Variance decomposition approach to the prediction of the seasonal mean circulation: Comparison with dynamical ensemble prediction using NCEP's CFS

Abstract

AbstractReanalysis data and a recently developed variance decomposition approach are applied to empirically predict seasonal mean 500 hPa geopotential height anomalies in both the Northern and Southern Hemisphere from the beginning of each season. The predictors are only several popular tropical SST indices and the annular mode indices at one month or one season lead. The statistical prediction is compared with the dynamical ensemble prediction of the National Centers for Environmental Prediction's coupled Climate Forecast System (CFS). It is found that their skills are quite comparable. The advantages and disadvantages of dynamical prediction and statistical prediction are discussed. A strategy for a new statistical–dynamical prediction which combines the advantages of both statistical and dynamical predictions is suggested. Copyright © 2008 Royal Meteorological Society