Elucidating the performance of data assimilation neural networks for chaotic dynamics
Abstract
Recent work has shown that the analysis operator in sequential data assimilation designed to track chaotic dynamics, can be learned with deep learning from the sole knowledge of a true state trajectory and observations thereof. This approach to learning the analysis is computationally more challenging, yet conceptually more fundamental than approaches that learn a direct mapping from forecasts and observations to the corresponding analysis increments. Such learned scheme has been demonstrated to achieve accuracy comparable to that of the ensemble Kalman filter when applied to low-order dynamics. Strikingly, the same accuracy can be reached with a single state forecast instead of an ensemble, hence bypassing the need to explicitly represent forecast uncertainty. In this study, we extend the investigation of such learned analysis operators beyond the preliminary experiments reported so far. First, we analyse the emergence of local patterns encoded in the operator, which accounts for the remarkable scalability of the approach to high-dimensional state spaces. Second, we assess the performance of the learned operators in stronger nonlinear regimes of the chaotic dynamics. We show that they can match the efficiency of the iterative ensemble Kalman filter, the baseline in this context, while avoiding the need for nonlinear iterative optimisation. Throughout the paper, we seek underlying reasons for the efficiency of the approach, drawing on insights from both machine learning and nonlinear data assimilation.
Related
- Ensemble Score Filter with Image Inpainting for Data Assimilation in Tracking Surface Quasi-geostrophic Dynamics with Partial Observations
- Accounting for model error in ensemble variational data assimilation through learned generative dynamics
- Measurement report: Global Total Ozone Records – part 1: ground-based monitoring networks performance assessment and status review
- On the Nudging Approach to Continuous Data Assimilation in the Limit of Infinite Error Feedback Gain
- LFS-SEnDA: a global ocean forecast system with an efficient Sliding-Ensemble-based Data Assimilation scheme
BibTeX
@article{bocquet2026elucidating,
title = {{Elucidating the performance of data assimilation neural networks for chaotic dynamics}},
author = {Marc Bocquet and Tobias Sebastian Finn and Sibo Cheng and Alban Farchi},
journal = {Nonlinear Processes in Geophysics},
year = {2026},
month = aug,
doi = {10.5194/npg-33-401-2026},
url = {https://hal.science/hal-05835061v1},
note = {CC BY},
}
RIS
TY - JOUR
TI - Elucidating the performance of data assimilation neural networks for chaotic dynamics
AU - Marc Bocquet
AU - Tobias Sebastian Finn
AU - Sibo Cheng
AU - Alban Farchi
PY - 2026
DA - 2026/08/11
JO - Nonlinear Processes in Geophysics
DO - 10.5194/npg-33-401-2026
UR - https://hal.science/hal-05835061v1
AB - Recent work has shown that the analysis operator in sequential data assimilation designed to track chaotic dynamics, can be learned with deep learning from the sole knowledge of a true state trajectory and observations thereof. This approach to learning the analysis is computationally more challenging, yet conceptually more fundamental than approaches that learn a direct mapping from forecasts and observations to the corresponding analysis increments. Such learned scheme has been demonstrated to achieve accuracy comparable to that of the ensemble Kalman filter when applied to low-order dynamics. Strikingly, the same accuracy can be reached with a single state forecast instead of an ensemble, hence bypassing the need to explicitly represent forecast uncertainty. In this study, we extend the investigation of such learned analysis operators beyond the preliminary experiments reported so far. First, we analyse the emergence of local patterns encoded in the operator, which accounts for the remarkable scalability of the approach to high-dimensional state spaces. Second, we assess the performance of the learned operators in stronger nonlinear regimes of the chaotic dynamics. We show that they can match the efficiency of the iterative ensemble Kalman filter, the baseline in this context, while avoiding the need for nonlinear iterative optimisation. Throughout the paper, we seek underlying reasons for the efficiency of the approach, drawing on insights from both machine learning and nonlinear data assimilation.
SN - 1023-5809
SN - 1607-7946
ER -