Hybrid iterative ensemble smoother for history matching of hierarchical models

TL;DR

This paper introduces a hybrid iterative ensemble smoother for hierarchical models in data assimilation, demonstrating improved sampling performance over standard methods in inverse problems and flow simulations.

Contribution

It proposes a novel hybrid approach combining RML and IES for better posterior sampling in hierarchical models, especially with small ensembles.

Findings

Hybrid IES samples effectively with small ensembles.

Standard IES performs poorly in flow examples due to local sensitivity issues.

Hybrid method outperforms standard IES in complex flow problems.

Abstract

The choice of the prior model can have a large impact on the ability to assimilate data. In standard applications of ensemble-based data assimilation, all realizations in the initial ensemble are generated from the same covariance matrix with the implicit assumption that this covariance is appropriate for the problem. In a hierarchical approach, the parameters of the covariance function, for example the variance, the orientation of the anisotropy and the ranges in two principal directions, may all be uncertain. Thus, the hierarchical approach is much more robust against model misspecification. In this paper, three approaches to sampling from the posterior for hierarchical parameterizations are discussed: an optimization-based sampling approach (RML), an iterative ensemble smoother (IES), and a novel hybrid of the previous two approaches (hybrid IES). The three approximate sampling methods are applied to a linear-Gaussian inverse problem for which it is possible to compare results with an exact "marginal-then-conditional" approach. Additionally, the IES and the hybrid IES methods are tested on a two-dimensional flow problem with uncertain anisotropy in the prior covariance. The standard IES method is shown to perform poorly in the flow examples because of the poor representation of the local sensitivity matrix by the ensemble-based method. The hybrid method, however, samples well even with a relatively small ensemble size.