On the Weights for an Ensemble-Averaged 6–10-Day Forecast

Abstract

A scheme to optimally weight the members of an ensemble of forecasts is discussed in the framework of calculating an as accurate as possible ensemble average. Results show, relative to a single member, a considerably improved 500-mb height forecast in the 6–10-day range for the Northern Hemisphere. The improvement is nontrivial and cannot be explained from simple smoothing. This method is used in operations at the National Meteorological Center.