Enhancing Robustness of Variational Data Assimilation in Chaotic Systems: An α-4DVar Framework with Rényi Entropy and α-Generalized Gaussian Distributions

Traditional 4-dimensional variational data assimilation methods have limitations due to the Gaussian distribution assumption of observation errors, and the gradient of the objective functional is vulnerable to observation noise and outliers. To address these issues, this paper proposes a non-Gaussia...

Full description

Saved in:
Bibliographic Details
Main Authors: Yuchen Luo, Xiaoqun Cao, Kecheng Peng, Mengge Zhou, Yanan Guo
Format: Article
Language:English
Published: MDPI AG 2025-07-01
Series:Entropy
Subjects:
Online Access:https://www.mdpi.com/1099-4300/27/7/763
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849303520730152960
author Yuchen Luo
Xiaoqun Cao
Kecheng Peng
Mengge Zhou
Yanan Guo
author_facet Yuchen Luo
Xiaoqun Cao
Kecheng Peng
Mengge Zhou
Yanan Guo
author_sort Yuchen Luo
collection DOAJ
description Traditional 4-dimensional variational data assimilation methods have limitations due to the Gaussian distribution assumption of observation errors, and the gradient of the objective functional is vulnerable to observation noise and outliers. To address these issues, this paper proposes a non-Gaussian nonlinear data assimilation method called α-4DVar, based on Rényi entropy and the α-generalized Gaussian distribution. By incorporating the heavy-tailed property of Rényi entropy, the objective function and its gradient suitable for non-Gaussian errors are derived, and numerical experiments are conducted using the Lorenz-63 model. Experiments are conducted with Gaussian and non-Gaussian errors as well as different initial guesses to compare the assimilation effects of traditional 4DVar and α-4DVar. The results show that α-4DVar performs as well as traditional method without observational errors. Its analysis field is closer to the truth, with RMSE rapidly dropping to a low level and remaining stable, particularly under non-Gaussian errors. Under different initial guesses, the RMSE of both the background and analysis fields decreases quickly and stabilizes. In conclusion, the α-4DVar method demonstrates significant advantages in handling non-Gaussian observational errors, robustness against noise, and adaptability to various observational conditions, thus offering a more reliable and effective solution for data assimilation.
format Article
id doaj-art-9cc2ad5ad3014f52b58bbbec906d5507
institution Kabale University
issn 1099-4300
language English
publishDate 2025-07-01
publisher MDPI AG
record_format Article
series Entropy
spelling doaj-art-9cc2ad5ad3014f52b58bbbec906d55072025-08-20T03:58:26ZengMDPI AGEntropy1099-43002025-07-0127776310.3390/e27070763Enhancing Robustness of Variational Data Assimilation in Chaotic Systems: An α-4DVar Framework with Rényi Entropy and α-Generalized Gaussian DistributionsYuchen Luo0Xiaoqun Cao1Kecheng Peng2Mengge Zhou3Yanan Guo4College of Meteorology and Oceanology, National University of Defense Technology, Changsha 410073, ChinaCollege of Meteorology and Oceanology, National University of Defense Technology, Changsha 410073, ChinaCollege of Computer Science, National University of Defense Technology, Changsha 410073, ChinaCollege of Meteorology and Oceanology, National University of Defense Technology, Changsha 410073, ChinaCollege of Computer Science, National University of Defense Technology, Changsha 410073, ChinaTraditional 4-dimensional variational data assimilation methods have limitations due to the Gaussian distribution assumption of observation errors, and the gradient of the objective functional is vulnerable to observation noise and outliers. To address these issues, this paper proposes a non-Gaussian nonlinear data assimilation method called α-4DVar, based on Rényi entropy and the α-generalized Gaussian distribution. By incorporating the heavy-tailed property of Rényi entropy, the objective function and its gradient suitable for non-Gaussian errors are derived, and numerical experiments are conducted using the Lorenz-63 model. Experiments are conducted with Gaussian and non-Gaussian errors as well as different initial guesses to compare the assimilation effects of traditional 4DVar and α-4DVar. The results show that α-4DVar performs as well as traditional method without observational errors. Its analysis field is closer to the truth, with RMSE rapidly dropping to a low level and remaining stable, particularly under non-Gaussian errors. Under different initial guesses, the RMSE of both the background and analysis fields decreases quickly and stabilizes. In conclusion, the α-4DVar method demonstrates significant advantages in handling non-Gaussian observational errors, robustness against noise, and adaptability to various observational conditions, thus offering a more reliable and effective solution for data assimilation.https://www.mdpi.com/1099-4300/27/7/763data assimilationnon-gaussian distributionRényi entropyLorenz-63 modelrobustness
spellingShingle Yuchen Luo
Xiaoqun Cao
Kecheng Peng
Mengge Zhou
Yanan Guo
Enhancing Robustness of Variational Data Assimilation in Chaotic Systems: An α-4DVar Framework with Rényi Entropy and α-Generalized Gaussian Distributions
Entropy
data assimilation
non-gaussian distribution
Rényi entropy
Lorenz-63 model
robustness
title Enhancing Robustness of Variational Data Assimilation in Chaotic Systems: An α-4DVar Framework with Rényi Entropy and α-Generalized Gaussian Distributions
title_full Enhancing Robustness of Variational Data Assimilation in Chaotic Systems: An α-4DVar Framework with Rényi Entropy and α-Generalized Gaussian Distributions
title_fullStr Enhancing Robustness of Variational Data Assimilation in Chaotic Systems: An α-4DVar Framework with Rényi Entropy and α-Generalized Gaussian Distributions
title_full_unstemmed Enhancing Robustness of Variational Data Assimilation in Chaotic Systems: An α-4DVar Framework with Rényi Entropy and α-Generalized Gaussian Distributions
title_short Enhancing Robustness of Variational Data Assimilation in Chaotic Systems: An α-4DVar Framework with Rényi Entropy and α-Generalized Gaussian Distributions
title_sort enhancing robustness of variational data assimilation in chaotic systems an α 4dvar framework with renyi entropy and α generalized gaussian distributions
topic data assimilation
non-gaussian distribution
Rényi entropy
Lorenz-63 model
robustness
url https://www.mdpi.com/1099-4300/27/7/763
work_keys_str_mv AT yuchenluo enhancingrobustnessofvariationaldataassimilationinchaoticsystemsana4dvarframeworkwithrenyientropyandageneralizedgaussiandistributions
AT xiaoquncao enhancingrobustnessofvariationaldataassimilationinchaoticsystemsana4dvarframeworkwithrenyientropyandageneralizedgaussiandistributions
AT kechengpeng enhancingrobustnessofvariationaldataassimilationinchaoticsystemsana4dvarframeworkwithrenyientropyandageneralizedgaussiandistributions
AT menggezhou enhancingrobustnessofvariationaldataassimilationinchaoticsystemsana4dvarframeworkwithrenyientropyandageneralizedgaussiandistributions
AT yananguo enhancingrobustnessofvariationaldataassimilationinchaoticsystemsana4dvarframeworkwithrenyientropyandageneralizedgaussiandistributions