applied math seminar
John Davis Jakeman, Sandia National Laboratories
Monday, November 13, 2017 -
3:30pm to 4:30pm
Title: Towards probabilistic predictions of future sea-level Authors: J.D. Jakeman, M.Perego, S.Price, I. Tezaur Abstract: Polar ice sheets present the largest potential for future sea-level change (~65 m of sea-level equivalent). Confidence in sea-level projections requires accurate simulations of ice sheet evolution using next-generation ice sheet and earth system models as well as analysis of the effect of system uncertainties on predictions. In this talk we present an overview of the ProSPect project, funded by the U.S. Department of Energy's Scientific Discovery through Advanced Computing (SciDAC) program, which been tasked with addressing the challenges of high-fidelity deterministic modeling and uncertainty analysis of ice-sheet evolution models. Focus will be given to research aimed at understanding and limiting the impact of uncertain ice-sheet initial conditions on future sea-level predictions. The approach presented employs Bayesian inference to condition initial estimates of uncertainty on observational data and employs techniques designed to overcome the challenges associated with large number of uncertain parameters and the computational expense of high-fidelity simulation. Here is a short biography. Dr Jakeman is a Foreign National Interim Technical Staff at Sandia National Laboratories in Albuquerque. His research focuses on efficient methods for quantifying uncertainty in high-dimensional, computationally expensive simulation models. Dr Jakeman has specific interest in methods that exploit structure in the model input-output map, including sparsity, low-rank, and lower-dimensional spaces. His research spans, theoretical advances in polynomial approximation, l1 minimization and inference to deployment of UQ methods to applications including, ice-sheet evolution and designing aerospace nozzles under uncertainty.
Contact Name: Pavel Lushnikov