Including correlated observation error statistics in upper atmospheric ensemble data assimilative modelling

Adil Siripatana, Sean Elvidge

Research output: Contribution to conference (unpublished)Posterpeer-review

Abstract

Utilising data assimilation (DA) techniques to improve the background estimates of a physics-based model has become increasingly common among the upper atmospheric modelling community. However, pragmatic implementations of the DA methods such as the Ensemble Kalman Filter (EnKF) and its variants often assume independent observations mainly for the ease of computation. As a result, the observation error covariance matrix R is diagonal, which results in filter performance being sub-optimal when the potentially correlated observation error statistics are neglected. In this work, we implement iterative estimation of the observation covariance matrix using the statistical averages of background and analysis innovations within the local ensemble Transform Kalman Filter (LETKF) algorithm of the Advanced Ensemble electron density (Ne) Assimilation System (AENeAS) to enhance its ionospheric state estimation and forecast capabilities. The total electron content (TEC) estimates of the developed model are statistically benchmarked against its baseline diagonal R counterpart, GPS TEC data, TIE-GCM, and NeQuick models. The results from the analysis show that taking into account the impact of correlated observation errors helps improve the estimation of the TEC and overall ionosphere/thermosphere structure.
Original languageEnglish
Publication statusPublished - 29 Oct 2021
Event17th European Space Weather Week - Technology Innovation Centre, Glasgow, United Kingdom
Duration: 25 Oct 202129 Oct 2021
http://esww17.iopconfs.org/home

Conference

Conference17th European Space Weather Week
Abbreviated titleESWW17
Country/TerritoryUnited Kingdom
CityGlasgow
Period25/10/2129/10/21
Internet address

Fingerprint

Dive into the research topics of 'Including correlated observation error statistics in upper atmospheric ensemble data assimilative modelling'. Together they form a unique fingerprint.

Cite this