the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 19 Dec 2023
GC Insights: Open R-code to communicate the impact of co-occurring natural hazards
Abstract. Hydro-meteorological hazard is often estimated by university-based scientists using publicly funded climate models, whilst the ensuing risk quantification uses proprietary insurance sector models, which can inhibit the effective translation of risk-related environmental science into modified practice or policy. For co-occurring hazards, this work proposes as an interim solution open R-code that deploys a metric (i.e., correlation coefficient r) obtainable from scientific research, usable in practice without restricted data (climate or risk) being exposed. This tool is evaluated for a worked example that estimates the impact on joint risk at an annual 1-in-200 year level of wet and windy weather in the UK co-occurring rather than being independent, and the approach can be applied to other multi-hazards and compound events in various sectors (e.g. road, rail, telecommunications).
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(824 KB)
-
Supplement
(4614 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(824 KB) - Metadata XML
-
Supplement
(4614 KB) - BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-2799', Anonymous Referee #1, 05 Feb 2024
The paper demonstrates a pragmatic approach to the combination of single-hazard losses. It is laudable to see that the underlying code is open-source and freely available. The paper would have benefited from reviewing existing literature of compounding hazards beyond cited Zscheischler (where I would rather suggest to cite https://doi.org/10.1038/s41558-018-0156-3), e.g. https://zenodo.org/records/7135138, https://doi.org/10.5194/esd-7-659-2016, https://doi.org/10.5194/nhess-22-1487-2022 and https://doi.org/10.1016/j.jenvman.2015.11.011 (most findings also hold rue w/o climate change) and to (if possible) include latest contributions based on event loss tables (e.g. https://eartharxiv.org/repository/view/5286), approaches quite close to the one presented here. In addition to what the paper covers, where r is determined from UKCP ’today’, what about climate change? You could determine the r for different GCMs under select RCPs at time-horizons.... and possibly contribute to physical risk disclosure. Might be worth a few sentences at least as an outlook.
In detail, a few minor points:
line 50: ..analysis with a second pair of commercial risk models… ‘second’ is hard to understand here. It becomes clear once one looks at Figure 1b.line 91: [AON] at the end of the sentence. You state above that the loss table stems from a model run by AON. AON here in brackets cannot be a citation, and with above statement(s) of caution, I do not see any need to reference AON here - what does it exactly stand for? Could it be you only missed to put it in italic font, so it would be a quote as you explain in lines 81-83?
line 96: Co-opetition. UK’s Flood Re is a good example of a public-private-partnership solution that also bears fruits to all market competitors. Might be worth mentioning (especially as you use flood as a demonstrating hazard)
line 104, [AON, Reto Office, Bank]: Again, what does this imply, as it is not in italic, hence does not look like a quote? Is it the entities that endorsed the statement? Did others (e.g. Verisk) not? Would it be an option to clarify this upfront (above, line 81ff, where you make the general statement, add that […] means endorsement of a statement) and the reader would then know how to interpret these listings. And as a minor detail, should it not be Bank of England throughout the text (or you state first time that you will abbreviate).
line 116/117: This statement holds for any tool
Citation: https://doi.org/10.5194/egusphere-2023-2799-RC1 - AC1: 'Reply to RC2', John K. Hillier, 20 Jun 2024
-
RC2: 'Comment on egusphere-2023-2799', Anonymous Referee #2, 29 May 2024
In this manuscript, the authors present an open R code that can transparently translate projected co-occurring natural climate-related hazards (illustrated with wind-flood events) into financial risk quantification for the reinsurance sector. This is an interesting manuscript and it makes an original contribution to the literature by tackling the important theme of open science-policy-practice. Further refining the objectives and methods will significantly enhance the presentation of this work. I hope you find the following comments useful in revising your manuscript for publication.
General comments:
- I find that the title doesn’t fully reflect the content of the paper. It could be enhanced by adding the keyword “(financial) risk”, and by changing “communicate” (which is very broad) to “translating” (as used on L14).
- Who are the intended users of the code you developed? I think that it would be great to clarify this in the manuscript, perhaps already in the introduction. It would appear that the primary users are people from the reinsurance sector, but I am wondering if your code could also be used by researchers, perhaps in collaboration with partners from the reinsurance sector, further fostering collaborations. Out of curiosity, do you have some information you could share on the partners’ perspectives on “anyone” (in theory) being able to use your code to calculate future financial risk?
- Could you please clarify throughout the manuscript what “flood risk” refers to. Do you use “flood proxies” as mentioned in the caption of Fig. 1 or are floods defined using a different method?
- How are the ensembles and error bands exemplified in Fig. 1a and 1c used to generate Fig. 1d? Is the median/mean used only?
- This leads me to wonder what exceedance probability curves might look like for future climate projections where the ensemble might cover a larger spectrum of possibilities. What would be the implications on this method if the uncertainty is so large that using the projected ensemble median/mean does not fully capture the broad spectrum of possibilities for financial risk information? A small reflection on this point in the manuscript would be great.
- To what extent is the methodology applicable to other regions of the world, different climate model outputs, and/or different hazards? Could you please give readers a brief overview of users’ level of involvement/changes they will need to make to the code to run it for a different case?
Specific comments:
- L13: Could “university-based scientists” be changed to “researchers” more generally, to include scientists working at research centres that are not academic institutes?
- L14: Please Specify that you are referring to financial “risk” (vs. other forms of risk).
- L15: If the word count permits, it would be great if you could briefly contextualize the focus on co-occurring hazards.
- L15-16: As this is an “interim solution”, could you please share some information in the manuscript on what the longer-term goal/tool might look like?
- L16: Please specify what the correlation coefficient is calculated on.
- L28-29: “For extreme weather such as flooding or wildfire, […], known as hydro-meteorological natural hazards”. Perhaps a more accurate rephrasing could be: “For hydro-meteorological natural hazards, such as flooding or wildfire, caused by extreme weather”.
- L31-33: How does your work tackle the issue related to large climate datasets that might not contain the necessary variables?
- L44-46: “This dependency exists in meteorological variables such as precipitation (e.g., Martius et al., 2016) and in impact data of losses and delays”. Can you please clarify this sentence? It is not clear what “dependency” you are referring to and what “losses and delays” means.
- L47: Is this the case everywhere in general or do you know of any compound risk models that you could reference here?
- L49-50: Related to the above, do you know if there are any other transparent codes for this type of work (and if so how your approach differs), or if this is the first of its kind? It would be some great context to have for readers less familiar with the literature.
- L50-52: “It also adds analysis with a second pair of commercial risk models to earlier results published in blogs […]”. I feel like this second part needs a bit of introduction.
- L56-57: It would be great if you could please clarify that “1-in-200 year level refer” refers to monetary losses here.
- L61-62: What kind of model are the simulated flood events from? i.e., is it a hydro-meteorological model that takes into account flood generating mechanisms such as rainfall and groundwater, but indeed not wind? I guess what I’m getting at is whether those events already account for wind if they’re based on a coupled ocean-atmosphere-land model.
- L65-66: The link between seasonal correlations and the 1-in-200 year return period is not clear to me. Please clarify in the manuscript.
- L67: Please clarify briefly in the text what these five statistical methods are used for. I can only see four methods mentioned in Fig. 1b. Is there one missing?
- L68-69: Could you clarify a bit further in the manuscript what input data (e.g., dates, magnitude, monetary losses?) are needed from the user?
- L75: I don’t understand how a correlation can be calculated between the number of events and the loss per flood event. Are the flood events ranked and correlated to the losses they incurred?
- L87: How is the Gaussian copula case assessed to be most realistic? It would be great to have a bit more information on this in the manuscript.
- Fig. 1: a) The right “0.4 ‘low’ case” y-label is not aligned with the left 0.4 y-label. d) It is not clear to me how the data is split into different boxplots with low and high correlations. Could you explain this briefly in the text/caption?
Citation: https://doi.org/10.5194/egusphere-2023-2799-RC2 - AC1: 'Reply to RC2', John K. Hillier, 20 Jun 2024
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-2799', Anonymous Referee #1, 05 Feb 2024
The paper demonstrates a pragmatic approach to the combination of single-hazard losses. It is laudable to see that the underlying code is open-source and freely available. The paper would have benefited from reviewing existing literature of compounding hazards beyond cited Zscheischler (where I would rather suggest to cite https://doi.org/10.1038/s41558-018-0156-3), e.g. https://zenodo.org/records/7135138, https://doi.org/10.5194/esd-7-659-2016, https://doi.org/10.5194/nhess-22-1487-2022 and https://doi.org/10.1016/j.jenvman.2015.11.011 (most findings also hold rue w/o climate change) and to (if possible) include latest contributions based on event loss tables (e.g. https://eartharxiv.org/repository/view/5286), approaches quite close to the one presented here. In addition to what the paper covers, where r is determined from UKCP ’today’, what about climate change? You could determine the r for different GCMs under select RCPs at time-horizons.... and possibly contribute to physical risk disclosure. Might be worth a few sentences at least as an outlook.
In detail, a few minor points:
line 50: ..analysis with a second pair of commercial risk models… ‘second’ is hard to understand here. It becomes clear once one looks at Figure 1b.line 91: [AON] at the end of the sentence. You state above that the loss table stems from a model run by AON. AON here in brackets cannot be a citation, and with above statement(s) of caution, I do not see any need to reference AON here - what does it exactly stand for? Could it be you only missed to put it in italic font, so it would be a quote as you explain in lines 81-83?
line 96: Co-opetition. UK’s Flood Re is a good example of a public-private-partnership solution that also bears fruits to all market competitors. Might be worth mentioning (especially as you use flood as a demonstrating hazard)
line 104, [AON, Reto Office, Bank]: Again, what does this imply, as it is not in italic, hence does not look like a quote? Is it the entities that endorsed the statement? Did others (e.g. Verisk) not? Would it be an option to clarify this upfront (above, line 81ff, where you make the general statement, add that […] means endorsement of a statement) and the reader would then know how to interpret these listings. And as a minor detail, should it not be Bank of England throughout the text (or you state first time that you will abbreviate).
line 116/117: This statement holds for any tool
Citation: https://doi.org/10.5194/egusphere-2023-2799-RC1 - AC1: 'Reply to RC2', John K. Hillier, 20 Jun 2024
-
RC2: 'Comment on egusphere-2023-2799', Anonymous Referee #2, 29 May 2024
In this manuscript, the authors present an open R code that can transparently translate projected co-occurring natural climate-related hazards (illustrated with wind-flood events) into financial risk quantification for the reinsurance sector. This is an interesting manuscript and it makes an original contribution to the literature by tackling the important theme of open science-policy-practice. Further refining the objectives and methods will significantly enhance the presentation of this work. I hope you find the following comments useful in revising your manuscript for publication.
General comments:
- I find that the title doesn’t fully reflect the content of the paper. It could be enhanced by adding the keyword “(financial) risk”, and by changing “communicate” (which is very broad) to “translating” (as used on L14).
- Who are the intended users of the code you developed? I think that it would be great to clarify this in the manuscript, perhaps already in the introduction. It would appear that the primary users are people from the reinsurance sector, but I am wondering if your code could also be used by researchers, perhaps in collaboration with partners from the reinsurance sector, further fostering collaborations. Out of curiosity, do you have some information you could share on the partners’ perspectives on “anyone” (in theory) being able to use your code to calculate future financial risk?
- Could you please clarify throughout the manuscript what “flood risk” refers to. Do you use “flood proxies” as mentioned in the caption of Fig. 1 or are floods defined using a different method?
- How are the ensembles and error bands exemplified in Fig. 1a and 1c used to generate Fig. 1d? Is the median/mean used only?
- This leads me to wonder what exceedance probability curves might look like for future climate projections where the ensemble might cover a larger spectrum of possibilities. What would be the implications on this method if the uncertainty is so large that using the projected ensemble median/mean does not fully capture the broad spectrum of possibilities for financial risk information? A small reflection on this point in the manuscript would be great.
- To what extent is the methodology applicable to other regions of the world, different climate model outputs, and/or different hazards? Could you please give readers a brief overview of users’ level of involvement/changes they will need to make to the code to run it for a different case?
Specific comments:
- L13: Could “university-based scientists” be changed to “researchers” more generally, to include scientists working at research centres that are not academic institutes?
- L14: Please Specify that you are referring to financial “risk” (vs. other forms of risk).
- L15: If the word count permits, it would be great if you could briefly contextualize the focus on co-occurring hazards.
- L15-16: As this is an “interim solution”, could you please share some information in the manuscript on what the longer-term goal/tool might look like?
- L16: Please specify what the correlation coefficient is calculated on.
- L28-29: “For extreme weather such as flooding or wildfire, […], known as hydro-meteorological natural hazards”. Perhaps a more accurate rephrasing could be: “For hydro-meteorological natural hazards, such as flooding or wildfire, caused by extreme weather”.
- L31-33: How does your work tackle the issue related to large climate datasets that might not contain the necessary variables?
- L44-46: “This dependency exists in meteorological variables such as precipitation (e.g., Martius et al., 2016) and in impact data of losses and delays”. Can you please clarify this sentence? It is not clear what “dependency” you are referring to and what “losses and delays” means.
- L47: Is this the case everywhere in general or do you know of any compound risk models that you could reference here?
- L49-50: Related to the above, do you know if there are any other transparent codes for this type of work (and if so how your approach differs), or if this is the first of its kind? It would be some great context to have for readers less familiar with the literature.
- L50-52: “It also adds analysis with a second pair of commercial risk models to earlier results published in blogs […]”. I feel like this second part needs a bit of introduction.
- L56-57: It would be great if you could please clarify that “1-in-200 year level refer” refers to monetary losses here.
- L61-62: What kind of model are the simulated flood events from? i.e., is it a hydro-meteorological model that takes into account flood generating mechanisms such as rainfall and groundwater, but indeed not wind? I guess what I’m getting at is whether those events already account for wind if they’re based on a coupled ocean-atmosphere-land model.
- L65-66: The link between seasonal correlations and the 1-in-200 year return period is not clear to me. Please clarify in the manuscript.
- L67: Please clarify briefly in the text what these five statistical methods are used for. I can only see four methods mentioned in Fig. 1b. Is there one missing?
- L68-69: Could you clarify a bit further in the manuscript what input data (e.g., dates, magnitude, monetary losses?) are needed from the user?
- L75: I don’t understand how a correlation can be calculated between the number of events and the loss per flood event. Are the flood events ranked and correlated to the losses they incurred?
- L87: How is the Gaussian copula case assessed to be most realistic? It would be great to have a bit more information on this in the manuscript.
- Fig. 1: a) The right “0.4 ‘low’ case” y-label is not aligned with the left 0.4 y-label. d) It is not clear to me how the data is split into different boxplots with low and high correlations. Could you explain this briefly in the text/caption?
Citation: https://doi.org/10.5194/egusphere-2023-2799-RC2 - AC1: 'Reply to RC2', John K. Hillier, 20 Jun 2024
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 362 | 103 | 44 | 509 | 36 | 35 | 37 |
- HTML: 362
- PDF: 103
- XML: 44
- Total: 509
- Supplement: 36
- BibTeX: 35
- EndNote: 37
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Adrian Champion
Tom Perkins
Freya Garry
Hannah Bloomfield
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(824 KB) - Metadata XML
-
Supplement
(4614 KB) - BibTeX
- EndNote
- Final revised paper