13B.4 Post-processing for multi-model ensembles

Friday, 29 June 2007: 11:15 AM
Summit B (The Yarrow Resort Hotel and Conference Center)
Thomas M. Hopson, NCAR, Boulder, CO; and J. P. Hacker and Y. Liu

A multi-model mesoscale ensemble constructed by swapping physical parameterization schemes, perturbing lateral boundary conditions, and perturbing observations in a cycling data assimilation system, is run for a 6-month period over the eastern U.S. As in many published studies, verification indicates that post-processing (calibrating) the ensemble may be necessary to provide meaningful probabilistic guidance to users. We apply a novel statistical correction approach by combining a selection of approaches used in the literature [e.g. Bayesian Model Averaging (BMA) and logistic regression] under the more general framework of quantile regression to improve mesoscale forecasts at specific probability intervals. We explore enhancing local forecast skill by including regional support for the nearest grid forecast (i.e. using nearest-neighbor forecast locations in the statistical correction procedures). We also introduce climatological quantile probabilities in the calibration so that our approach ensures that the forecast PDF, represented by the ensembles, has skill no worse than either a forecast of persistence or climatology. Results will be presented for a few selected locations along the eastern seaboard.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner