Ingredients-Based Examination of U.S. Severe Tornado Alleys Using Reanalysis Data

A new metric for identifying severe tornado (F3+) alleys is presented. Regions are distinguished by average F-scale values with a minimum of 15 severe tornado events within 200km. Five distinct "severe tornado alleys" are described across the United States: 1) the Carolinas, 2) "Dixie Alley," 3) Kansas and Oklahoma, 4) Iowa and Nebraska, and 5) Wisconsin, Illinois, Indiana and Michigan. These regions are evaluated using CFSR global reanalysis data to determine the presence of atmospheric variables relating to tornadogenesis (as described by prior work) in the pre-tornadic environments. Composite CFSR soundings and composite field charts are created for severe tornado events within each region. A robust comparison of pre-tornadic atmospheric conditions among regions is conducted using a Student t-test at 95% confidence.. An evaluation of the Sigtor parameter, a commonly used significant tornado forecast parameter developed by the Storm Prediction Center, is also conducted to identify regional variability. It is found that each region significantly differs at 95% confidence from each other in terms of many atmospheric variables from pre-tornadic environments, with the most significant differences existing between the two more eastern and two mid-western regions. SigTor values and the individual components, are calculated for severe tornado events and for every time-step for every grid-point within each region over a "peak severe tornado season" for U.S. severe tornadoes in order to compare the event cases to a background mean state. It is shown using linear correlations and covariances of the individual terms of the Sigtor parameter to the Sigtor value that not only is there discrepancy in Sigtor values between regions, but the different components of SigTor contribute to the value of Sigtor in different ways between regions. Some regions show background values of Sigtor terms to be favorable for tornadogenesis even when tornadoes have not occurred. Regional biases in presence of atmospheric variables preceding severe tornado events between regions are described.