We analyse seasonal averages of atmospheric variables across Antarctica (regions and catchments defined in Fig. ), calculated over December–February (DJF), March–May (MAM), June–August (JJA), and September–November (SON). All anomalies are calculated relative to the 1979–2018 mean (unless otherwise specified). All fields are linearly detrended prior to analysis.

We next consider whether SMB responses at the catchment scale are distinct from background and average conditions, i.e. outside the 5th or 95th percentiles of all SON SMB changes during the 1979–2018 period. We find that there are no consistent responses between extreme El Niño events (comparing red triangles in Fig. ) or strong La Niña events (comparing blue triangles in Fig. ), except in Enderby Land, East Antarctica (Fig. ). Here, the three extreme El Niño events all have cumulative SON SMB anomalies at approximately 8000 kgm-2 (Fig. ). In fact, the largest positive regional cumulative SMB anomalies in Enderby Land between 1979 and 2018 (other than in 2009, which our analysis identified as an outlier) occur during these extreme El Niño events (Fig. ). During La Niña events, we do not see evidence of a consistent SMB response in Enderby Land, with only the 1988/89 La Niña event SON SMB anomaly exceeding the 5th percentile at < -5000 kgm-2 (Fig. ).

In all other catchments, the magnitudes of the SMB anomalies during extreme/strong ENSO events are not consistently extreme, but, in numerous catchments, a single extreme/strong ENSO event is associated with an extreme SMB response (Fig. ). In the Wilkes Subglacial Basin and George VI catchment, the 2015/16 extreme El Niño event is associated with the largest positive cumulative SMB anomalies, at approximately 12 500 and 11 000 kgm-2, respectively (Fig. ). However, the other two extreme El Niño events are associated with SMB anomalies within interquartile ranges (between 2500 and -3000 kgm-2 and 3000 and -4500 kgm-2, respectively) (Fig. ). In Princess Elizabeth Land, the 2007/08 La Niña event is associated with the largest positive cumulative SMB anomaly in the region for the 1979–2018 period at 9000 kgm-2, but other strong La Niña events SMB anomalies are within the region's interquartile range (Fig. ). In the Getz and Amundsen Sea catchments, SMB anomalies during the 1998/99 and 1999/2000 strong La Niña events are the largest positive cumulative SMB anomalies in the time period around the 95th percentile at approximately 6000 and 11 000 kgm-2, respectively (Fig. ).

As explained in Sect. , we compare extreme El Niño events to non-extreme CP El Niño events, as show that CP El Niño events are associated with larger SMB responses than EP events. In the Aurora Subglacial Basin, significant SMB changes do not occur in any extreme El Niño event (anomalies are within the interquartile range, between -4500 and 4500 kgm-2), but three CP El Niño events are associated with large positive anomalies around the 95th percentile, between 9000 and 12 000 kgm-2 (Fig. ). In the Getz catchment, two strong La Niña events (1998/99 and 1999/2000) are associated with positive SMB anomalies outside the 95th percentile; however, the other strong La Niña events are associated with SMB anomalies within the region's interquartile range (Fig. ). In the Ross West region, two CP El Niño events (1991/92 and 2002/03) are associated with approximately 10 000 and 16 000 kgm-2, respectively, with all other SMB anomalies being < 6000 kgm-2 (Fig. ). For other catchments, responses vary by event, with no consistent or significant SMB changes (Fig. ).

We also consider whether moderate and strong El Niño events are associated with more consistent SMB responses. However, as for our analysis of extreme El Niño events, moderate and strong El Niño events do not cause significant and consistent regional SMB changes (Fig. S15). Interestingly, in Enderby Land (the sole catchment with a consistent positive SMB response to extreme El Niño events), we identify that the 1991/1992 strong El Niño event is associated with a positive SMB anomaly of 15 000 kgm-2, which is identified as an outlier in the region's density curves from Fig.  (Fig. S15).

Whilst this paper focuses primarily on SON, we also include DJF, MAM, and JJA results in Figs. S1–S3. An analysis of all seasons shows that SMB responses during extreme ENSO events are generally inconsistent, with individual events often exhibiting responses of opposing sign or of differing magnitudes during all seasons (Figs.  and S3–S5).

During DJF, we identify large positive cumulative SMB anomalies in Dronning Maud Land (2010/11), Enderby Land (1988/89), the Lambert–Amery System (1999/2000) and Princess Elizabeth Land (2007/08; Fig. S1). However, there is no consistent response during La Niña events (Fig. S1).

During both MAM and JJA, we find that there are no consistent responses between extreme El Niño events (comparing red triangles), strong La Niña events (comparing blue triangles), or CP El Niño events (comparing yellow circles; Figs. S4 and S5). During MAM, this is unsurprising because ENSO activity during these months tends to be insignificant . During JJA, positive SMB anomalies during different CP El Niño events are identified as outliers in Enderby Land and the Lambert–Amery System (Fig. S5). However, other CP El Niño events are associated with a range of SMB responses in these regions, reinforcing that this is not a consistent response (Fig. S5).

At the catchment scale, we find that SMB responses vary greatly between individual extreme ENSO events and that there are no consistent SMB responses between the three extreme El Niño events considered, except in Enderby Land, East Antarctica, where SMB anomalies are consistently positive (Fig. ).

Finally, we consider whether strong/extreme ENSO events are significantly different from moderate ENSO events in terms of SON SMB anomalies. Figures – show that there is no clear pattern indicating that strong/extreme ENSO events result in extreme SON SMB anomalies. That is, significant SON SMB responses to individual extreme El Niño events are evidenced in numerous individual catchments (Fig. ). However, these changes are not consistent between events, with the “centres of action” for increased SMB occurring in different catchments during each El Niño event in the record (Fig. ). Enderby Land is an exception, with statistically significant SON SMB increases occurring during all three extreme El Niño events (Fig. a). Enderby Land's average anomaly during these events is equivalent to approximately 32 % of the region's annual average SMB (Fig. a).

During the 2015/16 El Niño event, significant SMB changes are widespread across Antarctica: in the Enderby Land, Wilkes Subglacial Basin, Victoria Land, Ross East, George VI, and Ronne catchments (Fig. a, c–e, g, and h). The 2015/16 SMB probability distribution is shifted positively in all these catchments, except for in the Ross East catchment, where the SMB probability distribution is shifted negatively (Fig. a, c–e, g, and h). During the other two extreme El Niño events studied, significant SMB changes only occur in the Enderby Land (1982/83 and 1997/98; both probability distributions are shifted positively; Fig. a) and Ross West catchments (1997/98; probability distribution is shifted negatively; Fig. f). In all other catchments, SMB changes during extreme El Niño events are not significant (Figs. b and S9).

Significant SON SMB responses during CP El Niño are localised (in < 5 grid cells; Fig. ). In the Aurora Subglacial Basin, Victoria Land, Ross East, and Ross West catchments, a shift in the SMB probability distribution during a CP El Niño event (2002/03 or 2009/10 dependent upon catchment) is larger than the SMB response during the 1982/83, 1997/98, and 2015/16 events (Figs. b, d–f and S9). Most other catchments do not show significant results during CP El Niño events (Fig. a, c, g, and h). Regional results with each extreme El Niño event and CP El Niño event distinguished are included in Fig. S9. Extreme El Niño and CP El Niño results are also significant when compared with the SMB anomaly distribution of moderate El Niño events in most catchments (Fig. S10). This highlights that the SMB changes we identify during extreme or CP El Niño events are not necessarily statistically significantly different to those of moderate El Niño events (Figs. , , and S3–S5).

Few Antarctic catchments exhibit significant SMB anomalies during strong La Niña events (Figs.  and S12). Similar SMB changes occur during moderate and strong La Niña events (Fig. S13). Significant SMB changes during strong La Niña events only extend beyond the moderate La Niña SMB anomaly distribution in the Wilkes Subglacial Basin, Victoria Land, and Ross East catchments (Fig. S13). This highlights that the mass balance changes during strong La Niña events across almost all of Antarctica are not extreme when compared to those of moderate La Niña events (Figs. , , and S3–S5).

Annual SMB anomalies during extreme El Niño events are less significant than their SON equivalent (Fig. S11). Annual SMB anomalies during strong La Niña events are similar to those in SON (Fig. S14). This highlights that the contribution of extreme/strong ENSO events relative to the yearly SMB budget is not necessarily extreme or significant in many Antarctic catchments.

In summary, most SMB responses during the three extreme El Niño events identified are not significantly different from background and average conditions (Figs.  and ). However, SMB responses during the 2015/16 El Niño event are statistically different from both background conditions and moderate ENSO event SMB changes in numerous catchments in East Antarctica and West Antarctica and on the Peninsula (Figs.  and ). Conversely, during strong La Niña events, we do not see evidence of a consistent and significant SMB response (Fig. ).

In many Antarctic catchments, the SMB response differs greatly between the ENSO events that we examined. We find that the magnitude and spatial patterns of these differences are so great that these changes are not consistent between events and the composite is not representative. The addition of strong and moderate El Niño events to our analysis, although increasing the number of events analysed, does not result in a more consistent ENSO signal in SMB. All Antarctic catchments exhibit inconsistent and differing SMB responses between moderate, strong, and extreme El Niño events. Moderate and strong La Niña events are also associated with a range of inconsistent SMB responses during these events. However, when extreme El Niño event SMB changes are compared to SMB changes during moderate El Niño events, SMB responses are significantly different from one another in the Enderby Land, Wilkes Subglacial Basin, Victoria Land, Ross East, Ross West, Getz, Amundsen Sea, Abbott, George VI, and Ronne catchments. However, these results are not consistent across all three extreme El Niño events studied (other than in Enderby Land). and suggest that such findings relate to the non-linear nature of extreme ENSO events, driven by the convective response to extreme tropical Pacific SST anomalies and resultant Rossby wave propagation.

Conversely, numerous other catchments exhibit similar SMB responses when extreme and moderate El Niño event SMB changes are compared. That is, the SMB changes during extreme ENSO events do not differ from those during moderate events in the Dronning Maud Land, Lambert–Amery System, Princess Elizabeth Land, Aurora Subglacial Basin, West Graham Land, Larsen, Filchner, and Coats catchments. Analysis of moderate and strong El Niño events also identifies that regional SMB changes in Antarctica are indistinguishable during moderate, strong, and extreme El Niño events, except in Enderby Land. These findings support studies that have considered moderate and extreme ENSO event impacts together in analyses e.g., which suggests that the Antarctic response is not sensitive to the magnitude of ENSO events.

Smaller Antarctic SMB changes occur during strong La Niña events than during extreme El Niño events. This may be partially explained by the smaller amplitude and weaker Antarctic teleconnection of La Niña events when compared to El Niño events, which means these events are often also associated with smaller Antarctic impacts . Understanding the differences in the teleconnection dynamics between El Niño and La Niña events is an important area of research, but investigating the remote mechanisms of influence is beyond the scope of this study.

Our findings support previous work showing that snowfall increases in Enderby Land during El Niño events . We identify an SLP system and anomalous atmospheric circulation over the Indian sector of East Antarctica during each extreme El Niño event (Fig. e, f, i, j, m, and n). This circulation change causes warm onshore winds to extend into the Indian sector of Antarctica during these events, bringing moisture-laden air over the region, resulting in increased precipitation and SMB. Enderby Land is associated with SMB increases during each extreme El Niño event, despite differences in SLP and temperature anomalies between extreme El Niño events. This is because Enderby Land is located within the “centre of action” of an Indian sector circulation system that brings moist, warm air onshore . Differences in SLP and temperature are associated with this atmospheric circulation system migrating east, west, or poleward during each El Niño event (Fig. ).

During the 1982/83 event, a positive SLP anomaly extends from the Lambert–Amery System to Dronning Maud Land (Fig. e). This SLP anomaly shows a weakening of the low-pressure system in the Pacific sector, resulting in moisture-laden warm air extending across the whole Enderby Land catchment, causing precipitation and SMB increases (Fig. e–h). During the 1997/98 event, a stronger positive SLP anomaly extends over eastern Dronning Maud Land and across Enderby Land, and the SLP system further weakens along the Antarctic coastline relative to the 1982/83 event, resulting in an influx of moist air over the west of Enderby Land that drives precipitation and SMB increases in this region (Fig. i–l). Finally, during the 2015/16 event, a negative SLP anomaly extends across the Pacific sector of East Antarctica, inland and into the Lambert–Amery System, reaching the interior and eastern Enderby Land. This results from a strengthening and coastward shift in the low-pressure circulation system in the Indian sector , causing extensive precipitation across the Lambert–Amery System and the entire catchment of Enderby Land (Fig. m–p). Therefore, each extreme El Niño event is associated with similar SMB responses, despite differences in the SLP responses (Fig. ).

During the 1997/98 event, we find that a large positive SLP anomaly extends across the Indian sector, over Enderby Land and the eastern edge of the Dronning Maud Land catchment (Fig. m). This results in onshore moisture-laden winds over the region and subsequent precipitation (Fig. m–o), causing SMB increases across the entire Enderby Land catchment (extending inland) and the eastern edge of the Dronning Maud Land catchment (Fig. p). Our results agree with , who showed an SLP anomaly caused blocking over the Dronning Maud Land and Enderby Land catchments during the 1997/98 event, which enabled storm tracks to extend to Dome Fuji (inland Enderby Land) and brought sustained snow accumulation to the region.

During the 1998/99 and 2010/11 strong La Niña events, we also identify negative SLP anomalies and associated increases in precipitation and SMB over Enderby Land (Fig. i and u). show that SLP anomalies occur over the Enderby Land catchment during the 2010/11 strong La Niña event, resulting in sustained precipitation along the Indian sector of the East Antarctic coastline. Our findings support and highlight that similar SLP systems and blocking processes could occur in the Ross and Amundsen/Bellingshausen sectors, depending upon the ASL location, and in the Weddell and Pacific sectors, where SLP anomalies are identified during La Niña events in our analysis.

These findings also highlight that El Niño and La Niña event impacts are not always reciprocal . Therefore, if this teleconnection were to persist under future climate change, we expect that more frequent and intense ENSO extremes (no matter the polarity) will result in snow accumulation increases in Enderby Land.

Snow accumulation in inland Enderby Land is partially driven by short-term, high-magnitude precipitation events, which extend inland a few times each year . These events develop due to the large-scale circulation, which facilitates the development of atmospheric rivers that funnel warm, moist air across the southern mid-latitudes to Antarctica, resulting in high, sustained precipitation . The sheer volume of snowfall during extreme Antarctic precipitation events can, in a matter of days, offset regional melting . Local and regional variability in circulation can therefore play an important role in SMB. Future research should consider the interactions between teleconnection dynamics and atmospheric river development, determining if ENSO events facilitate such extensive snowfall.

The 2015/16 El Niño event was associated with significant SMB changes across Antarctica, consistent with previous research focusing only on this event . Our findings further show that the 2015/16 event stands out relative to previous events and that it is associated with more widespread and significant Antarctic SMB changes than during other extreme ENSO events. One reason for this may be because the 2015/16 extreme El Niño event displayed CP-type characteristics , which typically enhance snow accumulation in Antarctica compared with EP El Niño events .

In the Wilkes Subglacial Basin, the 2015/16 extreme El Niño was associated with SON SMB anomalies approximately 25 % larger than during any other year from 1979–2018. The 2015/16 event was the most extreme El Niño event during the past 40 years, with a larger SST anomaly magnitude than the 1982/83 and 1997/98 events . This magnitude difference was partially attributed to the 2015/16 event being initiated from a warmer tropical Pacific background state than the 1982/83 and 1997/98 events , resulting in the higher magnitude and more widespread anomalies in Antarctica.

We find no apparent anthropogenic trend in the magnitude of the extremes in our analyses (Fig. S7). Some studies have suggested that the larger-magnitude SST anomaly of the 2015/16 El Niño event could indicate a change in ENSO behaviour due to climate change e.g. . Furthermore, a previous study could not rule out an anthropogenic driver of recent regional increases in snow accumulation. Attributing whether there is an anthropogenic signal in these extreme ENSO events and SMB anomalies is beyond the scope of this study and requires centennial-scale datasets to fully characterise ENSO variability .

This study focuses on extreme ENSO events; however, we have not considered other patterns of climate variability that influence Antarctic climate on interannual, decadal, and interdecadal timescales . One such mode of climate variability is the Southern Annular Mode (SAM), which is known to influence the Antarctic climate and interact with ENSO . For example, during the 1982/83 and 1997/98 extreme El Niño events, the SAM was in a negative phase, with weaker circumpolar westerly winds around Antarctica , but, during the 2015/16 El Niño event, the SAM was in a very strong positive phase, leading to stronger and poleward-shifted westerly winds . These differing circumpolar wind anomaly patterns, alongside ENSO-driven anomalous atmospheric circulation, could explain some of the differences in precipitation and SMB anomalies we identify during the 1982/83, 1997/98, and 2015/16 events . A recent study shows that the SAM explains up to 24 % of Antarctic ice-mass trends since 2002, highlighting that the SAM almost certainly exerts an influence on our results and on extreme El Niño and strong La Niña events. Interestingly, the combination of an extreme El Niño event and positive-phase SAM in 2015/16 was unusual , and this 2015/16 El Niño event was also substantially warmer than previous extreme El Niño events, resulting in unusual regional impacts in southern South America and on the Antarctic Peninsula . Here, we find the positive precipitation and SMB anomalies do extend further inland during the 2015/16 event than during the other two extreme El Niño events. This is somewhat unexpected, given that other studies suggest increased poleward moisture transport during negative-phase SAM, as weakened westerly winds enable the advection of synoptic systems over the continent, bringing precipitation , and, in the austral spring of 2015, there was a positive SAM. Whether these anomalies may have been influenced by the warmer temperatures in 2015/16 – as a warmer atmosphere can hold more moisture and thus could be expected to transport it further inland – should be investigated.

Our analysis of Antarctic surface mass balance during extreme ENSO events is limited by the length of the datasets available. In this study, we utilise the full 40-year (1979–2018) temporal extent of RACMO2.3p3 . However, only a handful of extreme ENSO events occur in this period (three El Niños and five La Niñas). This period also includes both natural climate variability and anthropogenic forced change, which we are unable to separate due to the limited time series length of this analysis . Adequately characterising the statistics of ENSO requires a record on the order of 1–2 centuries , and this is not currently available. Future work utilising palaeoclimate reconstructions and climate model large ensemble simulations would increase the sample size and may reduce sampling bias . Despite these limitations, our examination of SMB responses during the satellite era provides useful and preliminary insights into the impact of extreme ENSO events on Antarctica.

Projections of future ENSO characteristics remain uncertain, including how the teleconnection between the tropics and the poles could change under future climate change . Some CMIP6 model outputs suggest that extreme ENSO events may become more frequent in the future . It is important to understand how Antarctica will respond in a future climate that is warmer and experiences greater extremes . Our results show that extreme ENSO events do not result in large SMB changes across Antarctica. Whilst we do identify consistent and significant SMB changes in Enderby Land, this is the result of atmospheric circulation changes specific to this region. Therefore, whether increases in extreme ENSO events in the future could lead to significant impacts on Antarctica's surface climate is dependent upon how the teleconnection to Antarctica changes . If future El Niño events were to resemble the 2015/16 El Niño event, the resulting increased accumulation could offset some of the ocean-driven melting in Antarctica, reducing mass loss .