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

The subseasonal-to-seasonal (S2S) variability of the Northern Hemisphere Stratospheric Polar Vortex (SPV) is primarily influenced by the vertical propagation of Rossby waves. These waves can trigger intense events known as Sudden Stratospheric Warmings (SSWs), characterized by warming and reversal of zonal winds in the SPV area. SSW effects extend beyond the stratosphere, impacting the troposphere for several weeks. This study investigates the predictability of the boreal winter stratosphere, focusing particularly on the relationship between lower-stratosphere wave activity (LSWA) represented by meridional eddy heat fluxes at 100-hPa which serves as an indicator of vertical propagation of Rossby waves, and the intensity of the wintertime SPV. Specifically, we hypothesize that improving the predictability of the North Atlantic Oscillation (NAO), a dominant climate mode influencing winter circulation in the Northern Hemisphere, will enhance prediction accuracy in the SPV region. A poor representation of model variability in the North Atlantic area, a known problem referred to as signal-to-noise paradox, limits the capacity of ensemble-based seasonal prediction systems (SPSs) in forecasting the NAO. To partially overcome this issue, we adopt a subsampling approach, already used in recent studies, with the idea of applying statistical methods to a dynamical prediction reducing the ensemble size of a SPS. This is performed with the use of four variables studied during the autumn season strongly correlated with the wintertime NAO, referred as predictors. The results demonstrate one more time the effectiveness of this approach in increasing the predictability of NAO index and its variability within SPSs. Furthermore, the analysis evidences an enhanced ability of the models in predicting the intensity of zonal-mean zonal winds in the SPV region with correlation prediction skill for wintertime SPV going from 0.39 to 0.64 for a 3 model multi-ensemble. We find that the hybrid statistical-dynamical approach improves the ability of SPSs in predicting the number of SSW days, defined as days during a winter season presenting an inversion of zonal-mean zonal winds in the SPV region, according to SSW definition. Additionally, our results suggest that these improvements are partly explained by an increased ability of SPSs in reproducing the vertical propagation of Rossby waves during December and January (DJ), especially over two critical regions across Eurasia. This result highlights the importance of the DJ LSWA for the seasonal prediction of the SPV.