Review on the Projections of Future Storminess over the North Atlantic European Region

Abstract

This is an overview of the results from previously published climate modeling studies reporting on projected aspects of future storminess over the North Atlantic European region (NAER) in the period 2020–2190. Changes in storminess are summarized for seven subregions in the study area and rated by a categorical evaluation scheme that takes into account emission scenarios and modeling complexity in the reviewed studies. Although many of the reviewed studies reported an increase in the intensity of high-impact wind speed and extreme cyclone frequency in the second half of the 21st century, the projections of aspects of future storminess over the NAER differed regionally. There is broad consensus that the frequency and intensity of storms, cyclones, and high-impact wind speed will increase over Central and Western Europe, and these changes will probably have the potential to produce more damage. In contrast, future extratropical storminess over Southern Europe is very likely to decrease. For Northern and Eastern Europe the results of the evaluation are inconclusive, because there is an indication of increasing as well as decreasing development of the evaluated aspects of future storminess. Concerning the storm track, we found indications of a likely north- and eastward shift in most assessed studies. Results from three studies suggest a northeastward shift of the North Atlantic Oscillation.

1. Introduction

Synoptic winter storms are a typical feature of the climate over the North Atlantic European region (NAER). Over continental Europe, they bear a high damage potential to property, infrastructure, and the environment and cause deaths and injuries. In the period 1980–2014, loss associated with the ten costliest European winter storms amounted to more than US$63 billion and close to 500 fatalities occurred through the passage of these storms [1]. In Central Europe, severe winter storms are the most loss-relevant natural hazard events [2]. They are also part of the natural disturbance dynamics of European forest ecosystems with severe winter storms currently representing the main hazard, because of their large expansion and high-impact near-surface flow field characteristics. In recent decades (1950–2010), they have caused 51% of all damaged timber and thus more than all other natural hazards (mainly fire, snow, and bark beetles) together [3]. Furthermore, severe storms chronically act on Atlantic coastlines and induce sand drift and erosion [4,5,6].

The North Atlantic Oscillation (NAO) is linked to changes in extratropical storm tracks and large-scale circulation patterns over Europe [7,8,9,10,11]. Since the NAO is more pronounced during winter, most severe synoptic storms occur between December and March [7,12,13]. In the period 1500–1995, 66% of all high-impact storm events occurred between October and March [14], with most storms passing over Europe during December. There is evidence that the NAO currently has at least two modes with periods of 10.4 and 62 years, the latter being related to long-term changes in Atlantic sea surface temperature [15].

Winter storms regularly disturb the planning of thinning and harvesting operations in forestry, and forest management seeks robust information on the future development of storminess, because goals of forest management are defined over a period of several decades. However, not only stakeholders in the forestry sector are interested in robust information on future storminess. Knowledge about the characteristics of future high-impact storms is also of great interest for a number of other socio-economic sectors. For the insurance sector, storms are of major importance, because (re)insurance companies usually hedge financial risk induced by storms and are keen to assess storm-induced loss potentials [16,17]. Wind turbine operators have to shut down their installations to avoid damage during the passage of storms [18,19]. In aviation, severe storms affect airline operations [20,21].

A large number of climate modeling studies report on the projected future development of storminess over the NAER. Due to the importance of the issue, main outcomes and various aspects from these studies have already been summarized and discussed in review articles in recent years: The study of [22] summarized the importance of high-impact winter storms for European forest ecosystems and forestry. A review of cyclone characteristics in the mid-latitudes under current and future climate in the southern and northern hemisphere was presented by [23], while [11] reported on the effect of sea ice variability on storm tracks and NAO-variability. The most recent review on storminess over the North Atlantic and Northwestern Europe analyzed the available literature for storm frequency trends, changes in cyclone activity, and storm tracks [24].

The present review is an update and further development of the reviews of [22,24]. It combines a refined version of the study evaluation methodology applied by [22] with the regional evaluation approach for different parts of the NAER presented by [24]. Furthermore, we defined and evaluated the reviewed studies for the categories “storms”, “cyclones”, and “wind speed” separately, which were often used to quantify aspects of projected future storminess. For these categories, we assessed changes in frequency, intensity, and/or activity. For the assessment, we divided the NAER into seven subregions, which we believe to represent regional differences in aspects of current storminess over the study area.

The goal of this review is to provide a comprehensive overview of the projected changes of measures that are commonly used to quantify storminess under future climate conditions over the NAER. The intention of using a categorical evaluation scheme is to meet the growing demand for information on the future development of storminess for strategic management processes and decision-making in various ecological and socio-economic sectors that are exposed to the atmospheric environment.

2. Methods

2.1. Literature Search

This review is based on an extensive literature search in the databases Google Scholar, ScienceDirect, Thomson Reuters Web of Science, Wiley Online Library, and Catalog, plus the library of the Albert-Ludwigs-University of Freiburg (Freiburg, Germany). We searched for articles that reported on aspects of projected future storminess over the NAER. Keywords related to “wind”, “storm”, “cyclone”, “storminess”, “North Atlantic European region”, “Europe”, “climate model”, and “projection” were used and combined during the search process. To assess the general suitability of the search results for this review, we started reading the abstracts and selected relevant articles for further analysis based on our expert knowledge. All 58 studies—including basic information on the models used to project aspects of future storminess over the NAER—that have finally been evaluated in this review are listed in Appendix A, Table A1.

2.2. Terms and Definitions

While reviewing studies on projected storminess over the NAER, it quickly became evident that the definition of the term “storminess” is ambiguous. In the evaluated studies it was either characterized directly by different measures of wind speed and/or pressure or indirectly by impact-related variables such as storm-induced damage and losses. Furthermore, the definition of “storm” was inconsistent. For example, “storm” has not only been defined as a strong gale (near-surface wind speed ≥ 9 Bft) but also as severe cyclone.

Due to the ambiguity of the definitions of storminess, we provide a number of definitions of indicators that we used to evaluate and interpret projections of aspects of future storminess:

• Extratropical cyclones can be defined as a minimum in the mean sea level pressure field or at the 1000 hPa geopotential height [25].
• Cyclone frequency can be quantified by counting the number of cyclones occurring in gridded datasets [23]. Changes in cyclone frequency are often expressed as changes in cyclone return period [26,27,28,29,30].
• Cyclone intensity is characterized by the depth of a pressure center at sea level or by the relative pressure minimum at the 1000 hPa geopotential height. Low pressure is always associated with high cyclone intensity. Alternative measures of cyclone intensity are magnitude of vorticity, precipitation intensity, or wind speed [23,31].
• Cyclone activity was defined and reported as the product of the number of cyclones and their mean intensity [32]. This entails that it was not always possible to attribute changes in cyclone activity to either changes in cyclone frequency or changes in cyclone intensity. Cyclone activity can also be quantified by cyclone track density [23].
• Cyclone track density was defined as the number of cyclone tracks which can be counted over time in gridded datasets [23]. Cyclone track density does not provide any information on the intensity of the counted cyclones [33]. We therefore interpreted information on cyclone track density as a measure of cyclone frequency.
• A storm track is the path of a storm over time. Originally, a storm track referred to preferred regions of storm/cyclone activity. “Storm track” was also used to describe clusters of cyclone tracks in different geographical regions [24]. We interpreted information on storm track activity as a measure of cyclone activity [34].
• Storm intensity can be characterized by wind speed measurements made at 10 m above the ground. In the reviewed literature, it was reported as (i) peak wind speed values averaged over 1–3 s; (ii) mean wind speed averaged over 10 min intervals; or (iii) percentiles (e.g., 90–98 percentiles) of maximum daily wind speed values [13,16,35,36,37,38].

Due to the numerous definitions of “storminess” it proved to be impossible to always unambiguously interpret the future development of stormy conditions based on the reviewed studies. Especially, the terms “storm” and “extreme cyclone” were often improperly defined, not only between studies, but also in one and the same study. When the definition of storm and extreme cyclone was ambiguous within the very same study, we interpreted the reported results as changes in cyclone characteristics.

Additionally ambiguous was the use of the term “wind speed”. Wind speed has either been used as a measure for storm intensity or cyclone intensity, or was investigated without relating to storms and/or extreme cyclones. In the following analysis, we use the term “wind” as an abbreviation for “high-impact wind speed”.

Table B1 in Appendix B summarizes the emission scenarios and the variables investigated in the evaluated studies and their assignment to categories (frequency, intensity, activity) of aspects (storms, extreme cyclones, high-impact wind speed) of projected storminess assessed in this review.

2.3. Study Evaluation and Rating

The present analysis is a modification of the approach of [22]. In their review of the past, present, and future storminess over the NAER, they interpreted findings from 32 studies on the basis of a categorical evaluation scheme of

• data volume differentiated for the categories (i) measured/observed data; (ii) data from reanalysis; (iii) combination of reanalysis and climate change scenarios, and
• methodology with the categories (i) trend analysis/simple statistics; (ii) output from an individual model/complex statistics; and (iii) output from multi-model ensembles.

We basically adopted this classification scheme, but focused on projections of the future development of storminess, which implies that we did not include measured/observed data and data from reanalysis into the present assessment. The rating of the significance of an evaluated study was completed with the type of emission scenario (S) and the number (N) of general circulation models (GCMs) used to project future storminess, with grades from 1 (lowest rating) to 3 (highest rating).

To assess S, we differentiated between (i) CO₂-experiments or IS92-scenarios [39] (S = 1); (ii) Special Report on Emission Scenarios (SRES) [40] (S = 2); and (iii) Representative Concentration Pathways (RCP) [41] (S = 3). The differentiation between the emission scenarios was based on the assumption that newer emission scenarios better represent the current state of scientific knowledge in emission scenario modeling.

The assessment of the applied methodology was based on the number of applied GCMs. The number of regional climate models (RCMs) was not considered because it has been demonstrated that GCM-forcing affects projection results more strongly than RCM-outputs [13,42,43]. Although the spread produced by the RCMs is smaller than the GCM-spread, it had an effect on the results reported in the evaluated studies. However, we did not take the number of RCMs into account, because this would have reduced the total number of evaluated studies. The categorical interpretation key for assessment of N differentiates between (i) single models (one GCM or one GCM together with one or more RCMs, N = 1); (ii) multiple GCMs (2 ≤ number of GCMs < 10, N = 2); and (iii) multi-model ensembles (number of GCMs ≥ 10, N = 3). The differentiation between the numbers of applied models was based on the assumption that the combined information of multiple models generally improves the accuracy, reliability, and consistency of projections [44]. The differentiation between the N-values of 2 and 3 was made at ten GCMs to balance the number of studies assigned to these two categories.

The evaluation score (ES) for individual studies was calculated emphasizing the effect of the applied GCMs on projections of storminess over the assumed future scenario [13] as follows:

ES = S + 2·N

(1)

After the comparison (data not shown) of several combinations of weighted and unweighted S- and N-values, it turned out that the applied weighting scheme allows for the intended differentiation between the evaluated studies without affecting the overall subregion-specific ES-pattern.

To better account for typical large-scale circulation patterns known to occur over the NAER, we divided the study area into the seven subregions (Figure 1) in a similar fashion, as has been done by [24]. The subregions we defined are (i) North Atlantic north of 60°N; (ii) North Atlantic south of 60°N; (iii) Northern Europe (including Baltic Sea); (iv) Central Europe; (v) Southern Europe; (vi) Western Europe (including North Sea); and (vii) Eastern Europe. Based on the individual study-specific ES-values, total ES-scores and mean ES-values ($\overline{ES}$) were calculated to provide the total and mean weight of all studies pooled for one of the subregions.

Figure 1. Differentiation of the study area (North Atlantic European region: NAER) into seven subregions. The horizontal dashed line at 60°N divides the North Atlantic into the subregions North Atlantic north of 60°N and North Atlantic south of 60°N. The vertical dashed lines indicate the eastern boundaries of the two North Atlantic subregions.

We assigned the projected and reported long-term evolution of aspects of storminess to the categories frequency, intensity, and activity of storms and extreme cyclones to these subregions based on the definitions provided above. This means that we do not always report on the meteorological variables actually presented in the reviewed studies (e.g., the future development of wind speed measured at 10 m above ground, change of wind speed at different pressure levels such as the 850 hPa pressure level) but interpreted them in terms of their projected long-term changes in frequency, intensity, and activity as either “storms” or “extreme cyclones” or “high-impact wind speed”. The reported tendency (T) in the long-term evolution of the reported variables was assessed and quantified using the categorical interpretation key (i) T = –1 for a decreasing tendency; (ii) T = 0 for no tendency; and (iii) T = +1 for an increasing tendency. Interestingly enough, in total, only four [45,46,47,48] out of the 58 evaluated studies reported no tendency in any one of the assessed categories of future storminess.

To allow for a quantitative comparison of the evaluation results obtained for the seven subregions, an effective tendency (eT) was computed based on subregions-related ES- and T-values:

$$eT={\displaystyle \sum _{j=1}^n\left(E{S}_j\cdot T_j\right)}/{\displaystyle \sum _{j=1}^{n}E{S}_j}$$

(2)

Results (data not shown) from a sensitivity analysis demonstrated that different combinations of weighted S- and N-values induced only minor changes in the eT-score (maximum ΔeT = 0.06) and did not change the structure of the presented inter-subregional results.

For the overall inter-subregional assessment, the eT-score puts more weight on studies with high ES-values and varies between −1.00 and 1.00. More negative values indicate a stronger tendency for a decreasing storminess in the future, while increasingly positive values indicate a stronger tendency for increasing storminess. With regard to the results from this study, we interpret eT-values as indicators for the reliability of the tendency of future projections, which implies that eT-values close to either −1.00 or 1.00 indicate clear decreasing or increasing tendencies in the evaluated aspects of projected storminess.

In addition to the assessment of future storminess in the seven subregions, we summarized results reported in the evaluated studies on projected changes of NAO- and storm track characteristics over the NAER. Although we also scored the applied emission scenario and methodology to calculate study-specific ES-values, we do not provide T- and eT-values, because projected changes in NAO- and storm track characteristics were spatially interpreted.

No differentiation of the evaluation results was made concerning the projection periods because the respective starting points in time, lengths, and endpoints were too diverse, leading to too many subgroups with too few members.

Provided that multiple projections were performed in a study, we evaluated only the study’s overall results and not the results obtained from individual projections included in the respective study. We focused on the overall results because synthesized results have been reported to be superior to results reported from individual projections [44]. Moreover, systematically considering the problems associated with individual projections in the evaluated studies, e.g., physical backgrounds, metrics of model validation, model dependences, experimental design, and model tuning [44], was beyond the scope of this paper. We, however, are aware that the chosen approach has an influence on our results and our interpretation of the future development of storminess.

Altogether, the reviewed studies cover the period 2020–2190. The spectral grid resolutions of the GCMs used in these studies varied at least between T21 and T213, which equal horizontal grid resolutions of 500 km to 49 km in the study area. With one exception [47], the evaluated studies have a focus on the European cold season including the months September to April. Typical terms and acronyms found in the reviewed studies referring to the cold season were “winter” without specifying individual months and “DJF” (December, January, and February). One reviewed study modeled future storminess for the month February representing winter conditions [49].

The evaluated studies summarized in tables in the Results section are primarily ranked in descending order based on the study-specific ES-value and secondarily based on the surname of the lead author.

3. Results

3.1. North Atlantic North of 60°N

In the North Atlantic region north of 60°N, highest N-ratings were awarded to two [50,51] out of eleven evaluated studies (Table 1, [9,13,25,27,37,50,51,52,53,54,55]), because they used multi-model ensembles with ten or more GCMs. In addition, the simulations of [51] were based on RCP-projections which, together with the large number of GCMs, yield ES = 9. Studies that reported results obtained from less than ten GCMs or from only one GCM (e.g., [25,52]) typically got lower N-ratings.

Table 1. Studies reporting on projections of aspects of future storminess in the North Atlantic subregion north of 60°N.

Study	Model(s)	S	N	ES	T	Category	Reference Period(s),	Region(s)
							Projection Period(s),
							Month(s), Season
Harvey et al. (2012) [50]	GCM: 19	3	3	9	−1	Storm activity	1976–2005,	Ocean basin north of 60°N, Barents Sea
	RCM: -						2070–2099,
							DJF
Zappa et al. (2013) [51]	GCM: 19	3	3	9	−1	Wind intensity	1976–2005,	Norwegian Sea
	RCM: -						2070–2099,
							DJF
Gregow et al. (2012) [53]	GCM: 9	2	2	6	−1	Wind intensity	1971–2000,	Norwegian Sea
	RCM: -						2046–2065, 2081–2100,
							September–April
Leckebusch et al. (2006) [13]	GCM: 4	2	2	6	−1	Frequency of extreme cyclones	1961–1989,	Norwegian Sea, Iceland
	RCM: 4						2071–2099,
							October–March
Schwierz et al. (2010) [27]	GCM: 2	2	2	6	−1	Storm intensity	1961–1990,	Iceland
	RCM: 2						2071–2100,
							October–March
Bengtsson et al. (2006) [9]	GCM: 1	2	1	4	−1	Cyclone intensity	1961–1990,	Northern part of the North Atlantic and Norwegian Sea
	RCM: -						2071–2100,
							DJF
Leckebusch and Ulbrich (2004) [54]	GCM: 1	2	1	4	−1	Frequency of extreme cyclones	1960–1989,	Norwegian Sea
	RCM: 1						2070–2099,
							October–March
Pinto et al. (2007) [37]	GCM: 1	2	1	4	−1	Wind intensity	1960–2000,	Polar latitudes
	RCM: -						2060–2100,
							October–March
Pinto et al. (2009) [55]	GCM: 1	2	1	4	−1, −1	Frequency of extreme cyclones, cyclone intensity	1960–2000,	Barents Sea, Iceland, Norwegian Sea
	RCM: -						2060–2100,
							October–March
Geng and Sugi (2003) [52]	GCM: 1	1	1	3	−1	Frequency of extreme cyclones	1978–1998,	Greenland Sea, Norwegian Sea
	RCM: -						~2050,
							DJF
Pinto et al. (2006) [25]	GCM: 1	1	1	3	+1, +1	Frequency of extreme cyclones, cyclone intensity	1880–1930,	Northern North Atlantic
	RCM: -						2039–2089,
							October–March

The majority of the studies reported on changes of projected frequency of extreme cyclones and cyclone intensity. It is obvious that, in all analyzed categories of storminess a prevailingly decreasing development is found. One exception is the study of [25], who expected an increase in both projected frequency of extreme cyclones and cyclone intensity.

The subregion-related total ES-score is 65, $\overline{ES}$-score is 5.0, and the eT-value equals −0.82, which, according to our evaluation scheme, implies a strong decreasing tendency of projected future storminess.

3.2. North Atlantic South of 60°N

Compared to the North Atlantic region north of 60°N, a different picture emerges in the North Atlantic region south of 60°N (Table 2, [2,9,13,31,45,47,50,52,54,56,57,58,59,60,61,62]). Results from 16 evaluated studies demonstrate that it is indisputable that storminess is projected to generally increase. Almost all studies indicated a positive change in at least one of the analyzed aspects of future storminess. Results from several studies suggested not only an increase in storm and cyclone frequency but also an increase in storm and cyclone intensity. From the results of the study of [2], it can be inferred that future wind intensity will also increase. Results of the study of [50], which scored ES = 9, showed an increase in cyclone activity in a large number of projections made in their study. Only one study did not identify a change in wind intensity [47].

Table 2. Same as Table 1 but for the North Atlantic subregion south of 60°N.

Study	Model(s)	S	N	ES	T	Category	Reference Period(s),	Region(s)
							Projection Period(s),
							Month(s), Season
Harvey et al. (2012) [50]	GCM: 38	3	3	9	1	Cyclone activity	1976–2005,	Ocean basin in the mid-latitudes
	RCM: -						2070–2099,
							DJF
Ulbrich et al. (2008) [56]	GCM: 15	2	3	8	1	Cyclone activity	1960–2000,	Eastern North Atlantic
	RCM: -						2081–2100,
							winter
Donat et al. (2010) [2]	GCM: 7	2	2	6	+1, +1	Cyclone intensity, wind intensity	Various model–	North Atlantic to Western Central Europe
	RCM: -						dependent periods,
							October–March
Leckebusch et al. (2006) [13]	GCM: 4	2	2	6	1	Frequency of extreme cyclones	1961–1989,	Northeast Atlantic
	RCM: 4						2071–2099,
							October–March
Leckebusch et al. (2008) [57]	GCM: 7	2	2	6	1	Storm frequency	1961–2000,	Northeast Atlantic
	RCM: -						2081–2100,
							winter
Bengtsson et al. (2006) [9]	GCM: 1	2	1	4	+1, +1	Frequency of extreme cyclones, cyclone intensity	1961–1990,	North Atlantic region between 45°N–60°N
	RCM: -						2071–2100,
							DJF
Bengtsson et al. (2009) [31]	GCM: 1	2	1	4	1	Cyclone intensity	1959–1990,	North Atlantic south of 60°N
	RCM: -						2069–2100,
							DJF
Hanson et al. (2003) [45]	GCM: 1	2	1	4	1	Frequency of extreme cyclones	1961–1990,	North Atlantic north and west of the UK
	RCM: 1						2070–2099,
							October–March
Leckebusch et al. (2008) [58]	GCM: 1	2	1	4	1	Storm intensity	1971–2000,	Northeast Atlantic
	RCM: -						2071–2100,
							October–March
Leckebusch and Ulbrich (2004) [54]	GCM: 1	2	1	4	+1, +1	Frequency of extreme cyclones, cyclone intensity	1960–1989,	Northeast Atlantic
	RCM: 1						2070–2099,
							October–March
Lionello et al. (2008) [59]	GCM: 1	2	1	4	+1, +1	Frequency of extreme cyclones, cyclone intensity	1961–1990,	Northeast Atlantic
	RCM: 1						2071–2100,
							winter
McDonald (2011) [60]	GCM: 1	2	1	4	1	Storm frequency	1960–1990,	North Atlantic southwest of the UK and west of France
	RCM: -						2070–2100,
							DJF
Pinto et al. (2007) [61]	GCM: 1	2	1	4	1	Frequency of extreme cyclones	1960–2000,	Northeast Atlantic
	RCM: -						2060–2100,
							October–March
Semmler et al. (2008) [47]	GCM: 1	2	1	4	0	Wind intensity	1985–2000,	Northeast Atlantic
	RCM: 1						2085–2100,
							May–December
Beersma et al. (1997) [62]	GCM: 1	1	1	3	1	Storm frequency	1980s,	Bay of Biscay
	RCM: -						2040s,
							-
Geng and Sugi (2003) [52]	GCM: 1	1	1	3	1	Frequency of extreme cyclones	1978–1998,	North Atlantic in the mid-latitudes, west of the British Isles
	RCM: -						~2050,
							DJF

The subregion-related total ES-score amounts to 95, the $\overline{ES}$value is 4.8, and the eT-value equals 0.96. It is the highest eT-value of this assessment and clearly points to an increasing tendency of the projected storminess in this subregion.

3.3. Northern Europe

The results of the 20 studies evaluated for Northern Europe are more diverse (eT = 0.54) than the results obtained for the two North Atlantic regions north and south of 60°N (Table 3, [17,27,28,30,31,37,46,49,53,54,55,62,63,64,65,66,67,68,69]). Northern Europe is, together with Eastern Europe, a subregion for which we found no study that achieved ES = 9. However, for the evaluated aspects of projected storminess, our scoring includes all categories of T, with the study of [46] being the only study that found no change in their analyzed aspect (wind intensity) of future storminess. The subregion-related total ES-score is 108, and the $\overline{ES}$-value equals 4.5.

Table 3. Same as Table 1 but for the Northern Europe subregion.

Study	Model(s)	S	N	ES	T	Category	Reference Period(s),	Region(s)
							Projection Period(s),
							Month(s), Season
Gastineau and Soden (2009) [64]	GCM: 16	2	3	8	1	Wind intensity	1995–2000,	Northern Europe
	RCM: -						2095–2100,
							-
Pryor et al. (2006) [46]	GCM: 10	2	3	8	0	Wind intensity	1961–1990,	Scandinavia
	RCM: -						2046–2065, 2081–2100,
							-
Donat et al. (2011) [17]	GCM: 7	2	2	6	1	Wind intensity	1960–2000,	Baltic Sea, Southern Scandinavia
	RCM: 9						2021–2050, 2071–2100,
							October–March
Gregow et al. (2012) [53]	GCM: 9	2	2	6	1	Wind intensity	1971–2000,	Eastern (20°E–40°E) and Northern Europe
	RCM: -						2046–2065, 2081–2100,
							September–April
Löptien et al. (2008) [65]	GCM: 2	2	2	6	1	Frequency of extreme cyclones	1978–1999,	Southern Sweden
	RCM: -						2070–2090, 2170–2190
							-
Nikulin et al. (2011) [66]	GCM: 6	2	2	6	1	Wind intensity	1961–1990,	Baltic Sea
	RCM: 1						2071–2100,
							–
Schwierz et al. (2010) [27]	GCM: 2	2	2	6	−1, −1	Frequency of extreme wind speed, wind intensity	1961–1990,	Northern and Central Scandinavia, Scandinavia
	RCM: 2						2071–2100,
							October–March
Bengtsson et al. (2009) [31]	GCM: 1	2	1	4	+1, +1	Frequency of extreme wind speed, wind intensity	1959–1990,	Northern Europe
	RCM: -						2069–2100,
							DJF
Giorgi et al. (2004) [63]	GCM: 1	2	1	4	1	Storm activity	1961–1990,	Northern Europe
	RCM: 1						2071–2100,
							DJF
Karremann et al. (2014) [30]	GCM: 1	2	1	4	−1	Storm frequency	1960–2000,	Scandinavia (parts of Sweden), Denmark
	RCM: -						2060–2100,
							October–March
Leckebusch and Ulbrich (2004) [54]	GCM: 1	2	1	4	−1	Frequency of extreme cyclones	1979–1989,	Northern Europe
	RCM: 1						2070–2099,
							October–March
Mizuta et al. (2011) [67]	GCM: 1	2	1	4	+1, +1	Frequency of extreme cyclones, frequency of extreme wind speed	1979–2003,	Scandinavia, Baltic Sea
	RCM: -						2075–2099,
							DJF
Pinto et al. (2007) [37]	GCM: 1	2	1	4	1	Wind intensity	1960–2000,	Baltic Sea
	RCM: -						2060–2100,
							October–March
Pinto et al. (2009) [55]	GCM: 1	2	1	4	1	Frequency of extreme cyclones	1960–2000,	Northern Europe
	RCM: -						2060–2100,
							October–March
Pinto et al. (2012) [28]	GCM: 1	2	1	4	1	Storm intensity	1960–2000,	Denmark
	RCM: -						2060–2100,
							October–March
Pryor et al. (2012) [68]	GCM: 1	2	1	4	1	Wind intensity	1961–1990,	Southwestern Scandinavia, Baltic Sea
	RCM: 2						2036–2065, 2070–2099,
							-
Beersma et al. (1997) [62]	GCM: 1	1	1	3	−1	Storm frequency	1980s,	Northern Europe
	RCM: -						2040s,
							-
Knippertz et al. (2000) [69]	GCM: 1	1	1	3	+1, +1, +1	Fequency of extreme cyclones, cyclone activity, wind intensity	1880–1930,	Northern Europe
	RCM: -						2039–2089,
							DJF
Walter et al. (2006) [49]	GCM: -	-	1	2	1	Wind intensity	1969–1989,	Northern Baltic Sea
	RCM: 4						2070–2099,
							February

Since only a few of the assessed studies reported on future development of “storms” no conclusive tendency can be identified for this variable. In two studies [30,62], expected future storm frequency decreased while two other studies projected an increase in storm intensity [28] and storm activity [63].

Most often studied were the future developments of extreme cyclone frequency and wind intensity. In both categories, model results support the conclusion that the number of cyclones as well as the wind intensity will increase towards the end of the 21st century. One study reported results that were obtained from projections of several RCMs without explicitly specifying a GCM [49].

3.4. Central Europe

The studies reviewed for Central Europe (Table 4, [2,17,27,28,30,31,37,38,49,50,51,54,55,61,63,64,70,71,72,73,74,75,76]) indicated an increase in almost all of the assessed aspects of storminess. However, although we reviewed 23 studies for this subregion, we found no clearly interpretable results on projected changes in future cyclone activity.

Table 4. Same as Table 1 but for the Central Europe subregion.

Study	Model(s)	S	N	ES	T	Category	Reference Period(s),	Region(s)
							Projection Period(s),
							Month(s), Season
Zappa et al. (2013) [51]	GCM: 19	3	3	9	+1, +1	Frequency of extreme cyclones, cyclone intensity	1976–2005,	Central Europe
	RCM: -						2070–2099,
							DJF
Gastineau and Soden (2009) [64]	GCM: 16	2	3	8	1	Wind intensity	1995–2000,	Central Europe
	RCM: -						2095–2100,
							-
Harvey et al. (2012) [50]	GCM: 19	2	3	8	1	Storm activity	1976–2005,	Central Europe
	RCM: -						2070–2099,
							DJF
Donat et al. (2010) [2]	GCM: 7	2	2	6	+1, +1	Storm frequency, storm intensity	Various model-	Central Europe (50°N, 10°E)
	RCM: -						dependent periods,
							October–March
Donat et al. (2011) [17]	GCM: 7	2	2	6	1	Wind intensity	1960–2000,	Northern Central Europe
	RCM: -						2021–2050, 2071–2100,
							October–March
Schwierz et al. (2010) [27]	GCM: 2	2	2	6	+1, +1	Frequency of extreme wind speed, wind intensity	1961–1990,	Germany, Poland, Baltic Sea, Central Europe in a belt south of Great Britain
	RCM: 2						2071–2100,
							October–March
Bengtsson et al. (2009) [31]	GCM: 1	2	1	4	+1, +1	Frequency of extreme cyclones, cyclone intensity	1959–1990,	Central Europe
	RCM: -						2069–2100,
							DJF
Beniston et al. (2007) [70]	GCM: 1	2	1	4	1	Storm intensity	1961–1990,	Central Europe (45°N–55°N)
	RCM: 7						2071–2100,
							DJF
Fink et al. (2009) [38]	GCM: 1	2	1	4	1	Storm frequency	1970–1999,	Central Europe, Germany, Poland
	RCM: 2						2070–2099,
							winter
Gerstengarbe et al. (2013) [71]	GCM: -	2	1	4	1	Storm frequency	1984–2008,	Germany
	RCM: 1						2011–2040, 2041–2070,
							-
Giorgi et al. (2004) [63]	GCM: 1	2	1	4	1	Storm activity	1961–1990,	Central Europe
	RCM: 1						2071–2100,
							DJF
Karremann et al. (2014) [30]	GCM: 1	2	1	4	1	Storm frequency	1960–2000,	France, Germany
	RCM: -						2060–2100,
							October–March
Klaus et al. (2011) [72]	GCM: -	2	1	4	1	Storm frequency	1971–2000,	Northwestern Germany
	RCM: 2						2031–2060,
							winter
Leckebusch and Ulbrich (2004) [54]	GCM: 1	2	1	4	1	Frequency of extreme cyclones	1979–1989,	Central Europe
	RCM: 1						2070–2099,
							October–March
Pinto et al. (2007) [37]	GCM: 1	2	1	4	1	Wind intensity	1960–2000,	Central Europe
	RCM: -						2060–2100,
							October–March
Pinto et al. (2007) [61]	GCM: 1	2	1	4	1	Wind intensity	1960–2000,	Central Europe
	RCM: -						2060–2100,
							October–March
Pinto et al. (2009) [55]	GCM: 1	2	1	4	1	Storm frequency	1960–2000,	Central Europe
	RCM: -						2060–2100,
							October–March
Pinto et al. (2010) [73]	GCM: 1	2	1	4	+1, +1	Storm frequency, wind intensity	1960–2000,	Western Germany
	RCM: 1						2060–2100,
							October–March
Pinto et al. (2012) [28]	GCM: 1	2	1	4	1	Storm intensity	1960–2000,	France, Germany
	RCM: -						2060–2100,
							October–March
Rauthe et al. (2010) [74]	GCM: 1	2	1	4	+1/−1	Wind intensity	1971–2000,	Northern, Central and Southern Germany
	RCM: 3						2021–2050,
							winter
Rockel and Woth (2007) [75]	GCM: 1	2	1	4	+1, +1	Frequency of extreme wind speed, wind intensity	1961–1990,	Central Europe
	RCM: 8						2071–2100,
							-
Leckebusch et al. (2008) [76]	GCM: 1	1	1	3	+1, +1, +1	Storm frequency, cyclone intensity, frequency of extreme wind speed	1880–1930,	Central Europe
	RCM: -						2039–2089,
							October–March
Walter et al. (2006) [49]	GCM: -	-	1	2	1	Wind intensity	1969–1989,	Germany
	RCM: 4						2070–2099,
							February

With respect to the applied methodology, we rated the studies of [50,51,64] as complex, multi-model approaches (M = 3). A shift towards a higher storm activity was identified by [50], while [51] projected an increase in the frequency of extreme cyclones, which are associated with high wind speed at the 850 hPa pressure level. Results from [64] suggested a shift of the 95–100 percentiles towards higher wind speed values at the 850 hPa pressure level; for the purpose of our study, we interpreted this shift as an increase in wind intensity.

Compared to most of the other subregions, for Central Europe the majority of the results were reported for storm frequency and wind intensity. In each case, the eight studies we evaluated, strongly indicated an increase. Furthermore, three studies projected an increasing storm intensity [2,28,70], and two studies suggested an increasing storm activity [50,63]. Results from [28] suggested an increase in wind-induced damage in the future. The study of [74] gave an overview of the future development of storminess over Germany. They used one GCM to drive three RCMs which projected gust speed as an indicator for storminess at a rather high horizontal resolution (<20 km). Their results basically implied a slight decrease of gust wind speed over Central and Southern Germany but suggested an increase of gust wind speed over Northern Germany.

The subregion-related total ES-score amounts to 151, $\overline{ES}$ = 4.7. The eT-value of 0.95 is the second highest eT-value of our evaluation and clearly indicates an increasing tendency of the future subregion-related storminess.

3.5. Southern Europe

The findings of the eleven studies for Southern Europe (Table 5) indicated, without exception, decreasing long-term behavior of storminess, including the two studies [51,64] for which we rated ES = 9. While the majority of the studies addressed projected future changes in wind intensity, three studies presented results on the expected future development of the frequency of extreme cyclones [31,69,77]. For all other assessed categories, a maximum of two studies was found that reported on projected aspects of storminess. Based on our evaluation scheme, the subregion-related total ES-score is 63, the $\overline{ES}$-value is 5.3, and eT = −1.00, which unambiguously suggests a decreasing tendency of the projected future extratropical storminess.

Table 5. Same as Table 1 but for the Southern Europe subregion.

Study	Model(s)	S	N	ES	T	Category	Reference Period(s),	Region(s)
							Projection Period(s),
							Month(s), Season
Zappa et al. (2013) [51]	GCM: 19	3	3	9	−1	Cyclone intensity	1976–2005,	Mediterranean region
	RCM: -						2070–2099,
							DJF
Gastineau and Soden (2009) [64]	GCM: 16	2	3	8	−1	Wind intensity	1995–2000,	Southern Europe
	RCM: -						2095–2100,
							-
Donat et al. (2011) [17]	GCM: 7	2	2	6	−1	Wind intensity	1960–2000,	Mediterranean region
	RCM: 9						2021–2050, 2071–2100,
							October–March
Nikulin et al. (2011) [66]	GCM: 6	2	2	6	−1	Wind intensity	1961–1990,	Europe south of 45°N
	RCM: 1						2071–2100,
							-
Schwierz et al. (2010) [27]	GCM: 2	2	2	6	−1	Wind intensity	1961–1990,	Mediterranean region
	RCM: 2						2071–2100,
							October–March
Bengtsson et al. (2006) [9]	GCM: 1	2	1	4	−1	Cyclone intensity	1961–1990,	Mediterranean region (30°N–45°N, 0°E–40°E)
	RCM: -						2071–2100,
							DJF
Bengtsson et al. (2009) [31]	GCM: 1	2	1	4	−1, −1	Frequency of extreme cyclones, cyclone intensity	1959–1990,	Southern Europe (30°N–47.5°N, 10°W–40°E)
	RCM: -						2069–2100,
							DJF
Beniston et al. (2007) [70]	GCM: 1	2	1	4	−1	Storm intensity	1961–1990,	Southern Europe (Alps and south of the Alps)
	RCM: 7						2071–2100,
							DJF
Fink et al. (2009) [38]	GCM: 1	2	1	4	−1	Storm frequency	1970–1999,	Mediterranean region
	RCM: 2
							2070–2099,
							winter
Giorgi et al. (2004) [63]	GCM: 1	2	1	4	−1	Storm activity	1961–1990,	Southern Europe
	RCM: 1						2071–2100,
							DJF
Muskulus and Jacob (2005) [77]	GCM: 1	2	1	4	−1	Frequency of extreme cyclones	1961–2099,	Mediterranean region
	RCM: 1						winter

3.6. Western Europe

The 33 studies available for this subregion (Table 6, [2,5,9,13,17,19,26,27,28,30,37,45,48,49,50,51,54,55,56,57,58,60,61,62,67,70,74,75,76,78,79,80,81]) allowed for the evaluation of all categories of aspects of projected future storminess. The majority of the evaluated studies projected an increase in frequency, intensity, and activity of storms and cyclones, as well as an increase in frequency and intensity of high-impact wind speed.

Table 6. Same as Table 1 but for the Western Europe subregion.

Study	Model(s)	S	N	ES	T	Category	Reference Period(s),	Region(s)
							Projection Period(s),
							Month(s), Season
Mizuta (2012) [78]	GCM: 11	3	3	9	1	Frequency of extreme cyclones	1979–2003,	British Isles
	RCM: -						2075–2099,
							DJF
de Winter et al. (2012) [48]	GCM: 12	3	3	9	0	Wind intensity	1950–2000,	North Sea
	RCM: -						2050–2100,
							-
Zappa et al. (2013) [51]	GCM: 19	3	3	9	1	Frequency of extreme cyclones	1976–2005,	United Kingdom
	RCM: -						2070–2099,
							DJF
Harvey et al. (2012) [50]	GCM: 19	2	3	8	1	Storm activity	1976–2005,	British Isles
	RCM: -						2070–2099,
							DJF
Ulbrich et al. (2008) [56]	GCM: 15	2	3	8	1	Cyclone activity	1960–2000,	Western Europe
	RCM: -						2081–2100,
							winter
Barstad et al. (2012) [19]	GCM: 4	2	2	6	−1	Wind intensity	1971–2001,	North Sea
	RCM: -						2020–2049,
							-
Donat et al. (2010) [2]	GCM: 7	2	2	6	+1, +1	Storm frequency, storm intensity	Various model–	Northeast Atlantic, British Isles, North Sea
	RCM: -						dependent periods,
							October–March
Donat et al. (2011) [17]	GCM: 7	2	2	6	1	Wind intensity	1960–2000,	Western Europe, North Sea region, British Isles
	RCM: 9						2021–2050, 2071–2100,
							October–March
Leckebusch et al. (2006) [13]	GCM: 4	2	2	6	+1, +1, +1	Frequency of extreme cyclones, frequency of extreme wind speed, wind intensity	1961–1989,	British Isles
	RCM: 4						2071–2099,
							October–March
Leckebusch et al. (2008) [57]	GCM: 7	2	2	6	1	Storm frequency	1961–2000,	Western Central Europe
	RCM: -						2081–2100,
							winter
Schwierz et al. (2010) [27]	GCM: 2	2	2	6	1	Frequency of extreme wind speed	1961–1990,	Northern Great Britain, English Channel, North Sea, Netherlands
	RCM: 2						2071–2100,
							October–March
Bengtsson et al. (2006) [9]	GCM: 1	2	1	4	+1, +1	Frequency of extreme cyclones, cyclone intensity	1961–1990,	British Isles
	RCM: -						2071–2100,
							DJF
Beniston et al. (2007) [70]	GCM: 1	2	1	4	1	Storm frequency	1961–1990,	North Sea coast
	RCM: 7						2071–2100,
							DJF
Della-Marta and Pinto (2009) [26]	GCM: 1	2	1	4	1	Frequency of extreme cyclones	1960–2000,	British Isles, North Sea/Western Europe (45°N–60°N, 10°W–30°E)
	RCM: -						2060–2100,
							October–March
Hanson et al. (2003) [45]	GCM: 1	2	1	4	0	Cyclone intensity	1961–1990,	United Kingdom
	RCM: 1						2070–2099,
							October–March
Karremann et al. (2014) [30]	GCM: 1	2	1	4	1	Storm frequency	1960–2000,	France, Netherlands, British Isles
	RCM: -						2060–2100,
							October–March
Leckebusch and Ulbrich (2004) [54]	GCM: 1	2	1	4	+1, +1, +1	Storm frequency, frequency of extreme cyclones, cyclone inten- sity	1960–1989,	Western Europe, Western Central Europe, British Isles
	RCM: 1						2070–2099,
							October–March
Leckebusch et al. (2008) [58]	GCM: 1	2	1	4	1	Storm intensity	1971–2000,	Western Central Europe
	RCM: -						2071–2100,
							October–March
McDonald (2011) [60]	GCM: 1	2	1	4	1	Storm frequency	1960–1990,	North Sea, British Isles
	RCM: -						2070–2100,
							DJF
Mizuta et al. (2011) [67]	GCM: 1	2	1	4	+1, +1	Frequency of extreme cyclones, frequency of extreme wind speed	1979–2003,	Scotland, North Sea
	RCM: -						2075–2099,
							DJF
Pinto et al. (2007) [37]	GCM: 1	2	1	4	+1, +1, +1	Storm frequency, frequency of extreme cyclones, wind intensity	1960–2000,	Western Europe, North Sea
	RCM: -						2060–2100,
							October–March
Pinto et al. (2007) [61]	GCM: 1	2	1	4	+1, +1	Frequency of extreme cyclones, wind intensity	1960–2000,	British Isles, Western Europe
	RCM: -						2060–2100,
							October–March
Pinto et al. (2009) [55]	GCM: 1	2	1	4	+1, +1	Frequency of extreme cyclones, cyclone intensity	1960–2000,	British Isles and environs, North Sea
	RCM: -						2060–2100,
							October–March
Pinto et al. (2012) [28]	GCM: 1	2	1	4	1	Storm intensity	1960–2000,	Belgium, France, Great Britain, Ireland, Netherlands
	RCM: -						2060–2100,
							October–March
Rauthe et al. (2010) [74]	GCM: 1	2	1	4	1	Wind intensity	1971–2000,	Netherlands, Northwestern North Sea
	RCM: 3						2021–2050,
							winter
Rockel and Woth (2007) [75]	GCM: 1	2	1	4	+1, +1	Frequency of extreme wind speed, wind intensity	1961–1990,	British Isles
	RCM: 8						2071–2100,
							-
Sterl et al. (2009) [79]	GCM: 1	2	1	4	1	Wind intensity	1950–2000,	Southern North Sea (53°N–55°N)
	RCM: -						2050–2100,
							-
Beersma et al. (1997) [62]	GCM: 1	1	1	3	1	Storm frequency	1980s,	North Sea
	RCM: -						2040s,
							-
Carnell et al. (1996) [80]	GCM: 1	1	1	3	+1, +1	Cyclone intensity, frequency of extreme wind speed	1983–1992,	British Isles
	RCM: -						-,
							DJF
Leckebusch et al. (2008) [76]	GCM: 1	1	1	3	1	Frequency of extreme wind speed	1880–1930,	British Isles
	RCM: -						2039–2089,
							October–March
Lozano et al. (2004) [5]	GCM: 1	1	1	3	+1, −1	Storm frequency, cyclone intensity	1970–1999,	Coastal zone of Ireland and Southern Scotland
	RCM: -						2060–2089,
							winter
Ulbrich and Christoph (1999) [81]	GCM: 1	1	1	3	+1	Cyclone activity	1860–2100,	Northwestern Europe
	RCM: -						DJF
Walter et al. (2006) [49]	GCM: -	-	1	2	1	Wind intensity	1969–1989,	North Sea, English Channel
	RCM: 4						2070–2099,
							February

Results of [50] obtained from an ensemble of 19 CMIP3-models, implied an increase in storm activity. From the results of [51,78], an increasing frequency of extreme cyclones can be inferred, while another study [56] reported an increase in cyclone activity. The results of the study of [5] showed an increase in cyclone intensity and at the same time a decreasing storm frequency.

Two of the evaluated studies also reported on the future development of storm damage. While the results from [17] indicated an increase of wind-induced damage, the study of [28] suggested a decrease in projected wind-induced damage. In two further studies, no change in analyzed long-term development of storminess was found: One study [48] did not identify a change in expected wind intensity over the North Sea, and another study [45] reported no change in cyclone intensity over the UK.

The subregion-related total ES-score is 221, the $\overline{ES}$-value equals 4.7, and the eT-value is 0.84; thus, we strongly assume that projected future storminess will increase in this subregion.

3.7. Eastern Europe

Altogether we found five studies for the Eastern Europe subregion (Table 7). Therefore, the results of our evaluation in this subregion are probably the least robust in comparison to the other subregions, not only because of the low number of available studies but also because of the limited (according to our evaluation scheme) methodological complexity of the evaluated studies. Among the studies available for evaluation, there was no study comprising RCP-scenarios and/or multi-model ensembles. Only two studies used more than one GCM [17,53]. Their results suggested an increase in future wind intensity. While one study [17] investigated the change of the ensemble mean of extreme wind speed, another study investigated the strength of extreme wind speed events with a return period of ten years [53].

Table 7. Same as Table 1 but for the Eastern Europe subregion.

Study	Model(s)	S	N	ES	T	Category	Reference Period(s),	Region(s)
							Projection Period(s),
							Month(s), Season
Donat et al. (2011) [17]	GCM: 7	2	2	6	1	Wind intensity	1960–2000,	Eastern Europe
	RCM: -						2021–2050, 2071–2100,
							October–March
Gregow et al. (2012) [53]	GCM: 9	2	2	6	1	Wind intensity	1971–2000,	Eastern Europe (20°E–40°E), Northern Europe
	RCM: -						2046–2065, 2081–2100,
							September–April
Lionello et al. (2008) [59]	GCM: 1	2	1	4	−1, −1	Frequency of extreme cyclones, cyclone intensity	1961–1990,	Russia
	RCM: 1						2071–2100,
							winter
Mizuta et al. (2011) [67]	GCM: 1	2	1	4	+1, +1	Frequency of extreme cyclones, frequency of extreme wind speed	1979–2003,	Parts of Russia
	RCM: -						2075–2099,
							DJF
Rockel and Woth (2007) [75]	GCM: 1	2	1	4	+1, +1	Frequency of extreme wind speed, wind intensity	1961–1990,	Eastern Europe
	RCM: 8						2071–2100,
							-

Overall, results concerning the future development of storminess over Eastern Europe are inconclusive. While findings from the study of [59] implied a decrease for Russia, another study projected an increasing number of cyclones over parts of Russia [67]. The subregion-specific total ES-score amounts to 36, $\overline{ES}$ = 4.5, and eT = 0.56.

3.8. Changes in the North Atlantic Oscillation and Storm Tracks

The studies summarized in Table 8 [9,25,31,36,38,48,50,51,52,54,56,59,60,61,63,64,65,69,80,81,82,83,84,85,86,87] report on projections of future NAO- and storm track characteristics over the NAER. The reported results are quite diverse. Sixteen out of the 26 assessed studies indicated that future storm tracks will at least partly shift either northward, northeastward, eastward, or poleward. However, findings from ten studies implied storm tracks shifting south- or southeastward. Furthermore, there were five studies that suggested no shifts in the long-term development of storm tracks. With respect to NAO-characteristics, the results from three of the studies indicated a northeastward shift of the NAO in the future. The NAO- and storm track-related total ES-score amounts to 129, and $\overline{ES}$ equals 5.0.

Table 8. Studies reporting on projections of future NAO- and/or storm track characteristics over the NAER.

Study	S	N	ES	Change in NAO	Change(s) in Storm Track
Harvey et al. (2012) [50]	3	3	9	-	No change
de Winter et al. (2012) [48]	3	3	9	-	Northward
Zappa et al. (2013) [51]	3	3	9	-	Eastward extension
Gastineau and Soden (2009) [64]	2	3	8	-	Northward
Lambert and Fyfe (2006) [82]	2	3	8	-	No change
Ulbrich et al. (2008) [56]	2	3	8	-	Slightly southward
Yin (2005) [83]	2	3	8	-	North- and eastward
Löptien et al. (2008) [65]	2	2	6	-	No change
Bengtsson et al. (2006) [9]	2	1	4	-	Northward
Bengtsson et al. (2009) [31]	2	1	4	-	Northward
Fink et al. (2009) [38]	2	1	4	Northeastward	-
Fischer-Bruns et al. (2005) [36]	2	1	4	-	Northward
Giorgi et al. (2004) [63]	2	1	4	-	Northward (JJA)
Laine (2009) [84]	2	1	4	-	Poleward and upward
Leckebusch and Ulbrich (2004) [54]	2	1	4	-	Southward
Lionello et al. (2008) [59]	2	1	4	-	Northward
McDonald (2011) [60]	2	1	4	-	Southward in eastern North Atlantic otherwise northward
Pinto et al. (2007) [61]	2	1	4	Northeastward	Eastward
Carnell et al. (1996) [80]	1	1	3	-	Northward
Catto et al. (2011) [85]	1	1	3	-	Northeastward (in 2 CO2-experiments), no change (in 4 CO2-experiments)
Geng and Sugi (2003) [52]	1	1	3	-	Southeastward
Knippertz et al. (2000) [69]	1	1	3	-	Northeastward
Lambert (1995) [86]	1	1	3	-	No change
Pinto et al. (2006) [25]	1	1	3	-	Northward
Schubert et al. (1998) [87]	1	1	3	-	Eastward (in 2 CO2-experiments), southeastward (in 3 CO2-experiments)
Ulbrich and Christoph (1999) [81]	1	1	3	Northeastward	-

4. Summary and Conclusions

Our research on the current literature yielded a total of 58 studies modeling aspects of the projected future storminess over the NAER. Through a categorical evaluation scheme based on the ranking of emission scenarios and modeling complexity, we attempted a joint evaluation of the methodologically highly diverse studies. The intention was to disclose to which degree the studies yielded either shared or differing general tendencies in the projected future development of storminess. In our evaluation, we did not attempt to treat the NAER as a whole but differentiated for seven subregions for the following two reasons: On the one hand, the spatial reference of the evaluated studies allowed for a differentiation, as they all clearly referred to subregions within the NAER. On the other hand, it is highly improbable that aspects of storminess will show consistent development throughout the NAER in the future.

Not unexpectedly, a heterogeneous picture of the projected future storminess emerges from the evaluation. From a total of 156 ratings, (i) 76% suggest an increase; (ii) 21% indicate a decrease; and (iii) 3% of the ratings do not indicate any change in the projected aspects of storminess. With respect to the assessed categories of storminess (Table 9), most of the ratings refer to results from studies addressing future changes in the intensity of high-impact wind speed (43 ratings) and frequency of extreme cyclones (35 ratings).

Table 9. Number of ratings assigned to categories of aspects of projected future storminess over the NAER.

Category	Number of Ratings
Wind intensity	43
Frequency of extreme cyclones	35
Storm frequency	23
Cyclone intensity	20
Frequency of extreme wind speed	14
Other categories	21

Despite the noteworthy differences in the number and quality of studies available for the evaluated subregions, it is apparent that the projected future development of storminess is quite heterogeneous and not consistent in the NAER. However, within the subregions, results on projected aspects of future storminess are clearer. For the three subregions Central Europe, North Atlantic south of 60°N, and Western Europe, the evaluated studies clearly indicate (>90% of the ratings) increasing tendencies for projected aspects of future storminess (Figure 2). For Eastern and Northern Europe, the majority of ratings (75% and 74%) also indicates increasing tendencies in projected aspects of future storminess; however, the studies available for these two subregions also indicate, in more than 20% of the ratings, decreasing tendencies of aspects of future storminess. Predominantly decreasing tendencies of aspects of storminess are indicated for subregions North Atlantic north of 60°N (85% of the ratings) and Southern Europe (100% of the ratings).

Figure 2. Relative proportions of ratings indicating either increasing (R_inc) or decreasing (R_dec) tendencies of aspects of projected future storminess, subregion-related eT-, and $\overline{ES}$-values for the seven subregions of the NAER. The $\overline{ES}$ -values were divided by 10 to adapt them to the scaling. The height of the columns corresponds to the numbers given in the subregion-related text sections.

The relative distribution of the subregion-related ratings is mirrored by the summary of the subregion-related eT-values . While eT-values computed for Central Europe, the North Atlantic region south of 60°N and Western Europe indicate clear tendencies for an increasing storminess in these subregions, eT-values computed for the North Atlantic region north of 60°N and the Southern Europe region strongly suggest decreasing tendencies in storminess. Based on the eT-values, we deemed the results on the projected future development of storminess over Eastern and Northern Europe as inconclusive. However, we believe that, in particular, the interpretation of our findings for Eastern Europe and Southern Europe is limited for two reasons: For Eastern Europe, there were only few studies available, and the evaluation of studies available for the Southern European subregion was restricted to the projected development of extratropical storms and cyclones. In the evaluation, we did not address the projected future development of tropical storms and cyclones, which might be of importance for the long-term evolution of storminess in this particular subregion. For example, results from [59] implied an increase in wind intensity associated with tropical cyclones, while there was indication that wind intensity associated with extratropical cyclones would simultaneously decrease. In addition, there is an indication that tropical cyclones might transform into extratropical cyclones as they move northward. Since tropical cyclones are expected to intensify in a warmer climate, this could also boost extratropical storms in the summer and autumn [31].

Across the evaluated subregions, the $\overline{ES}$-values vary between 4.5 and 5.3, which hints at a nearly balanced mean weighting of the seven analyzed subregions.

The results from the majority of studies that reported on the future development of storm tracks over the NAER point in a similar direction, as they mostly suggested a north- and eastward shift of storm tracks. Regardless of the results reported for the Northern Europe subregion, based on the storm track-related findings, it must be concluded that Northern Europe is probably the most strongly subject to potential increases in aspects of future storminess.

The comparison of the results presented in this review with results from previous reviews [11,22,23,24] on the storminess over NAER is limited. This is mainly due to methodological reasons. We did not only use a different evaluation scheme, but we have also assessed changes in categories of aspects of projected storminess in subregions of the NAER which are based on a pool of studies that differed from the pools of studies included in previous reviews.