dc.description.abstract | Tropical cyclone tornadoes (TCTORs) pose a warning challenge and are prone to having a high number of false-alarm warnings. Previous studies have found weather-radar attributes (i.e., rotational velocity (Vrot) or azimuthal shear (AzShear)) show some discriminating power between tornadic and nontornadic, but warned events. Given the time-consuming nature of a manual analysis, these previous studies often use a limited sample size. This work builds upon previous studies by developing and performing an automated method for determining radar attributes of 334 tornadic and 721 nontornadic initial events within 29 tropical-cyclone environments from 2013 to 2020. Rotation below 10 kft above radar level, vertically integrated liquid, and echo tops are examined to determine if these attributes can discriminate between tornadic and nontornadic events using an automated method. Results show that rotation, particularly rotation at lower elevation angles, is generally stronger in tornadic than nontornadic events whereas vertically integrated liquid and echo tops do not show much discriminating power. Warning statistics at 10 minutes before the tornado start time or nontornadic maximum rotation at an elevation angle of 0.5◦ show a best AzShear discriminator of about ~1 × 10^−2 s^−1 within 40 n mi. (74.1 km) of the nearest radar and ~8 × 10^−3 s^−1 beyond 40 n mi. Warning only for potentially more impactful events yields a higher probability of detection at higher rotation thresholds and the maximum Critical Success Index is shifted to higher thresholds as well, although with lower values. | |