The ensemble-transform scheme adapted for the generation of stochastic forecast perturbations

Abstract

A method for generating stochastic perturbations is derived from the ensemble-transform analysis-error sampling scheme. The method produces stochastic perturbations that not only have multivariate spatial correlation, balance, and prescribed mean and variance statistics, but also conditioning for growth. The perturbations can serve as modification for numerical weather prediction (NWP) ensemble forecasts either during integration or post-processing. A practical demonstration of the method is given, in which operational global NWP ensemble forecasts are modified with the stochastic perturbations as a post-processing exercise. The results suggest that the method can produce stochastic perturba..