The response of Antarctic sea ice to anthropogenic climate change, from model and satellite observations

Global coupled climate models currently do not reproduce observed regional and seasonal trends in Antarctic sea ice extent. Given that sea ice is a fundamental component of the global climate system, it is imperative that climate models accurately capture the processes driving sea ice trends in order to produce reasonable future climate projections. This thesis investigates causes of the disparity between observed and simulated Antarctic sea ice trends. Interactions between large-scale patterns of atmospheric variability and Antarctic sea ice in observation-based estimates and the output of 16 coupled models are examined. The models reasonably reproduce observed interactions during the season of sea ice advance, which is when the influence of the atmosphere on Antarctic sea ice is strongest and also when observed and simulated sea ice trends most strongly diverge. However, during sea ice retreat in the spring (when the ocean heat contribution to sea ice change is larger), the models do not reproduce observed interactions. Large-scale atmospheric variability is strongly linked to sea ice in some sectors but not in others, indicating that in some sectors sea ice is more strongly influenced by ocean heat exchange or atmospheric variability unexplained by the major modes. The dominance of dynamic or thermodynamic processes driving monthly changes in Antarctic sea ice is then examined through a sea ice volume budget in 10 coupled models. Low-magnitude mean sea ice mass transport in 4 of the 10 models is associated with the dominance of thermodynamic processes, while the dominant drivers in the remaining 6 models with higher-magnitude sea ice transport were more spatially and seasonally variable. However, while sea ice transport biases were associated with the dominant drivers of sea ice trends, biases in sea ice concentration did not show a clear link to these drivers. The impact of biases in the Southern Ocean on model representation of mean Antarctic sea ice in the same 10 models is investigated through an analysis of the surface heat budget, freshwater fluxes, and the stability of the ocean seasonal layer. The relative magnitude of mean sea ice extent (compared to observations) separates the models as to the rate of ice formation and melt. For example, models that have relatively low mean sea ice delay sea ice production in favour of cooling the ocean, while models with relatively high mean sea ice rapidly produce sea ice early in the autumn months. Models in which thermodynamic processes dominate sea ice volume change have warmer surface ocean temperatures, and are more sensitive to the ocean thermal state than other models with comparative amounts of sea ice. These models also have higher levels of near-surface ocean stratification. The results of this thesis highlight priority areas for future model development in order to improve representation of mean Antarctic sea ice and thus simulation of future sea ice trends. While accurate representation of the seasonal evolution and magnitude of sea ice concentration, extent and volume is important, it is also necessary that models produce a realistic balance between the dynamic and thermodynamic drivers of sea ice evolution, as well as realistic sea ice motion. Furthermore, the evaluation of model representation of the aforementioned sea ice aspects is limited by our understanding of observed sea ice processes. Therefore, the acquisition of a long-term, continuous, reliable record of sea ice thickness observations is pivotal to more accurately measure seasonal sea ice production and melt, to improve understanding sea ice processes and interactions with the atmosphere, ocean and ice shelf, and to better quantify ocean-atmosphere heat exchange in the sea ice zone of the Southern Ocean.