Non-Stationarity of Wintertime Atmospheric Circulation Regimes in the Euro-Atlantic Sector

Abstract:

Atmospheric circulation regimes can be used to study links between regional weather and other climate processes, such as sudden stratospheric warmings. For these studies it is important to know whether there is any background non-stationarity in the regimes themselves. In the identification of circulation regimes mostly anomalies with respect to a seasonal climatology are considered, even when only one season is studied. Using a fixed climatology has been found to not significantly alter the regimes, whereas it does introduce a seasonal cycle in the occurrence rates of the different regimes.

We use ECMWF SEAS5 hindcast data to study intra-seasonal non-stationarity of circulation regimes. A (persistent) k-means clustering algorithm using six clusters is applied to the data to identify the circulation regimes for winter (DJFM). In a previous study six has been identified as the optimal number of regimes via an information criterion and allows for more variability than the common-used four regimes (NAO+/-, Atlantic Ridge, Scandinavian Blocking). The two additional regimes are labelled as the negative phase of the Atlantic Ridge and Scandinavian Blocking, respectively.

We discuss whether changing occurrence rates of these six regimes throughout winter can be fully attributed to a changing background state or whether other processes play a role. Furthermore, we discuss changes of the regimes on a decadal timescale using the same hindcast data. The hypothesis in literature is that changes in the regimes predominantly occur as changes in the occurrence rate and transition probability of the regimes, whereas the regimes themselves remain unchanged. Early results indicate that for six regimes this may not necessarily be the case and the regimes themselves change as well. This suggests that the additional variability of the six regimes allows for identifying a non-stationary signal which cannot be picked up using four regimes.