Calibration of an Antarctic Ice Sheet model

Time-averaged or snapshot observations of contemporary ice sheet geometry and surface velocity are commonly used in numerical ice-sheet simulations to infer information about ice viscosity and basal sliding. The solution generally depends on the form of the basal sliding law, ice rheology and some form of regularization. On the other hand, close relationships between observed changes in ice sheet geometry and surface velocity are not systematically examined, yet they contain valuable information about the laws that govern ice-sheet dynamics. For example, it can be shown that the functional relationship between perturbations in ice thickness and ice speed depends on the sliding law exponent in a monotonic way. Hence, by harnessing the information contained in successive measurements of ice-sheet geometry and velocity, one can plausibly derive constraints on the form of the sliding law. Here we use a high-resolution numerical setup of the modern-day West Antarctic Ice Sheet to simulate the response of ice speed to contemporary changes in geometry (ice front location and ice thickness between 2000 and 2020). The simulated changes in ice speed are compared to observations over the same period and used in a Bayesian framework to derive constraints on the form of the basal sliding law, ice rheology, and regularization parameters. The a-posteriori distribution of model parameters is used to construct an ensemble of initial states for the whole of the Antarctic Ice Sheet in the year 2000. The ensemble serves as a starting point for hindcast and forecast simulations, with quantified uncertainties for key model parameters.