Exploration of the northern stratospheric polar vortex extremes via archetype analysis and tropospheric imprint

Abstract The winter stratospheric polar vortex is known to go through morphological change yielding various extreme states of its dynamic and thermodynamic state variables. These extreme states, including sudden stratospheric warming, generally have longer timescale compared to the chaotic limit of weather predictability, and may have significant impact on tropospheric variability. This offers the opportunity to boosting the predictability in the subseasonal-to-seasonal timescales especially regarding extreme events. Here extreme states of polar vortex are identified using archetypal analysis; a powerful unsupervised machine learning method. Polar vortex archetypes represent corners of the data convex hull and yield various extreme states including strong, split and displaced vortices. These states are shown to yield tropospheric and surface climate delayed response through teleconnections including North Atlantic Oscillation and Arctic Oscillation with lag of more than four to five weeks offering potential for predictability in the subseasonal-to-seasonal window.