Efficiently constraining parameter uncertainty in a General Circulation Model using targeted data

O.Dunbar, Caltech (USA)

Abstract: Climate prediction relies upon closure models for subgrid scale processes that are unfeasible to resolve globally. These closures feature model parameters, and there is a strong interest in quantifying the uncertainty of these parameters. Regional data may available and can be used to learn about these parameters. The research activities are focused on efficiently finding regions of optimally informative data, useful for learning. The talk considered a closure for moist convection within an idealized aquaplanet general circulation model (GCM). A Calibrate -- Emulate -- Sample (CES) philosophy was used to feasibly perform uncertainty quantification on closure parameters, making use of Gaussian process emulation. The talk introduced Bayesian design to the CES framework to locate optimally informative data subsets.