Extratropical cyclones are important phenomena to regulate the daily weather in mid-latitudes and to transport energy and moisture in the atmosphere. They are mainly driven by baroclinicity , which is characterized by a strong temperature difference between the poles and Equator. They are associated with precipitation and strong winds, which can lead to high impacts in society e.g., for example floods, power outages, and forest and property damage. Since extratropical cyclones play a crucial role in the atmospheric circulation and weather variability and as a cause for societal and economical impacts, the knowledge and understanding of extratropical cyclone climate and dynamics are essential.

Extratropical cyclones and their tracks have been widely studied in the Northern Hemisphere e.g.. It is well known that the main storm track regions in the Northern Hemisphere are located over the North Atlantic and the North Pacific, with secondary storm track regions over the Mediterranean and Siberia e.g.. The North Atlantic storm track region starts east of the Rocky Mountains in North America and ends in northern Europe e.g.. In this study, we consider northern Europe to be a region covering Norway; Sweden; Finland; Estonia; and parts of the Baltic, Norwegian, and Barents seas. In general, extratropical cyclones at the end of all storm track regions are less studied. However, the structure and evolution of such extratropical cyclones may differ from extratropical cyclones elsewhere. noted that extratropical cyclones at the end of the North Atlantic storm track generally resemble more of the Norwegian cyclone model than the Shapiro–Keyser cyclone model , which is reasonable since the Norwegian cyclone model was developed based on cyclones that occur in north-western Europe. found that there are many differences in the dynamical and thermal structure and evolution of strong explosive cyclones in the north-eastern Atlantic compared to the north-western Atlantic. For example, the explosive cyclones in the north-eastern Atlantic have lower static stability but less environmental baroclinicity than explosive cyclones in the north-western Atlantic .

There are two main ways to identify extratropical cyclones: the Eulerian approach and the Lagrangian approach . The Eulerian approach uses basic statistics such as the variance or standard deviation of a chosen meteorological field (for example mean sea level pressure (MSLP) or 850 hPa relative vorticity), which is bandpass-filtered with synoptic timescales, usually 2–6 d . While the Eulerian approach is straightforward and gives a general description of the storm track activity it does not provide information about individual extratropical cyclones and their characteristics. The Lagrangian approach is a feature tracking method where a localized minimum or maximum of a meteorological field (for example MSLP or 850 hPa relative vorticity) is identified, and their location in time is tracked e.g.. With the Lagrangian approach it is possible to analyse cyclone-specific characteristics such as the genesis and lifetime. Moreover, additional meteorological variables can be determined along the cyclone track to investigate how they vary during the cyclone evolution.

The spatial and temporal evolution of extratropical cyclones can be analysed by producing cyclone composites. Cyclone compositing allows us to statistically find key features of the structure and development of the chosen set of cyclones. Cyclone composites have been used in extratropical cyclone research for example for creating an extratropical cyclone atlas for the 200 most intense historical extratropical cyclones in the North Atlantic and to investigate how moisture is transported into extratropical cyclones . Commonly cyclone composites are used to study the structures in extratropical cyclones, for example for precipitation and clouds and for air–sea turbulent heat fluxes and heat and moisture content . Cyclone composites have also been used to examine extratropical cyclones and their structures in climate models and in warming climate .

Since extratropical cyclones are responsible for the daily weather, and their associated heavy rainfall and strong winds can cause damage, it is important that extratropical cyclones and windstorms are accurately predicted. In this study, we define a windstorm to be an extratropical cyclone that has exceptionally strong 10 m winds (see Sect.  for a full definition). In addition to producing forecasts directly for the parameters of extratropical cyclone intensity (for example MSLP or wind gust), we can use other variables as proxies to predict the windstorm intensity . The use of proxies in forecasting is especially helpful regarding parameters that are small in scale and therefore harder for numerical weather prediction models to accurately predict . For forecasting purposes, the proxies should be descriptive of the evolution of windstorms earlier than the most intense phase; i.e. they should act as precursors to windstorms. One way to find precursors to windstorms is to use ensemble sensitivity analysis (which is described in detail in Sect. ), which allows us to quantify how sensitive the subsequent windstorm intensity is to precursor fields. The ensemble sensitivity analysis method was examined by , who analysed the linear connection between perturbed initial conditions and forecast metrics of the wintertime flow pattern on the west coast of North America. The ensemble sensitivity method has also been used to find precursors to intense Mediterranean cyclones and to extratropical cyclones in the western and eastern North Atlantic . This method is a computationally simple way to identify sensitivities, but it has a limitation of assuming a linear relationship between the precursor and the characteristics of the windstorm being predicted. Nonetheless, the results can provide guidance for forecasters and indicate to which variables the most attention should be paid. From an application point of view, the sensitivities can also be used to assess, for example, likely error propagation or future climate projections. Since the resulting sensitivity shows how much, for example, the windstorm intensity changes if the precursor changes, a similar magnitude change in the precursor caused by an error in the numerical weather model would therefore result in a similar change in the windstorm intensity as the sensitivity results show. Likewise, future studies potentially could estimate how warmer climates would change the windstorm intensity by applying the sensitivity results to how the precursors are projected to change.

In this study, we aim to increase the knowledge of the extratropical cyclone and windstorm climate in northern Europe, which occurs at the end of the North Atlantic storm track. Although many extratropical cyclone studies focus only on the cold season or winter months e.g., we investigate the results separately for the cold (October–March) and warm (April–September) seasons in order to analyse the seasonal differences. In addition to the climatological properties as temporal frequencies and cyclone characteristics, we extend the study to examine what precursors are relevant in northern Europe windstorms by using ensemble sensitivity analysis. These results may help forecasters to better predict and prepare for windstorms by increasing their understanding of what environments windstorms typically develop in.

The research questions we aim to answer in this paper are as follows. (1) What are the annual and monthly frequencies of extratropical cyclones and windstorms in northern Europe? (2) Are there differences in cyclone characteristics between extratropical cyclones and windstorms and between the cold and warm seasons? (3) How does the spatial and temporal structure of northern Europe windstorms evolve? (4) What precursor has the strongest impact on the minimum MSLP and the maximum wind gust in northern Europe windstorms? The remainder of the paper is structured as follows: first, we introduce the data in Sect.  and the methods in Sect. . The annual and monthly frequencies of the three cyclone classes (all extratropical cyclones, windstorms, and non-windstorms) are shown in Sect.  , and different cyclone characteristics are analysed in Sect. . Section examines the spatial and temporal evolution of windstorms, and Sect.  investigates the precursors to windstorms. The final conclusions are given in Sect. .

ERA5 is the newest atmospheric reanalysis produced by the European Centre for Medium Range Weather Forecasts , and it is based on the Integrated Forecasting System (IFS; cycle 41r2). ERA5 has a horizontal resolution of 31 km, and vertically it has 137 levels from the surface up to approximately 80 km. ERA5 provides hourly analysis and forecast fields, and while conducting the analysis for this study, ERA5 covered years from 1979 onwards. Recently, there was a release of a back extension, making ERA5 available from 1950 onwards.

In our study, ERA5 data are obtained every 3 h from 1980–2019 (40 years). The variables we obtain are MSLP, the maximum 10 m wind gust (maximum since previous post-processing), 850 hPa temperature, total column water vapour (TCWV), 300 hPa horizontal wind speed components, and 300 hPa potential vorticity (PV). We consider 10 m wind gusts rather than 10 m wind speeds since the impacts and damage regarding extratropical cyclones are usually caused by the strong wind gusts .

The annual frequency of extratropical cyclones during 1980–2019 shows a large year-to-year variability in all three cyclone classes (Fig. ). There are on average 149 extratropical cyclones every year that reach northern Europe (Fig. a), of which on average 35 cyclones, i.e. 23 %, are windstorms (Fig. c). Although the threshold for windstorms is the 90th percentile of wind gusts, the wind gust distribution can have multiple wind gust values per cyclone since we include all wind gust values at each time step when the cyclone centre is inside the northern Europe box. Hence, the ratio of windstorms can be higher than 10 %. The 40-year linear trends are not statistically significant in any of the three cyclone classes by using the Mann–Kendall test at the 5 % level. However, there is a decreasing trend (-3.7 cyclones per decade) in all extratropical cyclones (Fig. a and b) with a p value of 0.067, and therefore the trend would be significant with a 7 % level or higher. However, some time periods stand out with high or low annual number of extratropical cyclones. In the 1980s, there were overall more extratropical cyclones compared to the 40-year average (Fig. b), which was mainly due to the increased numbers of non-windstorms (Fig. f). In contrast, a longer-term drop in all extratropical cyclones (Fig. b) occurred between 1995–2005, during which time the number of windstorms (Fig. d) decreased more than the number of non-windstorms. The highest annual numbers of windstorms occurred in 1983 and 1990 (Fig. c and d).

In addition to large inter-annual variability, we also investigated the seasonal cycle of the number of extratropical cyclones, windstorms, and non-windstorms (Fig. ). Regarding all extratropical cyclones in northern Europe (Fig. a), the months from October to January and April have the highest median, between 13–15 extratropical cyclones per month, which means that on average almost every other day in these months an extratropical cyclone travels through northern Europe. The variability in extratropical cyclone numbers within each month (length of whiskers) is the highest during autumn and winter. Nevertheless, the monthly variation in the number of all extratropical cyclones does not have a significantly large seasonal cycle. In contrast, the number of windstorms (Fig. b) shows a strong annual cycle, with more windstorms in the cold season and less in the warm season. The highest median values of 5–6 windstorms per month are in January and December, whereas in May, June, and July, there are on average no windstorms at all. The variation in windstorm numbers within the cold-season months is, however, quite large; for example in November, the number of windstorms ranges from 0 to 14. The seasonal cycle for non-windstorms (Fig. c) is the opposite of that for windstorms with the highest number of non-windstorms occurring during the warm season from April to August. Therefore, we can conclude that while windstorms are more common in winter, and extratropical cyclones in summer are weaker, the overall number of extratropical cyclones per month in northern Europe does not differ considerably depending on season.

The genesis and track densities are computed by the TRACK code using the spherical kernel method , and the unit of the densities is number of cyclones per season (i.e. 6 months) per unit area, where the unit area is a 5∘ radius spherical cap (≈ 106 km2). In both seasons, the genesis region of all extratropical cyclones that affect northern Europe extends from the east coast of the United States to western Russia and from the Greenland Sea to the Mediterranean Sea (Fig. a and b). Most of the extratropical cyclones in the warm season originate over the land areas inside the northern Europe box (Fig. b). In the cold season (Fig. a), the most dense genesis region is broader, and, in addition to the land areas in northern Europe, most of the extratropical cyclones originate from the Norwegian Sea and the Barents Sea. The genesis over the east coast of the United States is more poleward in the warm season than in the cold season, which is in agreement with , who found that extratropical cyclones occur more equatorward in the winter than in the summer in the eastern North Atlantic.

When considering windstorms and non-windstorms separately, there are distinct differences. Most of the non-windstorms originate over land areas in northern Europe (Fig. e and f), whereas windstorms have their genesis mostly over the Norwegian Sea and the Barents Sea (Fig. c and d). This also means that windstorms tend to originate farther away from northern Europe than non-windstorms. In addition, non-windstorms can originate over southern Europe while windstorms do not. In the cold season, the genesis region of windstorms is more zonally spread (Fig. c), whereas non-windstorms are more meridionally spread (Fig. e). In contrast, in the warm season the genesis region of non-windstorms is zonally spread and includes tracks from the eastern United States (Fig. f).

The track densities of all extratropical cyclones are, on a large scale, similar in both seasons, and the highest track density occurs inside the northern Europe box (Fig. a and b). Cold-season extratropical cyclones have high track densities over the eastern part of northern Europe and over the Barents Sea (Fig. a). There are somewhat fewer tracks over the Scandinavian Mountains, which is likely caused because the Scandinavian Mountains can, to some degree, modify the track of passing extratropical cyclones and even split the cyclone into two low centres . When comparing windstorms and non-windstorms, it is evident that most windstorms have their cyclone centre at higher latitudes than non-windstorms (Fig. c–f). The difference is more pronounced in the cold season, when windstorms mostly occur over the Barents Sea (Fig. c), while non-windstorms occur over the southern parts of northern Europe (Fig. e).

The majority of extratropical cyclones, regardless of the cyclone class (windstorm or non-windstorm) or season, propagate towards the east inside the northern Europe box (not shown). However, each year there is a small portion of extratropical cyclones that propagate towards the west, and this is slightly more visible in the warm season than in the cold season. This is also seen in the genesis density (Fig. ), where some extratropical cyclones originate on the eastern edge of the northern Europe box.

The median duration of the cyclone lifetime with all extratropical cyclones is 3.4 d in the cold season and 3.6 d in the warm season (not shown). Considering windstorms, the median lifetime is 4 d in both the cold and warm seasons, while non-windstorms have a longer median lifetime in the warm season (3.6 d) than in the cold season (3 d, not shown). Therefore, extratropical cyclones in northern Europe are on average longer-lived in the warm season compared to the cold season and in windstorms compared to non-windstorms. The lifetime of extratropical cyclones, especially in the extremely long-lasting cases, can also be interpreted from the width of the histograms in Fig. . In the warm season the histogram is wider (from -13 to 17 d; Fig. b), indicating that the lifetimes are longer than in the cold season (from -9 to 10 d; Fig. a).

Figure  shows the maximum 10 m wind gusts in northern Europe (i.e. the maximum 10 m wind gust when the cyclone centre is inside the northern Europe box) in relation to the time of the minimum MSLP. The highest wind gust values on the y axis are concentrated close to zero on the x axis, which means that the maximum wind gusts occur close to the time of the minimum MSLP. The median occurrence time of the maximum gusts relative to the time of the minimum MSLP is 0 h for all extratropical cyclones and non-windstorms and +3 h for windstorms. There is no differences between the seasons on the median occurrence times. Overall, the occurrence time of the wind gusts is practically very close to the minimum MSLP, and the difference between windstorms and non-windstorms is small. In the cold season, the maximum 10 m wind gusts are higher (up to 52 m s-1) than in the warm season (up to 38 m s-1), which was also seen in the wind gust distributions (Fig. ). Figure  also shows that while the wind gusts are the strongest at the time of the minimum MSLP, after that the wind gusts decrease. This happens especially in windstorms (Fig. c and d) where a negative tilt is visible after the time of the minimum MSLP. The negative tilt is not evident in non-windstorms (Fig. e and f), but non-windstorms by definition do not have strong wind gusts (there are no wind gust values above 27.2 m s-1).

The composites of 10 m wind gust, MSLP and 850 hPa potential temperature are shown in Fig.  for the cold season and in Fig.  for the warm season, and these fields allow us to determine the spatial structure of windstorms in northern Europe. The offset times are relative to the time of the minimum MSLP of each windstorm (defined in Sect. ). We investigate here the offset times at -48, -24, 0, and +24 h since Fig.  shows that these times cover the evolution before and after the strongest wind gusts and additionally include enough windstorms to ensure a reasonable number of cyclones in each composite; at the time of the minimum MSLP there are 1129 windstorms in the cold season and 245 windstorms in the warm season.

Already 48 h before the minimum MSLP (Figs. a and a), a closed low-pressure centre is evident in both seasons. The cold-season composite has a deeper central pressure (994 hPa) and a stronger pressure gradient than the warm-season composite (998 hPa). The frontal structure is visible from the 850 hPa potential temperature composite with a warm front downstream of the cyclone centre and a cold front upstream of the cyclone centre. The strongest wind gusts occur in the warm sector in the region of the strongest pressure gradient. In addition, the strongest wind gusts are located within a 6∘ radius from the cyclone centre, which supports the choice of selecting the maximum wind gust within 6∘. The strongest composite gusts are 17.2 m s-1 in the cold season and 14.2 m s-1 in the warm season. The region of the strongest gusts is more widely spread in the cold season than in the warm season. This indicates that the windstorms in the cold season are larger in spatial scale than in the warm season.

At 24 h before the minimum MSLP (Figs. b and b), the central pressure decreases, and the pressure and frontal gradients increase compared to the -48 h offset time. The wind gusts get stronger, while the maximum gust location spreads extending from the warm sector to the cold front. At the time of the minimum MSLP (Figs. c and c), the central pressure in the cold season is 972 hPa and in the warm season 981 hPa, and the pressure gradient is the strongest equatorward of the cyclone centre. The warm front starts to wrap around the cyclone centre, indicating that the windstorm is reaching its mature stage. The wind gusts are the strongest at this offset time, and the maximum gust values are 20.5 m s-1 in the cold season and 18.6 m s-1 in the warm season. The location of the maximum wind gusts occurs in a region extending from the warm sector to behind the cold front, where the pressure gradient is the strongest. The maximum wind gusts are still within a 6∘ radius from the cyclone centre. At 24 h after the minimum MSLP (Figs. d and d), the central pressure has risen and the frontal structure weakened. The maximum wind gusts have decreased and now occur behind the cold front.

To conclude, the location of the maximum wind gusts moves from the warm sector to behind the cold front following the location of the strongest pressure gradient during the evolution of the composite windstorm. The wind gusts, pressure gradient, and potential temperature gradient are all stronger and the minimum MSLP deeper in the cold season than in the warm season. In addition, the spatial scale of the windstorm is larger in the cold season than in the warm season.

In this section, we use the ensemble sensitivity analysis to investigate how two response functions, (1) the minimum MSLP and (2) the maximum 10 m wind gust, are sensitive to four different precursors at different offset times. The precursors we examine are 850 hPa potential temperature anomaly, TCWV, 300 hPa wind speed, and 300 hPa PV, and these are analysed 72, 48, and 24 h before the offset time.

This study investigated extratropical cyclones and windstorms in northern Europe and their characteristics, spatial and temporal evolution, and precursors. We tracked all extratropical cyclones in northern Europe during 1980–2019 and classified them into three classes: (1) all extratropical cyclones in northern Europe, (2) windstorms (i.e. extratropical cyclones which have strong wind gusts in northern Europe), and (3) non-windstorms (i.e. extratropical cyclones with no strong wind gusts in northern Europe). We investigated the cold season (October–March) and the warm season (April–September) separately to identify seasonal differences. We created cyclone composites to examine the cyclone evolution and used an ensemble sensitivity method to analyse precursors.

During the 40-year period, there is a large year-to-year variation in the annual number of extratropical cyclones, windstorms, and non-windstorms in northern Europe. The 40-year linear trends are not significant at the 95 % level; however, there is a decreasing trend in all extratropical cyclones (-3.7 cyclones per decade) that is significant at the 93 % level. reviewed the long-term trends of extratropical cyclone numbers over the North Atlantic and north-western Europe. They found that in their “Baltic Sea” region (covering the Baltic Sea; the Baltic states; and the major parts of Finland, Sweden, and Norway) there is no consensus on the long-term trends in reanalysis-, model- or, observation-based studies. Nonetheless, the sign and magnitude of linear trends depend strongly on the chosen time period and region, and therefore comparing different studies and results is difficult. In addition, it is important to note that a low number of windstorms in a specific year or decade does not mean that there could not be individual, strong windstorms. For example, our study shows that there were fewer windstorms than average between 1995–2005, while during this period a few extremely strong and damaging windstorms occurred in northern Europe: Storm Anatol in December 1999 and Storm Gudrun in January 2005 .

We found the well-known seasonality that windstorms are more common in winter, and extratropical cyclones in the warm season are weaker. However, a more unexpected result was that the overall number of extratropical cyclones per month in northern Europe does not differ much between months. This is contradictory to the general claim that there are less extratropical cyclones in the Northern Hemisphere during summer than winter (see for example the review by ). However, this seasonal difference is most pronounced in the core of the North Atlantic storm track, while at the end of the storm track, i.e. in northern Europe, the difference is smaller . If considering the track densities in northern Europe, show mainly no difference between winter and summer, and show only a small difference. Therefore, our finding does agree with other extratropical cyclone climatologies, but the other studies simply highlight more the general seasonality in the core of the storm track, although the seasonality has regional differences.

We investigated a set of cyclone characteristics to detect possible differences between windstorms and non-windstorms and between the cold and warm seasons. Windstorms tend to originate and occur over the sea areas (in the Norwegian and Barents seas), while non-windstorms originate and occur mostly over land in northern Europe. Partially, this is likely related to surface friction since we define windstorms based on their wind gusts, and winds are stronger over smooth surfaces, i.e. over sea areas. Moreover, the highest genesis densities of windstorms are co-located with the climatological left exit of the jet stream , which is a favourable location for cyclone development. Our result also means that windstorms occur at higher latitudes than non-windstorms. This agrees with our composites, which showed that windstorms are located on the poleward side of the jet stream in both seasons. We additionally found that the maximum wind gusts occur approximately at the same time as the minimum MSLP regardless of the cyclone class or season.

The maximum wind gusts in windstorms move from the warm sector to behind the cold front during the cyclone evolution following the strongest pressure gradient. This shift in the maximum gust location during the cyclone evolution is in agreement with the shift in the maximum 900 hPa wind speeds of the 200 strongest extratropical cyclones in an aqua-planet study . In addition, the spatial patterns of MSLP and 850 hPa potential temperature in our composites are similar to the spatial patterns found by when they created composites of the 100 strongest extratropical cyclones in the Northern Hemisphere. This suggests that northern Europe windstorms have similar spatial and temporal structures to those found elsewhere in the Northern Hemisphere. Our results show that the wind gusts, pressure gradient, and temperature gradient are stronger, and the minimum MSLP is deeper in the cold season than in the warm season. The cold-season windstorms have their associated wind gusts covering a larger area and thus are spatially larger than warm-season windstorms. This is relevant when informing and preparing for likely impacts, yet it is important to note that our study only includes large-scale extratropical cyclones, whereas other weather systems (e.g. thunderstorms) can also cause strong winds and impacts.

Lastly, we investigated precursors to northern Europe windstorms for the minimum MSLP and the maximum 10 m wind gust. The first main conclusion of our precursor results is that cold-season windstorms have higher sensitivities and thus are potentially easier to forecast than warm-season windstorms in terms of both the minimum MSLP and the maximum wind gusts. Possible reasons for this are a higher case-to-case variability and more rare occurrence of warm-season windstorms compared to cold-season windstorms. The second main conclusion is that the minimum MSLP of a windstorm has higher sensitivities than its associated maximum wind gust. This is possibly caused by the different nature of these variables: MSLP is a larger-scale field, while wind gusts are more localized and turbulent. The third and last main conclusion is that the best precursor for northern Europe windstorms is the 850 hPa potential temperature anomaly, i.e. the temperature gradient. This precursor outperformed the TCWV (moisture), the 300 hPa wind speed (jet stream), and the 300 hPa PV (upper-level trough and tropopause steepness). This means that the 850 hPa temperature gradient has the strongest impact on the resulting minimum MSLP and the maximum wind gusts, and thus the temperature gradient is an important variable to follow when forecasting windstorms in northern Europe. If we apply our sensitivity results to climate change, the low-level temperature gradient is expected to decrease , which based on our results would indicate weaker windstorms in terms of the minimum MSLP and the maximum wind gusts. However, this is counteracted by the atmospheric moisture, which is predicted to increase and hence would, based on our results from the TCWV sensitivities, indicate stronger windstorms.

The 300 hPa PV shows sensitivities to the downstream ridge, which is not found in , who used the ensemble sensitivity analysis to investigate precursors to extratropical cyclones in the west and east North Atlantic. This difference between our and their results may be because of the difference in the geographical location: blocking is more common over northern Europe and Russia than over the North Atlantic . The 300 hPa PV also gives weak or nonexistent sensitivities in the warm season. and developed a cyclone development classification based on the upper- and lower-level forcing mechanisms, and they identified three types of cyclones. Type A cyclones develop due to lower-level thermal advection in baroclinic regions without a pre-existing upper-level trough; type B cyclones develop due to upper-level vorticity advection, where a pre-existing upper-level trough moves over an area of warm advection; and type C cyclones are characterized by strong latent heat release with initiation of strong upper-level forcing and weak low-level baroclinicity . Therefore, our result may indicate that warm-season windstorms in northern Europe are type A or type C cyclones since they are not dominated by upper-level forcing. Our finding that the TCWV gives weak sensitivities in the warm season was surprising since typically there is more diabatic heating in the warm season than in the cold season. However, this may be a result of a low number of windstorms and a high case-to-case variability in the warm season. Our results also showed that the sensitivities are generally higher at -48 h than at -24 h. One possible explanation could be that 24 h is too short of a time for the atmospheric available potential energy to be converted to kinetic energy, an energy conversion known as the Lorenz energy cycle .

Our study has certain limitations. There are multiple ways to classify extratropical cyclones and to define windstorms , and therefore, our results may depend on our windstorm definition. Due to different extratropical cyclone and windstorm definitions, direct comparison of our climatology to previous studies is not simple. We also made a subjective choice about what geographic area to study, and our results are likely to be sensitive to our choice of the box over northern Europe. The area we considered included sea areas, and most windstorms we identified occurred over sea, making our results less applicable to windstorms over land areas. Furthermore, our analysis was based solely on ERA5 reanalysis, which may not accurately capture wind gusts due to the coarse resolution of the reanalysis (∼31 km) and also because wind gusts are parameterized, not directly resolved. Another limitation is that composites (and climatological studies overall) smooth the case-to-case variability, and therefore there is still need for case studies to investigate different evolution paths in more detail. For example, Storm Aila occurred in Finland in September 2020 and developed in the right entrance of the jet stream , which is contradictory to our result that northern Europe windstorms occur poleward of the jet stream. On the other hand, Storm Aila formed in an area of a strong temperature gradient , which agrees with our results.

Despite the limitations noted above, our composite and sensitivity results highlight the common features of windstorms in northern Europe. This enhances our dynamical understanding of these potentially damaging weather systems and furthermore provides valuable guidance in the form of data-based “conceptual models” to weather forecasters in the region.