An ensemble-based four-dimensional variational (En4DVar) algorithm has been presented (Liu et al 2008), which uses the flow-dependent background error covariance constructed by ensemble forecasts and performs 4D-Var optimization based on the incremental approach and preconditioning algorithm. We conducted Observing System Simulation Experiments (OSSE) for En4DVar that showed promising results (Liu et al 2009).

The current study extends the En4DVar to assimilate real observations for an explosive cyclone happened in the Antarctic and South Ocean on 2-5 October 2007. We performed an intercomparison of four different WRF-variational approaches for the case, including three-dimensional variational data assimilation (3DVar), first guess at the appropriate time (FGAT), ensemble-based three-dimensional variational (En3DVar) and En4DVar. It is found that all data assimilation approaches can give some positive impact in this case study even few in situ observations are assimilated. When the flow-dependent background error covariance is applied, the forecast skills obtained by using En3DVar and En4DVar are improved relative to using 3DVar and FGAT with the homogeneous and isotropic background error covariance respectively. The impacts of the inflation factor and the ensemble mean background on En4DVar are examined in this case. The results show the forecast from En4DVar isn't sensitive to inflation factor and is degraded by the ensemble mean background