A Stochastic Perturbation Scheme For Representing Model Related Uncertainty


A stochastic parameterization scheme is designed for representing model related errors in ensemble forecasting. The strategy involves the addition of a stochastic term with structured noise in the tendencies of each model variable. For each member of the ensemble, the stochastic forcing is constructed as a random linear combination of tendencies of ensemble members and the control forecast. The flexible software infrastructure developed in the Earth System Modeling Framework (ESMF) collaboration can be employed to synchronize the integration of all ensemble members and the control run, and facilitate the information exchange among these runs.

A simplified version of the scheme is tested with the NCEP Global Ensemble Forecast System (GEFS). Experimental forecasts for October, 2004 show that the scheme can significantly increase the ensemble spread, with reduced number of outliers and reduced systematic error in the ensemble mean forecast. The performance of probabilistic forecasts based on the NCEP global ensemble system is also improved, especially in terms of Ranked Probability Skill Score (RPSS).

The scheme could also be used in other Earth system modeling applications such as hydrologic ensemble forecasting.

Dingchen Hou, Zoltan Toth and Yuejian Zhu