| Summary: | Frozen sea water, called sea ice, is an important actor of the climate system. It covers about 12% of the world's oceans. More than reflecting the incoming light, it regulates the exchanges of heat, momentum and matter between the ocean and the atmosphere in polar regions. More extended than its Arctic counterpart, the Antarctic sea ice actively participates in the redistribution of water masses in the world’s major ocean basins. Contrary to what is commonly believed, the Antarctic sea ice has been relatively unaffected by global warming. Until recently, satellite observations even showed a small positive sea ice cover trend. This trend is punctuated by large interannual variations, with a record-high cover in 2014 and a record-low cover in 2017. This makes the Antarctic climate unique and sea ice predictions challenging. At short timescales, predictions are subject to errors originating from incorrect initial conditions (ICs), model imperfections, and by “chaosâ€. While we can act to reduce the first two sources of errors, chaos is inherent to fully coupled climate models. Focusing on this source of error using an idealised protocol, this thesis demonstrates that such models can provide skilful sea ice edge predictions. The predictability is accounted for by the ocean with its great thermal inertia. Unfortunately, we showed that there is still a large predictability gap between idealised and operational predictions. The dearth of observations is problematic to start a prediction. Our results suggest that the errors in the ocean–sea ice ICs could even dominate the errors coming from an incorrect representation of the atmospheric conditions. The imperfect representation of the Antarctic climate in models is another major obstacle. A better observational coverage would certainly help to fix both issues. (SC - Sciences) -- UCL, 2021
|
|---|