Testing Linear Diagnostics of Ensemble Performance on a Simplified Global Circulation Model

Abstract

Ensemble weather forecast systems are used to account for the uncertainty in the initial conditions of the atmosphere and the chaotic dynamics of the models. It has been previously found that forecast performance of an ensemble forecast system is inherently flow dependent and that the ensemble predicts potential patterns of forecast errors more reliably than the magnitudes of the errors. A low-resolution global circulation model is implemented to calculate linear diagnostics in the vector space of the ensemble perturbations. It is confirmed that the ensemble provides a good representation of the pattern of forecast uncertainties, but the magnitude of the error projected onto the space of ensemble perturbations is underestimated for the 6 to 120 hour forecast times.