Statistical Analysis of Stratospheric-Tropospheric Coupling in Extended Range Ensemble Forecasts

Abstract:

Each winter, strong cooling of the polar stratosphere gives rise to a strong stratospheric jet, known as the polar vortex. However, in about two out of three winters this polar vortex gets abruptly destroyed, associated with a drastic stratospheric temperature increase of up to 50°C: Sudden Stratospheric Warmings. It is well established, that these dynamically impressive changes in the circulation can on the one hand at least partly be elicited by the troposphere (upward coupling). On the other hand, they are also able to influence subsequent tropospheric weather in mid-latitudes (downward coupling). This coupling is highly variable across sudden warming events, though; hence, large statistics are highly beneficial to investigate the downward coupling, including associated tropospheric extreme states. Unfortunately, due to their sporadic nature only a limited number of sudden warming events is available from the observational record.

In this study, ECMWF’s extended range ensemble forecasts from the S2S database with a 46-day lead-time are used to study the connection between stratospheric and tropospheric extreme events in more detail. Importantly, these ensembles greatly boost the statistics of sudden warming events by including not only those events that took place in the real atmosphere, but also those that could have happened (were forecast) but did not materialize. The ECMWF ensemble includes realtime and hindcast forecasts initialized twice a week. In total, this creates a database of 7859 forecast runs, in which we diagnosed 1550 sudden warming events.

The following key results can be concluded for downward coupling in the Northern Hemisphere as quantified by the Arctic Oscillation (AO) index. The distribution of daily AO values is robustly shifted toward negative values following sudden warming events, consistent with previous findings. Likewise, an anomalously strong polar vortex results in a positive shift of the AO distribution. The probability of extreme negative AO events (below -3 standard deviations) is enhanced by a factor of three to four following sudden warmings. Extremely negative AO values are more often preceded than followed by sudden warmings (19% versus 7%). In contrast, sudden warmings appear to be suppressed during periods of extreme positive AO values.