Computer Model Calibration with High and Low Fidelity Model Output for Spatio-Temporal Data
March 8, 2012
3:45 p.m.
William Kleiber
Abstract
We discuss a framework for computer model calibration based on high and low fidelity model output for large spatio-temporal datasets. The physical model in use is a coupled magnetohydrodynamic-ionospheric model for solar storms occurring in Earth’s upper atmosphere. The statistical model is informed by physical equations and is tailored to large datasets. Given initial runs of the computer model, we propose a sequential design based on Expected Improvement. We end with extensions to multiple variables, developing a nonstationary multivariate covariance function approach.