefore using a modified version of it for reservoir categorization for flood/drought was a natural first choice and a reasonable hypothesis. Our modified AF estimates how much water exists above the streamflow drought threshold in a typical year. When conceptualizing a study to determine relevant characteristics that would indicate suitability for storage of flood volume for drought usage, we hypothesized that a reservoir that had more available water to store—in other words, a higher AF—would have a greater impact on the downstream drought conditions. This is based on the assumption that more water available would also mean more water to release in drought conditions.

   Testing this hypothesis also gave us another criteria to refine our reservoir selection without having to study all 600 reservoirs: by summarizing the hydrology of each reservoir by its inflow and normalizing it by the capacity, we gain an indicator that describes the hydrological conditions in a way that is comparable across different reservoir sizes. Our study showed, however, that this was not entirely the case (please see related test in the conclusions). While reservoirs with high AF were indeed able to release significant volumes of water during drought conditions, they were ultimately unable to significantly reduce total drought deficits (due to the total deficit being quite high) or drought penalty (due to shortages in summer being penalized more heavily than those in winter). Thus, while AF was ultimately not helpful as an indicator for potential drought penalty, it remains useful for characterizing water availability.
4. The conclusion “reservoirs can release up to 80 times their capacity and reduce drought penalties and water deficits by almost 95% over a 24-year simulation period” is mentioned twice (abstract and conclusions). That is the exception and not the rule and therefore misleading.
   
   
   
   The statement about releases and penalty / deficit reductions is not meant to be a general conclusion valid for each reservoir, but rather a summary of the range of results. The actual conclusions follow these statements in the abstract (“there seems to be a trade-off between relative water availability to the reservoir and ability to alleviate drought conditions”, lines 23-24) and conclusions (“… the relative water ability… did not have a strong relationship to a reservoir’s ability to curtail drought conditions”, lines 491-492). We also qualify that there are other reservoirs with different results in the abstract (“Certain reservoirs were near-optimal, others could be improved further, and still others were not very effective at reducing drought conditions”, line 24-25). However, to avoid misunderstandings, we will further stress that this is a range of results rather than a generally valid value in a revised version of the manuscript.

Further comments in detail:

1. Title: “…Estimating the maximum dual-use benefit of small flood reservoirs in Southern Germany” à That is not correct. The authors also analyzed  medium and large reservoirs.
   
   
   
   We concede that the wording in our title can create a misunderstanding, as we do study reservoirs that are technically “large”. However, even the “large” reservoirs here are, in comparison to the typical reservoir studied for this sort of usage, quite small—in the manuscript we note that the size descriptors large, medium, and small follow the German reference standard DIN19700 categories (Line 113-116). According to common definitions of small reservoirs in the literature which we introduced in the introduction ( <= 15 m dam height, capacity of 1-2 million m³; line 45), the reservoirs studied here are indeed primarily small reservoirs—the capacity and dam heights for our DIN19700 “medium” reservoirs shown in Table 1 are well within this definition. The largest reservoir in this study has a capacity of 4.3 million m³, whereas typical reservoirs for this research have billions of m³. In other words, the “typical” reservoir in this field are several orders of magnitude greater than ours. We do, however, note this size definition discrepancy in our abstract (line 15), our introduction (line 72), our reservoir selection section (line 113-115), and our conclusion (line 483). However, in the interest of clarity, we suggest removing "small” from the title and further emphasize the working definitions in the introduction.
2. Introduction: The sections on small reservoirs in Africa and worldwide do not fit the main content of the manuscript. Why do the authors write so much about agricultural (rainwater harvesting) reservoirs?
   
   
   
   The section on small reservoirs worldwide serve to illustrate that reservoirs of that capacity are used for water resources. This is to show that there is precedent for this kind of work, as well as to introduce the challenges associated with reservoirs of this size. The references to agricultural reservoirs serve to demonstrate potential uses for the water and examples of existing reservoirs. We therefore propose keeping this section in the manuscript.
3. The hypothesis “that the reservoirs providing the most benefit in drought conditions will be those that have high inflow relative to the reservoir capacity” is not justified. Why did the authors formulate this hypothesis?
   
   
   
   Please also see our related reply to reviewer main comment 3. The hypothesis is built on the assumption that high inflow relative to reservoir capacity means that there is a lot of water that can be retained and released throughout the year. This is especially important in our reservoirs, as they must be completely empty before flood events in order to guarantee flood protection and since not every flood event will completely refill the reservoir. More water available implies that more water can be stored—in other words, potentially more water can be delivered in drought conditions, thus reducing streamflow drought. We will state this more clearly in a revised version of the manuscript.
4. Line 127 until 135: The “slot” definition is unclear. The final selection of the reservoirs is not explained.
   
   
   
   The “slots” simply refer to number of allocated positions for the final selection. Selecting numbers of reservoirs based on proportion to the whole would result in either an overrepresentation of rather similar reservoirs or an unwieldy number of reservoirs; hence, we assigned each of the highly populated categories several “slots” or positions in the final selection. We further adjusted the number based on stakeholder interest (increasing the number of “large” reservoirs) and proportionality (increasing the number of MF, MM, and SF reservoirs). The final selection is, per line 146, based on the AF—we aimed to select reservoirs across the spectrum of AF. This is related to the use of AF as a hypothesizing factor: in order to test the AF equally among varying size and use categories, we selected reservoirs of varying AF within their category. If the editor agrees, we will seek to emphasize this more clearly in the next revision.
5. The storage factor SF should be renamed as “availability factor AF”. That would also avoid confusion with the variable designation SF already assigned for “small flood protection” (see Table 1). However, it is unclear why the storage factor can support drought mitigation. It is not logical that smaller C value (i.e. a higher SF value ) will “reduce drought conditions more effectively”.

   
   
   Thank you for this suggestion for a change of naming, we will adopt it in a revised version of the manuscript and have adopted it for these responses.

   The storage factor is the number of times per year that the reservoir can be filled to capacity—in other words, there are more chances to save water for drought conditions. Please also see our related reply to the previous comment. A smaller C in and of itself does not reduce conditions more effectively. Rather, that it can store and release more water relative to its inflow is its hypothesized basis for drought mitigation. A reservoir that can be filled 50 times in a year, for example, was hypothesized be able to affect more relative change on streamflow than one that is filled only 10 times in a year because it can store and release more water. This is again important due to the complete drawdown of the reservoir before a flood event without guarantee that it will be entirely refilled: by looking for reservoirs with higher water available for storage, we increase the chances of water being stored for delivery in drought. We will seek to emphasize this more clearly in the next revision.

6. Table 2: the volume of C should be added.
   
   
   
   Thank you for the suggestion. We will move the capacity from Table 3 to Table 2 and move Table 3 to the appendix, as suggested in a later comment.
7. Equation 3: Why “-5” as a penalty factor?
   
   
   
   The only constraint for the penalty factor is that it should be negative; the exact number is arbitrary for its functionality, as it is a simple linear transformation. We will add this information to the manuscript in a future revision.
8. Equation 3 and 4: Create a diagram for a better interpretation of the penalty functions.
   
   
   
   We feel that the equations already provide all necessary information, but if the editor also agrees that this would significantly improve the interpretation of the penalty functions, we will add a supporting diagram.
9. Table 3: Move it to the appendix.
   
   
   
   Will do; please also see our response to comment 6.
10. Figure 4: Why is the penalty benefit of SMALL multipurpose reservoirs so much lower than flood-only reservoirs? Please explain and discuss!
    
    
    
    Thank you for this question. This is due to the limitations of the reservoirs in this category. Nonnenbach falls into the same trap as other high AF reservoirs in the high AF, low improvement grouping (i.e. Gottswald)—namely, not having a capacity that is able to cope with the high water variability, thus limiting the range (and reducing the median) of the small multipurpose category. Similarly, Lennach and Hoelzern are limited in the same way as Wollenberg—they are unable to fill often enough to supply the water necessary for drought deficit reduction. Moreover, the capacities of these reservoirs are smaller than the typical capacity of the DIN 19700 “small” category (but remain in that category due to their dam height), resulting in reduced water stored for drought conditions. We have mentioned these points briefly in the manuscript, but not apparently not in enough detail. We will add a more detailed discussion to a revised version of the manuscript.
11. Figure 5 until 9: The flood penalty can be skipped because the precondition was that flood protection must not be impaired. Furthermore, the graphs related to outflow and drought penalties are not clearly recognizable.
    
    
    
    We would like to maintain that the flood penalty carries relevant information for the reader about the complete interaction of the system (how flooding affects volume, which affects penalty), e.g. in Figure 8. However, if the editor agrees, we can remove the flood penalty. We understand the concern with the outflow and drought penalties and will explore options to more improve readability.
12. Figure 10: Can be skipped (flood statistics are not changed due to the authors’ precondition)
    
    
    
    We would like to maintain that the flood penalty statistics provide context for high flood pre-releases and differences in performances between reservoirs. In a future revision, we will seek to make this context more explicit. However, if the editor agrees, we can move this figure to the appendix.
13. Line 440 ff: The argumentation with the IWD is very weak. Either omit or elaborate in detail (seasonal influence, PET, soil conditions etc).
    
    
    
    The discussion with the IWD is intended to contextualize how much water is made available through our strategy and what real world impacts it could have with a callback to the agricultural small reservoirs in the introduction. We think this has value for the reader and therefore prefer to keep it in the manuscript.
14. Line 473: “Changing the model so that the timing of reservoir releases such that water is given at the drought peaks could improve the penalty benefit further, though at the cost of complicating the model and the release rules.” This is recommended for the resubmission of the paper.
    
    
    
    While we agree that this would be useful and interesting, we would like to argue that this is beyond the current scope of the paper. In this paper, we aim to a) determine whether or not the idea of reusing flood basins for drought protection is viable; and to b) determine whether or not water availability (via AF) is a suitable indicator for a reservoir’s potential impact under these schemes. We do, however, plan to explore this potential change in releases—as well as more realistic demand targets tailored to local conditions—in a further paper. We will add this information to a revised version of the manuscript.

Minor comments:

1. Line 253: “is used only used to”--> “is only used to”
   
   
   
   Thank you; this was an oversight. We will correct this in the next revision.
2. Equation 11 “V = C >=0” --> “V-C >=0”
   
   
   
   Thank you for noting the error; the condition should be “V = C”. We will correct this in the next revision.
3. Line 330: “former” --> “latter”
   
   
   
   Thank you; this was an oversight. We will correct this in the next revision.
4. Line 338: “flood droughts” --> “flood”
   
   
   
   Thank you; this was an oversight. We will correct this in the next revision.
5. Figure 11: Are the x- and y-axis swapped? “combined operation model” should be on the y-axis?
   
   
   
   Thank you; this was an oversight. We will correct this in the next revision.
6. Better to use Bp (Benefit for Drought) instead of Vd,nor (y-axis, left figure)
   
   
   
   If this is in reference to Figure 14, we respectfully disagree—a plot of Bp against AF already exists in Figure 4. Moreover, this pair of figures serves to demonstrate that, while increasing water availability is strongly correlated with a higher Vd,nor, the total volume benefit remains variable among different reservoirs. We therefore will keep the figure as is.

References:

Brunner, M. I.: Reservoir regulation affects droughts and floods at local and regional scales, Environmental Research Letters, 16, 10.1088/1748-9326/ac36f6, 2021.

Chang, J., Guo, A., Wang, Y., Ha, Y., Zhang, R., Xue, L., and Tu, Z.: Reservoir Operations to Mitigate Drought Effects With a Hedging Policy Triggered by the Drought Prevention Limiting Water Level, Water Resources Research, 55, 904-922, 10.1029/2017wr022090, 2019.

Hashimoto, T., Loucks, D. P., and Stedinger, J. R.: Reliability, resiliency, robustness, and vulnerability criteria for water resource systems, Water Resources Research, 18, 1982.

Hisdal, H., Tallaksen, L. M., Gauster, T., Bloomfield, J. P., Parry, S., Prudhomme, C., and Wanders, N.: Hydrological drought characteristics, in: Hydrological Drought, Elsevier, 157-231, 2004.

Karamouz, M. and Araghinejad, S.: Drought Mitigation through Long-Term Operation of Reservoirs: Case Study, Journal of Irrigation and Drainage Engineering - ASCE, 134, 2008.

Knight, R. R., Gain, W. S., and Wolfe, W. J.: Modelling ecological flow regime: an example from the Tennessee and Cumberland River basins, Ecohydrology, 5, 613-627, 10.1002/eco.246, 2011.

Knight, R. R., Murphy, J. C., Wolfe, W. J., Saylor, C. F., and Wales, A. K.: Ecological limit functions relating fish community response to hydrologic departures of the ecological flow regime in the Tennessee River basin, United States, Ecohydrology, 7, 1262-1280, 10.1002/eco.1460, 2013.

LUBW: NIZ Pegel Klassifizierung,  2024.

Padiyedath Gopalan, S., Hanasaki, N., Champathong, A., and Tebakari, T.: Impact assessment of reservoir operation in the context of climate change adaptation in the Chao Phraya River basin, Hydrological Processes, 35, 10.1002/hyp.14005, 2020.

Shih, J.-S. and ReVelle, C.: Water supply operations during drought: A discrete hedging rule, European Journal of Operational Research, 82, 1995.

Van Loon, A. F., Van Huijgevoort, M. H. J., and Van Lanen, H. A. J.: Evaluation of drought propagation in an ensemble mean of large-scale hydrological models, Hydrology and Earth System Sciences, 16, 4057-4078, 10.5194/hess-16-4057-2012, 2012.

Van Loon, A. F., Van Lanen, H. A., Hisdal, H., Tallaksen, L. M., Fendeková, M., Oosterwijk, J., Horvát, O., and Machlica, A.: Understanding hydrological winter drought in Europe, Global Change: Facing Risks and Threats to Water Resources, IAHS Publ, 340, 189-197, 2010.

Vigiak, O., Lutz, S., Mentzafou, A., Chiogna, G., Tuo, Y., Majone, B., Beck, H., de Roo, A., Malago, A., Bouraoui, F., Kumar, R., Samaniego, L., Merz, R., Gamvroudis, C., Skoulikidis, N., Nikolaidis, N. P., Bellin, A., Acuna, V., Mori, N., Ludwig, R., and Pistocchi, A.: Uncertainty of modelled flow regime for flow-ecological assessment in Southern Europe, Sci Total Environ, 615, 1028-1047, 10.1016/j.scitotenv.2017.09.295, 2018.

Wu, Y., Sun, J., Hu, B., Xu, Y. J., Rousseau, A. N., and Zhang, G.: Can the combining of wetlands with reservoir operation largely reduce the risk of future flood and droughts?, EGUsphere [preprint], 10.5194/egusphere-2022-1103, 2022.

Yarnell, S. M., Stein, E. D., Webb, J. A., Grantham, T., Lusardi, R. A., Zimmerman, J., Peek, R. A., Lane, B. A., Howard, J., and Sandoval‐Solis, S.: A functional flows approach to selecting ecologically relevant flow metrics for environmental flow applications, River Research and Applications, 36, 318-324, 10.1002/rra.3575, 2020.

Dear editor, dear reviewer 2,

We thank you for your time and comments. Please find our point-by-point reply to the reviewer’s comments below. For easier reading, the reviewer comments are in italics and our replies are in standard font.

General comments:

1. The drought definition of Q_70 as a static concept without consideration of dynamics, such as river flashiness, leads to the selection of unrealistic discharges, as the authors state in line 333f, on which the entire model optimization is based. A reconsideration of this definition, e.g. based on catchment type or other catchment characteristics, may be necessary.

   • We acknowledge that there are limitations associated with using only Q70 for drought, and that this definition is insufficient for a genuine application of our strategy. However, for the purposes of building up and testing the potential of this revised strategy, we felt that a preliminary definition based on statistics that could be calculated for many different locations would be a reasonable place to start, as local low-water warnings are based on a similar threshold. It has also been used as a threshold for water scarcity in previous studies (Van Loon et al., 2010; Hisdal et al., 2004; Van Loon and Van Lanen, 2012; Cammalleri et al., 2016) and to characterize flow regimes, including for ecological flows (Vigiak et al., 2018; Knight et al., 2011). We use this under the assumption that water levels below this threshold means there is some user of the water—whether human or otherwise—is suffering due to a reduced water level. More refined demand curves are currently under investigation as a next step in this work. In our revision, we will further emphasize that this Q70 method is meant as a preliminary investigative step.

     We also believe that the Q70 as applied in our study does allow for consideration of streamflow variability and flashiness. The Q70 is time-dynamic and is different for every hour of the year. By considering all discharges within a 30-day window centered on each time step, the resulting exceedance probability curves contain enough data as to provide a baseline for what is “normal” for that day and time. This would include river flashiness. However, we acknowledge that if the river is extremely flashy (as in the context of line 333), the exceedance probability curve will shift and a 70^th percentile exceedance may not be strict enough. While for a redesign of such a reservoir it might be necessary to select a stricter definition (e.g. Q90 a.k.a. 90^th percentile exceedance), we continue the use of the Q70 in order to compare the different reservoirs using the same low flow definitions.

2. The definition of SF is associated with some misleading assumptions. The hypothesis that reservoirs with a high SF are less efficient does not seem surprising, as small reservoirs with a large catchment area (leading to a high SF) are unlikely to have a significant impact on water supply.

   • We agree that reservoirs that are small for their catchment area are rather unlikely to have significant impact on overall water supply, but we assert that it is worth seeing if they are able to have an impact on low flow supplementation directly downstream due also, for example, to timing. This is also the motivation behind considering penalty and volume benefit separately (for more, please see the answer to 12). The SF is modified such that it describes the typical amount of water above the low flow condition relative to capacity—in other words, it is a description of how much water we can store without causing water shortages. We feel it is reasonable to assume that reservoirs that can be refilled more often will be able to give more water in critical times.

     Additionally, testing this hypothesis also gave us a criteria we could use to refine our reservoir selection without having to study all 600 reservoirs: by summarizing the hydrology of each reservoir by its inflow and normalizing it by the capacity, we gain an indicator that describes the hydrological conditions in a way that is comparable across different reservoir sizes. Selecting reservoirs of varying SF allowed us to test the potential of this strategy using cases that span a broad range of flow conditions and reservoir sizes.
3. Since a perfect knowledge scenario for the future is not realistic, it would be interesting to define a time window of X days that the authors would need for optimal operating decisions. How tenable are the operating proposals in climate change? If there is no flooding in the simulation period, the following interpretation does not have any meaning.

   • While a perfect knowledge scenario is indeed not realistic, it is useful in that it gives us upper bounds of benefit based on past scenarios. We emphasize that this is a potential study, and having a benchmark best-case scenario (presented in this work) is useful for our next studies.

     We also agree that defining a time window for optimal decision-making would be an interesting question. We plan to address this in another phase of this work, in which we use (historical) flood forecasts at these locations to explore the effect of uncertain flood operation on the potential benefits to the reservoirs. This next phase will simulate a more realistic operation, including the effects of false alarms introducing more potential flooding events. The results of this study will be used as a benchmark for comparison, and give useful insights on the value of good flood predictions for this strategy.
4. The selection of 30 of the more than 800 retention basins seems to be biased. It would be of great benefit to define a workflow to identify all retention basins with high penalty benefit (perhaps for future work).

   • We agree with the importance of defining a workflow to determine penalty benefit—indeed, we had initially hypothesized that SF would be a way to identify high-benefit reservoirs, hence its emphasis in this paper. Unfortunately, this did not work. However, in order to develop such a workflow, we needed to test a broad range of possible conditions.

     We concede that there is an element of bias to our selection, as large reservoirs are statistically overrepresented. This is due to the assumption that larger reservoirs would be able to impact water supply more effectively, as well as due to increased interest from relevant stakeholders after introducing our ideas. However, outside of selection of categories and numbers of reservoirs, our selection was based on estimates of SF from available statistics from the local environmental agency with the aim of selecting a wide range of water regimes in comparison to reservoir sizes (approximated by SF). This allows us to test the potential of our strategy in a broad range of conditions.
5. I understand that the ecosystem functions of the river downstream of the reservoir are out of scope. However, the impact on the ecosystem in the reservoir, where e.g. the current land use is severely affected due to the prolonged inundation, should at least be mentioned.
   • We agree and will prepare a brief discussion on this point in the next revision, though we stress that this is beyond the scope of this potential study.
6. The limitation to 30 reservoirs could at least for some allow a pictorial representation with some additional features such as morphology.

   • We are uncertain how this would help, but if the editor insists, we can include this in the appendix for the next revision.

7. To avoid distortions at the beginning of the simulation period, the first year should be taken as a lead time, for example, or the starting condition of the empty basin should be redefined.
   • We agree that these possible distortions should be considered and will seek to address them in the next revision. This will most certainly result in modified figures and perhaps some modified conclusions.
8. Model uncertainties and statistical evaluations of trends are missing.

   • We are uncertain which statistical evaluations of trends are desired here, but in the case that streamflow trends are meant, we feel this is not entirely relevant, as trend analysis when calculating Q70 is not typically done. If statistics regarding the best-fit lines in Figure 14 were meant, we can add that if the editor agrees it would bring significant value to the paper.

     We are also uncertain what model certainties are desired. For LARSIM uncertainties, please refer to the answer to 10; for uncertainties related to our assumption of perfect knowledge, please refer to the answer to 3.
9. In the introduction, the context of the effects of drought in rivers is missing.
   • We agree and will prepare a brief discussion on this point in the next revision.
10. The authors state that LARSIM is typically not used for small catchment sizes. Could the authors explain the uncertainty arising from this and how they deal with it?

    • While LARSIM is not typically used for small catchment sizes, the pre-calibrated that we are using is distributed on a 1 km² grid and therefore is, in principle, capable of doing small catchments. We further clarify that it is not used for operational forecasts—its main purpose—for catchments less than ~100 km² because of high associated uncertainties from weather forecasts, local conditions, etc.

      We thank the reviewer, however, for this comment—after looking into the catchment sizes to ensure they aligned, we noticed a few errors in our model setup. For the next revision, we will map the reservoirs to the relevant LARSIM subcatchments, seeking to match the catchment areas as closely as possible, and rerun the hydrological model. However, due to different catchment delineation processes related to the 1 km² grid, a difference of a few km² can sometimes be unavoidable. We will adjust for these by scaling the model output by the area ratio of delineated catchment area to the LARSIM catchment area. This will likely affect the figures (and potentially also some conclusions) in the next revision.

Minor comments:

1. A lot of reading disruption due to hyphens.
   • We will seek to resolve this in the next revision.
2. Line 33ff: Sentence hard to understand.
   • We will seek to resolve this in the next revision.
3. Table 2: Horizontal lines between size categories would be needed.
   • We will add this in the next revision.
4. Line 141f: Q_in and Q_70 are not flow volumes but volume rates.
   • We will correct this in the next revision.
5. Line 153f.: Large Area Runoff Simulation Model.
   • We will correct this in the next revision.
6. Sentence line 192 is for discussion not methods
   • We will consider this in the next revision, but we feel it is well-placed as is.
7. Figure 2: Please update the figure without red underlining.
   • We will correct this in the next revision.
8. Chapter 2.2.1 not quite logically structured (sentence line 216 could be after the introducing sentences in line 201 etc.).
   • We will seek to resolve this in the next revision.
9. Line 153: double used
   • We were unable to find the issue and would appreciate clarification.
10. Equation 3: Where does 5 come from?
    • In this study, the exact factor in the flood penalty is arbitrary for its functionality, as it is only relevant for comparison to the flood-only operation. We will add this information to the manuscript in a future revision.
11. Line 233f: How does this fit in with integrated flood management?
    • The flood protection offered by the reservoir is the total empty volume—we cannot do any better (as far as flood protection is concerned) than that. In order for us to maintain the planned flood protection of the reservoir, we must guarantee that the reservoir is empty before the event. We will clarify this in the next revision.
12. Line 267f: How is weighting integrated in B_p?

    • Thank you for the question; we will clarify this point better in the next manuscript.

      Penalty at each time step is calculated based on the Q70 of the same time step, which contains seasonal streamflow information. In drier seasons, the Q70 will be lower; in wetter seasons, it will be higher. The penalty at each time step is a square root function that crosses the x-axis at the day’s Q70, stretching asymptotically to negative infinity. Therefore, a penalty associated with lacking 1 m³/s will be more negative if the Q70 on this day is 2 m³/s (p = -1/√1 + 1/√2 = -0.293) as opposed to 10 m^3/s (p = -1/√9 + 1/√10 = -0.0171). In this way, the penalty for not meeting the streamflow is worse in lower-flow periods and accounts for seasonal flow variations. It follows that the reduction of this penalty—the benefit—associated by providing 1 m³/s will be higher in low-flow seasons than in high-flow seasons. This is different from volume benefit in that volume benefit only cares about how much water is given in comparison to the overall deficit, not when and how impactful it may be.

      As an aside, we would like to note that the equation for Bp has a typo. It should read:

      B_p = 100 x (∑P_d,f - ∑P_d,c)/∑P_d,f

13. Table 3: Decimal places could be reduced.
    • We will correct this in the next revision.
14. Line 338: Flood droughts?
    • This is a typo and should read "floods". We will correct this in the next revision.
15. Line 432: Figure reference doubled.
    • We will correct this in the next revision.
16. Line 475: The trend that the benefit increases with increasing SF only applies to flood-only basins (Figure 14).
    • We respectfully disagree. Though not as dramatic as flood-only basins, small multipurpose reservoirs do see increasing benefit with increasing SF.
17. Line 476ff: Replicate of line 362ff
    • Thank you; we will revise accordingly.

Cammalleri, C., Vogt, J., and Salamon, P.: Development of an operational low-flow index for hydrological drought monitoring over Europe, Hydrological Sciences Journal, 1-13, 10.1080/02626667.2016.1240869, 2016.

Hisdal, H., Tallaksen, L. M., Gauster, T., Bloomfield, J. P., Parry, S., Prudhomme, C., and Wanders, N.: Hydrological drought characteristics, in: Hydrological Drought, Elsevier, 157-231, 2004.

Knight, R. R., Gain, W. S., and Wolfe, W. J.: Modelling ecological flow regime: an example from the Tennessee and Cumberland River basins, Ecohydrology, 5, 613-627, 10.1002/eco.246, 2011.

Van Loon, A. F. and Van Lanen, H. A. J.: A process-based typology of hydrological drought, Hydrology and Earth System Sciences, 16, 1915-1946, 10.5194/hess-16-1915-2012, 2012.

Van Loon, A. F., Van Lanen, H. A., Hisdal, H., Tallaksen, L. M., Fendeková, M., Oosterwijk, J., Horvát, O., and Machlica, A.: Understanding hydrological winter drought in Europe, Global Change: Facing Risks and Threats to Water Resources, IAHS Publ, 340, 189-197, 2010.

Vigiak, O., Lutz, S., Mentzafou, A., Chiogna, G., Tuo, Y., Majone, B., Beck, H., de Roo, A., Malago, A., Bouraoui, F., Kumar, R., Samaniego, L., Merz, R., Gamvroudis, C., Skoulikidis, N., Nikolaidis, N. P., Bellin, A., Acuna, V., Mori, N., Ludwig, R., and Pistocchi, A.: Uncertainty of modelled flow regime for flow-ecological assessment in Southern Europe, Sci Total Environ, 615, 1028-1047, 10.1016/j.scitotenv.2017.09.295, 2018.