An overview of the variational assimilation in the ALADIN/France numerical weather-prediction system

We present an overview of the 3D-Var data assimilation in the framework of the ALADIN/France model. The purpose of this system is to provide improved precipitation forecasts at mesoscale and in the short range, up to 18 hours. The goal of the paper is threefold. Firstly, we present initial considerations for the design of the 3D-Var system. Secondly, we discuss in more detail the specification of the background-error covariance matrix, by comparing three different error simulation techniques, namely two variants of the NMC method and an ensemble-based approach. The formal, diagnostic and impact studies have led to the selection of the ensemble-based covariances for the ALADIN/France assimilation. Thirdly, scores of quantitative precipitation forecasts are shown in order to illustrate the robustness and the preliminary meteorological performance of the ALADIN/France assimilation suite. The results indicate that the tested configuration improves some aspects of the precipitation forecast, while being neutral for others, when compared with the spin-up model. We conclude the paper by providing a more explicit insight into the future evolution of limited-area variational analysis towards convective-scale data assimilation.