Title:How to optimally treat large scale information in limited area ensemble-based data assimilation?

Authors: Jean-François Caron (Met Office, Reading, UK)

The Met Office is developing a convective scale version of the Met Office ensemble prediction system, MOGREPS, where the analysis perturbations are produced using an Ensemble Transform Kalman Filter (ETKF). The primary goals of this work are 1) to conduct examinations of 1-hour forecast error covariances for the benefit of a NWP-based nowcasting system currently under development at the Met Office and 2) to perform predictability studies of localized weather at very short time scale. This convective scale ETKF system uses a model with a grid resolution of 1.5 km covering southern United Kingdom and has an hourly cycle. The lateral boundaries are provided by the 18 km and 12-hourly cycling regional component of MOGREPS which cover the North Atlantic and Europe.

In this poster we will present the issues we faced due to the discontinuities introduced by the current ETKF method between the analysis perturbations and the perturbations at the lateral boundaries. To reduce theses discontinuities, we developed and tested a so-called scale-selective ETKF approach where the ETKF transform matrix is applied only to the small scale component of the 1.5km forecast perturbations while the large scale component of the analysis perturbations is taken directly from parent EPS. The changes implied by this new approach in terms of both ensemble forecast performance and forecast error covariances will be presented. Finally, we will discuss on the implication of our results for limited area ensemble-based data assimilation.

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

Curator: Nikki Privé Last Updated: May 27 2011