The seasonal Antarctic sea ice concentration anomalies related to the Atlantic Niño index

Antarctic sea ice concentration anomalies (ASICA) have been found to be linked to sea surface temperature anomalies in tropical oceans. However, it is not clear whether and how ASICA is linked to the Atlantic Niño mode (ANM). This study demonstrates a significant relationship between ASICA and ANM. The relationships vary by season, with a peak in austral winter and a secondary one in spring. Significant sea ice anomalies associated with a positive phase of ANM are mostly negative in austral winter and spring, and mostly positive in austral summer and autumn. This teleconnection is established by atmospheric wavetrains that are excited over the tropical southwestern Pacific and Indian Oceans and the southern Atlantic Oceans and propagate over the Southern Ocean. These wavetrains induce anomalous near-surface circulations, which generate dynamic and thermodynamic forcing on sea ice, resulting in the observed ASICA patterns. The absence of El Niño Southern Oscillation weakens the connection.


Introduction
The polar sea ice plays a crucial role in regulating the Earth's climate by reflecting solar radiation back to space, reducing heat and moisture exchange between the ocean and atmosphere, and influencing ocean currents and atmospheric circulations.Therefore, understanding the expansions and contractions of the polar sea ice cover is crucial for improving climate predictions.One notable feature of the Antarctic sea ice is its slow expansion from 1970s until recent years, which is contrary to the decreasing trend observed in the Arctic sea ice (Parkinson 2019).Recent studies have focused on this trend and shed light on potential factors behind it (Hobbs et al 2016, Eayrs et al 2021, Yu et al 2022a, 2023a, 2023c).
Additionally, Antarctic sea ice shows significant interannual variability, which has been linked to large-scale climate modes at southern mid-high latitudes (Hobbs et al 2016).The Southern Annular Mode (SAM), the leading mode of the southern mid-and high-latitude atmosphere circulation, is recognized as a crucial driver of the interannual variability of Antarctic sea ice, primarily associated with sea ice anomalies in the Weddell and Ross Seas (Kwok andComiso 2002, Lefebvre andGoosse 2005).The Pacific-South American pattern (Mo and Higgins 1998), representing the second and third circulation modes, also influences the Antarctic sea ice (Kohyama and Hartmann 2016).The Antarctic sea ice anomalies related to the Zonal Wave Three mode mainly occur during austral autumn and early winter (Raphael 2007).The Amundsen Sea Low anomaly is also related to the dipole variability of sea ice in the Ross Sea and the Antarctic Peninsula/Bellingshausen Sea (Turner et al 2016).Moreover, the South Pacific Oscillation is linked to opposite austral autumn Antarctic sea ice variability in areas to the west and east of 150 • W longitude in the Pacific sector of the Southern Ocean (Yu et al 2021).
Apart from southern mid-latitude climate modes, climate modes related to sea surface temperature (SST) anomalies in the tropical oceans have also been identified as important drivers for interannual variability of the Antarctic sea ice (Yuan et al 2018, Li et al 2021).Specifically, the SST anomalies in the equatorial Pacific Ocean known as the El Niño Southern Oscillation (ENSO) can significantly influence Antarctic sea ice anomalies in the Pacific sector of the Southern Ocean, and the location and areal extent of the influence as well as the degree of influence can vary depending on the type of ENSO (Cisato et al 2015, Zhang et al 2021).The sea ice anomalies associated with ENSO tend to reach their peak in the late austral winter and spring (Turner 2004).The teleconnection between Antarctic sea ice and ENSO can be modulated by the state of the SAM (Fogt and Bromwich 2006, Stammerjohn et al 2008, Fogt et al 2010).
The two leading SST interannual variability modes in the tropical Indian Ocean, the Indian Ocean Basin mode (IOBM) and the Indian Ocean Dipole (IOD) mode (Chambers et al 1999, Klein et al 1999, Saji et al 1999), have also been linked to the Antarctic sea ice anomalies.Yu et al (2022b) have shown that IOBM is particularly influential in austral autumn and spring, while Nuncio and Yuan (2015) and Yu et al (2022b) have found that IOD can significantly influence sea ice in the region west of the Ross Sea.These teleconnections can be modulated by ENSO (Yu et al 2022b).The large-scale circulation patterns typically set boundary conditions for the occurrence and tracks of synoptic-scale cyclones (Uotila et al 2013), which then generate the direct dynamic and thermodynamic forcing on sea ice (Uotila et al 2014).
Tropical Atlantic SST anomalies are also found to affect Antarctic sea ice.Specifically, the long-term warm trend in SST associated with the positive phase of Atlantic Multidecadal Oscillation (AMO) has been linked to the increase of sea ice extent in the Ross Sea and decrease in the Amundsen, Bellingshausen, and Weddell Seas (Li et al 2014, Simpkins et al 2016).On the interannual time scale, the two leading modes of the SST variability in the tropical Atlantic Ocean are the Atlantic Meridional Mode (AMM) and the Atlantic Equatorial Mode also known as the Atlantic Niño mode (ANM) (Lübbecke et al 2018, Jiang andLi 2021).While AMM is found to significantly affect Antarctic sea ice in certain regions such as the Ross Sea, the Bellingshausen Sea, and east of the Weddell Sea during austral autumn when AMM reaches its seasonal peak (Ren et al 2022), the effect of ANM on Antarctic sea ice remains unclear.
The ANM typically develops in austral autumn and peaks in austral winter mainly through the Bjerknes feedback mechanism (Zebiak 1993, Keenlyside and Latif 2007, Lübbecke et al 2018).While there are interactions between tropical ocean basins (Cai et al 2019, Wang 2019), there is no significant correlation between ENSO and ANM (Chang et al 2006, Lübbecke and McPhaden 2012, Martín-Rey et al 2014, Tokinaga et al 2019, Jiang et al 2023).ANM has been found to strongly influence precipitation in several regions around the world, including West Africa (Rodríguez-Fonseca et al 2015), northeastern South America (Torralba et al 2015), the Indian monsoon region of Southeast Asia (Kucharski et al 2008, Losada andRodríguez-Fonseca 2016) and Europe (Losada et al 2012, Mohino andLosada 2015).Although ANM has been linked to the South Atlantic Anticyclone (Lübbecke et al 2014), its impact on Antarctic sea ice remains unclear.The goal of this study is to investigate the relationship between ANM and Antarctic sea ice concentration anomalies (ASICA) for each season and to examine whether and how this relationship is modulated by ENSO.

Methods
The ANM index is calculated as the average of SST anomalies in the equatorial Atlantic Ocean region between the latitudes of 3 • S and 3 • N and the longitudes of 20 • W and 0 • W, as defined by Zebiak (1993).To investigate how ENSO may modulate the influence of ANM on the Antarctic sea ice, the ANM index is computed both with and without the ENSO signal.The ENSO signal, represented by the Niño3.4index defined as SST anomalies in the region (5 (Trenberth 1997), is removed using the equation specified in An (2003): where ANM and ANM * represent the time series of the ANM index with and without the ENSO signal, respectively, cov(ANM,Niño3.4) indicates the temporal covariance between ANM and the Niño 3.4 indices, and var(Niño3.4)denotes the variance of the Niño 3.4 index.
The ANM index is computed using version 5 of the Extended Reconstructed SST dataset, which is generated by the US National Oceanic and Atmospheric Administration (NOAA) and detailed in Huang et al (2017).This global SST dataset has a horizontal resolution of 2 degree in both latitude and longitude and spans from 1854 to the present day.The SST dataset is only analyzed from 1979 through 2020.
The variability of the Antarctic sea ice is characterized by the monthly global sea ice concentration data, which are accessible at the US National Snow and Ice Data Center and described in Cavalieri et al (1996).The sea ice concentration data are projected on a polar stereographic grid with 25 km grid spacing and cover the period from 1979 to the present day.
The relationship between the Antarctic sea ice and the ANM is explained primarily with the help of atmospheric circulation anomalies, which are characterized using data from the European Centre for Medium-Range Weather Forecasts fifth-generation reanalysis, also known as ERA5.This new global reanalysis, detailed in Hersbach et al (2020), has been demonstrated to describe the realistic atmospheric fields over the Antarctic ice sheet and the Southern Ocean (Gossart et al 2019, Ramon et al 2019).
In addition, monthly outgoing longwave radiation (OLR) data at the nominal top of the atmosphere in NOAA's Interpolated OLR dataset, described in Liebmann and Smith (1996), is utilized to characterize convective activities over tropical oceans.
The relation between ANM and Antarctic sea ice anomalies is analyzed primarily through linear regression.Linear regression is utilized to identify anomalous atmospheric circulations corresponding to ANM, with and without the ENSO signal.A linear regression between Y (y 1 , y 2 , . . ., y i , . . ., y n ) and X (x 1 , x 2 , . . ., x i , . . ., x n ) can be constructed using where α is the intercept, β is the slope, and ε i is a residual term.A least squares approach can be used to minimize the sum of squared residuals.The statistical significance of linear correlation and regression is determined by a two-tailed Student's t test.Prior to the correlation and regression analyses, all variables are detrended.Furthermore, preceding the regression analysis, the ANM index is normalized as well.This normalization involves dividing the index's deviation from its climatological value by its standard deviation.
Similar analyses are conducted for each of the four seasons, namely austral summer (JFM), austral autumn (AMJ), austral winter (JAS), and austral spring (OND).
The sea ice anomalies associated with the ANM are compared to those linked with other significant SST modes represented by various indices including the AMM, Tropical Southern Atlantic (TSA) and the South Atlantic Subtropical Dipole (SASD) indices for the South Atlantic Ocean, as well as the IOBM and Subtropical IOD (SIOD) indices for the Indian Ocean.For definition and calculation of these indices, refer to table S1 in supplemental materials.

Anomalous SST patterns related to ANM
ANM-regressed SST patterns (figure 1) show significant positive or warm SST anomalies in the equatorial Atlantic Ocean.There is also an area of significant negative or cold anomalies in the western mid South Atlantic Ocean between approximately 10-40 • W and 25-45 • S.This anomalous SST pattern in the Atlantic, which appears in all seasons, share some similarities to another Atlantic SST interannual mode represented by the SASD index in its negative phase (Venegas et al 1997) (not shown), and in fact, the two indices exhibits a significant inverse correlation of −0.50, −0.53, −0.53 and −0.35 (p < 0.05) for austral summer, autumn, winter, and spring, respectively.This result is consistent with Nnamchi et al (2016) (not shown).
The ANM index also shows a significant negative correlation with the Niño 3.4 index, but only for austral autumn (−0.41, p < 0.05) and winter (−0.34, p < 0.05), likely due to the strong La Niña signal in the equatorial Pacific Ocean during these seasons (figures 1(b) and (c)).The correlation, however, is insignificant for austral spring and summer, despite the presence of El Niño in the equatorial Pacific during austral summer (figure 1(a)).
When the ENSO signal is removed during austral autumn and winter, the ANM signal becomes more pronounced, as suggested by an increase in the area of significant warm SST anomalies in the tropical Atlantic (figures 1(e) and (f)).The area of significant cold SST anomalies in the western mid-South Atlantic, however, is somewhat reduced (figures 1(e) and (f)).
For austral summer (figure 1(a)), two Indian Ocean SST Modes, IOBM and SIOD are present.Previous studies have established the co-variability of the SIOD and SASD in austral summer (Fauchereau et al 2003, Hermes and Reason 2005, Lin 2019, Yu et al 2023b).

Anomalous Antarctic sea ice patterns related to ANM
The anomalous Antarctic sea ice patterns associated with the ANM index reveal considerable seasonal variability (figure 2).During the austral summer, significant positive sea ice anomalies are observed over the western Weddell and Ross Seas, while negative values occur primarily along the coast of Dronning Maud Land and in a small part of Southern Atlantic Ocean (figure 2(a)).In the austral autumn, the extent of decreased sea ice over Southern Atlantic Ocean expands, and the Bellingshausen and Ross Seas also have

Anomalous atmospheric circulation patterns related to ANM
The sea ice anomaly patterns and their seasonal variations can largely be elucidated by examining ANM-regressed surface and upper-level atmospheric circulation patterns, along with their seasonal disparities.Starting with winter, a period characterized by the largest ANM-related sea ice anomalies (figure 2(c)), the anomalous mean sea level pressure (MSLP) indicates a zonal wave three structure, with positive MSLP anomalies over the Southern Atlantic Ocean, the western Southern Pacific Ocean and the eastern Southern Indian Ocean (figure 3(a)).A similar structure of surface air temperature and wind field is evident (figure 3(c)).Antarctic sea ice is typically in the state of free drift, i.e. the momentum equation is governed by a balance between the air-ice stress, air-water stress, and the Coriolis force, resulting in an ice drift vector that deviates some 30 • to the left of the wind vector (Vihma and Launiainen 1993).Hence, the negative ANM-related sea ice anomalies at 90 • E, 160 • E and 90 • W (figure 2(c)) can be attributed to the northerly-northwesterly wind anomalies (figure 3(c)).These generate ice drift towards approximately southeast, decreasing sea ice concentration in the northern parts of the sea ice zone.When the ENSO signal is removed, the northwesterly wind anomaly in the Bellingshausen Sea is much weaker (figure 3(d)), and there is no sea ice concentration anomaly in the region (figure 2(f)).
Using anomalous OLR as a proxy for convection, with negative OLR anomalies indicating stronger convection, we show that the ANM-related OLR anomalies in the tropics has a pattern (figure 4(a)) that is almost the inverse of the anomalous SST pattern (figure 1(c)), where positive (negative) SST anomalies correspond to negative (positive) OLR anomalies, indicating an increase (decrease) in convection.In the extratropical regions, increased convective activity occurs over several regions, including the southwestern Indian Ocean, eastern Australia and the southwestern Pacific Ocean, while the opposite is true in other regions, such as the southeastern Pacific Ocean, southern Brazil and Southern Atlantic Ocean.Increased  wavetrain that travels southeastward over the Amundsen and Bellingshausen Seas and then northeastward over the southern Atlantic Ocean.The negative wave sources over the southeastern Pacific Ocean excite a wavetrain that propagates eastward into the southern Atlantic Ocean.The negative wave sources over the southern Atlantic Ocean also aid the propagation of these waves from high latitudes to low latitudes.
The removal of the ENSO signal amplifies convection over the southwestern Indian Ocean but reduces it over the tropics, eastern Australia and southwestern Pacific Ocean (figure 4(b)).Although the wave sources and wavetrains discussed earlier still exist, these are much weaker, with the exception of those over the southwestern Indian Ocean, which influences the structure of zonal wave three in the upper-level height anomalies (figures 4(d) and (f)).
Anomalous MSLPs over the Southern Ocean shows a wave two structure in austral autumn and wave three structure in both austral spring and summer (figures S1 and S2).The anomalous surface wind and temperature patterns associated with these anomalous MSLP patterns are consistent with the anomalous Antarctic sea ice patterns in summer.The negative ANM-related sea ice anomaly at the coast of the Dronning Maud Land (figure 2(a)) can be attributed to the offshore wind anomalies probably opening a coastal polynya (figure S2(c)).The positive sea ice anomaly in the central Ross Sea (figure 2(a)) is probably due to low temperatures and anomalous northeasterly winds (figure S2(c)).In spring, negative sea ice cover at the In austral autumn, the magnitude of the OLR anomalies in the tropics is smaller without the ENSO signal than with the ENSO signal (figures S3(a) and (b)), which is similar to austral winter.The anomalous wave sources over the regions east of Australia and the southwestern Pacific Ocean excite a wavetrain that propagates into the Southern Ocean.Suppressed convective activity over the southern South America yields negative wave sources, strengthening the wavetrain (figures S3(c)-(f)).In austral summer and spring, anomalous convective activities over the southwestern Pacific Ocean can still produce wavetrains into the Southern Ocean (figure S4).Additionally, convective activities over the southern Atlantic Ocean also excite some wavetrains.The ANM-related sea ice anomaly is most pronounced close to the northern margin of the sea ice zone at 130-160 • W (figure 2(b)).Anomalous southerly winds (figure S1(c)) transport sea ice and cold air to the region, explaining the positive sea ice anomaly.The absence of ENSO strongly reduces the sea ice anomaly (figure 2(e)).Another region with similar wind and temperature conditions but a smaller extent of ANM-related sea ice anomaly is found around 30 • E (figure 2(b)).

Effects of other SST modes on Antarctic sea ice
In addition to ANM, we also investigated the relationship between Antarctic sea ice and the other leading Atlantic SST variability modes represented by the AMM and TSA indices.While a recent study by Ren et al (2022) found the most significant relation between Antarctic sea ice and AMM to occur in austral autumn, our results suggested that spring as the season of largest areal coverage of significant AMM-related sea ice anomalies (figure S5).This difference in findings may be due to the use of lagged maximum covariance analysis in Ren et al (2022), compared to simultaneous regression analysis in our study.Nevertheless, the spatial pattern of anomalous sea ice in austral autumn (AMJ) in our study (figure S5 As for the TSA index, while it exhibits insignificant correlation with the AMM index, it is closely correlated with the ANM index, with correlation coefficients of 0.78, 0.82, 0.94, and 0.80 for austral summer, autumn, winter and spring, respectively (p < 0.01).Hence, the influence of the TSA index on Antarctic sea ice closely resembles that of the ANM index, as illustrated in figure S6.However, a substantial disparity arises for positive anomalies over the Pacific Ocean sector and negative anomalies over the Indian Ocean sector in the austral winter.Notably, the ANM index demonstrates a significant negative correlation with the SASD index, with values of −0.50, −0.53, −0.53, −0.35 for austral summer, autumn, winter and spring, respectively (p < 0.05), and this correlation remains independent of the ENSO influence.The Antarctic sea ice anomalies linked with the SASD index exhibits a nearly inverse relationship to those associated with the ANM index.Only during the austral autumn does the IOBM index exhibit a noteworthy correlation with the ANM index (0.33, p < 0.05).Interestingly, the spatial patterns of Antarctic sea ice anomalies align in the Ross Sea, while opposing patterns emerge in the Atlantic sector (Yu et al 2022b).

Conclusion and discussion
In this study, we investigated the teleconnection between seasonal ASICA and the ANM, the leading mode of the tropical Atlantic SST anomalies.We explained this teleconnection through anomalous atmospheric circulations, with particular attention given to the modulation by ENSO, the dominant mode of the equatorial Pacific SST anomalies.The seasonal ANM index is significantly correlated with the Niño3.4index only in austral autumn and winter.The significant correlation between the two indices contradicts with previous studies that reported insignificant correlation between them (Chang et al 2006, Lübbecke and McPhaden 2012, Martín-Rey et al 2014, Tokinaga et al 2019, Jiang et al 2023).The disparity could potentially arise from variations in the study period and the definition of different seasons.Furthermore, prior research primarily investigated lag-lead correlations between the two indices, while the present study focuses on evaluating their simultaneous correlations.
The linkage between the Antarctic sea ice and the ANM index varies with season.Significant sea ice anomalies associated with positive phase of ANM are mostly negative in austral winter and spring and mostly positive in austral summer and autumn, and the areas covered by significant sea ice anomalies are larger in the former than the latter two seasons.Specifically, positive sea ice anomalies occur mainly in the Southern Pacific Ocean in austral autumn and the Ross Sea in austral summer.Negative sea ice anomalies appear in the Weddell and Bellingshausen Seas, the regions north of Oates Land and the regions near 90E • during austral winter and over the Dronning Maud Land and the regions north of Oates Land in austral spring.The largest extent of Antarctic sea ice anomalies related to the ANM index occurs in austral winter when the ANM peaks.Without the ENSO signal, the extent of significant sea ice anomalies reduces, except for the area of negative sea ice anomalies near 90E • .
In austral winter, the wavetrains that travel to the Southern Ocean, and the impact on sea ice appear to be excited by convective activity over the southwestern Pacific Ocean with ENSO and over the southwestern Indian Ocean without ENSO.Additionally, convective activity over the southern Atlantic Ocean supports the development of the wavetrains, especially in austral summer and autumn.These wavetrains induce anomalous MSLP patterns with structures of zonal wave three or two, which yield anomalous surface air temperature and wind with patterns largely consistent with those of Antarctic sea ice anomalies.The negative sea ice concentration anomalies, northerly/northwesterly wind anomalies, and warm anomalies were approximately collocated (figures 2(c) and 3(c); and 2(f) and 3(d)) but it is not guaranteed that the warm anomalies always have a causal contribution to the negative sea ice anomalies.Even if the warm anomalies affect sea ice thickness, a decrease in sea ice concentration would require melt of the entire ice layer, which is not common in winter.Hence, the contribution of warm anomalies to changes in sea ice concentration is probably restricted to regions of very thin ice close to the northern margin of the sea ice zone, and in other cases their collocation with sea ice anomalies may be due to northerly winds that transport warm air south and generate the sea ice anomalies via ice drift.
This study is, to the best of our knowledge, the first to focus on the linkage between the interannual variability of Antarctic sea ice anomalies and the ANM index.The ANM exhibits distinct evolutions at different phases of the AMO and diverse spatial patterns of the equatorial Atlantic SST (Losada andRodrígue-Fonseca 2016, Martín-Rey et al 2018).The phase of the AMO also affects the ENSO annual cycle and its variance (Dong et al 2006, Dong and Sutton 2007, Levine et al 2017).Previous research has shown that the phase of the AMO modulates the relation between the ENSO and ANM indices (Martín-Rey et al 2014).However, it remains unclear whether and how the phase of AMO may affect the connection between Antarctic sea ice anomalies and the ANM index, which requires further investigation.Additionally, future studies should assess the impact of increasing greenhouse gas emissions on this relationship.Studies with idealized numerical experiments are needed to confirm the relationship and to disentangle the roles of Pacific and Atlantic Niño in this connection.
Data processing and analysis codes are available upon request to the corresponding author.

Figure 1 .
Figure 1.Regression map of anomalous seasonal sea surface temperature (SST) ( • C) onto the detrended Atlantic Niño mode index for austral summer (JFM) (a), austral autumn (AMJ) (b), austral winter (JAS) (c), and austral spring (OND) (d) over the 1979-2020 period.Panels (e), (f) are the same as panels (b), (c) except that the ENSO signal is removed from the Atlantic Niño mode index.Dotted regions indicate above 90% confidence level.The color bar indicates the ANM-regressed SST anomalies.

Figure 2 .
Figure 2. Regression map of anomalous seasonal sea ice concentration onto the detrended Atlantic Niño mode index for austral summer (JFM) (a), austral autumn (AMJ) (b), austral winter (JAS) (c), and austral spring (OND) (d) over the 1979-2020 period.Panels (e), (f) are the same as panels (b), (c) except that the ENSO signal is removed from the Atlantic Niño mode index.Only values with above 90% confidence level are shown.Red lines denote climatological Antarctic sea ice extent.The color bar indicates the ANM-regressed sea ice anomalies.

Figure 3 .
Figure 3. Regression map of anomalous seasonal mean sea level pressure (hectopascal) (a) and (b), surface wind field (vector) and surface air temperature ( • C) (c) and (d) onto the detrended Atlantic Niño mode index for austral winter (JAS) over the 1979-2020 period.Panels (b), (d) are the same as panels (a), (c), except that the ENSO signal is removed from the detrended Atlantic Niño mode index.Dotted regions and green vectors indicate above 90% confidence level.The color bar indicates the ANM-regressed surface air temperature anomalies.
(b)) is consistent with that in May in Ren et al (2022) (figure 1(b)).