The West Antarctic ice mass has been on a steady decline over the past few decades due to global warming, resulting in global sea level rise1, 2. The ice mass change is attributed to ice discharge from ice sheets by accelerated ice flow due to disintegration of ice shelves3, 4. On the other hand, snow accumulation associated with atmosphere variations entails short-term increase in the ice mass. The El Niño–Southern Oscillation (ENSO), which is the most powerful climate phenomenon on interannual timescales, leads to weakening of the ASL by teleconnection and causes extreme snowfall over West Antarctica by encouraging moisture transport, in the positive phase (El Niño)5. In 2016, when a strong El Niño occurred, West Antarctica suffered heavy snowfall of about 274.5 Gton, which temporarily delayed the ice mass reduction that had been steadily maintained by global warming (Fig. 1a and Supplementary Text 1)6, 7, which can be said to be a self-defense effect in response to warming. However, since lots of climate variabilities have potential to change nonlinearly related with tipping points8–10, it is unclear whether the short-term delay effect on the ice loss due to strong ENSO will continue in the future.
Earth system models (ESMs) agree that atmospheric response to ENSO intensifies and shifts eastward under global warming because of enhanced ENSO-related convection over the central and eastern Pacific11–13, but changes in high latitudes remains uncertain. Since a formation of responsive teleconnection is regulated by interaction with the mean state and global warming has altered the mean state14, understanding the future change in teleconnection is quite challenging. Furthermore, reliability of models in simulating Antarctic precipitation remains insufficient. Although it has been possible to quantitatively measure the Antarctic ice mass change by GRACE satellites since the early 2000s, unknown physics related to Antarctic precipitation and lack of observations due to geographical specificity make precipitation projections in ESMs differ widely. However, despite the limitations, deliberate adoption of models with regard to model performance can provide more reliable prediction15, 16. The aim of this study is to identify the future change in the El Niño effect on West Antarctic precipitation as a function of degree of global warming with adopted ESMs.
Relationship Between El Niño And West Antarctic Precipitation
The relationship between ENSO and West Antarctic precipitation over the last 41 years (1979 − 2019) can be summarized as follows by considering the seasonal cycle. In 2016, West Antarctica gained ice mass by at least 102.1 Gton in April to May (AM) against ice discharge, mainly from mass accumulation by cumulative precipitation of approximately 107.6 Gton (Supplementary Fig. 1). In an anomaly sense, this is 48.5 Gton higher precipitation than normal years, which exceeds 1.6 standard deviations of the precipitation variation (Fig. 1a). Similarly, in 1998 when another strong El Niño occurred, severe anomalous precipitation of 85.9 Gton was observed in AM, corresponding to 2.8 standard deviations of the precipitation change.
West Antarctic precipitation obviously responds to El Niño, with a correlation coefficient of 0.46, but there is a time lag of a season in the relationship17 (Supplementary Table 1) due to the seasonally varying atmospheric mean state, since establishment of teleconnection depends on the mean state configuration18, 19. In austral summer, although ENSO intensity is strong, the teleconnection is weak in high latitudes due to absence of the subtropical jet that acts as an amplifier and a direction guide of wave response20. On the other hand, in austral autumn, the wave response to ENSO becomes stronger and reaches high latitudes owing to appearance of the subtropical jet forming a double-jet structure with the polar jet21. The seasonal feature in the mean state allows high mean and large variability of West Antarctic precipitation for austral autumn17 (Supplementary Fig. 2). Additionally, ENSO is more highly correlated with the precipitation in El Niño years than in La Niña years by asymmetry of ENSO teleconnection, with a correlation coefficient of 0.53 in the former and 0.33 in the latter (Supplementary Table 1). Since the ENSO-induced circulation anomaly near the Amundsen Sea is more strengthened and pole-eastward positioned in El Niño years22, moisture is efficiently transferred to West Antarctica (Fig. 1b). Therefore, we examined variations of the Niño3 index during December to February (DJF) and West Antarctic precipitation during AM in El Niño years.
To verify simulation ability of the relationship in the latest climate models, 23 Coupled Model Intercomparison Projection Phase 6 (CMIP6) ESMs were used (see Methods). Based on regression analysis, 11 models were chosen in which capability of reproducing the response of West Antarctic precipitation and ASL anomalies to El Niño is appropriate (Supplementary Fig. 3). The models effectively capture the horizontal structure of them as well, although the precipitation anomaly is weaker and the ASL anomaly is biased southward compared to the observational data (Supplementary Fig. 4). Despite the adoption, mean and variability of the West Antarctic precipitation in the models are somewhat underestimated comparing with the observational data (Supplementary Fig. 2), suggesting that a role of surface mass balance, regulated by precipitation, in the West Antarctic ice mass change has been reduced in the model simulation. Nevertheless the adopted models demonstrate reasonable performance in simulating the physical process associated with the relationship.
Vanishing Of The Relationship Under High-emissions Future Scenarios
Previous research suggested that increase in moisture availability in Antarctica by global warming can bring to more precipitation under high-emissions scenarios15, 23. The physics-based perspective leads to an inference that, in the same manner, precipitation caused by El Niño increases. However, multi-model ensemble (MME) means of future projections, i.e., SSP1-2.6, SSP3-7.0, and SSP5-8.5, show that the response of West Antarctic precipitation to El Niño is expected to weaken and disappear under the high-emissions scenarios, and the precipitation anomaly even turns to negative (Fig. 2a). The reversal implying vanishing of the relationship begins to occur in the 2050s under SSP5-8.5 and 2060s under SSP3-7.0. In turn, occurrence of extreme precipitation over West Antarctica driven by strong El Niño disappears (Fig. 2b). As El Niño intensifies, the precipitation response becomes stronger in the present climate, but the linearity is broken in the high-emissions scenarios, indicating that the delay effect does not exist in the warmer climate and resilience of the climate variability to the loss of the West Antarctic ice mass is lost. On the other hand, no remarkable change exists under SSP1-2.6.
The vanishing is originated from a change in dynamical process. In the higher emission scenarios, El Niño-driven ASL anomaly is positioned more east-equatorward and the migration prevents the water vapor at low latitude from advecting to the interior of West Antarctica (Fig. 3a, b), consequently suppressing precipitation there. The east-migrating is known to be due to enhanced ENSO convection in the central and eastern Pacific under global warming24, 25, which is seen as eastward-shift of the convection. The modified mean state in warmer climate is also responsible for the migration. Greenhouse warming gives rise to poleward shift and strengthening of the jet stream in the Southern Hemisphere26, 27 and positively enhanced SAM28, 29, although the effect of greenhouse gases is expected to be decelerated by stratospheric ozone recovery in the near future30. However, the mitigation will be temporary in high-emissions scenarios29, 31. Uncontrolled greenhouse gas emission is expected to cause more positive SAM trend, which is represented by strengthening and eastward extension of the polar jet in the high-emissions scenarios (Fig. 3a). The change in the mean state, in particular the jet stream, causes a variation in teleconnection by modifying phase speed and direction of propagating Rossby wave emitted from the tropical forcing18.
To explicitly understand the migration of the El Niño-driven ASL anomaly, shown in Fig. 3a, an additional analysis was conducted on how the positioning varies according to extension of the polar jet. For this, four experiments (i.e., the historical and three future projections) and all 23 models were used regardless of simulation ability (see Supplementary Text 2 and Supplementary Fig. 5 for advantage of the undiscerning use). In addition, 1) we acquired horizontal maps of El Niño-driven ASL anomaly for each of 92 cases on the last 41 years of each projection, 2) divided them into four 23-case groups according to the degree of eastward extension of the polar jet in the mean state, 3) calculated multi-case ensemble mean of the regression maps, and lastly 4) detected the center of the ASL anomaly in the ensemble means. The result is exhibited in Fig. 3c, which demonstrates a nonlinear behavior in the meridional migration that the El Niño-driven ASL anomaly shifts poleward and then returns to the equator with jet extension. The arc-like migration is likely to be associated with a waveguide near the Drake Passage, which is responsible for curving wave propagation patterns such as the Pacific South America (PSA) teleconnection18. This obviously shows that the positioning of the El Niño-driven ASL anomaly is dependent on a condition of the polar jet32.
In the future change, strengthening of westerly wind in the mean state can increase phase speed of propagating waves and make their zonal wavelength large due to the Doppler effect18. This is identified in the power spectrum analysis showing the power of each zonal wavenumber (Supplementary Fig. 6). The power of the lower zonal wavenumber is larger in the higher emission scenarios; thus, the teleconnection gets slightly relocated with the change. But, it exhibits inconsistent relocation near a ‘reflecting latitude’ at which poleward propagating wave is reflected back toward the equator (Supplementary Fig. 7). Phase of waves in front of the latitude is expected to migrate toward the pole, similar to results in previous studies11, 24, whereas phase of reflected waves is more likely to shift to the equator, which is in agreement with our results. Since the El Niño-driven ASL anomaly is placed near the reflecting latitude, prediction of the migration in the meridional direction is somewhat ambiguous in near future, but it evidently shows equatorward shift in far future.
Global warming by the industrial revolution is a main factor affecting climate change, and the increase in global-mean surface air temperature (GMT) is relevant to positive trend in SAM31, 33 (Fig. 4a). The El Niño-driven ASL anomaly seems to linearly migrate eastward with global warming but not in the meridional direction (Fig. 4b, c). The shift of the ASL anomaly heads toward the pole by 1.6°C GMT warming, while it turns in the opposite direction above 1.6°C warming. Interestingly, even in the low-emission SSP1-2.6 scenario, the latitude positioning the ASL anomaly is almost identical to that in the higher emission scenarios in far future (2059–2099), since the GMT is over 1.6°C. The result indicates that keeping GMT below 1.6°C is essential to prevent the reversal.