Richard Menard, Environment and Climate Change Canada

Chemical data assimilation using tuning of error covariances

Sergey Skachko, Richard Menard, Quentin Errera, Yves Christophe and Simon Chabrillat


The study discusses the BASCOE chemistry transport model (CTM) coupled to ensemble Kalman filter (EnKF) and 4D-Var data assimilation systems. Both systems assimilate multiple species retrieved by Aura MLS satellite into a CTM representing a full set of stratospheric chemistry. The error covariances of both data assimilation systems are estimated using the Desroziers method based on innovation statistics. The 4D-Var uses the estimation of background error and observation error covariance estimation, whereas EnKF uses only the latter. The background error covariance matrix of EnKF evolves in time following the ensemble evolution. We discuss the fundamental differences between two data assimilation algorithms and how each systems benefits from the covariance error estimation. The performance of each system is assessed using a case study with and without a strong model bias. The results of this comparison are shown using the Observations-minus-Forecasts (OmF) statistics. And the advantages and disadvantages of each system are discussed

File available here