E4.8 SciDAC ProSPect 2: Optimization, V&V, and UQ

Abstract

The contribution to sea-level rise from ice sheets is increasing. Observed acceleration in the rate of ice loss from the Greenland and Antarctic ice sheets is a concern, particularly for the West Antarctic Ice Sheet, which is largely grounded below sea level. This geometric configuration makes Antarctica susceptible to a dynamic instability that could result in a catastrophic collapse of one or more ice shelves as a result of relatively small perturbations at the ice sheet's margins. While ice sheet models have become significantly more advanced over the past decade, they are still incomplete in many ways; hence, there are large uncertainties in these models' projections of future sea-level rise. The goal of the ProSPect SciDAC partnership is to address current limitations of DOE ice sheet and Earth system models that limit their use for making accurate sea-level projections. These include inadequate or missing model physics, incomplete couplings between models, and deficiencies in methods for model initialization, validation, and uncertainty quantification (UQ). This poster focuses on recent ProSPect efforts towards: (1) improved model initialization using PDE-constrained optimization methods constrained by the ice sheet flow model and observations, (2) Verification and Validation through ongoing development of the Land Ice V&V toolkit (LIVVkit), and (3) the development and initial deployment of an end-to-end UQ framework consisting of the solution of an approximate Bayesian inference problem followed by forward simulations of posterior samples.