Temporal Variations in Predictability

Abstract

A large ensemble of predictability runs made during the course of a long equilibrium run in a two-level nonlinear quasi-geostrophic model with orography is examined in order to elucidate characteristics contributing to temporal variations in error growth. After the initial dissipation of the small-scale error, an error spectrum is developed wherein all scales grow with about the same doubling time until saturation is reached first at the smallest scales. Toward the end of the predictability runs, the error spectrum steepens toward the equilibrium energy spectrum. This error growth is largest during times of large equilibrium kinetic energy. Because of a lag relationship between the equilibrium kinetic energy and the available potential energy, it is possible to marginally predict times of large and small error growth. Removal of the orography during a forecast produces much larger and more linear growth rates characteristic of present operational forecast model errors.