Seasonal Prediction of Wintertime Teleconnections—Empirical Model Compared to CFSv2

Dynamical and empirical seasonal forecasts over the Pacific-North America domain tend to be dominated by signals related to long-term trends and low-frequency SST forcing from ENSO. There has been a great deal of literature over recent years related to the prediction of patterns of variability confined to the middle and polar latitudes, especially the North Atlantic Oscillation (NAO). A recently published empirical model of wintertime seasonal NAO prediction (Wang et al. 2017) is tested using real-time data sets at CPC. Using autumn SST, sea ice concentration, and stratospheric geopotential height to derive the pool of potential predictors, the model is able to replicate published results qualitatively, but not at the same magnitude. The empirical model is extended to forecast seasonal values of the NAO and North Pacific Oscillation-West Pacific pattern (NPO/WP) that are derived from a traditional EOF-based teleconnection analysis. The results are also promising with respect to the NPO/WP even after rigorous cross validation.

The results of the empirical model are compared to CFSv2 forecasts with favorable results. The CFSv2 demonstrates only marginal skill at predicting the seasonal NAO, and zero skill at predicting the seasonal NPO/WP. These results suggest that extending operational empirical prediction at CPC to include initial conditions in sea ice concentration and stratospheric circulation may improve improve winter forecasts by better supplementing our suite of dynamical model guidance with information beyond long-term trends and canonical ENSO impacts.