Adaptive observations in ensemble data assimilation

An important question in ensemble based data assimilation scheme is how to configure our observations to correctly capture the important features in either our atmospheric or oceanic models given a set of ensembles. In this paper a systematic approach for effective sensor placement is formulated to evaluate how to target our observations. This method is based on a criterion of Shannon information entropy and condition number of our ensemble subspace covariance matrix yielding adaptive observation configurations. The theory behind this method is presented as well as an example illustrated with a global shallow water equations model on the sphere.