1.6.2.1 Cryosphere: Ice sheets and sea ice

Ice sheets

Tipping points in the Greenland and West Antarctic ice sheets are detailed in Chapter 1.2, and several studies have looked for EWS. In West Antarctica, Rosier et al. (2021) searched for EWS for marine ice sheet instability on the Pine Island Glacier, identifying changes in recovery time and looking at the variance of the system state in a model. The EWS were applied to model output and successfully used to pinpoint tipping points. They find the tipping point that leads to total collapse of the glacier occurs at a +1.2ºC ocean temperature increase, relative to initial conditions. 

In Greenland, Boers and Rypdal (2021) found significant increases in variance and autocorrelation in detrended ice core-derived melt records from the central-western part of the Greenland Ice Sheet (GrIS), suggesting that this part of the ice sheet might be close to a tipping point. While they rule out that these EWS are directly caused by changes in temperature or precipitation, the exact mechanisms leading to the observed signs of stability decline remain unclear. The melt-elevation feedback (Levermann and Winkelmann, 2016) acts mostly on timescales longer than what can be captured by the data used by Boers and Rypdal (2021), so other positive/amplifying feedbacks related to much shorter timescales likely dominate. As mentioned by Boers and Rypdal (2021) these include the melt-albedo feedback, related to snowline migration and albedo reductions once the uppermost, white firn layer has melted and the darker grey ice is exposed (Ryan et al., 2019), as well as thinning of outlet glaciers, which accelerates the ice flow upstream (Aschwanden et al., 2019). 

Sea ice

Arctic sea ice loss has previously been proposed as a potential tipping system, but in this report both summer and winter sea ice loss are categorised as unlikely to feature tipping thresholds beyond which feedback-driven self-sustaining loss occurs, with other factors driving abrupt losses (1.2.2.2). In simplified models, however, Merryfield et al. (2008) found increasing variance and AR(1) in sea ice area before abrupt summer loss in a single column, two-season model. In another single-column model with a continuous seasonal cycle (Eisenman and Wettlaufer, 2009), Moon and Wettlaufer (2011) found that the destabilising ice-albedo feedback leads to CSD before the loss of winter sea ice.

While these results are apparently in agreement with expectations from simple dynamical systems, Arctic sea ice is an example of how additional caveats can obscure EWS, leading to ‘false alarms’. Although attempts have been made using empirical data (Livina and Lenton, 2013), using total Arctic sea ice area as a variable could lead to misleading EWS, due to the different amounts of area masked by the continents in different climates (Goose et al., 2009; Eisenman, 2010). Moreover, several nonlinear feedbacks can dominate the recovery time and obscure CSD far from tipping; once ice gets thinner, its heat conductivity decreases, making its response to atmospheric temperature anomalies much faster (Thorndike et al., 1975). Also, a warmer Arctic means a longer period of open water after summer sea ice loss, which introduces a longer timescale. This effect is independent of the nonlinearity of the winter sea ice loss, and could cause EWS false alarms (Wagner and Eisenman, 2015; Bathiany et al., 2016a).Alternative EWS which can also work in seasonal systems where the balance of feedbacks can change during the year (Moon and Wettlaufer, 2011) include measuring the amplitude and phase lag relative to the forcing (Williamson et al., 2016). Also, there are indications that abrupt loss of winter sea ice are still possible, but could potentially be predicted on the basis of the homogeneity in the ice-thickness distribution (Bathiany et al., 2016b; see 1.2.2.2).