Characterizing Stratospheric Polar Vortex Variability With Computer Vision Techniques

Abstract

AbstractComputer vision techniques are used to characterize the Arctic stratospheric polar vortex in 38 years of reanalysis data. Such techniques are typically applied to analyses of digital images, but they represent powerful tools that are more widely applicable: basic techniques and considerations for geophysical applications are outlined herein. Segmentation, descriptive, and tracking algorithms are combined in the Characterization and Analysis of Vortex Evolution using Algorithms for Region Tracking (CAVE‐ART) package, which was developed to comprehensively describe dynamical and geometrical evolution of polar vortices. CAVE‐ART can characterize and track multiple vortex regions through time, providing an extensive suite of region, moments, and edge diagnostics for each. CAVE‐ART is valuable for identifying vortex‐splitting events including, but not limited to, previously cataloged vortex‐split sudden stratospheric warmings. An algorithm for identifying such events detects 52 potential events between 1980 and 2017; of these, 38 are subjectively classified as distinct “split‐like” events. The algorithm based on CAVE‐ART is also compared with moment‐based methods previously used to detect split events. Furthermore, vortex edge‐averaged wind speeds from CAVE‐ART are used to define extreme weak and strong polar vortex events over multiple vertical levels; this allows characterization of their occurrence frequencies and extents in time and altitude. Weak and strong events show distinct signatures in CAVE‐ART diagnostics: in contrast to weak events, strong vortices are more cylindrical and pole centered, and less filamented, than the climatological state. These results from CAVE‐ART exemplify the value of computer vision techniques for analysis of geophysical phenomena.