the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 09 Jan 2025
The impact of climate change on dam overtopping flood risk
Abstract. There is unequivocal evidence that climate change will change the risk profile of dams, which are critical pieces of infrastructure that safeguard water supply and provide flood mitigation for populated areas. However, the challenges involved in estimating the probability of extreme floods under climate change have meant that few studies have estimated the plausible changes in the risk of extreme floods that have the potential to overtop dams. A recent examination of contemporary scientific findings pertinent to climate change impacts on flood risk has informed the projection of extreme flood risk and dam overtopping risk estimates made here. We project changes in the exceedance probabilities of overtopping risk for 18 large dams in Australia under a range of global warming assumptions, where consideration is given to the impacts of climate change on rainfall depth, rainfall temporal pattern, and rainfall losses resulting from changes in antecedent catchment wetness. We used event-based flood modelling and Monte Carlo sampling to appropriately represent the range of uncertainties associated with projecting estimates of extreme flood risk. Our results are presented in terms of changes per degree of global warming, which facilitates their interpretation in terms of different greenhouse gas emission scenarios and future time horizons. We found that increases in rainfall depth had the largest impact on increasing dam overtopping flood risk for all 18 dams under climate change. Under 4 °C of global warming, which approximates conditions towards the end of this century under a high emissions scenario, the risk of overtopping floods was between 2.4–17 times that of historical conditions for the 18 dams investigated. We also found that the risk of overtopping has more than doubled compared to the historical baseline for four of the dams investigated here as a result of global warming that has already occurred.
- Preprint
(2956 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on hess-2024-403', Anonymous Referee #1, 27 Jan 2025
General comments
This paper addresses a critical and underexplored issue—the implications of climate change on dam hydrological safety. The topic is highly relevant given the increasing frequency and severity of extreme weather events, making the findings applicable to dam owners, policymakers, and climate adaptation professionals globally. The study offers a robust method for addressing uncertainties, and its analysis of 18 large dams across varied Australian climate regions is particularly noteworthy.
However, while the study’s objectives are clear and the methodology is sound, there are areas for improvement:
- First of all, the title of the article clearly refers to the impact of climate change on dam overtopping. However, most of the paper focuses on the impact on the hydrological loads to dams, neglecting the specific aspects related to dam safety. Furthermore, the only considerations regarding dams are taken into account in a very simplified manner (e.g., the assumption of the reservoirs being at a full supply level prior to the storms). The article title does not embody the methodology and results presented, and cannot be accepted as a valid title.
- Throughout the entire manuscript, the term “risk” is misused (even in the title). In the dam safety context (and in any context related to natural hazards), the risk is defined as the combination of a potential hazard and its consequences. However, in the manuscript only the occurrence of the hazard and its probability are studied and quantified. Therefore, the manuscript should be reviewed and the term “risk” should be adequately replaced by “probability” or “hazard”, depending on the case.
- The methodology is quite dependent on the Australian context. The introduction and discussion sections could better frame the global implications of the findings, as they currently focus primarily on Australia. Specify what assumptions and methods can be replicated to other regions or contexts. In line 342, the authors say that the approach “could be translated globally for estimating flood risk under climate change elsewhere in the world”, which is not clear.
- The graphical abstract includes a graph representing the dam crest flood level vs the exceedance probability:
- This graph is not represented in the article.
- The y-axis does not correspond to the exceedance probability.
Specific comments
Introduction
- The introduction provides a strong rationale for the study. However, it lacks a succinct statement of the research gap. Explicitly contrast the current study with previous works on flood risk and dam safety.
- The global context is underdeveloped. Adding examples from other regions (e.g., Europe or Asia) could broaden the impact.
Materials and methods
- Line 121: the description of the R2ORB emulator needs more detail. A brief explanation of how it works and its advantages compared to other tools would be helpful.
- Line 129: “we assumed that the reservoir was at a full supply level prior to the storm”: this is a strong simplification of the methodology. Please justify:
- The reasons (lack of data, lack of time…)
- The potential impact (have you done an example calculation?)
In terms of dam safety, this is equivalent to not considering antecedent catchment wetness when calculating floods. I strongly recommend the authors to at least perform one example analysis of this effect. This is important in an article that focuses on dam safety (it’s in the title).
- Line 134: “outflow hydrograph”: do you mean the catchment’s outflow hydrograph, or the reservoir’s outflow hydrograph? Please harmonize the vocabulary throughout the paper.
- More information on the Monte Carlo simulation framework is needed in Appendix A and in the description of the methodology.
- Line 162: impacts of climate change on what?
- No mention to dam operations is presented here. What are the assumptions? How dams are considered operated in the baseline period and in the future periods? Just a short indication is given in line 138.
Results
- The results are presented effectively, but their practical implications could be expanded to aid decision-makers.
- Line 263: specify the three flood drivers.
- Line 272: “the reservoir outflow flood frequency curve” should be “the reservoir inflow flood frequency curve” because it refers to the DCF, which is the flood entering the dam’s reservoir.
- The RS factor does not illustrate the importance of the underlying AEP (i.e., how unsafe is a dam under historical conditions). When presented, results should simultaneously show the base AEP and the RS factor. Moreover, given that the results are anonymized and Figure 5 is just a summary of the results for the 18 dams analyzed, maybe an anonymized figure (similar to Figure 4, but only for the combined effect) in an Appendix could help shedding light on this issue.
- Synthesize somewhere the simplifications assumed in the methodology and the potential improvements.
- The Results section is somehow succinct and could benefit from more detailed analysis. For instance, the authors haven’t studied in more detail the hydrological response of each catchment to the changes in the rainfall parameters (rainfall losses, storm temporal patterns…). Figure 5 is presented as a black box without details and hardly exploitable.
Discussion
- Line 367: the assumption made by the authors that climate change will lead to increasing the potential for dams to attenuate floods is not backed by the findings presented in this paper. I recommend to replace “will result” by “could result”.
- Paragraph lines 384 to 399: authors justify the use of outputs from global climate models instead of regional ones, while in reality none of these outputs have been used here. The authors have simply applied a series of temperature increases to the hydrological drivers, without relying on any climate model. This could lead to misunderstanding the process followed. I recommend to replace this paragraph by a justification of the simplified methodology proposed.
Conclusion
- Line 429: it is the first time that this result appears in the text. It should be mentioned before.
- Line 440: replace “practical approach” by “simplified approach”.
Technical corrections
- Line 42: rephrase “moisture delivered”.
- Lines 56-57: consider revising the citation format.
- Line 59: repetition “that that”.
- Line 89: reference for the Australian Rainfall and Runoff.
- Table 1: instead of indicating dam owners, indicate dam type.
- Line 169: replace “The rates of changed” by “The rates of change”.
- Figure 4:
- replace “notional overtopping” in legend by “DCF”
- replace “Shift in overtopping risk” by “Shift in overtopping AEP”
- y-axis does not represent the AEP, but the return period.
- Figure 5: legend indicates that outliers are not shown, but it is the case in plots (b), (c) and (d)
- Line 325: change “appear” to “appears”
Citation: https://doi.org/10.5194/hess-2024-403-RC1 -
AC1: 'Reply on RC1', Michelle Ho, 08 May 2025
Referee 1
General comments
This paper addresses a critical and underexplored issue—the implications of climate change on dam hydrological safety. The topic is highly relevant given the increasing frequency and severity of extreme weather events, making the findings applicable to dam owners, policymakers, and climate adaptation professionals globally. The study offers a robust method for addressing uncertainties, and its analysis of 18 large dams across varied Australian climate regions is particularly noteworthy.
However, while the study’s objectives are clear and the methodology is sound, there are areas for improvement:
- First of all, the title of the article clearly refers to the impact of climate change on dam overtopping. However, most of the paper focuses on the impact on the hydrological loads to dams, neglecting the specific aspects related to dam safety. Furthermore, the only considerations regarding dams are taken into account in a very simplified manner (e.g., the assumption of the reservoirs being at a full supply level prior to the storms). The article title does not embody the methodology and results presented, and cannot be accepted as a valid title.
To provide a more precise description of the study, we intend to replace the title of “The impact of climate change on dam overtopping flood risk” with “The impact of climate change on dam overtopping floods”. Please refer to our response in point 2 where we provide further explanation.
- Throughout the entire manuscript, the term “risk” is misused (even in the title). In the dam safety context (and in any context related to natural hazards), the risk is defined as the combination of a potential hazard and its consequences. However, in the manuscript only the occurrence of the hazard and its probability are studied and quantified. Therefore, the manuscript should be reviewed and the term “risk” should be adequately replaced by “probability” or “hazard”, depending on the case.
The estimation of flood exceedance probabilities is the basis of assessing hydrological risk-based designs in engineering and we had been using the term “risk” in this context. However, we did not clarify this context and we recognise that this application of the term is used variously in the literature. We therefore intend to replace this terminology throughout the manuscript, as well as the title, as suggested to provide a more precise description of our analysis.
- The methodology is quite dependent on the Australian context. The introduction and discussion sections could better frame the global implications of the findings, as they currently focus primarily on Australia. Specify what assumptions and methods can be replicated to other regions or contexts. In line 342, the authors say that the approach “could be translated globally for estimating flood risk under climate change elsewhere in the world”, which is not clear.
We thank the reviewer for pointing this out. The manuscript would be improved with providing specifics with respect to model and data requirements to make these methods globally applicable and we will do so.
- The graphical abstract includes a graph representing the dam crest flood level vs the exceedance probability:
- This graph is not represented in the article.
- The y-axis does not correspond to the exceedance probability.
The reviewer is correct – the graphical abstract is not from the article. It was specially designed for the purpose of providing an overview of the research and general results rather than to convey specific results and we believe that it serves this purpose.
The reviewer is also correct that the y-axis does to correspond to the exceedance probability – the y-axis represents the reservoir level and this was communicated using an icon of a dam with water levels along the y-axis. To clarify this, we will add a text label to the y-axis in addition to the icon and relocate the x-axis label to improve clarity.
Specific comments
Introduction
- The introduction provides a strong rationale for the study. However, it lacks a succinct statement of the research gap. Explicitly contrast the current study with previous works on flood risk and dam safety.
We believe that we have provided an explicit contrast between our study and previous work in the last two paragraphs of the introduction. However, we believe that addressing the reviewer’s recommendation of using precise terminology with respect to the word “risk” will improve the clarity of our statement explaining the research gap.
- The global context is underdeveloped. Adding examples from other regions (e.g., Europe or Asia) could broaden the impact.
The two studies we reference that have quantified changes in the probability of dam overtopping floods were located in Taiwan and Spain and additional reference to these locations will be added. We will clarify that there are few other studies that specifically investigate the changes in dam outflow flood frequencies under climate change.
Materials and methods
- Line 121: the description of the R2ORB emulator needs more detail. A brief explanation of how it works and its advantages compared to other tools would be helpful.
We will include more detail on the R2ORB emulator in Section 2.2. when R2ORB is first mentioned. We will revise the text to explain that R2ORB uses data inputs and performs the calculations outlined in Fig 2, which shows a schematic of the event-based modelling process. We will add that by using an emulator of RORB, we were able to focus our calculations on the aspects of flood hydrology modelling that are most relevant in the exploration of climate change impacts on dam hydrology, namely the catchment runoff-routing and reservoir routing to estimate peak reservoir outflows. Our approach to calibrating and validating the R2ORB models will also be included. We will also add that R2ORB follows the basic generic modelling structure of event-based conceptual rainfall runoff models, namely, partitioning of rainfall into losses and excess and attenuation of the flood through channels and reservoirs. There are many hydrological tools that can model runoff responses to rainfall at an event timescale and we do not think that a listing and comparison to these tools would be of material value to this paper.
- Line 129: “we assumed that the reservoir was at a full supply level prior to the storm”: this is a strong simplification of the methodology. Please justify:
- The reasons (lack of data, lack of time…)
- The potential impact (have you done an example calculation?)
In terms of dam safety, this is equivalent to not considering antecedent catchment wetness when calculating floods. I strongly recommend the authors to at least perform one example analysis of this effect. This is important in an article that focuses on dam safety (it’s in the title).
We will include in our justification for assuming a full supply level in all the dams that it provides a worst-case scenario for estimating the probability of a dam crest flood. We will also add that the modelling of reservoir levels considering climate change impacts had only been conducted for two of the dams and these were based on specific future scenarios making them incompatible with our analysis approach, which is based on changes in global temperature. While it is very easy to include initial reservoir level in a Monte Carlo analysis, the assumptions underlying the shifts in the marginal distribution are subject to deep uncertainty due to the future changes in operating conditions that are required, which in our experience have a much greater influence on reservoir levels than the change in antecedent conditions. We will provide some additional discussion on this point in the revised version of the manuscript.
- Line 134: “outflow hydrograph”: do you mean the catchment’s outflow hydrograph, or the reservoir’s outflow hydrograph? Please harmonize the vocabulary throughout the paper.
We will clarify early on the in the manuscript that the “inflow hydrograph” refers to the reservoir inflow hydrograph while the “outflow hydrograph” refers to the reservoir outflow hydrograph.
- More information on the Monte Carlo simulation framework is needed in Appendix A and in the description of the methodology.
Agreed. This will be addressed.
- Line 162: impacts of climate change on what?
This will be revised to read: “The impacts of climate change on the AEP of the DCF…”
- No mention to dam operations is presented here. What are the assumptions? How dams are considered operated in the baseline period and in the future periods? Just a short indication is given in line 138.
We will add that dam operations in response to a dam crest flood under climate change are assumed to remain the same as historical operations and that the operational impacts are encompassed in the relationships between reservoir height and outflow provided by the dam owners. This point is also relevant to our response to item 8) above.
Results
- The results are presented effectively, but their practical implications could be expanded to aid decision-makers.
We will include in the discussion some of the practical implications such as potentially re-evaluating spillway capacities and updating floodplain inundation maps.
- Line 263: specify the three flood drivers.
Thank you. This will be addressed.
- Line 272: “the reservoir outflow flood frequency curve” should be “the reservoir inflow flood frequency curve” because it refers to the DCF, which is the flood entering the dam’s reservoir.
The dam crest flood is dependent on the outflow rate from the dam and we are in fact showing the reservoir outflow flood frequency curves in Fig 4, referenced in L272. We will include that the outflow rate corresponding to dam crest reservoir levels is indicated in these figures.
- The RS factor does not illustrate the importance of the underlying AEP (i.e., how unsafe is a dam under historical conditions). When presented, results should simultaneously show the base AEP and the RS factor. Moreover, given that the results are anonymized and Figure 5 is just a summary of the results for the 18 dams analyzed, maybe an anonymized figure (similar to Figure 4, but only for the combined effect) in an Appendix could help shedding light on this issue.
We will show a summary of the outflow frequency curves across all dams in response to all three flood factors combined with the y-axis standardised by the dam crest flood level. We will also include in the text the range of historical dam crest flood AEPs across the 18 dams. We do not think it would be helpful or relevant to provide comment on how unsafe the dams are under historical conditions: apart from the fact that the owners would not want to see this point highlighted, the assessment of “dam safety” is a complex task that needs to take into account dam-specific engineering factors that lie outside the scope of this paper.
- Synthesize somewhere the simplifications assumed in the methodology and the potential improvements.
We will clarify that we have included the key aspects of modelling a catchment’s rainfall-runoff relationship and dam operations relevant to assessing climate change impacts on dam crest flood frequency and will expand on the second to last paragraph of our discussion where we’ve outlined our simplifications. We will also include here that dam owners would be able to implement our analysis approaches for assessing climate change impacts on changes in the likelihood of dam crest floods using their models that could include more detailed representations of the catchment, dam operations, and initial reservoir levels.
- The Results section is somehow succinct and could benefit from more detailed analysis. For instance, the authors haven’t studied in more detail the hydrological response of each catchment to the changes in the rainfall parameters (rainfall losses, storm temporal patterns…). Figure 5 is presented as a black box without details and hardly exploitable.
We have derived detailed hydrologic results for each dam, and Figure 5 is intended to provide a summary that shows the variation in outcomes due to differences in dam and catchment configuration. We can prepare a summary table that shows the relative impacts of the different components on a dam-by-dam basis.
Discussion
- Line 367: the assumption made by the authors that climate change will lead to increasing the potential for dams to attenuate floods is not backed by the findings presented in this paper. I recommend to replace “will result” by “could result”.
We will replace “will result” with “likely result” given we follow up this statement with references to studies that have shown that climate change will increase demands for stored water resulting in lower initial reservoir levels.
- Paragraph lines 384 to 399: authors justify the use of outputs from global climate models instead of regional ones, while in reality none of these outputs have been used here. The authors have simply applied a series of temperature increases to the hydrological drivers, without relying on any climate model. This could lead to misunderstanding the process followed. I recommend to replace this paragraph by a justification of the simplified methodology proposed.
We will add an explicit explanation of how our results, using covariates of global temperature, can then be related to various scenarios of climate change for any future time horizon as shown in Fig. 3. We will also clarify that the rates of change used in our analysis are based on the results from previous studies that employed outputs from global and regional climate models to calculate the impact on hydrological drivers and their dependencies on global temperature change.
Conclusion
- Line 429: it is the first time that this result appears in the text. It should be mentioned before.
This comment is in relation to the statement that “current levels of global warming relative to the period used to inform historic flood risk estimates in Australia mean that the risk of floods exceeding the dam crest flood is already more than twice as probable for four of the 18 dams investigated”. This finding will be included in the results section rather than making a first appearance in the conclusion.
- Line 440: replace “practical approach” by “simplified approach”.
This comment is in relation to the final line: “Our study provides a practical approach for estimating extreme flood and dam overtopping risk under climate change that aligns with approaches widely used by practitioners making it feasible to be adopted globally.”
While we acknowledge that our implementation in representing rainfall and runoff for each dam was simplified, our method for assessing climate change on changes in the dam crest flood frequency, which is the primary focus of this paper, was not simplified. It was designed to be tractable in contrast to many existing top-down, scenario-driven methods of assessing climate change impacts on large floods. So, while it is, in comparison, simple, it is not a simplified approach. We will clarify this statement by adding: “Our study provides a practical and tractable approach”
Technical corrections
Unless a response is provided below, all the technical corrections will be made as recommended.
- Line 42: rephrase “moisture delivered”.
- Lines 56-57: consider revising the citation format.
- Line 59: repetition “that that”.
- Line 89: reference for the Australian Rainfall and Runoff.
- Table 1: instead of indicating dam owners, indicate dam type.
- Line 169: replace “The rates of changed” by “The rates of change”.
- Figure 4: replace “notional overtopping” in legend by “DCF”
- replace “Shift in overtopping risk” by “Shift in overtopping AEP”
- y-axis does not represent the AEP, but the return period.
We will revise the x-axis tick marks to be the AEP as the 1/Y terminology is obstructing the clarity of the figure.
- Figure 5: legend indicates that outliers are not shown, but it is the case in plots (b), (c) and (d)
Thank you for noticing this. The outliers should be shown for all plots including (a) and this will be corrected and the caption will be updated.
- Line 325: change “appear” to “appears”
Citation: https://doi.org/10.5194/hess-2024-403-AC1 -
AC3: 'Reply on RC1', Michelle Ho, 08 May 2025
Referee 1
General comments
This paper addresses a critical and underexplored issue—the implications of climate change on dam hydrological safety. The topic is highly relevant given the increasing frequency and severity of extreme weather events, making the findings applicable to dam owners, policymakers, and climate adaptation professionals globally. The study offers a robust method for addressing uncertainties, and its analysis of 18 large dams across varied Australian climate regions is particularly noteworthy.
However, while the study’s objectives are clear and the methodology is sound, there are areas for improvement:
- First of all, the title of the article clearly refers to the impact of climate change on dam overtopping. However, most of the paper focuses on the impact on the hydrological loads to dams, neglecting the specific aspects related to dam safety. Furthermore, the only considerations regarding dams are taken into account in a very simplified manner (e.g., the assumption of the reservoirs being at a full supply level prior to the storms). The article title does not embody the methodology and results presented, and cannot be accepted as a valid title.
To provide a more precise description of the study, we intend to replace the title of “The impact of climate change on dam overtopping flood risk” with “The impact of climate change on dam overtopping floods”. Please refer to our response in point 2 where we provide further explanation.
- Throughout the entire manuscript, the term “risk” is misused (even in the title). In the dam safety context (and in any context related to natural hazards), the risk is defined as the combination of a potential hazard and its consequences. However, in the manuscript only the occurrence of the hazard and its probability are studied and quantified. Therefore, the manuscript should be reviewed and the term “risk” should be adequately replaced by “probability” or “hazard”, depending on the case.
The estimation of flood exceedance probabilities is the basis of assessing hydrological risk-based designs in engineering and we had been using the term “risk” in this context. However, we did not clarify this context and we recognise that this application of the term is used variously in the literature. We therefore intend to replace this terminology throughout the manuscript, as well as the title, as suggested to provide a more precise description of our analysis.
- The methodology is quite dependent on the Australian context. The introduction and discussion sections could better frame the global implications of the findings, as they currently focus primarily on Australia. Specify what assumptions and methods can be replicated to other regions or contexts. In line 342, the authors say that the approach “could be translated globally for estimating flood risk under climate change elsewhere in the world”, which is not clear.
We thank the reviewer for pointing this out. The manuscript would be improved with providing specifics with respect to model and data requirements to make these methods globally applicable and we will do so.
- The graphical abstract includes a graph representing the dam crest flood level vs the exceedance probability:
- This graph is not represented in the article.
- The y-axis does not correspond to the exceedance probability.
The reviewer is correct – the graphical abstract is not from the article. It was specially designed for the purpose of providing an overview of the research and general results rather than to convey specific results and we believe that it serves this purpose.
The reviewer is also correct that the y-axis does to correspond to the exceedance probability – the y-axis represents the reservoir level and this was communicated using an icon of a dam with water levels along the y-axis. To clarify this, we will add a text label to the y-axis in addition to the icon and relocate the x-axis label to improve clarity.
Specific comments
Introduction
- The introduction provides a strong rationale for the study. However, it lacks a succinct statement of the research gap. Explicitly contrast the current study with previous works on flood risk and dam safety.
We believe that we have provided an explicit contrast between our study and previous work in the last two paragraphs of the introduction. However, we believe that addressing the reviewer’s recommendation of using precise terminology with respect to the word “risk” will improve the clarity of our statement explaining the research gap.
- The global context is underdeveloped. Adding examples from other regions (e.g., Europe or Asia) could broaden the impact.
The two studies we reference that have quantified changes in the probability of dam overtopping floods were located in Taiwan and Spain and additional reference to these locations will be added. We will clarify that there are few other studies that specifically investigate the changes in dam outflow flood frequencies under climate change.
Materials and methods
- Line 121: the description of the R2ORB emulator needs more detail. A brief explanation of how it works and its advantages compared to other tools would be helpful.
We will include more detail on the R2ORB emulator in Section 2.2. when R2ORB is first mentioned. We will revise the text to explain that R2ORB uses data inputs and performs the calculations outlined in Fig 2, which shows a schematic of the event-based modelling process. We will add that by using an emulator of RORB, we were able to focus our calculations on the aspects of flood hydrology modelling that are most relevant in the exploration of climate change impacts on dam hydrology, namely the catchment runoff-routing and reservoir routing to estimate peak reservoir outflows. Our approach to calibrating and validating the R2ORB models will also be included. We will also add that R2ORB follows the basic generic modelling structure of event-based conceptual rainfall runoff models, namely, partitioning of rainfall into losses and excess and attenuation of the flood through channels and reservoirs. There are many hydrological tools that can model runoff responses to rainfall at an event timescale and we do not think that a listing and comparison to these tools would be of material value to this paper.
- Line 129: “we assumed that the reservoir was at a full supply level prior to the storm”: this is a strong simplification of the methodology. Please justify:
- The reasons (lack of data, lack of time…)
- The potential impact (have you done an example calculation?)
In terms of dam safety, this is equivalent to not considering antecedent catchment wetness when calculating floods. I strongly recommend the authors to at least perform one example analysis of this effect. This is important in an article that focuses on dam safety (it’s in the title).
We will include in our justification for assuming a full supply level in all the dams that it provides a worst-case scenario for estimating the probability of a dam crest flood. We will also add that the modelling of reservoir levels considering climate change impacts had only been conducted for two of the dams and these were based on specific future scenarios making them incompatible with our analysis approach, which is based on changes in global temperature. While it is very easy to include initial reservoir level in a Monte Carlo analysis, the assumptions underlying the shifts in the marginal distribution are subject to deep uncertainty due to the future changes in operating conditions that are required, which in our experience have a much greater influence on reservoir levels than the change in antecedent conditions. We will provide some additional discussion on this point in the revised version of the manuscript.
- Line 134: “outflow hydrograph”: do you mean the catchment’s outflow hydrograph, or the reservoir’s outflow hydrograph? Please harmonize the vocabulary throughout the paper.
We will clarify early on the in the manuscript that the “inflow hydrograph” refers to the reservoir inflow hydrograph while the “outflow hydrograph” refers to the reservoir outflow hydrograph.
- More information on the Monte Carlo simulation framework is needed in Appendix A and in the description of the methodology.
Agreed. This will be addressed.
- Line 162: impacts of climate change on what?
This will be revised to read: “The impacts of climate change on the AEP of the DCF…”
- No mention to dam operations is presented here. What are the assumptions? How dams are considered operated in the baseline period and in the future periods? Just a short indication is given in line 138.
We will add that dam operations in response to a dam crest flood under climate change are assumed to remain the same as historical operations and that the operational impacts are encompassed in the relationships between reservoir height and outflow provided by the dam owners. This point is also relevant to our response to item 8) above.
Results
- The results are presented effectively, but their practical implications could be expanded to aid decision-makers.
We will include in the discussion some of the practical implications such as potentially re-evaluating spillway capacities and updating floodplain inundation maps.
- Line 263: specify the three flood drivers.
Thank you. This will be addressed.
- Line 272: “the reservoir outflow flood frequency curve” should be “the reservoir inflow flood frequency curve” because it refers to the DCF, which is the flood entering the dam’s reservoir.
The dam crest flood is dependent on the outflow rate from the dam and we are in fact showing the reservoir outflow flood frequency curves in Fig 4, referenced in L272. We will include that the outflow rate corresponding to dam crest reservoir levels is indicated in these figures.
- The RS factor does not illustrate the importance of the underlying AEP (i.e., how unsafe is a dam under historical conditions). When presented, results should simultaneously show the base AEP and the RS factor. Moreover, given that the results are anonymized and Figure 5 is just a summary of the results for the 18 dams analyzed, maybe an anonymized figure (similar to Figure 4, but only for the combined effect) in an Appendix could help shedding light on this issue.
We will show a summary of the outflow frequency curves across all dams in response to all three flood factors combined with the y-axis standardised by the dam crest flood level. We will also include in the text the range of historical dam crest flood AEPs across the 18 dams. We do not think it would be helpful or relevant to provide comment on how unsafe the dams are under historical conditions: apart from the fact that the owners would not want to see this point highlighted, the assessment of “dam safety” is a complex task that needs to take into account dam-specific engineering factors that lie outside the scope of this paper.
- Synthesize somewhere the simplifications assumed in the methodology and the potential improvements.
We will clarify that we have included the key aspects of modelling a catchment’s rainfall-runoff relationship and dam operations relevant to assessing climate change impacts on dam crest flood frequency and will expand on the second to last paragraph of our discussion where we’ve outlined our simplifications. We will also include here that dam owners would be able to implement our analysis approaches for assessing climate change impacts on changes in the likelihood of dam crest floods using their models that could include more detailed representations of the catchment, dam operations, and initial reservoir levels.
- The Results section is somehow succinct and could benefit from more detailed analysis. For instance, the authors haven’t studied in more detail the hydrological response of each catchment to the changes in the rainfall parameters (rainfall losses, storm temporal patterns…). Figure 5 is presented as a black box without details and hardly exploitable.
We have derived detailed hydrologic results for each dam, and Figure 5 is intended to provide a summary that shows the variation in outcomes due to differences in dam and catchment configuration. We can prepare a summary table that shows the relative impacts of the different components on a dam-by-dam basis.
Discussion
- Line 367: the assumption made by the authors that climate change will lead to increasing the potential for dams to attenuate floods is not backed by the findings presented in this paper. I recommend to replace “will result” by “could result”.
We will replace “will result” with “likely result” given we follow up this statement with references to studies that have shown that climate change will increase demands for stored water resulting in lower initial reservoir levels.
- Paragraph lines 384 to 399: authors justify the use of outputs from global climate models instead of regional ones, while in reality none of these outputs have been used here. The authors have simply applied a series of temperature increases to the hydrological drivers, without relying on any climate model. This could lead to misunderstanding the process followed. I recommend to replace this paragraph by a justification of the simplified methodology proposed.
We will add an explicit explanation of how our results, using covariates of global temperature, can then be related to various scenarios of climate change for any future time horizon as shown in Fig. 3. We will also clarify that the rates of change used in our analysis are based on the results from previous studies that employed outputs from global and regional climate models to calculate the impact on hydrological drivers and their dependencies on global temperature change.
Conclusion
- Line 429: it is the first time that this result appears in the text. It should be mentioned before.
This comment is in relation to the statement that “current levels of global warming relative to the period used to inform historic flood risk estimates in Australia mean that the risk of floods exceeding the dam crest flood is already more than twice as probable for four of the 18 dams investigated”. This finding will be included in the results section rather than making a first appearance in the conclusion.
- Line 440: replace “practical approach” by “simplified approach”.
This comment is in relation to the final line: “Our study provides a practical approach for estimating extreme flood and dam overtopping risk under climate change that aligns with approaches widely used by practitioners making it feasible to be adopted globally.”
While we acknowledge that our implementation in representing rainfall and runoff for each dam was simplified, our method for assessing climate change on changes in the dam crest flood frequency, which is the primary focus of this paper, was not simplified. It was designed to be tractable in contrast to many existing top-down, scenario-driven methods of assessing climate change impacts on large floods. So, while it is, in comparison, simple, it is not a simplified approach. We will clarify this statement by adding: “Our study provides a practical and tractable approach”
Technical corrections
Unless a response is provided below, all the technical corrections will be made as recommended.
- Line 42: rephrase “moisture delivered”.
- Lines 56-57: consider revising the citation format.
- Line 59: repetition “that that”.
- Line 89: reference for the Australian Rainfall and Runoff.
- Table 1: instead of indicating dam owners, indicate dam type.
- Line 169: replace “The rates of changed” by “The rates of change”.
- Figure 4: replace “notional overtopping” in legend by “DCF”
- replace “Shift in overtopping risk” by “Shift in overtopping AEP”
- y-axis does not represent the AEP, but the return period.
We will revise the x-axis tick marks to be the AEP as the 1/Y terminology is obstructing the clarity of the figure.
- Figure 5: legend indicates that outliers are not shown, but it is the case in plots (b), (c) and (d)
Thank you for noticing this. The outliers should be shown for all plots including (a) and this will be corrected and the caption will be updated.
- Line 325: change “appear” to “appears”
Citation: https://doi.org/10.5194/hess-2024-403-AC3
-
RC2: 'Comment on hess-2024-403', Anonymous Referee #2, 01 Apr 2025
The topic is interesting and the manuscript is generally presented well. I found the consideration of temporal pattern innovative as this rainfall property is typically overlooked. Assumptions and methods are unclear though. Additionally, the results are not interesting/surprising and the discussion do not offer much insights. Here are my detailed comments:
- Title: Add “Australia” to the title for clarity.
- Graphical abstract: Write “Australia” on the map for clarity. Also, please clarify what variables you consider in your overtopping analyses (dam crest height, flood pool, maximum water level etc.) in the dam schematic (bottom right plot).
- Abstract: Please clarify the future period and variables you consider in your overtopping analyses (dam crest height, flood pool, maximum water level etc.).
- Discussion of threats posed by aging dams (e.g., Ferdowsi et al. 2024; Shirzaei et al. 2025) may strengthen the introduction of your manuscript.
- Past research about dam overtopping, water level and inflow should be further acknowledged (e.g., Kwon & Moon 2006; Kuo et al. 2007; Hsu et al. 2011; Michailidi & Bacchi 2017; Wang & Zhang 2017; Cho et al. 2024, 2025).
- The future period should be mentioned in the last paragraph of introduction section.
- Any reason for selecting 1961-1990 as the historic period and not a more recent period?
- L96: How did you classify the dams as large? Was it based on the ICOLD classification? Please clarify.
- The methodology is unclear. I suggest adding a schematic view of your overall methodology for estimating the dam overtopping. Also add a short write-up about how your methods connect to each other.
- Any reason for the selection of the 18 dams among other dams in Australia?
- Are the 18 dams dependent on each other in terms of the inflows and water levels?
- Were the R2ORB models calibrated and validated? How did you account for changes in the land cover and surface roughness in the catchments? It would be helpful to discuss the changes in the historic period.
- Did you use the rainfall at the dam location or across the upstream catchment? The same question applies to antecedent soil moisture.
- Please show the study subcatchments in a supplement figure.
- How were future rainfall time series generated? Did you use any GCMs? Any bias correction and downscaling? Details are needed.
- PMP can be estimated via different methods. How sensitive your analyses with respect to the selected method?
- Rate of change factors were estimated based on climate zones. Would these remain stable under future climate?
- Please discuss your methodology for combing AEPs of the three rainfall characteristics.
- 14a: Can you elaborate on the zone that shows a shift in overtopping risk?
- 14b: Can you show the temporal patterns of rainfall events as dimensionless time series showing the fraction of total rainfall?
- The term overtopping risk should be replaced with overtopping probability or hazard as you do not investigate the consequences of dam overtopping.
- Please add a map that shows the overtopping probability under historic and future conditions.
- In general, I do not find any interesting results. Can you highlight the key findings of your paper?
- As acknowledged by the authors, the overtopping is a result of multiple factors acting together. As such, the reliability of this study is questionable.
- L245: Please revise the sentence.
- Sources of uncertainty and their impact on your results should be discussed.
- Section 5: Conclusion->Conclusions
Citation: https://doi.org/10.5194/hess-2024-403-RC2 -
AC2: 'Reply on RC2', Michelle Ho, 08 May 2025
Referee 2
General comments
The topic is interesting and the manuscript is generally presented well. I found the consideration of temporal pattern innovative as this rainfall property is typically overlooked. Assumptions and methods are unclear though. Additionally, the results are not interesting/surprising and the discussion do not offer much insights. Here are my detailed comments:
- Title: Add “Australia” to the title for clarity.
We intend to revise the title to “The impact of climate change on dam overtopping floods in Australia”.
- Graphical abstract: Write “Australia” on the map for clarity. Also, please clarify what variables you consider in your overtopping analyses (dam crest height, flood pool, maximum water level etc.) in the dam schematic (bottom right plot).
“Australia” will be added to the map. As also recommended by Reviewer 1, we will add “reservoir level” to the y-axis of the figure.
- Abstract: Please clarify the future period and variables you consider in your overtopping analyses (dam crest height, flood pool, maximum water level etc.).
We did not limit our analysis to a specific future period as our analysis was dependent on degrees of global warming.
- Discussion of threats posed by aging dams (e.g., Ferdowsi et al. 2024; Shirzaei et al. 2025) may strengthen the introduction of your manuscript.
Thank you for these references – they will be included in our introduction. The article by Ferdowsi et al. (2024) provides a succinct argument for proactive dam safety assessments in response to climate change and increasing hydrologic extremes particularly in the Global South. A reference to this article would provide a better discussion of the global context in our introduction. The article by Shirzaei et al. (2025) provides context for how climate change is but one component in dam risk assessments and disaster mitigation.
- Past research about dam overtopping, water level and inflow should be further acknowledged (e.g., Kwon & Moon 2006; Kuo et al. 2007; Hsu et al. 2011; Michailidi & Bacchi 2017; Wang & Zhang 2017; Cho et al. 2024, 2025).
Thank you for recommending these papers. I’m afraid at this stage we were unable to locate the article by Cho et al. 2025 but will continue searching, but the remaining papers are all relevant to assessing dam overtopping floods and will be referred to in our introduction and discussion.
- The future period should be mentioned in the last paragraph of introduction section.
This will be included.
- Any reason for selecting 1961-1990 as the historic period and not a more recent period?
Yes, we provide a justification for using this period as the historic period around L90. Namely: “The historic period approximates the mid-point for much of the information used to derive the design information provided in Australian Rainfall and Runoff (the national flood guidelines for Australia (Ball et al., 2019)), which establishes a baseline of historic flood risk with which to compare climate change impacts.”
- L96: How did you classify the dams as large? Was it based on the ICOLD classification? Please clarify.
We will clarify under Section 2.1.Case study locations that the 18 case study dams are classified as “large” based on the ICOLD classification of large dams as all the dams have a foundation to crest height exceeding 15 m.
- The methodology is unclear. I suggest adding a schematic view of your overall methodology for estimating the dam overtopping. Also add a short write-up about how your methods connect to each other.
A schematic is provided in Figure 2. We will provide additional reference to this figure throughout the description of the methodology to improve clarity on how the different method sections connect.
- Any reason for the selection of the 18 dams among other dams in Australia?
The 18 dams are owned by authorities who expressed an interest in examining the change in their dam exposure to hydrological risk under climate change and provided support to this project with respect to financial support and sharing of data and models. The sharing of data and models will be stated in Section 2.1. Case study locations, while the financial support has been disclosed in the Acknowledgements.
- Are the 18 dams dependent on each other in terms of the inflows and water levels?
These dams are all independent of each other with the exception of Somerset and Wivenhoe and a statement reflecting this will be included in Section 2.1.
- Were the R2ORB models calibrated and validated? How did you account for changes in the land cover and surface roughness in the catchments? It would be helpful to discuss the changes in the historic period.
We will provide additional detail on how the R2ORB models were calibrated and validated to RORB models used in practice for design flood estimation. We will include our assumption that changes in land cover and catchment surface roughness were not considered as this was beyond the scope of the study. Furthermore, we are assessing relative changes resulting from climate change impacts rather than changes in catchment morphology over time. Please also see our response to reviewer 1, comment number 7.
- Did you use the rainfall at the dam location or across the upstream catchment? The same question applies to antecedent soil moisture.
Design rainfall depths were applied at the approximate centroid of each catchment sub-area. The rainfall and antecedent soil moisture data used to inform the rates of change in initial and continuing loss were based gridded data that were spatially averaged across the over 200 catchments used to inform the rates of change.
- Please show the study subcatchments in a supplement figure.
We will provide maps showing how we represented the subcatchments for all the case studies in a supplement figure.
- How were future rainfall time series generated? Did you use any GCMs? Any bias correction and downscaling? Details are needed.
We did not use projections of rainfall time series. We used a rate of change applied to a design rainfall depth described in section 2.3.1. The rate of change that we adopted is based on a published systematic review of observed historical trends, relationships between extreme rainfall and temperature, and findings modelled using both general circulation models and regional models. We will include this additional detail in section 2.3.1 that describes how we consider rainfall depth.
- PMP can be estimated via different methods. How sensitive your analyses with respect to the selected method?
Our results suggested that the change in overtopping flood probability was most influenced by changes in rainfall depth. The historic estimate of PMP is therefore a baseline of our analysis and different estimates of PMP using different methods will consequently change our results. The PMP method used here is based on generalised hydrometeorological methods as advocated by the WMO (2009), which is commonly used across the world. It is worth noting that published studies of the impacts of climate change on statistical estimates of PMP (eg Herschfield) yield similar results as hydrometeorological methods. These points are discussed in Visser et al (2022) and Wasko et al (2024), and we will make reference to these in our Discussion.
- Rate of change factors were estimated based on climate zones. Would these remain stable under future climate?
The rate of change factors for storm temporal patterns were based on Köppen-Geiger zones, while the rate of change factors for rainfall losses are based on Natural Resource Management region. The climate zones will likely change into the future (Beck et al. 2018). However, the rate change factors are based on trends in temporal patterns with current climate zones used as a method for classifying results. We do not have information on how these scaling rates would change if the results were to be classified by future climate zone definitions. We will include this caveat when rate of change factors for temporal patterns are introduced in Section 2.3.2.
Beck, H. E., Zimmermann, N. E., McVicar, T. R., Vergopolan, N., Berg, A., and Wood, E. F.: Present and future Köppen-Geiger climate classification maps at 1-km resolution, Sci Data, 5, 180214, https://doi.org/10.1038/sdata.2018.214, 2018.
- Please discuss your methodology for combing AEPs of the three rainfall characteristics.
Thank you for noting this omission. We will include a description of how the three flood characteristics under climate change are assessed individually and in combination in Section 2.3. Assessing impacts of climate change.
- 14a: Can you elaborate on the zone that shows a shift in overtopping risk?
We assume this comment is in reference to Figure 4a showing the shift in the AEP of the overtopping flood between historic flood probabilities and those under 5°C of global warming. We will reference the different degrees of global warming and related shifts in overtopping flood AEP within this zone in the paragraph under Figure 4.
- 14b: Can you show the temporal patterns of rainfall events as dimensionless time series showing the fraction of total rainfall?
The figure attached shows the temporal patterns used for the different catchments. While this figure could be included in an appendix or Supplementary Material, we believe that providing these temporal patterns could distract the reader from the focus of the paper.
- The term overtopping risk should be replaced with overtopping probability or hazard as you do not investigate the consequences of dam overtopping.
We will revise the terminology. Please refer to our response to Reviewer 1 point 2 for more detail.
- Please add a map that shows the overtopping probability under historic and future conditions.
The shift in overtopping risks is a function of differences in catchment characteristics, configuration of a dam’s outlet works, and hydroclimatic region. Proving a map of the differences only relates to one of these factors, and thus is potentially misleading. We will provide some additional comment on this point in the Discussion.
- In general, I do not find any interesting results. Can you highlight the key findings of your paper?
The results may not be “interesting” in the sense that they are consistent with the general expectation that the probability of dam overtopping floods will increase with global warming. Our results show that on average dams are seven times more likely to be overtopped under a plausible climate scenario and we believe that these results are of considerable scientific and practical interest. To our knowledge this is the first time that projections have been systematically quantified for a wide range of hydroclimatic regions giving consideration to rainfall depths, antecedent conditions, and temporal patterns. This is both of scientific interest, as the projections are based on our most current understanding of climate science as published in Wasko et al. (2024), and of direct relevance to industry, as they are based on models and procedures that are currently being used in engineering design practice. We will ensure that these aspects are highlighted in the abstract and conclusions.
- As acknowledged by the authors, the overtopping is a result of multiple factors acting together. As such, the reliability of this study is questionable.
Our approach explicitly considers the joint probabilities of the factors impacting floods using the best available climate science. We agree that there are deep uncertainties associated with the trajectory of global warming and the dynamic factors that vary by location, and that there are additional uncertainties associated with hydroclimatic dependencies. That said, we need some basis to understand how climate change may impact on dam safety in the future. We consider the adopted methodology and derived results to be a useful contribution, despite the current irreducible uncertainties involved. We will ensure that these uncertainties are highlighted, noting that such uncertainties undermine confidence in all such climate-impact studies.
- L245: Please revise the sentence.
This original sentence was: “The catchments used in the study by Ho et al. (2023) were selected where a statistically significant relationship (at a significance level of a = 0.05) could be found between losses and antecedent soil moisture for 3-day rainfall events that were equalled or exceeded, on average, 5 times per year (a 5 EY event).”
This will be revised to: “The catchments included in the study by Ho et al. (2023) were those where a statistically significant relationship (at a significance level of a = 0.05) was found between losses and antecedent soil moisture.”
- Sources of uncertainty and their impact on your results should be discussed.
We believe that we have provided a discussion of the sources of uncertainty and likely impacts on our analysis in the discussion: We discuss plausible impacts of shifts in temporal patterns in the second paragraph, the assumption of reservoirs starting at a full supply level in the fourth paragraph, and the use of approximate central tendencies for the rates of change in the sixth paragraph
- Section 5: Conclusion->Conclusions
This will be corrected as suggested
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 355 | 89 | 13 | 457 | 11 | 9 |
- HTML: 355
- PDF: 89
- XML: 13
- Total: 457
- BibTeX: 11
- EndNote: 9
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1