Latent Autoencoder Ensemble Kalman Filter for Nonlinear Data Assimilation
SIAM Journal on Scientific Computing · 31.12.2026 · 48 (6) · C1403-C1436
DOI: 10.1137/26m1850230 ↗Kurz gefasst
Fachartikel in der Zeitschrift SIAM Journal on Scientific Computing, online erschienen am 03.11.2026. Verfasst von Xin T. Tong und 2 weiteren Autor:innen. Die Arbeit nennt 51 Literaturangaben. Gefördert von: Jiangsu Provincial Scientific Research Center of Applied Mathematics, National Natural Science Foundation of China, Ministry of Education, Singapore.
Abstract (Originalsprache)
The ensemble Kalman filter (EnKF) is widely used for data assimilation in high-dimensional systems, but its performance often deteriorates for strongly nonlinear dynamics due to the structural mismatch between the Kalman update and the underlying system behavior. In this work, we propose a latent autoencoder ensemble Kalman filter (LAE-EnKF) that mitigates this limitation by reformulating the assimilation problem in a learned latent space with stable linear dynamics. The proposed method learns a nonlinear encoder–decoder together with a stable linear latent evolution operator and a consistent latent observation mapping, yielding a closed linear state-space model in the latent coordinates. This construction restores compatibility with the Kalman filtering framework and allows both forecast and analysis steps to be carried out entirely in the latent space. Compared with existing autoencoder-based and latent assimilation approaches that rely on unconstrained nonlinear latent dynamics, the proposed formulation emphasizes structural consistency, stability, and interpretability. We provide a theoretical analysis of learning linear dynamics on low-dimensional manifolds and establish generalization error bounds for the proposed latent model. Numerical experiments on representative nonlinear and chaotic systems demonstrate that the LAE-EnKF yields more accurate and stable assimilation than the standard EnKF and related latent-space methods under partial, sparse, and nonlinear observations.
Förderung
- Jiangsu Provincial Scientific Research Center of Applied Mathematics
- National Natural Science Foundation of China
- Ministry of Education, Singapore
Zitieren
APA
Tong, X. T., Wang, Y., & Yan, L. (2026). Latent Autoencoder Ensemble Kalman Filter for Nonlinear Data Assimilation. SIAM Journal on Scientific Computing, 48(6), C1403-C1436. https://doi.org/10.1137/26m1850230
Vancouver
Tong XT, Wang Y, Yan L. Latent Autoencoder Ensemble Kalman Filter for Nonlinear Data Assimilation. SIAM Journal on Scientific Computing. 2026;48(6):C1403-C1436. doi:10.1137/26m1850230
BibTeX
@article{Tong2026,
author = {Tong, Xin T. and Wang, Yanyan and Yan, Liang},
title = {Latent Autoencoder Ensemble Kalman Filter for Nonlinear Data Assimilation},
journal = {SIAM Journal on Scientific Computing},
year = {2026},
volume = {48},
number = {6},
pages = {C1403--C1436},
doi = {10.1137/26m1850230}
}
EUVIMED