Thursday, 13 January 2005: 4:15 PM
Seasonal Forecast Skill Comparison of Cluster mean, Ensemble Mean and EOF Mode patterns
Seasonal Forecast Skill are compared between cluster mean, ensemble mean, and EOF modes patterns obtained from the NCEP SFM hindcast dataset over the Pacific-North America (PNA) sector.The anomalous correlation coefficients (ACCs) are computed for all cluster mean patterns at all hierarchical levels, and the highest cluster mean ACCs for top 4 clusters exceed the ensemble mean ACCs. These results were obtained when the cluster analysis was applied to the ensembles of both each 1-year and of the 21-year. The results indicated the existence of multiple atmospheric regimes. Multimodal regimes were found during normal episodes, while single regimes were found during warm/cold episodes. The dissimilarity and the number of the cluster members were dependent on the tropical sea surface temperature (SST), suggesting that the external forcing regulated the frequency of the occurrence of the regimes. The ACC improvements from the ensemble mean ACCs to the cluster mean ACCs are statistically significant for the top two clusters. These improvements are useful as a supplementary forecasting material in addition to the ensemble mean.