Efficient quadratures for high-dimensional Bayesian data assimilation

doi:10.1016/j.jcp.2024.112945

Efficient quadratures for high-dimensional Bayesian data assimilation

2024-06-01

0

PRE

AI

Daniel M. Tartakovsky *

DOI：10.1016/j.jcp.2024.112945

Original

Original request for help

Save

Abstract

En 中文

Bayesian update is a common strategy used to combine (uncertain) model predictions and (noisy) observational data. A computational bottleneck in this data assimilation technique is the evaluation of high -dimensional quadratures involving multivariate probability density functions (PDFs) of system states. We explore designed quadratures as a means to reduce the computational cost of Bayesian update of multivariate joint PDFs. A series of numerical experiments demonstrate that our method outperforms stochastic collocation on sparse grids, a popular technique used to perform high -dimensional integration in the context of uncertainty quantification, in terms of both accuracy and computational efficiency.

Keywords:

Data assimilation

Designed quadrature

Joint probabilistic density function

Efficient quadratures for high-dimensional Bayesian data assimilation

Abstract

Journal

Organization