Applying Stratospheric Polar Vortex Geometry to Subseasonal Forecasts

Stratospheric variability can influence the likelihood of cold air outbreaks and winter weather across the Northern Hemisphere on subseasonal-to-seasonal (S2S) timescales. This analysis considers the potential for windows of opportunity in S2S forecast afforded by stratospheric variability, as measured through variability in the geometry of the stratospheric polar vortex. Previous research calculated the statistical relationships between metrics, accounting for the geometry of the stratospheric polar vortex and wintertime U.S. tropospheric temperature, in reanalysis data. We revealed that the vortex center location, ellipticity, size, and rotation have distinct correlations with U.S temperature, even outside of periods of extreme stratospheric variability like sudden stratospheric warming events. The relationships between stratospheric vortex geometry metrics and U.S. temperature are of the same magnitude, or stronger in some regions, as those between the zonal-mean zonal wind metric of the stratospheric vortex and U.S. temperature.

This research extends the climatological analysis and explores the applications of stratospheric polar vortex geometry metrics in S2S forecasts of winter extremes. A combination of ERA-Interim reanalysis data and S2S reforecast data from the S2S prediction project dataset is used to examine the potential of the vortex geometry metrics from mixed dynamical-statistical forecasts by climatological analogs. The analogs are obtained by creating a decision tree of stratospheric vortex geometry metrics to produce week 2 temperature outlooks from the reanalysis data. The dynamical-statistical week 3-4 forecasts are created by using the decision tree outcomes but with the input of the week 1 and 2 forecasts of stratospheric vortex geometry. A comparison of skill from the week 3-4 temperature forecast based on vortex geometry forecast analogues and the S2S ensemble mean temperature forecasts at week 3-4 are discussed. The analysis highlights the potential for more comprehensive machine learning approaches, like neural networks, to be applied to windows of opportunity forecasts from stratospheric variability.