Northern Pacific extratropical cyclone variability and its linkage with Arctic sea ice changes

Extratropical cyclones are critical weather systems affecting climate change in mid and high-latitude regions. Researching the characteristics, patterns, and movements of extratropical cyclones is helpful for improved prediction and understanding of global climate change. Currently, there are still great difficulties in predicting extratropical cyclones. Also, few prior studies have investigated the potential impact of Arctic sea ice on extratropical cyclone activity (ECA). This study utilizes updated ERA5 data and an improved ECA identification method to reveal ECA in the Pacific. The results demonstrate that the Pacific ECA primarily occurs during the cold season (November to March), and the North Pacific region has the maximum ECA. More remarkably, a strong linkage exists between the preceding summer-fall anomalous changes in the Arctic sea ice and the cold season Pacific ECA. We discover that Arctic sea ice could modify the local pressure field, changing the southern boundary of the Pacific sector polar vortex in winter, which in turn influences the intensity of the westerly jet stream and eventually impacts the Pacific ECA during the cold season. Our exploration will provide references for further study and prediction of ECA in the Pacific Ocean.


Introduction
Extratropical cyclones are critical weather systems, and their activities cause devastating weather in middle and high latitudes with powerful winds, heavy rain, cold, and snowstorms, which are important large-scale weather systems influencing the human living environment (Averkiev and Klevannyy 2010;Feser et al. 2015).However, it is challenging to predict extratropical cyclones' activities precisely because of the abruptness.As a result, experts from all over the world are paying close attention to it, researching and analyzing it using various techniques in the hopes of precisely predicting it.
Extratropical cyclone activity (ECA) prediction has received much attention and made advancements in recent decades in terms of long-term projection and short-term forecasting.However, because extratropical cyclones are unpredictable in their physical properties and certain operational forecast techniques are sophisticated, it is tough to make accurate predictions of these storms (Marshall et al. 2009;Froude 2011;Maycock et al. 2011).Although satellite cloud images, weather maps, and other numerical forecasting products have undergone extensive research and testing, the prediction capacity still needs to be improved because of limitations in data, observations, and models, among other factors (Pinto et al. 2006;Raible 2007).
In recent decades, scientists have primarily adopted sea level pressure (Mendes et al. 2010;Pepler and Dowdy 2020) or geopotential height fields (Hirschberg et al. 1991;Lu et al. 2020) to track and analyze extratropical cyclones.But these analytical methods have limitations, such as large-scale systems (e.g., Icelandic low pressure) affecting the surface pressure field and strong background currents (e.g., subtropical jet streams) obscuring fast-moving weak systems.These methods can cause significant bias in the identification of extratropical cyclones.They must first experience considerable growth before being assessed and monitored objectively.Only slowly moving, large-scale low-pressure systems can be adapted by the geopotential height field and 1 3 sea level pressure index (Hoskins and Hodges 2002).However, the abovementioned issues can be solved using a more objective and reasonable ECA definition.To define ECA objectively, Zheng et al. (2019) employed mean sea level pressure (MSLP) data and a filter with a 24-h difference, known as ECApp, in Eq. 1.
The 24-h MSLP variance data have been used to assess "storminess" widely because of the benefit that the bandpass filter gives half power points at intervals of 1.2 and 6 days, which is an extratropical cyclone life cycle and highlights synoptic time scale variability (Colle et al. 2013).
Prior study has shown that the maximum from this filter usually occurs over locations where extratropical cyclones are frequently passing (Zheng et al. 2019).Therefore, this method could be used as an appropriate ECA indicator in our investigations.Here, we make a further improvement by removing the corresponding values when the value of the latter moment is greater than that of the previous one, thus eliminating the anticyclonic activities.
Several phenomena have been demonstrated in previous investigations, which may be the causes of ECA modulation (Guo et al. 2017;Attard and Lang 2019;Zhao et al. 2020).El Niño-Southern Oscillation (ENSO) in the Northern Hemisphere (NH) may significantly alter ECA (Eichler and Higgins 2006;Ulbrich et al. 2009).El Niño events are connected to the eastward and equatorward migrations of the boreal winter ECA in the Pacific and a weakening of ECA in North America (Yoshiike and Kawamura 2009), whereas La Niña occurrences are connected to the opposite developments (Eichler and Higgins 2006).Furthermore, the Rossby waves caused by the Madden Julian Oscillation (MJO) that propagate into the mid-latitudes significantly impact ECAs across the North Pacific (NP; Lee and Lim 2012).This may occur when tropical diabatic heating takes occurring in different locations.The ECA fluctuates in the NH upper troposphere due to the quasi-biennial oscillation (QBO), which suggests that the NH polar vortex also impacts ECA (Kidston et al. 2015).
As noted, air pressure, temperature, and atmospheric circulation influence ECA in the NH, especially at high latitudes.On the other hand, the variability of Arctic sea ice considerably affects circulation, air pressure fields, and highlatitude air temperature (Vihma 2014;Chen andSun 2022, 2023).For instance, Prior studies have shown how Arctic sea ice can affect global atmospheric circulation (Overland and Wang 2010;Budikova 2009), winter storms and sea ice in relation to the North Atlantic Oscillation (NAO) (Dawson et al. 2002;Bader et al. 2011), and strong correlation between winter sea ice concentration anomalies in southwest Greenland and summer Eurasian atmospheric circulation in (1) ECApp = {MSLP(t + 24 hr) − MSLP(t)} 2 .the coming year (Wu et al. 2013).Consequently, Arctic sea ice variability probably has an effect on the ECA in the NH through certain "bridges", whereas these "bridges" have not yet received enough attention.
Despite the fact that there are extensive ECAs in the NP and Atlantic ( Lambert 1996;Mailier et al. 2007;Villamil-Otero et al. 2018), there are few studies specifically on ECAs in the NP, and even fewer studies have been conducted on how Arctic sea ice affects ECA.Therefore, as a starting point for investigation, this paper concentrates on the multi-year variability of ECA in the NP and the connection with the rapid decline in Arctic sea ice.
We also consider the variability of rapidly diminishing Arctic sea ice, particularly the unprecedented pace of Arctic summer and fall sea ice melting in recent decades under global warming (Chen et al. 2016).As early as 2012, it was predicted that with the acceleration of global warming, the Arctic summer sea ice would likely disappear by 2030 (Wang andOverland 2009, 2012).According to some research, ice-free summers may occur in approximately 2040 (Holland et al. 2006).The rapid disappearance of Arctic sea ice will have a significant impact on the Arctic environment and mid-latitude climate change (Blackport et al. 2019), particularly on the frequency of harsh winter weather (cold waves and intense cold air incursion) across Eurasia (Tang et al. 2013).Extratropical cyclone activity in the Pacific Ocean will also be impacted and will undoubtedly change as a result, although there has yet to be much research done on how sea ice influences ECA.This study will focus on the potential link between Arctic sea ice and the NP ECA, which is anticipated to offer another method and point of reference for ECA forecasting and prediction.

Data
We utilized ERA5 data from the European Centre for Medium Range Weather Forecasts (ECMWF) for the years 1979 to 2021 with a resolution of 0.25° × 0.25° (Hersbach et al. 2020), which has the greatest skill level for almost all cyclone features, including cyclone position, strength, and propagation speed.The variables incorporated include geopotential height, zonal wind u, meridional wind v, and MSLP.Additionally, we used Hadley Centre Arctic sea ice concentration (ASIC) with a spatial resolution of 1.0° × 1.0° from 1979 to 2021 as the data to study the changes in Arctic sea ice (Rayner et al. 2003;Johnson et al. 2007).

ECApp calculation
As the introduction mentions, ECApp is described using a 24-h difference filter on MSLP data (Eq.1).The 24-h MSLP variation statistics have been used extensively to quantify "storminess" due to the benefit that the bandpass filter contains half power points at intervals of 1.2 and 6 days, which is an extratropical cyclone life cycle that shows synoptic time scale variability.According to the previous study, the Northern Hemisphere, which includes much of Europe and North America, may experience precipitation and high wind events that are significantly correlated with the ECApp (Ma and Chang 2017).

Calculation of the polar vortex index
Using the definition of the polar vortex by the National Climate Center of the China Meteorological Administration, the southern boundary of the polar vortex in the Pacific sector was determined using the 550 gpm contour at 500 hPa geopotential height (http:// cmdp.ncc-cma.net/).
In addition, conventional analysis methods such as singular value decomposition (SVD) analysis (Akritas and Malaschonok 2004), empirical orthogonal function (EOF) analysis (Hannachi et al. 2007), composite analysis (Bauer Del Genio 2006), and wavelet analysis (Torrence and Compo 1998) are also involved in this study.
All the data in our investigation was converted into anomaly fields by removing the 1979-2021 mean and then detrended prior to analyses.

Characteristics of Pacific extratropical cyclone changes
Based on the mentioned calculation method, the daily Pacific ECA changes from 1979 to 2021 were calculated, and the daily climatology changes were obtained accordingly (Fig. 1).The figure shows that Pacific extratropical cyclones are most frequent from November to March, with summer having a lower probability of occurrence.Based on the findings of Fig. 1, we focus on the relationship between ECA from November to March (hereinafter referred to as the "cold season") and preceding summer-fall Arctic sea ice variation in this research.
Figure 2 displays the cold season changes and multi-year climatology variations of ECA in the Pacific area (including the Arctic sector).In the northern Pacific Ocean, the maximum ECA is situated south of the Aleutian Islands, and the axis of maximum activity is distributed along the latitudinal direction.This result generally agrees with the analysis using the number of NP extratropical cyclones (Zhang et al. 2012), which indicates that the abnormal variation of ECA in this region is very significant.The ECA anomaly in the 100° E-110° W, 10-80° N region is chosen for EOF analysis in order to retrieve the primary variation information of extratropical cyclones in the Pacific region.According to the findings of the EOF analysis, the leading two modes contribute 23.6% and 14.9% of the total variance, respectively.The first mode has a dipole distribution, i.e., the NP is characterized by changes in the opposite phase to the polar region, and when extratropical cyclones south of 60° N appear more, extratropical cyclones north of 60° N show a decreasing trend.In contrast, extratropical cyclones north of 60° N appear more, while extratropical cyclones south of 60° N appear less (Fig. 3a).The second mode is characterized by a tripole, i.e., a tripole formed by the NP, Central Pacific, and Polar Regions that exhibit spatial variance in extratropical cyclones.On the other hand, their northern and southern latitudinal belts have a trend of more extratropical cyclones when the range of extratropical cyclones between 40 and 60° N appears to be smaller (Fig. 3b).The long-term variation of the first mode exhibits a weakly growing tendency, whereas the variation of the second mode tends to decrease weakly, as can be shown from the regression analysis of the time series of the first two modes (Fig. 3c and d).Overall, the ECA variation in the Pacific sector is characterized by significant interannual variability.The spatial distribution is slightly different from that in the paper (Zhang et al. 2012), but the overall variation trend is the same.
To compare with the results from Zhang et al. (2012), the variation characteristics of ECA were calculated for different latitudes in the Pacific sector of the Northern Hemisphere. Figure 4 shows the average variation of ECA along the latitudes in 30-60° N and 60-90° N, respectively (110-270° E).It can be seen that the mid-and highlatitude ECAs are characterized by significant interannual and interdecadal variability, and the ECA at middle and high latitudes shows opposite variation, with an increasing trend of ECA at middle latitudes (30-60° N) and a decreasing trend of ECA at high latitudes (60-90° N).In addition, the wavelet analysis and power spectrum show that the significant period of ECA at mid-latitudes is 10 years, while the significant period of ECA activity at high latitudes is 16 years (Fig. 5).Therefore, it can be concluded that there is a considerable difference between mid-latitude and high-latitude ECAs.Overall, our results are basically consistent with those of Zhang et al. (2012), except for weaker ECAs in East Asia and South America probably due to different time series and datasets.

Potential mechanism of Pacific extratropical cyclone change
Arctic sea ice is characterized by remarkable seasonal changes, freezing in winter (December to March) and melting in summer and fall (June to November).Previous studies have suggested that the increase in Arctic surface temperature may be the result of a decrease in Arctic sea ice.This will result in reduced temperature gradients between mid and high latitudes and weakened meridional circulation (Oudar et al. 2017).The latter will to some extent weaken the baroclinic action of the mid-high latitude atmosphere (Jaiser et al. 2012; Simmonds and Li 2021), which would have implications for the strength of the polar vortex and the jet stream (Barnes and Screen 2015).Mid-and highlatitude variations of the polar vortex and high-altitude jet Previous research has shown that changes in Arctic sea ice may alter the phase of NAO and Arctic Oscillation (AO) patterns, which in turn can impact the temperature in Eurasia (Hopsch et al. 2012;Vihma 2014).According to the results and further supported by the numerical simulations, when Arctic sea ice melt is considerable in summer and fall, it is likely to trigger the NAO's negative phase (Wu and Zhang 2010) and favor cold extremes over midlatitudes (Peings and Magnusdottir 2014).
Arctic summer-fall sea ice has been shrinking at an alarming rate over the past few decades, according to satellite assessments (http:// nsidc.org/ arcti cseai cenews/), and this phenomenon will undoubtedly have a substantial influence on the climate in the middle and high latitudes.To this end, based on previous analyses, we use the summerfall Arctic sea ice (all the following are summer-fall sea ice, referred to as sea ice) changes and the NP cold-season ECA (all the following are cold-season ECA, referred to as ECA) to analyze and try to reveal the possible connection between Arctic sea ice changes and ECA in the NP region.
To understand the connection between the cold season ECA and summer-fall Arctic sea ice, we conducted an SVD analysis of ECA and Arctic sea ice.The first coupled mode of the SVD analysis of the Pacific ECA and Arctic sea ice is depicted in Fig. 6a.The first mode variance contribution is 32.7%.It can be seen that the sea ice in the left field of the Pacific sector (Arctic sea ice), the East Siberian Sea, the Chukchi Sea, and the Beaufort Sea are positive anomalies, while the sea ice in Baffin Bay, the Greenland Sea, and the Barents Sea around Greenland are negative anomalies.The Pacific and Atlantic sectors make up an opposing phase variation, which aligns with the Arctic dipole pattern.This spatial distribution pattern is similar to the first mode obtained from the EOF decomposition in Fig. 3, which indicates that the Pacific temperate cyclone is still characterized by mid and high-latitude antiphase variation.According to this result, when there is a negative sea ice anomaly in the Pacific sector of the Arctic and a positive sea ice anomaly in the Atlantic sector during the summer and fall, there is a negative anomaly in the Pacific The coupled spatial distributions of summer-fall Arctic sea ice and cold season NP ECA both show dipole distributions (Fig. 6a and b).Namely, when the summer and autumn sea ice changes show an anomalously high distribution in the Pacific sector and an anomalously low distribution in the Atlantic sector, the Pacific cold season ECA is more in the mid-latitude and less in the low-latitude, and vice versa.The coupled spatial distribution, along with the time coefficients in Fig. 6c), shows that the Arctic summer and fall sea ice was primarily in a positive phase change before 2004, while it was in a negative phase change after 2004.Correspondingly, the Pacific ECA was also in a positive phase change before 2004 and a negative one after 2004.That is, when the Arctic sea ice was in a positive phase in summer and autumn before 2004, the Pacific cold season ECA was more at mid-latitudes and less at high latitudes.After 2004, when the Arctic sea ice was in a negative phase in summer and autumn, the Pacific cold season ECA was less at mid-latitudes and more at high latitudes.
To clarify the underlying connection between Pacific ECA and Arctic sea ice, the Pacific ECA corresponding to more (less) sea ice is selected for composite analysis.Figure 7 reveals the spatial distribution characteristics of the Pacific ECA corresponding to the years with more and less Arctic sea ice.The figure shows that the ECA at high Pacific latitudes is less, the ECA at mid-latitudes is more, and the reversed-phase change happens at middle and high latitudes when the Arctic sea ice shows positive anomalies.When there is a negative anomaly in the Arctic sea ice, there is more ECA at high Pacific latitudes and less ECA at midlatitudes, which is a reversed-phase change of ECA at midand high latitudes.
To further explore the internal linkage between Pacific ECA and Arctic sea ice, we selected the anomalous mean sea ice variation in the Pacific sector in the first mode of Arctic sea ice was obtained from the SVD analysis in Fig. 6 as the key sea ice area.For composite analysis, the atmospheric circulation field corresponding to the more (less) key area's sea ice concentration is extracted.We will reveal the impact mechanism between Arctic sea ice and Pacific ECA.
Figure 8 shows the spatial distribution of the anomalously more (less) sea ice in the key Arctic region, the composite 500 hPa geopotential height field, and the variations in the polar vortex and westerly jet stream.The 500 hPa geopotential height field anomalies show opposite spatial distributions when the sea ice in the Arctic key area is anomalously more or less.When the sea ice in the critical Arctic region is positive, the 500 hPa geopotential height field in the NP region has a negative anomaly.The Eurasian, North American, and Nordic regions have positive anomalies.The 500 hPa geopotential height field over the NP region is a noticeably positive anomaly, while the sea ice in the Arctic key area is a negative anomaly.The Eurasian, North American, and Nordic regions are negative anomalies.The amount of ECA in the Pacific area will be directly impacted by the distribution of this kind of quasi-AO anomaly.In other words, anomalous low pressure at the sea surface will occur and trigger ECA when there is a negative anomaly in the 500 hPa geopotential height field in the NP area; When positive anomalies appear in the 500 hPa geopotential height field in the NP region, anomalously high sea surface pressure will appear, suppressing ECA.In addition, when there is a positive sea ice anomaly in the key area, the position of the southern boundary of the polar vortex in the Pacific sector is southward, and the westerly jet stream in the Pacific enhances and expands southward and eastward; When there is a negative sea ice anomaly in the key area, the position of the southern boundary of the polar vortex in the Pacific sector is northward.The Pacific westerly jet is weakened and moves northward and westward.The strengthening and weakening of the Pacific westerly jet stream also play a critical role in the Pacific ECA (Nakamura et al. 2004;Luo et al. 2007), and the NP ECA will be more frequent when the westerly jet stream strengthens and moves southward; When the Pacific jet stream weakens and withdraws northward, the NP ECA decreases.
According to the analysis's findings, the key area's Arctic sea ice fluctuations have a critical impact on the Pacific ECA, while the Pacific's quasi-AO distribution has a considerable excitation and suppression effect on the Pacific ECA.When the quasi-AO phase is positive, the Pacific ECA decreases, and when it is negative, it increases.The East Asian westerly jet stream and the Pacific sector polar vortex are the crucial factors directly affecting the ECA in the Pacific area, which is a significant mechanism causing the ECA in the Pacific region and a vital connection between the Arctic sea ice and the Pacific ECA.

Summary
The latest ERA5 reanalysis data and the more objective approach for estimating ECA are used in this research to determine the frequency of Pacific ECA during the cold season.The EOF and SVD analyses, as well as composite analysis, are used to reveal the multi-year variation characteristics of ECA in the Pacific region during the cold season and the variation characteristics of different phases at middle and high latitudes, and to give the respective cycles and multi-year variation trends of ECA at the middle and high latitudes.Using this as a foundation, the discussion turns to the Pacific ECA variance throughout the cold season and its potential relationship with the summer and fall Arctic sea ice.
(1) ECA in the Pacific Ocean mainly occurs from November to March, with the largest activity area located in the NP, south of the Aleutian Islands.
(2) There is a north-south (mid-and high-latitude) reversed-phase pattern of ECA in the Pacific Ocean during the cold season, and there is a significant difference in the variation of ECA south and north of 60 (5) However, there are still some issues to be solved.In the middle and high latitudes, for instance, what is the process causing the several phase shifts of the Pacific cold season ECA?What is the mechanism of the influence of Arctic sea ice on ECA activity in different phases at middle and high latitudes in the summer and autumn of the previous year?All these issues need to be further explored in the follow-up work.

Fig. 1 Fig. 2 aFig. 3
Fig. 1 The daily climatology of ECA in the North Pacific.ECA is the calculation by daily mean sea level pressure (MSLP) data in ERA5 from 1979-2021 in the Pacific area (100-270° E; 20-90° N)

Fig. 6
Fig. 6 SVD analysis of the Arctic sea ice in the summer-fall and Pacific area ECA in the cold season(Nov-Mar).a Arctic sea ice.b ECA.c Time series of the Arctic sea ice and ECA.The time correlation coefficient of the two fields is 0.84

Fig. 7
Fig.7ECA anomalies distribution corresponding to Arctic sea ice variation anomaly using composite analysis a more ice, b less ice.The green dots are significant at the 95% confidence level (t-test)

Fig. 8
Fig. 8 Composite 500 hPa geopotential height corresponding to the sea ice variation anomaly in the key region.a Distribution of positive sea ice anomalies.b Distribution of negative sea ice anomalies.The color shadings are the 500 hPa anomaly height (gpm), the purple line is the southern boundary (550 gpm) of the mean polar vortex in winter, and the green line is the southern boundary (550 gpm) of the