Forecasting the North Atlantic Oscillation Using North Pacific Surface Pressure

Abstract

A statistical forecasting scheme for the North Atlantic Oscillation (NAO) is presented. The forecasts are based on Pacific Ocean surface pressure. With 21 months lead time, skills are significant and useful but modest. At 15-month lead, forecasts are robust and skillful under stringent cross validation, and further improve at 12-month lead. Cross-validated 1-yr forecasts correlate with 1925?2002 observations at ?0.45. Other performance measures indicate similar skills, out performing the best autoregressive models of the NAO index. Importantly, skills of North Pacific?based forecasts easily exceed those of North Atlantic?based ones using identical machinery. The results suggest that NAO variability is not exclusively internal to the North Atlantic, but is also a response to upstream forcing from the North Pacific.