Iterative eignsolvers based on Arnoldi and Krylov subspace methods [1, 4, 9] are developed and applied to normal mode and finite-time normal mode instability problems. These techniques, which are generalizations of the Lanczos method [6], yield the leading eigenvalues and eigenvectors of large non-symmetric linear operators and matrices. They are used to generate the faster growing normal modes and finite-time normal modes for the barotropic vorticity equation linearized about both frozen instantaneous and time-dependent atmospheric basic states. We aim to show that initially random errors in barotropic weather forceasts take up evolved structures similar to the dominant finite-time normal modes during the periods of rapidly developing atmospheric blocks.
展开▼
