The advantage of the local ensemble square root Kalman filter (LEnSRF) based on covariance localisation over the local ensemble transform Kalman filter (LETKF) based on local domain updates is its ability to assimilate non-local observations, such as radiances. However, the analysis step of the LEnSRF is numerically challenging. In this new paper, my colleague Alban Farchi and I investigate the detailed implementation of the update step of the LEnSRF using augmented ensembles . In particular the ensemble modulation approach is compared to a randomised singular value decomposition tested for the first time in this context. The parallel nature of the latter method gives it a significant edge. Using twin simulations of a multilayer extension of the Lorenz-1996 model, we show that this approach is adequate to assimilate satellite radiances, for which domain localisation alone is insufficient.

The paper, entitled On the Efficiency of Covariance Localisation of the Ensemble Kalman Filter Using Augmented Ensembles, is published (open access) in Frontiers in Applied Mathematics and Statistics.