the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 22 Jul 2024
Brief Communication: Stay local or go global? On the construction of plausible counterfactual scenarios to assess flash flood hazards
Abstract. Spatial counterfactuals are gaining attention to address the lack of robust flood frequency analysis in small catchments. However, the credibility of counterfactual scenarios decreases with the distance rain fields are shifted across space. We limit that distance by a local counterfactual search design, and compare the corresponding scenarios to recently published results from large spatial shifts. We then put all scenarios in context with 200-year return levels, and with flood peaks simulated for the June 2024 flood event in southern Germany. We conclude that local counterfactuals scenarios are transparent and credible, and could complement the anticipation of low probability events.
- Preprint
(1176 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on nhess-2024-119', Anonymous Referee #1, 08 Aug 2024
The authors present a brief study of what they term local vs. global counterfactual flood scenarios, in which observed high-resolution precipitation from extreme storms in Germany are transposed a short distance (local) and a long distance (global) and the resulting floods are resolved in a large number of small-to-medium-sized watersheds in Germany. The study builds upon several previous papers (Merz et al., 2024 and Voit and Heistermann, 2024) that explore storm transposition in Germany.
The study is well-written and could, with careful consideration of the major issue described below, be worthy of publication. I will note also that my criticism applies to Merz et al. (2024) and Voit and Heistermann (2024).
In what I consider to be a failure of the literature review and peer review process, the present study and those previous ones neglect the century-long massive body of relevant research that in multiple respects is much more advanced what the authors present in terms of level of sophistication and importantly, real-world application. Namely, geographic transposition of storms to generate counterfactual flood scenarios has been a cornerstone of flood risk management in North America, Australia, and elsewhere (perhaps not in Europe) for about as long as flood risk management has been quantified. As explained in Wright et al. (2020), Fuller’s pioneering 1914 work on flood frequency described transposition of flood quantiles; this quickly evolved into transposing rainstorms (e.g., Morgan, 1916) and associated rainfall-runoff modeling, including for Probable Maximum Precipitation (PMP; e.g., WMO 2009, Hansen 1987) and probable maximum flood estimation. It is worth emphasizing that this PMP work was not just academic in nature; literally tens to hundreds of billions of dollars of high-value, high-risk infrastructure decisions in various continents have been based upon it.
Furthermore, as described in Wright et al. (2020), the notion of transposition to create counterfactuals was further developed into a rigorous probabilistic rainfall and flood frequency analysis framework via Stochastic Storm Transposition (SST), beginning in the early 1960s and continuing to this day. While SST has found less (though not zero) real-world application compared with PMP, this is changing in both the USA and Australia (see Nathan et al. 2016 for some relatively recent details on the latter). Recent SST studies have included applications with high-resolution long-term radar rainfall datasets (e.g., Wright et al. 2014). SST remains a very active research area in the United States and elsewhere, and is in the process of being deployed for nationwide flood mapping in the U.S. via FEMA and the widely-used HEC-HMS modeling software.
A key consideration in both PMP and SST-based storm transposition work is the question of how far to transpose. In fact, the answer, which remains elusive despite decades of work, is less about seeking “X kilometers is fine, Y kilometers is too far” and much more about under what meteorological, topographical, and other physiographic conditions a storm that occurred in one place could occur in a location of interest, and whether a transposed storm needs to be altered in some way, e.g. to reflect orographic differences. There are no simple answers here, and much of the prior work is buried in “grey literature” (particularly old reports from the US Weather Bureau that can be difficult to find online). Statistical approaches can be found in Nathan et al. (2016), Yu et al. (2021), Zhou et al. (2019), and elsewhere. It is important to note that the authors’ work of local vs. global transposition is quite simplistic compared to this prior work, some of which provides solutions that largely though not completely obviate the whole question.
Additional work using transposition that employs somewhat similar language to that employed by the authors (e.g., bottom-up analysis of possible/plausible flood scenarios) include Hayden et al. (2016).
I am not suggesting that any author intentionally overlooked this previous work. In light of the very large body of highly-related and arguably more advanced methods, however, the present study, as well as Merz et al. (2024) and Voit and Heistermann (2024) should be re-evaluated in terms of the novelty of their contributions. Proper acknowledgement of the efforts of earlier researchers is certainly needed. Unfortunately, the scope for doing so in the case of the two already-published studies is limited. In order for the present study to be published in NHESS or any other publication, the authors would need to adequately review and cite existing research and practice, as well as carefully place their own work within this much broader 100+-year long body of work.
Minor comments:
Generally: there’s no reason why this approach needs to be restricted to flash floods
Generally: Given the major criticism above, it would be appropriate to adopt existing language (e.g., “transposing” instead of “shifting”)
L15: the meaning of “small-scale observational records” is unclear
L23: Should the language be clarified to make clear that it is an observed heavy precipitation event that is transposed?
Figures 2 and 3: these figures didn’t render properly in the PDF I downloaded, using two different widely-used PDF viewers on a Mac. Given that, I can’t properly assess these figures.
References:
Fuller, W.E. “Flood Flows.” ASCE Trans. 77 (1914): 567–617.
Hansen, E Marshall. “Probable Maximum Precipitation for Design Floods in the United States.” Journal of Hydrology 96, no. 1–4 (1987): 267–78. https://doi.org/10.1016/0022-1694(87)90158-2.
Hayden, Nicholas G., Kenneth W. Potter, and David S. Liebl. “Evaluating Infiltration Requirements for New Development Using Extreme Storm Transposition: A Case Study from Dane County, WI.” JAWRA Journal of the American Water Resources Association 52, no. 5 (October 1, 2016): 1170–78. https://doi.org/10.1111/1752-1688.12441.
Morgan, A., 1916. Official Plan for the Protection of the District From Flood Damage. Miami Conservation District, Dayton, Ohio.
Nathan, Rory, Phillip Jordan, Matthew Scorah, Simon Lang, George Kuczera, Melvin Schaefer, and Erwin Weinmann. “Estimating the Exceedance Probability of Extreme Rainfalls up to the Probable Maximum Precipitation.” Journal of Hydrology 543 (2016): 706–20. https://doi.org/10.1016/j.jhydrol.2016.10.044.
WMO. “Manual on Estimation of Probable Maximum Precipitation (PMP).” World Meteorological Organization, 2009.
Wright, D.B., James A. Smith, and Mary Lynn Baeck. “Flood Frequency Analysis Using Radar Rainfall Fields and Stochastic Storm Transposition.” Water Resources Research 50, no. 2 (February 2014): 1592–1615. https://doi.org/10.1002/2013WR014224.
Wright, Daniel B., Guo Yu, and John F. England. “Six Decades of Rainfall and Flood Frequency Analysis Using Stochastic Storm Transposition: Review, Progress, and Prospects.” Journal of Hydrology 585 (June 1, 2020): 124816. https://doi.org/10.1016/j.jhydrol.2020.124816.
Yu, Guo, Daniel B. Wright, and Kathleen D. Holman. “Connecting Hydrometeorological Processes to Low-Probability Floods in the Mountainous Colorado Front Range.” Water Resources Research 57, no. 4 (April 1, 2021): e2021WR029768. https://doi.org/10.1029/2021WR029768.
Zhou, Zhengzheng, James A. Smith, Daniel B. Wright, Mary Lynn Baeck, Long Yang, and Shuguang Liu. “Storm Catalog-Based Analysis of Rainfall Heterogeneity and Frequency in a Complex Terrain.” Water Resources Research 55, no. 3 (March 1, 2019): 1871–89. https://doi.org/10.1029/2018WR023567.Citation: https://doi.org/10.5194/nhess-2024-119-RC1 - AC1: 'Reply on RC1', Paul Voit, 23 Aug 2024
-
RC2: 'Comment on nhess-2024-119', Anonymous Referee #2, 20 Sep 2024
Review of the manuscript „Brief communication: Stay local or go global? On the construction of plausible counterfactual scenarios to access flash flood hazards by Voit and Heistermann.
In this manuscript, Voit and Heistermann compared two types of counterfactual analysis. First, the global counterfactual analysis as known from a previous study based on global shifting across the study domain. Second, a local method that focuses on extreme rainfall in neighbour catchments to the catchment under study (local counterfactual analysis) was applied.
This study is highly relevant, the research focus is clearly stated and motivated. The article is well written and I recommend only a few changes.
Comments:
L.23: Please define the abbreviation HPE, preferably at the first occurrence of the full name.
L.52: Formatting error in „HQextrem”
L.75: The SCS-CN method is often criticized for its simplicity and its empirical character. Could you justify why this method is appropriate for calculating infiltration and surface runoff in your study?
L.79: More information about the hydrologic model is needed. Which hydrological model is used? If an own model is used, please mention this.
L.91: “We model the quick runoff” is a too short for the methods. Could you elaborate more how you have modelled quick runoff?
L.93: In FFA studies, at least 30 years are usually used to calculate the annual maxima series and to generate the distribution function. Please explain in more detail why 23 years are sufficient in your analysis.
Fig.2: Please add a,b,c to the subplots.
Fig. 2: Description of X-axis and Y-axis is missing
Fig. 2: What is the meaning of “0 50” and so on? Even if it is explained in the caption, the subplot titles should be more specific.
Fig. 2: And add an explanation for two lines (green, blue) to the legend. At the moment, only one line (red) is explained.
Fig. 3: Description of X-axis and Y-axis is missing.
Fig. 3: What does “0 50” and so on mean? Even if it is explained in the caption, the subplot titles should be more precise.
Fig. 3: Why m³/s/km² 0 6 and Why 2 0 6?
Fig. 3: The legend needs to be reformatted. The lines are partly covered by the legend. The green line has no explanation. The explanation for the other two lines (red, blue) should be presented consistently, either on the left or the right-side
Results and discussion: Could you perhaps add sub-chapters to make it easier to read?
Citation: https://doi.org/10.5194/nhess-2024-119-RC2 - AC2: 'Reply on RC2', Paul Voit, 02 Oct 2024
Status: closed
-
RC1: 'Comment on nhess-2024-119', Anonymous Referee #1, 08 Aug 2024
The authors present a brief study of what they term local vs. global counterfactual flood scenarios, in which observed high-resolution precipitation from extreme storms in Germany are transposed a short distance (local) and a long distance (global) and the resulting floods are resolved in a large number of small-to-medium-sized watersheds in Germany. The study builds upon several previous papers (Merz et al., 2024 and Voit and Heistermann, 2024) that explore storm transposition in Germany.
The study is well-written and could, with careful consideration of the major issue described below, be worthy of publication. I will note also that my criticism applies to Merz et al. (2024) and Voit and Heistermann (2024).
In what I consider to be a failure of the literature review and peer review process, the present study and those previous ones neglect the century-long massive body of relevant research that in multiple respects is much more advanced what the authors present in terms of level of sophistication and importantly, real-world application. Namely, geographic transposition of storms to generate counterfactual flood scenarios has been a cornerstone of flood risk management in North America, Australia, and elsewhere (perhaps not in Europe) for about as long as flood risk management has been quantified. As explained in Wright et al. (2020), Fuller’s pioneering 1914 work on flood frequency described transposition of flood quantiles; this quickly evolved into transposing rainstorms (e.g., Morgan, 1916) and associated rainfall-runoff modeling, including for Probable Maximum Precipitation (PMP; e.g., WMO 2009, Hansen 1987) and probable maximum flood estimation. It is worth emphasizing that this PMP work was not just academic in nature; literally tens to hundreds of billions of dollars of high-value, high-risk infrastructure decisions in various continents have been based upon it.
Furthermore, as described in Wright et al. (2020), the notion of transposition to create counterfactuals was further developed into a rigorous probabilistic rainfall and flood frequency analysis framework via Stochastic Storm Transposition (SST), beginning in the early 1960s and continuing to this day. While SST has found less (though not zero) real-world application compared with PMP, this is changing in both the USA and Australia (see Nathan et al. 2016 for some relatively recent details on the latter). Recent SST studies have included applications with high-resolution long-term radar rainfall datasets (e.g., Wright et al. 2014). SST remains a very active research area in the United States and elsewhere, and is in the process of being deployed for nationwide flood mapping in the U.S. via FEMA and the widely-used HEC-HMS modeling software.
A key consideration in both PMP and SST-based storm transposition work is the question of how far to transpose. In fact, the answer, which remains elusive despite decades of work, is less about seeking “X kilometers is fine, Y kilometers is too far” and much more about under what meteorological, topographical, and other physiographic conditions a storm that occurred in one place could occur in a location of interest, and whether a transposed storm needs to be altered in some way, e.g. to reflect orographic differences. There are no simple answers here, and much of the prior work is buried in “grey literature” (particularly old reports from the US Weather Bureau that can be difficult to find online). Statistical approaches can be found in Nathan et al. (2016), Yu et al. (2021), Zhou et al. (2019), and elsewhere. It is important to note that the authors’ work of local vs. global transposition is quite simplistic compared to this prior work, some of which provides solutions that largely though not completely obviate the whole question.
Additional work using transposition that employs somewhat similar language to that employed by the authors (e.g., bottom-up analysis of possible/plausible flood scenarios) include Hayden et al. (2016).
I am not suggesting that any author intentionally overlooked this previous work. In light of the very large body of highly-related and arguably more advanced methods, however, the present study, as well as Merz et al. (2024) and Voit and Heistermann (2024) should be re-evaluated in terms of the novelty of their contributions. Proper acknowledgement of the efforts of earlier researchers is certainly needed. Unfortunately, the scope for doing so in the case of the two already-published studies is limited. In order for the present study to be published in NHESS or any other publication, the authors would need to adequately review and cite existing research and practice, as well as carefully place their own work within this much broader 100+-year long body of work.
Minor comments:
Generally: there’s no reason why this approach needs to be restricted to flash floods
Generally: Given the major criticism above, it would be appropriate to adopt existing language (e.g., “transposing” instead of “shifting”)
L15: the meaning of “small-scale observational records” is unclear
L23: Should the language be clarified to make clear that it is an observed heavy precipitation event that is transposed?
Figures 2 and 3: these figures didn’t render properly in the PDF I downloaded, using two different widely-used PDF viewers on a Mac. Given that, I can’t properly assess these figures.
References:
Fuller, W.E. “Flood Flows.” ASCE Trans. 77 (1914): 567–617.
Hansen, E Marshall. “Probable Maximum Precipitation for Design Floods in the United States.” Journal of Hydrology 96, no. 1–4 (1987): 267–78. https://doi.org/10.1016/0022-1694(87)90158-2.
Hayden, Nicholas G., Kenneth W. Potter, and David S. Liebl. “Evaluating Infiltration Requirements for New Development Using Extreme Storm Transposition: A Case Study from Dane County, WI.” JAWRA Journal of the American Water Resources Association 52, no. 5 (October 1, 2016): 1170–78. https://doi.org/10.1111/1752-1688.12441.
Morgan, A., 1916. Official Plan for the Protection of the District From Flood Damage. Miami Conservation District, Dayton, Ohio.
Nathan, Rory, Phillip Jordan, Matthew Scorah, Simon Lang, George Kuczera, Melvin Schaefer, and Erwin Weinmann. “Estimating the Exceedance Probability of Extreme Rainfalls up to the Probable Maximum Precipitation.” Journal of Hydrology 543 (2016): 706–20. https://doi.org/10.1016/j.jhydrol.2016.10.044.
WMO. “Manual on Estimation of Probable Maximum Precipitation (PMP).” World Meteorological Organization, 2009.
Wright, D.B., James A. Smith, and Mary Lynn Baeck. “Flood Frequency Analysis Using Radar Rainfall Fields and Stochastic Storm Transposition.” Water Resources Research 50, no. 2 (February 2014): 1592–1615. https://doi.org/10.1002/2013WR014224.
Wright, Daniel B., Guo Yu, and John F. England. “Six Decades of Rainfall and Flood Frequency Analysis Using Stochastic Storm Transposition: Review, Progress, and Prospects.” Journal of Hydrology 585 (June 1, 2020): 124816. https://doi.org/10.1016/j.jhydrol.2020.124816.
Yu, Guo, Daniel B. Wright, and Kathleen D. Holman. “Connecting Hydrometeorological Processes to Low-Probability Floods in the Mountainous Colorado Front Range.” Water Resources Research 57, no. 4 (April 1, 2021): e2021WR029768. https://doi.org/10.1029/2021WR029768.
Zhou, Zhengzheng, James A. Smith, Daniel B. Wright, Mary Lynn Baeck, Long Yang, and Shuguang Liu. “Storm Catalog-Based Analysis of Rainfall Heterogeneity and Frequency in a Complex Terrain.” Water Resources Research 55, no. 3 (March 1, 2019): 1871–89. https://doi.org/10.1029/2018WR023567.Citation: https://doi.org/10.5194/nhess-2024-119-RC1 - AC1: 'Reply on RC1', Paul Voit, 23 Aug 2024
-
RC2: 'Comment on nhess-2024-119', Anonymous Referee #2, 20 Sep 2024
Review of the manuscript „Brief communication: Stay local or go global? On the construction of plausible counterfactual scenarios to access flash flood hazards by Voit and Heistermann.
In this manuscript, Voit and Heistermann compared two types of counterfactual analysis. First, the global counterfactual analysis as known from a previous study based on global shifting across the study domain. Second, a local method that focuses on extreme rainfall in neighbour catchments to the catchment under study (local counterfactual analysis) was applied.
This study is highly relevant, the research focus is clearly stated and motivated. The article is well written and I recommend only a few changes.
Comments:
L.23: Please define the abbreviation HPE, preferably at the first occurrence of the full name.
L.52: Formatting error in „HQextrem”
L.75: The SCS-CN method is often criticized for its simplicity and its empirical character. Could you justify why this method is appropriate for calculating infiltration and surface runoff in your study?
L.79: More information about the hydrologic model is needed. Which hydrological model is used? If an own model is used, please mention this.
L.91: “We model the quick runoff” is a too short for the methods. Could you elaborate more how you have modelled quick runoff?
L.93: In FFA studies, at least 30 years are usually used to calculate the annual maxima series and to generate the distribution function. Please explain in more detail why 23 years are sufficient in your analysis.
Fig.2: Please add a,b,c to the subplots.
Fig. 2: Description of X-axis and Y-axis is missing
Fig. 2: What is the meaning of “0 50” and so on? Even if it is explained in the caption, the subplot titles should be more specific.
Fig. 2: And add an explanation for two lines (green, blue) to the legend. At the moment, only one line (red) is explained.
Fig. 3: Description of X-axis and Y-axis is missing.
Fig. 3: What does “0 50” and so on mean? Even if it is explained in the caption, the subplot titles should be more precise.
Fig. 3: Why m³/s/km² 0 6 and Why 2 0 6?
Fig. 3: The legend needs to be reformatted. The lines are partly covered by the legend. The green line has no explanation. The explanation for the other two lines (red, blue) should be presented consistently, either on the left or the right-side
Results and discussion: Could you perhaps add sub-chapters to make it easier to read?
Citation: https://doi.org/10.5194/nhess-2024-119-RC2 - AC2: 'Reply on RC2', Paul Voit, 02 Oct 2024
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 239 | 70 | 96 | 405 | 10 | 10 |
- HTML: 239
- PDF: 70
- XML: 96
- Total: 405
- BibTeX: 10
- EndNote: 10
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1