Identifying Flavors of Lee Cyclogenesis over the Central U.S. Using Self-Organizing Maps

Previous regional studies of extratropical cyclone characteristics and predictability focus predominantly on coastal North America due, in part, to its larger population density. Yet, the central United States is also a favorable region for cyclogenesis to the east of the Rocky Mountains. This region’s distinctive location in the center of the continental United States and juxtaposition with complex topography contributes to unique forecast challenges, including uncertainties in the prediction of cyclone intensity and location, as well as concomitant areas of heavy precipitation. Consequently, this study utilizes machine learning to identify cyclone characteristics across the comparatively understudied central U.S that may be overlooked using traditional analysis methods.

Central U.S. cyclones were identified within the ERA5 reanalysis between 1980—2021 using a method for cyclone identification adopted from previously published work. This cyclone dataset facilitated an analysis of the monthly and seasonal frequency of Central U.S. cyclones, for which an increase in cyclone frequency between October—May 1980—2021 was found for winter and springtime cyclones. A Self-Organizing Map (SOM) was trained using mean sea-level pressure anomalies from ERA5 during lee cyclogenesis events and subsequently used to examine the variability in large-scale weather regimes conducive to lee cyclogenesis across the Central U.S. The extent to which the characteristics of cyclone environments (i.e., frontal structure, moisture sources, upper-level jet stream structure, quasi-geostrophic forcing, etc.) vary as a function of the prevailing SOM weather regime are also investigated.