Large-scale thermo-mechanical modelling of Greenland ice sheet

Antarctic and Greenland ice sheets lose most of their mass by a few corridors of rapidly flowing ice. These ice conveyor belts constitute fast drainage routes whose flow velocities are undoubtedly sensitive to climate perturbations directly impacting sea-level. Observations suggest the ice is rather sliding than flowing, the key being where sliding is accommodated. Commonly, sliding occurs at the ice-bedrock interface, but recent studies favour englacial sliding to explain data from Western Margin of Greenland. 

Our aim is to demonstrate that the mechanisms controlling the spontaneous formation of englacial sliding explains the transition from slow flowing to fast sliding ice over Greenland. We employ a new thermo-mechanical ice flow model to predict thermally activated creep instability leading to the spontaneous rearrangement of ice motion in three dimensions. Accurately resolving these nonlinear interactions on regional to ice sheet scales requires high spatial and temporal resolution which can only be achieved using a supercomputer.

We present a new thermo-mechanical ice flow model, FastIce.jl, that is capable of predicting the evolution of ice sheet at unprecedented scale. The model uses the full-Stokes formulation for the ice flow and the enthalpy method for describing the polythermal ice behaviour. FastIce.jl uses GPU acceleration for solving the flow equations, resulting in close to ideal scaling in distributed computing benchmarks. We compare our simulation results with other full-Stokes models, and present the results of simulating the 120x120 km regions of Greenland ice sheet at 10m resolution. High resolution allowed us to capture the transition from slow to sliding flow regimes without any simplifying assumptions.