Some Characteristics of GEFS Stochastic Perturbations

A Stochastic Perturbation Scheme (SPS) will be implemented into the Global Ensemble Forecast System (GEFS) at the National Centers for Environmental Predictions (NCEP) in 2009. The scheme represents the uncertainty associated with the forecast model itself by adding stochastic terms, known as Stochastic Perturbations (SP), to the tendencies of all model variables. To rescale the SPs, the current version of SPS employs a prescribed formula to determine a coefficient which varies with latitude, season and forecast lead time. This over simplification, especially the one-dimensional variation of the coefficient in space, set a limit to the benefit of SPS and thus need to be modified.

In order to develop an algorithm to automatically and objectively determine the size of the Stochastic Perturbations, their spatial structure, temporal evolution and statistical characteristics are investigated. Also studied are the relations of these SPs with the ensemble spread, errors of the ensemble mean and their rates of temporal variations. Based on the major characteristics of SP, a simple adaptive algorithm is designed to rescale their sizes with three dimensional variations and automatic seasonal adjustment.

This article will present the major results of the investigations and the preliminary tests of the adaptive algorithm.