Antarctica's uncertain contribution to sea level rise

The Antarctic Ice Sheet (AIS) is the largest ice mass on Earth and the largest freshwater reservoir, containing about 26,000 km³ of ice—equivalent to roughly 58 m of global mean sea level^1,2. It is divided into the West Antarctic Ice Sheet (WAIS), containing about 4 m sea-level equivalent (SLE), and the East Antarctic Ice Sheet (EAIS), containing about 54 m SLE, separated by the Transantarctic Mountains^1,2 (Figure 1a). Much of the AIS is grounded above sea level, but large sectors—particularly in West Antarctica—rest on bedrock below sea level (Figure 1b), making them vulnerable to both atmospheric and oceanic warming (Figure 1c).

As ice flows from the interior toward the coast, it thins and extends into floating ice shelves that form cavities underneath, within which ocean water circulates, delivering heat and salt to the ice sheet base and driving basal melt (Figure 1c). The resulting meltwater, together with ambient salinity, modifies density gradients (i.e., differences in water salinity and temperature) and influences circulation within the cavity. The last point of the ice sheet grounded on bedrock before it starts floating onto the ocean is called the “grounding line”, or more realistically “grounding zone”. In contrast to Greenland, surface melt remains limited over most of Antarctica, and mass loss primarily occurs through basal melting of ice shelves and calving³. When surface melting does occur, it is uneven: in some areas it is more intense than in others. It is influenced by teleconnections (long-distance, large-scale climate patterns connecting weather anomalies) and extreme events⁴, including atmospheric rivers⁵ that can drive both enhanced surface melting and increased snowfall and even rain on the ice sheet (Figure 1c). Ocean-driven basal melting, particularly through the intrusion of warm Circumpolar Deep Water (CDW), has been identified as a dominant process driving ice-shelf thinning in several regions^6–8.

Until the early 2000s, the Antarctic contribution to sea-level rise was assessed to be close to zero⁹, with models suggesting a near-equilibrium or slightly positive mass balance (i.e., gaining more ice than it loses to the ocean). The launch of satellite gravimetry and altimetry missions (such as GRACE and ICESat) transformed this understanding, revealing sustained and increasing mass loss^10,11 from both the Antarctic and Greenland ice sheets (Figure 2b). In particular, early observations showed that most Antarctic mass loss had occurred in West Antarctica and the Antarctic Peninsula, while East Antarctica appeared to be in balance¹¹.

Recent GRACE/GRACE-FO observations as of 2025 also show that some sectors of East Antarctica are losing mass, in particular in the Aurora basin (Totten and Denman glaciers) (Figure 2c).

Over the period 1993–2018, Antarctica contributed approximately 8% of total GMSL rise, compared to about 15% from Greenland and 20% from mountain glaciers¹² (Table 1, Figure 2a). Together, this land ice accounted for about 43% of GMSL rise. Ocean thermal expansion contributed an additional 46% (up to 56% according to Liang et al., 2025), and land water storage changes contributed the remaining 11%^12,13. While ocean thermal expansion dominated GMSL rise until the early 2000s, land ice loss has become the primary driver since 2005^11,14. Although Antarctica contributes to sea level rise, snowfall over the 20th century was large enough to compensate for ice mass loss¹⁵. Accelerating ice mass loss observed since the 2000s indicates that calving and ice shelf thinning now outpace the rate of snow and ice accumulation in West Antarctica^11,16 (Figure 2b). Satellite observations have, however, revealed that increased snowfall over East Antarctica continues to reduce the potential overall mass loss since 2020¹⁷ (Figure 2b). This reduction is compatible with the projected increase in Antarctic snowfall due to global warming (warmer air holding more moisture, but still below freezing, allowing for more snow) during the 21st century^18,19.

Table 1: Contribution to global mean sea-level rise over 1993-2018 (IPCC AR6 Chapter 09¹²). Left column: cumulative contribution (mm); Right column: rate of contribution (mm yr^-1). Also provided are the measurement uncertainties, represented here as the 17th and 83rd percentiles.

The drivers of mass loss from the West Antarctic Ice Sheet were for a long time uncertain, but the collapse of the Larsen B Ice Shelf (along the Antarctic Peninsula) in 2002 demonstrated the critical role of atmospheric heat waves and ocean warming resulting in accelerating land ice discharge into the ocean^20,21. In several cases, such as the collapses of the Larsen A ice shelf in 1995^22,23 and the Wilkins ice shelf in 2008²³ along the Antarctic Peninsula, melting from the warm ocean underneath weakened the ice from below. This made the shelves fragile, so when unusually warm air caused surface melting and forced meltwater into cracks, the water deepened and widened them through hydrofracturing, leading to rapid collapse. In other cases, collapse was triggered by extreme weather events, as for the collapse of the Conger Ice Shelf (East Antarctica) in 2022²⁴, which was pre-conditioned by sea-ice loss and pre-existing ice shelf structural weakness²⁵ enhanced by tidal flexure for example²⁶.

Ice shelves play a critical stabilizing role by exerting buttressing forces (acting like a dam) that restrain ice flow from the interior grounded ice sheet into the ocean^27,28. Where ice sheets are grounded below sea level, the intrusion of warm ocean water enhances ice shelf basal melting²⁹, thinning ice shelves⁶ and weakening their buttressing effect. This process is currently observed in the Amundsen and Bellingshausen Sea sectors^7,30. Ice-shelf thinning can trigger grounding-line (the point where ice rests on the seabed) retreat on beds sloping towards the ice sheet interior, e.g where the seabed gets deeper further inland (Figure 3a). This leads to a process known as Marine Ice Sheet Instability (MISI), a self-reinforcing process that drives accelerating ice discharge, irreversible retreat, and long-term sea-level rise^31,32. For some future scenarios this process can be slowed down as the bedrock rises after ice loss (a process known as the slow rise of the bedrock as the weight of the ice decreases)³³ (Figure 1), providing a stabilising effect³⁴. As the ice sheet melts and the grounding-line retreats, uplift of the bedrock provides new pinning points³⁵ that can stabilize the ice and may favour subsequent readvance of the grounding line³⁶. A second proposed mechanism, Marine Ice Cliff Instability (MICI), may occur after ice-shelf loss, when the newly exposed ice front becomes too tall and steep to support its own weight and begins to collapse, potentially accelerating ice loss^37–39 (Figure 3b). The extent to which this process operates, and how rapidly it can unfold, remains debated because several negative feedbacks may limit cliff height and reduce instability^1,40. However, the key implication is that once thresholds for both instabilities are crossed, marine-based sectors of the ice sheet may become committed to long-term and potentially irreversible mass loss, with consequences for multi-meters sea-level rise over multi-centennial to millennial timescales, as shown by paleo-evidence^41,42 and ice sheet modeling^43,44.

On some glaciers such as Pine Island Glacier⁴⁵ or Thwaites Glacier⁴⁶, MISI could already be operating. While the prevalence of this process in EAIS remains poorly constrained, several ice shelves buttressing major outlet glaciers (which drain inland ice through valleys), including Totten⁴⁷ and Denman⁴⁸, show signs of weakening over the past decades, as do some ice shelves in the Wilkes Subglacial Basin (for example, Cook ice shelf)³⁰.

Since satellite altimeters began observing sea surface height in 1993, GMSL has risen faster than during any preceding century of the last three millennia¹². This acceleration is robustly detected in tide-gauge records and satellite altimetry and was unequivocally attributed to anthropogenic forcing in IPCC AR6⁴⁹. Since 1993 GMSL has risen by about 111 mm at an accelerating rate⁵⁰, with nearly a quarter of this rise occurring between 2019 and 2024 alone⁵¹. The annual rate of rise has increased from about 3.4 mm yr⁻¹ over 1993–2024 to nearly 6 mm yr^-1 in 2024, reflecting accelerating land ice loss combined with recent exceptional atmospheric and ocean warming^50,52 (Figure 2a). Satellite-based analyses suggest that if current trends merely continue, GMSL could rise by an additional ~17 cm over the next three decades⁵⁰.

Although Antarctica’s historical contribution has been smaller than that of other components, it represents the largest source of uncertainty in future sea-level projections mostly because of its unpredictable dynamical behaviour. Projections of GMSL rise and future Antarctic mass loss are indeed highly sensitive to greenhouse gas emissions pathways (currently the Shared Socio-economical Pathways, SSP) and the rate of warming, underscoring the importance of mitigation policies and their timely implementation⁵³. By 2100, sea level projections diverge substantially (Figure 4a), with much of the spread attributable to uncertainty in the Antarctic Ice Sheet accelerating contribution especially on the long term¹² (Figure 4a & 4b).

If global warming remains below the +2°C target of the Paris Agreement adopted in 2015 (SSP1-1.9 & SSP2-2.6), GMSL is projected to rise by 0.28–0.62 m by 2100, with Antarctica contributing roughly 26–45% of the total rise, and some glaciers committed to retreat (Figure 4c). Under a +3°C warming pathway (SSP4.5), GMSL rise increases to 0.44-0.76 m, with Antarctica contributing 19–38%. Under high emissions scenarios exceeding +4.4°C (SSP5-8.5), GMSL rise is projected to reach 0.63-1.01 m by 2100, with Antarctica contributing 16–33%. If ice-sheet instabilities (e.g., MISI, MICI) are triggered, rapid disintegration of large sectors of the AIS could lead to GMSL rise approaching or exceeding 1.6 m by 2100 (Figure 4c), although confidence in these high-end estimates remains low^12,54,55, and recent studies^56–58 suggest that such high estimates are unlikely to happen this century.

Projected sea-level rise refers to the rise reached by a given time horizon under a specific greenhouse-gas emissions scenario. In contrast, committed sea-level rise refers to the unavoidable rise effectively locked in by past and ongoing warming, even if part of the response unfolds only over longer timescales. This is because ice sheets, like the ocean, absorb and release heat on a millennial timescale. Owing to this large thermal inertia (the slow response of huge ice masses to temperature changes), ice sheets and glaciers will continue to contribute to GMSL rise well beyond 2100^49,59,60. Some model studies suggest that even if carbon emissions stabilise at current levels, committed GMSL rise could be around 0.5 m by 2100 and exceed 1 m by 2300. Multi-centennial and multi-millennial projections suggest that under high-emissions scenarios, collapse of West Antarctic Ice Sheet and parts of East Antarctic Ice Sheet could contribute up to several meters of sea-level rise by 2300^12,61–63(Figure 4b). Some studies suggest that parts of the Amundsen Sea sector in West Antarctica may already be committed to long-term retreat, even without further warming⁶⁴. Paleoclimate evidence supports these projections, indicating multi-meter Antarctic contributions to sea level during past warm periods with global temperatures only 1.5-3°C above pre-industrial levels and atmospheric CO₂ concentrations below 400 ppm (parts per million)^65–69. These records highlight the high sensitivity of the AIS to sustained moderate warming⁷⁰.

Current-generation models struggle to fully represent these instabilities, leading to their exclusion from standard “likely” ranges in IPCC assessments. As a result, low-probability but high-impact scenarios—such as GMSL rise exceeding 2 m by 2100 or more than 5 m by 2150 under high-emissions pathways—cannot be ruled out¹². Despite advances in climate and ice-sheet modeling, large uncertainties persist due to limited on-site observations, short data records, and differences in how models represent key physical processes⁷¹. This hampers a better understanding of the potential for self-amplifying cycles and tipping points^72,73. An additional source of uncertainty is the change in the Antarctic surface mass balance, which is expected to increase with warming due to enhanced moisture transport and extreme precipitation events^24,74 (e.g., atmospheric rivers), potentially offsetting part of dynamic mass loss as observed recently^72,74 (Figure 2) for the 21st century.

Reducing uncertainty in future sea-level projections thus urgently requires improved observations and modeling of key processes at the ocean–ice, atmosphere–ice and ice-bedrock interfaces, particularly those governing ice-shelf melt, grounding-line dynamics, and non-linear instabilities^72,75. To address these challenges, the ice-sheet modelling community is preparing a new Ice Sheet Model Intercomparison Project (ISMIP7, https://www.ismip.org/research/ismip7-protocol), supported by the WCRP Climate and Cryosphere (CliC) project, which will contribute to the IPCC AR7. In parallel, the SCAR Scientific Research Programme INSTANT (https://scar.org/science/research- programmes/instant) is working to resolve key knowledge gaps in Antarctic ice-sheet dynamics.

Sea level will continue to rise under all emission scenarios beyond the end of this century, but both the magnitude and rate of rise remain uncertain (Figure 4b). The long-term impacts on global coastlines are linked to the fate of the AIS, which represents the largest source of uncertainty in GMSL projections. Uncertainty in Antarctica’s future contribution to sea-level rise has direct societal consequences. For many coastal communities and megacities (defined as cities with more than 10 million inhabitants), 0.5 m of GMSL rise represents the upper limits of adaptation, with some regions already experiencing increasing damage and loss^76–78. Such a limit is committed to happen within this century and is unavoidable (Figure 4).

In global and regional sea-level projections, differences in the drivers of climate change used in ice-sheet models dominate uncertainties until about 2070³⁸. Beyond this timeframe, uncertainty increasingly depends on future carbon emissions pathways and poorly constrained ice mass loss from the Antarctic ice sheet³⁸. The stability of large ice shelves is critical in constraining how quickly the ice sheet will lose mass⁷¹. Uncertainty on regional sea-level projections is further compounded by local processes affecting vertical land movement. These can occur at rates comparable to global sea-level rise and significantly offset or amplify relative sea-level change^79–81. These processes remain poorly accounted in current sea-level rise projections and are often insufficiently addressed in coastal hazard and risk assessments. Integrating improved physics into next-generation ice-sheet models is essential to narrow the current low-confidence, high-impact uncertainty range and better inform risk assessment and policy decisions. A closer collaboration between scientists, practitioners, and end users is essential (e.g., Practitioner Exchange for Effective Response to Sea level rise – PEERS network, https://peerscoastal.org/) to improve access to consistent, existing actionable sea-level information^82,83 and to better communicate uncertainty.

As GMSL rises, the frequency of extreme sea-level events, historically occurring once every 100 years, is projected to increase to at least once per year by the end of the century in many regions, even under low-emissions scenarios^82,83. Populations at risk of extreme sea-level events doubles after just 0.75 m GMSL rise⁸⁴ which is projected to happen before 2100 in the case of SSP5-8.5 scenario and around 2150 in the SSP2-4.5 scenario (Figure 5). Higher GMSL would amplify the impacts of storms and high tides, increasing coastal flooding and economic damages, which could reach up to 14 trillion USD per year globally by 2100 under high-emissions scenarios, without additional adaptation⁸⁵, and impacting as many as 77 to 132 million people worldwide⁸⁶ and potentially up to more than 900 million⁸⁷ (Figure 5). Paleo-evidence underscores that the irreversible loss of the main Antarctic ice shelves will cause a multi-meter long-term sea-level rise, impacting many generations in the future⁸⁸. These impacts highlight that uncertainty in Antarctic-driven sea-level rise is not merely a scientific challenge, but a critical constraint on effective coastal planning and long-term multi-generational adaptation.