Uncertainty Quantification for Complex Multiscale Systems

Multiscale modeling efforts must adequately quantify the effect of both parameter uncertainty and model discrepancy across scale. Advancements in uncertainty quantification using Bayesian calibration are described; a dynamic discrepancy approach to upscale uncertainty, functional inputs and extrapolation uncertainty, and a large parameter space. For emulation and discrepancy modeling, a Bayesian Smoothing Spline ANOVA (BSS-ANOVA) approach is utilized. These approaches are applied here to applications in chemical kinetics and carbon capture technology, with wide ranging impact.