Introduction
Rapid-onset hazards such as flash floods, tornadoes, wildfires, earthquakes, and tsunamis pose significant challenges for disaster risk reduction because they develop rapidly, leaving limited time for protective action. 1 -Reference Hannel, Ellis and First 4 Effective warning systems are critical for public safety because they enable individuals to seek shelter, evacuate, or undertake other protective actions before hazardous conditions occur.Reference Ashley and Ashley 5 However, warning reception may be particularly challenging during overnight hours when individuals are asleep and less likely to receive hazard information.Reference Ellis, Mason and Hurley 2 -Reference Hannel, Ellis and First 4 Fatalities from rapid-onset hazards, particularly flash floods and tornadoes, are disproportionately higher during overnight hours when much of the public is sleeping.Reference Strader, Ashley and Haberlie 6 , Reference Luo, Cong and Liang 7 Because receiving a warning is a prerequisite for protective action, missed or delayed warnings may substantially reduce opportunities to undertake life-saving actions.Reference First, Ellis and Held 3 , Reference Ashley and Ashley 5 , Reference Lindell and Perry 8
This study examines public confidence in receiving flash flood and tornado warnings during overnight hours among residents of the southeastern United States (N = 1,008), a region with elevated risk for overnight flash flood and tornado events.Reference Hannel, Ellis and First 4 , Reference Luo, Cong and Liang 7 Specifically, this brief report presents preliminary findings examining whether nighttime emergency preparedness plans, access to warning sources, perceptions of nocturnal hazard risk, and trust in weather information are associated with increased confidence in receiving overnight flash flood and tornado warnings. The following research questions guided the study:
RQ1: To what extent are individuals confident in their ability to receive flash flood and tornado warnings during overnight hours (12:00 AM-6:00 AM)?
RQ2: Are nighttime emergency plans, warning source reliance, perceptions of nocturnal hazard risk, trust in weather information, and other situational factors associated with increased confidence in overnight warning reception?
Methods
Data Collection Methods
This study employed a cross-sectional online survey to examine confidence in receiving flash flood and tornado warnings during overnight hours across 12 states in the U.S. Southeast. A total of 1,041 adult participants were recruited through a Qualtrics online survey panel. Quota sampling was used to approximate the regional distribution of state of residence, housing type, household income, and race. Although quotas improved the demographic composition of the sample, women and individuals with higher educational attainment were overrepresented relative to regional population estimates. Prior to analysis, the dataset was screened for accuracy, completeness, and consistency. Cases with more than 20% missing data or respondents residing outside the 12 target states were excluded, resulting in a final sample of 1,008 participants. Eligibility criteria included being 18 years or older and residing in the southeastern United States, a region at elevated risk for nocturnal tornadoes and flash floodsReference Hannel, Ellis and First 4 , Reference Luo, Cong and Liang 7 Participants completed the survey online via Qualtrics and were compensated through the survey panel. All participants provided informed consent, and study procedures were approved by the University of Tennessee Institutional Review Board.
Measures
The survey questions used to assess study variables are provided in the Supplementary Appendix.
Confidence in overnight warning reception was assessed separately for flash flood and tornado hazards by asking respondents how confident they were that they would receive a flash flood and tornado warning issued between 12:00 AM and 6:00 AM. Responses were measured on a 5-point Likert scale ranging from 1 (not at all confident) to 5 (extremely confident), with higher scores indicating greater confidence.
Warning sources were assessed by asking respondents how much they relied on multiple weather information sources during severe weather, including television, radio, Wireless Emergency Alerts, landline phone alerts, social media, outdoor warning sirens, weather radios, family or friends, schools, faith-based leaders, community organizations, and other sources. Responses ranged from 1 (not at all), 2 (somewhat), and 3 (a great deal). Responses were summed to create an overall warning source reliance index (M = 20.51; SD = 4.43), with higher scores indicating greater reliance on weather information received during severe weather.
Nocturnal flash flood and tornado risk perception were each measured using four items assessing perceived likelihood, potential property damage, potential injury to household members, and fear associated with nocturnal flash floods and tornadoes. Responses ranged from 1 (strongly disagree) to 5 (strongly agree), with higher scores indicating greater perceived nighttime risk.
Nighttime household emergency planning was assessed separately for flash flood and tornado hazards by asking respondents whether their household had developed and discussed a nighttime emergency plan that included instructions regarding what to do and where to go during a nighttime hazard event (0 = no, 1 = yes).
Trust in weather information was measured separately for flash flood and tornado hazards using a single item assessing agreement with the statement, “I trust weather forecasts on [hazard] threats when they are issued,” with responses ranging from 1 (strongly disagree) to 5 (strongly agree).
Prior hazard experience was assessed by asking respondents whether they had previously been under a nighttime flash flood warning or nighttime tornado warning (0 = no, 1 = yes).
Demographic characteristics included age, gender, race, household income, and mobile home residence.
Analyses
All analyses were conducted using SPSS version 29. Descriptive statistics were used to summarize respondent characteristics, confidence in overnight warning reception, warning source use, nighttime risk perception, trust in weather information, household emergency planning, and prior warning experience. To answer RQ1, responses of “very confident” or “extremely confident” were combined to calculate the proportion of respondents reporting high confidence in receiving overnight flash flood and tornado warnings. To address RQ2, separate ordinary least squares (OLS) regression models were estimated for predicting confidence in flash flood and tornado warning reception. Independent variables included the warning source reliance index, nocturnal hazard risk perception, trust in weather information, nighttime household emergency planning, prior warning experience, age, gender, household income, race, and mobile home residence. Standardized regression coefficients (β), significance tests, and model fit statistics were examined for each model. Education and state of residence were evaluated but excluded from the final models because they were not significantly associated with the outcomes or model fit.
Results
Descriptive statistics for the demographic variables are presented in Table 1. Respondents reported higher nighttime risk perception for tornadoes (M = 15.35, SD = 3.31) than for flash floods (M = 12.29, SD = 4.20). Trust in weather forecasts was also slightly higher for tornadoes (M = 4.14, SD = 0.78) than for flash floods (M = 3.99, SD = 0.79). A greater proportion of respondents reported having a household nighttime emergency plan for tornadoes (61.7%) compared with flash floods (36.1%). Prior warning experience was more common for tornadoes (87.3%) than for flash floods (65.3%). Less than half of respondents reported confidence in their ability to receive overnight flash flood warnings (47.8%), and a little more than half reported confidence in their ability to receive tornado warnings (51.5%) during overnight hours.
Participant demographic information
Table 1. Long description
Three columns report demographic category, number of respondents (n), and percentage. Gender: Female: 670 respondents (66.5%). Male: 312 respondents (31.0%). Missing: 26 respondents (2.6%). Racial and ethnic background: American Indian/Alaska Native: 11 respondents (1.1%). Asian/ Asian American: 16 respondents (1.6%). Black/ African American/ Afro-Caribbean: 135 respondents (13.74%). Hispanic/ Latino / Latina / Latinx: 27 respondents (2.7%). Native Hawaiian/ Other Pacific Islander: 4 respondents (0.4%). White/ not of Hispanic origin: 778 respondents (77.2%). Other: 13 respondents (1.3%). Missing: 24 respondents (2.4%). Education: Grade school: 1 respondents (0.1%). Some high school: 31 respondents (3.1%). High School/ GED: 230 respondents (22.8%). Some college, but no degree: 214 respondents (21.2%). College degree (AA, BA, BS, etc.): 373 respondents (37.0%). Advanced degree (MA, PhD, JD, etc.): 135 respondents (13.4%). Missing: 24 respondents (2.4%). Income: Less than $15,000: 85 respondents (8.4%). $15,000 to under $29,999: 144 respondents (14.3%). $30,000 to under $44,999: 160 respondents (15.9%). $45,000 to under $59,999: 152 respondents (15.1%). $60,000 to under $74,999: 116 respondents (11.5%). $75,000 to under $104,999: 142 respondents (14.1%). $105,000 or more: 172 respondents (17.1%). Missing: 37 respondents (3.7%). Housing structure: A mobile home: 90 respondents (8.9%). A one-family house, detached from other buildings: 706 respondents (70.0%). A one-family house, attached to other buildings: 69 respondents (6.8%). A building with apartments: 106 respondents (10.5%). A boat, RV, van, etc.: 13 respondents (1.3%). Missing: 24 respondents (2.4%). State: Alabama: 74 respondents (7.3%). Arkansas: 88 respondents (8.7%). Florida: 78 respondents (7.7%). Georgia: 75 respondents (7.4%). Kentucky: 112 respondents (11.1%). Louisiana: 65 respondents (6.4%). Mississippi: 100 respondents (9.9%). North Carolina: 78 respondents (7.7%). South Carolina: 76 respondents (7.5%). Tennessee: 81 respondents (8.0%). Virginia: 88 respondents (8.7%). West Virginia: 93 respondents (9.2%).
To address RQ2, separate OLS regression models were estimated to identify factors associated with confidence in receiving overnight flash flood and tornado warnings. Results found the flash flood model was statistically significant, F(10, 997) = 25.95, P < 0.001, explaining 24.8% of the variance in confidence (adjusted RReference Ellis, Mason and Hurley 2 = .238). Greater warning source reliance (β = .218, P < 0.001), greater nocturnal flash flood risk perception (β = .117, P < 0.001), having a nighttime household emergency plan (β = .210, P < 0.001), and trust in flash flood forecasts (β = .230, P < 0.001) were associated with greater confidence in receiving overnight flash flood warnings. Prior warning experience and demographic characteristics were not significantly associated with confidence (Table 2a).
Predictors of confidence in overnight flash flood warning reception
Table 2a. Long description
Five columns report independent variable, unstandardized coefficient, standard error, standardized coefficient (beta), and p value. N = 1,008. Warning sources (sum): unstandardized coefficient 0.093, standard error 0.014, standardized beta .218***, p < .001. Nocturnal flash flood risk perception: unstandardized coefficient 0.038, standard error 0.011, standardized beta .117***, p < .001. Nighttime flash flood emergency plan: unstandardized coefficient 0.833, standard error 0.130, standardized beta .210***, p < .001. Trust in flash flood forecasts: unstandardized coefficient 0.561, standard error 0.074, standardized beta .230***, p < .001. Prior flash flood warning experience: unstandardized coefficient 0.020, standard error 0.123, standardized beta .063, p .090. Gender: unstandardized coefficient −0.112, standard error 0.116, standardized beta −.028, p .337. Age: unstandardized coefficient 0.003, standard error 0.004, standardized beta .023, p .474. Race: unstandardized coefficient −0.046, standard error 0.328, standardized beta −.004, p .888. Household income: unstandardized coefficient 0.039, standard error 0.032, standardized beta .040, p .216. Mobile home residence: unstandardized coefficient −0.060, standard error 0.184, standardized beta −.010, p .745. Model summary: F(10, 997) = 25.95, p < .001; adjusted R squared = .238.
Note. N = 1,008. F(10, 997) = 25.95, P < 0.001; adjusted R² = .238.
The tornado model was also statistically significant, F(10, 997) = 17.88, P < 0.001, explaining 18.5% of the variance in confidence (adjusted R2 = .185). Greater warning source reliance (β = .268, P < 0.001), greater nocturnal tornado risk perception (β = .112, P < 0.001), having a nighttime household emergency plan (β = .155, P < 0.001), and trust in tornado forecasts (β = .144, P < 0.001) were associated with greater confidence in receiving overnight tornado warnings. Most demographic characteristics and prior warning experience were not significantly associated with confidence. However, mobile home residence was negatively associated with confidence (β = −.087, P = 0.006), with mobile home residents reporting lower confidence in receiving overnight tornado warnings than respondents residing in other housing types (Table 2b).
Predictors of confidence in overnight tornado warning reception
Table 2b. Long description
Five columns report independent variable, unstandardized coefficient, standard error, standardized coefficient (beta), and p value. N = 1,008. Warning sources (sum): unstandardized coefficient 0.107, standard error 0.013, standardized beta .268***, p < .001. Nocturnal tornado risk perception: unstandardized coefficient 0.065, standard error 0.019, standardized beta .112***, p < .001. Nighttime tornado emergency plan: unstandardized coefficient 0.565, standard error 0.114, standardized beta .155***, p < .001. Trust in tornado forecasts: unstandardized coefficient 0.323, standard error 0.075, standardized beta .144***, p < .001. Prior tornado warning experience: unstandardized coefficient 0.152, standard error 0.166, standardized beta .029, p .359. Gender: unstandardized coefficient −0.064, standard error 0.113, standardized beta −.037, p .573. Age: unstandardized coefficient 0.002, standard error 0.004, standardized beta .014, p .638. Race: unstandardized coefficient −0.140, standard error 0.320, standardized beta −.013, p .662. Household income: unstandardized coefficient 0.001, standard error 0.031, standardized beta .002, p .963. Mobile home residence: unstandardized coefficient −0.494, standard error 0.180, standardized beta −.087**, p .006. Model summary: F(10, 997) = 17.88, p < .001; adjusted R squared = .185.
Note. N = 1,008. F(10, 997) = 17.88, P < 0.001; adjusted R² = .185.
Discussion
Across both flash flood and tornado hazards, greater warning source reliance, higher nocturnal hazard risk perception, having a nighttime household emergency plan, and trust in weather information were consistently associated with greater confidence in receiving overnight warnings. These findings identify several modifiable cognitive and behavioral factors that may strengthen overnight warning reception and preparedness during rapid-onset hazard events.
First, warning source reliance emerged as one of the strongest predictors of confidence for both hazards. Public education and outreach efforts should emphasize maintaining multiple warning sources, particularly active alerting technologies capable of waking individuals during sleep (e.g., Wireless Emergency Alerts), rather than relying on passive sources (e.g., television and social media) that are not designed to awaken users.Reference Ellis, Mason and Hurley 2 Reliance on a single warning source may increase the likelihood of missed or delayed warnings during overnight hazard events.Reference Sadiq, Dougherty and Tyler 9
In addition, our findings suggest that framing tornado and flash flood hazard risk through a nocturnal lens may be effective for improving overnight warning reception. Psychometric studies of risk perception have found that characteristics including hazard likelihood and severity are important attributes of individuals’ judgments of risk.Reference Sadiq, Dougherty and Tyler 9 Emergency management and public health messaging that emphasizes the frequency and likelihood of occurrences of nocturnal hazards and the heightened severity and consequences (e.g., increased fatalities) may elevate public risk perception. In addition, having a nighttime household emergency plan was significantly associated with increased confidence in overnight warning reception. Developing and discussing plans that explicitly prepare for overnight scenarios, including how warnings will be received while household members are sleeping, how individuals will be alerted, and how protective actions will be carried out, may strengthen warning reception during overnight hours.Reference First, Ellis and Held 3 , Reference First, Ellis and Strader 10 In addition, trust in weather information was also associated with confidence across both hazards, highlighting the importance of maintaining public confidence and trust in warning systems and forecast information.Reference Lindell and Perry 8
Furthermore, while we did not see a significant difference in respondents’ confidence in warning reception for tornadoes versus flash floods during overnight hours (51% vs. 47%), results found that respondents reported higher risk perception for nocturnal tornadoes than for nocturnal flash floods, and fewer respondents indicated that they had developed and discussed a household nighttime emergency plan for a flash flood threat compared to a nighttime tornado threat. Prior research has suggested that the public assigns greater perception of risk to tornado hazards in comparison to flash flood hazards.Reference First, Ellis and Strader 10 Overall, these findings highlight the need for cross-hazard research that identifies the shared and hazard-specific behavioral, cognitive, and social factors influencing overnight warning reception and protective decision-making to inform more effective warning communication and preparedness strategies across extreme weather hazards.Reference Ashley and Ashley 5
Finally, most demographic characteristics were not associated with confidence in receiving overnight warnings. However, mobile home residents reported lower confidence in receiving overnight tornado warnings than respondents residing in other housing types. Given the elevated vulnerability of manufactured housing during tornadoes, these findings underscore the need for targeted communication strategies, improved access to overnight-capable warning technologies, and preparedness initiatives that support feasible protective actions and sheltering options for mobile home residents and other socially vulnerable populations.Reference First, Ellis and Strader 10
Several limitations should be acknowledged and provide opportunities for future research. First, this study relied on self-reported perceptions of confidence and was conducted within a single region of the United States. Confidence in warning reception may not correspond to actual warning receipt during overnight hazard events. Future research should examine objective warning receipt during actual overnight events and compare warning receipt with perceived confidence. In addition, much of the existing research on overnight warning reception has focused on nocturnal tornado hazards and U.S. populations.Reference Ellis, Mason and Hurley 2 , Reference First, Ellis and Held 3 , Reference Ashley and Ashley 5 , Reference Luo, Cong and Liang 7 Future research should examine nocturnal warning reception across a broader range of hazards, including floods, wildfires, tsunamis, and earthquakes, as well as across diverse international warning systems and populations.
Abbreviations
- M =
Mean
- OLS =
ordinary least squares
- RQ =
research question
- SD =
standard deviation
- SPSS =
Statistical Package for the Social Sciences
- U.S. =
United States
- WEA =
Wireless Emergency Alert
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/dmp.2026.10467.
Author contribution
All authors contributed to the study conception and design. Material preparation, data collection, and analysis were performed by the listed authors. The first draft of the report was written by the first author, and all authors commented on and edited the report. All authors read and approved the final report.
Funding statements
This study was funded by the National Oceanic and Atmospheric Administration, U.S.A. award (NA230AR4590367).
Competing interests
None.
IRB Statement
All study procedures were reviewed and approved by the University of Tennessee Institutional Review Board.
Open access
