Examining Nocturnal Hazard Risk Perception and Protective Decision-Making in the Southeast

Background: The U.S. Southeast region experiences a high proportion of rapid-onset hazards (e.g., tornadoes, flash floods) that can occur quickly with little warning. When rapid-onset hazards occur at night, they increase public fatality odds as they are difficult for the weather enterprise to forecast and difficult for the public to respond to, as much of the population may be sleeping when they occur. Furthermore, the Southeast contains a high percentage of socially vulnerable populations who may not have access to resources to receive overnight alerts (e.g., NOAA weather radio, smartphone). The primary goal of the study was to identify factors that influence an individual personalizing risk related to nocturnal tornadoes and flash floods and developing protective actions. First, in consultation with NWS Weather Forecast Office partners in the Southeast region and NOAA Weather Program Office subject matter experts, we created two nocturnal hazard risk scales (6 items each) to measure the concept of nocturnal risk perception related to tornadoes and flashfloods events. We also developed additional questions in a survey to assess: 1) multiple modes of nocturnal hazard alerts; 2) protective decision-making, actions, behaviors for nocturnal hazards; and 3) socio-demographics, household characteristics, and resources. The current study examined the reliability and validity of the adapted scales created to measure nocturnal risk perception factors and protective decision-making among members of the public (N= 1,041) living in geographic areas with nocturnal hazard event frequencies and overlapping with socially vulnerable populations.

Methods: Data analyses were completed using R statistical software and packages. First, we conducted a Confirmatory Factor Analysis (CFA) to examine the indicator reliabilities for the nocturnal hazard risk perception constructs developed. Cronbach’s alpha was also used to assess internal reliability of the scales used to measure perceived nocturnal tornadoes and flash flood risk. In addition, a multivariate regression model was estimated to examine the associations between different factors hypothesized to influence people’s protective actions during nocturnal hazards (e.g., nocturnal risk perception, resources, socio-demographics, multiple warning sources).

Results: Our CFA model achieved acceptable fit, χ2(12) = 77.359, p <.01, CFI = .96, TLI = 0.92, RMSEA = 0.08, SRMR = 0.07. CFA results found acceptable to moderate factor loadings (λ values above 0.50), indicating they represented the latent construct of nocturnal risk perception well. Cronbach’s alpha results for the nocturnal tornado risk scale were a= 0.85 and a= 0.88 for the nocturnal flash flood risk scale, also indicating high internal consistency and reliability for measuring the constructs. In addition, our regression model achieved acceptable fit, χ2(40) = 273.501, p <.01, CFI = .93, TLI = 0.90, RMSEA = 0.08, SRMR = 0.07. Regression results found that for the public, having multiple modes of hazard-related information sources (β = 0.221, p<.001), nocturnal tornado risk perception (β = 0.139, p<.001), and resources (β = 0.127, p<.01), increased nocturnal tornado protective actions. Likewise, for nocturnal flash floods protective actions, having multiple modes of hazard-related information sources (β = 0.260, p<.001), nocturnal flash flood risk perception (β = 0.316, p<.001), and resources (β = 0.109, p<.01), were found to increase nocturnal flash flood protective actions.

Conclusion: The current study provides new insights into measuring nocturnal hazard risk perception along with identifying factors that support protective decision-making during nocturnal tornadoes and flash flood events in the Southeast. These findings have implications for the operational community, the public, and the scientific community living in geographic areas in the Southeast with high nocturnal hazard event frequencies.