Yu et al. (2025) A nudging-based data assimilation method coupled with bidirectional gated neural networks for error correction

Identification

• Journal: Environmental Modelling & Software
• Year: 2025
• Date: 2025-09-08
• Authors: Qinghe Yu, Yulong Bai, Manhong Fan, Chunlin Huang, Xinan Yue, Kun Yang
• DOI: 10.1016/j.envsoft.2025.106670

Research Groups

• College of Physics and Electrical Engineering, Northwest Normal University, Lanzhou, Gansu, China
• Northwest Institute of Eco-Environment and Resources, Chinese Academy of Sciences, Lanzhou, Gansu, China

Short Summary

This study develops a novel nudging-based data assimilation method that integrates Bidirectional Gated Recurrent Units (BiGRU) with the Ensemble Kalman Filter (EnKF) to enhance accuracy and stability in error correction. Numerical experiments using the Lorenz-96 model demonstrate its improved resilience to noise interference and greater robustness with sparse observations.

Objective

• To develop and evaluate a novel nudging-based data assimilation method that integrates Bidirectional Gated Recurrent Units (BiGRU) with the Ensemble Kalman Filter (EnKF) for enhanced accuracy and stability in nonlinear error prediction and correction.

Study Configuration

• Spatial Scale: Conceptual (Lorenz-96 nonlinear system, a low-dimensional chaotic model).
• Temporal Scale: Conceptual (Lorenz-96 nonlinear system, with numerical experiments conducted over discrete time steps not explicitly specified).

Methodology and Data

• Models used: Bidirectional Gated Recurrent Units (BiGRU), Ensemble Kalman Filter (EnKF), Lorenz-96 model (as the system under study).
• Data sources: Synthetic data generated from the Lorenz-96 model, with simulated noise and sparse observations.

Main Results

• The novel coupled BiGRU-EnKF approach demonstrates enhanced resilience to noise interference in data assimilation.
• The method exhibits greater robustness in generating assimilation outcomes from sparse observations.

Contributions

• Introduction of a novel nudging-based data assimilation method that integrates Bidirectional Gated Recurrent Units (BiGRU) with the Ensemble Kalman Filter (EnKF).
• Leveraging machine learning (BiGRU) for improved nonlinear error prediction and correction within a data assimilation framework.
• Demonstration of enhanced resilience to noise and greater robustness with sparse observations in data assimilation compared to traditional methods.

Funding

• Not specified in the provided text.

Citation

@article{Yu2025nudgingbased,
  author = {Yu, Qinghe and Bai, Yulong and Fan, Manhong and Huang, Chunlin and Yue, Xinan and Yang, Kun},
  title = {A nudging-based data assimilation method coupled with bidirectional gated neural networks for error correction},
  journal = {Environmental Modelling & Software},
  year = {2025},
  doi = {10.1016/j.envsoft.2025.106670},
  url = {https://doi.org/10.1016/j.envsoft.2025.106670}
}