the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 27 Mar 2026
Effect of Design Storm Characterization on Flood Exposure and Structure Damage Estimates: A Case Study in South Louisiana, USA
Abstract. Quantitative flood risk assessments rely on rainfall frequency analysis to define Annual exceedance probability (AEP) storms, commonly referred to as design storms, for generating flood hazard maps and expected annual damage curves. Current engineering practice typically employs spatially uniform design storms derived from point-based gauge statistics; however, this approach suppresses the spatial organization and intensity gradients present in observed storms. Stochastic storm transposition (SST) offers an alternative by preserving the spatial structure of observed rainfall and stochastically repositioning full storm fields across a watershed. Although recent work shows that SST-based design storms can alter peak discharge estimates relative to uniform storms, their implications for flood inundation mapping and for estimating structural damage and their resulting monetary losses remain understudied. This study addresses that gap by comparing flood exposure and damages produced by both design-storm approaches in the Vermilion River Basin, a low-gradient inland–coastal watershed in south-central Louisiana. The comparison reveals that SST identifies 11,518 buildings inundated in at least one SST realization but missed entirely by Atlas 14, representing over ∼$110 M in cumulative structural damages. The divergence between the two approaches is concentrated in mid-elevation urban neighborhoods, where spatial rainfall variability activates flooding thresholds that uniform storms cannot trigger. These results demonstrate that uniform design storms systematically underestimate both the catastrophic tail and the breadth of flood exposure in low-gradient basins.
- Preprint
(2480 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 08 May 2026)
- RC1: 'Comment on egusphere-2026-1388', Anonymous Referee #1, 16 Apr 2026 reply
-
RC2: 'Comment on egusphere-2026-1388', Daniel Wright, 19 Apr 2026
reply
The authors present a comparison of flood inundation and damage estimates from 100-year storms developed using two independent techniques: the conventional design storm approach using Atlas 14 rainfall frequency estimates, and fifty realizations of rainstorms from stochastic storm transposition (SST). The manuscript is well written and the findings are useful. I do have some suggestions, including at least one new hydrologic/hydraulic simulation and some additional discussion that could improve the manuscript prior to publication. Of particular note in this study is the demonstration of localized high rainfall rates driving flooding in “intermediate areas” that are neither adjacent to rivers nor in upland areas. This is important. It is striking how different the basin-average SST depths are so much less than the Atlas 14 depths, but the flood damage resulting from the former are greater. I have some comments below to improve the related discussion.
One notable if understandable omission from the works cited is the new study by Baer et al. (2026), which was recently accepted for publication in NPJ Natural Hazards (see preprint citation below; the paper will likely be published shortly, providing a better citation to used). That paper has broadly similar goals, but addresses one of the shortcomings of this manuscript (described below) in that it provides an (almost) fully probabilistic deployment of SST, allowing it to use both SST and design storm methods to estimate average annual loss. The authors will need to cite the Baer et al. study. It can also facilitate description of the main limitation of this study. Another omission is Perez et al. (2024), who used inundation and damage together with SST to examine the relationships between rainfall and flood responses.
In my opinion, the main limitation of this study is that it doesn’t fully explore the capabilities of SST (or other stochastic rainfall methods) to examine the translation of rainfall into flood response and damage in probability terms. To do that, one would generate a large number (1,800+ in the Baer et al. case) of rainfall scenarios of varying magnitudes/AEPs, run them all through the hydrologic/hydraulic/damage estimation chain, estimate resulting damage AEPs, and then compare rainfall and damage AEPs. To be clear: I don’t think that the authors need to do this. But they should acknowledge that it is something they didn’t do, that it would be a useful extension of their work, cite appropriate studies (such as Baer et al. and Perez et al.), and explain why they didn’t go this route. One reasonable justification for not doing it is that the computational burden of thousands of hydraulic simulations is oftentimes too great to be feasible. This was the main factor in both Baer et al. and Perez et al. that limited the number of simulations they could run.
I recommend that the authors carry out one or two more simulations; using the upper (and less importantly, lower) bound of the 90% confidence interval from Atlas 14. This would help contextualize the results even further in terms of conventional design storm uncertainties.
Regarding the useful analysis of the importance of localized rainfall rates: it really seems to me like these findings are driving home the issue of pluvial vs. fluvial flooding, and how the former is an important contributor to flood damage that is omitted from FEMA floodplain assessments. Of course, that opens up the necessity to cite some sources from the growing pluvial flooding literature. In other words, the authors seem to be describing an important and increasingly recognized phenomenon, one that SST and other space-time stochastic rainfall methods can address, that design storms manifestly fail to address, and they should use the right terminology when describing it.
Minor comments:
L9: “Annual” should not be capitalized
L75: we didn’t use Stage IV in Wright et al. (2014). We used a radar rainfall dataset developed ourselves using the HydroNEXRAD algorithm
Generally: details on the SST “configuration,” such as the transposition domain, etc. are missing. I imagine that stuff will be included in the publicly available data and so perhaps it isn’t necessary to describe it in the paper. You should highlight that these details will be available (perhaps you already do in the paper and I missed it).
L164 and L172: Wright et al. (2014) is not the correct citation for RainyDay. It should be Wright et al. (2017).
L168-169: Please cite our ongoing work on objective transposition domain delineation: (preprint) FitzGerald et al. (2025)
Fig 4 caption and/or figure: make clear that “count” refers to number of inundated structures
L351-354: I would recommend leaning heavily into this striking quantitative result. For example, I think you can gain the readers’ attention if you contrast the basin average rainfall differences (238 vs. 310 mm) and flood damage differences quantitatively in the paper’s abstract.
L381: “localized extreme rainfall over highly populated areas” this is a clear example of where using the “pluvial flooding” terminology is appropriate.
L402: I think you mean “spatial”, not “special”Regards,
Daniel Wright
References:
John A. Baer, Antonia Sebastian, Lauren E. Grimley et al. Neglecting Spatiotemporal Rainfall Variability Underestimates Flood Hazard and Risk, 23 January 2026, PREPRINT (Version 1) available at Research Square [https://doi.org/10.21203/rs.3.rs-8593870/v1]
FitzGerald, Benjamin and Wright, Daniel B. and Yan, Lei and Dietrich, Alyssa Hendricks and Sebastian, Antonia, An L-Moments-Based Hypothesis Test to Identify Homogeneous Storm Transposition Regions. Available at SSRN: https://ssrn.com/abstract=5292988 or http://dx.doi.org/10.2139/ssrn.5292988
Perez, Gabriel, Ethan T. Coon, Saubhagya S. Rathore, and Phong V. V. Le. “Advancing Process-Based Flood Frequency Analysis for Assessing Flood Hazard and Population Flood Exposure.” Journal of Hydrology 639 (August 2024): 131620. https://doi.org/10.1016/j.jhydrol.2024.131620.
Wright, D. B., Ricardo Mantilla, and Christa D. Peters-Lidard. “A Remote Sensing-Based Tool for Assessing Rainfall-Driven Hazards.” Environmental Modelling & Software 90 (2017): 34–54. https://doi.org/10.1016/j.envsoft.2016.12.006.Citation: https://doi.org/10.5194/egusphere-2026-1388-RC2 -
RC3: 'Comment on egusphere-2026-1388', Anonymous Referee #3, 20 Apr 2026
reply
Review of egusphere-2026-1388
Effect of Design Storm Characterization on Flood Exposure and Structure Damage Estimates: A Case Study in South Louisiana, USA
By Mohamed ElSaadani, Emad Habib, and Mohamed M. Morsy
This study compares flood exposure and structure damage estimates produced by a conventional spatially uniform NOAA Atlas 14 design storm against an ensemble of 50 stochastic storm transposition (SST) realizations in the Vermilion River Basin, a low-gradient inland-coastal watershed in south-central Louisiana. The topic is relevant and the methodological framework is well-developed. The use of a calibrated coupled HEC-HMS/HEC-RAS model, the HAND-based topographic analysis, and the exposure gap quantification are meaningful contributions. The paper, however, requires moderate revisions before it can be considered for publication. The comments are provided below.
- The abstract leads with methodology rather than the problem motivating the study. A clearer statement of why this comparison matters for flood risk management and engineering practice would strengthen the opening.
- The introduction is long and could be streamlined without losing substance. Several paragraphs covering prior work (lines 40–70) overlap in their message. The specific contribution of this paper relative to Wright et al. (2013, 2014, 2019) and Zhu et al. (2018) should be stated more sharply. It is not immediately clear whether the novelty lies in the hydraulic model fidelity, the damage estimation framework, or the HAND-based diagnostic,or all three.
- Section 3.1: The SST domain selection is described as relying on "familiarity with the study region," and the authors acknowledge that no formal procedure exists for this selection. This is a meaningful source of uncertainty that deserves more explicit discussion. How sensitive are the results to reasonable alternative domain definitions? This should be addressed either in the methods or in the limitations section.
- Section 3.1: The number of parent storms selected for the SST catalog is not stated. The total count should be reported explicitly, along with the years from the 2002–2020 archive that contributed the most events.
- Section 3.2: The acceptability criterion for model calibration, requiring that at least 50% of evaluated gauges meet performance thresholds for each event, is relatively lenient. The rationale for this threshold should be justified, and the actual performance statistics (PPCC, RMSE, PBIAS) for calibration and validation events should be reported, at minimum in a summary table.
- Section 3.2: The antecedent soil moisture condition is set to 36.8% of maximum storage for all 51 simulations. This value appears precise but its derivation is not explained. Is this a calibrated parameter, a median of observed pre-event conditions, or an assumed value? This should be clarified.
- Section 4.1: The discussion of the lognormal distribution parameters is informative, but the claim that the dispersion parameter σ_ln is "largely governed by the basin's topography" is stated as a conclusion without supporting analysis. A brief justification or citation would strengthen this claim.
- Section 4.2: The Atlas-14 scenario produces approximately $0.48B in damages, which the authors describe as comparable to the SST 90th-percentile composite. This alignment is then attributed to the smoothing effect of percentile composites. However, a more direct explanation, that Atlas-14 represents a relatively high areal rainfall depth after TP40 reduction (310 mm) while the average SST basin accumulation is only 238 mm, is introduced later in Section 4.1 and not clearly integrated into the Section 4.2 discussion. The relationship between total rainfall volume and damage ranking across the 50 SST realizations should be examined more explicitly.
- Section 4.3: The median inundation depths for the exposure gap structures (0.08 to 0.13 m) are described as "significant," resulting in "a significant amount of potential losses." This language is imprecise. At those shallow depths, depth-damage functions typically produce modest damage fractions. The authors should clarify what fraction of the $110M is attributable to shallow inundation versus the subset of structures experiencing deeper flooding, or at minimum avoid characterizing these depths as significant without qualification.
- Section 4.5: The HAND analysis is the most diagnostic element of the paper and is well-executed. However, the bandwidth h used in the KDE analysis (Equation 5) is never reported. A single bandwidth was stated to have been applied across all maps, the value should be stated and its selection justified.
- Section 5: The discussion largely restates the results rather than interpreting them in the context of engineering practice. A more focused discussion of what these findings imply for FEMA floodplain mapping, NOAA Atlas 14 applications, or design standards in low-gradient coastal watersheds would strengthen the paper's practical contribution.
- Section 5, line 505: There is a typographic error, "(SST and Atlas-14-based uniform design storms)" is missing an opening parenthesis and reads awkwardly.
- The conclusions are presented as a numbered list. While the findings are clearly stated, points 1 and 2 are closely related and could be combined. The practical recommendation, what practitioners should do with these findings, is absent from the conclusions and should be added.
Citation: https://doi.org/10.5194/egusphere-2026-1388-RC3
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 31 | 12 | 6 | 49 | 7 | 10 |
- HTML: 31
- PDF: 12
- XML: 6
- Total: 49
- BibTeX: 7
- EndNote: 10
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This manuscript compares a conventional Atlas-14-based uniform design storm with an SST-based ensemble and examines how the two rainfall representations affect inundation extent, building exposure, and structural damage estimates in the Vermilion River Basin, Louisiana. The topic is relevant, and the attempt to connect rainfall representation with consequence analysis at the building scale is worthwhile. The results are potentially useful, especially the finding that SST realizations identify additional exposed structures and larger upper-tail losses than the deterministic Atlas-14 case. The paper is not ready for publication in its current form. My main concerns are with the framing of the study, the incomplete treatment of the SST literature, and the lack of methodological clarity in the SST implementation. I also think the flood exposure analysis is not yet sufficiently developed. The manuscript currently reads more like a case-study demonstration than a well-supported analysis of “design storm characterization” in a broader sense.
Major comments are as follows.
The SST literature review is too limited. For a paper built around SST, the review of previous work is not sufficient. The manuscript cites some core SST studies, but the broader literature on storm transposition, rainfall spatial structure, and basin-scale sensitivity is not adequately covered. This matters because the paper frames itself as addressing a gap in how storm representation affects inundation and damage, yet the introduction does not fully establish that context. In its current form, the novelty is not entirely convincing.
The SST methodology is not described in enough detail for the analysis to be fully convincing. Section 3.1 gives only a general description of how SST was applied. Several choices that could affect the results are either missing or treated too briefly. The discussion of domain selection is particularly weak. The manuscript notes that no exact procedure exists and then states that the domain was selected largely based on familiarity with the region and inclusion of major storms. That may be practical, but it remains subjective, and the implications of that subjectivity are not sufficiently addressed. Overall, the SST setup is not transparent enough at present.
The title is broader than the actual analysis. The paper is presented as a study of the “effect of design storm characterization,” but in practice it compares two specific storm representations: one Atlas-14-based uniform storm and one SST-based ensemble. That is a narrower question. The manuscript does not clearly state which storm characteristics are actually under investigation. As a result, the framing promises more than the analysis delivers.
The Atlas-14 vs. SST comparison is not fully controlled. I was not fully convinced by some of the causal interpretation. The two setups differ in several respects at the same time, not only in spatial rainfall variability. This makes it difficult to attribute the reported differences to one factor alone. The manuscript later emphasizes the role of localized rainfall cores, but the comparison itself does not isolate that effect clearly enough. At minimum, this issue should be acknowledged more carefully in the interpretation of results.
The flood exposure analysis is not fully convincing in its current form. This part of the paper needs to be stronger. The “exposure gap” is defined in a one-sided way, focusing on buildings inundated in at least one SST realization but not under Atlas-14. That framing is useful, but it is also selective. Exposure is mostly treated as a binary outcome, and the analysis does not yet provide a sufficiently balanced picture of how exposure differs across the two hazard representations. Since flood exposure is a central part of the paper, this weakens one of the main arguments.
The conclusions are somewhat overstated. The study considers one basin, one storm duration, one return period, one deterministic Atlas-based scenario, and 50 SST realizations. That is enough for an interesting case study, but not enough to support broad claims about low-gradient basins in general. The conclusions should stay closer to what has actually been demonstrated here. There is also still some misalignment between the title, the analysis, and the final claims.
Minor comments