Abstract:The use of model order reduction techniques in combination with ensemble-based methods for estimating the state of systems described by nonlinear partial differential equations has been of great interest in recent years in the data assimilation community. Methods such as the multi-fidelity ensemble Kalman filter (MF-EnKF) and the multi-level ensemble Kalman filter (ML-EnKF) are recognized as state-of-the-art techniques. However, in many cases, the construction of low-fidelity models in an offline stage, before solving the data assimilation problem, prevents them from being both accurate and computationally efficient. In our work, we investigate the use of {\it{adaptive}} reduced basis techniques in which the approximation space is modified online based on the information that is extracted from a limited number of full order solutions and that is carried by the past models. This allows to simultaneously ensure good accuracy and low cost for the employed models and thus improve the performance of the multi-fidelity and multi-level methods.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to improve the accuracy and computational efficiency of the set - based Kalman filtering method in dealing with the state estimation problems of systems described by nonlinear partial differential equations (PDEs). Specifically, the paper focuses on how to improve the performance of multi - fidelity and multi - level ensemble Kalman filters (MF - EnKF and ML - EnKF) through adaptive reduced - basis models (RBM) techniques. ### Main Problems 1. **Limitations of Off - line Construction of Low - Fidelity Models**: - In many cases, the methods for constructing low - fidelity models in the off - line phase cannot guarantee high precision and computational efficiency simultaneously. - Due to the lack of sufficient understanding of the initial state of the system and the complexity of long - term dynamics, the off - line - constructed reduced - order space may not be able to effectively capture the changes of the unknown "true" state trajectory. 2. **The Need for Dynamic Adjustment of the Reduced - Order Space**: - Existing methods usually rely on static or pre - determined reduced - order spaces, which perform poorly in the face of complex, non - periodic, or chaotic dynamics. - A method that can adjust the reduced - order space online during the data assimilation process is required to ensure that it can maintain both good approximation accuracy and low computational cost. ### Proposed Solutions The paper proposes two adaptive algorithms: the Adaptive Reduced - Order Multi - level Ensemble Kalman Filter (aRB - ML - EnKF) and the Adaptive Reduced - Order Multi - fidelity Ensemble Kalman Filter (aRB - MF - EnKF). These algorithms solve the above problems in the following ways: 1. **Online Re - training**: - Use high - fidelity state trajectories to expand the approximation space (model inflation), thereby continuously updating the reduced - order space. 2. **Preserving Past Information**: - Use past low - fidelity simulation results to determine which subspace directions can be discarded without losing accuracy (model contraction). 3. **Local - Time Reduced - Order**: - Construct the reduced - order space only within a relatively short time range to reduce the size of the required training set and ensure that the dimension of the approximation space is small, thereby saving computational time. ### Specific Applications The paper conducts numerical experiments through the quasi - geostrophic equation (QGE), compares the performance under different surrogate models, and verifies the effectiveness of the proposed methods. ### Summary This research aims to overcome the limitations of existing methods in constructing reduced - order spaces by introducing adaptive reduced - order techniques, thereby enhancing the performance of multi - fidelity and multi - level ensemble Kalman filters in dealing with complex system state estimation problems.

An adaptive hierarchical ensemble Kalman filter with reduced basis models

A Reduced Basis Ensemble Kalman Method

Adaptive Multi-Fidelity Probabilistic Collocation-Based Kalman Filter for Subsurface Flow Data Assimilation: Numerical Modeling and Real-World Experiment

A two-stage ensemble Kalman filter based on multiscale model reduction for inverse problems in time fractional diffusion-wave equations

Multiscale-finite-element-based ensemble Kalman filter for large-scale groundwater flow

Data Assimilation for Unsaturated Flow Models with Restart Adaptive Probabilistic Collocation Based Kalman Filter

A Multi‐Timescale EnOI‐like High‐Efficiency Approximate Filter for Coupled Model Data Assimilation

A Stochastic Collocation Based Kalman Filter for Data Assimilation

Dynamic Reservoir Data Assimilation with an Efficient, Dimension-Reduced Kalman Filter

A Bayesian Consistent Dual Ensemble Kalman Filter for State-Parameter Estimation in Subsurface Hydrology

Adaptive Multifidelity Data Assimilation for Nonlinear Subsurface Flow Problems

Deterministic ensemble Kalman filter based on two localization techniques for mitigating sampling errors with a quasi-geostrophic model

Sequential Ensemble-Based Optimal Design for Parameter Estimation

A multigrid/ensemble Kalman Filter strategy for assimilation of unsteady flows

Model error covariance estimation in particle and ensemble Kalman filters using an online expectation-maximization algorithm

A Kalman filter powered by $\mathcal{H}^2$-matrices for quasi-continuous data assimilation problems

High‐dimensional ensemble Kalman filter with localization, inflation, and iterative updates

Ensemble Randomized Maximum Likelihood Method as an Iterative Ensemble Smoother

Relevance of random biopsy at the transformation zone when colposcopy is negative.

A Hybrid Nudging-Ensemble Kalman Filter Approach to Data Assimilation. Part I: Application in the Lorenz System

Localizing the Ensemble Kalman Particle Filter