JZUS - Journal of Zhejiang University SCIENCE

Frontiers of Information Technology & Electronic Engineering

Accepted manuscript available online (unedited version)

DRL-EnVar: an adaptive hybrid ensemble-variational data assimilation method based on deep reinforcement learning

Author(s): Lilan HUANG^1, ², Hongze LENG^1, ², Junqiang SONG^1, ², Dongzi WANG¹, Wuxin WANG¹, Ruisheng HU², Hang CAO²
Affiliation(s): ¹College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, China ²College of Meteorology and Oceanography, National University of Defense Technology, Changsha 410073, China
Corresponding email(s): huanglilan18@nudt.edu.cn, hzleng@nudt.edu.cn, junqiang@nudt.edu.cn
Key Words: Adaptive data assimilation; Hybrid ensemble-variational method; Background error covariance; Deep reinforcement learning

Share this article to： More <<< Previous Paper \|Next Paper >>>

Lilan HUANG^1,², Hongze LENG^1,², Junqiang SONG^1,², Dongzi WANG¹,Wuxin WANG¹, Ruisheng HU², Hang CAO². DRL-EnVar: an adaptive hybrid ensemble-variational data assimilation method based on deep reinforcement learning[J]. Frontiers of Information Technology & Electronic Engineering,in press.https://doi.org/10.1631/FITEE.2401063

@article{title="DRL-EnVar: an adaptive hybrid ensemble-variational data assimilation method based on deep reinforcement learning",
author="Lilan HUANG^1,², Hongze LENG^1,², Junqiang SONG^1,², Dongzi WANG¹,Wuxin WANG¹, Ruisheng HU², Hang CAO²",
journal="Frontiers of Information Technology & Electronic Engineering",
year="in press",
publisher="Zhejiang University Press & Springer",
doi="https://doi.org/10.1631/FITEE.2401063"
}

%0 Journal Article
%T DRL-EnVar: an adaptive hybrid ensemble-variational data assimilation method based on deep reinforcement learning
%A Lilan HUANG^1
%A²
%A Hongze LENG^1
%A²
%A Junqiang SONG^1
%A²
%A Dongzi WANG¹
%A Wuxin WANG¹
%A Ruisheng HU²
%A Hang CAO²
%J Frontiers of Information Technology & Electronic Engineering
%P
%@ 2095-9184
%D in press
%I Zhejiang University Press & Springer
doi="https://doi.org/10.1631/FITEE.2401063"

TY - JOUR
T1 - DRL-EnVar: an adaptive hybrid ensemble-variational data assimilation method based on deep reinforcement learning
A1 - Lilan HUANG^{1
A1 -}²
A1 - Hongze LENG^{1
A1 -}²
A1 - Junqiang SONG^{1
A1 -}²
A1 - Dongzi WANG¹
A1 - Wuxin WANG¹
A1 - Ruisheng HU²
A1 - Hang CAO²
J0 - Frontiers of Information Technology & Electronic Engineering
SP -
EP -
%@ 2095-9184
Y1 - in press
PB - Zhejiang University Press & Springer
ER -
doi="https://doi.org/10.1631/FITEE.2401063"

Abstract
Chinese Summary
Academic Network
Reviewer Comment

Abstract: Accurate estimation of the background error covariance matrix (B) remains a critical challenge in numerical weather prediction (NWP), directly influencing data assimilation (DA) performance and forecast accuracy.Although hybrid ensemble-variational (EnVar) methods combine static and flow-dependent B matrices to improve assimilation, their effectiveness is constrained by empirically fixed weights. To address this limitation, we propose DRL-EnVar, an adaptive hybrid EnVar DA method enhanced with deep reinforcement learning. DRL-EnVar integrates deep learning (DL) components, including a novel cyclic convolution module to extract abstract features from data, and employs reinforcement learning (RL) to dynamically optimize hybrid weighting strategies. The system adaptively combines multiple ensemble-based flow-dependent B matrices with one or more static B matrices to construct a time-varying hybrid B matrix that better reflects real-time background errors. Experimental results demonstrate that DRL-EnVar performs better than traditional ensemble Kalman filter (EnKF) and hybrid covariance DA (HCDA) methods, especially under sparse observations or transitional changes in state variables. It achieves competitive or superior assimilation accuracy with lower computational cost and can be flexibly integrated into both three-dimensional variational assimilation (3DVar) and four-dimensional variational assimilation (4DVar) frameworks. Overall, DRL-EnVar offers a novel and efficient approach to adaptive DA, particularly valuable for improving forecast skill during transitional weather regimes.

Darkslateblue:Affiliate; Royal Blue:Author; Turquoise:Article

Reference

Open peer comments: Debate/Discuss/Question/Opinion

<1>

- Go to

Darkslateblue:Affiliate; Royal Blue:Author; Turquoise:Article

Reference