Climate Change Hazard Impact
The 2010 CoreLogic Climate Change Impact Study
CoreLogic assessed the potential impact of Climate Change on European Windstorm risk in 2010, with the publication of a white paper Activity Of Catastrophic Windstorm Events In Europe In The 21st Century.
Using climate scenario data presented in the IPCC SRES report, CoreLogic collaborated with experts at the Freie Universität Berlin to investigate the effects of climate change on extra-tropical windstorms in Europe as simulated by Atmosphere Ocean General Circulation Models (AOGCMs).
In this study, several climate scenarios for the year 2100 set by the IPCC were modelled to evaluate the sensitivity of several characteristics of windstorms to increasing CO2 concentrations, with an emphasis on the countries covered by the Eurowind™ model.
The impacts of two different SRES scenarios, namely A1B and A2, were investigated
- The A1B scenario assumes an increase of CO2 concentrations up to the middle of the 21st century and a gradual levelling off afterward
- The A2 scenario uses a continuously increasing level of concentrations throughout the entire 21st century (IPCC, 2007).
Comparisons were made with respect to a control run, i.e. a historical simulation, spanning the years 1901-2000 and forced using measured CO2 levels
A storm tracking algorithm developed by Leckebusch et al. (2008), using the 98th percentile value of the local wind speed as a threshold was applied to the 10-meter wind speed field of the AOGCM model simulations to create a catalogue of storms for each climate scenario. The resulting storm catalogues were then analysed leading to the key findings discussed in the following sections.
CoreLogic Study: Storm Frequency Changes
Both scenarios A1B and A2 lead to an overall decrease in storm frequency across the European domain. However, the picture is more complicated in terms of changes as a function of storm intensity and location.
The storm severity index (SSI) is a measure of a storm’s damage potential. For the purposes of this study, stochastic storms were partitioned into three categories of SSI: low, medium, and high, using the Eurowind™ historical catalogue to define the portioning bins.
- Low storms span the SSI range corresponding to historical storms that have caused economic losses smaller than several hundred million euros.
- Medium storms span the SSI range of so-called historical ‘named storms’ which have typically caused economic damage of at least several hundred million euros.
- High SSI storms are those which obtain at least the meteorological severity of storm Daria (1990).
Figure 4 Possible Climate Change Impacts on European Windstorm Frequencies Using Eurowind and IPCC-SRES Scenarios. Source: CoreLogic
The ratio of frequencies, or rates, for each of the SSI groups is shown in Figure 4. Low SSI storms decrease by ~8% under both A1B and A2. By contrast, Medium and High SSI storms increase under both scenarios. In particular, the rates of storms as strong as Daria increase by over three times under both scenarios. These very large storms are of significant concern to insurers.
Northern Hemisphere Frequency Changes
In this section we provide a review of further studies of the impact of climate change on storm frequencies across the entire Northern Hemisphere, i.e. not confined to the European domain.
Northern Hemisphere Frequency Changes (Independent of Storm Intensity)
According to both AR5 and AR6 reports, the number of ETCs composing the storm tracks in the Northern Hemisphere (NH) is projected to weakly decline in future projections, but by no more than a few percent. This is supported by studies based on earlier generations of GCMs (e.g., König et al., 1993; Bengtsson et al., 2009 Michaelis et al., 2017; Zappa et al., 2013) as well as the CMIP6 results published after AR6 of Priestly & Catto 2022 (reproduced in Figure 5), where the frequency decrease becomes larger with increasing end-of-century climate forcing.
Figure 5 CMIP Rates of Northern Hemisphere Extra-tropical Cyclones during 1979–2014 (historical) and 2080–2100 (projected) for the different SSPs. Adapted version of Figure 1 of Priestly & Catto 2022
The reduction in ETC rates can be understood as arising from the reduced baroclinity in the lower troposphere which is a consequence of Arctic amplification during global warming (see, for example, Catto et al. 2019). In the amplification process, surface albedo and temperature feedbacks lead to increased warming of the Arctic relative to the global mean. This, in turn, reduces the pole-to-equator temperature gradient, and it is this gradient which is the catalyst of ETCs.
Figure 6 CMIP Rates of intense Northern Hemisphere Extra-tropical Cyclones during the historical period 1979–2014 and 2080–2100 (projected) for the different SSPs. Intense cyclones are defined as those that exceed the historical 90th percentile of peak cyclone vorticity. Adapted version of Figure 1 of Priestly & Catto 2022.
Northern Hemisphere Frequency changes of intense windstorms
From a catastrophe risk management perspective, understanding the impact of climate change on intense windstorms is of particular importance due to their potential financial impact on the sector. The AR6 report states that there is only a medium level of confidence in the projected impact on such storms, which, nevertheless, are predicted to strongly decrease in rate (as evidenced in Harvey et al., 2014; Seiler and Zwiers, 2016; J. Wang et al., 2017a). Other newer research, however, such as Priestly & Catto 2022, whose study was performed after the publication of the AR6 report, predict an increase in the rates of intense NH ETCs, albeit with a very large associated uncertainty (Figure 6).
A large contributory factor to the uncertainty associated with intense ETC rates resides in the formulation of the numerical climate models themselves. Computational expense means that decades-long simulations require the horizontal grid-spacing of climate models to be relatively large (i.e., 100km or greater). Such coarse resolutions are unable to adequately model intense events. Low resolution limits the accuracy of the physical processes depicted and introduces biases, so that the intensity of storms is underestimated. Explosive cyclones, for example, tend to be too weak in climate models (Seiler and Zwiers, 2016; Priestley et al., 2020).
As air parcels rise within a storm, water vapour condenses into liquid form, releasing latent heat. This heat, in turn, increases that rate at which air rises, leading to increased surface wind speeds. Low-resolution models fail to properly depict this intensification process (see, for e.g., Michaelis et al., 2017).
The results of the studies considered in this section broadly agree with those of the CoreLogic study, with the added benefit of providing the latest estimate of uncertainty in frequency changes. In the next section, we focus in on how storm frequencies are likely to change by location, both according to the CoreLogic study and subsequent research.
Storm Frequencies by Location
In the earlier Corelogic study, to determine the impact of climate change on the frequencies of storms as a function of storm track position, three latitude bands were defined: the southern region (35°N-48°N), central region (48°N-61°N), and northern region (61°N-74°N).
Each simulated storm was then assigned to one of these regions according to where its track crossed the Greenwich meridian. Hence, for example, storms with zonal tracks crossing the southern parts of the United Kingdom and the BENELUX countries would fall into the central latitude band.
The ratio of storm frequencies by region group is shown Figure 7 for both climate change scenarios.
Figure 7 Possible Climate Change Impacts on European Windstorm Frequencies by latitudinal bands using Eurowind and IPCC-SRES Scenarios. Source: CoreLogic
Compared to the control run, all SRES scenarios display a shift in storm tracks from the “southern latitude” band to the “central latitude” band of Europe. The shift is most pronounced in the A1B_1 run. This shift is also partially observable between the “northern” and “central” bands.
The shift from northern to “central latitude” band is again most pronounced in the
A1B_1 scenario. On the other hand, the A2 scenario displays the largest shift, with a decrease of ~30% in number of storm tracks from the “southern latitude”.
AR6 (see Chapter 4 of Lee et al. 2021) concludes that under scenario SSP5-8.5, a ‘tripolar’ signal of change in ETC track density will occur across Europe in the years spanning 2080-2100, relative to the historical period of 1979-2014. This pattern can be seen in Figure 8, where the tripolar signal is formed by increases over Northwest Europe (the areas coloured pink, including the UK and North Sea), and decreases (shown in blue) over Southwest Europe and Northern Scandinavia.1
The changes pertain to the mean of 13 CMIP6 models. The hatched areas indicate where less than 80% of the CMIP6 models used shown the same sign of track density change.
1 Most likely due to changes in North-Atlantic ocean currents caused by the melting of sea ice and ice sheets (see Bengtsson et al., 2009; Gervais et al., 2018, 2019; Oudar et al., 2020).
This is broadly consistent with the CoreLogic results shown in Figure 7 under SRES scenario A1B where the rates increase in the central region while decrease in the northern and southern regions.
According to Priestley & Catto 2020, the tri-polar pattern strengthens as the severity of emission scenario increases (Figure 9). The figure shows mean track densities calculated by averaging over all CMIP6 models used in the study, with the historical period spanning 1979-2014 shown in the top left plot (a). The remaining plots show the changes in density with respect to this historical period, with increases shown in red and decreases in blue. The future projections span 2040-2100 using the following SSPs: SSP1-26 (b), SSP2-45 (c), SSP3-70 (d), and SSP5-85 (e).
Figure 8 Changes in extratropical storm track density. December-February (DJF) Northern Hemisphere, 2080–2100 for SSP5‑8.5 relative to 1979–2014 based on 13 CMIP6 models.Source: adapted from Figure 4.27 of Lee et al. 2021.
Figure 9 Storm track density changes under increasing emissions for the SSP1-26, SSP2-45, SSP3-70, and SSP5-85 scenarios for the years 2040–2100 Source: Figure 2 of Priestly & Catto 2022
Due to the speed and cyclonic motion of an ETC windstorm, the distribution of winds within the storm (i.e., the so-called wind field) tends to be asymmetric with the highest surface wind speeds typically 80-240 km (i.e., 50-150 miles) south of the centre of the storm. Hence, the tripolar pattern exhibited by changes in track density is associated with a spatial change in the maximum wind illustrated in Figure 10.
Accordingly, climate change increases the mean wind hazard in a key region of insured exposure, i.e., Western Europe, as well as in central and Eastern Europe as the storm tracks have penetrated further east (Figure 8). The mean hazard decreases in Scandinavia and the Iberian Peninsula.
In this section we have seen how location-dependent hazard changes due to climate change are primarily driven by changes in storm-track density. In the next, and final, section of hazard impact, we move attention to storm size and intensity.
Figure 10 Multi-model mean 850-hPa wind change (L) and signal-to-noise (R) Source: Figure 2 of Zappa et al. 2013.
Changes in Storm Size & Intensity
CoreLogic investigated the impact of climate change on the maximum wind speed of storms, by defining three different equally-spaced wind speed ranges (“low-range,” “medium-range,” and “high-range”), covering the entire range of modelled maximum storm wind speeds. The group of “high-range” wind speeds corresponds approximately to the upper 10% of all modelled wind speeds.
In the control run, about 80% of all tracked storms can be found in the “medium-range” wind speed category and approximately 15% in the “high-range” wind speed values. This distribution is a function of the storm tracking algorithm threshold, which has been chosen in such a way as to eliminate all weaker systems.
Compared to the control run, all SRES scenarios display a decrease in the number of storms in the “low-range“ and “medium-range” wind speed categories and an increase in the “high-range” wind speed group.
Multiple modelled storms can be combined into a single representative ‘composite’ storm. The impact of climate change on the intensity and spatial extent of storm windspeeds can then be analysed from this perspective. Both intensity and spatial extents of wind fields are likely to increase under climate change.
Figure 11 . Composite storm analysis of historical wind speeds (ms-1) at 850hPa (left), and the projected differences under climate change(centre and right). See text for detailed explanation. Source: Priestly & Catto 2022
Example composites formed of CMIP6 extreme storms in the North Atlantic by Priestly & Catto 2022 are shown in (Figure 11). The left plot shows winds at 850mb in m/s for the composite storm created using the 1979-2014 historical CMIP6 runs, as a function of latitudinal and longitudinal distance from the centre of the storm. The storm direction is to the right, with maximum winds located in the lower sector. The change in windspeed under SSP2-45 (centre plot) SSP5-85 (right plot) correspond to the 2080-2100 period. Winds increase by up to 0.9m/s under SSP5-85, and increases in both speed and area are concentrated in the lower forward sector of the composite storm footprint.
Figure 12 shows the impact of climate change on storm area as a function of time for several scenarios considered by Priestly & Catto 2022. The metric used is the area of 850mb windspeeds exceeding 17m/s normalised with respect to the 1979–2014 historical seasonal average (represented by the black line). The boxplots to the right show the distribution of annual averaged cyclone areas with the yellow lines corresponding to the median values and the limits of each box spanning the inter-quartile range. By 2050, even the strongest scenario (SSP5-85, purple) leads to only a small increase in area of ~3%. The model uncertainty, as represented by the length of the boxes in the plot, is very large with respect to the mean changes (yellow lines). In other words, the mean signals are dwarfed by the variation in predicted change across the CMIP6 models.
In this part of the article, we have seen how predicted changes in hazard according to the CoreLogic study broadly agree with mean predictions using CMIP6 (and earlier model) studies. Furthermore, model uncertainty is shown to be large compared to the mean changes.
In the final part of this article, we consider how these changes in hazard may translate into changes in loss.
Figure 12 . Evolution of cyclone area with Earth relative wind speeds. SSP1-26 (red), SSP2-45 (cyan), SSP3-70 (green), and SSP5-85 (purple). Source: Figure 12 of Priestly & Catto 2022.