More hurricanes to hit western Europe due to global warming



[1] We use a very high resolution global climate model (~25 km grid size) with prescribed sea surface temperatures to show that greenhouse warming enhances the occurrence of hurricane-force (> 32.6 m s–1) storms over western Europe during early autumn (August–October), the majority of which originate as a tropical cyclone. The rise in Atlantic tropical sea surface temperatures extends eastward the breeding ground of tropical cyclones, yielding more frequent and intense hurricanes following pathways directed toward Europe. En route they transform into extratropical depressions and reintensify after merging with the midlatitude baroclinic unstable flow. Our model simulations clearly show that future tropical cyclones are more prone to hit western Europe, and do so earlier in the season, thereby increasing the frequency and impact of hurricane force winds.

1 Introduction

[2] Severe hurricane-force (> 32.6 m s–1) storms can cause floods in west-European coastal regions and inflict large-scale damage on infrastructure and agriculture [Dorland et al., 1999; Sterl et al., 2009]. In the present climate, such storms primarily occur in winter and are associated with midlatitude baroclinic instability. However, global warming and increased sea surface temperatures (SSTs) have the potential to alter the intensity and cause of west-European storms.

[3] In the present-day climate, the vast majority of west-European storms originate from baroclinic instability in the midlatitudes, which is driven by the north-south atmospheric temperature gradient. In a warmer climate, due to increased concentrations of greenhouse gases, this meridional temperature gradient will decrease because the Arctic will warm faster than the equatorial regions, implying that baroclinic instability will reduce [Gitelman et al., 1997]. In itself this would result in reduced wintertime storms (in frequency and intensity). This is, however, counteracted by the increase of tropopause height and increased latent heat release, which tend to intensify storms [Lorenz and DeWeaver, 2007]. Recent studies suggest that these two opposing effects may approximately balance, as it has been found that global warming only induces minor changes in the frequency and intensity of midlatitude storms [Lambert and Fyfe, 2006; Bengtsson et al., 2006, 2009; Leckebusch and Ulbrich, 2004; Loptien et al., 2008; Ulbrich et al., 2008]. On the other hand, greenhouse warming will result in rising SSTs, which will extend the breeding ground of tropical hurricanes. Recent research indeed suggests that climate warming causes a poleward and eastward extension of the hurricane genesis area [Zhao and Held, 2012; Murakami et al., 2012]. This implies that future hurricanes would be increasingly able to affect extreme midlatitude storm conditions. As a result, the likelihood of strong storms hitting western Europe could become larger. Generally, climate warming may thus result in changes in the frequency, intensity, and seasonality of western European storms, for which tropical conditions may become increasingly responsible.

[4] To investigate possible changes in the development and pathways of tropical hurricanes in future climates, climate models are required. Unfortunately, the horizontal resolution of most current climate models is too coarse (100 km or more) to accurately simulate relatively small-scale weather systems such as tropical cyclones. Simulations using high-resolution models, however, indicate that the intensity and frequency of small-scale severe weather systems change significantly as a result of global warming. In particular, the intensity of tropical cyclones and heavy precipitation events have been projected to increase [Zhao and Held, 2012; Murakami et al., 2012; Knutson et al., 2010; Bender et al., 2010]. High-resolution climate models are thus required to adequately address possible changes in the occurrence of intense storms in western Europe.

[5] Here we investigate the change in the occurrence of severe midlatitude storms caused by global warming using the global climate model EC-EARTH [Hazeleger et al., 2010] at very high resolution to adequately resolve small-scale extreme weather systems, including tropical cyclones. We simulate the beginning and end of 21st century conditions, and compare both to assess the 21st century trends.

2 Model and Experimental Setup

[6] The atmospheric component of EC-EARTH is derived from the European Centre for Medium-Range Weather Forecasts numerical weather prediction model, which is also used operationally for forecasting tropical cyclones. The resolution in this experiment is T799 L91 (~25 km horizontal resolution, 91 vertical levels), which was the operational resolution at the European Centre for Medium-Range Weather Forecasts from February 2006 to January 2010.

[7] The experiment consisted of two sets of 5 year 6 member ensemble simulations for 2002–2006 (present) and 2094–2098 (future), respectively, resulting in a 30 year data set for each period. In the present simulations the observed greenhouse gases and aerosol concentrations were used, whereas for the future simulations the concentrations were derived from the RCP 4.5 scenario [Van Vuuren et al., 2011]. The SSTs are prescribed as lower boundary conditions. For the present simulations daily SSTs were taken from the NASA data set [] for the indicated period on a 0.25 degree horizontal resolution. Future SSTs were computed by adding to the present SST field the ensemble mean change in SST as simulated by the ECHAM5/MPI-OM model used in the ESSENCE project [Sterl et al., 2008]. This project consisted of a 17 member ensemble under the SRES A1B scenario, which is compatible with the RCP4.5 scenario. The increase in the global mean temperature at the end of the 21st century is in RCP4.5 about 1°C less than in SRESA1B [Rogelj et al., 2012]. Observed sea-ice coverage was taken for the present simulations. For the future simulations, the sea-ice coverage was computed by using a linear regression using the present SST and sea-ice cover fields.

[8] A 10 year spin-up run at low resolution (T159) was made for both the present and the future, followed by a 9 month (from January to October) spin-up run at T799 resolution. The 6 member ensemble was made by taking the atmospheric state of one of the first 6 days of October as initial state for each member. Thereafter, the model was run for another 3 months until 1 January before the data were used for the analysis. After this spin-up the spread in the atmospheric states was sufficient to treat the 6 runs as independent members.

[9] The choice of the specific years for the present and future period was motivated by the inclusion of dominant modes of variability like El Niño for the present period and a separation of a multiple of 4 years between present and future for obtaining the same number of leap years in the present and future period.

[10] The storms were tracked backward in time by computing the location of the minimum pressure in the preceding time step (3 h). This computation was carried out in an area of 3° by 3° surrounding the present location of minimum pressure. Because our aim is to locate the genesis region, the tracking algorithm ended when the intensity of the storm fell below that of a tropical depression (Beaufort 7; 13.9 m s–1).

3 Results

[11] We focus on changes in severe storms characteristics in four regions along the coast of western Europe (Figure 1): Norway, North Sea, western UK, and Gulf of Biscay. For each of these regions we determine the frequency distribution of 3-hourly wind speeds at 10 m height. They show that in the North Sea and Gulf of Biscay there is a clear increase in the frequency of severe winds (Beaufort 11–12, > 28.4 m s–1) during autumn (Figures 2a–2d). The signal is less clear for Norway and the western UK. Additionally, the season of highest occurrence shifts from winter to autumn in the North Sea and Gulf of Biscay regions. The maximum wind speed in early autumn increases substantially along the western European coast (Figure 2e), emphasizing the intensification of extreme winds during that season.

Figure 1.

SST (°C) change between future (2096–2098) and present (2002–2006) for August–October. The SST change is derived from the ESSENCE model (see Model and Experimental set-up). The colored squares indicate the different analysis regions: Norway (60°N–70°N, 0°E–15°E) [blue], North Sea (50°N–60°N, 3°W–8°E) [red], western UK (50°N–60°N, 3°W–15°W) [green], Gulf of Biscay (43°N–50°N, 0°W–15°W) [light blue].

Figure 2.

Wind speed changes over the 21st century along the western European coast. (a–d) Frequency distribution of the 3-hourly 10 m windspeed (U10) area-averaged over the different analysis regions for Beaufort 11–12 (U10 > 28.4 m s–1) for present (blue) and future (red) conditions. The frequency is multiplied by 102. The 95% significance intervals based on a Poisson distribution for the Future peaks in October for North Sea and Biscay are indicated by vertical bars. (e) Wind speed change (m s–1) for the period (August–October) between future and present conditions. Results are significant at 95% confidence level. Return period computations are done for the location of the black cross (see auxiliary material, Figure S4). (f) Number of hurricane force storms during (August–October) in the different basins for the present (blue) and future (red) climates.

[12] The occurrence of extreme wind speeds can be attributed to a few storms during the 30 year integration period (Figure 2f ). Summed over Norway, the North Sea, and the Gulf of Biscay the number of hurricane force storms (Beaufort 12, > 32.6 m s–1) in early autumn (August–October) increases from 2 to 13 over the 21st century. We have traced back the origin of these events (see Model and Experimental set-up), showing that virtually all future west-European hurricane-force storms originate as hurricanes or tropical storms in the tropics (Figure 3b). In contrast, the few hurricane-force storms in the present climate have predominantly an extratropical origin (Figure 3a). This suggests that the climatological mechanisms driving western Europe hurricane-force winds are likely to change dramatically over the 21st century.

Figure 3.

Tracks and frequency of hurricane-force storms. (upper panels) Backward tracks of storms with Beaufort 12 (> 32.6 m s–1) in one of the four analysis regions of Figure 1 during early autumn (August–October) for (a) present and (b) future climate. The storm intensity is indicated by the colors of the tracks. The SSTs (°C) during early autumn (August–October) for present and future climates are plotted as background. The black solid line denotes the 27 °C isotherm. (lower panels) Frequency (multiplied by 2×104) of 3-hourly Beaufort 12 (> 32.6 m s–1) wind speeds during (July–October) for (c) present and (d) future climate. The vertical wind shear (m s–1) between 850 and 200 hPa, computed as the length of the difference wind vector between these two levels, is depicted in (Figure 3e). Contours: present climate. Shaded: future minus present climate.

[13] In the auxiliary material we show that tropical cyclones will increase the probability of present-day extreme events over the North Sea and the Gulf of Biscay with a factor of 5 and 25, respectively, with far reaching consequences especially for coastal safety.

[14] In the current climate, the main genesis region for hurricanes is confined to the western tropical Atlantic, where SSTs are above the threshold (27°C) required for tropical cyclones to develop. (Note that the SST threshold for tropical storm development will be affected by changes in the stability characteristics of the atmosphere, with the projected tropospheric warming hampering tropical cyclone genesis.) Future tropical storms that reach western European coasts (and cause hurricane-force storms) predominantly originate from the eastern part of the tropical Atlantic (east of 50°W; Figure 3b). This is because climate warming in the eastern tropical Atlantic causes SSTs to rise well above the 27°C threshold (Figure 3b). The occurrence of hurricane-force winds indeed shows that, notwithstanding the increased atmospheric stability, the eastern boundary of the hurricane development region shifts about 10° to the east in response to the SST rise (Figures 3c and 3d). The present development region exhibits a decrease in hurricane frequency (see Figure S1e in the auxiliary material), which, incidentally, has significant implications for hurricanes reaching North America. This agrees with the finding that the number of Atlantic hurricanes to make landfall in North America is reduced during years with an anomalous warm tropical Atlantic and consequent eastward extension of the development region [Wang et al., 2011].

[15] After their formation, tropical cyclones are pushed poleward by the beta-effect [Rossby, 1948]. Combined with the ambient prevailing westward trade winds, this causes the initial path of Atlantic tropical cyclones to be in a northwesterly direction. When they reach the midlatitudes they are caught by the predominant westerly winds, thereby veering their track in a northeasterly direction, with the possibility of reaching western Europe. Geometrically, this likelihood increases if their genesis region in the tropical Atlantic is further to the east. In addition, the shorter travel distance in the midlatitudes will enable the “tropical” characteristics of hurricanes to be better preserved along their journey to western Europe. Hence, the likelihood of these storms maintaining their strength when reaching western Europe will increase, because there is simply less time for them to dissipate [Hart and Evans, 2001].

[16] The number and strength of hurricanes forming and reaching the midlatitudes is affected by many factors. An important one is vertical wind shear, associated with differences in SST between the Pacific and Atlantic oceans. In our simulations the change in vertical shear between upper and lower troposphere over the Atlantic basin is small, generally less than 2 m s–1 (Figure 3e), and is not likely to strongly affect the life cycle of hurricanes. Although there is large uncertainty about the wind shear and other factors that affect the genesis and development of hurricanes, our results for the change in frequency and intensity of hurricanes (Figure S1, auxiliary material) are compatible with other studies [Knutson et al., 2010].

[17] Before reaching the western European coastal regions the tropical storms undergo extratropical transition [Jones et al., 2003; Hart and Evans, 2001]. Their evolution can be depicted using the cyclone phase space diagrams of Hart [2003]. Figure 4 shows the ensemble mean evolution of the western Europe hurricane-force storms with tropical origin. Positive (negative) values of –VLT and –VUT indicate a warm (cold) lower and upper core, respectively. For a tropical symmetrical cyclone, B is approximately zero, whereas an asymmetrical developing cyclone has a large positive value of B. For details of the computation of these variables see Hart [2003] and the auxiliary material. Tropical cyclones have a warm core and an axial symmetric structure. During extratropical transition these typical characteristics disappear [Hart, 2003]: the horizontal size of the storm increases, they become asymmetrical, and the core temperature decreases (Figure 4). Significantly, though, these transformed storms intensify again, but only after they merge with an unstable background flow. In their final stage, one day before hitting western Europe, they exhibit a rapid and strong intensification (reddish points in Figure 4). The average MSLP drops from 970 to 952 hPa and the average maximum wind speeds increases from 27 ms–1 (Bf 10) to 33.9 ms–1 (Bf 12). This is accompanied by a renewed warming of especially the lower part of the core, indicating that diabatic heating due to the release of latent heat plays a crucial role. The anomalously large moisture content of these storms of tropical origin and the increased midlatitude humidity caused by global warming thus appear to be crucial in their reintensification. Figures S2 and S3 show a telling example of a reintensifying storm, its merging with the midlatitude baroclinic unstable flow, and the reemergence of its tropical storm characteristics.

Figure 4.

Mean evolution of western Europe hurricane-force storms during the last 6 days before they hit the western European coast, represented in cyclone phase space (a) –VLT vs B and (b) –VLT vs –VUT. A circular marker is placed every 3 h. The numbers next to the markers indicate the days before hitting the western European coast. The color of the markers indicates the mean sea level pressure intensity and the size the mean radius R of the cyclone.

[18] For the current climate, the model simulates 18.3 hurricane days (> 32.6 m s–1) per season, which is slightly less than the observed long-term mean of 23.8 [Gray and Landsea, 1992]. The number of simulated intense hurricane days (IHD) (> 49.2 m s–1, Cat 3–5) is 0.7, much smaller than the observed long-term mean of 5.7, illustrating the inability of the model to simulate the most severe hurricanes. The model simulations depict a more than fourfold increase in major hurricanes over the 21st century, with IHD increasing to 3.1 (see also Figure S1). More intense hurricanes are generally able to survive longer in a nonsupportive environment [Hart and Evans, 2001]. The consequence of the underestimation of IHD in the present climate is that even more hurricanes could hit western Europe, but higher resolutions would be required to properly investigate this.

[19] Another caveat is that the model uses fixed SSTs, which prevents the possibility of negative feedback on intensity due to SST cooling associated with vertical ocean mixing or latent heat fluxes, which might significantly reduce the intensity of hurricanes.

[20] The results of this study are based on two samples of 30 years. Although we argue that the main conclusions are not affected by the small sample size, we realize that more extended simulations are needed to further test our results.

4 Conclusions

[21] We have shown that in our model simulations global warming yields an eastward extension of the development region of tropical storms. Together with higher SSTs this implies that tropical storms are more likely to reach the midlatitude baroclinic region before they dissipate. After merging with the baroclinic unstable flow in the midlatitudes they reintensify, aided by upper-level diabatic heating owing to the release of latent heat. As a result, the reintensified storm is likely to regain hurricane-force winds along the coast of western Europe. Such revived hurricanes are known to cause widespread damage along the North American east coast in the current climate. Our model simulations suggest that tropical hurricanes might become a serious threat for western Europe in the future. Hence, we anticipate an increase in severe storms of predominantly tropical origin reaching western Europe as part of 21st global warming.