Numerical Stabilisation of Anisotropic Ice-Sheet Models

Objective

The Antarctic Ice Sheet is the largest source of uncertainty in sea-level projections, highlighting the need for significant improvements in model physics, including the internal friction of ice, known as its viscosity. Large-scale ice-sheet models currently misrepresent the viscosity of ice by neglecting key physical properties, which can significantly affect ice-sheet simulations. Similar to how cards in a deck can easily slip against each other in one direction compared to trying to split the deck in another direction, ice slips up to 60 times more easily in one direction due to its crystal structure. Despite this substantial difference, large-scale ice-sheet models do not account for this direction-dependent sliding, known as anisotropy. Although extensive literature on the implementation of anisotropy in ice-flow models exists, development and widespread application has been delayed by numerical issues and a lack of observational data. However, recent advances in radar technology, capable of measuring ice crystal structure have shifted the focus towards resolving the remaining numerical issues associated with anisotropic ice-flow modelling. This project builds on existing anisotropic models by implementing numerical stabilisation schemes and using radar data to constrain the model, enabling simulations of areas where current anisotropic models face numerical instability, particularly in fast-flowing and floating ice. A long-term impact is to reduce sea-level rise uncertainty by providing an ice-sheet model that is physically more accurate and numerically stable. The project will benefit from BAS’s strong network of ice-sheet modelling experts, providing the necessary support to effectively tackle the complexity of anisotropic ice-flow modelling. Building on expertise in Numerical Mathematics, Scientific Computing and Geophysics, the project will position the researcher as an emergent expert in anisotropic ice-flow and next-generation forecast modelling.