Adaptive error covariances estimation methods for ensemble Kalman filters

Yicun Zhen, John Harlim

Research output: Contribution to journalArticle

7 Citations (Scopus)

Abstract

This paper presents a computationally fast algorithm for estimating, both, the system and observation noise covariances of nonlinear dynamics, that can be used in an ensemble Kalman filtering framework. The new method is a modification of Belanger's recursive method, to avoid an expensive computational cost in inverting error covariance matrices of product of innovation processes of different lags when the number of observations becomes large. When we use only product of innovation processes up to one-lag, the computational cost is indeed comparable to a recently proposed method by Berry-Sauer's. However, our method is more flexible since it allows for using information from product of innovation processes of more than one-lag.Extensive numerical comparisons between the proposed method and both the original Belanger's and Berry-Sauer's schemes are shown in various examples, ranging from low-dimensional linear and nonlinear systems of SDEs and 40-dimensional stochastically forced Lorenz-96 model. Our numerical results suggest that the proposed scheme is as accurate as the original Belanger's scheme on low-dimensional problems and has a wider range of more accurate estimates compared to Berry-Sauer's method on L-96 example.

Original languageEnglish (US)
Pages (from-to)619-638
Number of pages20
JournalJournal of Computational Physics
Volume294
DOIs
StatePublished - Aug 1 2015

Fingerprint

Kalman filters
time lag
Innovation
products
costs
linear systems
Covariance matrix
nonlinear systems
Linear systems
Costs
Nonlinear systems
estimating
estimates

All Science Journal Classification (ASJC) codes

  • Numerical Analysis
  • Modeling and Simulation
  • Physics and Astronomy (miscellaneous)
  • Physics and Astronomy(all)
  • Computer Science Applications
  • Computational Mathematics
  • Applied Mathematics

Cite this

@article{e62e7c89654049a0b2c1dd7528861a5c,
title = "Adaptive error covariances estimation methods for ensemble Kalman filters",
abstract = "This paper presents a computationally fast algorithm for estimating, both, the system and observation noise covariances of nonlinear dynamics, that can be used in an ensemble Kalman filtering framework. The new method is a modification of Belanger's recursive method, to avoid an expensive computational cost in inverting error covariance matrices of product of innovation processes of different lags when the number of observations becomes large. When we use only product of innovation processes up to one-lag, the computational cost is indeed comparable to a recently proposed method by Berry-Sauer's. However, our method is more flexible since it allows for using information from product of innovation processes of more than one-lag.Extensive numerical comparisons between the proposed method and both the original Belanger's and Berry-Sauer's schemes are shown in various examples, ranging from low-dimensional linear and nonlinear systems of SDEs and 40-dimensional stochastically forced Lorenz-96 model. Our numerical results suggest that the proposed scheme is as accurate as the original Belanger's scheme on low-dimensional problems and has a wider range of more accurate estimates compared to Berry-Sauer's method on L-96 example.",
author = "Yicun Zhen and John Harlim",
year = "2015",
month = "8",
day = "1",
doi = "10.1016/j.jcp.2015.03.061",
language = "English (US)",
volume = "294",
pages = "619--638",
journal = "Journal of Computational Physics",
issn = "0021-9991",
publisher = "Academic Press Inc.",

}

Adaptive error covariances estimation methods for ensemble Kalman filters. / Zhen, Yicun; Harlim, John.

In: Journal of Computational Physics, Vol. 294, 01.08.2015, p. 619-638.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Adaptive error covariances estimation methods for ensemble Kalman filters

AU - Zhen, Yicun

AU - Harlim, John

PY - 2015/8/1

Y1 - 2015/8/1

N2 - This paper presents a computationally fast algorithm for estimating, both, the system and observation noise covariances of nonlinear dynamics, that can be used in an ensemble Kalman filtering framework. The new method is a modification of Belanger's recursive method, to avoid an expensive computational cost in inverting error covariance matrices of product of innovation processes of different lags when the number of observations becomes large. When we use only product of innovation processes up to one-lag, the computational cost is indeed comparable to a recently proposed method by Berry-Sauer's. However, our method is more flexible since it allows for using information from product of innovation processes of more than one-lag.Extensive numerical comparisons between the proposed method and both the original Belanger's and Berry-Sauer's schemes are shown in various examples, ranging from low-dimensional linear and nonlinear systems of SDEs and 40-dimensional stochastically forced Lorenz-96 model. Our numerical results suggest that the proposed scheme is as accurate as the original Belanger's scheme on low-dimensional problems and has a wider range of more accurate estimates compared to Berry-Sauer's method on L-96 example.

AB - This paper presents a computationally fast algorithm for estimating, both, the system and observation noise covariances of nonlinear dynamics, that can be used in an ensemble Kalman filtering framework. The new method is a modification of Belanger's recursive method, to avoid an expensive computational cost in inverting error covariance matrices of product of innovation processes of different lags when the number of observations becomes large. When we use only product of innovation processes up to one-lag, the computational cost is indeed comparable to a recently proposed method by Berry-Sauer's. However, our method is more flexible since it allows for using information from product of innovation processes of more than one-lag.Extensive numerical comparisons between the proposed method and both the original Belanger's and Berry-Sauer's schemes are shown in various examples, ranging from low-dimensional linear and nonlinear systems of SDEs and 40-dimensional stochastically forced Lorenz-96 model. Our numerical results suggest that the proposed scheme is as accurate as the original Belanger's scheme on low-dimensional problems and has a wider range of more accurate estimates compared to Berry-Sauer's method on L-96 example.

UR - http://www.scopus.com/inward/record.url?scp=84927766836&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84927766836&partnerID=8YFLogxK

U2 - 10.1016/j.jcp.2015.03.061

DO - 10.1016/j.jcp.2015.03.061

M3 - Article

AN - SCOPUS:84927766836

VL - 294

SP - 619

EP - 638

JO - Journal of Computational Physics

JF - Journal of Computational Physics

SN - 0021-9991

ER -