the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Probabilistic Methods for Projecting Average Recurrence Intervals of Coastal Flooding with Sea Level Rise
Timothy Hall
James P. Kossin
Terence Thompson
Seung Hun Baek
Kimberley Drouin
Danielle Montagne
Abstract. We illustrate efficient methods to estimate future projected average return intervals (ARIs) of flood depths in coastal regions from storm-tide data and sea-level rise (SLR) projections. A flood-water path-finding algorithm is applied to digital elevation models (DEMs) in coastal regions to determine possible flood depths at interior points, given storm-tide levels at ARIs and local SLR with uncertainty on the coast. We show that the distribution of projected ARIs of a historical baseline flood-depth is truncated log-normal in the Gumbel extreme-value approximation, and we provide analytic expressions for the means. With this approximation, projected change in flood damage over a range of ARIs can be estimated by analysis at any single ARI. Compared to flood-depth distributions, ARI distributions are less directly related to flood damage, but they have the advantage of relative insensitivity to uncertainties in DEMs and other granular details of flood-risk modeling. We illustrate with applications to Miami, Florida, and coastal North Carolina using two different DEMs.
- Preprint
(8671 KB) - Metadata XML
- BibTeX
- EndNote
Timothy Hall et al.
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2023-425', Anonymous Referee #1, 13 May 2023
I confess to being puzzled by this paper. It presents a computational approach for hydrologically connected bathtub modeling of the flooding associated with changing storm tides under sea-level rise, under the (questionable) assumption that storm tides are Gumbel-distributed.
The outcome of this approach does not seem substantially different than that employed in widely available products, for example Climate Central's Coastal Risk Screening Tool (https://coastal.climatecentral.org/) or (albeit without the probabilistic sea-level rise projections, and limited to US in scope) NOAA's Sea Level Rise Viewer (https://coast.noaa.gov/slr/). Compare Figure 13, 2090s 100-yr ARI to https://coastal.climatecentral.org/mapview/14/-77.4225/34.5065/cbfbfdb22ebc86e961e058b6dd870721a7e507c2ceb73fbd672fc0354e8a1d5f. It is unclear what the innovation here is. The particular applications shown to Miami and coastal North Carolina provide no generalizable insights, nor indeed are they contextualized in a manner to provide local insights.
This paper does not present a computational model for doing the analyses described (there is no Code Availability section). An open-source model doing the analyses doing the analyses described might well be useful, but a paper describing a computational approach for what appears to be a fairly routine type of analysis seems less useful. Without code, there is indeed no way to adequately review the computational approach.
Nor does the paper present the output of such a model; the Data Availability section describes only the publicly available data sets used as inputs. The data used are, in fact, somewhat dated: Kopp et al. (2014) has been supplanted by the IPCC AR6 sea-level rise projection data set (see https://github.com/Rutgers-ESSP/IPCC-AR6-Sea-Level-Projections and https://sealevel.nasa.gov/ipcc-ar6-sea-level-projection-tool). Muis et al. (2016) was supplanted by the Coastal Dataset for the Evaluation of Climate Impact (CoDEC) dataset (Muis et al., 2020, doi:10.3389/fmars.2020.00263). The MERIT DEM does not remove buildings, but other global DEMs (e.g., COPDEM30, Hawker et al. 2022, 10.1088/1748-9326/ac4d4f; CoastalDEM, Kulp & Staruss 2021, https://www.climatecentral.org/uploads/media/CoastalDEM_2.1_Scientific_Report_.pdf) do; no rationale for using MERIT is presented.
I note that the failure to provide either code or data appears to be inconsistent with NHESS's data policy (https://www.natural-hazards-and-earth-system-sciences.net/policies/data_policy.html).
Given the affiliation of the authors, with (as they peculiarly state in a self-advertisement in the text of the paper itself) "The Climate Service (TCS), an S&P Global company providing decadal climate risk projections globally for a range of climate and weather hazards", I am concerned that this paper represents an attempt at 'peer-review washing' of an internal, private sector model, in an attempt to give it the sheen of up-to-date, peer-reviewed science. I would be happy to be proved wrong, but I found nothing compelling in this paper that would merit anything other than a corporate white paper.
This paper does not appear to offer to the scientific community either novel scientific insights or new tools or datasets of broad use. Rather, I am concerned that The Climate Science (TCS), an S&P Global company, is trying to exploit the labor of volunteer journal peer-reviewers so that it can claim to its customers that it uses recently published, peer-reviewed methods, without the expense of hiring expert reviewers to critique their methodologies. (S&P Global has annual review of $12 billion; they should pay for the labor they use.)
I am also concerned that The Climate Science (TCS), an S&P Global company, will claim to their customers that the output of this approach represents "latest generation geospatial data [and] ownership mapping" (to use the language on their website), when the methods presented offer only modest advances on the 2008 state-of-the-art, and many of the underlying data sets are themselves not the most recent versions available.
Other points:
- The authors assume extreme sea levels are Gumbel distribution (and mistakingly at times refer to a 'Gumbel generalized extreme value distribution'; a Gumbel distribution is explicitly is not generalized). This is well known not to be the case in tropical cyclone-influenced regions (e.g., Buchanan et al., 2016, 10.1007/s10584-016-1664-7).
- The authors say that "bathtub" models consider only a site's elevation and not hydrological connectivity; if so, very few people still use "bathtub" models, as what I would consider modern "bathtub" models have long considered hydrological connectivity (e.g., Poulter and Halpin 2008, 10.1080/13658810701371858, as an initial paper introducing this, and Buchanan et al. 2020, 10.1088/1748-9326/abb266, as a recent example of an economic risk analysis applying such a model).
Citation: https://doi.org/10.5194/egusphere-2023-425-RC1 -
AC1: 'Reply on RC1', Timothy Hall, 16 May 2023
We thank the reviewer for his/her comments on our manuscript. We agree with the reviewer that the flood model itself, while useful in its global coverage and probabilistic formulation, does not break new ground, and that the driving data sets from 2014 (sea level rise) and 2016 (storm tide) have more recent versions.
However, we were disappointed that the reviewer didn’t comment on the probabilistic methodology and analysis, which we view as the primary novel part of our work. Perhaps our lengthy model description obscured the main points of the manuscript, concerning this probabilistic analysis. Pending comments from other reviewers, we see merit in significantly revising and paring down the manuscript to focus more clearly on the probabilistic analysis, comparing the pros and cons of flood depth versus flood frequency as hazard metrics in a probabilistic context. These points are largely independent of the particular flood model employed. In this context, our particular model is just a convenient tool to illustrate these points with examples, Florida and North Carolina. We will certainly make the model data from these examples available, along with the analysis code.
On another topic, we vehemently disagree with the reviewer’s insinuation that we are somehow unfairly trying to “exploit the labor of volunteer peer-reviewers” for free corporate advertising. The peer-review system functions to evaluate the merit of scientific work regardless of the researchers’ affiliation. The annual revenue of our parent organization has no bearing whatsoever on participation in peer-review, any more than the annual tax revenues of governments that fund academic work.
In addition, please be aware that we have peer-reviewed numerous papers ourselves in the past few years. In other words, we contribute to the system on both sides.
Citation: https://doi.org/10.5194/egusphere-2023-425-AC1 -
RC2: 'Reply on AC1', Anonymous Referee #1, 16 May 2023
I would be delighted to review a new submission that made the novelty of the contribution clear. This contribution is deeply buried, and to the extent I can make out hints of it, it does not seem substantially different than other reduced-form, probabilistic, (hydrologically connected) bathtub-model based approaches to coastal flood risk assessment (consider, for example, Rasmussen et al., 2020, 10.1029/2019EF001340, which is not cited).
All scientists, regardless of affiliation, are, of course, welcome to contribute to the peer-reviewed literature. (And I am well aware of the distinguished records of some of the authors, which is why I was initially excited to read the paper and subsequently disappointed upon reading it.) However, as previously noted, the paper as currently read seems more like a corporate white paper than a novel contribution to the literature -- if it does indeed advance the frontier, it needs to be much clearer about where it does.
And, of course, all scientists, regardless of affiliation, should be expected to comply with norms of open science, including NHESS's data policy, which states:
"The output of research is not only journal articles but also data sets, model code, samples, etc. Only the entire network of interconnected information can guarantee integrity, transparency, reuse, and reproducibility of scientific findings. Moreover, all of these resources provide great additional value in their own right. Hence, it is particularly important that data and other information underpinning the research findings are "findable, accessible, interoperable, and reusable" (FAIR) not only for humans but also for machines."
Please include a statement of Code Availability when you revise the manuscript.
Citation: https://doi.org/10.5194/egusphere-2023-425-RC2
-
RC2: 'Reply on AC1', Anonymous Referee #1, 16 May 2023
-
RC3: 'Comment on egusphere-2023-425', Anonymous Referee #2, 20 Aug 2023
The manuscript introduced me to the concept of average return interval (ARI), defined as the average time between events exceeding a specified flood depth. The paper explored changes in the ARI due to sea level rise. The authors show that fractional change in mean loss is a single factor regardless of ARI magnitude and develop an approach to find this factor. The approach boils down to finding the distribution of the future ARI that corresponds to some reference flood depth (100-year in this case). The distribution of the future ARI can be found relatively easily thanks to the normality of the sea level rise projections and assumptions relating to the flood depth – ARI relationship and distribution of the annual maximum. The authors then apply a simple inundation model to find the projected changes in the ARI for two case study sites.
I enjoyed following the mathematics which are rather elegant and in the mainly very well explained, however certain assumptions within the methodology are concerning.One is the assumption that the Food depth–average return interval (ARI) relationship is log-linear with respect to the ARI is particularly concerning. The assumption is based on studies at two locations. At one location the relationship is stated to break downs for ARIs greater than 100 years. At the other location e.g., New York in the study by Lin et al. (2016) only two of the five projected scenarios appear to give an approximately log-linear depth-ARI relationship. Furthermore, Lin et al. (2016) account for changes in storm climatology as well as sea level rise which is not mentioned in the text.
Another questionable assumption is that the annual maximum water levels are distributed according to a GEV with a shape parameter of zero. In fairness to the authors the limitations of this assumption are discussed. I agree with the authors that GEV shape parameters are difficult to determine for short records but would the extra effort to attempt to estimate them more robustly really add significant complexity to the method? Fixing the GEV parameter in the current way really means it is debatable whether the study is deserving of the term ‘probabilistic method’. I understand that this assumption is an important component of simplifying the parameter estimation for the distribution of the projected ARI.
In my view, the compounding effect of these assumptions combined with the simple flood model somewhat brings into question the credibility of the results. To me the proposed framework is reminiscent of coarse global scale assessments which rely on similar assumptions and are plagued with uncertainties stemming from all parts of the modeling chain (Vousdoukas et al. 2018). Nevertheless, these large-scale analyses are useful to rating agencies, insurers, and the World Bank that compare the relative flood risk in different regions/cities. In this paper, however, the approach is implicitly recommended for a local-scale assessment where more granular probabilistic–numerical approaches with fewer assumptions are regularly undertaken. If I need a relatively high-resolution DEM to apply the approach why not apply a traditional statistical approach combined with a hydraulic model? If the proposed framework could be applied at a continental or global scale it could be useful as part of a suite of methods for assessing the relative flood risk in different over larger spatial scales.
In summary, there are merits to the methodology, however, in its current state I am concerned about its utility i.e., it is too data intensive for a large-scale analysis yet there are too many simplification/assumptions for it to be an approach that is applied at a local scale. At the very least, I strongly encourage the authors to consider the users such a framework could be aimed at and if necessary, tweak the framework and revise the manuscript accordingly. I am also concerned about the simplifying assumptions and whether it is reliable for any type of application.
Specific comments
There are more recent sea level rise projections for the U.S., see Sweet et al. (2022). See also the Coastal Dataset for the Evaluation of Climate Impact (CoDEC) dataset (Muis et al., 2020).
According to the caption of Figure 9, ‘x’ should appear below ‘hEL’ rather than ‘hMET'.
References
Muis, S., Apecechea, M. I., Dullaart, J., de Lima Rego, J., Madsen, K S., Su, J., Yan, K., and Verlaan, M. (2020). A High-Resolution Global Dataset of Extreme Sea Levels, Tides, and Storm Surges, Including Future Projections, Frontiers in Marine Science, 7, https://doi.org/10.3389/fmars.2020.00263.
Vousdoukas, M. I., Bouziotas, D., Giardino, A., Bouwer, L. M., Mentaschi, L., Voukouvalas, E., and Feyen, L. (2018). Understanding epistemic uncertainty in large-scale coastal flood risk assessment for present and future climates, Nat. Hazards Earth Syst. Sci., 18, 2127–2142, https://doi.org/10.5194/nhess-18-2127-2018.
Sweet, W.V., B.D. Hamlington, R.E. Kopp, C.P. Weaver, P.L. Barnard, D. Bekaert, W. Brooks, M. Craghan, G. Dusek, T. Frederikse, G. Garner, A.S. Genz, J.P. Krasting, E. Larour, D. Marcy, J.J. Marra, J. Obeysekera, M. Osler, M. Pendleton, D. Roman, L. Schmied, W. Veatch, K.D. White, and C. Zuzak (2022). Global and Regional Sea Level Rise Scenarios for the United States: Updated Mean Projections and Extreme Water Level Probabilities Along U.S. Coastlines. NOAA Technical Report NOS 01. National Oceanic and Atmospheric Administration, National Ocean Service, Silver Spring, MD, 111 pp. https://oceanservice.noaa.gov/hazards/sealevelrise/noaa-nostechrpt01-global-regional-SLR-scenarios-US.pdf
Citation: https://doi.org/10.5194/egusphere-2023-425-RC3
Timothy Hall et al.
Timothy Hall et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 228 | 90 | 15 | 333 | 4 | 4 |
- HTML: 228
- PDF: 90
- XML: 15
- Total: 333
- BibTeX: 4
- EndNote: 4
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1