Statistical Correction of Dynamical Prognoses: The Decision Problem

Abstract

A statistical correction scheme significantly reduces the mean S1 skill score of the Australian Region Primitive Equation Model. However, on a number of days, the statistically predicted corrections degrade the dynamical prognoses. Empirical Bayes? methods are used here to construct decision rules which reject a priori most of the deleterious corrections, at the expense of rejecting some of the beneficial corrections.