Assessment of the Impact of Climate Change on Dam Hydrological Safety by Using a Stochastic Rainfall Generator

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This manuscript presents a comprehensive analysis of climate change impact on inflows to the Eugui Reservoir in Spain. The study builds on the modelling chain that includes a stochastic rainfall generator STORAGE, a continuous hydrological model of the reservoir’s catchment COSMO4SUB, and the VEM to translate the change in the inflows to the maximum water levels in the reservoir. The methodology is sound, and the topic of the study is highly relevant for the readers of the journal.

However, there are three major concerns about this manuscript: namely the explanation of the adopted methodology, presentation of the results, and the discussion.

1. The calibration of the hydrological model is not fully described (see specific comments below), whereas the model evaluation is not explained at all. I suggest the authors to evaluate the model, especially in terms of reproducing design floods. Additionally, I suggest the authors to concentrate the description of the models and their application in a single catchment, rather than in two separate sections (see e.g., comments related to the calibration of COSMO4SUB). Lastly, the authors infer the impacts of climate change solely from visual inspection of the probability plots. I suggest the authors to provide quantitative comparison on the distributions by applying an appropriate statistical test.

1. RESULTS. The manuscript contains too many figures. Some of the figures can be merged together (see specific comments below), but many of them (showing the probability plots) can be replaced by the results of formal statical tests that compare future probability plots to the corresponding probability plots obtained in the baseline period. Such way of presenting the results may be clearer and thus more easily followed by the readers, so I strongly suggest the authors to consider a different presentation of the results.
2. DISCUSSION. The paper simply lacks a discussion of the results. I strongly recommend the authors to add a Discussion section to the manuscript, and to discuss uncertainties stemming from the methodology (e.g., application of STORAGE to generate series with 15-minute resolution, although the sub-daily data were not included in the estimation of model parameters, extension of series from 30 to 500 years, sensitivity of the results to the temporal resolution and the length of the synthetic series, etc.), and in the results (e.g., some climate modelling chains result in an increase of maximum water levels, and how these discrepancies in the results can be taken into account when deciding on dam adaptation). Furthermore, this section may also present the outlook, including the progress mentioned in lines 507-509. The manuscript would greatly benefit from a designated section aimed at properly discussing the results.

 

In addition to these major comments, specific comments are outlined below. Provided that the comments are properly addressed, I shall recommend the editor to accept the revised manuscript for publication in Hydrology.

 

SPECIFIC COMMENTS

1. INTRODUCTION

• Sentence in lines 65-68 is a bit unclear. Please explain “in perturbed conditions”.
• Lines 85-87. Please explain flood analysis at monthly or annual scale, or consider rephrasing the sentence.
• Line 95. I suggest outlining the conclusions of Lehner et al (2006), or omitting this reference.
• Line 96. Perhaps it would be convenient to explain that there are many reservoirs in Spain (it is a country with most reservoirs in Europe), and justify selection of a reservoir in this country as a case in this study before diving into the available climate change projections.
• Lines 108-111. This sentence implicates that SRG do simulate flood events, which is misleading. Please consider rephrasing this sentence.

2. MODELS

• VEM is not used for inferring optimal operation of gates (i.e., no gate operation is included in section 2.3), so I strongly suggest the authors to modify the manuscript accordingly (e.g., sentence in lines 127-129 or 181-182). The correct explanation is provided in lines 250-252.
• Line 134. Please explain application of STORAGE under changing climate. Does it imply non-stationarity? Here you can refer to pseudo-non-stationary application in this study.
• Explain the COSMO4SUB model in more details. How does it simulate rainfall (as stated in line 148), and how it treats baseflow, evapotranspiration and snow, considering that it is a continuous hydrological model. Please, explain how it treats spatial heterogeneity. I suggest the authors to explain model calibration in this section (rather than in lines 317-322). It is not explained how the hydrological groups for the initial CN assessment were inferred based on CORINE maps (lines 234-236).
• Line 147. Perhaps is better to explain that the model simulates three processes or that it comprises three modules, rather than that it is characterised by them.
• Line 175. Please explain “tc” (here, not just in line 319).

3. CASE STUDY AND DATA

• Line 235. Please add a reference for CORINE 2018.
• Line 239. Terms like “reference” or “baseline” period are more common than the “control period”.
• Line 237. According to Table 1, 12 GCM-RCM modelling chains are used (not just regional climate models).
• Line 234, Table1. Please correct type (GCM instead of CGM), and explain the abbreviations in the table caption.
• It would be beneficial if the authors would provide the time of concentration of the catchment.

4. METHODOLOGY

• Order of subsection of section 4 does not follow the workflow outlined in the first paragraph of the section. For the sake of consistency, I suggest the authors to consider reshuffling the order of subsections, or presenting the workflow differently.
• Line 256. Please explain 500-year period and the three future periods mentioned in e.g., line 241.
• Line 268. I suggest referring to Table 1 when mentioning the 12 climate modelling chains. It also appears that some words are missing in this sentence, as only the reference period is mentioned.
• Figure 3. I suggest the authors to use the same scales of Y-axes when showing the same variable. The authors should also consider presenting the future projections in the Results section.
• Figure 4 and Figure 5 can be combined in a single figure with four panels. It would facilitate presentation of seasonal projections, and decrease the number of figures in the manuscript.
• Line 299. The number of generated series in total is a bit unclear. It seems like the authors used each future subperiod to create extended series of finer temporal resolution, but this has to be explicitly stated in the manuscript.
• Lines 317-322. The model calibration can be explained in a section that describes the hydrological model. Additionally, the calibration has to be explained in more details: for example, what was the calibration period, which optimisation algorithm was used, was a multi-objective calibration applied, since three objectives are mentioned in line 323, was any regularisation method applied if the model was gridded (?), how were initial estimates of the CN used in the calibration? The explanation given in lines 360-362 (The results section) do not clearly explain the calibration protocol. Furthermore, the model evaluation has to be elaborated, in particular its ability to simulate design floods.
• Lines 341-342. The statement that “The model allows to calibrate the sub-daily parameters…”, differs from the explanation in section 4.1 and in the next three lines, which clearly state that no sub-daily parameters could be optimised. Please consider rephrasing for the sake of clarity.
• Lines 345-350 are basically repetition of lines 261-267. Consider removing these lines to reduce redundancy.

5. RESULTS

• Figure 6. Please provide unit to tc in the panels, and MAE and RMSE. Please state the calibration period in the figure caption.
• The results presented in Figure 7 and Figure 8 are a bit difficult to follow, particularly because the flow ranges differ from one panel to another. I recommend the authors to consider providing a condensed presentation of these results in the main text (e.g., by providing the results of formal statistical tests to compare the distributions in a future period to the corresponding distribution in the baseline period and indicate an increase/decrease in flows), and moving the probability plots to the Supplementary material. Additionally, the way in which the points in the panels were obtained has to be explain (annual maxima of inflows at Eugui Reservoir of the 500-year synthetic flow series?). Should the authors opt for presenting the results of statistical tests, the testing procedure has to be explained in the Methodology section.
• Lines 382-385 and Figure 9. Please explain the analysis of the median/average flows, considering that the study deals with flood flows.
• Presentation of the probability plots of outflows in Figures 10, 12 and 14 is not informative and a bit different to follow (similarly as in Figures 7 and 8). For the sake of clarity, the authors can present results of the formal statistical tests to compare the outflow distributions in a future to the corresponding in the baseline period and to indicate the sign of change (an increase or a decrease), and to show the probability plots in the Supplementary material. For the sake of consistency, the same diagram type should be used to present all probability plots (Figure 7, 8m 10, 12 and 14). Similar comment can be made about presentation of the maximum water levels shown in Figures 11, 13 and 15.
• Line 414. There is a typo: “...as much time AS possible” (or, alternatively, “as long as possible”).

6. Conclusions

• Lines 481-494 read more like Results than Conclusions.
• Lines 507-509. These sentences do not present conclusions of the present study, so they may be removed, or replaced in a designated part of the Discussion section.

OTHER COMMENTS

• The references in the manuscript are doubled (even with two different styles of referencing), probably due some confusion with the reference manager. Please correct.
• Many figures are misaligned, thus not entirely visible (Figure 1, Figure 2, Figures 4 and 5 (SON, MAM), Figure 9).
• I suggest the authors to consider using “climate modelling chains” instead of “climate models”, and “current conditions” instead of “current situation”.

 

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Major Recommendations for Improvement:

1. The manuscript would benefit from a dedicated pre-analysis section evaluating the sensitivity of key model parameters (e.g., Curve Number [CN], time of concentration [tₐ] in COSMO4SUB, and STORAGE rainfall parameters). This would enhance understanding of their influence on hydrological outcomes.
2. The use of established sensitivity analysis techniques (e.g., Monte Carlo simulation, perturbation methods) is recommended to rigorously assess output uncertainty arising from input variability.
3. Strengthen the validation of STORAGE and COSMO4SUB by employing independent datasets, such as split-sample validation or k-fold cross-validation, to ensure robustness.
4. A systematic comparison between synthetic rainfall/runoff series and observed extreme events should be included to evaluate model reliability in simulating high-return-period conditions.
5. The ensemble spread of the 12 climate models should be analysed more explicitly—visual aids such as box plots or ensemble dispersion diagrams for key performance metrics would improve interpretability.
6. The assumption of stationarity in the 500-year synthetic series (Section 4.1) warrants further scrutiny, including a discussion of its implications and limitations in non-stationary climate regimes.
7. Several figures (e.g., Figures 3–5, 10–15) exceed page margins, necessitating resizing. Ensuring proper scaling would enhance readability.
8. Axes labels, legends, and font sizes should be standardised across all figures to improve visual consistency and clarity.
9. A clear rationale for the 500-year simulation period should be provided, including an assessment of its adequacy for extreme event analysis.

Minor Recommendations for Improvement:

1. The assumption of uniform rainfall distribution (Section 4.1) should be critically discussed, particularly regarding its potential impact on dam safety assessments.
2. The spillway capacity (270 m³/s) should be introduced earlier in the abstract to better frame safety thresholds.
3. The rationale for selecting the Eugui Dam (e.g., data availability, regional hydrological significance) should be explicitly stated.
4. The "wet day" threshold (0.1 mm) should be defined upon its first mention to avoid ambiguity.
5. A brief justification for omitting sub-daily parameters, given the daily resolution of climate projections, would strengthen methodological transparency.
6. The final calibrated values of CN and tₐ should be reported in the main text alongside Figure 6 for immediate reference.
7. Model calibration accuracy should be supported with standard metrics (e.g., Nash-Sutcliffe Efficiency [NSE], Root Mean Square Error [RMSE]).
8. The "optimal" gate operation strategy should clarify how competing objectives (e.g., dam safety vs. water conservation) are balanced.
9. The sensitivity of results to the fixed initial reservoir level (conservation pool) should be examined.
10. Figure 9 should include explicit legends or labels to distinguish RCP scenarios, as color-coding alone (e.g., blue/red) may be insufficient.
11. An explanation for the apparent reduction in peak outflows under RCP 8.5 (2041–2070, Figure 12) compared to earlier projections (e.g., Figure 10) is warranted.
12. Table 1 (climate model inventory) and Figure 1 (location map) should include more informative captions to aid reader comprehension.
13. A uniform colour scheme for RCP scenarios across all figures (e.g., RCP 4.5 in blue, RCP 8.5 in red) would improve cross-referencing.
14. Findings should be contextualized within existing flood risk literature (e.g., Alfieri et al., 2015), with discrepancies explicitly addressed.
15. The apparent stability of median reservoir levels despite climate variability merits discussion.
16. The exclusion of land-use change impacts (e.g., urbanization) should be acknowledged, with implications for long-term projections discussed.
17. Potential biases from using CORDEX data instead of national projections (e.g., AEMET) should be evaluated.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

I appreciate your careful and comprehensive edits in response to my feedback.  The manuscript has been greatly strengthened by the addition of box plots to visualize ensemble spread, the expanded discussion of stationarity assumptions, the additional paragraphs on parameter sensitivity (Sections 4.1 and 4.2), and the clarifications regarding model validation and methodological decisions (e.g., uniform rainfall distribution, 500-year simulation rationale).  Additionally, I value how the figures are standardized, how they are better contextualized with flood risk literature, and how restrictions (such as land-use change exclusion and CORDEX data selection) are handled transparently.

The revisions address all my major and minor concerns, and I now support the publication of the manuscript in its current form. This work provides a valuable framework for assessing climate change impacts on dam safety, and the rigorous responses to reviewer feedback further enhance its contribution to the field.

Author Response

The Authors would like to thank the Reviewer for the positive comments about the revised manuscript. The Authors would also like to thank the Reviewer for the comments and suggestions of the first-round review, as they have been very useful to improve the paper quality.