The SEIK (Singular 'Evolutive' Interpolated Kalman) filter is anensemble-based Kalman filter. It has been developed for sequentialdata assimilation with large-scale non-linear numerical models. TheSEIK filter is particularly efficient compared, e.g., to the morecommon Ensemble Kalman filter (EnKF) due to the combination of theensemble-representation of forecast errors with an explicit operationon a low-dimensional error space. A localized variant of the SEIKfilter has been developed to further improve the performance of thefilter algorithm. I will review aspects of sequential dataassimilation and discuss the SEIK and local SEIK filters as well astheir implementation within the assimilation framework PDAF for dataassimilation on parallel computers. The application of the filters isdiscussed on the basis of experiments assimilating satellite data ofeither sea surface height or ocean color into ocean andocean-biogeochemical models.