the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Understanding the Compound Flood Risk along the Coast of the Contiguous United States
Dongyu Feng
Donghui Xu
L. Ruby Leung
Abstract. Compound flooding is a type of flood events caused by multiple flood drivers. The associated risk has usually been assessed using data-based statistical analyses or physics-based numerical models. This study proposes a compound flood (CF) risk assessment (CFRA) framework for coastal regions in the contiguous United States (CONUS). In this framework, a large-scale river model is coupled with a global ocean reanalysis dataset to (a) evaluate the CF exposure risk related to the coastal backwater effects on river basins, and (b) generate spatially distributed data for analyzing the CF hazard risk using a bivariate statistical model of river discharge and storm surge. The two kinds of risk are also combined to achieve a holistic understanding of the continental-scale CF comprehensive risk. The estimated CF risk shows remarkable inter- and intra-basin variabilities along the CONUS coast with more variabilities in the CF hazard risk over the US West and Gulf coastal basins. Different risk assessment methods present significantly different patterns in a few key regions, such as San Francisco Bay area, lower Mississippi River and Puget Sound. Our results highlight the needs to weigh different CF risk measures and avoid using single data-based or physics-based CFRAs. Uncertainty sources in these CFRAs include the use of gauge observations, which cannot account for the flow physics or resolve the spatial variability of risks, and underestimations of the flood extremes and the dependence of CF drivers in large-scale models, highlighting the importance of understanding the CF risks for developing a more robust CFRA.
- Preprint
(4546 KB) - Metadata XML
-
Supplement
(3263 KB) - BibTeX
- EndNote
Dongyu Feng et al.
Status: final response (author comments only)
-
RC1: 'Comment on hess-2023-31', Anonymous Referee #1, 07 Mar 2023
Summary
In this manuscript, the authors apply a compound flood risk assessment (CFRA) to the coastal regions of the contiguous United States. In contrast to several previous studies, they do not only perform a bivariate extremes analysis between storm surge and river streamflow (referred to in their manuscript as data-based CFRA), but they couple their statistical analysis to a large-scale river-routing model and population exposure information (referred to in their manuscript as physics-based CFRA). Their main finding appears to be that their metrics, such as the marginal and joint exceedance probabilities and Kendall’s rank correlation, differ substantially depending on the type of CFRA. The authors use their results to argue that ‘data-based’ CFRA alone does not provide a holistic view of CFRA and provides a biased view, hence different types of CFRA need to be considered. Although my experience in this field is limited, I think that this contribution is welcome and useful, and I also found it well written. I do have several comments, though, that I hope will help to improve the manuscript.
Comments:
-‘Data-based CFRA’: I am not sure if this is a commonly used term, but it is ambiguous to me because ‘physics-based CFRA’ is also based on data. Would it be possible to find terminology that more clearly distinguished the two types of CFRA? Perhaps ‘statistics-based’ and ‘dynamics-based’ or so?
-Figures: for the maps, color schemes are used that I doubt are color-blind friendly/perceptually uniform. Could the authors please check and modify their figures where necessary?
L64-L92: I am missing an introduction of existing large-scale compound flood risk studies with similar aims, such as https://nhess.copernicus.org/articles/23/823/2023/. I would encourage the authors to include a paragraph on such studies and explain the similarities and differences with their own manuscript. The same comment applies to the Discussion section.
L144: Kendall’s rank correlation is computed, but it is not clear to me if/what lag is allowed between the two variables. Could the authors please also discuss how the samples for which the rank correlation is computed are conditioned, i.e., are cases in which one of the variates but not necessarily both variables are extreme (two-sided) also considered or not, and why (not)?
Section 2.1: information about declustering of the peak events seems to be missing, although declustering is necessary to avoid dependence in the extremes. Could the authors please explain if/how their data was declustered?
L318-332: this paragraph discusses figures that are not part of the main manuscript. I would either include both the discussion and those figures in the manuscript or put both in the supplements, and only briefly refer to them. Its current form is inconvenient to read in my opinion.
Section 3.2.1: A comparison with the results of previous (‘data-based’) studies in this region (e.g., Wahl et al., 2015; Nasr et al., 2021) would provide useful context, but is missing. Could this be added?
L400-402: while I appreciate this remark, data that covers a longer period cannot simply be obtained. Could the authors please explain/discuss how they imagine a dataset such as the GTSM reanalysis being extended back in time with reasonable confidence? In the absence of such data, how can we get a sense of how uncertain the results of the present study are given the relatively short period used? This seems especially relevant given the underrepresentation of tropical cyclones in the observations/reanalysis data. And what about the influence of temporal variability on the dependence between the two variates examined? (e.g., Wahl et al., 2015)
L440-445: The open access to the code used for the paper is commendable. I strongly recommend doing the same for the output data instead of sharing it ‘upon request’.
Minor issues:
L55: ‘across rivers and estuaries’: I suggest changing this to ‘different rivers and estuaries’, unless variation within a river or estuary is meant.
L103 could the authors please comment on using a river routing model v.s. using a hydrological model that also includes groundwater? I am not an expert, but I can imagine that changes in groundwater also affect the propagation time between upstream and the coastal interface?
L114: It only becomes clear later why the 2nd boundary condition is used (i.e., in eq. 6). It would be helpful to briefly motivate the 2nd boundary condition here already. Besides, it is not clear where this fixed level is derived from – is it the mean level in GTSM?
L119: ‘statistical model’ please specify what this refers to.
L175: ‘using an R-package’ please specify which package.
L192: is this necessary? Can the authors not just change the scale of the relevant plots?
L200: to avoid confusing this section for a results section, I suggest replacing ‘investigate’ with ‘review’ or ‘discuss’. Additionally, I found this section not as well connected to the rest of the manuscript. Perhaps some sentences could be added as to why these uncertainty sources are introduced here.
L206-207: would be more consistent to define the different types of uncertainty in the same order as they are introduced.
L252: ‘shift’ is this the same as ‘lag’? If so, I think that would be a clearer term.
L258: I suggest to change +/- 5 days to: a window of 10 days around the extreme, if this is indeed the intended meaning.
Section 2.2.4: please consider putting this in a table, for a clearer overview.
L279: ‘elevation gradient’ – would it be worth plotting the gradient instead of the elevation itself in figure?
L286-287: could use some more explanation, in my opinion.
L298: I think this needs to be ‘Section 2.2.4 at 24 river basins’, correct?
Citation: https://doi.org/10.5194/hess-2023-31-RC1 - AC1: 'Reply on RC1', Dongyu Feng, 25 Jul 2023
-
RC2: 'Comment on hess-2023-31', Anonymous Referee #2, 16 Jun 2023
This work has proposed a compound flood (CF) risk assessment (CFRA) framework for coastal regions in the contiguous United States (CONUS). Compound flood is a significant research topic, and this study is timely to contribute to expand the literature. Overall, this work reads smooth and has included great simulations and analysis. Below are some comments on the manuscript.
1. The scientific question was not quite clear. Is the objective to build a framework for assessing compound flood risk or to understand the risk across the U.S. coast? When reading the manuscript, it feels like the former was the main objective.
2. Line 160-165: “We consider 𝑄 and 𝑆𝑆 to be dependent of each other when they display a significant positive correlation (p-value<0.05).” Can we ignore the dependence if the p-value is smaller than 0.05. The authors need to better justify the selection of p-value.
3. Line 170-175: Which copula did the authors choose to quantify the dependence and why?
4. Line 53: change “not available” to “unavailable”
5. Line 87: “has accounted”
6. Line 108-111: Why did the authors use GRFR runoff forcing? ERA5 land and NLDAS also provide runoff.
7. Fig.7’s caption is unclear. “Figure 7: The Kendall's correlation coefficient (𝝉).” What is this 𝝉 and how to define it?
8. Fig.8’s caption is also unclear. Please specify each term and variable for (𝑷𝑸,𝑺𝑺) in the figure.
9. In terms of storm surge events, is it possible to identify storm surge caused by tropical cyclones, extratropical cyclones etc.? How well is GTSM data validated by observations? Given the GTSM products are forced by ERA5 data, it would be useful to validate the GTSM’s storm surge.
10. Figs 2-4: I would suggest that the authors add labels for geographical locations/states. It is hard to know where the locations are in these figures.
11. While it is not reported in this work, the authors could add discussions on the compound flood risk in the future given the climate projection.
12. In Section 3.2 CFRA, it would be better to add more discussions on the results across different coastal regions and the reason why there is spatial discrepancy in the results such as joint exceedance probability. Is the discrepancy in the risk across different regions due to differences in extremes Q or in SS (e.g., storm type, frequency or track), or in both?Citation: https://doi.org/10.5194/hess-2023-31-RC2 - AC2: 'Reply on RC2', Dongyu Feng, 25 Jul 2023
-
RC3: 'Comment on hess-2023-31', Anonymous Referee #3, 16 Jun 2023
SummaryThe paper assesses compound flooding for CONUS based on a statistical analysis of compound flood drivers and an analysis of population exposed to flooding from coastal backwater. It also investigates the skill of the model to represent observed peak timing as well as the correlation between, and (joint) exceedance probabilities of surge and discharge. The paper potentially provides an interesting contribution to current literature. However, I have some major comments that I would like to see addressed before the paper is accepted for publication.Main comments- The paper would really benefit from a comparison, both in the introduction and discussion on results, with other large-scale coupled river-coast models (e.g. Ikeuchi et al., 2017 and Eilander et al., 2020), CONUS flood risk models (e.g. Bates et al., 2021) and statistical CF analysis (e.g. Wahl et al., 2015 and Nasr et al., 2021).- The difference between "data-based CFRA" (line 40) and "physics-based CFRA" (line 60) is not clear but essential to understand the introduction. It would help to start with defining both concepts before discussion pros and cons of both approaches. My first guess was that physics-based would refer to numerical models and data-based to observations, but models are also mentioned under data-based CFRAs (line 45) and observations under the physics-based approach (line 61). Also, the approach of Ikeuchi et al. 2017 and Eilander et al. 2020 are basically similar (analysis of simulated estuarine water levels in a coupled CaMa-Flood and GTSM model) but here mentioned as different approaches.- The terms "hazard risk" and "exposure risk" (line 73) are confusing in my opinion as risk is usually referred to as the combination of hazard, exposure and vulnerability. "hazard risk" seems to only consider the hazard and seems similar to what Couasnon et al. (2020) call "compound flood potential". In short, I think usage of the term "risk" here is confusing and the term "comprehensive risk" (line 98) is even a bit misleading.- The computation of the "risk" metrics is not entirely clear due several ambiguities in the methods section. Specifically for the sampling of events for the calculation of both metrics, see specific comments below.- The Figures with maps and the overlying bars or dots are very difficult to read (Fig 3, 4 & 6).In Fig 6, 7 and 8 the color bars don't have a title or unit and there is no legend for the colors or x-axis label for the subplots .Specific comments- line 28: "A CF event [...] occurs when the associated drivers exceed their respective thresholds (Zscheischler et al., 2020)." Zscheischler et al. (2020) actually argue that not all associated drivers need to exceed their respective thresholds to have large impact CF events. I suggest to rephrase this sentence.- line 78: Do you mean "A robust CFRA should provide a thorough understanding of the uncertainties related to the risk analysis such as uncertainties associated with flood frequency and possible flood damages" ? Would it be possible to make this more specific for *Compound Flood* Risk Analyses?- line 89: Please clarify what is meant by "variability in the fluvial process".- line 103-118 (CFRA model framework): could you provide some more details of the models. I.e. what is the temporal resolution of the MOSART and GTMS model outputs used here? How are the channel widths, depth and lengths defined? What equations are solved in the model (Full Saint-Venant, Local inertial, other)?- line 107: Is the slope of a 15 arcsec DEM (~25km) adequate to estimate the channel river slope?- line 133: Towner et al. (2019) study the skill of several GFMs (not including MOSART or GTSM) for peak flows in the Amazon, how does that relate to the skill of your model framework?- line 139 (step a): The sampled events should be independent. How is this ensured?- line 142 (step b): How are Q events sampled? These are not mentioned under step a.- line 144 (step c): It is not clear how bivariate variables are defined. Is this based on AND or OR sampling of the variables? And do you allow for any time lag between the variables? Please clarify.- line 179-184 (exposure risk): The exposure risk metric accounts not only for surge but also tide as the 'baseline' downstream boundary conditions is based on MSL only (and not MSL+tide). This has several consequences on the analysis in my opinion which are not included nor discussed. For instance, the tidal amplitude is in many locations probably an important predictor of backwater volumes (section 2.2.2) but not accounted for. It could also explain some of the differences between both CF metrics which is not discussed (section 3.2.2). And it should not be referred to as "surge-induced backwater effects" (line 405).- line 193: "the CF hazard index (CFHI) and the CF exposure index (CFEI) are obtained by normalizing Pq,ss and Wp with their corresponding 95th percentile values". How are the 95th percentile values calculated? If I understand correctly, both indicators are a single value per cell right?- line 196: "Our approach transforms the probability of occurrence into a direct measure of human exposure." Could you explain how?- line 239-250 (Impact of riverbed elevation): Is this analysis done per cell or per basin? And where does the riverbed elevation data come from? (see also earlier comment on the CFRA framework)- line 384: "In summary, CFRA should not rely on any single method; more comprehensive thinking is needed considering the different characteristics among the different risk types." How does the CF comprehensive risk metric compare to an actual risk analysis (i.e., combining the hazard and its potential consequences to derive e.g., annual expected losses or people exposed)ReferencesBates, P. D., Quinn, N., Sampson, C., Smith, A., Wing, O., Sosa, J., Savage, J., Olcese, G., Neal, J., Schumann, G., Giustarini, L., Coxon, G., Porter, J. R., Amodeo, M. F., Chu, Z., Lewis-Gruss, S., Freeman, N. B., Houser, T., Delgado, M., Hamidi, A., Bolliger, I., McCusker, K., Emanuel, K., Ferreira, C. M., Khalid, A., Haigh, I. D., Couasnon, A., Kopp, R., Hsiang, S., and Krajewski, W. F.: Combined modeling of US fluvial, pluvial, and coastal flood hazard under current and future climates, Water Resour. Res., 57, e2020WR028673, https://doi.org/10.1029/2020wr028673, 2021.Couasnon, A., Eilander, D., Muis, S., Veldkamp, T. I. E., Haigh, I. D., Wahl, T., Winsemius, H. C., and Ward, P. J.: Measuring compound flood potential from river discharge and storm surge extremes at the global scale, Nat. Hazards Earth Syst. Sci., 20, 489-504, https://doi.org/10.5194/nhess-20-489-2020, 2020.Eilander, D., Couasnon, A., Ikeuchi, H., Muis, S., Yamazaki, D., Winsemius, H. C., and Ward, P. J.: The effect of surge on riverine flood hazard and impact in deltas globally, Environ. Res. Lett., 15, 104007, https://doi.org/10.1088/1748-9326/ab8ca6, 2020.Ikeuchi, H., Hirabayashi, Y., Yamazaki, D., Muis, S., Ward, P. J., Winsemius, H. C., Verlaan, M., and Kanae, S.: Compound simulation of fluvial floods and storm surges in a global coupled river-coast flood model: Model development and its application to 2007 Cyclone Sidr in Bangladesh, J. Adv. Model. Earth Syst., 9, 1847–1862, https://doi.org/10.1002/2017ms000943, 2017.Nasr, A. A., Wahl, T., Rashid, M. M., Camus, P., and Haigh, I. D.: Assessing the dependence structure between oceanographic, fluvial, and pluvial flooding drivers along the United States coastline, Hydrol. Earth Syst. Sci., 25, 6203-6222, https://doi.org/10.5194/hess-25-6203-2021, 2021.Wahl, T., Jain, S., Bender, J., Meyers, S. D., and Luther, M. E.: Increasing risk of compound flooding from storm surge and rainfall for major US cities, Nat. Clim. Chang., 5, 1-6, https://doi.org/10.1038/nclimate2736, 2015.Citation: https://doi.org/
10.5194/hess-2023-31-RC3 - AC3: 'Reply on RC3', Dongyu Feng, 25 Jul 2023
Dongyu Feng et al.
Dongyu Feng et al.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 467 | 167 | 20 | 654 | 45 | 5 | 6 |
- HTML: 467
- PDF: 167
- XML: 20
- Total: 654
- Supplement: 45
- BibTeX: 5
- EndNote: 6
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1