Application of Ensemble Kalman Filter With Covariance Localization in Data Assimilation of Radiation Belt Electrons
Abstract Data assimilation aims to enhance the system state estimate by merging sparse and diverse measurements with physical models, while considering their individual uncertainties. As a highly promising data assimilation technique, Ensemble Kalman Filter (EnKF) is well‐suited for addressing nonli...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2025-06-01
|
| Series: | Space Weather |
| Subjects: | |
| Online Access: | https://doi.org/10.1029/2025SW004387 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Data assimilation aims to enhance the system state estimate by merging sparse and diverse measurements with physical models, while considering their individual uncertainties. As a highly promising data assimilation technique, Ensemble Kalman Filter (EnKF) is well‐suited for addressing nonlinear, complex, and high‐dimensional problems. As a result, EnKF has emerged as an increasingly important tool for the reanalysis and prediction of highly dynamic variations of energetic electrons in Earth's radiation belt. However, EnKF often encounters the challenge of spurious correlations between observations and state variables due to insufficient ensemble size. In this study, we introduce a covariance localization method to address this issue in the radiation belt data assimilation. We validate the covariance localization method through a twin experiment, which demonstrates its effectiveness in truncating spurious correlations between variables along three‐dimensional grids (radial distance, pitch angle, and energy). Furthermore, EnKF with covariance localization enables usage of fewer ensemble members to achieve reanalysis results that closely approximate the “truth,” thereby significantly reducing computational costs. Finally, using the localized EnKF with 15 ensemble members, in combination with measurements from Van Allen Probes and GOES satellites, we reconstruct the spatial and temporal evolution of radiation belt electrons in March 2013 as an illustrative purpose. Our study demonstrates that the application of covariance localization method can effectively improve the performance of ensemble data assimilation of radiation belt electrons, particularly for finite ensemble sizes. This sheds important light on future efforts of nowcast and forecast of Earth's radiation belt electron dynamical variability. |
|---|---|
| ISSN: | 1542-7390 |