SEIK - the unknown ensemble Kalman filter
Lars Nerger, Tijana Janjic Pfander, Wolfgang Hiller, Jens Schröter
Abstract
Open-access reader
Lars Nerger, Tijana Janjic Pfander, Wolfgang Hiller, Jens Schröter
Abstract
Open-access reader
The SEIK filter (Singular "Evolutive" Interpolated Kalman filter) hasbeen introduced in 1998 by D.T. Pham as a variant of the SEEK filter,which is a reduced-rank approximation of the Extended KalmanFilter. In recent years, it has been shown that the SEIK filter isan ensemble-based Kalman filter that uses a factorization rather thansquare-root of the state error covariance matrix. Unfortunately, theexistence of the SEIK filter as an ensemble-based Kalman filter withsimilar efficiency as the later introduced ensemble square-root Kalmanfilters, appears to be widely unknown and the SEIK filter is omittedin reviews about ensemble-based Kalman filters. To raise the attentionabout the SEIK filter as a very efficient ensemble-based Kalmanfilter, we review the filter algorithm and compare it with ensemblesquare-root Kalman filter algorithms. For a practical comparison theSEIK filter and the Ensemble Transformation Kalman filter (ETKF) areapplied in twin experiments assimilating sea level anomaly data intothe finite-element ocean model FEOM.
A significance statement is not available in the OpenAlex record.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
The SEIK filter (Singular "Evolutive" Interpolated Kalman filter) hasbeen introduced in 1998 by D.T. Pham as a variant of the SEEK filter,which is a reduced-rank approximation of the Extended KalmanFilter. In recent years, it has been shown that the SEIK filter isan ensemble-based Kalman filter that uses a factorization rather thansquare-root of the state error covariance matrix. Unfortunately, theexistence of the SEIK filter as an ensemble-based Kalman filter withsimilar efficiency as the later introduced ensemble square-root Kalmanfilters, appears to be widely unknown and the SEIK filter is omittedin reviews about ensemble-based Kalman filters. To raise the attentionabout the SEIK filter as a very efficient ensemble-based Kalmanfilter, we review the filter algorithm and compare it with ensemblesquare-root Kalman filter algorithms. For a practical comparison theSEIK filter and the Ensemble Transformation Kalman filter (ETKF) areapplied in twin experiments assimilating sea level anomaly data intothe finite-element ocean model FEOM.
Key concepts: Ensemble Kalman filter, Alpha beta filter, Invariant extended Kalman filter, Fast Kalman filter, Extended Kalman filter, Kalman filter, Mathematics, Filter (signal processing)