Antarctic Sea Ice #4: Record lows between 2022 and 2025

Observations from a series of satellites provide a high-quality dataset of near daily sea ice coverage from the late 1970s to the present day. These observations show a large annual cycle in Antarctic sea ice cover, ranging from around 2-4 million km² in late February (summer) to a seasonal maximum of 18-19 million km² in late September (winter) (Fig. 1a). Moreover, from the beginning of the satellite record until 2014, total Antarctic sea ice extent was increasing slowly (Fig. 1b) even though the global atmosphere and ocean were warming. Various hypotheses for this sea ice increase were investigated [3]. However, rapid loss of Antarctic sea ice coverage, beginning in winter 2015 and continuing in 2016, led to a then record low summer extent in 2016/17 (e.g. [4]; Fig. 1a,b; [3]). Since then, Antarctic sea ice coverage has generally been below the long-term average (Fig 1b) [5], with a series of new record lows set in the years since 2021. The most recent austral summers (2021/22, 2022/23, 2023/24, and 2024/25) had the four lowest levels of sea ice cover on record (Fig 1a), and 8 of 12 months in 2023 were record lows for those months [6]. The winters of 2023 and 2024 were the two lowest on record (Fig. 1a). These were particularly notable since, until this point, the winter maximum extent was relatively consistent from year to year (Fig. 1a); the sea ice extent anomaly during winter freeze in 2023 reached 1.8 million square kilometres, or 10%, below any previous observations (Fig. 1a). The sea ice changes in autumn and spring caused changes in the length of the biologically important ice-covered season, with the season shortened by 3-4 months in most areas [6].

With the recent recurrently low sea ice conditions, the trend in annual mean sea ice coverage went from being positive, and statistically meaningful, over 1979-2018 [4] to a trend which is very weakly negative (-1000 km²/yr), but statistically indistinguishable from zero [7], for 1979-2023. However, a linear trend (i.e. a straight line drawn through the data) is not necessarily the best description of long-term changes given the apparent abrupt shift which began in 2016/17 [5].

Analysis of Antarctic sea ice is typically divided into five regions (Fig. 1c). Until approximately 2000, the satellite record was marked by differing regional trends [4, 8]; increases in the Ross and Weddell Sea sectors contrasted with decreases in the Amundsen-Bellingshausen Sea (ABS), especially in summer. Since around 2004 sea ice coverage around Antarctica has shown less regional differences, though some sectors have nonetheless had higher than average sea ice coverage occasionally [9]. The winter 2023 record had low sea ice in all regions except the Amundsen-Bellingshausen Seas [10, 11] (Fig. 1d), while the subsequent near-record low sea ice in winter 2024 had markedly high sea ice concentrations in the ecologically important Scotia Sea region as far north as South Georgia, northeast of the Antarctic Peninsula.

Sea ice exists at the interface of the atmosphere and the ocean and so its variability is driven by a complex interplay of both domains (see [1]). Over short timescales (days to weeks), winds move sea ice around and waves contribute to the breakup of consolidated pack ice, with both driven by weather changes. Over monthly to seasonal timescales, ocean currents affect sea ice both by transporting the ice and by bringing ocean water of different temperatures into the sea ice zone.  For the atmosphere, on average, sea ice is transported north-eastward by prevailing westerly winds (i.e., winds that blow from the west). This happens because, in the Southern Hemisphere, the rotation of the Earth pushes sea ice to the left of the wind direction by 15-30 degrees (i.e., the Coriolis effect). Variability in both south-to-north and west-to-east winds thus changes the sea ice cover by controlling its movement around and away from the continent. These winds also affect air temperature by moving warm or cold air masses that affect the melting or formation of sea ice. For example, the north-westward transport of sea ice by northward winds (i.e., ‘southerlies’) is typically accompanied by cold air from farther south.

Atmospheric features which are important for wind variability include:

• The Southern Annular Mode (SAM), which quantifies the strength and latitude of the circumpolar westerly (i.e. eastward) winds.
• The depth and location of the Amundsen Sea Low (ASL) (a low pressure system in the Ross-Amundsen-Bellingshausen area), and a pattern of three alternating high and low pressure centres around Antarctica called Zonal Wave 3 (ZW3). Both affect sea ice through ice motions and temperature changes associated with north-south winds.

Moreover, the ASL and other local features are heavily modulated both by the SAM and by the El Niño Southern Oscillation (ENSO), which describes tropical Pacific sea surface variability. Coastal polynyas (i.e. regions of open ocean within the sea ice cover) form in locations where intense, cold winds called “katabatic winds” flow off the Antarctic continent and push newly formed ice away from the coast.

In comparison to the atmosphere, the ocean plays a relatively larger role in driving sea ice variability over longer timescales (months and longer), because of its substantially larger ability to store heat and the fact that it changes more slowly. Changes in the saltiness of the ocean and the temperature below the surface can impact sea ice by changing the locations and speed at which the ocean brings heat up towards the sea ice (the surface temperature must be at the freezing temperature of salt water for sea ice to exist).

In the context of these known drivers, a robust picture is emerging of the specific causes of individual recent record low sea ice extents. Identified contributions to the 2016/17 sea ice reduction include warm sea surface temperatures, record strength low pressure systems (storms), a negative SAM (typically resulting in relatively weak and north-shifted eastward winds), and regionally strong north-south wind anomalies. These latter were linked both to variability in the tropics and in the stratosphere (see [3], [15], and references therein.)

The then-record low sea ice extent of summer 2021/2022 was found to be driven by both monthly-to-seasonal atmospheric circulation anomalies [16, 17], with the ASL contributing around three-fifths of the total anomaly [18], as well as individual atmospheric storms [16]. Similarly, the 2022/23 summer sea ice reduction and record low was again linked to a deep ASL driven by a third successive year of ENSO being in its La Niña phase (cold equatorial Pacific ocean), especially La Niña lasting throughout the sea ice melt season [15]. Notably, the sea ice coverage lows of 2021/22 and 2022/23 happened in the context of positive SAM conditions, which would normally be expected to contribute to sea ice expansion in most sectors. There is therefore emerging evidence of a changing relationship of summer sea ice with the atmosphere [15, 19].

As sea ice expanded from 1979 to 2014, the surface of the Southern Ocean cooled [20]. In contrast, the subsurface Southern Ocean (i.e. around 100m-500m depth) has warmed since the 1960s due to greenhouse gas emissions from human activity [21]. This subsurface ocean warming has been implicated in the decrease in Antarctic sea ice coverage since 2014 [5, 22-25], with a modelling study suggesting that 70% of the winter 2023 sea ice anomaly was due to these pre-established warm ocean conditions [25].  The remainder of the winter 2023 anomaly was due to atmospheric conditions, including a strong ZW3 [26]. A change in the atmospheric circulation in the Amundsen Sea with a weakening of the ASL, likely driven by a tropical transition in austral autumn 2023, contributed to the local sea ice expansion in the Amundsen-Bellingshausen Sea after summer low sea ice conditions [11]. Storms in the Weddell and Ross Seas are likely to have further contributed to the record low conditions [10].

Evidence of impacts of the low sea ice state since 2017 on the physical system are emerging, in accordance with our expectations based on its physical role and function [1, 2]. The dark surface of the open ocean reflects less than a tenth of incoming solar radiation, while the sea ice surface is five to nine times more reflective, with the highest reflectivity occurring when the sea ice is covered with snow. Therefore, losses in sea ice increase the amount of heat absorbed by the planet, termed the albedo feedback effect. The additional energy retained by the Earth due to this effect is estimated as 12‑15% of the additional heat energy trapped in the Earth system due to greenhouse gases over the period 1979-2023 [27, 28].

Along the coastline, some sea ice is held motionless by interaction with the land or grounded icebergs. This `landfast’ sea ice is important for coastal processes and ecosystems, as well as infrastructure and logistics. The extent of landfast ice reached a then-record minimum in March 2022 [29]. During the record conditions in 2023, low summer sea ice led to 154 days of record high coastal exposure [6], defined as the length of coastline lacking any protective sea ice buffer from waves and swell [30]. However, while this may be expected to contribute to increased iceberg formation and ultimately ice-shelf collapse [31], with potential indirect impacts on sea level rise, impacts on such events in 2023 have yet to be assessed. Later in 2023, the remarkable lack of winter sea ice has been linked to increased midwinter heat transfer from the ocean to the atmosphere [32], driving an increase in atmospheric storm frequency. Other potential impacts, such as impacts on globally important ocean circulations [33], the amount of carbon that is stored in the oceans, and the balance of snowfall to melt on the surface of the Antarctic ice sheet, have yet to be assessed and will be challenging to quantify.

Antarctic sea ice is also one of the most extensive, seasonally variable, and productive habitats on Earth [34]. The ice and its snow cover create a unique and highly dynamic environment at the atmosphere-ocean interface. This ice provides a crucially-important home, refuge, breeding platform and food source for many species that are highly adapted to, and reliant on, its presence and seasonal rhythms [35]. These range from microorganisms (including microscopic algae and bacteria), through open-ocean herbivores such as krill, to fish, seabirds (including penguins), seals and whales [36]. Due to its critical role in Southern Ocean ecosystems [2], the occurrence since 2022 of the four lowest summer sea ice extents on record has had implications for many of the species that are dependent upon the sea ice for foraging, moulting, breeding and shelter.

At the lowest levels of the food-chain, phytoplankton have strong connections with sea ice extent and melt. Sea ice melt contributes to phytoplankton blooms by isolating a thin layer of sea ice meltwater at the surface and delivering dissolved iron and algae [37]. Under-ice phytoplankton blooms may account for 10% of net Southern Ocean primary productivity [38], but are particularly difficult to quantify as they cannot be measured by satellite. Highly productive under ice blooms have been detected by autonomous floats [39] highlighting the potential for these blooms to contribute to Southern Ocean primary productivity and the potential for sea ice changes to impact ecological systems. A satellite record [40] shows an increase in the seasonal duration of phytoplankton blooms in the seasonally ice-covered Southern Ocean between 1997 and 2022, as well as a trend towards a more variable, less predictable seasonal cycle. Sea-ice decline also appears to be associated with bigger and longer phytoplankton blooms on the western Antarctic Peninsula [41]. However, this does not necessarily translate into more food availability for krill, as there is evidence of a shift in phytoplankton composition away from their preferred diet, diatoms [42-44].

For all species the consequences of the sudden reduction in sea ice extent over this period has not been fully quantified, but for some species the impacts are already visible. One such is the emperor penguin, which has been assessed to be the Antarctic vertebrate most vulnerable to climate change [45] (see also emperor penguin Portal article [46]). Satellite analysis has recorded an increase in breeding failure due to early sea-ice break-up at a number of colonies [47]. In 2022, sea ice break-up before the onset of the fledging season in December led to 19 out of the 66 known colonies (29%) being impacted. The worst affected region was the Bellingshausen Sea (Fig. 1c) where four out of five colonies had total or near-total breeding failure [48]. Although the time series of satellite observations that was used to observe these events is short (since 2018), there appears to be an increasing number of early ice break-out events. A recent satellite based circum-Antarctic study monitoring emperor penguin population change between 2009-2018 suggested a probable 9.5% decrease in the whole population (even though the time period does not include the records lows in 2022-2024) [49]. Updated demographic projections of future emperor populations incorporating uncertainties and testing a variety of climate models suggest that the emperor penguin may meet criteria from the International Union for Conservation of Nature (IUCN)’s  listing of at least Vulnerable and show that populations in five of the eight identified genetically distinct geographical regions may meet IUCN criteria for Critically Endangered [50].

The impacts on other species can be more geographically variable, and in some cases, a reduction in sea ice can be beneficial. Work from long term monitoring at the Pointe-Géologie archipelago (140°E) showed south polar skuas display higher breeding success and adult survival in years of low local spring and summer sea ice extent [51], while extensive sea ice led to low breeding success in Adélie penguins. Large decreases in Adélie population along Mawson Coast (around 63°E) have also been attributed to heavy sea ice years [52]. However, evidence from a 17-year tagging dataset also at Pointe-Géologie found that juvenile Adélie penguin survival probability was negatively impacted by low sea ice cover near their natal colony right after fledging [53].

Unfortunately, the satellite record is not long enough to provide multi-decadal or centennial context for the recent extreme lows. Multiple reconstructions of sea ice coverage have been produced using atmospheric and ice-core data [54-56]. These independent reconstructions extend back in time over 100 years. They show that, although 1979-2015 were not representative of the full 20^th century, the current extreme lows in Antarctic sea ice are unprecedented in recent history, and that without a structural change in the sea ice system, the probability of three record sea ice extent lows in the last six years is only 0.1% [55]. Model results also suggest the low winter sea ice extent was extremely rare, but up to four times more likely due to climate change, although models do tend to overestimate year-to-year variability when compared with satellite observations [57].

This evidence, coupled with an increase in variability and repetitive year-to-year values (i.e. statistical persistence) since 2007 [5, 19] has led researchers to suggest that Antarctic sea ice may be undergoing a regime shift [5, 19, 55], since these changes in variability and persistence are regarded as (statistical) early warning signs of a critical transition. The extent to which such a shift is due to climate change is as yet unclear, although existing evidence of warming in the subsurface ocean due to rising greenhouse gases [21] driving the shifts, as well as model results [57] and expectations of centennial decline [58], point to a strong role.

The ice-ocean system is known to exhibit poorly understood variability on decadal timescales, which when combined with long-term trends would result in pauses and surges in sea ice decline over many years [59]. However, it is not clear that this interpretation explains the observed changes in sea ice variability or persistence, highlighting the need for further work to understand and interpret recent changes.

While sea ice extent variability and change are relatively well quantified during the satellite era, the sea ice volume remains a key unknown. This means we only have a partial knowledge of the state of Antarctic sea ice, which limits our ability to validate the models upon which future projections depend. Recent progress has been made in estimating long-term records of both snow depth [60] and sea ice thickness [61]. However, both records rely on large assumptions about, for example, the rate of snow loss due to snow-ice formation [60] or the exact surface from which radar altimetry is measuring. Furthermore, calibrating and validating snow and ice thickness datasets is extremely challenging due to a lack of in situ (airborne or surface) observations. Moreover, icebreaker-based observations from the SCAR ASPeCt (Expert Group on Antarctic Sea-ice Processes and Climate) programme, which provide an important validation dataset, are likely biased to provide low measurements, since icebreakers do not travel through the thickest ice [62].

The upcoming Antarctica InSync (Antarctica International Science & Infrastructure for Synchronous Observation) effort, planned for 2027-2030, aims to perform co-ordinated international observational campaigns. This may provide an opportunity to increase the quantity of observations of key variables such as snow and ice thickness and critical sea ice processes, as well as under-ice ocean observations from autonomous underwater robots known as Argo floats [63]. However, specific plans are still evolving.

In the biosphere, while effects on some species have been observed, the impact of the abrupt recent change in sea ice is yet unclear for most species. Urgent effort is needed to assess the implications of sea ice change on the ice obligate species across trophic levels, but also those ecosystems that depend indirectly on sea ice. For longer lived vertebrate species, the implications of reduced reproduction due to changing ice conditions — for example the loss of breeding habitat for seals and seabirds — may not become evident in the adult population for several years, so immediate understanding of the consequences of the physical changes may be challenging. 

Our ability to project future changes relies on having reliable climate models that can accurately represent sea ice. A prominent indicator of whether a given model is reliable is whether it can capture observed changes in sea ice, including its mean state, variability, and trends. For the latter, most models simulate negative trends in Antarctic sea ice since 1979, while historical observations show positive trends from 1979-2015. This discrepancy reduces confidence in the reliability of models’ forward projections.  However, with the recent decline in Antarctic sea ice, the observed trend when extended to 2023 is no longer positive, and is no longer inconsistent with the models [7]. If models can capture the historical changes in sea ice, then their projections of large declines in the 21^st century may be more reliable than previously thought. Nonetheless, climate models are known to have some limitations in simulating the Antarctic sea ice mean state as well as its variability [58] and wider aspects of the Southern Ocean climate in general. Therefore, until models can better capture the physics underlying complex ice-ocean-atmosphere interactions, including the physics driving the likely regime change and the different timescales of variability driving sea ice, confidence in the details of model projections (including the magnitude of change) will remain low.