Extremely values of the Arctic Oscillation (AO) Index in the stratosphere have recently been linked to anomalous weather regimes at the surface for periods of 60 days following the anomalous events. This has lead to the suggestion that stratospheric variability might be useful for extending weather prediction beyond current time scales. However, the cause of extreme stratospheric events has not been identified.

In order to elucidate the cause of the extreme AO index events, we examine, using NCEP reanalyses from 1958 to 2001, the upward wave flux near the tropopause levels, integrated of periods of 30 days prior to extreme stratospheric events. We find that the AO timeseries the stratosphere (10hPa) and 30-day-integrated wave flux time series at 100 hPa correlate with an extremely high coefficient (typically 0.8) over the entire record, and that there is a clear difference in the upward wave flux from the troposphere between the periods preceding extremes in the phases of AO index at 10hPa. In particular, the eddy heat flux at 100 hPa integrated over a month is anomalously large (small) preceding the onset of a weak (strong) polar vortex. Furthermore, the evolution of the AO index following periods when the integrated 100 hPa wave fluxes are anomalous is very similar to that of the AO events discussed in Baldwin and Dunkerton (2001): The anomalous AO index occurs first in the upper stratosphere, migrates down into lower stratosphere and troposphere, and persists for 30-60 days at the surface.

Our results thus indicate that, although the AO signal appears first in the upper stratosphere, the "forcing" for these events can be detected in the upper troposphere first, and that the 30-day integrated eddy heat flux at 100 hPa is the key quantity for understanding the occurrence of extreme stratospheric vortex episodes, and their related impacts on surface weather.