Parallel numerical modelling of ice flow in Antarctica
This thesis describes the parallel implementation of a three-dimensional numerical ice flow model of the whole of the grounded part of the Antarctic Ice Sheet at a grid resolution of 20km. Numerical modelling of ice flow is computationally intensive as it requires the solution of non-linear equations over long time scales. A parallel model was developed to overcome these restrictions, and it is demonstrated that the model runs more quickly on multiple processors than on a single processor (70% efficiency on four processors). The model was successfully validated against published benchmarks and compared against other models and remote sensing work. The main ice flow features are well reproduced, including some newly observed fast flow features in East Antarctica. The optimal run-time versus efficiency was exploited to run a series of detailed sensitivity tests on parameters that may affect the resulting ice sheet volume and basal thermal regime. Compared with the effects of surface air temperature, the accumulation rate and tuning parameter m in flow parameter A., geothermal heat flux was found to have the strongest effect on basal melting. It is shown that use of different geothermal heat flux values can affect the inclusion of sub-glacial lakes in the zone of basal melting. Topographic smoothing may reduce the model’s ability to locate subglacial lakes. Fast flow features appear in the modelled ice sheet despite the lack of basal slip conditions in the model. Use of a new topography data set improved the model’s ability to locate subglacial lakes in zones of basal melting, and revealed additional fast flow features in East Antarctica.