the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 14 Feb 2023
Projections and uncertainties of future winter windstorm damage in Europe
Abstract. Winter windstorms are among the most significant natural hazards in Europe linked to fatalities and substantial economic damages. However, projections of windstorm impact in Europe under climate change are highly uncertain. This study combines climate projections from 30 general circulation models participating in CMIP6 with the climate-risk assessment model CLIMADA to obtain projections of future change in windstorm-induced damages over Europe. We conduct an uncertainty-sensitivity analysis, and find large uncertainties in the projected changes in the damages, with climate model uncertainty being the dominant factor of uncertainty in the projections. We investigate spatial patterns of the future changes in windstorm damages and find an increase in the damages in northwestern and northern-central Europe, and a decrease over the rest of Europe, in agreement with an eastward extension of the North Atlantic storm track into Europe. We combine all 30 available climate models in an ensemble of opportunity approach and find evidence for an intensification of future windstorm damages, with damages with return periods of 100 years under current climate conditions becoming damages with return periods of 28 years under future SSP585 climate scenarios. Our findings demonstrate the importance of climate model uncertainty for the CMIP6 projections of windstorms in Europe, and emphasize the increasing need for risk mitigation due to extreme weather in the future.
- Preprint
(1778 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on egusphere-2023-205', Anonymous Referee #1, 25 Apr 2023
This paper investigates future changes and uncertainties in the damage associated with European extreme wind events using CMIP6 models and the CLIMADA impact modelling framework. This combines the hazard, exposure, and vulnerability factors to create the impact. This is a very nice addition to the literature on windstorm damages.
The result that the scenario is not important for the uncertainty of the projections of damages is quite surprising, and as the authors point out, is not consistent with previous studies. The model variability is the largest uncertainty here. However, the use of multi-model ensembles is good because the multi-model mean will give a better representation of observations than any one realisation (e.g. IPCC 2007). There is of course a lot of uncertainty between the models, and this may be larger than the difference shown by using different SSPs. But, given the multi-model mean, it would be very informative to know the variation between the different scenarios. This is especially true given the larger changes in the storm tracks projected for higher emissions scenario (e.g. Priestley and Catto 2022). It would be good if the authors could give more information about and interpretation of this result.
Other points:
- Line 140: Could the area threshold of 15000km2 be spread over wide regions? Or is there some criterion that says this is a contiguous area? I’m wondering if winds from multiple different storm features could be combined together.
- The exposure data is interesting, but I’m curious to know how this compares to a simple population density.
- Lines 488-490: I wonder if the dynamical downscaling would actually reduce biases. Surely this depends on the large-scale/lower-resolution input? Unless the pattern is correct, but it’s only the intensity that is misrepresented.
Reference:
IPCC (2007) https://archive.ipcc.ch/publications_and_data/ar4/wg1/en/ch10s10-5-4-1.html
Priestley MDK, Catto JL. (2022) Future changes in the extratropical storm tracks and cyclone intensity, wind speed, and structure, Weather and Climate Dynamics, volume 3, no. 1, pages 337-360, DOI:10.5194/wcd-3-337-2022.
Citation: https://doi.org/10.5194/egusphere-2023-205-RC1 -
AC1: 'Reply on RC1', Luca Severino, 11 Jul 2023
We are very grateful for the insightful comments and the positive feedback. We appreciate your suggestion concerning the more detailed analysis of the sensitivity of our projections to the climate scenario and will gladly include it in the revised version of our manuscript. Please find in the attached .pdf our response to your comments and suggestions.
-
RC2: 'Comment on egusphere-2023-205', Anonymous Referee #2, 30 May 2023
I have reviewed the manuscript ‘Projections and uncertainties of future winter windstorm damage in Europe’, by Luca G. Severino et al.. The manuscript combines CMIP6 historical and scenario simulations and the CLIMADA impact model to obtain projections of winter windstorm damage in Europe. The manuscript is well-written, conclusions are relevant and the methodology is potentially reasonable. Some points require substantial clarifications and perhaps modifications. A far clearer discussion on the assumptions made and the framework for the assessment of the uncertainty should be provided.
-It appears that the presented framework accounts only for future hazards and assumes a stationary exposure and vulnerability. In view of the current understanding of future risk (see for instance Cremen et al. 2022), this assumption appears not reasonable. If this is the case then, in my opinion, the authors are essentially quantifying the risk in a warmer climate. Can the authors clarify and modify the manuscript (and the title) accordingly?
-Lines 158-164, We only consider one single impact function for the entire domain and do not differentiate between the different asset types (e.g. residential, industrial). Considering only one single impact function allows us to avoid a complex calibration procedure requiring the tuning of numerous parameters. We use primarily one impact function, which has been designed in Schwierz et al. (2010, hereafter Sw2010), and which is already implemented in CLIMADA. This function has been directly derived from an insurer’s loss model and is based on past claim data in the United Kingdom and cross-validated with other European countries. ‘.
Here the motivation behind the choice is apparently based on avoiding a ‘complex calibration procedure’. This poses questions on the soundness of the methodology. The authors should clarify the assumptions behind their choice and its implications. Given the focus on the uncertainty, this is crucial to properly assess the robustness of the conclusions.-Lines 242, following my previous comment, here the authors state ‘We quantify the uncertainty associated with the vulnerability by using two different impact functions to model the damages: the Sw2010 and the CubEOT impact functions, as this allows us to account for the uncertainty associated with the functional form of the impact function used to model the damages’. This is unclear. Is it correct that you use two functions and in both cases you assume that they apply everywhere in the European domain?’ Please clarify the methodology for modelling of vulnerability and associated uncertainty.
-I am puzzled by statements around intrinsic and epistemic uncertainty in climate projections. This is mostly due to the fact that the study relies on small ensembles that are arguably inadequate to separate the two components of the uncertainty. For instance, caption of Figure 6, ‘The effects of internal variability on the EFCs are simulated by generating multiple EFCs via bootstrapping. ‘Does this include resampling/subsampling over different models? The same comment applies to line 417.
-Can you comment on the interpretation of Figs. 5 and 6. Fig.5 suggests that damage is reduced in many regions while Fig.6 indicates that overall it increases over Europe in a warmer climate. This is perfectly reasonable but more information to interpret the results could be given. A baseline value of the damage in present climate should be shown in the body of the manuscript.
-Following my previous comment, how can one use the ranking provided in Tab. A1 if the signal and its robustness vary substantially from one region to another? Can you provide an example application?
-In section 4, the authors suggest climate model uncertainty to be the dominant factor of uncertainty in the projections. I have two points here. 1) How do you separate intrinsic and epistemic uncertainty in climate projections. 2) If indeed you have assumed a stationary exposure and vulnerability, is this conclusions essentially a consequence of your assumptions? Further comments by the authors around the interpretation of their results would be helpful .
ReferencesCremen, Gemma, Carmine Galasso, and John McCloskey. "Modelling and quantifying tomorrow's risks from natural hazards." Science of The Total Environment 817 (2022): 152552.
Citation: https://doi.org/10.5194/egusphere-2023-205-RC2 - AC2: 'Reply on RC2', Luca Severino, 11 Jul 2023
Status: closed
-
RC1: 'Comment on egusphere-2023-205', Anonymous Referee #1, 25 Apr 2023
This paper investigates future changes and uncertainties in the damage associated with European extreme wind events using CMIP6 models and the CLIMADA impact modelling framework. This combines the hazard, exposure, and vulnerability factors to create the impact. This is a very nice addition to the literature on windstorm damages.
The result that the scenario is not important for the uncertainty of the projections of damages is quite surprising, and as the authors point out, is not consistent with previous studies. The model variability is the largest uncertainty here. However, the use of multi-model ensembles is good because the multi-model mean will give a better representation of observations than any one realisation (e.g. IPCC 2007). There is of course a lot of uncertainty between the models, and this may be larger than the difference shown by using different SSPs. But, given the multi-model mean, it would be very informative to know the variation between the different scenarios. This is especially true given the larger changes in the storm tracks projected for higher emissions scenario (e.g. Priestley and Catto 2022). It would be good if the authors could give more information about and interpretation of this result.
Other points:
- Line 140: Could the area threshold of 15000km2 be spread over wide regions? Or is there some criterion that says this is a contiguous area? I’m wondering if winds from multiple different storm features could be combined together.
- The exposure data is interesting, but I’m curious to know how this compares to a simple population density.
- Lines 488-490: I wonder if the dynamical downscaling would actually reduce biases. Surely this depends on the large-scale/lower-resolution input? Unless the pattern is correct, but it’s only the intensity that is misrepresented.
Reference:
IPCC (2007) https://archive.ipcc.ch/publications_and_data/ar4/wg1/en/ch10s10-5-4-1.html
Priestley MDK, Catto JL. (2022) Future changes in the extratropical storm tracks and cyclone intensity, wind speed, and structure, Weather and Climate Dynamics, volume 3, no. 1, pages 337-360, DOI:10.5194/wcd-3-337-2022.
Citation: https://doi.org/10.5194/egusphere-2023-205-RC1 -
AC1: 'Reply on RC1', Luca Severino, 11 Jul 2023
We are very grateful for the insightful comments and the positive feedback. We appreciate your suggestion concerning the more detailed analysis of the sensitivity of our projections to the climate scenario and will gladly include it in the revised version of our manuscript. Please find in the attached .pdf our response to your comments and suggestions.
-
RC2: 'Comment on egusphere-2023-205', Anonymous Referee #2, 30 May 2023
I have reviewed the manuscript ‘Projections and uncertainties of future winter windstorm damage in Europe’, by Luca G. Severino et al.. The manuscript combines CMIP6 historical and scenario simulations and the CLIMADA impact model to obtain projections of winter windstorm damage in Europe. The manuscript is well-written, conclusions are relevant and the methodology is potentially reasonable. Some points require substantial clarifications and perhaps modifications. A far clearer discussion on the assumptions made and the framework for the assessment of the uncertainty should be provided.
-It appears that the presented framework accounts only for future hazards and assumes a stationary exposure and vulnerability. In view of the current understanding of future risk (see for instance Cremen et al. 2022), this assumption appears not reasonable. If this is the case then, in my opinion, the authors are essentially quantifying the risk in a warmer climate. Can the authors clarify and modify the manuscript (and the title) accordingly?
-Lines 158-164, We only consider one single impact function for the entire domain and do not differentiate between the different asset types (e.g. residential, industrial). Considering only one single impact function allows us to avoid a complex calibration procedure requiring the tuning of numerous parameters. We use primarily one impact function, which has been designed in Schwierz et al. (2010, hereafter Sw2010), and which is already implemented in CLIMADA. This function has been directly derived from an insurer’s loss model and is based on past claim data in the United Kingdom and cross-validated with other European countries. ‘.
Here the motivation behind the choice is apparently based on avoiding a ‘complex calibration procedure’. This poses questions on the soundness of the methodology. The authors should clarify the assumptions behind their choice and its implications. Given the focus on the uncertainty, this is crucial to properly assess the robustness of the conclusions.-Lines 242, following my previous comment, here the authors state ‘We quantify the uncertainty associated with the vulnerability by using two different impact functions to model the damages: the Sw2010 and the CubEOT impact functions, as this allows us to account for the uncertainty associated with the functional form of the impact function used to model the damages’. This is unclear. Is it correct that you use two functions and in both cases you assume that they apply everywhere in the European domain?’ Please clarify the methodology for modelling of vulnerability and associated uncertainty.
-I am puzzled by statements around intrinsic and epistemic uncertainty in climate projections. This is mostly due to the fact that the study relies on small ensembles that are arguably inadequate to separate the two components of the uncertainty. For instance, caption of Figure 6, ‘The effects of internal variability on the EFCs are simulated by generating multiple EFCs via bootstrapping. ‘Does this include resampling/subsampling over different models? The same comment applies to line 417.
-Can you comment on the interpretation of Figs. 5 and 6. Fig.5 suggests that damage is reduced in many regions while Fig.6 indicates that overall it increases over Europe in a warmer climate. This is perfectly reasonable but more information to interpret the results could be given. A baseline value of the damage in present climate should be shown in the body of the manuscript.
-Following my previous comment, how can one use the ranking provided in Tab. A1 if the signal and its robustness vary substantially from one region to another? Can you provide an example application?
-In section 4, the authors suggest climate model uncertainty to be the dominant factor of uncertainty in the projections. I have two points here. 1) How do you separate intrinsic and epistemic uncertainty in climate projections. 2) If indeed you have assumed a stationary exposure and vulnerability, is this conclusions essentially a consequence of your assumptions? Further comments by the authors around the interpretation of their results would be helpful .
ReferencesCremen, Gemma, Carmine Galasso, and John McCloskey. "Modelling and quantifying tomorrow's risks from natural hazards." Science of The Total Environment 817 (2022): 152552.
Citation: https://doi.org/10.5194/egusphere-2023-205-RC2 - AC2: 'Reply on RC2', Luca Severino, 11 Jul 2023
Model code and software
CLIMADA project gabrielaznar; Samuel Eberenz; Thomas Vogt; Emanuel Schmid; Carmen B. Steinmann; Yue Yu; Thomas Röösli; Samuel Lüthi; Inga J. Sauer; Evelyn Mühlhofer; Jan Hartman; Chahan M. Kropf; Benoit P. Guillod; Zélie Stalhandske; Alessio Ciullo; David N. Bresch; Chris Fairless; Pui Man (Mannie) Kam; wjan262; Lukas Riedel; Simona Meiler; Rachel_B; veronicabozzini; DarioStocker https://doi.org/10.5281/zenodo.6807463
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 697 | 434 | 33 | 1,164 | 22 | 24 |
- HTML: 697
- PDF: 434
- XML: 33
- Total: 1,164
- BibTeX: 22
- EndNote: 24
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
