1.2.2.2 Sea ice

Sea ice is frozen sea water that floats on the sea surface. It forms in the polar oceans whenever the temperature of the sea water drops below its freezing point of around -1.8°C. The formation and growth of sea ice therefore requires a sufficient heat loss from the ocean to the atmosphere, which in today’s climate occurs in both polar regions from autumn to spring. During this period, sea ice is expanding, while during summer it is retreating. 

While the formation of sea ice through heat loss to the atmosphere is similar in both polar regions, the dominating process for sea ice decay in summer differs between the two hemispheres. In the North, where the sea ice is largely landlocked by the land masses surrounding the pole, the loss of sea ice is primarily driven by atmospheric heat input that melts the sea ice. In the southern hemisphere, however, the summer loss of sea ice is primarily governed by the export of sea ice through northward winds that move the ice into regions of warmer sea water, which then melts the ice from below. The freeze-melt cycle of sea ice gives rise to substantial seasonal variations in the polar sea ice coverage (Figure 1.2.7 and Figure 1.2.9), whose magnitude is an indicator for the very fast response time of sea ice, in particular relative to other cryospheric systems such as permafrost, glaciers and ice sheets.

Given the different processes that are relevant for the regional and seasonal response of sea ice to global warming, in the following we differentiate our assessment of tipping potential between Arctic summer sea ice, Arctic winter sea ice, Barents Sea ice, and Southern Ocean sea ice.

Arctic summer sea ice

Evidence for tipping dynamics

In summer, the retreating sea ice cover in the Arctic exposes the much darker ocean surface to the atmosphere, giving rise to the ice-albedo feedback: Less ice implies an additional uptake of heat, implying further ice loss. This mechanism was hypothesised to give rise to a nonlinear tipping point behaviour for the loss of Arctic summer sea ice (e.g., Lenton et al., 2008). However, a large variety of studies based on both conceptual models and coupled Earth system models have provided convincing evidence that the summer ice-albedo feedback is compensated by damping feedbacks in winter that minimise the long-term memory of the Arctic summer sea ice cover (Figure 1.2.8). This dominance of negative/damping feedbacks gives rise to a linear retreat of the Arctic summer sea ice cover with ongoing global warming (e.g., Gregory et al., 2002; Winton, 2006; Winton, 2008; Notz, 2009; Tietsche et al., 2011; Mahlstein and Knutti, 2012: Wagner and Eisenman, 2015). 

Based on this understanding, the response of the sea ice cover to global warming is expected to remain linear as a function of global mean temperature (e.g., Gregory et al., 2002; Winton, 2011; SIMIP 2020) and thus as a function of CO₂ emissions (Zickfeld et al., 2012; Notz and Stroeve, 2016) until the complete loss of the summer sea ice cover that is expected to occur for the first time before 2050 in all future climate scenarios (SIMIP, 2020; Kim et al., 2023). If, in the future, atmospheric CO₂ were to decrease, for example by the technological removal of CO₂, there would be some time lag before global temperature would decrease in response. This hysteresis then carries over to the relationship between CO₂ concentration and sea ice area. The relationship between sea ice area and hemispheric mean temperature, however, has been found to remain linear also for a cooling climate (e.g., Armour et al., 2011; Li et al., 2013; Jahn, 2018). 

Assessment and knowledge gaps

The assessment of a linear, threshold-free loss of Arctic summer sea ice is in line with recent assessments (Fox-Kemper et al., 2021; Armstrong McKay et al., 2022). Given the very broad evidence base, we have high confidence in the assessment of Arctic summer sea ice not being a tipping system. This confidence could be increased further if climate models would more reliably capture the observed evolution of the Arctic sea ice cover – for example regarding its linear sensitivity to observed global warming (SIMIP, 2020). A comprehensive assessment of climate model performance is, however, hampered to some degree by the difficulty to obtain reliable, long-term observations of the sea ice thickness distribution (SIMIP, 2020). Some progress in this regard can be expected in the near future, with the recent development of an approach to retrieve sea ice thickness throughout the entire seasonal cycle using remote sensing (Landy et al., 2022).

Arctic winter sea ice

For the loss of summer sea ice, the existing ice cover needs to be melted completely, which is a gradual process. The loss of winter sea ice, however, is governed by a different mechanism: given that the Arctic will already be ice-free in summer, the formation of new ice needs to become impossible to lose the winter sea ice cover. Winter sea ice will form in the Arctic Ocean as long as the water temperature at the ocean surface drops below the freezing point – around -1.8⁰C for typical saline ocean water – but will no longer form once the water temperature remains above freezing all year round. This binary behaviour of the Arctic Ocean lies at the heart of the analysis of the ongoing loss of the Arctic winter sea ice cover. 

Evidence for tipping dynamics

Both in some simple models and in some complex climate models, the loss of Arctic winter sea ice area accelerates drastically once a given warming threshold has been reached (e.g., Winton, 2006; Eisenman and Wettlaufer, 2009; Bathiany et al., 2016). However, this acceleration is simply a consequence of the geometry of the Arctic Ocean: as the climate warms, the winter sea ice edge moves northward. As long as the ice edge is located in the narrow straits that connect the Arctic Ocean to the south, the freely moving ice edge is short and only a little ice is lost by its northward movement. Once the ice edge becomes located in the central Arctic Ocean, more sea ice area is lost for a given retreat of the ice edge, and ice loss accelerates. This acceleration therefore occurs in most models as soon as the winter maximum sea ice area drops below around 8m sq km, which is roughly the area of the Arctic Ocean and its adjacent seas (Goosse et al., 2009; Eisenman, 2010). 

Beyond this threshold, the loss of the winter sea ice cover occurs faster than the loss of the summer sea ice in CMIP5 models. This can be explained by the fact that the future formation of winter sea ice from a largely ice-free ocean will lead to a geographically rather homogenous distribution of winter sea ice thickness, such that larger areas can become ice-free simultaneously (Bathiany et al., 2016). 

In modelling studies, the faster loss in winter compared to summer has additionally been found to be related to the increased humidity and the related increased downward longwave radiation, for example from convective clouds in areas of open water (Abbot and Tziperman, 2008; Abbot et al., 2009; Li et al., 2013; Hankel and Tziperman, 2021). While this process could potentially imply hysteresis behaviour of the loss of Arctic winter sea ice, the loss of winter sea ice has been shown to be fully reversible in a number of dedicated modelling studies (Armour et al., 2011; Ridley et al., 2012; Li et al., 2013). In particular, for a cooling of the climate induced by the removal of CO₂, studies have found no hysteresis of Arctic winter sea ice area as a function of hemispheric mean temperature, while they found a time lag between the decrease of atmospheric CO2 concentration and the resulting increase of Arctic winter sea ice area. This can be explained by the delayed response of atmospheric temperature to the removal of CO₂, and the potential nonlinear response of oceanic heat transport (Li et al., 2013; Schwinger et al., 2022).

Assessment and knowledge gaps

Based on this assessment, there is currently only very limited support for a dominating role of self-perpetuating processes that would make Arctic winter sea ice a tipping system. Given the difficulty of climate models to realistically simulate the processes that govern the loss of winter sea ice and the related oceanic response, we have medium confidence in the assessment of Arctic winter sea ice not being a tipping system.

Barents Sea ice

Sea ice in the Barents Sea – the sector of the Arctic Ocean north of Scandinavia and Western Russia – is treated as a sub-case of Arctic winter sea ice in Armstrong et al., (2022), who categorised it as a regional impact climate tipping system with medium confidence.

Evidence for tipping dynamics

In the Barents Sea, which is only ice-covered in winter, sea ice loss is primarily driven by an increase in lateral oceanic heat inflow of warm Atlantic water (Docquier et al., 2020; Smedsrud et al., 2021; Muilwijk et al., 2023). Because of this tight coupling, in almost all models the sea ice loss is largely linearly related to changes in oceanic heat transport (Docquier et al., 2020) with only one model showing an abrupt loss of the Barents Sea sea ice cover in winter in a dedicated study (Drijfhout et al., 2015). The loss of the Barents Sea winter sea ice cover might reinforce itself through related changes in atmospheric circulation, but there is no consensus among studies that examined these linkages (e.g., Haarsma et al., 2021; Smith et al., 2022 and references therein). The sea ice loss could also reinforce itself through a related increase in the inflow of warm Atlantic water (Lehner et al., 2013) but very few studies have examined this in detail. 

Assessment and knowledge gaps

In summary, there is currently no clear support for the Barents Sea winter sea ice cover being a tipping system. We have low confidence in this assessment, given the very low number of respective studies.

Southern Ocean sea ice

In the Southern Ocean, the amount of sea ice is much more dominated by the combination of oceanic and atmospheric processes than in the Arctic, which gives rise to a much more pronounced seasonal cycle of the Antarctic sea ice area compared to the Arctic (Figure 1.2.9). Generally, the area of sea ice in the Southern Ocean is determined by the balance of ice formation near the continent and ice melt through oceanic heat further away from the coast, where the ice is advected by the prevailing winds and currents. Variations in ice coverage can therefore largely be explained by weaker northward transport of the ice, by increased melting from increased upward oceanic heat transport, and/or by weakened ice formation (e.g., Maksym, 2019). The regional distribution of sea ice growth with its related brine release, and sea ice melt with the release of freshwater, in turn affects the stratification and circulation of the Southern Ocean (see Chapter 1.4 and e.g., Abernathey et al., 2016).  

Over the full satellite record from 1979 onwards, there is no significant trend in Antarctic sea ice coverage (e.g., Fox-Kemper et al., 2021). The maximum sea ice coverage of the observational record was recorded in 2014, while the minimum sea ice coverage was recorded in 2022/2023 (Figure 1.2.9). The low ice coverage of the past two years can be linked to changes in the prevailing wind patterns that are caused by changes in the prevailing large-scale atmospheric modes ( e.g., Zhang and Li, 2023; Wang et al., 2023), and 2023’s historic low has been suggested to represent a new low ice regime resulting from ocean warming (Purich and Dodderidge, 2023). However, given the shortness of the signal, it is currently unclear whether this change in the sea ice forcing will persist, which then could cause a significant, long-term decline of the Antarctic sea ice cover. 

Evidence for tipping dynamics

Given the very long response time of the Southern Ocean to climatic changes, and given the potential long-term changes in the Southern Ocean circulation in response to irreversible changes in ice sheet dynamics, hysteresis behaviour can be expected to exist for the long-term loss of Southern Ocean sea ice. Such hysteresis is indeed identified in a number of dedicated studies (Ridley et al., 2012; Li et al., 2013), but is explained by a lagged response of the sea ice cover to the imposed warming and cooling. This dynamic hysteresis behaviour is therefore a consequence of the long response time of the Southern Ocean. Whether or not one considers this behaviour truly hysteretic is a question of the timescales of relevance. 

Assessment and knowledge gaps

There is currently limited evidence for a self-amplification of Southern Ocean sea ice loss, and we cannot estimate a related temperature threshold. We have low confidence in the assessment of the future evolution of Antarctic sea ice given the difficulties of large-scale climate models to reproduce its observed evolution. This shortcoming of the models might be related to the dominating impact of small-scale eddies in the ocean which low-resolution climate models cannot explicitly resolve. Another shortcoming is the current absence of reliable satellite retrievals of Southern-Ocean sea ice thickness that would be crucial for a detailed model evaluation. This is expected to be addressed with new satellite technologies including, for example, the Surface Water and Ocean Topography (SWOT) mission (Armitage and Kwok, 2021).