Pre- and Post- Processing in Data Assimilation

This paper provides an overview of the details of the data assimilation system implementation, at both pre and post-processing levels, which are, more often than not, mentioned in passing in data assimilation talks or articles. Data assimilation is the art of combining model and observations. It relies on a set of equations grounded on sound statistics. Some theoretical studies investigate ways to define optimally various quantities (observation operators, observation and background error covariances) and how to combine all the flow-dependent information (time evolution of the model, using linearized models in 4D-Var or ensembles in EnKF).