1. Introduction
Sequences of intense extratropical cyclones have led to large socioeconomic damages in Europe in the last few decades. For example, the storm series of December 1999, including Anatol, Lothar, and Martin (Ulbrich et al. 2001), caused approximately €18.5 billion in economic damage over continental Europe (Munich Re 2002; Alert Air Worldwide 2009). The stormy winter of 2013/14 was the stormiest on record for the British Isles (Matthews et al. 2014) and was characterized by clustered periods, in particular in late December 2013 and mid-February 2014 (Priestley et al. 2017). Many of these storms brought a lot of precipitation to Europe and resulted in some notable flood events, for instance, in southern England (Kendon and McCarthy 2015; Schaller et al. 2016). A recent example occurred on 16–20 February 2022, when three major storms (Ylenia, Zeynep, and Antonia) hit the European continent and caused much destruction over large areas of northwestern and central Europe (Mühr et al. 2022). In this storm series, more than 20 storm-related deaths have been reported across Europe, and the insured wind losses from the first two winter storms range between €3 and €5 billion, primarily in Germany, the United Kingdom, and the Netherlands (Alert Air Worldwide 2022a).
The concept of cyclone series (also known as cyclone sequences, cyclone families, temporal clustering, or temporally compounding cyclones) is not new (Schultz et al. 2019). For example, Bjerknes and Solberg (1922) had already recognized the importance of cyclone families about a century ago, and described their synoptic characteristics. In the present study, we focus on temporal clustering of extratropical cyclones (i.e., sequences of cyclones affecting a certain area in a specific time window). Following Dacre and Pinto (2020), there are several methods for quantifying clustering. The first corresponds to absolute metrics based on synoptic knowledge, such as defining a threshold of 4 or more cyclones over a certain location within a week as cyclone clustering (Pinto et al. 2014). The second option is to consider relative metrics, where clustering is defined as a positive deviation from a stochastic process, where more cyclones occur than expected by chance, such as by using a dispersion statistic (Mailier et al. 2006) and nonexponential return times (Blender et al. 2015). The last possibility is to assume impact metrics and determine the relationship between the highest single loss per year [occurrence exceedance probability (OEP)] and the cumulative loss per year [annual exceedance probability (AEP)] (Priestley et al. 2018).
For the North Atlantic basin, there is a large consensus that clustered cyclone activity is found primarily on the flanks and exit region of the North Atlantic storm track, including the United Kingdom, the Benelux countries, France, Germany, and occasionally Scandinavia, while regions like the entrance of the North Atlantic storm track are characterized by independent cyclone occurrence (Dacre and Pinto 2020). There is evidence that clustering increases for more intense cyclones (Vitolo et al. 2009; Pinto et al. 2013; Economou et al. 2015). While fast-moving storms may produce strong wind gusts in a short period of time, slow-moving storms can be just as damaging, since these storms can accumulate huge amounts of precipitation during their development, and hence, result in severe flooding over the same location (e.g., winter 2013/14 in the United Kingdom; Priestley et al. 2017). Accordingly, the clustering of extratropical cyclones may be stronger for more extreme storms, and countries further away from the storm track are likely to be more affected than others (Cusack 2016). Other regions such as the northeast Pacific, including the west coast states of the United States and the western Canadian provinces, are also regions affected by clustering due to the northeastern Pacific storm track (Mesquita et al. 2010). Composite analyses show that the serial clustering of Rossby waves and extratropical cyclones near the U.S. west coast are supporting water vapor flux and therefore prolonging heavy precipitation events (Moore et al. 2021). This region also experiences extreme precipitation from successive atmospheric river events (Fish et al. 2022). As strong cyclones often pass across the Aleutian Island chain, the Gulf of Alaska is another important storm-prone region, which might provide information on the future changes in cyclone clustering (Mesquita et al. 2010). Still, there is a large knowledge gap in studies assessing cyclone clustering outside of the North Atlantic basin. In fact, only two studies have investigated cyclone clustering in the North Pacific in climatological terms (Mailier et al. 2006; Blender et al. 2015). While both studies agree that the preferred regions for cyclone clustering are the Gulf of Alaska and the west coast of Canada and the United States, large uncertainties remain.
Concerning historical periods (e.g., the last 40 years) clustered extratropical cyclone activity has been examined only a few times and exclusively in the ERA-Interim reanalysis or older reanalyses (Mailier et al. 2006; Pinto et al. 2013; Economou et al. 2015; Priestley et al. 2018). The ERA5 reanalysis (Hersbach et al. 2020) has recently been made available, and features both higher spatial and time resolution and an improved data assimilation methodology. Thus, ERA5 is one of the best reanalysis products suitable for reliable cyclone tracking and storm analyses, for example, as in Karwat et al. (2022).
Another important question is how extratropical cyclone clustering will change in a future climate under different greenhouse gas emissions scenarios. In phase 5 of the Coupled Model Intercomparison Project (CMIP5), future changes were found to be minor and inconsistent between models, whereas for more extreme storms, clustering increases over northern Europe and Scandinavia in the RCP4.5 scenario (Economou et al. 2015). This is in line with Pinto et al. (2013), who focused on a single GCM large ensemble with 20 members. Other studies find shorter return times for severe winter storms and related losses for western Europe in the future (Della-Marta and Pinto 2009; Pinto et al. 2012; Karremann et al. 2014). Therefore, the number of (clustered) cyclones is expected to increase, while overall the total number of all cyclones is projected to decrease in the future (e.g., Bengtsson et al. 2006; Pinto et al. 2009; Ulbrich et al. 2009; Zappa et al. 2013; Priestley and Catto 2022). However, many of these studies that address clustering focus on Europe.
Large ensembles have become more important over the last few years due to their ability to better quantify uncertainty by increasing the sample size, which is important for analyzing rare events such as clustering of storms and extreme events. Overall, they can improve the accuracy, robustness, and reliability of cyclone clustering projections, making them a valuable tool. The 100-member CESM2 Large Ensemble (CESM2-LE) (Rodgers et al. 2021) provides thus an excellent opportunity to investigate how cyclone clustering may change from preindustrial times (e.g., 1850–90) until the end of the twenty-first century in many different realizations, which are essential to quantify rare events, such as cyclone clustering. To our best knowledge, it is currently the largest existing ensemble with a sufficiently high spatial and temporal resolution to examine cyclone clustering. There is also a substantial lack of available large ensembles of future climate simulations from CMIP6 models with sufficient temporal and spatial high resolution to perform cyclone tracking from 2050 onward, and a general lack of studies assessing all ocean basins except the North Atlantic (Dacre and Pinto 2020). Moreover, this analysis will provide us with more insights into clustering from 1850 to 2100 focusing on both the North Pacific and the North Atlantic basins. Therefore, the specific aims of this paper can be summarized by the following three questions:
-
How is clustered cyclone activity in the Northern Hemisphere midlatitudes characterized in the ERA5 reanalysis and CESM2-LE in 1980–2020?
-
How is extratropical cyclone clustering projected to change by 2060–2100 compared to 1980–2020?
-
Can we identify reasons in terms of large-scale dynamics and cyclone characteristics for these projected changes?
The remainder of this paper is organized as follows: in section 2, we introduce the ERA5 and the CESM2-LE data and the methodology; in section 3, we analyze the cyclone statistics and mean sea level pressure (mslp) fields for recent and future climate conditions. Section 4 examines clustering based on absolute measures. Possible links to large-scale patterns and cyclone characteristics are explored in section 5. Finally, we discuss the conclusions of this study in section 6.
2. Data and methods
a. ERA5 reanalysis data
We use the European Centre for Medium-Range Weather Forecasts reanalysis (ERA5) data for the winter months December–February (DJF) from 1980/81 to 2019/20 (Hersbach et al. 2020). ERA5 has a horizontal resolution of 0.25°. For the Lagrangian cyclone tracking we interpolate the mean sea level pressure (mslp) fields on a 2° × 2° grid and use 6-hourly data. We analyze clustered cyclones in the North Atlantic and North Pacific sectors. Cyclones are tracked over the North Atlantic and Europe (80°W–30°E and 25°–75°N) and in the North Pacific (120°E–120°W and 25°–65°N). We consider the areas of Europe (10°W–15°E and 50°–60°N), the west coast of the U.S. states and western Canadian provinces (120°–135°W and 30°–60°N), and the Gulf of Alaska (135°–165°W and 50°–60°N) (Fig. 1).
Map of the cyclone cluster areas in this study: (a) northwest Europe (10°W–15°E and 50°–60°N), (b) the west coast of the U.S. states and western Canadian provinces (120°–135°W and 30°–60°N), and (c) the Gulf of Alaska (135°–165°W and 50°–60°N). Map credit: own creation.
Citation: Journal of Climate 37, 4; 10.1175/JCLI-D-23-0160.1
b. CESM2 Large Ensemble
We use model simulations with the Community Earth System Model, version 2 (CESM2) (Danabasoglu et al. 2020). The large ensemble (CESM2-LE) is documented in Rodgers et al. (2021). The CESM2-LE has a horizontal resolution of 1° and is forced by the CMIP6 historical forcing from 1850 to 2014 and the Shared Socioeconomic Pathways forcing scenario (SSP3-7.0) from 2015 to 2100. The SSP3-7.0 scenario relates to an additional radiative forcing of 7 W m−2 by 2100 (IPCC 2021). On this SSP, temperatures rise steadily and CO2 emissions roughly double from current levels by 2100. By the end of the century, global average temperatures have risen by approximately 3.6°C compared to preindustrial temperatures of 1850–1900 (IPCC 2021). This scenario is in the upper-middle part of all scenarios in CMIP6. It was introduced to close the gap between the RCP6.0 and RCP8.5 scenarios in CMIP5. We use the 50 of 100 CESM2-LE ensemble members that have 6-hourly mslp output fields chosen from a set of different combinations of atmosphere and ocean initial states (Rodgers et al. 2021). Overall, we focus on three different time slices: preindustrial times (1850–90), recent past (1980–2020), and future (2060–2100).
c. Cyclone tracking methodology
Several algorithms for identifying cyclone tracks have been developed, for example, by Murray and Simmonds (1991), Sinclair (1994), Hodges (1995), Wernli and Schwierz (2006). Here, we use the Lagrangian cyclone tracking algorithm of Blender et al. (1997). In the tracking algorithm cyclones are defined as local minima of the mslp field. The cyclone tracks are computed by a nearest-neighbor search in the mslp field where the trajectories of individual cyclones are connected at subsequent time steps. In our study, we only consider cyclones with a minimum lifetime of 48 h and which have traveled at least a distance of 1000 km. The cyclone statistics from this tracking method compare well with others (Neu et al. 2013), also in terms of metrics related with cyclone clustering (Pinto et al. 2016). We use the same conditions for the cyclone tracking for ERA5 and CESM2-LE.
d. Cyclone statistics and clustering metrics
We compare ERA5 and CESM2-LE to investigate if the overall number of cyclones per winter in CESM2-LE is comparable for recent climate conditions and if there are any significant changes between the different CESM2-LE time slices (1850–90, 1980–2020, and 2060–2100). In particular, we analyze the storm characteristics for ERA5 and CESM2-LE in 1980–2020. Therefore, we look at the ensemble spread (defined as twice the standard deviation of all ensemble members) of the cyclone counts, lifetime, mslp, radius, depth (difference between the cyclone center and the synoptic environment) (Schneidereit et al. 2010), and distance (overall length of the traveled cyclone path from beginning to end). Since our tracking depends on the mslp fields, any trends in mslp are likely to affect our analysis of clustering. Thus, we examine the differences in the mslp fields of the North Atlantic and North Pacific storm tracks in 2060–2100 compared to 1980–2020 and 1850–90, respectively. To quantify our results we use a Welch two-sample t test (Wilks 2011). We compare the local cyclone mean counts between the future winter climate and the current climate from the large ensemble. Here we apply a bootstrapping with resampling (n = 1000) (Wilks 2011).
To assess cyclone clustering, we use the number of cyclone occurrences per time interval, for example, 3–6 cyclones in 7 days [following, e.g., Pinto et al. (2014)]. For this metric we use 5° × 5° grid boxes [similar to Mailier et al. (2006)] in which we count the number of cyclone occurrences: cyclone clustering occurs when at least 3 cyclones in a 7-day period pass over the area. We then categorize the intensity of a cluster dependent on whether more than 3 cyclones (e.g., 4, 5, and 6 cyclones) occur in that week at the respective grid location. This approach and thresholds are used by insurance companies when considering extreme cyclones (Alert Air Worldwide 2019, 2022b). The larger thresholds are necessary, e.g., because of the high number of 2.8 cyclone passages in weekly mean cyclone occurrences over Europe (see other regions in the appendix). We define “ensemble agreement” as the cumulative percentage of ensemble members that show a trend of the same direction (i.e., positive, negative, or none) in the different types of cluster events and specifically analyze the lower (more extreme) 5th, 10th, 15th, and 20th percentiles of the mslp and the median of the mslp. These percentiles characterize different levels of storm intensity (mslp) (similar to e.g., Priestley and Catto 2022; Karwat et al. 2022) and are calculated globally for all cyclone clusters. Finally, we summarize the cyclone clustered activity on average (using the arithmetic mean of the mslp) and relate their significance to our ensemble agreement, corresponding to 60% (medium), 70% (high), and 80% (very high) of all individual ensemble members in 2060–2100 compared to 1980–2020.
Moreover, we examine the large-scale patterns that are associated with cyclone clustering in 2060–2100 and compare them to patterns from historical clustering during 1980–2020. Here, we concentrate on the 3 cyclones within 7 days scenario to capture the most severe cyclone cluster events using CESM2-LE. Finally, we investigate specific cyclone characteristics that help us understand whether cyclones are more intense in the future by evaluating the mslp, the pressure gradient, the storms’ radii and depths, lifetimes, and the distances traveled. With this aim, we use the Welch two-sample t test (Wilks 2011) to show significant difference between 2060 and 2100 and the recent climate in 1980–2020.
3. Cyclone and large-scale evaluation by ERA5 and CESM2-LE
a. Cyclone statistics for recent climate conditions
Our analysis starts with comparing the number of all cyclone occurrences in ERA5 and CESM2-LE. We find that for 1980–2020 CESM2-LE underestimates the cyclone frequency in the North Atlantic by approximately 10% and by about 4% in the North Pacific when compared to ERA5 (Table 1). While the mean North Pacific cyclone count in CESM2-LE is consistent with that in ERA5, the mean North Atlantic cyclone frequency is underestimated in the climate model, as seen by the ensemble spread of CESM2-LE (Table 1). The slightly underestimated number of cyclones from the ensemble mean of CESM2-LE could be related to ERA5 being more sensitive toward tracking cyclones or a bias of the large ensemble. Overall, there is evidence that the number of all cyclones is decreasing in the Northern Hemisphere between 1850–90 and 2060–2100 (Table 1). This trend is clearer when considering the three time slices, that is, 2060–2100 in comparison with 1980–2020 and 1850–90, respectively (p values ≤ 0.05, Table 1). Only in the North Pacific between 1980–2020 and 1850–90 does the number of cyclones remain stationary (Table 1). The significant trends in cyclone occurrences suggest that their numbers have been decreasing since preindustrial times and that there will be fewer cyclones by the end of the current century. These findings are in line with other studies, which also have identified a reduction in the number of cyclones in future decades (e.g., Bengtsson et al. 2006; Pinto et al. 2009; Ulbrich et al. 2009; Zappa et al. 2013; Priestley and Catto 2022).
Statistical significance of the cyclone frequency in the North Atlantic and North Pacific in ERA5 and in CESM2-LE. The last column shows the p value of the trend.
The characteristics of the North Atlantic storm track are almost identical in ERA5 and CESM2-LE in 1980–2020 in terms of the typical northeast–southwest tilt of the main storm track (Figs. 2a,b). In both datasets the maximum in the number of seasonal cyclone counts is located between Greenland and Iceland (e.g., Denmark Strait). Furthermore, the number of cyclones is decreasing from northern to central Europe, with the lowest numbers occurring between 25° and 40°N (Figs. 2a,b). In CESM2-LE, a higher number of up to 2 cyclones per grid point is detected between Iceland and Great Britain (Fig. 2c). Here, ERA5 identifies fewer cyclones (Fig. 2a). With the exception of a few minor differences (e.g., Baffin Bay and eastern Mediterranean), CESM2-LE reproduces the North Atlantic storm track rather well.
North Atlantic storm track in (a) ERA5 and (b) CESM2-LE. (d),(e) As in (a) and (b), but for the North Pacific storm track. (c),(f) Storm track differences between CESM2-LE and ERA5 in 1980–2020.
Citation: Journal of Climate 37, 4; 10.1175/JCLI-D-23-0160.1
The observed storm track features in the North Pacific are very well replicated by CESM2-LE in 1980–2020 (Figs. 2d,e). Although the maximum in seasonal cyclone counts may be slightly more specific in terms of where the maxima are located in ERA5 (Fig. 2d) as in CESM2-LE (Fig. 2e), the storm track is well captured. A higher number of up to 2 cyclones in the difference field is detected over Siberia and Alaska in ERA5 (Fig. 2f).
Moreover, we compare ERA5 to the ensemble spread (defined as twice the ensemble standard deviation) of the large ensemble in 1980–2020 to ascertain how cyclone characteristics are represented in both datasets. As previously stated, the ensemble spread of cyclone counts matches well with the number of cyclones in ERA5, especially for North Pacific cyclones (Table 2). Northern Hemispheric winter cyclones have slightly longer lifetimes and therefore larger radii and depths, and travel longer distances in the large ensemble than in ERA5 (Table 2). However, these differences are comparably small, as is the ensemble spread. This indicates a high level of confidence in the ensemble’s ability to precisely depict cyclone characteristics.
Mean frequency of cyclones per winter, lifetime, minimum and mean mslp, mean radius, depth, and distance in ERA5 and CESM2-LE during 1980–2020. We compare ERA5 to the ensemble spread of CESM2-LE.
To conclude, CESM2-LE provides an accurate estimate on the North Pacific storm track and storm characteristics in 1980–2020 compared with the track and characteristics in ERA5, but seems less accurate on the North Atlantic storm characteristics due to some regional differences.
b. Mean sea level pressure conditions in ERA5 and CESM2-LE
We further evaluate the mslp fields of both Northern Hemisphere storm tracks to show whether ERA5 and CESM2-LE are comparable during 1980–2020. The most noticeable differences in the North Atlantic mslp are found between Iceland and Scotland, where ERA5 shows higher mslp of up to 3 hPa and lower mslp of up to 5 hPa over the western Mediterranean (Fig. A1a in the appendix). In the recent climate of the North Pacific, mslp in ERA5 is higher over the Bering Sea than in CESM2-LE (Fig. A1b in the appendix). Thus, some of these differences might account for changes in the mslp or dynamical changes in storms that originate in these regions.
Having shown that ERA5 and CESM2-LE are overall compatible in cyclone frequency, characteristics and mslp, we investigate the differences in the mslp fields of both storm tracks in 2060–2100 compared to 1980–2020 to see whether they can (partly) explain changes in storm intensities (Fig. 3).
Differences in mean sea level pressure between 2060–2100 and 1980–2020 in the (a) North Atlantic and (b) North Pacific storm track regions using CESM2-LE. The black dots mark grid points where differences are significant at the 95% confidence level.
Citation: Journal of Climate 37, 4; 10.1175/JCLI-D-23-0160.1
We find that in the North Atlantic region mslp is decreasing over the Hudson Bay and central to northern Scandinavia by up to 4 hPa (Fig. 3a). A t test shows that these trends are statistically significant as indicated by the black dots. A strengthening eastern trough associated with cyclogenesis over the eastern Hudson Bay implies that storms in or emerging from these areas are more intense (Fazel-Rastgar 2020). In the future, there may be either more storms in these regions or storms may be more intense. The melting of ice caps in Canada might also contribute to the changes in pressure in these regions (Serreze et al. 2017). Moreover, pressure is significantly increasing over Greenland by up to 3 hPa. These trends indicate that storms that have their cyclogenesis, for example, over Greenland or in the Denmark Strait, might have a higher initial pressure before they intensify as they are propagating toward the European mainland. Greenland is one of the fastest-warming regions on Earth and its ice sheet is melting more rapidly in the higher global warming SSP3-7.0 scenario (Xie et al. 2022). As a result of Arctic warming, this might lead to an increase in mslp. Furthermore, it might also affect the baroclinicity and therefore the intensity of cyclones. Note that changes in mslp should be treated carefully due to the complex orography of Greenland.
There is an increased pressure gradient over Europe in the future that might affect the cyclone strength in this region and also result in continued propagation into the continent, which could enhance cyclone clustering (Fig. 3a). In the Mediterranean, pressure is increasing (Fig. 3a). A t test confirms that there is a significant difference between the two periods (p values < 0.05). This region has been identified as drying up with fewer numbers of cyclones in the future (Ulbrich et al. 2009; Reale et al. 2022). With the exception of a very few minor shifts and more extended areas, the future changes in mslp we have seen here between 2060–2100 and 1980–2020 are almost identical with those of 2060–2100 compared to 1850–90. This means that climatological mslp in 1980–2020 is very similar to 1850–90 (not shown).
Significant changes in mslp are also found in the North Pacific region (see black dots in Fig. 3b). Here, the central part of the North Pacific Ocean experiences an increase in pressure by approximately 3 hPa (Fig. 3b). Higher greenhouse gas emissions in the SSP3-7.0 scenario might cause this amplifying effect (e.g., through changes in atmospheric large-scale circulation patterns or weather patterns). Over Siberia and in the Bering Sea we find that pressure is significantly decreasing by up to 5 hPa (Fig. 3b), which implies a higher potential that cyclones propagating over the northern part of the North Pacific and toward Alaska and Canada are more intense on average. This has already been observed for Arctic cyclones in this region (Day and Hodges 2018). There is a decline in the sea ice extent over this particular region (Jeong et al. 2022; Huang et al. 2022), similar to the ice sheet melting over Greenland (Zheng et al. 2022). Interestingly, the trend in mslp here is negative. One possible reason could be changes in atmosphere stratification or in the meridional temperature gradient, which in turn could affect the baroclinicity. This could explain alterations in the cyclones’ intensities. Future research is needed to identify the processes behind these opposite trends between a decline in sea ice mass and changes in mslp. We find that the differences here are similar to those of 2060–2100 and 1850–90 with the exception of one small area near the U.S. West Coast (located approximately at 130°W and 40°N) where pressure is decreasing by up to 2 hPa. This might be a derivative of the climate response between 1980–2020 and 1850–90 (not shown). Thus, cyclones in this particular region might be more intense, and the eastern part of the North Pacific storm track more prone to cyclone clustering.
While a statistically significant decrease in mslp is found for some areas (e.g., northeastern Canada, Scandinavia, the Bering Sea, and Alaska) under future climate conditions, other regions (e.g., Greenland, the Mediterranean, and central part of the North Pacific Ocean) show an increase, thus leading to substantial changes in pressure gradients in the midlatitudes. In conclusion, our findings suggest that the climatological background state, the mslp fields, might generally influence the intensity of cyclones under future conditions, however, more research is needed to quantify whether the changes lead to, e.g., stronger wind anomalies associated with the storms.
c. Mean frequency of future cyclones in CESM2-LE
We are also interested in the future cyclone mean counts, since they are highly affected by the changes in the intensity of the climatological background states. To study the number of cyclones in more detail, we examine the mean number of cyclone transits in 2060–2100 and compare them to 1980–2020 (Fig. 4). We find that the mean number of cyclone transits is decreasing over Greenland and in the eastern Mediterranean (Fig. 4a), where mslp is higher in the future (Fig. 3a). These are the only differences identified by the large ensemble. Comparing the different confidence intervals shows that the decrease in the number of cyclone occurrences is, however, only statistically significant for cyclones in the eastern Mediterranean. The future changes in mean cyclone counts we have seen here between 2060–2100 and 1980–2020 are similar to those of 2060–2100 and 1850–90 (not shown). In the latter period, the decreases in cyclone counts are statistically significant for both southern Greenland and the eastern Mediterranean. The decreased cyclone occurrences are in line with other studies, e.g., Bengtsson et al. (2006), Zappa et al. (2013), Priestley and Catto (2022).
Differences in the mean number of seasonal cyclone counts per grid point of (a) North Atlantic and of (b) northeast Pacific cyclone transits during 2060–2100 compared to 1980–2020 using CESM2-LE. The black dots mark grid points where differences are significant at the 95% confidence level.
Citation: Journal of Climate 37, 4; 10.1175/JCLI-D-23-0160.1
Now, we only focus on the Pacific Northwest, since we are particularly interested in cyclone clustering in the eastern part of the North Pacific including the Gulf of Alaska and the west coast regions of the United States and Canada. CESM2-LE shows fewer cyclone transits in the future compared to recent climate conditions (Fig. 4b). This large basinwide reduction in the North Pacific differs from other studies where a poleward shift across the North Pacific is observed (e.g., Bengtsson et al. 2006; Priestley and Catto 2022; Karwat et al. 2022). This might be due to the SSP3-7.0 scenario that is applied in CESM2-LE. The specific details and magnitude of the poleward shift can vary depending on the model used, the emissions scenario considered, the region, and the time period analyzed. A reduction in cyclone transits is argued to be a result of changing temperature gradients and a higher amount of atmospheric moisture available due to global warming (e.g., Bengtsson et al. 2006; Pinto et al. 2009; Ulbrich et al. 2009; Zappa et al. 2013; Priestley and Catto 2022). The only exception is western Canadian provinces, where it is indicated that the numbers of cyclones are slightly increasing (Fig. 4b). Both results are statistically significant as indicated by the black dots. There are no significant differences between 2060–2100 and 1850–90 that differ from 1980 to 2020 (not shown). This suggests robust findings in future North Pacific cyclone counts.
To conclude, climate simulations show that the number of cyclones decreases particularly over rapidly warming regions, such as Greenland and the eastern Mediterranean. In addition, the central northeast Pacific will experience significantly fewer cyclone transits in the future. It is likely that dynamic changes such as the earlier discussed alterations in the temperature gradients and baroclinicity are responsible for these changes.
4. Clustering analysis based on storm numbers
a. ERA5 and CESM2-LE in recent climate conditions
Now we examine clustered events from the cyclone tracking data directly. First, we compare how many cluster type events were identified by ERA5 and CESM2-LE during 1980–2020, respectively. The intensity is measured by different percentiles of mslp where lower percentiles refer to more extreme clusters. For 3 cyclones within 7 days, ERA5 identifies 17 clustered events on average in the entire 40-yr period; 14 in CESM2-LE (Fig. 5a). There have only been up to 5 clustered events in the most extreme 5th and 10th percentiles of mslp over Europe in 1980–2020, since these are very rare events. Along the west coast of the United States and Canada we find that the intensity of clustered cyclones is clearly lower (indeed, there has been only one occurrence in the 20th percentile of mslp), while the median intensity is characterized by approximately 10 clustered events during 1980–2020 (Fig. 5b). In contrast, cyclones in the Gulf of Alaska cluster on average almost 10 times more often than along the west coast of the United States and Canada and 5–6 times more often than over Europe: the deviation between CESM2-LE and ERA5 is less than 6% (Fig. 5c). Extremely intense cyclone clusters, e.g., 5th percentile of mslp, occur about every 5–6 years in this region.
Total number n of the “3 cyclones in 7 days” cluster type event over (a) Europe, (b) along the west coast of the United States and Canada, and (c) in the Gulf of Alaska as identified by ERA5 and CESM2-LE during 1980–2020.
Citation: Journal of Climate 37, 4; 10.1175/JCLI-D-23-0160.1
To summarize, ERA5 and CESM2-LE are highly comparable over different percentiles of mslp in 1980–2020. Thus, the large ensemble shows reasonable skill compared to ERA5 in identifying clustered events with different cyclone intensities.
b. Future projections of different cyclone cluster events in CESM2-LE
We now look at possible trends in the different types of cyclone cluster events between 1980–2020 and 2060–2100. To quantify clustering we consider 3, 4, 5, and 6 cyclone passages within 7 days as the cumulative percentage of ensemble members using the same thresholds (different percentiles of mslp) as in the previous section.
For 3 cyclones within 7 days, the ensemble members agree by 72% in the median of mslp that there will be more clustering over Europe (1.3 events per winter in 2060–2100; Fig. 6a). Furthermore, the large ensemble finds no changes in extreme cyclone quadruples concerning the 5th percentile of mslp in 7 days (up to 90% agreement, 0.1 events per winter in 2060–2100; Fig. 6b). However, the ensemble shows medium confidence in an increase of 4 cyclones within 7 days in the future (62% agreement, 0.5 events per winter in 2060–2100; Fig. 6b). In terms of 5 or 6 cyclones within 7 days, values that rarely occur in this area, there is very high ensemble confidence across almost all percentiles of mslp that there are no changes expected by 2060–2100 (Figs. 6c,d). Comparing 1850–90 with 2060–2100, we find similar projections (not shown). In conclusion, the large ensemble predicts more clustering over Europe for cyclone triplets and quadruples within 7 days, but no changes in the most extreme percentile of mslp and concerning 5 or 6 cyclones.
Ensemble agreement for trends in cyclone clustering over Europe for (a) 3 cyclones, (b) 4 cyclones, (c) 5 cyclones, and (d) 6 cyclones in 7 days. The ensemble agreement is defined as the cumulative percentage of all 50 ensemble members (ems) of CESM2-LE that either predict an increase (orange bars), decrease (beige bars), or no trend (gray bars) between 2060–2100 and 1980–2020.
Citation: Journal of Climate 37, 4; 10.1175/JCLI-D-23-0160.1
For the west coast states of the United States and Canada, 74% of the ensemble members agree on a decrease in the median of mslp for 3 cyclones within 7 days (0.6 events per winter in 2060–2100; Fig. 7a). For cyclone quadruples we find no changes across different percentiles of mslp except for an overall decrease in clustering indicated by the median of mslp for 7 days (0.2 events per winter in 2060–2100, 58% ensemble agreement; Fig. 7b). The trends we have seen in the median of mslp are more confident between 1850–90 and 2060–2100, and agree with the current climate (not shown). Considering 5 or 6 cyclones, which also rarely occur in this area, the large ensemble shows overall no trend across different percentiles of mslp between 2060–2100 and 1980–2020 (Figs. 7a–d). This is an important finding, since it suggests that extreme cyclone clusters might not necessarily become more common in the Pacific Northwest.
Ensemble agreement for trends in cyclone clustering at the west coast of the U.S. and Canada for (a) 3 cyclones, (b) 4 cyclones, (c) 5 cyclones, and (d) 6 cyclones in 7 days. The ensemble agreement is defined as the cumulative percentage of all 50 ensemble members (ems) of CESM2-LE that either predict an increase (orange bars), decrease (beige bars), or no trend (gray bars) between 2060–2100 and 1980–2020.
Citation: Journal of Climate 37, 4; 10.1175/JCLI-D-23-0160.1
In the Gulf of Alaska, the clustering of cyclone triplets within 7 days decreases in the 20th percentile of mslp according to 98% of all ensemble members (2 events per winter in 2060–2100; Fig. 8a). Cyclone quadruples are also shown to be decreasing overall in the future (up to 72% agreement, 1.3 events per winter in 2060–2100; Fig. 8b). In addition, no trend is identified in the 5th percentile of mslp concerning 5 cyclones in 7 days (Fig. 8c). However, there seems to be a clear cut when comparing 6 cyclones in 7 days: here, we detect high ensemble confidence for less clustered cyclones in the 20th percentile of mslp and especially in the median of mslp (Fig. 8d). We find that the ensemble members are even more confident in the decrease of clustering across all percentiles of mslp when comparing these projections to 1850–90 and 2060–2100 (not shown). This might result from a stronger climate signal between preindustrial times and the SSP3-7.0 scenario in the future.
Ensemble agreement for trends in cyclone clustering in the Gulf of Alaska for (a) 3 cyclones, (b) 4 cyclones, (c) 5 cyclones, and (d) 6 cyclones in 7 days. The ensemble agreement is defined as the cumulative percentage of all 50 ensemble members (ems) of CESM2-LE that either predict an increase (orange bars), decrease (beige bars), or no trend (gray bars) between 2060–2100 and 1980–2020.
Citation: Journal of Climate 37, 4; 10.1175/JCLI-D-23-0160.1
To summarize, the majority of ensemble members agree that enhanced cyclone clustering will occur over Europe for 3 and 4 cyclones in 7 days in the future, but only for average cyclones rather than for those in the most extreme percentiles of mslp. Clustering along the west coast of the United States and Canada is seen to generally decline or remain steady in lower percentiles of mslp for extreme storms; the ensemble agreement on clustering to decrease in the Gulf of Alaska in a future winter climate is similarly confident.
c. Ensemble agreement on projections of future cyclone clustering in CESM2-LE
For different types of clustering we can now compare the mean cyclone clustered activity as defined by the arithmetic mean of the mslp to the significance derived from our ensemble agreement (Table 3). Concerning Europe, we find an increase of 21.5% for 3 cyclones within 7 days by 2060–2100 compared to 1980–2020 (high confidence) (Table 3). Similarly, cyclone quadruples are increasing over Europe within 7 days by 25% during 2060–2100 compared to 1980–2020 (medium confidence). Along the west coast of the United States and Canada cyclone triplets within 7 days decrease by 24.3% during 2060–2100 compared to 1980–2020 (medium ensemble confidence; Table 3). However, the actual numbers here are small, since these are very rare events. Over the Gulf of Alaska, clustering is more common and decreases by 10.1% in the same period (high confidence; Table 3).
Trends in cyclone clustering of 3 or 4 clustered cyclones in 7 days as mean numbers per winter and as mean percentage change for Europe, the west coast regions of the United States and Canada, and the Gulf of Alaska during 2060–2100 compared to 1980–2020 and 1850–90 in CESM2-LE. Ensemble confidence is defined as the agreement of the same trend that is identified by at least 60% (medium), 70% (high), and 80% (very high) of all individual ensemble members in 2060–2100 compared to the reference period 1980–2020.
To conclude, there is high confidence in the ensemble for more cyclone clustering over Europe and for less clustering in the Pacific Northwest regions. Overall, this direct approach of comparing cyclone triplets and quadruples and the other cyclone cluster types appears to be quite straightforward and easily interpretable when determining actual changes in future cyclone clustering compared to the theoretical approach (e.g., using a dispersion statistic; see Mailier et al. 2006). The dispersion is a much coarser metric where it is only possible to distinguish between regular and clustered cyclones without the information on different cyclone intensities. Taking absolute numbers of clustered cyclones into consideration, that is, different types of cyclone cluster events defined by the intensity (mslp) in relation to the ensemble confidence, yields robust results.
5. Large-scale features of clustered cyclones
a. Possible links between cyclone clustering and large-scale patterns in CESM2-LE
There are considerable knowledge gaps in the physical mechanisms that underlie possible changes in cyclone clustering, which is thus quite uncertain (Dacre and Pinto 2020). Changes in baroclinicity brought on by global warming and substantial variations in large-scale conditions are often considered as the main two factors to influence (clustered) cyclone occurrences (Walz et al. 2018). Pinto et al. (2009) showed that a positive North Atlantic Oscillation (NAO) phase is more favorable for the growth conditions of extreme cyclones than a negative NAO phase. Additionally, the impact of the NAO on cyclone clustering is largely independent from the cyclone tracking algorithm used (Pinto et al. 2016). For the positive phase of the Pacific–North American (PNA) pattern, Mailier et al. (2006) found a decrease in cyclone counts across the Canadian west coast but an overall increase of cyclones in the North Pacific.
Here, we examine the large-scale patterns that are associated with cyclone clustering in 2060–2100 and compare them to patterns from 1980 to 2020 (Fig. 9). We focus on the 5th percentile of mslp; that is, we consider only the most severe cyclones to build the 3 cyclones within 7 days clusters in CESM2-LE. A minimum in mslp over Northern Ireland was shown during cyclone clustering in 1980–2020 (Fig. 9a). By 2060–2100, the minimum has moved further east toward Scotland (Fig. 9b).
Ensemble mean of the large-scale patterns during historical clustering of cyclone triplets within 7 days over (a) Europe, (c) along the west coast of Canada and the United States, and (e) in the Gulf of Alaska in 1980–2020; (b),(d),(f) as in (a), (e), and (e), but for 2060–2100. Here, only the strongest cyclones are considered, e.g., the 5th percentile of mslp of 3 cyclones within 7 days. We find similar patterns for other intensities of clustered cyclones using CESM2-LE (not shown).
Citation: Journal of Climate 37, 4; 10.1175/JCLI-D-23-0160.1
In 1980–2020, cyclones cluster along the west coast of the United States and Canada when a minimum in mslp is located near Vancouver Island (Fig. 9c). There may also be a second low pressure system present north of the Aleutian Islands (Fig. 9c). In the future, however, these two low pressure systems have merged into one overall system (Fig. 9d). This suggests that the storm track is more focused.
Cyclones in the Gulf of Alaska cluster during the occurrence of a mslp minimum south of the Aleutian Islands (see the 970-hPa isobar in Fig. 9e). We find that this minimum shifts farther east and slightly expands by 2060–2100 (Fig. 9f). However, there is little change to the impact regions.
Thus, the large-scale patterns might not project onto the NAO, PNA, or other teleconnections. When considering different intensities (other percentiles of mslp) or types of cyclone clusters, similar patterns are found (not shown). This confirms that the overall large-scale conditions are rather persistent during times of cyclone clustering (Hauser et al. 2023), which could be useful not only for weather predictions but also for future projections.
b. Trends in cyclone characteristics in CESM2-LE
Finally, we investigate possible changes in cyclone characteristics in a warming climate, considering both nonclustered and clustered cyclones. For the North Atlantic, we find evidence that cyclones are decreasing in mslp, lifetime, radius, and depth during 2060–2100 compared to 1980–2020 (p values ≤ 0.05; Table 4). Regularly occurring cyclones also travel longer distances in the future (p value ≤ 0.05; Table 4). These trends are in line with a previous study where cyclones have been shown to be more intense in recent decades and traveling longer distances (Karwat et al. 2022). In contrast, we find that the mean pressure gradient is significantly decreasing for nonclustered North Atlantic cyclones (Table 4), which indicates a general weakening of the surface winds associated with these storms.
Trends in characteristics of nonclustered cyclones over the North Atlantic (NA) and North Pacific (NP) using the ensemble mean of CESM2-LE. We compare the cyclone characteristics of intensity (mslp), pressure gradient, lifetime, radius, depth, and distance traveled during 2060–2100 to the reference period 1980–2020.
Next we focus on clustered cyclones in the North Atlantic, but specifically on “3 cyclones in 7 days” as this seems to be the most representative scenario, similar to what is used by insurance companies to evaluate the socioeconomic losses of clustered storms (Alert Air Worldwide 2019, 2022b). We find that clustered cyclones in the North Atlantic are more intense than regularly occurring storms, have larger radii, and travel longer distances during 2060–2100 compared to 1980–2020 (median, p values ≤ 0.05; Table 5). When considering the most extreme clustered cyclones, the 5th percentile of mslp, we find that mslp is decreasing during 2060–2100 (p values ≤ 0.05; Table 5). At the same time the most extreme cyclone clusters have larger radii and depths in the future, while they also travel longer distances than during 1980–2020 (p values ≤ 0.05, Table 5). As a result, trends in the 5th percentile of mslp are similar to the median, except for an increased pressure gradient (Table 5). This increased pressure gradient implies stronger surface winds associated with more extreme storms. Thus, North Atlantic clustered cyclones with especially high intensity will intensify in the future. This may imply an increase in socioeconomic losses for Europe.
Trends in characteristics of clustered cyclones over the North Atlantic (NA) and North Pacific (NP) using the ensemble mean of CESM2-LE. We compare the cyclone characteristics of intensity (mslp), pressure gradient, radius, depth, and distance traveled during 2060–2100 to the reference period 1980–2020. Here, the “3 cyclones in 7 days” scenario is considered for the median and the 5th percentile of mslp.
Nonclustered North Pacific cyclones, like North Atlantic cyclones, are decreasing in mslp, pressure gradient, and depth, but are showing longer lifetimes, radii, and distances traveled during 2060–2100 compared to 1980–2020 (p values ≤ 0.05; Table 4). This trend is also seen in clustered cyclones, which have larger radii, but show an increased pressure gradient during 2060–2100 in comparison with 1980–2020 (median, p values ≤ 0.05; Table 5). This implies stronger surface winds associated with clustered cyclones on average. The most extreme cyclones show similar trends in radius and depth; however, the trends are not statistically significant (Table 5). Hence, North Pacific clustered cyclones reveal the same trends as North Atlantic clustered cyclones, but are significant in only a few variables.
In conclusion, all Northern Hemispheric cyclones are intensifying, but especially clustered cyclones reveal larger radii and depths that might be associated with stronger surface winds in future decades. Currently it is not clear what may cause the enhanced growth rates in clustered cyclones. It is theorized that cyclones may enhance baroclinicity through diabatic heating and moisture availability, resulting in preferred times of cyclone clustering (Weijenborg and Spengler 2020). Therefore, they may develop better in a warmer and wetter climate. This might explain why certain cyclones are intensifying more rapidly than others under the SSP3-7.0 scenario.
6. Conclusions
Using the ERA5 and CESM2-LE data, we analyzed the representation of Northern Hemisphere extratropical cyclone characteristics and specifically of cyclone clustering under recent and future climate conditions for boreal winter. With this aim, we have considered absolute measures based on storm numbers for clustering. The main conclusions are as follows:
-
During 1980–2020, we find different cyclone characteristics for the three regions studied: over Europe, we find that there have only been up to 5 clustered events of the 3 cyclones in 7 days scenario in the most extreme 5th and 10th percentiles of mslp in the past, since these are very rare events. Along the U.S. and Canadian west coast, cyclone clustering is characterized by approximately 10 clustered events on average during 1980–2020. In contrast, we find that clustered cyclones occur 10 times more often in the Gulf of Alaska than along the west coast of the United States and Canada, and 5 to 6 times more frequently than across Europe. CESM2-LE shows reasonable skill compared to ERA5 in identifying clustered events with different cyclone intensities of mslp for the 3 cyclones in 7 days scenario.
-
Compared to ERA5, the CESM2-LE provides an accurate estimate on the typical southwest to northeast tilt of the North Atlantic storm track and the more zonally oriented North Pacific storm track. Some uncertainties on the precise locations (e.g., between Iceland and Great Britain) remain. In terms of cyclone characteristics, the CESM2-LE replicates the average cyclone count and mean intensity especially for the North Pacific reasonably, but slightly overestimates the mean cyclone lifetimes in both Northern Hemispheric ocean basins. We assume that these differences are due to ERA5’s high resolution and a model bias of CESM2-LE.
-
Using the absolute metric, clustering based on storm numbers, in the CESM2-LE, we find robust results across different percentiles of mslp and thus can identify significant trends in cyclone clusters with varying intensities: during 2060–2100, clustering increases by up to 25% on average for 3 and 4 cyclones within 7 days over Europe, but decreases by 24.3% along the west coast of the United States and Canada and by 10.1% in the Gulf of Alaska compared to 1980–2020. The absolute count metric takes into account the actual storms by numbers, individual trajectories and intensities. Thus, it produces robust results in relation to the ensemble agreement and confidence.
-
In the SSP3-7.0 scenario, the large-scale patterns do not change under cyclone clustering in 2060–2100 in comparison with 1980–2020, which may be useful for predicting future cluster events. At the same time cyclones are intensifying by having lower mslp minima but especially clustered cyclones may develop even better in a warmer winter climate, since they have much larger radii and depths than nonclustered cyclones in the future. Moreover, clustered cyclones reveal an enhanced pressure gradient that is likely associated with stronger surface winds and consequently more damage. This highlights the need for new studies addressing the changes in the dynamics behind clustering.
Projected changes by the count metric could be related to a poleward shift of the storm tracks on zonal average (e.g., Bengtsson et al. 2006; Harvey et al. 2020; Karwat et al. 2022). It is likely that clustering is also affected by cyclone tracks shifting northward and therefore affecting the spatial distribution of storms. Recent studies based on CMIP5 models have shown that while cyclone clustering may increase over northern and western Europe, trends are often small, insignificant, and inconsistent between the different climate models (Pinto et al. 2013; Economou et al. 2015). Using CMIP6 data, we have shown that the CESM2-LE’s signal on trends in clustering based on our count metric is robust as shown by a strong ensemble agreement across different intensities of cyclone triplets, quadruples, and even higher numbers of cyclones in the future. Moreover, this direct approach appears to be quite straightforward and easily interpretable when comparing socioeconomic losses and quantifying future changes in cyclone clustering.
These results are in line with studies by Pinto et al. (2013) and Economou et al. (2015), who found that clustering is likely to increase over Europe in the future. However, here we quantify the storm changes in CMIP6 data for the first time. A general significant increase in cyclone clustering over Europe across different percentiles of mslp implies more socioeconomic damage for countries such as Great Britain, France, the Netherlands, and Germany. Karremann et al. (2014) have already shown shorter return periods of western European storm series in the future.
Even though the decline in future cyclone clustering in the Gulf of Alaska is highly intriguing from a climatological perspective, it also implies a reduced socioeconomic impact on the cities nearby (i.e., Seattle, Vancouver, and Anchorage) due to less populated regions along the Canadian west coast than the more densely populated urban areas affected in Europe. Thus, the new insights presented here may facilitate a better probability of future cyclone clustering in different regions. There is an urgent need for further research on the mechanisms behind the observed regional differences.
Our study has only considered the SSP3-7.0 scenario where higher greenhouse gas emissions lead to more clustering over Europe, and to less clustering on part of the Pacific Northwest. Thus, the obtained probabilities may be different for other scenarios.
To conclude, our study provides evidence that cyclone clustering over the Northern Hemisphere may change in a warmer climate depending on the region, with implications also for the associated hazards, primarily strong wind gusts, heavy rainfall, and storm surges. Studies that combine clustering of storms and impacts are urgently needed (Franzke 2022). Thus, an adequate quantification of cyclone clustering and related wind and precipitation extremes in a warming climate for different scenarios is very important. This is research we plan to pursue in the future. Finally, the uncertainties in clustering can be reduced when utilizing large ensembles of data, like in the present study, which ideally would be CMIP6-type multimodel ensembles.
Acknowledgments.
We thank the reviewers for their constructive comments to improve our manuscript. The simulations were conducted on the IBS/ICCP supercomputer “Aleph,” a 1.43 petaflops high-performance Cray XC50-LC Skylake computing system with 18 720 processor cores, 9.59 PB storage, and 43 PB tape archive space. We also acknowledge the support of KREONET. AK, JGP and RB were supported by the Federal Ministry of Education and Research (BMBF) consortium ClimXtreme, projects B3.6 and A6 (01LP1902F, 01LP1901A). CF and SSL were supported by the Institute for Basic Science (IBS), Republic of Korea, under IBS-R028-D1, and AK and CF by the National Research Fund of Korea (NRF-2022M3K3A1097082). JGP thanks the AXA Research Fund for support.
Data availability statement.
The ERA5 reanalysis is openly available in the Climate Data Store of the ECMWF at https://cds.climate.copernicus.eu/cdsapp\#!/search?type=dataset. The CESM2-LE ensemble members are available in the Climate Data Gateway at NCAR https://www.earthsystemgrid.org/dataset/ucar.cgd.cesm2le.atm.proc.6hourly_ave.PSL.html and at request at the IBS/ICCP in Busan, South Korea: https://ibsclimate.org/research/cesm2-large-ensemble-community-project/.
APPENDIX
Differences in mslp between ERA5 and CESM2-LE during 1980–2020
We define cyclone clustering using the absolute measure based on storm counts within 7 days (following e.g., Pinto et al. 2014). To assess the effect of clustered cyclones it is necessary to consider higher numbers of cyclones (e.g., 3–6 storms in 7 days), since the weekly mean cyclone occurrences are already high in CESM2-LE (see Fig. A1 and Table A1).
Differences in mean sea level pressure between ERA5 and CESM2-LE in the (a) North Atlantic and (b) North Pacific storm track regions during 1980–2020.
Citation: Journal of Climate 37, 4; 10.1175/JCLI-D-23-0160.1
Number of weekly mean cyclone occurrences and standard deviation in the three study regions using CESM2-LE for the different time slices 1850–90, 1980–2020, and 2060–2100.
REFERENCES
Alert Air Worldwide, 2009: Looking back, looking forward: Anatol, Lothar and Martin ten years later. Accessed 22 November 2022, https://www.air-worldwide.com/publications/air-currents/looking-back-looking-forward-anatol-lothar-and-martin-ten-years-later/.
Alert Air Worldwide, 2019: Modeling fundamentals: Accounting for the hours clause. Accessed 22 November 2022, https://www.air-worldwide.com/publications/air-currents/2019/Modeling-Fundamentals--Accounting-for-the-Hours-Clause/.
Alert Air Worldwide, 2022a: Europe hit by a storm triplet within six days. Accessed 22 November 2022, https://www.air-worldwide.com/blog/posts/2022/02/europe-storm-triplet-in-six-days/.
Alert Air Worldwide, 2022b: Storms Dudley and Eunice: Post-event analysis—Summary. Accessed 22 November 2022, https://alert.air-worldwide.com/extratropical-cyclone/2022/storms-dudley-and-eunice/post-event-analysis/.
Bengtsson, L., K. I. Hodges, and E. Roeckner, 2006: Storm tracks and climate change. J. Climate, 19, 3518–3543, https://doi.org/10.1175/JCLI3815.1.
Bjerknes, J., and H. Solberg, 1922: Life cycle of cyclones and the polar front theory of atmospheric circulation. Geofys. Publ., 3, 468–473.
Blender, R., K. Fraedrich, and F. Lunkeit, 1997: Identification of cyclone-track regimes in the North Atlantic. Quart. J. Roy. Meteor. Soc., 123, 727–741, https://doi.org/10.1002/qj.49712353910.
Blender, R., C. C. Raible, and F. Lunkeit, 2015: Non-exponential return time distributions for vorticity extremes explained by fractional Poisson processes. Quart. J. Roy. Meteor. Soc., 141, 249–257, https://doi.org/10.1002/qj.2354.
Cusack, S., 2016: The observed clustering of damaging extratropical cyclones in Europe. Nat. Hazards Earth Syst. Sci., 16, 901–913, https://doi.org/10.5194/nhess-16-901-2016.
Dacre, H. F., and J. G. Pinto, 2020: Serial clustering of extratropical cyclones: A review of where, when and why it occurs. npj Climate Atmos. Sci., 3, 48, https://doi.org/10.1038/s41612-020-00152-9.
Danabasoglu, G., and Coauthors, 2020: The Community Earth System Model version 2 (CESM2). J. Adv. Model. Earth Syst., 12, e2019MS001916, https://doi.org/10.1029/2019MS001916.
Day, J. J., and K. I. Hodges, 2018: Growing land-sea temperature contrast and the intensification of Arctic cyclones. Geophys. Res. Lett., 45, 3673–3681, https://doi.org/10.1029/2018GL077587.
Della-Marta, P. M., and J. G. Pinto, 2009: Statistical uncertainty of changes in winter storms over the North Atlantic and Europe in an ensemble of transient climate simulations. Geophys. Res. Lett., 36, L14703, https://doi.org/10.1029/2009GL038557.
Economou, T., D. B. Stephenson, J. G. Pinto, L. C. Shaffrey, and G. Zappa, 2015: Serial clustering of extratropical cyclones in a multi-model ensemble of historical and future simulations. Quart. J. Roy. Meteor. Soc., 141, 3076–3087, https://doi.org/10.1002/qj.2591.
Fazel-Rastgar, F., 2020: Seasonal analysis of atmospheric changes in Hudson Bay during 1998–2018. Amer. J. Climate Change, 9, 100–122, https://doi.org/10.4236/ajcc.2020.92008.
Fish, M. A., J. M. Done, D. L. Swain, A. M. Wilson, A. C. Michaelis, P. B. Gibson, and F. M. Ralph, 2022: Large-scale environments of successive atmospheric river events leading to compound precipitation extremes in California. J. Climate, 35, 1515–1536, https://doi.org/10.1175/JCLI-D-21-0168.1.
Franzke, C. L. E., 2022: Changing temporal volatility of precipitation extremes due to global warming. Int. J. Climatol., 42, 8971–8983, https://doi.org/10.1002/joc.7789.
Harvey, B. J., P. Cook, L. C. Shaffrey, and R. Schiemann, 2020: The response of the Northern Hemisphere storm tracks and jet streams to climate change in the CMIP3, CMIP5, and CMIP6 climate models. J. Geophys. Res. Atmos., 125, e2020JD032701, https://doi.org/10.1029/2020JD032701.
Hauser, S., S. Mueller, X. Chen, T.-C. Chen, J. G. Pinto, and C. M. Grams, 2023: The linkage of serial cyclone clustering in western Europe and weather regimes in the North Atlantic-European region in boreal winter. Geophys. Res. Lett., 50, e2022GL101900, https://doi.org/10.1029/2022GL101900.
Hersbach, H., and Coauthors, 2020: The ERA5 global reanalysis. Quart. J. Roy. Meteor. Soc., 146, 1999–2049, https://doi.org/10.1002/qj.3803.
Hodges, K. I., 1995: Feature tracking on the unit sphere. Mon. Wea. Rev., 123, 3458–3465, https://doi.org/10.1175/1520-0493(1995)123<3458:FTOTUS>2.0.CO;2.
Huang, L., A. Timmermann, S.-S. Lee, K. B. Rodgers, R. Yamaguchi, and E.-S. Chung, 2022: Emerging unprecedented lake ice loss in climate change projections. Nat. Commun., 13, 5798, https://doi.org/10.1038/s41467-022-33495-3.
IPCC, 2021: Climate Change 2021: The Physical Science Basis. Cambridge University Press, 2391 pp., https://doi.org/10.1017/9781009157896.
Jeong, H., H.-S. Park, M. F. Stuecker, and S.-W. Yeh, 2022: Record low Arctic sea ice extent in 2012 linked to two-year La Niña-driven sea surface temperature pattern. Geophys. Res. Lett., 49, e2022GL098385, https://doi.org/10.1029/2022GL098385.
Karremann, M. K., J. G. Pinto, M. Reyers, and M. Klawa, 2014: Return periods of losses associated with European windstorm series in a changing climate. Environ. Res. Lett., 9, 124016, https://doi.org/10.1088/1748-9326/9/12/124016.
Karwat, A., C. L. E. Franzke, and R. Blender, 2022: Long-term trends of Northern Hemispheric winter cyclones in the extended ERA5 reanalysis. J. Geophys. Res. Atmos., 127, e2022JD036952, https://doi.org/10.1029/2022JD036952.
Kendon, M., and M. McCarthy, 2015: The UK’s wet and stormy winter of 2013/2014. Weather, 70, 40–47, https://doi.org/10.1002/wea.2465.
Mailier, P. J., D. B. Stephenson, C. A. T. Ferro, and K. I. Hodges, 2006: Serial clustering of extratropical cyclones. Mon. Wea. Rev., 134, 2224–2240, https://doi.org/10.1175/MWR3160.1.
Matthews, T., C. Murphy, R. L. Wilby, and S. Harrigan, 2014: Stormiest winter on record for Ireland and UK. Nat. Climate Change, 4, 738–740, https://doi.org/10.1038/nclimate2336.
Mesquita, M. S., D. E. Atkinson, and K. I. Hodges, 2010: Characteristics and variability of storm tracks in the North Pacific, Bering Sea, and Alaska. J. Climate, 23, 294–311, https://doi.org/10.1175/2009JCLI3019.1.
Moore, B. J., A. B. White, and D. J. Gottas, 2021: Characteristics of long-duration heavy precipitation events along the West Coast of the United States. Mon. Wea. Rev., 149, 2255–2277, https://doi.org/10.1175/MWR-D-20-0336.1.
Mühr, B., L. Eisenstein, J. G. Pinto, P. Knippertz, S. Mohr, and M. Kunz, 2022: CEDIM Forensic Disaster Analysis Group (FDA): Winter storm series: Ylenia, Zeynep, Antonia (int: Dudley, Eunice, Franklin)-February 2022 (NW & Central Europe). CEDIM Research Rep. 1, 21 pp., https://doi.org/10.5445/IR/1000143470.
Munich Re, 2002: Winter storms in Europe (II). Analysis of 1999 losses and loss potentials. Münchener Rückversicherungs-Gesellschaft Rep., 72 pp., https://library.metoffice.gov.uk/Portal/Default/en-GB/RecordView/Index/47780.
Murray, R. J., and I. Simmonds, 1991: A numerical scheme for tracking cyclone centers from digital data. Part I: Development and operation of the scheme. Aust. Meteor. Mag., 39, 155–166, https://www.bom.gov.au/jshess/docs/1991/murray1.pdf.
Neu, U., and Coauthors, 2013: IMILAST: A community effort to intercompare extratropical cyclone detection and tracking algorithms. Bull. Amer. Meteor. Soc., 94, 529–547, https://doi.org/10.1175/BAMS-D-11-00154.1.
Pinto, J. G., S. Zacharias, A. H. Fink, G. C. Leckebusch, and U. Ulbrich, 2009: Factors contributing to the development of extreme North Atlantic cyclones and their relationship with the NAO. Climate Dyn., 32, 711–737, https://doi.org/10.1007/s00382-008-0396-4.
Pinto, J. G., M. K. Karremann, K. Born, P. M. Della-Marta, and M. Klawa, 2012: Loss potentials associated with European windstorms under future climate conditions. Climate Res., 54 (1), 1–20, https://doi.org/10.3354/cr01111.
Pinto, J. G., N. Bellenbaum, M. K. Karremann, and P. M. Della-Marta, 2013: Serial clustering of extratropical cyclones over the North Atlantic and Europe under recent and future climate conditions. J. Geophys. Res. Atmos., 118, 12 476–12 485, https://doi.org/10.1002/2013JD020564.
Pinto, J. G., I. Gómara, G. Masato, H. F. Dacre, T. Woollings, and R. Caballero, 2014: Large-scale dynamics associated with clustering of extratropical cyclones affecting western Europe. J. Geophys. Res. Atmos., 119, 13 704–13 719, https://doi.org/10.1002/2014JD022305.
Pinto, J. G., S. Ulbrich, T. Economou, D. B. Stephenson, M. K. Karremann, and L. C. Shaffrey, 2016: Robustness of serial clustering of extratropical cyclones to the choice of tracking method. Tellus, 68A, 32204, https://doi.org/10.3402/tellusa.v68.32204.
Priestley, M. D. K., and J. L. Catto, 2022: Future changes in the extratropical storm tracks and cyclone intensity, wind speed, and structure. Wea. Climate Dyn., 3, 337–360, https://doi.org/10.5194/wcd-3-337-2022.
Priestley, M. D. K., J. G. Pinto, H. F. Dacre, and L. C. Shaffrey, 2017: The role of cyclone clustering during the stormy winter of 2013/2014. Weather, 72, 187–192, https://doi.org/10.1002/wea.3025.
Priestley, M. D. K., H. F. Dacre, L. C. Shaffrey, K. I. Hodges, and J. G. Pinto, 2018: The role of serial European windstorm clustering for extreme seasonal losses as determined from multi-centennial simulations of high-resolution global climate model data. Nat. Hazards Earth Syst. Sci., 18, 2991–3006, https://doi.org/10.5194/nhess-18-2991-2018.
Reale, M., and Coauthors, 2022: Future projections of Mediterranean cyclone characteristics using the Med-CORDEX ensemble of coupled regional climate system models. Climate Dyn., 58, 2501–2524, https://doi.org/10.1007/s00382-021-06018-x.
Rodgers, K. B., and Coauthors, 2021: Ubiquity of human-induced changes in climate variability. Earth Syst. Dyn., 12, 1393–1411, https://doi.org/10.5194/esd-12-1393-2021.
Schaller, N., and Coauthors, 2016: Human influence on climate in the 2014 southern England winter floods and their impacts. Nat. Climate Change, 6, 627–634, https://doi.org/10.1038/nclimate2927.
Schneidereit, A., R. Blender, and K. Fraedrich, 2010: A radius–depth model for midlatitude cyclones in reanalysis data and simulations. Quart. J. Roy. Meteor. Soc., 136, 50–60, https://doi.org/10.1002/qj.523.
Schultz, D. M., and Coauthors, 2019: Extratropical cyclones: A century of research on meteorology’s centerpiece. A Century of Progress in Atmospheric and Related Sciences: Celebrating the American Meteorological Society Centennial, Meteor. Monogr., No. 59, Amer. Meteor. Soc., https://doi.org/10.1175/AMSMONOGRAPHS-D-18-0015.1.
Serreze, M. C., B. Raup, C. Braun, D. R. Hardy, and R. S. Bradley, 2017: Rapid wastage of the Hazen Plateau ice caps, northeastern Ellesmere Island, Nunavut, Canada. Cryosphere, 11, 169–177, https://doi.org/10.5194/tc-11-169-2017.
Sinclair, M. R., 1994: An objective cyclone climatology for the Southern Hemisphere. Mon. Wea. Rev., 122, 2239–2256, https://doi.org/10.1175/1520-0493(1994)122<2239:AOCCFT>2.0.CO;2.
Ulbrich, U., A. H. Fink, M. Klawa, and J. G. Pinto, 2001: Three extreme storms over Europe in December 1999. Weather, 56, 70–80, https://doi.org/10.1002/j.1477-8696.2001.tb06540.x.
Ulbrich, U., G. C. Leckebusch, and J. G. Pinto, 2009: Extra-tropical cyclones in the present and future climate: A review. Theor. Appl. Climatol., 96, 117–131, https://doi.org/10.1007/s00704-008-0083-8.
Vitolo, R., D. B. Stephenson, I. M. Cook, and K. Mitchell-Wallace, 2009: Serial clustering of intense European storms. Meteor. Z., 18, 411–424, https://doi.org/10.1127/0941-2948/2009/0393.
Walz, M. A., D. J. Befort, N. O. Kirchner-Bossi, U. Ulbrich, and G. C. Leckebusch, 2018: Modelling serial clustering and inter-annual variability of European winter windstorms based on large-scale drivers. Int. J. Climatol., 38, 3044–3057, https://doi.org/10.1002/joc.5481.
Weijenborg, C., and T. Spengler, 2020: Diabatic heating as a pathway for cyclone clustering encompassing the extreme Storm Dagmar. Geophys. Res. Lett., 47, e2019GL085777, https://doi.org/10.1029/2019GL085777.
Wernli, H., and C. Schwierz, 2006: Surface cyclones in the ERA-40 dataset (1958–2001). Part I: Novel identification method and global climatology. J. Atmos. Sci., 63, 2486–2507, https://doi.org/10.1175/JAS3766.1.
Wilks, D. S., 2011: Statistical Methods in the Atmospheric Sciences. Academic Press, 704 pp.
Xie, A., J. Zhu, S. Kang, X. Qin, B. Xu, and Y. Wang, 2022: Polar amplification comparison among Earth’s three poles under different socioeconomic scenarios from CMIP6 surface air temperature. Sci. Rep., 12, 16548, https://doi.org/10.1038/s41598-022-21060-3.
Zappa, G., L. C. Shaffrey, K. I. Hodges, P. G. Sansom, and D. B. Stephenson, 2013: A multimodel assessment of future projections of North Atlantic and European extratropical cyclones in the CMIP5 climate models. J. Climate, 26, 5846–5862, https://doi.org/10.1175/JCLI-D-12-00573.1.
Zheng, L., X. Cheng, X. Shang, Z. Chen, Q. Liang, and K. Wang, 2022: Greenland ice sheet daily surface melt flux observed from space. Geophys. Res. Lett., 49, e2021GL096690, https://doi.org/10.1029/2021GL096690.
