Localization

Under construction.

Truncation of covariance matrix

Domain-based localization

Under this scheme, the global state is decomposed into local domains based on the (geographic) location that each state vector element represents. Furthermore, only observations within the de-correlation length are used for computing the analysis update of each local domain. The de-correlation length is given either as a radius (i.e. a single scalar) or more typically via a taper function (endas.localization.TaperFn).