Self-Organizing Maps for Tornadic Near-Storm Environments of the Southeastern United States

The distribution of tornadic near-storm environments in the Southeastern United States represents a shift away from the "textbook" environments of Great Plains tornadoes: Southeastern tornadoes are, on average, characterized by more marginal instability (as measured by Convective Available Potential Energy), as well as stronger wind shear and greater storm-relative helicity than their Great Plains counterparts. Given that the percentage of deadly tornadoes in the Southeast is nearly twice as high as in the continental United States as a whole, these relatively high-shear and low-CAPE (HSLC; typically defined as having mixed-layer CAPE values less than 1000 J/kg and 0-6-km shear values greater than 35 kt) environments have been a primary research focus in recent severe storms literature, although past studies' datasets have frequently contained fewer than five years of data.

Using self-organizing maps to cluster together similar spatial patterns of environmental parameters, we investigate a dataset consisting of the 3,670 tornado events that occurred in the Southeastern United States between 2003 and 2015, over half of which are considered HSLC events, as defined above. In this way, "prototypical" tornadic near-storm environments are described for all Southeastern tornado events as well as for the HSLC subset of Southeastern tornado events, and the statistics of these summary environments are discussed and compared, including in the context of simple metrics of warning skill such as probability of detection (POD; the percentage of all tornadoes with positive lead time). The spatial information provided by self-organizing maps in this application allows for a deeper and more nuanced understanding of regional differences in tornadic near-storm environments.