May 22 - July 28
Characteristics and Statistics of Tropical Cyclone Tornadoes and Warning Skill
Walter McGavran, Benjamin A. Schenkel, Thea N. Sandmæl, Kristin M. Calhoun, and A. Addison Alford
What is already known:
- Forecasts for TC tornadoes are less skillful than those elsewhere in the continental US for all lead times.
- Tropical cyclone tornadic supercells tend to be more marginal with shallower updrafts, more compact mesocyclone diameters, and shorter lifespans than non-TC tornadoes elsewhere in the continental US.
- A single prior case study showed fluctuations in warning skill based on time of day and distance from TC center.
What this study adds:
- Forecasters issue false alarm tornado warnings at a much higher relative rate for tropical cyclone tornadoes than other tornadoes in the continental US.
- Demonstrated significant difference in azimuthal shear and divergent shear distributions and medians for tornadoes which were accurately forecast versus those which were not.
- Demonstrated relatively low variability in forecasting performance across large sample size parameters such as distance from the TC center, time relative to local sunset, TC intensity, and tornado damage rating.
Abstract:
Landfalling tropical cyclones (TCs) often spawn tornadoes, which yield conflicting call-to-actions between tornado warnings and those issued for other hazards (e.g., flash flooding). There have only been a limited number of TC case studies examining tornado warning forecast skill within TC environments, often yielding differing results. Hence, this study conducts a climatological analysis of tornado warning skill within TCs from 2011–2021, using observed TC and tornado data in conjunction with single-radar data quantifying low-level rotation and convergence. Tornado warning skill was lower for TC tornadoes than values of non-TC tornadoes cited in prior work. Warning skill was generally insensitive to parameters like distance from the TC center, time of day, and TC intensity. However, despite the low number of tornadoes far from the TC center (i.e., ≥600 km), warning issuance was still very high, resulting in poor forecasting performance. Additionally, tornadoes that occurred after local sunset were forecast with greater skill than had previously been demonstrated, with tornado detection comparable to daytime rates. Probability of detection was substantially increased with increasing tornado damage rating. To explore forecasting thresholds that could increase skill, low-level azimuthal (i.e., quantifying low-level rotation) and divergent shear (i.e., quantifying low-level convergence) were examined which showed that warned tornadic supercells typically had strong values com- pared to both non-warned tornadic supercells and warned nontornadic supercells. Together, this work provides a groundwork for potentially improving tornado forecast skill in landfalling TCs.