Atmosphere Ocean Science Colloquium

Ice sheet model initialization as a Bayesian inverse problem

Speaker: Georg Stadler, CIMS

Location: Warren Weaver Hall 1302

Date: Wednesday, December 13, 2017, 3:30 p.m.

Synopsis:

Due to the amount of water stored in the Antarctica and Greenland ice sheets, they play a crucial role in predictions of future sea level change. Estimation of the present day ice sheet state (e.g., temperature distribution, basal boundary conditions) from surface satellite (and other) observations is crucial for these predictions.  I will present this ice sheet model initialization problem as a Bayesian inverse problem.  This problem is challenging due to the high dimension of the inversion field and the expensive-to-solve governing equations (which are nonlinear Stokes equations describing the gravity-driven viscous flow of ice). These challenges increase if one is interested in quantifying the degree of uncertainty---due to limited surface observations and the imperfect model equations--- in the model initialization. These uncertainties translate to uncertainties in predictions of sea level change. I will discuss approximations and solution methods for the inverse problem, and solvers for the governing nonlinear Stokes equations.