Bayesian Inversion

Predicting future sea level rise

Modeling the ice sheet dynamics in Antarctica is critical to predicting future sea level rise.

Learn more ↓

Oden Institute Center for Computational Geosciences and Optimization

Bayesian inversion for basal friction in Antarctica

InSAR observational data provide ice surface velocity observations. To make predictions about future behavior, we must somehow estimate the dynamics of the full ice sheet, even though our data are limited to surface observations.

The basal friction field is a critical parameter that feeds into these predictions – it characterizes just how quickly the base of the ice sheet is slipping in each spatial location. Since we cannot observe the basal friction directly, we must use a physics-based model that relates basal friction to ice surface velocity. A large-scale Bayesian inverse problem formulation then provides the mechanism to learn the basal friction—and its uncertainty—from the surface velocity data, through the lens of the physics-based ice dynamics model.

Once the basal friction is characterized, we then use the physics-based model to issue predictions with quantified uncertainties on the contribution of Antarctic ice sheets to future sea-level rise.

These physics-based models use advanced discretization methods, mesh adaptivity, and scalable parallel solvers to resolve the complicated dynamics of ice sheet flow.

InSAR-based ice surface velocity observations

Dynamics of ice sheet flow

Predictions with quantified uncertainties on the contribution of Antarctic ice sheets to future sea-level rise

Convergence