1 Introduction

One of the most interesting topics regarding polar climate change is the sudden reversal of the Antarctic sea ice extent from a slow expansion that lasted more than three decades starting in late 1970s to a rapid contraction since the mid-2010s (Parkinson 2019). Sea ice cover directly influences not only the ice shelf (Massom et al. 2018) and ecosystem (Labrousse et al. 2017) locally in Antarctica, but also global ocean circulations (Heuzé 2021), ocean-atmosphere heat, moisture and momentum exchanges (Yu 2019), and ocean carbon uptake (Gupta et al. 2020). Consequently, changes in the Antarctic sea ice extent have significant implications that go far beyond the Antarctic region.

Numerous studies have examined the increasing trend in Antarctic sea ice extent and have attributed the increase to anthropogenic forcing, notably increased greenhouse gas emission and stratospheric ozone depletion, and to natural forcing that arises primarily from sea surface temperature (SST) oscillations. The latter include global-scale oscillations in the Pacific or the Atlantic Oceans, particularly the Pacific Decadal Oscillation (PDO), the Interdecadal Pacific Oscillation (IPO), and the Atlantic Multidecadal Oscillation (AMO) (Li et al. 2014; Meehl et al. 2016; Yu et al. 2017; Yu et al. 2022a), the Zonal Wave Three (ZW3) (Raphael 2007) and the South Pacific Oscillation (Yu et al. 2021), as well as oscillations, such as the Antarctic Oscillation (AAO) or the Southern Annular Mode (SAM) (Thompson et al. 2011; Yu et al. 2023) and the Amundsen Sea Low (ASL) (Raphael et al. 2017), which are more local to the Antarctic. Other important local factors, e.g., local wind patterns (Holland and Kwok 2012), ocean-ice interactions (Zhang 2007; Goosse and Zunz 2014), and freshwater input (Bintanja et al. 2013, 2015), have also been identified to affect Antarctic sea ice cover. The complex interactions among forces of different nature and scales render a complete explanation of the increasing trends in the Antarctic sea ice extent elusive (Hobbs et al. 2016).

Compare to our knowledge about the mechanisms for the increasing trend in the Antarctic sea ice extent in the three decades prior to the mid-2010s, less is known about the forces behind the decreases in recent years. Eayrs et al. (2021) provided a summary of the mechanisms proposed to explain the recent decline in the Antarctic sea ice extent, which included the teleconnection to SST anomalies in the tropical oceans (Stuecker et al. 2017; Wang et al. 2019; Yu et al. 2022b; Bonan et al. 2023; Zhang and Li 2023) and the southern high-latitude ocean and atmosphere circulation anomalies (Turner et al. 2017; Schlosser et al. 2018; Meehl et al. 2019; Wang et al. 2022; Turner et al. 2022; Yadav et al. 2022; Zhang et al. 2022; Purich and Doddridge 2023; Wang et al. 2023). Wang et al. (2019) also identified the Antarctic stratospheric polar vortex anomalies as a contributor to the Antarctic sea ice anomalies in 2016 (Wang et al. 2019).

Previous studies about the Antarctic sea ice extent have focused on SST role in either the increasing period prior to the mid-2010’s or the decreasing period in recent years. What is SST role for the recent four decades including two periods remains to be solved. In the following, we seek for global SST and atmospheric circulations associated with Antarctic sea ice variations for 1979 through 2020.

2 Dataset and methods

The monthly Antarctic sea ice concentration data are available from the US National Snow & Ice Data Center (NSIDC) for the 1979–2020 period (Cavalieri et al. 1996). The sea ice extent is defined as the sum of the areas in the 25 km × 25 km polar stereographic grid cells with sea ice concentration of at least 15%. Monthly sea ice anomalies in are calculated as the subtraction the 42-year average of the monthly data for each month from the original monthly data. The anomalous atmospheric circulations related to the Antarctic sea ice anomalies are from the European Centre for Medium Range Weather Forecasts (ECMWF) fifth generation reanalysis (ERA5) (Hersbach et al. 2020); the ERA5 dataset has a good performance of representing the atmospheric conditions over the Antarctic continent and the Southern Ocean than other contemporary reanalysis datasets (Gossart et al. 2019; Tetzner et al. 2019; Dong et al. 2020). Specific atmospheric variables include mean sea level pressure (MSLP), 2-m air temperature, 10-m wind field, surface downward longwave radiation, and 200-hPa geopotential height and wind. Another important variable in the analysis is the top-of-the-atmosphere outgoing longwave radiation (OLR) obtained from the US National Oceanic and Atmospheric Administration (NOAA) Interpolated OLR dataset (Liebmann and Smith 1996; Henley et al. 2015). In addition to these atmospheric variables, SST anomalies are derived from NOAA’s Extended Reconstructed SST V5 (Huang et al. 2017).

The anomalous SST and atmospheric circulations related to Antarctic sea ice anomalies are analyzed through regression analysis. The level of statistical significance is estimated by the Student’s t-test. Furthermore, the Rossby wave source (RWS) suggested by Sardeshmukh and Hoskins (1988) and wave activity flux (WAF) defined in Takaya and Nakamura (2001) are evoked to demonstrate the generation and propagation of planetary waves, respectively. Finally, the monthly AMO Index (Enfield et al. 2001) is obtained from the US National Center for Atmospheric Research (NCAR) Climate Data Guide and the IPO Index (Power et al. 1999) is obtained from NOAA’s Physical Sciences Laboratory database.

3 Results

As shown in Fig. 1 and similar to the results of previous studies (Parkinson 2019), the overall Antarctic sea ice extent shows a slightly upward trend for three and a half decades beginning in 1979 when satellite observations became widely available, and a sharp downward trend since 2014 (2013 for spring). The rates of increase prior are 2.4, 2.9, 2.2, and 2.2× 104 km2 yr−1 (p < 0.05) for austral summer, autumn, winter, and spring, respectively. The rates of declining in recent years are an order of magnitude larger, at −3.3, −3.5, −1.7, and −1.4 105 km2 yr−1, for the corresponding seasons (p>0.05).

Fig. 1
figure 1

The time series of Antarctic sea ice extent for the 1979–2020 period in austral summer (JFM) (a), autumn (AMJ) (b), winter (JAS) (c), and spring (OND) (d)

Full size image

Through regression analysis, we show that these trends, including the recent reversal, are largely consistent with the changes in the atmospheric circulation patterns and the anticipated response of sea ice to the dynamical and thermodynamic effects associated with these circulations. We further demonstrate that the changes in the atmospheric circulation patterns are related to SST anomalies through the generation of planetary waves of different amplitudes that propagate along different paths depending on seasons.

To explain Antarctic sea ice extent change, we examine the regression patterns of the sea ice concentration anomaly field onto the time series of normalized Antarctic sea ice extent, calculated as the time series of anomalous Antarctic sea ice extent divided by its standard deviation. The spatial regression patterns shown in Fig. 2 correspond to a positive slope or increasing trend with magnitude of one standard deviation of the sea ice extent. Negative slopes of the same magnitude are represented by the inverted patterns (not shown).

Fig. 2
figure 2

Regression maps of Antarctic sea ice concentration onto the time series of normalized Antarctic sea ice extent for austral summer (a), autumn (b), winter (c), and spring (d) for the 1979–2020 period. Dotted regions indicate the above 95% confidence level. The thick red lines indicate the Antarctic sea ice extent (sea ice concentration is equal to 0.15)

Full size image

For austral summer, positive sea ice concentration anomalies prevail across the Southern Ocean with the exception of the Bellingshausen Sea and southwestern Weddell Sea. The pattern is similar for austral autumn, but the areas of positive anomalies are somewhat larger and the areas of negative sea ice anomalies have also expanded to include the northeast part of the Antarctic Peninsula. Austral winter and spring are characterized by positive sea ice anomalies over most of the Sothern Ocean.

Previous studies have suggested the existence of a teleconnection between the Antarctic sea ice anomalies and the global SST anomalies (Li et al. 2014; Meehl et al. 2016; Yu et al. 2017; Yu et al. 2022a). To explore this teleconnection, we show in Fig. 3 the spatial regression patterns of the anomalous SST onto the time series of normalized sea ice extent. The SST regression patterns, which correspond to the sea ice regression patterns (Fig. 2), are characterized by a prominent IPO signal (negative phase) in austral autumn and spring, with negative SST anomalies in the tropical central and eastern Pacific Ocean flanked by positive SST anomalies in both the central and the tropical western Pacific Ocean. A positive phase AMO is also noticeable, but only in austral winter and spring. The 10-year filtered IPO index and the time series of the Antarctic sea ice extent anomalies are inversely correlated, with the correlation coefficients varying by season, at −0.59 (p<0.01) for austral summer, −0.40 (p<0.05) for austral autumn, −0.47 (p<0.02) for austral winter, and −0.68 (p<0.01) for austral spring. The 10-year filter is applied to obtain the decadal and interdecadal signal in the IPO index. Similarly, significant correlations are also found between the 10-year filtered AMO index and the normalized time series of Antarctic sea ice extent for austral winter (0.53, p<0.01) and spring (0.56, p<0.01).

Fig. 3
figure 3

The same as Fig. 2, but for SST anomalies (°C)

Full size image

To explore the physical processes behind the statistically significant correlations between the anomalous SST and the Antarctic sea ice concentration patterns corresponding to sea ice extent, we turn our attention to the spatial regression patterns of the anomalous atmospheric circulation variables (Figs. 4, 5, and 6). For austral summer, negative OLR anomalies (Fig. 4a) appear over the tropical western Pacific Ocean indicating enhanced convection as a result of the positive SST anomalies (Fig. 3a) in the region. The deep convections generate or strengthen 200-hPa divergent wind and positive RWS (Fig. 4b) over the southwestern Pacific Ocean. An anomalous Rossby wavetrain propagates eastwards into the eastern Pacific and then southeastwards into the Bellingshausen Sea (Fig. 4c and d). The wavetrain is related to the positive 200-hPa height anomalies over the southeastern Pacific Ocean, in contrast to negative anomalies over the southern high-latitude regions (Fig. 4d). The pattern of the MSLP anomalies, which is similar to that of the anomalous 200-hPa height, resembles a positive phase of the SAM (Fig. 5a). The positive SAM index is associated with stronger westerly winds (Fig. 5b) and decreased surface air temperature and downward longwave radiation across the Southern Ocean with the exception of the Bellingshausen Sea (Fig. 5c and d). These near-surface wind, temperature, and radiation patterns resemble those of anomalous sea ice pattern. Lower temperature/radiation corresponds to positive sea ice anomalies, and vice versa (Li et al. 2014).

Fig. 4
figure 4

Regression map of OLR (W m−2) (a), 200-hPa divergent wind (vectors) and Rossby wave source (RWS) (10−10 s−2) (b), stream function (107m2 s−1) and wave activity flux (WAF) (vectors) (c), and 200-hPa geopotential height (gpm) (d) onto the time series of normalized Antarctic sea ice extent for austral summer. Dotted regions in panels (a) and (d) denote above 95% confidence level

Full size image
Fig. 5
figure 5

Regression map of a mean sea level pressure (Pascal), b 10-m wind field (vectors), c 2-m air temperature (°C), and d surface downward longwave radiation (105 W s−1), onto the time series of normalized Antarctic sea ice extent for austral summer (JFM). Dotted in panels (a), (c), (d) and shaded in panel (b) regions indicate above 95% confidence level

Full size image
Fig. 6
figure 6

Regression map of OLR (W m−2) (a), 200-hPa divergent wind (vectors) and Rossby wave source (RWS) (10−10 s−2) (b), stream function (107m2 s−1) and wave activity flux (WAF) (vectors) (c), and 200-hPa geopotential height (gpm) (d) onto the time series of normalized Antarctic sea ice extent for austral autumn. Dotted regions in panels (a) and (d) denote above 95% confidence level

Full size image

For austral autumn, negative OLR anomalies induced by positive SST anomalies in the southwestern Pacific Ocean produce positive 200 hPa divergence wind and RWS (Fig. 6a–d) (Fig. 3b). An anomalous Rossby wavetrain triggered by the positive RWS propagates southeastwards into the Amundsen and Bellingshausen Seas and then into the southern Atlantic Ocean (Fig. 6c and d). The anomalous 200-hPa heights (Fig. 6d) display a negative phase of the Pacific South America (PSA) mode (Mo and Higgins 1998), which has been shown to influence the climate of West Antarctica as suggested by Mo and Paegle (2001). The significantly negative MSLP anomalies in the Amundsen and Bellingshausen Seas strengthens the Amundsen Sea Low (ASL) (Fig. 7a), which generate anomalous northerly winds over the Bellingshausen and western Weddell Seas and southerly winds over the Amundsen and Ross Seas (Fig. 7b). The positive MSLP anomalies induce southwesterly winds over the southern Atlantic Ocean. From the dynamic perspective, anomalous southerly (northerly) winds push sea ice offshore (onshore), which increase (decrease) sea ice extent. From the thermodynamic perspective, positive downward longwave radiation anomalies favor increased surface air temperature and reduced sea ice cover over the Bellingshausen and western Weddell Seas, while the opposite occurs across the rest of the Southern Ocean (Fig. 7c and d). For austral winter, a Rossby wavetrain appears over the tropical western Pacific Ocean where positive RWS exists due to enhanced convection (Fig. 8a and b) by the positive SST anomalies in the region (Fig. 3c). The wavetrain splits into two branches, with one branch propagating southeastwards into West Antarctica and the other propagating eastwards into the eastern Pacific Ocean, South America, and the southern Atlantic Ocean (Fig. 8c and d). Similarly, a wavetrain is generated over the western Indian Ocean, and it propagates southeastwards into the southern Indian Ocean. Afterwards, a branch of the wavetrain reflects back into Australia and the other branch continues to propagate southeastwards into East Antarctica. The two wavetrains result in stronger ASL and positive 200-hPa height and MSLP anomalies over the southern high-latitude regions, with a spatial pattern resembling the negative phase of the SAM (Figs. 8d, 9a). An anomalous cyclone over the Amundsen and Ross Seas transports sea ice onshore in the Bellingshausen Sea and offshore in the Ross Sea (Fig. 9b). Similarly, the anomalous southerly winds over the Weddell Sea and southeasterly winds over the coast of East Antarctica contribute to the expansion of sea ice cover in these regions. The patterns of surface air temperature and downward longwave radiation anomalies are consistent with the patterns of sea ice anomalies (Fig. 9c, d).

Fig. 7
figure 7

Regression map of a mean sea level pressure (Pascal), b 10-m wind field (vectors), c 2-m air temperature (°C), and d surface downward longwave radiation (105 W s−1), onto the time series of normalized Antarctic sea ice extent for austral autumn. Dotted in panels (a), (c), (d) and shaded in panel (b) regions indicate above 95% confidence level

Full size image
Fig. 8
figure 8

Regression map of OLR (W m−2) (a), 200-hPa divergent wind (vectors) and Rossby wave source (RWS) (10−10 s−2) (b), stream function (107m2 s−1) and wave activity flux (WAF) (vectors) (c), and 200-hPa geopotential height (gpm) (d) onto the time series of normalized Antarctic sea ice extent for austral winter. Dotted regions in panels (a) and (d) denote above 95% confidence level

Full size image
Fig. 9
figure 9

Regression map of a mean sea level pressure (Pascal), b 10-m wind field (vectors), c 2-m air temperature (°C), and d surface downward longwave radiation (105 W s−1), onto the time series of normalized Antarctic sea ice extent for austral winter. Dotted in panels (a), (c), (d) and shaded in panel (b) regions indicate above 95% confidence level

Full size image

For austral spring, positive 200-hPa divergent wind and RWS appear over the tropical eastern Indian Ocean and the southwestern Pacific Ocean where positive SST anomalies (Fig. 3d) boost convection (Fig. 10a, b). An anomalous wavetrain excited over the region to the southeast of Australia propagate southeastwards, but most reflect back into the central and eastern Pacific Ocean (Fig. 10c and d). A small branch of the wavetrain propagates eastwards into the Amundsen and Bellingshausen Seas and the southern Atlantic Ocean. Near the surface, the wavetrain is accompanied by a spatial pattern resembling the positive polarity of the SAM (Figs 10d, 11a), which produces anomalous southerly and southwesterly surface winds (Fig. 11b) and negative surface air temperature and downward longwave radiation anomalies (Fig. 11c and d) in the southwestern Pacific Ocean and thus positive sea ice anomalies in the region. Similarly, over the southern Indian Ocean, the expanded sea ice cover is in accord with the negative surface air temperature and downward radiation anomalies there. The anomalous southwesterly winds over the region of 0–30°E also help the expansion of sea ice there.

Fig. 10
figure 10

Regression map of OLR (W m−2) (a), 200-hPa divergent wind (vectors) and Rossby wave source (RWS) (10−10 s−2) (b), stream function (107m2 s−1) and wave activity flux (WAF) (vectors) (c), and 200-hPa geopotential height (gpm) (d) onto the time series of normalized Antarctic sea ice extent for austral spring. Dotted regions in panels (a) and (d) denote above 95% confidence level

Full size image
Fig. 11
figure 11

Regression map of a mean sea level pressure (Pascal), b 10-m wind field (vectors), c 2-m air temperature (°C), and d surface downward longwave radiation (105 W s−1), onto the time series of normalized Antarctic sea ice extent for austral spring. Dotted in panels (a), (c), (d) and shaded in panel (b) regions indicate above 95% confidence level

Full size image

4 Conclusion

We have demonstrated, through regression analysis of sea ice, SST and atmospheric data, that the Antarctic sea ice trend from 1979 through 2020, including the abrupt change in direction around 2014, can be largely explained by SST oscillations and the associated anomalous atmospheric circulations over the Pacific and the Atlantic Oceans.

In particular, we show that the Antarctic sea ice changes are closely related to two SST interdecadal variability modes, the IPO in the Pacific Ocean, and the AMO in the Atlantic Ocean. The connection to IPO is significant for all four seasons. However, the connection to AMO is significant only for austral winter and spring. The two modes correspond to different spatial patterns and magnitude of the SST anomalies, which excite planetary wavetrains of different amplitude and propagation path. The differences in the wavetrains induce season-dependent responses in the MSLP fields in the Southern Oceans that are characterized specifically by positive phase of the SAM for austral summer and spring, negative phase of the SAM for austral winter, and negative phase of the PSA1 for austral autumn. The patterns of the anomalous surface wind, air temperature, and downward longwave radiation fields corresponding to the anomalous MSLP fields are largely consistent with the anomalous sea ice patterns in different seasons and different regions of the Southern Ocean.

5 Discussion

Previous studies investigated the role of the IPO and AMO in the Antarctic sea ice concentration anomalies during the sea ice expansion period from late 1970s to mid-2010’s (Li et al. 2014; Meehl et al. 2016; Yu et al. 2017; Chung et al. 2022). By extending the period to 2020, the current study reveals that the two interdecadal SST modes are related not only to the slow sea ice expansion prior to 2014, but also to the rapid contraction since then. While the results here show how Antarctic sea ice respond to natural forcing within the climate system, as represented by the two SST modes, it is important to note that external forces, such as anthropogenic greenhouse gas emission and ozone depletion, also play an important role in driving the changes in the Antarctic sea ice extent, especially in austral summer (Sigmond and Fyfe 2010; Bitz and Polvani 2012; Landrum et al. 2017). A serious limitation of the current study is the relatively short time series (42 years) of the atmospheric and oceanic data for adequately capturing the interdecadal signal. In future studies, model data from the Coupled Model Intercomparison Project Phase 6 (CMIP6) and longer-term reanalysis and reconstruction data may be used to confirm and further elucidate the connection between the IPO and AMO indices and the Antarctic sea ice change on interdecadal scale. Idealized numerical experiments using coupled atmosphere and ocean models should also prove useful for such purpose.