Filtering methods for finite-time normal modes and atmospheric error growth

Abstract

Krylov subspace methods are employed to generate the leading normal modes (NMs) and dominant finite-time normal modes (FTNMs) for atmospheric flows. The filtering techniques that play essential roles in our applications are described. The statistics of 100 evolved error fields are studied and compared with the growth and structures of fast growing NMs and FTNMs in a barotropic tangent linear model, with time-dependent basic states taken from observations.