20Society Diurnal and Seasonal Patterns of Convective Weather Hazards across Tennessee and Their Likelihood of Causing Harm

Tennessee is prone to three prevalent severe convective hazards: floods, tornadoes, and nontornadic thunderstorm winds. The climatology of these hazards varies seasonally, diurnally, and spatially. For example, flash flood frequencies are known to be greatest in warm season in much of the Southeast, though areas prone to tropical cyclones are known to have fall flood signals as well. Similarly, the likelihood of the hazards causing harm to people has seasonal and diurnal variability that is partly associated with their climatological patterns but also dependent on how human activity and vulnerability vary across space and time. For example, flash flood fatalities peak during twilight and darkness, most often occurring when people are in vehicles.

Because hazards are often studied separately, and studies use different methods, periods, and regions, there is not a clear picture on how the climatology and harm from convective hazards shift over space and time for a given location. The primary goal of this work was to assess the historical frequency of hazards across the state and their likelihood for causing harm to determine how their frequency varied seasonally and diurnally, and to highlight areas where vulnerabilities may attribute to a greater likelihood of an injury or fatality at a specific time.

Data: Data between 1996 and 2022 from the National Centers for Environmental Information Storm Events Database, Version 3.1, were used to gather hazard, injury, and fatality information. Events were divided into categories based on whether they caused harm and when they occurred. Hazards that resulted in at least one direct fatality or injury were labeled as harmful hazards (0 = not harmful; 1 = harmful). Hazards were divided into two diurnal categories––daytime and nighttime––and into the four meteorological seasons: spring (March, April, May), summer (June, July, August), fall (September, October, November), and winter (December, January, February). Daytime and nighttime categories were determined using sunset and sunrise times.

Methods: We used chi-square tests to identify binary relationships between variables, testing whether there was a significant association between hazard types, harm, season, harm likelihood season, and diurnal timing. Cramer’s V test was used as a post-hoc test to assess the strength of each significant relationship. Then, we used logistic regression to model whether a hazard caused harm based on hazard type, time of day, and season. The independent variables were tested for multicollinearity using the Variation Inflation Factor (VIF). VIF for all variables was close to 1.00, indicating little correlation between variables. We tested model fit by comparing the full regression models to an intercept-only model, and they were significantly different (p<0.001), confirming that the predictors significantly improved the model.

Results: A total of 19,100 hazardous events occurred in Tennessee between 1996 and 2022. Nontornadic thunderstorm winds were the most common hazard (80 %), followed by flash floods (15 %), and tornadoes (5 %). The annual frequency of the hazards varied, but there was no trend in annual frequencies during the period. Approximately 2 % of the hazards were harmful.

Hazards were most common and most harmful in Middle Tennessee and densely populated counties. Generally, spatial patterns indicate an urban bias of reports for hazards and harm. Chi-square tests identified differences in hazard and harmful hazard occurrences between seasons and from day to night. A logistic regression model indicated that being classified as a tornado and occurring at night significantly increases the likelihood of causing at least one injury or fatality, and being a nontornadic thunderstorm wind and occurring in summer makes a hazard less likely to cause harm.

Conclusion: These findings demonstrate how hazard climatology and human vulnerability operate in different patterns over a year and a day across Tennessee and point to the increased danger of hazards occurring at night.