Strengthened impact of late autumn Arctic sea ice on Asian winter cold extremes after 1999/2000

Winter cold extremes (WCEs) frequently plague densely populated areas of Asia, leading to substantial economic losses and even fatalities. It has been found that the late autumn sea ice concentration (SIC) anomalies in the northern (SICN) and southern Arctic (SICS) are significantly positively and negatively correlated with the occurrence frequency of WCE in Asia, respectively (Wang and Su 2024). Our study demonstrates that the impacts of SICN and SICS have strengthened after 1999/2000. Specifically, before 1999/2000, the influences of SICN and SICS on the Asian WCE (AWCE) were relatively weak, possibly related to the weak intensity of SICS and the limited correlation between SICN and SICS. After 1999/2000, the interannual variability of SICS became larger and anti-correlated with that of SICN, resulting in a stronger teleconnection between the Arctic SIC and AWCE. It is revealed that after 1999/2000, the greater loss of SICS modified atmospheric stability through changes in surface heat fluxes and surface upward longwave radiation fluxes. This alteration weakened the magnitudes of westerly winds and increased the frequency of blocking events over the northern Eurasian continent, leading directly to a higher occurrence of cold extremes in Asia. These interdecadal differences in the influence of Arctic SIC on AWCE may be associated with long-term climate change.


Introduction
Asia is one of the regions that is densely populated and frequently suffers from winter cold extremes (WCEs).When WCE occur, they often cause significant damage to various aspects of society, particularly during the mid-winter period (January-February, shortened as JF), such as the freezing of 2008 over the Middle East-East Asia, rare cold events of 2011 and record 'boss level' cold wave over East Asia in 2016 [1][2][3][4].These events are often associated with a dominant circulation component called atmospheric blocking, which refers to a quasi-stationary anticyclone over the mid-high latitudes and plays a crucial bridging role in the influence of other factors on cold extremes [5,6].By intensifying the downstream cold advection, the blocking impacts the occurrence of WCE in the midlatitudes [2,[7][8][9].Furthermore, by using both reanalysis data and simulations, some studies have attributed more WCE to the Arctic sea ice anomalies through inducing an increase in higher latitudinal blocking events [8,10,11].The potential physical process underlying this linkage is as follows: Arctic sea ice anomalies influence local atmospheric stability via affecting surface outgoing longwave radiation and upward heat fluxes [12][13][14].These changes in atmospheric stability then induce alterations in the westerly jet stream [15][16][17].A more wave westerly jet stream favors the occurrence of blocking events over higher latitudes [10,11], which impacts the WCE occurrence in the mid-high latitudes of the Northern Hemisphere [10,11,18,19].
By anchoring a blocking anticyclone, Arctic sea ice plays a crucial role in some teleconnection patterns leading to enhanced cold anomalies over East Asia [20].And the Eurasian cooling during Ural blocking, and high-latitude European blocking events is associated with a decrease in sea ice concentration (SIC) over the Barents-Kara Seas [21].On the other hand, there exist several studies that do not support the view that Arctic sea ice affects winter cooling in mid-high latitudes [22,23].
Interdecadal changes in the impacts of Arctic sea ice on mid-high latitudinal climate may be one of the reasons for the divergence of viewpoints [24].Specifically, the impact of autumn Arctic SIC around the Kara-Laptev seas on the subsequent winter Siberian High has disappeared since the late 1990s [25]; while autumn Arctic SIC around the Lapteveastern Siberian-Beaufort Sea has made an enhanced impact on the following spring Arctic Oscillation after the mid-1990s [12], as well as the East Asian winter monsoon after the early 1990s [26].About this time, the Arctic sea ice extent started to decrease very rapidly, which changed the climate background conditions and may have contributed to the interdecadal changes in its impacts [27].Thus, it is conjectured that the impact of autumn Arctic SIC anomaly on the WCE in mid-latitudes may experience an interdecadal change.Exploring the interdecadal change in the linkage of autumn Arctic SIC anomaly with WCE in midlatitudes can facilitate our understanding, which in turn can help predictions of WCE occurrence based on preceding SIC variations.
Many previous studies have used Arctic sea ice anomalies as a starting point to investigate their impact on WCE in different regions.For example, the autumn Arctic sea ice area anomaly has been shown to have a significant impact on cold extremes in the Middle East and midwestern North America [11], while a decrease in sea ice in Baffin Bay, Davis Strait, and the Labrador Sea will lead to more cold air outbreaks in eastern North America [28].Anomalous September sea ice in the Barents to East Siberian Seas has been linked to low temperature anomalies in the Far East region in December and in high-latitude regions from Europe to East Asia in February [29].Recent studies have also found the impact of Arctic sea ice on one side of the Pacific hemisphere; for example, anomalous sea ice conditions in the East Siberian-Chukchi-Beaufort region during September and October have been found to affect low temperature events in central and western China in winter [30,31].The different regional anomalies in Arctic sea ice can have different regional impacts on mid-latitude weather [32].Unlike these previous studies, we have used the occurrence of Asian WCE (AWCE) as the starting point for our analysis, and found in a separate study that November SIC anomalies in the northern and southern Arctic (SICN and SICS, respectively) are significantly positively and negatively correlated with the occurrence of WCE in Asia on an interannual timescale from 1979 to 2022, respectively.We further discovered that these interannual relationships between late autumn Arctic SICN and SICS with the occurrence of AWCE exhibit interdecadal changes.Thus, the present study focuses on examining the interdecadal changes in the teleconnections between late autumn Arctic SICN and SICS with the frequency of AWCE occurrence, as well as investigating the physical mechanisms involved in these teleconnection changes.

Data
The study utilizes ERA5 reanalysis data from the European Centre for Medium-Range Weather Forecasts [33], including both hourly and monthly atmospheric variables, as well as monthly surface heat fluxes.The hourly atmospheric variables encompass 2 m air temperature, 1000 hPa air temperature and 500 hPa geopotential height, while the monthly atmospheric variables comprise geopotential height and horizontal winds at standard pressure levels.The monthly surface heat fluxes encompass surface latent heat, surface sensible heat, and surface upward longwave radiation fluxes.The monthly Arctic SIC is obtained from UK Met Office Hadley Centre [34] and US National Snow and Ice Data Center (NSIDC) [35].In this study, late autumn refers specifically to the month of November and winter of a particular year refers to JF of the following year, such as the winter of 1979 refers to JF of 1980.The study period covers the autumn and winter of 1979-2022.
A high-pass filtering is applied to the data to exclude the long-term trend and interdecadal variations [36], as this study focuses on the changes in an interannual time scale impact of Arctic SIC on WCE.The high-pass filtering was performed using weights of −0.27, 0.53 and −0.27, with a cutoff frequency of 0.1.The response function of the highpass filtering indicates that it is capable of effectively removing signals with periods longer than a decade while retaining signals at the interannual timescale.After applying a high-pass filtering, the first and last year of the data will be lost.Thus, the dataset covering the autumn and winter of 1980-2021 is analyzed after applying a high-pass filtering.The statistical significance for the linear regression and correlation is assessed using a two-tailed Student's t-test.

Winter cold extremes (WCEs)
WCE is defined as the days when the daily air temperature at 1000 hPa drops below 1.0 standard deviation lower than the local climatological seasonal mean in the winter of 1979-2022.WCE denotes the frequency of WCEs at each grid point.The greater WCE at certain grid points, the more frequent WCEs at those grid points.

Asian WCEs (AWCEs) index
The AWCE index is defined as the days when the regional average daily air temperature over 25 • -45 • N, 50 • -120 • E at 1000 hPa drops below 1.0 standard deviation lower than the local climatological seasonal mean in the winter of 1979-2022.The greater the AWCE index, the more frequent the WCE in the mid-latitudes of Asia.The region of AWCE is indicated by the dashed green box in figure S1.

Blocking events
Blocking events are defined as intervals in which daily geopotential height at 500 hPa exceeds 1 standard deviation above the local climatological seasonal mean in the winter of 1979-2022 for five consecutive days at each grid point [10,11].The greater the number of blocking events at a given grid point, the more frequent the blocking events are at that location.
It is worth noting that there are alternative methods for defining blocking events [37,38].We also define the blocking index (BI) for each grid point as another way to represent atmospheric blocking, based on the following formulation [38]: For each grid point, MZ represents the daily geopotential height value at 500 hPa, while Zd and Zu are the minimum of daily geopotential height value at 500 hPa within 60 • downstream and upstream, respectively, at the same latitude as the selected grid point.

Results
The study on changes in the interannual time scale impact of Arctic SIC anomaly on WCE is built on the results of the study investigating the interannual time scale impact of Arctic SIC anomaly on WCE.In section 3.1, we will provide a brief summary of the impact of Arctic SIC anomaly on AWCE on an interannual time scale.

Relationship between the AWCE and Arctic SIC on an interannual time scale
To explore the regions of the Arctic where SIC anomalies have an effect on the frequency of AWCEs occurrence, the correlation coefficients between Arctic SIC and the AWCE index from 1980 to 2021 are calculated, as shown in figure S2.The analysis reveals two distinct regions in the Arctic with SIC anomalies that are closely related to the AWCE index.One region is characterized by significant positive correlations in the northern Arctic (79 • -90 • N, 90 • E-180 • -90 • W; SICN), while the other region shows significant negative correlations in the southern Arctic (72 • -76 • N, 160 • E-140 • W; SICS), as highlighted by the two green boxes in figure S2.
Moreover, to represent the climatic variability of SIC anomalies in these two regions, SICN index and SICS index are defined as the regional averaged SIC anomalies over the northern Arctic (79 , respectively, as shown in figure 1(a).The correlation coefficient between the AWCE index and the SICS index is −0.54 (p < 0.05), while the correlation between the AWCE index and the SICN index is 0.45 (p < 0.05) during 1980-2021.This suggests that both the SICS and SICN indices exhibit significant correlations with the AWCE index on an interannual timescale throughout the study period.To evaluate the potential impact of autocorrelation on the significance, we recalculate the significance of the correlation between the AWCE index and the SICN/SICS index, following the approach described in Xue et al (2022) [39].Table S1 indicates that the p-values of the correlations between the AWCE index and the SICN/SICS index remain below the 0.05 significance level during the 1980-2021 period.
However, as shown in figure 1(a), the relationship between the SICN/SICS and AWCE indices displays varying phases prior to the late 1990s, indicating a weak correlation.In contrast, starting from the late 1990s, the SICN (SICS) and AWCE indices consistently exhibit same (opposite) phases in most years, implying a strong positive (negative) correlation.This study specifically focuses on exploring the interdecadal differences in teleconnections between Arctic SIC anomalies and the AWCE before and after the late 1990s.In the next subsection, we will utilize statistical methods such as sliding correlations to quantitatively analyze the interdecadal changes in these relationships.

Strengthened impact of SICS and SICN on the AWCE 3.2.1. Changes in their relationships
To further illustrate this interdecadal shift around the late 1990s, the 15 year sliding correlation coefficients between the AWCE index and the SICS/SICN index are shown in figure 1(b).Taking the changes in relationship between the AWCE and SICS indices as an example, we will illustrate how the entire study period is divided into two distinct periods based on the changes in their relationships.
As shown by the black line in figure 1(b), the 15 year sliding correlation coefficients between the AWCE and SICS indices are insignificant during To further verify these shifts, scatter plots and their correlation coefficients are shown in figure S3.The correlation coefficient between the AWCE and the SICS (SICN) indices during the P1 period is not significant, measuring only −0.35 with p > 0.05 (0.39 with p > 0.05), implying a weak relationship.However, the correlation strengthens to −0.61 with p < 0.05 (0.51 with p < 0.05) during the P2 period, which is statistically significant, indicating enhancements in the teleconnections.To further evaluate the potential impact of autocorrelation on these enhanced decadal relationships, the significance of the correlation coefficients between the AWCE and SICN/SICS for the two distinct periods are recalculated, as shown in table S1.Although the significance decreases to some extent when considering autocorrelation of the indices, the enhanced decadal relationships remain robust.Subsequently, we will investigate the potential reasons and underlying physical mechanisms behind these enhancements.

Possible reasons of their strengthened relationships
Previous studies have suggested that an increase in the interannual variability of Arctic SIC could cause strong atmosphere anomalies in the mid-to-high latitudes by inducing strong Arctic warming.Thus, the strengthened impacts of Arctic on the mid-latitude climate could be related to increases in the interannual variability of Arctic SIC, while weakened impacts may be associated with decreased interannual variability [12,13,25].We observe that the interannual variation of the SICS index (blue line of figure 1(a)) is relatively small during the P1 period, but becomes larger during the P2 period.Specifically, the standard deviations (quantifying interannual variability) of the SICS index increases from 0.02 during the P1 period to 0.07 during the P2 period.This indicates that the intensity of the SICS anomaly has increased since the late 1990s, which could lead to amplified Arctic warming according to the previous studies [13,25].
To further illustrate the changes in the amplitude of the SICS anomaly and its impact on local Arctic atmosphere anomalies, regressions of the late autumn SIC, surface sensible heat, latent heat, upward longwave radiation fluxes, and 1000 hPa air temperature anomalies onto the inverted SICS index are conducted during the P1 and P2 periods, as shown in figure 2. The inverted SICS index described below corresponds to SICS loss.The amplitude of Arctic SIC loss in the East Siberian Sea-Chukchi Sea-Beaufort Sea is notably greater during the P2 period compared to the P1 period (figures 2(a) and (b)).During both periods, there are significant increases in surface upward latent heat, sensible heat, and longwave radiation fluxes associated with the SICS loss around the corresponding region.However, the amplitudes of these flux anomalies are larger during the P2 period, consistent with the greater extent of SICS loss compared to the P1 period (figure 2).Consequently, there are significant positive air temperature anomalies associated with the SICS loss in the corresponding region during the P2 period; in contrast, the Arctic warming anomalies are weak and insignificant during the P1 period (figures 2(i) and (j)).By analyzing the near-surface air temperature anomalies in the Arctic caused by different intensities of SICS loss during the two periods, we have discovered that the greater amplitude of SICS loss during the P2 period leads to stronger Arctic near-surface warming anomalies compared to the P1 period, which is consistent with previous studies [13,25].The anomalous exposed warm ocean surface, resulting from SICS loss, contributes to increased heat and radiation fluxes from the ocean surface to the atmosphere during late autumn; then these heat fluxes, in turn, have significantly impacts on the atmospheric circulation anomalies during the following winter season [10,15].In the following analysis, we will investigate the impact of different intensities of SICS loss on the subsequent winter atmospheric circulations by regressing the atmospheric conditions of subsequent winter onto the inverted SICS index in late autumn during the P1 and P2 periods.
Figure 3 displays the regression of winter zonal wind at 200 hPa and blocking event anomalies onto the inverted SICS index in late autumn.During the P2 period, the Rossby wave induced by SICS reaches the Asia region, which could be partly attributed to the strong and significant Arctic warming anomalies induced by the larger amplitude of SICS loss (figures 2(b), (j) and S4(d)).This significantly strengthens the cyclone and anticyclone anomalies over the northern Eurasian continent in the midtroposphere.Between these two anomalous systems, there are significant negative westerly wind anomalies in association with the autumn SICS loss over northern Eurasia (figure 3(b)).However, due to the weak and insignificant Arctic warming anomalies caused by the smaller amplitude of SICS loss during the P1 period, the propagation of the wave activity flux associated with SICS was weaker, inducing only weak anomalous cyclone and anti-cyclone over northern Eurasia (figures 2(a), (i) and S4(c)).Consequently, the significant westerly wind anomalies were hardly observed during the P1 period (figure 3(a)).
Furthermore, a decrease in westerly winds often enhance broader meanders in the atmospheric flow, which can potentially lead to the formation of blocking events [10,11].As a result of the significant negative westerly wind anomalies during the P2 period, there has been a notable increase in the occurrence of blocking events in association with SICS loss over northern Eurasia (figure 3(d)).These significant positive blocking event anomalies have contributed to an increase in the frequency of WCE occurrences in the mid-latitudes of Asia by intensifying the downstream transport of cold air [2,[7][8][9].This implies that the SICS could make strong impact on the AWCE.However, during the P1 period, there were no significant changes observed in the occurrence of blocking events associated with SICS loss in the same region (figure 3(c)), implying a weak connection between SICS and AWCE indices.We also conducted a regression analysis between the BI index, the other index representing atmospheric blocking, and the SICS index during the two periods.The results of this analysis are consistent with those shown in figures 3(c) and (d), as depicted in figure S5.
Therefore, the strengthened impact of SICS on AWCE could be partly attributed to the increased interannual variability of SICS.The larger amplitude of SICS loss during the P2 period leads to significant near-surface Arctic warming anomalies due to positive surface latent heat, sensible heat and upward longwave radiation fluxes anomalies.These anomalies intensify the impact of Arctic SICS on the atmospheric circulations.In the subsequent winter, the significant deceleration of westerly wind associated with SICS loss leads to a notable increase in the occurrence of blocking events, resulting in a significant increase in the occurrence of AWCE.However, during the P1 period, the relatively weak amplitude of SICS loss did not induce a series of significant atmospheric circulation anomalies, and thus did not significantly impact the occurrence of AWCE.
On the other hand, the significant increase in the interannual variability of SICN from the P1 to P2 periods is not observed.However, the Rossby wave associated with SICN also enhanced during the P2 period, as shown in figures S4(a) and (b).Why has the impact of SICN on AWCE also strengthened after the late 1990s? Figure S6 illustrated that the correlation coefficient between SICS and SICN was not significant during the P1 period, but it became significantly anti-correlated during the P2 period, indicating a stronger relationship between SICS and SICN from the P1 to P2 periods.Moreover, considering the strengthened influence of SICS on AWCE mentioned above, it is conjectured that the enhanced correlation between SICS and SICN might contribute to an enhanced impact of SICN on AWCE.
In order to eliminate the influence of the SICS (SICN) on the strengthened relationship between SICN (SICS) and AWCE during the P2 period, we recalculate the correlation coefficient between SICN_res (SICS_res) and AWCE during this period, as shown in table S2.The SICN_res and SICS_res are defined as the residual components of the SICN and SICS indices, respectively, after removing the linearly correlated signals between them, following the approach used in previous studies [25,40].As the relationship between SICS and SICN during the P1 period was statistically insignificant, while it became statistically significantly anticorrelated during the P2 period, we only perform the recalculation of the correlation coefficient between SICN_res/SICS_res and AWCE during the P2 period.
The correlation coefficient between SICN and AWCE increased from 0.39 during the P1 period to 0.51 during the P2 period.However, the correlation coefficient between SICN_res and AWCE during the P2 period is only 0.27, which is not statistically significant (table S2).This decrease in the correlation coefficient between SICN and AWCE from 0.51 (significant) to 0.27 (insignificant) after removing the linearly correlated signal with SICS suggests that SICS has a great influence on the relationship between SICN and AWCE during the P2 period.Furthermore, the fact that the correlation coefficient between SICN and AWCE during the P2 period, after removing the linearly correlated signal with SICS, did not increase any more compared to the P1 period suggest that SICN alone cannot enhance its impact on AWCE from the P1 to P2 periods without the influence of SICS.Put differently, the strengthened relationship between SICN and AWCE from the P1 to P2 periods is closely related to the enhanced relationship between SICN and SICS.
Similarly, to evaluate the influence of SICN on the relationship between SICS and AWCE during the P2 period, the correlation coefficient between SICS_res and AWCE is calculated.According to table S2, the correlation coefficient between SICS and AWCE during the P2 period decreases from −0.61 to −0.49 after removing the linearly correlated signal with SICN.This suggests that, to some extent, SICN also affects the relationship between SICS and AWCE.
However, even after removing the linearly correlated signal with SICN, the relationship between SICS and AWCE during the P2 period (significant correlation coefficient: −0.49) remains statistically significant and is still stronger than that (insignificant correlation coefficient: −0.35) during the P1 period.This finding suggests that even in the absence of the linearly correlated signal with SICN, SICS can still make a strengthened impact on AWCE after the late 1990s and the role of increased SICS intensity in the strengthened impact of SICS on AWCE is to some extent independent of SICN.It also indirectly supports the significant role of increased SICS intensity mentioned above in enhanced impact of SICS on AWCE.
Figure 4 provides further evidence regarding the impact of the significant correlation between SICS and SICN on the relationship between SICS/SICN and the frequency of WCE occurrence during the P2 period.Prior to removing the linearly correlated signal with SICS, SICN-related significant WCE anomalies were mainly observed in East Asia; however, after removing the linearly correlated signal with SICS, the significant WCE anomalies associated with SICN become weaker in East Asia, indicating a substantial reduction in the impact of SICN on WCE in the absence of SICS.
Regarding the SICS-related significant WCE anomalies, before removing the linearly correlated signal with SICN, they were mainly concentrated in the Middle East, with scattered occurrences in East Asia.After removing the linearly correlated signal with SICN, the SICS-related significant WCE anomalies still persist in the Middle East region of Asia, suggesting that even after removing linearly correlated signal with SICN, the influence of SICS on WCE, although somewhat weakened, remained notable.
Overall, the strengthened influence of SICS/SICN on the frequency of WCE occurrence in Asia after 1999/2000 is closely tied to the increased interannual

Summary and discussion
Arctic sea ice has undergone rapid changes in recent decades, which have also affected the climate in midlatitudes.This study found that after the late 1990s, the impact of SIC anomalies in the northern (SICN) and southern Arctic (SICS) on the occurrence of WCEs in mid-latitude Asia has intensified.To assess the reliability of the interdecadal changes, we calculated the correlation maps between SIC and AWCE index during the P1 and P2 periods using both the Hadley and NSIDC datasets (figure S7).The NSIDC dataset provides additional evidence supporting the conclusion that the impact of Arctic SIC on AWCE has strengthened to some extent.
To further assess the impact of the selected threshold for defining extreme cold events on the conclusions, we selected the 90th percentile as the new threshold to redefine the AWCE index, denoted as AWCE_90th.Figure S8 compares the AWCE and AWCE_90th indices, with a high correlation coefficient of 0.97, indicating strong consistency between them.Additionally, the correlation coefficients between AWCE_90th and SICN/SICS were not significant during the P1 period, but they strengthened and became statistically significant during the P2 period (figure S9).These results suggest that using the 90th percentile as the new threshold to redefine cold extremes is consistent with the original approach.We also refined the definition of AWCE by using surface 2 m air temperature, denoted as AWCE_2m.The high correlation coefficient of 0.95 (p < 0.05) between AWCE and AWCE_2m indices indicates a strong consistency between the two.
On one hand, this is due to an increased interannual variability of SICS.The strong SICS loss modulates the local atmospheric stability by positive surface latent heat, sensible heat and upward longwave radiation fluxes anomalies, and then lead to a strong mid-latitudinal atmospheric response, including negative westerly wind anomalies and positive blocking event anomalies.These anomalies result in more WCE occurrences in mid-latitudinal Asia.On the other hand, the strengthened correlation between SICS and SICN since 1999/2000 amplifies the impact of both on frequency of WCE occurrences.
Previous studies have studied the impact of Arctic sea ice anomalies on the Arctic polar vortex one month later [30,41,42].During the P2 period, the November SIC anomalies cause local temperature anomalies in the Arctic, ultimately affecting the occurrence of cold extremes.To discuss whether this process includes interactions between the stratosphere and troposphere, the 50 hPa geopotential height anomalies in December were regressed onto the inverted SICN and SICS indices, as shown in figure S10.There is a weakening of the polar vortex associated with SICN and SICS.Furthermore, Song et al [18] confirms that through the interaction between the stratosphere and troposphere during the intraseasonal to cross-seasonal processes, the low autumn sea ice in the Kara-East Siberian Sea leads to the prolongation and enhancement of the Ural blocking, exacerbating the Eurasian cold current.Therefore, troposphere-stratosphere interaction may play a role in the present study, albeit not a primary one.
Another question is why, in comparison to the P1 period, the late autumn Arctic SICS loss was larger during the P2 period (figures 2(a) and (b)).Figure S11 shows the climatological mean of the late autumn SIC in Arctic during the P1 and P2 periods.During the P1 period, the climatological mean of SICS is mostly above 0.9 (dashed green box in figure S11(a)).In other words, the Arctic sea ice during the P1 period is predominantly in a relatively stable and close to a fully ice-covered state, which restricts the extent of SICS loss.In contrast, during the P2 period, the climatological mean of SICS in some regions is less than 0.9, which may be related to the greater SICS losses compared to P1 period (dashed green box in figure S11(b)).Moreover, the decrease in the climatological mean of SICS from the P2 period to the P1 period is consistent with the ongoing decline trend in Arctic sea ice during recent several decades, as previous studies found [43,44].Therefore, the increased interannual variability of SICS may be a result of the sustained decline in the Arctic sea ice.
These changes in Arctic sea ice may be related to the long-term climate change.In the context of global and Arctic warming, understanding the changing state of Arctic SIC and analyzing its evolving impacts is crucial for improving the prediction of mid-latitudinal extreme events.

Figure 1 .
Figure 1.Time series of the standardized AWCE index (solid gray line), the late autumn SICN index (dashed red line), and the late autumn SICS index (dashed blue line) during 1980-2021.(b) The 15 year sliding correlation coefficients between the AWCE and the late autumn SICN indices (solid gray line), and between the AWCE and the late autumn SICS indices (solid black line) displayed for the central year from 1987 to 2014.The r and p in (a) represent the correlation coefficient and significance level, respectively.The dashed red line in (b) represents the correlation coefficient at the significance level of p < 0.05 and the red dots in (b) represent the central years with 15-year sliding correlation coefficients at the significance level of p < 0.05.A high-pass filtering is applied to the AWCE, SICN, and SICS indices.

Figure 3 .
Figure 3. Anomalies of winter 200 hPa zonal winds ((a), (b); unit: m s −1 ) and incidence of winter blocking events (c), (d) regressed on the standardized inverted SICS index in late autumn during the P1 period (a), (c) and the P2 period (b), (d).

Figure 4 .
Figure 4. Anomalies of WCE incidence regressed on the standardized SICN index (a), SICN_res index (b), inverted SICS index (c) and inverted SICS_res index (d) in late autumn during the P2 period.The SICN (SICS) index after removing the linearly correlated signal with SICS (SICN) index is shortened as SICN_res (SICS_res) index.