Full-window branch discovery and loss-selected EnKF continuation for data assimilation
Abstract
We develop a framework for offline full-window branch discovery, optionally followed by online continuation with an ensemble Kalman filter (EnKF). Three mechanisms drive the branch search: adjoint path-kernel (APK) differentiation balances kernel differentiation and correction-stabilized path perturbation, shifting the optimization from exploration to exploitation; an optimized Gaussian initial law broadens the search over initial-state basins; and loss-weighted mixing across independent runs recombines successful path components. We may then select an interior state using a local loss and continue online with an EnKF. In 40-dimensional Lorenz-96 experiments, the mean offline path RMSE of APK is 4.3 times smaller than that of population weak-4D−Varx. The resulting APK-EnKF method has a mean online RMSE 64 times smaller than that of ordinary EnKF.
Related
- Quantitative Unique Continuation on Simplex and ℤn
- Strong Unique Continuation for Fractional Schrödinger Operators
- Unique continuation at infinity for potentials with arbitrary radial growth
- Nonlinear Assimilation via Score-Based Sequential Langevin Sampling
- Continuation and threshold global strong solutions for the two-dimensional Landau--Lifshitz--Gilbert equation
BibTeX
@misc{ni2026fullwindow,
title = {{Full-window branch discovery and loss-selected EnKF continuation for data assimilation}},
author = {Angxiu Ni},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.10949},
eprint = {2609.10949},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.10949},
}
RIS
TY - UNPB
TI - Full-window branch discovery and loss-selected EnKF continuation for data assimilation
AU - Angxiu Ni
PY - 2026
DA - 2026/09/10
PB - arXiv
DO - 10.48550/arxiv.2609.10949
UR - https://arxiv.org/abs/2609.10949
AB - We develop a framework for offline full-window branch discovery, optionally followed by online continuation with an ensemble Kalman filter (EnKF). Three mechanisms drive the branch search: adjoint path-kernel (APK) differentiation balances kernel differentiation and correction-stabilized path perturbation, shifting the optimization from exploration to exploitation; an optimized Gaussian initial law broadens the search over initial-state basins; and loss-weighted mixing across independent runs recombines successful path components. We may then select an interior state using a local loss and continue online with an EnKF. In 40-dimensional Lorenz-96 experiments, the mean offline path RMSE of APK is 4.3 times smaller than that of population weak-4D−Var_x. The resulting APK-EnKF method has a mean online RMSE 64 times smaller than that of ordinary EnKF.
ER -