Title: Variational ensemble data assimilation at Meteo-France for error covariance modelling and ensemble prediction

Authors: Loik Berre (Meteo-France)
Gerald Desroziers (Meteo-France)
Laure Raynaud (Meteo-France)
Hubert Varella (Meteo-France)
Laurent Descamps (Meteo-France)
Carole Labadie (Meteo-France)

Since July 2008, a variational ensemble data assimilation system (EnVar) has been operational at Meteo-France. It is used to calculate flow-dependent background error covariances for optimizing data assimilation. It also provides perturbed initial states for the Meteo-France ensemble prediction system. Prominent features and developments of this EnVar will be presented.

The ensemble approach has been chosen to be consistent with an error simulation technique of the deterministic 4D-Var data assimilation cycle. Spectral filtering techniques based on objective signal and noise estimates are applied for modelling background error standard-deviations, and wavelet filtering techniques are considered for correlations. Moreover, innovation-based variance estimates are compared to ensemble-based variances to get information about model errors. Acceleration and parallelization opportunities will eventually be discussed.

GMAO Head: Michele Rienecker Global Modeling and Assimilation Office NASA Goddard Space Flight Center

Curator: Nikki Privé Last Updated: March 1 2011