Ensemble data assimilation systems are routinely run at all major NWP Centres for the initialization of ensemble prediction and for the cycling of error covariances to be used in a deterministic assimilation cycle.

In this presentation we will first describe recent upgrades to the operational ensemble data assimilation system at ECMWF, which is based on an ensemble of adjoint-based, perturbed observations 4D-Var analyses.

We will then report on recent results obtained with a quasi-operational configuration of an alternative hybrid EnKF-Var ensemble data assimilation system based on the Hybrid-Gain approach and discuss current areas of research.

---

*email: massimo.bonavita@ecmwf.int
*Preference: Oral