| Summary: | The Antarctic sea ice widely impacts the Earth climate through its interactions with the ocean and the atmosphere. Unlike its Arctic counterpart, the sea ice surrounding Antarctica has been slightly expanding since 1979. A possible explanation for this observed increase relies on the internal variability of the Antarctic sea ice, particularly large at multi-decadal timescales. Unfortunately, reliable observations of sea ice do not span a long enough time period to properly examine how the sea ice varies between decades. The goal of this doctoral thesis is first to investigate the mean state and the internal variability of the Antarctic sea ice simulated by general circulation models (GCMs). This allows checking if these models reproduce reasonably well the observations and if the observed positive trend in sea ice extent is compatible with the internal variability of the models. Second, we assess the impact of the initialisation of a simulation using various data assimilation methods on the skill of the predictions of the Antarctic sea ice at decadal timescales. According to our results, the observed increase in Antarctic sea ice extent is compatible with the internal variability simulated by current GCMs, although it is a relatively rare event. Unfortunately, these models generally display systematic biases in the way they reproduce the mean state and the internal variability of the Antarctic sea ice. In an idealised framework, we show that the interannual variations of the Antarctic sea ice are not predictable beyond three years. Besides, an adequate initialisation clearly improves the skill of the prediction of the multi-decadal trends in sea ice extent, in both idealised and realistic conditions. (SC - Sciences) -- UCL, 2014
|
|---|