Abstract

Any numerical model of Antarctic ice sheet needs to address the possibility of grounding line migration. That can be computationally difficult because the horizontal scale length of features close to the grounding line is tiny—on the order of a square kilometre—compared to the millions of square kilometres covered by the ice sheet. Conventional discretization methods can take advantage of non-uniform meshes, with fine resolution in exciting parts of the domain and coarse resolution in dull parts, but are then faced with the possibility that the grounding line might move and sweep out much of the domain, making all of it potentially exciting.

The BISICLES ice sheet model is a conventional finite volume model that employs a block-structured mesh of rectangular cells to obtain high resolution at the grounding line. In some ways that is restrictive compared to (say) an unstructured mesh of triangles, but it has three points in its favour. Most importantly, it is fairly straightforward to compute and use new meshes as the simulation progresses, but on top of that there is an obvious domain decomposition that can be exploited for parallel computing. It is also natural to construct a geometric multigrid solver for the stress-balance equations, although recent developments point us towards a more generic class of multigrid methods.