Transformed and generalized localization for ensemble methods in data assimilation (Q2802656)

The authors study and analyze transformation localization and generalized localization for ensemble methods in data assimilation. Localization is important in data assimilation, since it guarantees a sufficient number of degrees of freedom when using a small number of ensembles. It has been successful when the observation operator is local, but very limited for nonlocal observation operators. The authors propose some transformation to change nonlocal operators to local ones in the transformed space, analogous to preconditioning in numerical linear algebra. The authors study the general algebraic formulations, and provide several examples to illustrate the approach, including temperature profile reconstruction from infrared measurements given by the infrared atmospheric sounding interferometer.