A new augmented Ensemble of Data Assimilations (EDA) technique to estimate background error covariances (B-matrix) has been developed for the Copernicus European Regional Re-Analysis (CERRA). The B-matrix is modelled on a bi-Fourier limited area model. Background errors are assumed isotropic, homogeneous and non-separable. Linearised geostrophic and hydrostatic balances are incorporated as multivariate relationships, coupling vorticity and geopotential extended to mass-wind and specific humidity fields via the f-plane approximation. The B-matrix is estimated by a new 10-member CERRA-EDA system, temporally tethered to real-time meteorological situations. The EDA forecast differences comprise two main pools: seasonal and daily. The seasonal component is pre-prepared at reanalysis-resolution (5.5km). The new augmentation governs real-time mixture of winter and summer differences. The daily component is an 11km moving 2.5 day average. B-matrix re-estimation occurs every 2 days, with a fixed split of 80-20\% seasonal-daily. We consider a case study to illustrate the potential of CERRA-EDA to estimate weather regime change. The most influential factors are temporal evolution of spatial observation coverage, and varying the seasonal-daily split. Background error statistics, improvements in analysis and forecast skill scores and overall assimilation system performance are shown and discussed.