[4-3]

Development and test of a 4D-Var assimilation for the convective-scale model AROME-France.

P. Brousseau (CNRM-GAME (Meteo-France/CNRS))

AROME-France is the operational convective-scale Numerical Weather Prediction system in operations at Meteo-France since the end of 2008. It uses a 3D-Var Data Assimilation scheme in a 3-h continuous assimilation cycle in order to determine its initial conditions at the horizontal resolution of the model (2.5km). The weakness of the representation of the temporal dimension in this system leads to an inefficient use of observations with high temporal and spatial coverage (screen-level observations, radial wind and reflectivity from radar measurements,...) very informative at the convective-scale.

Different ways to improve the use of such observations have been investigated. First the increase on the assimilation cycle frequency (using a 1 h period) , allowed by the reduction of the numerical noise in the first ranges of the model forecast, has been tested and will become operational in 2015. Secondly, a 4D-Var scheme in a 3-h assimilation cycle has been developed in a research framework. The present study deals with the description of the configuration and the numerical cost of this 4D-Var system. Then, results obtained with 3D or 4D-var in a 3-h or 1-h assimilation cycle during a long period assimilation experiment are compared. Finally the potential of a one-hour non-cycled data assimilation for now-casting applications will be shown.