An Ensemble Kalman-Particle Predictor-Corrector Filter for Non-Gaussian Data Assimilation

TL;DR

This paper introduces a novel ensemble Kalman-particle predictor-corrector filter that effectively handles non-Gaussian data in high-dimensional state estimation, combining strengths of both EnKF and PF methods.

Contribution

The paper proposes a new hybrid filter that integrates EnKF and PF using nonparametric density estimation, suitable for high-dimensional PDE-based models.

Findings

Effective in high-dimensional state spaces

Combines advantages of EnKF and PF

Demonstrated on numerical examples

Abstract

An Ensemble Kalman Filter (EnKF, the predictor) is used make a large change in the state, followed by a Particle Filer (PF, the corrector) which assigns importance weights to describe non-Gaussian distribution. The weights are obtained by nonparametric density estimation. It is demonstrated on several numerical examples that the new predictor-corrector filter combines the advantages of the EnKF and the PF and that it is suitable for high dimensional states which are discretizations of solutions of partial differential equations.