B0202
Title: Non-Gaussian emulation of climate models via scalable Bayesian transport maps
Authors: Matthias Katzfuss - University of Wisconsin-Madison (United States) [presenting]
Abstract: A multivariate distribution can be described by a triangular transport map from the target distribution to a simple reference distribution. Bayesian nonparametric inference is proposed on the transport map by modelling its components using Gaussian processes. This enables regularization and accounting for uncertainty in the map estimation while resulting in a closed-form invertible posterior map. The focus is on inferring the distribution of a spatial field from a small number of replicates. Specific transport-map priors are developed that are highly flexible but shrink toward a Gaussian field with Matern-type covariance. The approach is scalable to high-dimensional fields due to data-dependent sparsity and parallel computations. Numerical results are presented to demonstrate the accuracy, scalability, and usefulness of the generative methods, including emulation of non-Gaussian climate-model output.
