The Predictors and Forecast Skills of Northern Hemisphere Teleconnection Patterns for Lead Times of 3-4 Weeks

The Pacific/North American (PNA) pattern, North Atlantic Oscillation (NAO), and Arctic Oscillation (AO) and are three dominant North Hemisphere teleconnection patterns known to strongly affect wintertime surface weather in North America. A partial least-squares regression (PLSR) method is adopted in this study to generate wintertime 2-week statistical forecasts of these three teleconnection pattern indices for lead times of up to five weeks over for the 1980-2013 period.  The PLSR approach generates forecasts for the teleconnection pattern indices through maximizing the variance explained by predictor indices determined as linear combinations of predictor fields, which include gridded outgoing longwave radiation (OLR), 300hPa geopotential height (Z300), and 50hPa geopotential height (Z50). Overall, the PLSR models yield statistically significant skill at all lead times up to five weeks.  In particular, the correlations between the weeks 3-4 PLSR forecasts and observations for the PNA, NAO and AO indices are 0.34, 0.28 and 0.41, all of which are statistically significant at the 1% level. The persistence of significant skill through five weeks provides promise for developing a seamless subseasonal-to-seasonal (S2S) prediction system. 

The PLSR approach also allows the authors to isolate a small number of predictor patterns that help shed light on the source of prediction skill for each teleconnnection pattern. As expected, the results reveal the importance of tropical convection (OLR) for forecast skill in weeks 3-4, but the initial atmospheric flow (Z300) accounts for a substantial fraction of the skill as well.