Uncertainty quantification of flood mitigation predictions and implications for interventions

Berends, Koen D.; Straatsma, Menno W.; Warmink, Jord J.; Hulscher, Suzanne J. M. H.

doi:https://doi.org/10.5194/nhess-19-1737-2019

Articles | Volume 19, issue 8

https://doi.org/10.5194/nhess-19-1737-2019

© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

Special issue:

Flood risk assessment and management

https://doi.org/10.5194/nhess-19-1737-2019

© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 19, issue 8

Research article

|

13 Aug 2019

Research article |

| 13 Aug 2019

Uncertainty quantification of flood mitigation predictions and implications for interventions

Koen D. Berends, Menno W. Straatsma, Jord J. Warmink, and Suzanne J. M. H. Hulscher

Download

Final revised paper (published on 13 Aug 2019)
Supplement to the final revised paper
Preprint (discussion started on 14 Nov 2018)
Supplement to the preprint

Interactive discussion

Status: closed

AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment

- Printer-friendly version

- Supplement

RC1: 'An interesting analysis involving approximating uncertainty', Joseph Guillaume, 12 Dec 2018
- AC1: 'Reply to the comments by Joseph Guillaume', Koen Berends, 20 Feb 2019
RC2: 'comment on “Uncertainty quantification of flood mitigation predictions and implications for decision making” by Koen D. Berends et al.', Jan Kwakkel, 19 Dec 2018
- AC2: 'Reply to the comments by Jan Kwakkel', Koen Berends, 20 Feb 2019

Peer-review completion

AR: Author's response | RR: Referee report | ED: Editor decision

ED: Reconsider after major revisions (further review by editor and referees) (09 Mar 2019) by Cristina Prieto

AR by Koen Berends on behalf of the Authors (17 Apr 2019) Author's response Manuscript

ED: Referee Nomination & Report Request started (25 Apr 2019) by Cristina Prieto

RR by Joseph Guillaume (13 May 2019)

Suggestions for revision or reasons for rejection

On the whole, the authors' responses adequately address my concerns, and I support publication of this article.

I would, however, like to comment on one of my criticisms. I had originally suggested that a major contribution was in approximating uncertainty, but that 1) some extra support was needed for the argument that linear interpolation was an appropriate approximation, and 2) the paper would benefit from additional discussion of the uncertainty in the approximation.

Rather than providing these additions, the authors have weakened their argument, emphasising that this linear interpolation is only a first approximation that cannot (yet) be used as a technical tool for obtaining uncertainty estimates:
p16 L11 "By linearly interpolating between the expected values of the low and high intensities, we obtain a first approximation of the uncertainty"
p17 L5 "Given that we only have two points (low and high intensity), a linear transformation is the only model we can support. Although we acknowledge that other models are possible, we included this as a first approximation, because it illustrates information available for decision makers."

I find this to be an adequate solution, as I was particularly impressed during the first review that this uncertainty approximation method "made the results of the analysis more accessible to a broader audience." The authors have played to their strength in emphasising the conceptual rather than technical use of their results.

Nevertheless, there are a couple of points where I think the selected wording could better reflect the authors's decision in this regard:

p1 L7 "establish relationships to aid in decision making"
-> "establish approximate relationships to aid in decision making" or "investigate relationships that may aid decision making"

p3 L11 "Our second objective is to establish a relationship"
-> "establish an approximate relationship" or "explore the relationship"

p21 L25 "The second objective was to establish a relationship between the expected effect and its uncertainty, to aid in the decision making process."
-> "establish an approximate relationship" or "explore the relationship"

p1 L10 we show that intervention effect uncertainty behaves like a traditional backwater curve with a constant relative uncertainty value
-> "behaves approximately like" or "with an approximately constant"

The paper could also benefit from a final proof-read. Here are a couple of typos/errors I noticed:

p2 L16 "2.3 billion" -> euro?
p4 L15 "gradually increasing resolution" -> increasing grid size or decreasing resolution?
p5 L9 "Each interventions" -> each intervention
p6 L21 "as decreases vegetation friction" -> and decreases...
p8 L1 "are designed" -> are designated
p10 L6 "is that a" -> is a
p14 "The novelty of Fig. 6 these results" -> "The novelty of the Fig. 6 results"
p14 "coral estimation" -> CORAL
p20 L2 "intervention the given case study." -> for the given

Hide

RR by Jan Kwakkel (28 May 2019)

Suggestions for revision or reasons for rejection

The authors have made a substantial effort to address my comments and concerns. The modifications and additions largely have addressed my initial concerns. Given the updated manuscript, I still have a few questions and suggestions that might be of use in further developing the paper.

First, you introduce relative uncertainty as a useful metric. I still think that the manuscript could benefit from a broader discussion and preferably a comparison with other metrics. For example, relative uncertainty is quite close to a signal to noise ratio. On theoretical grounds, why would you prefer one over the other, or empirically do they produce different insights for this specific case? Alternatively, perhaps look at (McPhail, Maier et al. 2018) and position relative uncertainty in light of the more conceptual description of robustness metrics provided there.

I struggle with the interpretation given by the authors of figure 8. You write that MINEMBLOW and FLPSMOOTH are most uncertain. I guess this has to do with the bandwidth of outcomes covered by the densities in figure 8. However, based on visual comparison, the difference between MINEMBLOW and GROYNELOW seems to be quite marginal. So how did you arrive at this statement and might it be possible to annotate figure 8 in such a way that it becomes easier for the reader to come to the same conclusion?

Minor comments
• In the abstract, you write “affects predictions of flood mitigation strategies”, I suggest rephrasing this to “affects predictions of the effects of flood mitigation strategies”
• on page 2 you claim that more detailed models are needed given the increasingly complicated nature of intervention designs. I am not sure whether this actually follows.
• On page five, line 10 you touch on the problem of implementation uncertainty. Realizing that it is outside scope for this paper, I am still curious whether you have any thougths on how to bring it into this kind of analysis. For example, you could include an additional source of uncertainty with some variation on how something is implemented. That would provide you with insight into the sensitivity of your designs to this implementation uncertainty.
• The focus of the paper is on the uncertainty in the estimated effects, which during design is quite relevant. However, what also matters is whether the ranking of solutions changes due to this uncertainty.
• P14, line 9, coral should probably be capitalized.

McPhail, C., H. R. Maier, J. H. Kwakkel, E. Giuliani, A. Castelletti and S. Westra (2018). "Robustness metrics: How are they calculated, when should they be used and why do they give different results?" Earth's Future.

Hide

RR by Jonathan Remo (01 Jun 2019)

ED: Publish subject to minor revisions (review by editor) (09 Jun 2019) by Cristina Prieto

AR by Koen Berends on behalf of the Authors (18 Jun 2019) Author's response Manuscript

ED: Publish as is (19 Jun 2019) by Cristina Prieto

AR by Koen Berends on behalf of the Authors (28 Jun 2019) Manuscript

Short summary

River flooding is a major safety concern. Sophisticated models are used to design ways to decrease flood risk, but until recently it was not feasible to calculate how uncertain these model predictions are. Using a new approach, we have now quantified the uncertainty of 12 interventions along the River Waal. Results show significant but not problematically high uncertainty. We demonstrate that the choice between interventions can be different when uncertainty is taken into account.