Preprints
https://doi.org/10.5194/npg-2021-35
https://doi.org/10.5194/npg-2021-35

  12 Nov 2021

12 Nov 2021

Review status: this preprint is currently under review for the journal NPG.

Brief communication: An innovation-based estimation method for model error covariance in Kalman filters

Eviatar Bach1 and Michael Ghil1,2 Eviatar Bach and Michael Ghil
  • 1Geosciences Department and Laboratoire de Météorologie Dynamique (CNRS and IPSL), École Normale Supérieure and PSL University, Paris, France
  • 2Department of Atmospheric and Oceanic Science, University of California at Los Angeles, Los Angeles, United States

Abstract. We present a simple innovation-based model error covariance estimation method for Kalman filters. The method is based on Berry and Sauer (2013) and the simplification results from assuming known observation error covariance. We carry out experiments with a prescribed model error covariance using a Lorenz (1996) model and ensemble Kalman filter. The prescribed error covariance matrix is recovered with high accuracy.

Eviatar Bach and Michael Ghil

Status: open (until 07 Jan 2022)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse

Eviatar Bach and Michael Ghil

Eviatar Bach and Michael Ghil

Viewed

Total article views: 187 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
159 27 1 187 1 1
  • HTML: 159
  • PDF: 27
  • XML: 1
  • Total: 187
  • BibTeX: 1
  • EndNote: 1
Views and downloads (calculated since 12 Nov 2021)
Cumulative views and downloads (calculated since 12 Nov 2021)

Viewed (geographical distribution)

Total article views: 187 (including HTML, PDF, and XML) Thereof 187 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
1
 
 
 
 
Latest update: 01 Dec 2021
Download
Short summary
Data assimilation (DA) is the process of combining model forecasts with observations in order to provide an optimal estimate of the system state. When models are imperfect, the uncertainty in the forecasts may be underestimated, requiring inflation of the corresponding error covariance. Here, we present a simple method for estimating the magnitude and structure of the model error covariance matrix. We demonstrate the efficacy of this method with idealized experiments.