A surrogate-based approach to nonlinear, non-Gaussian joint state-parameter data assimilation

TL;DR

This paper introduces Emu-PF, a novel data assimilation method combining statistical emulators with particle filters, enabling efficient and accurate joint state-parameter estimation in complex nonlinear models.

Contribution

It presents a new emulator-particle filter approach that reduces computational cost while maintaining accuracy in non-Gaussian, high-dimensional data assimilation problems.

Findings

Emu-PF achieves well-resolved posterior distributions with fewer model runs.

Dimension reduction techniques improve emulator fitting efficiency.

Demonstrated effectiveness on a Lorenz-96 system simulation.

Abstract

Many recent advances in sequential assimilation of data into nonlinear high-dimensional models are modifications to particle filters which employ efficient searches of a high-dimensional state space. In this work, we present a complementary strategy that combines statistical emulators and particle filters. The emulators are used to learn and offer a computationally cheap approximation to the forward dynamic mapping. This emulator-particle filter (Emu-PF) approach requires a modest number of forward-model runs, but yields well-resolved posterior distributions even in non-Gaussian cases. We explore several modifications to the Emu-PF that utilize mechanisms for dimension reduction to efficiently fit the statistical emulator, and present a series of simulation experiments on an atypical Lorenz-96 system to demonstrate their performance. We conclude with a discussion on how the Emu-PF can be paired with modern particle filtering algorithms.