Effects of large-scale changes in environmental factors on the genesis of Arctic extreme cyclones

The Arctic cyclone is an active weather system in the Arctic, and the Arctic extreme cyclone (AEC) strongly influences polar weather. Thus, many studies have focused on the activity of AEC and its relationship with large-scale circulation in the Arctic. In this study, Arctic cyclones were detected using the ERA5 Reanalysis data from 1979 to 2020, and the AECs of cold and warm semesters were defined using the 5th percentile of sea level pressure. An Arctic cyclone genesis index, GPIArctic, is established by using the least square fitting of absolute vorticity, omega, wind shear, and long-wave radiation with Eady growth rate. Our findings show that wind shear and long-wave radiation strongly influence AECs. In the cold semester, the high value of GPIArctic mainly occurs in the south of Greenland, while in the warm semester, the high value of GPIArctic also occurs in northeast Eurasia. The results of the multi-model integration of CMIP6 show that more AECs will be formed in the Arctic in the future (2021–2050), and their seasonal contrast will be stronger in northeast Eurasia.


Introduction
The Arctic is closely associated with the global climate system, having significant impacts on the global water cycle, atmospheric circulation, and thermohaline circulation (Brown and Caldeira 2017). Increased global warming, with the Arctic warming almost four times faster than the global average, which is manifested as the Arctic amplification effect (Cohen et al 2014, Rantanen et al 2022. The life span of Arctic cyclones, one of the most active weather systems in the Arctic, ranges from a few hours to several days (Keegan 1958, Roebber 1984, Serreze and Barry 1988, Serreze 1995, Wei et al 2017. The center of the Arctic cyclone is often accompanied by strong upward movement of the air, resulting in stormy weather that greatly affects the polar weather and climate (Lang andWaugh 2011, Tanaka et al 2012). Arctic cyclones have two main sources: extratropical cyclones at lower latitudes that move northward into the Arctic and cyclones directly formed in the Arctic (Zhang et al 2004, Serreze andBarrett 2008). The number of Arctic cyclones from these two sources is similar in spring and summer seasons, but more Artic cyclones are directly formed in the polar region in autumn and winter seasons compared to the other cyclones that move into the region. In winter, Arctic cyclones mainly occur near the Iceland Low, while in other seasons, they are also distributed in Eurasia and its nearby waters (Zhang et al 2004, Serreze andBarrett 2008). Charney (1947) and Eady (1949) described the development speed of unstable disturbance waves using the baroclinic instability theory. Arctic cyclones are mainly caused due to baroclinic instability (Nielsen 1997, Descamps et al 2007. The maximum Eady growth rate (EGR) derived from the Eady mode is often used to measure the possibility of cyclone development and intensification. In summer, the temperature difference between the snowless land in northeast Eurasia and the Arctic Ocean causes baroclinic instability; in particular, cyclone formation in the Arctic region is influenced by the Arctic frontal zone (AFZ) Rudeva 2012, Crawford andSerreze 2016). The large-scale circulation system is also closely related to Arctic cyclones. The positive phase of the Arctic oscillation (AO) that enhances the jet stream is conducive to the development of cyclones (Serreze and Barrett 2008, Rudeva and Simmonds 2015, Akperov et al 2018. However, few researchers seem to have paid as much attention to the environmental factors that play an important role in the formation of Arctic cyclones as to the causes of tropical cyclones (Murakami et al 2011, Zhang et al 2016, Suneeta and Sadhuram 2018. In recent years, the Arctic has witnessed more intense and long-lasting cyclones, with more frequent occurrences of Arctic extreme cyclones (AECs). Rinke et al (2017) studied the winter trend of AECs from 1979 to 2015 and found that the number of AECs showed a positive trend in the North Atlantic. Simmonds and Rudeva (2014) believed that the Arctic cyclone activity is linked to the distributions of lower troposphere temperature and sea ice and they reported that a great Arctic cyclone appeared in the summer of 2012 Rudeva 2012, 2014). The global climate models of Coupled Model Intercomparison Project (CMIP) showed Arctic cyclone activity under historical and future climates (Vavrus 2013, Nishii et al 2015, Song et al 2021. These models have estimated that the number and intensity of most AECs will increase in the future (Tao et al 2017, Priestley andCatto 2022), but they have not focused on environmental factors that will contribute to AEC formation in the future.
Therefore, in this study, we established GPI Arctic , which is a genesis potential index for AEC, to assess the possible development potential of AECs and analyze the contribution of specific environmental factors to the development of AECs. In section 2, the data and methods used in this research work are introduced. In section 3, the basic characteristics of AECs and the possible causes of AECs are analyzed, and the GPI Arctic is established. In section 4, the historical and future contribution of various environmental factors of GPI Arctic is discussed.

Data
In this study, the 1979-2020 reanalysis data from European Center for Medium-range Weather Forecasts ERA5 Reanalysis (www.ecmwf.int/en/forecasts/ dataset/ecmwf-reanalysis-v5) was used to detect AECs and calculate necessary large-scale environmental variables. Some of these variables include sixhourly mean sea level pressure (SLP), monthly wind velocity, omega, and long-wave radiation data, with a spatial resolution of 2.5 • (Kanamitsu et al 2002). The AO index was obtained from the NOAA Physical Sciences Laboratory (https://psl.noaa.gov/data/ correlation/ao.data). Song et al (2021) evaluated the ability of CMIP6 global climate models to simulate Arctic cyclones and showed that EC-Earth3, EC-Earth3-Veg, MMRI -ESM2-0, and NorESM2-MM outperformed other models. Thus, we used monthly large-scale environmental data, i.e. wind velocity, omega and long-wave radiation (https://esgf-node.llnl.gov/projects/cmip6/), of the mid-21st century (2021-2050) from these four models. The basic information of the models is summarized in table 1. These models include four scenarios, i.e. SSP126, SSP245, and SSP585, that represent different combinations of possible future mitigation and adaptation challenges (Huang et al 2019). We only selected the first variant label (r1i1p1f1) of the ensembles for each model. All the model data were interpolated onto a 1 • × 1 • grid before analysis. The multi-model mean (MMM) is the arithmetic mean of the four model simulations.

Methods
The detection and tracking of cyclones rely on previously described algorithms (Zhang et al 2004, Crawford andSerreze 2016). With respect to AECs, the intensity of Arctic cyclones has significant seasonal variability; for example, the intensity of winter Artic cyclones is significantly stronger than that in summer (Keegan 1958, Serreze and Barry 1988, Serreze 1995, Zhang et al 2004. If a unified threshold is used to define extreme cyclones in each season, almost no extreme cyclones will be reported in summer . Therefore, we defined the thresholds of the cold semester (October-March) and the warm semester (April-September) separately. In this study, we considered the minimum SLP of 6 h in the Arctic (north of 60 • N) during the cold and warm semesters from 1979 to 2020, and the threshold of AEC was obtained by the pressure corresponding to the 5th percentile. Accordingly, the cold semester threshold is 953.7 hPa, and the warm semester threshold is 972.3 hPa.
Although the (thermal) dynamics of Arctic and tropical cyclones are very different, similar parameters and methods are considered and used (Murakami et al 2011, Li et al 2013, Zhang et al 2016. Regarding parameter selection, some factors, such as absolute vorticity, omega, wind shear, long-wave radiation, sea ice concentration, relative humidity, mixed layer depth, etc., that generally affect cyclone development as candidate factors are considered in GPI Arctic . Four parameters, namely absolute vorticity, omega, wind shear, and long-wave radiation, related to the development of Arctic cyclones in cold and warm semesters were screened out by empirical and regression methods (Ballabrera-Poy et al 2002). Among these parameters, long-wave radiation denotes downward longwave radiation, which is more related to AECs trend change.
EGR is used as a measure of baroclinicity, and therefore, the ability of the environment to foster cyclonic systems is associated with baroclinic instability to vertical wind shear and static (in)stability Lim 2009, Simmonds andRudeva 2012).
EGR was calculated as follows (Simmonds and Lim 2009): ( 1) where f denotes the Coriolis parameter, V (z) denotes the vertical profile of the eastward wind component, and N denotes the Brunt-Väisälä frequency. In this study, we calculated EGR at 500 hPa. By taking EGR (500 hPa) as the control index, the exponential coefficients a, b, c, and d are calculated over the Arctic using the least square fitting, and the coefficient p enables GPI Arctic has the best least square fit with the EGR. The terms in the equation are in the normalized format after calculating the probability density Here, η denotes absolute vorticity at 850 hPa, ω denotes omega at 500 hPa, v s denotes wind shear, and F denotes long-wave radiation. In the above equation, a = 2.7, b = 5.4, c = 1.86, d = 0.13, and p = 0.17.

Frequency and distribution characteristics of AECs
A total of 711 AECs were detected by different thresholds in cold and warm semesters from 1979 to 2020, made up of 359 AECs in the cold semester and 352 AECs in the warm semester. Figures 1(a) and (b) show the track distribution of AECs in the cold and warm semesters, respectively. As shown in the figure, AECs in the cold semester are concentrated in box A, which is mainly located in the Arctic-Atlantic sector, including the East Greenland Sea, the Norwegian Sea, and the Barents Sea. In the warm semester, AECs in box A were active but with significantly smaller track densities than those in the cold semester. In comparison, warm semester AECs in box B (the northern side of Eurasia) were more active than those in box A. In addition, the number of AECs for the whole year was also exhibited considerable variability ( figure 1(c)). There were years in which AECs were active in both cold and warm semesters (such as 1989, 1990, and 2020, etc), and there were also years in which AECs were more active only in the cold semester (warm semester). In particular, AECs were detected only in the warm semesters, while the AECs were not detected in the cold semesters in 1987 and 2010. The number of AECs was not statistically significant over the 42 years, although there was a slight trend (whole year: 0.06 events/year, cool semester: −0.02 events/year, warm semester: 0.08 events/year).

Relationship between AECs and large-scale circulation background
In the cold semester, box A is the region with the most frequent Arctic cyclone events, i.e. a large number of cyclones are formed in this region. In addition, this region is associated with Icelandic low pressure, which is a semi-permanent low pressure in the North Atlantic Ocean that is described as the statistical manifestation of many cyclones migrating into the region, forming in the region and dying in the region (Serreze 1995). Although the semi-permanent Aleutian Low is observed in the North Pacific, very few cyclones enter the polar circle from the North Pacific because the region is further south than the Iceland Low and is blocked by the North American continent. Figure 2(a) shows the distribution of the 42 year average SLP in the cold semester, which is obviously controlled by the Icelandic low in box A. The Icelandic low is an important factor influencing the formation and development of Arctic cyclones during the cold semester. In the warm semester, the temperature gradient between the pole and the equator weakens, resulting in weaker mid-tropospheric standing waves and Icelandic low (Serreze et al 1993, Serreze 1995. Moreover, the activity of AECs in box A decreases. However, box B in the warm semester is no longer controlled by the Siberian cold high pressure, resulting in increased temperature, decreased SLP ( figure 2(b)), and increased activity in the AECs.
The Arctic cyclone activity is controlled by the large-scale air circulation in the middle and high latitudes of the northern hemisphere, and AO is a key factor in the air-sea interaction at high latitudes. The empirical orthogonal function (EOF) and AO index of the SLP were calculated in both cold and warm semesters, respectively, in order to show the intra-seasonal changes more clearly (figures 2(c)-(f)), and all the EOF calculations shown in this study passed the North test (North et al 1982). In the cold semester, the EOF1 explains 54.8% of the domainintegrated SLP variability and captures the positive AO phase pattern and the decrease in SLP in the Arctic. This decrease in SLP is particularly obvious in box A, where its maximum decrease is 2 hPa. The time series of principal component (PC) 1 has a similar trend change with the AO index and AECs in the cold semester, and the correlation between AECs, AO index, and PC1 reaches more than 0.7 (table 2). In the warm semester, the EOF1 with an explained variance of 38.1% is characterized by a general decrease in SLP during the positive AO phase pattern, but the degree is weaker than that in the cold semester. Near Iceland, the SLP decreased by a maximum of 0.94 hPa. The correlation between the AECs, AO index, and PC1 in the warm semester were 0.42 and 0.52.

Possible factors affecting summer AECs
As shown in figure 2, the explained variance of EOF1 is smaller in the warm semester than in the cold semester the warm semester AECs will be affected by more other factors, as shown in figures 3(a) and (b). The warm semester EOF2 also explains 16.5% of the Arctic SLP variability, with a decrease in SLP on the northern side of Eurasia, up to 0.74 hPa. A strong horizontal temperature gradient extending from 41 • E to 126 • W along the Arctic coastline of Eurasia and western North America during the warm semester is defined as the Arctic front region (AFZ) (Crawford and Serreze 2015Serreze , 2016. The generation of AFZ is related to the obvious sea-land thermal difference in warm semester, and the strong baroclinic instability associated with AFZ contributes to the generation of AECs near box B. Figures 3(c) and (e) show the spatial distribution of mean meridional temperature gradient and EGR in typical years of EOF2. It can be found that mean meridional temperature gradient and EGR are significantly higher than those in atypical years (figures 3(d) and (f)), which means that there will be more AEC generation in warm semesters. We attempted to establish a genesis potential index for AEC to quantify the effects of various factors on AEC generation in both cold and warm semesters.

A genesis potential index for AECs
GPI Arctic (equation (2)) is established by fitting parameters with EGR. Figure 4 shows the interannual variation of GPI Arctic in the past 42 years.  GPI Arctic in the cold semester is generally higher than that in the warm semester, which is more conducive to the formation of AECs. In addition, GPI Arctic and AEC frequencies show similar interannual variability (figure 1). The correlation between GPI Arctic and AECs in the cold semester is 0.81, while  this correlation in the warm semester is lower than 0.3, probably because many non-locally formed cyclones enter the Arctic from Eurasia in the warm semester (Zhang et al 2004, Serreze and Barrett 2008, Crawford and Serreze 2016. Moreover, GPI Arctic has obvious seasonal and spatial distribution differences (figure 5), which is obviously similar to the distribution of EGR pointed out by (Simmonds and Li 2021). The high GPI Arctic values in the cold semester are concentrated in southern Greenland of box A, and the GPI Arctic is generally weakened in the warm semester. High GPI Arctic values still exist in southern Greenland in the warm semester, but the range and intensity of AECs are significantly reduced. GPI Arctic along the Eurasian coastline of box B increases significantly ( figure 5(c)), and the frequency of AECs increases.

Relative contributions of each individual environmental factor
Equation (2)  . ( Applying the total differential to both sides of equation (3) yields: Integral equation (4) yields the following expression from the annual average to a specific month: where,α 1 ,α 2 ,α 3 ,α 4 are assumed to be constant coefficients, which are expressed as follows: where the bars denote the annual average, and δ denotes the difference between an individual month and the annual average in the equation. The four terms on the right of equation (5) represent the contribution of each environmental factor to GPI Arctic .
As shown in figure 6, the monthly contribution of each environmental factor to box A and box B is significantly different. In Box A, the δDPI Arctic in the cold semester is generally positive compared with the climatology, and its maximum value in January is 0.19. Wind shear, absolute vorticity, and omega offset the adverse conditions of long-wave radiation, which is conducive to the development of AECs. The average δDPI Arctic in the warm semester is less than 0, indicating weak conditions for the formation of AECs. Other environmental factors, except for longwave radiation, inhibit the development of AECs. The variation in δDPI Arctic in Box B is smaller than that in box A, and the maximum value of δDPI Arctic in July is only 0.02. However, it is associated with greater and more complex variations of environmental factors. There are favorable and unfavorable months for the formation of AECs in both cold and warm semesters. In the warm semester, long-wave radiation plays a positive role in the development of AECs, while in the cold semester, wind shear plays a positive role in the development of AECs. In recent years, the rate of Arctic warming and sea ice melting has accelerated (Simmonds and Li 2021, Rantanen et al 2022, Zheng et al 2022, which may have contributed to the greater positive effect of long-wave radiation on the formation of Arctic cyclones during the warm semester ( figure 8(b)). And many researchers have stated that that Arctic cyclone activity is closely related to the sea ice and atmospheric circulation (Simmonds and Keay 2009, Koyama et al 2017, Valkonen et al 2021.

CMIP6 predicts the change of GPI Arctic in the mid-21st century
In the CMIP6 global climate model, EC-Earth3, EC-Earth3-Veg, MRI-ESM2-0, and NorESM2-MM  Figures  (a)-(c) are cold semester, warm semester, and the difference between cold and warm semesters under the SSP126 scenario respectively; (d)-(f) are cold semester, warm semester, and the difference between cold and warm semesters under the SSP245 scenario respectively; (g)-(i) are cold semester, warm semester, and the difference between cold and warm semesters under the SSP370 scenario respectively; (j)-(l) are cold semester, warm semester, and the difference between cold and warm semesters under the SSP585 scenario respectively. models showed better performance of Arctic cyclone characteristics. Their MMM shows the possible changes in GPI Arctic and environmental factors in the mid-century (2021-2050) under the scenarios SSP126, SSP245, SSP370, and SSP585, respectively (figures 7 and 8). Figure 7 shows that, compared with the reference period of 1979-2020 (figure 5), the GPI Artic values of box A and box B will increase to a certain extent in both semesters in the next 30 years, and more extreme cyclones may be formed. The maximum number of AECs may be formed under the SSP585 scenario (figures 7(j) and (k)). The future monthly contribution distribution (figure 8) is generally similar to that in figure 6, but the amplitude of δGPI Arctic in box A smaller, and that of δGPI Arctic in box B is larger. This indicates that the comparison of environmental factors in the cold and warm semesters in box B may be more obvious in the future, and the seasonal differences in AECs frequency will be greater. Among them, the increase in δDPI Arctic in Box B in July and August, respectively, is more than 0.05, indicating that the overall environment may be more conducive to the formation of AECs (figure 8).

Conclusions
In this study, the reanalysis data provided by ERA5 Reanalysis were used to detect Arctic cyclones from 1979 to 2020, and two different thresholds were set to define the AECs in the cold and warm semesters. A total of 359 and 352 AECs were detected in the cold and warm semesters, respectively. An Arctic cyclone genesis index, GPI Arctic , was established through the least squares fitting of parameters such as absolute vorticity, omega, wind shear, and long-wave radiation to EGR, and the relative contribution of each largescale environmental factor was studied.
AECs are mainly concentrated in southern Greenland within box A during the cold semester, while AECs also appear in box B during the warm semester. AECs are closely related to AO, and their genesis is mainly influenced by EGR. In the cold semester, the correlation between GPI Arctic , and AECs reached 0.81. The disturbance variability of AECs becomes larger in the warm semester, and the correlation between GPI Arctic and AECs is weak. Among large-scale environmental factors, wind shear and long-wave radiation strongly affect δGPI Arctic . The combined effects of various environmental factors determine the different δGPI Arctic in box A and box B (figure 6), which leads to significant differences in the space and time of AECs formation. The MMM of the four CMIP6 models with a better ability to reflect the characteristics of Arctic cyclones shows that more AECs may be formed by the mid-century. The seasonal contrast in the next climatology will be stronger in northeast Eurasia.
All data that support the findings of this study are included within the article (and any supplementary files).