the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Insurance loss model vs meteorological loss index – How comparable are their loss estimates for European windstorms?
Abstract. Windstorms affecting Europe are among the natural hazards with the largest socio-economic impacts. Therefore, many sectors like society, economy or the insurance industry are highly interested in reliable information on associated impacts and losses. There are different metrics to quantify windstorm-related losses, ranging from simple natural hazard databases over loss indices based on meteorological variables to more complex insurance loss (catastrophe) models. In this study, we compare estimated windstorm losses using the meteorological Loss Index (LI) with losses obtained from the European Windstorm Model of Aon Impact Forecasting. To test the sensitivity of LI to different meteorological input data, we furthermore contrast LI based on the reanalysis dataset ERA5 and its predecessor ERA-Interim. We focus on similarities and differences between the datasets in terms of loss values and storm rank for specific storm events in the common reanalysis period across 11 European countries.
Our results reveal higher LI values for ERA5 than for ERA-Interim for all of Europe, coming mostly from a higher spatial resolution in ERA5. The storm ranking is comparable for Western and Central European countries for both reanalyses. Compared to Aon’s Impact Forecasting model, LI ERA5 shows comparable storm ranks. However, LI seems to have difficulties in distinguishing between extreme windstorms with high losses and those with only moderate losses. The loss distribution in LI is thus not steep enough and the tail is probably on the short side, yet it is an effective index, precisely because of its simplicity, suitable for estimating the impacts and ranking storm events.
- Preprint
(2107 KB) - Metadata XML
-
Supplement
(685 KB) - BibTeX
- EndNote
Status: open (until 28 Apr 2024)
-
RC1: 'Comment on nhess-2024-16', Gerard van der Schrier, 29 Mar 2024
reply
Dear authors,
When the request of the editor came in, I asked my colleague from the department to join me in this review. While my perspective is from the observational/climatology side, hers is from the data science side with a strong focus on the link between weather extremes and impact. We are both iinterested in the subject of your study and we were both enthusiastic to provide a review. However, two reviews may be a bit more than what you bargained for. Nevertheless, I hope the reviews are of some help.
Kind regards,
Gerard van der Schrier & Irene Garcia Marti
-
CC1: 'Comment on nhess-2024-16', Mathias Raschke, 10 Apr 2024
reply
Dear authors,
I am very interested in any kind of quantification of aggregated natural catastrophes (natcat) losses from such as you present, as I am employed in (re)insurance industry for natcat modelling and research and publish as freelancer in this field. Unfortunately, I realize many gaps and shortfalls in and questions regarding your draft about event losses from windstorms (in parts of Europe):
- Why is the renaming (storm severity index [SSI] to meteorological loss index [LI]) not mentioned in the abstract?
- As far as I know, the original purpose of the SSI was not to estimate loss, but to formulate a size measure (for spatial extent (RMS 2024)). The SSI is also used to quantify spatial correlation (Bonazzi et al. 2012).What means a “metrics to quantify windstorm-related losses”? An event loss can be recorded or estimated. Why so you change this focus?
- The natcat models in (re)insurance industry (such as the applied Aon’s IF Euro WS model) are less explained in the draft in contrast to your reference Mitchell-Wallace et al. (2017). Alternative models and vendors are not even mentioned. The Natcat model with thousands of stochastic events estimates event losses for a defined exposure and (high) return period.
- What sense does it make to compare two model results on event losess when data for (insured) event losses and corresponding exposure (provided by the Perils AG, mentioned in your draft) are available? There is also information about market penetration (proportion of insured exposure to insurable exposure).
- Why is my model (Raschke 2022) not even mentioned? The agreement between estimated and observed event damage (windstorm Germany) is significantly better in my results (plot) than in yours.
- The loss/damage function applied should be discussed. The power parameter of 3 might be unrealistic. The wind speed does not cause damage, but rather it creates a damage generating wind pressure/load (at the buildings) that is proportional to the squared wind speed (details see Raschke 2022).
- Correlation measures has been formulated for random variables. The maximum event loss per year or the annual sum of such losses are random variables (drawn once per year). However, event losses don’t be random variables but point events of a stochastic process. Therefore, you can't just apply a correlation measure to it.
Besides, the results are not perfectly presented (e.g., Figure 8, colour scale for correlation measure form blue [-0.9] to red [0.9] although only positive correlations are mapped).
With kind regards,
Mathias Raschke
References
Bonazzi, A. et al. The spatial structure of European wind storms as characterized by bivariate extreme-value Copulas. Nat. Hazards Earth Syst. Sci. 12, 1769–1782 (2012).
Natural Catastrophes – Risk Management and Modelling – A Practioner’s Guide (eds Mitchell-Wallace, K. et al.) (Wiley Blackwell, 2017).
Raschke, M. About the return period of a catastrophe. Nat. Hazards Earth Syst. Sci. 22, 245–263 (2022).
RMS Verisk. Storm Severity Index (SSI), Help Centre Risk Link Storm Severity Index (SSI), (2023).
Citation: https://doi.org/10.5194/nhess-2024-16-CC1 -
RC2: 'Comment on nhess-2024-16', Anonymous Referee #2, 16 Apr 2024
reply
Dear editor,
thank you for giving me a chance to review this manuscript. The paper is well written and has the potential to shed light on the differences between a rather simple, but well documented open access approach and a more refined proprietary commercial product. This is very welcome contribution that could inform the community on the differences between approaches pursued by the academic community and private sector.I do however have several critical comments on the quality of the analysis and on some of the conclusions drawn from the analysis.
I have attached my comments.
-
RC3: 'Comment on nhess-2024-16', Anonymous Referee #3, 26 Apr 2024
reply
Dear editor, authors,
Thanks for giving me the chance to read and review this paper on loss estimates for European windstorms. The paper is well written and can be a valuable contribution to the better understanding of the impact of European windstorms and methods to asses these.
I do have a number of comments, though, which hopefully will help the authors to improve the manuscript - which are attached.
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
327 | 84 | 22 | 433 | 27 | 14 | 16 |
- HTML: 327
- PDF: 84
- XML: 22
- Total: 433
- Supplement: 27
- BibTeX: 14
- EndNote: 16
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1