A hybrid statistical–dynamical approach for seasonal prediction of the boreal winter stratosphere

The subseasonal–to–seasonal variability of the Northern Hemisphere Stratospheric Polar Vortex (SPV) is primarily influenced by the vertical propagation of Rossby waves, which can trigger intense events known as Sudden Stratospheric Warmings (SSWs). These events are characterized by a rapid warming and reversal of zonal winds in the SPV area, with effects extending beyond the stratosphere, impacting the troposphere for several weeks. However, current seasonal prediction systems (SPSs) exhibit low to moderate skill in SPV prediction, with both deterministic and probabilistic predictability for SSWs still challenging to achieve. This study examines a new approach to improve SPV seasonal prediction: we hypothesize that enhancing forecast ability for the North Atlantic Oscillation (NAO) could improve prediction accuracy in the SPV region by enabling a better representation of lower-stratosphere wave activity within SPSs. To achieve this, we implement a subsampling approach that applies statistical methods to dynamically generated ensembles, reducing ensemble size to increase signal-to-noise ratios in SPS predictions. This method uses autumn variables strongly correlated with NAO, referred to as predictors, as a reliable NAO statistical estimation. Our results show consistent improvements in both NAO and SPV prediction skills across individual models and multi-model ensembles, supporting our initial hypothesis.  These improvements are partly explained by a better representation of extratropical lower-stratosphere wave activity during December and January (DJ), especially over critical regions exhibiting higher variability. We also assess the representation of troposphere–stratosphere coupling in both reanalysis data and SPSs, revealing that most models overestimate the strength of the NAO–SPV link. Finally, we examine the role of autumn predictors in influencing both tropospheric and stratospheric dynamics, highlighting their potential as joint predictors for NAO and SPV variability in seasonal forecasts.