For 50 years (1949-1998), the August-September Arctic Oscillation (AO) index served as a good predictor for the November-December circumpolar vortex (CV). The correlation coefficient between the unaltered values of AO and CV for the above period is -0.59, which is highly statistically significant (p < 0.0001), given that the data satisfies all the assumptions for the Pearson correlation coefficient. In recent years, however, the relationship deteriorated to the point that the correlation coefficient switched the sign.
The SLP-based AO data is available from Colorado State University and Climate Prediction Center. The CV is defined here as the mean 50 hPa heights north of 80N. Applying STARS to this data reveals a shift in the mean in CV that occurred in 1970 (Fig. 1). No regime shift in the mean was found in AO. The variance in both AO and CV was quite stable over the period of observation, 1949-2014.
The result of regime shift analysis in the correlation coefficient is presented in Fig. 2. It shows an abrupt shift in the correlation coefficient from -0.51 in 1949-1998 to 0.57 in 1999-2014. The difference between these correlation coefficients is statistically significant at p = 0.0001.
A dramatic alteration in the seasonal progression of atmospheric circulation over the Northern Hemisphere from summer-fall to prewinter that occurred in the late 1990s is clearly seen in Fig. 3. Prior to 1999 (Fig. 3a), a strong (weak) westerly flow at the surface in August-September was typically followed by a strong (weak) circumpolar vortex, meaning a zonal (meridional) type of circulation aloft in November-December. Since 1999 (Fig. 3b), the pattern flipped, and now a strong (weak) westerly flow at the surface in August-September is more often followed by a meridional (zonal) type of atmospheric circulation aloft in November-December.