Abstract

Tornado warnings are currently issued an average of 13 min in advance of a tornado and are based on a warn-on-detection paradigm. However, computer model improvements may allow for a new warning paradigm, warn-on-forecast, to be established in the future. This would mean that tornado warnings could be issued one to two hours in advance, prior to storm initiation. In anticipation of the technological innovation, this study inquires whether the warn-on-forecast paradigm for tornado warnings may be preferred by the public (i.e., individuals and households). The authors sample is drawn from visitors to the National Weather Center in Norman, Oklahoma. During the summer and fall of 2009, surveys were distributed to 320 participants to assess their understanding and perception of weather risks and preferred tornado warning lead time. Responses were analyzed according to several different parameters including age, region of residency, educational level, number of children, and prior tornado experience. A majority of the respondents answered many of the weather risk questions correctly. They seemed to be familiar with tornado seasons; however, they were unaware of the relative number of fatalities caused by tornadoes and several additional weather phenomena each year in the United States. The preferred lead time was 34.3 min according to average survey responses. This suggests that while the general public may currently prefer a longer average lead time than the present system offers, the preference does not extend to the 1–2-h time frame theoretically offered by the warn-on-forecast system. When asked what they would do if given a 1-h lead time, respondents reported that taking shelter was a lesser priority than when given a 15-min lead time, and fleeing the area became a slightly more popular alternative. A majority of respondents also reported the situation would feel less life threatening if given a 1-h lead time. These results suggest that how the public responds to longer lead times may be complex and situationally dependent, and further study must be conducted to ascertain the users for whom the longer lead times would carry the most value. These results form the basis of an informative stated-preference approach to predicting public response to long (>1 h) warning lead times, using public understanding of the risks posed by severe weather events to contextualize lead-time demand.

1. Introduction

Tornado prediction capabilities have advanced significantly over the past few decades. The first tornado forecasts (the term “warning” had not yet been used) were issued in 1948 (Doswell et al. 1999). By 1978, the average tornado warning lead time was 3 min and the probability of detection was 22%. Twenty years later there was a 65% probability of tornado detection with a 13-min lead time on average (Golden and Adams 2000); some of the improvement was attributed to the implementation of the network of Weather Surveillance Radar–1988 Doppler (WSR-88D) radars in the 1990s (Bieringer and Ray 1996).

Currently, warnings for tornadoes are based on observed (not predicted) weather information (i.e., warn-on-detection; Erickson and Brooks 2006). However, the National Weather Service (NWS) is considering the possibility of issuing warnings based on storm forecasts (i.e., warn-on-forecast), which would extend tornado warning lead times to as much as one to two hours. This new method of early tornado prediction is possible by using convection-resolving models, which assimilate Doppler radar data into mesoscale numerical models operating at the convective scale of resolution in time and space. This allows for a more reliable means of providing improved analysis and prediction of thunderstorms and their associated severe weather to the public. The National Oceanic and Atmospheric Administration (NOAA) expects this new technology to be available by the year 2020 (Stensrud et al. 2009).

Although the science behind this new warning technology is underway and rapidly advancing, the social implications of a longer lead time have not yet been rigorously evaluated. Meteorologists and social scientists do not have a clear understanding of how the general public would respond to a 1- to 2-h lead time. In fact, Simmons and Sutter (2008) question whether longer lead times would necessarily be advantageous for the general public. Controlling for sociodemographic variables, they found that as lead time increases, generally morbidity decreases in a linear fashion until locationally specific time thresholds are reached. Simmons and Sutter’s (2008) empirical investigation of tornado casualties showed that an optimal lead time is about 15 min, and lead times greater than 15 min did not reduce fatalities or injuries compared to a weather event with no warning (Simmons and Sutter 2008). Another study that focused on administrators of schools and nursing homes found that the ideal tornado warning lead time was no more than 30 min (Ewald and Guyer 2002). Additionally, lead times may decrease public confidence in the scientific community, especially if the longer lead time is associated with greater uncertainty about tornado path and intensity (Ewald and Guyer 2002; Doswell 1999). People may associate longer lead times with false alarms since warnings might not be immediately followed by a tornado. Given that the false alarm rate is over 75% and the public has little familiarity with longer lead times, this is a reasonable reaction (Simmons and Sutter 2006).

Despite the lack of knowledge pertaining to extremely long lead times, there has been considerable research pertaining to public response to warnings in the current system. Factors such as education, sex, ethnicity, age, prior tornado experience, and number of children are known to influence warning responses (Riad et al. 1999; Balluz et al. 2000; Sorensen 2000). An individual’s response to weather warnings may also be influenced by how dangerous or risky he or she perceives an event to be. This perception of risk is complex, dependent on a combination of objective criteria for evaluation (e.g., hazard climatology and warning forecast quality) and subjective factors, discussed in the next section. Someone who believes an event to be of little risk to them personally may be less compelled to take immediate action than someone who feels an event to be life threatening.

This study investigates 1) how the public perceives hazardous weather risks, particularly tornadoes, 2) the amount of lead time the public prefers for a tornado warning, and 3) the relationship between the public’s perception of tornado risks and preferred tornado warning lead time. This research evaluates the benefits and limitations of a possible future warning paradigm, warn-on-forecast, by seeing if the 1–2-h target lead time is demanded by the public, and identifying risk perception factors associated with the demand.1

2. Background

Severe weather events are risky based on their ability to cause harm to people and property, a premise about which a federal agency, the NWS, has devised its mission statement (NWS 2010). Assessing the probability of tornado risk exposure for different geographical areas (Brooks et al. 2003), as well as mortality/morbidity across geographical regions and vulnerable populations (Sims and Baumann 1972; Ashley 2007; Donner 2007; Simmons and Sutter 2006, 2008) has been an important part of this effort to understand where we may be able to reduce risk and vulnerability. These studies reveal patterns of societal impact from tornadoes, a few using aggregate statistical measures to attempt to identify a lead time that minimizes mortality and morbidity (Simmons and Sutter 2008; Ashley 2007). These studies uncover what demand for lead time might be like given the impact of tornadoes on different groups of people, but such an approach makes presumptions about the manner in which these risks are understood during the warning. When considering what societal response might be like with a longer lead time, this level of understanding regarding the judgments and preferences of individuals is insufficient to make predictions. The perception of risk, or the way the risks are framed by the individual, must be considered. In situations of uncertainty and potential loss, individuals increase their reliance on heuristics and biases (Kahneman and Tversky 1979). What seems optimal to an individual may appear unnecessary or irrational from a scientific risk assessment perspective. Theories of risk perception have not yet been tied explicitly to lead-time preference, but their foundational principles can be used to formulate ideas about how lead time can be understood in such a context.

Weather-related damage does not occur in a black box; rather, severe weather events (and tornadoes in particular) have many attributes that influence interpretation. Storms are large, fast, and menacing. Tornadoes may seem nearly impossible to imagine; in fact, tornadoes are among a select class of risks whose threat of mortality is significantly overestimated, and thus the risk is considered to be “dreaded” (Schmidt 2004). Some combination of the attributes of tornadoes leads to a unique way their risk is perceived, and thus, what sort of lead time may be demanded to mitigate the risk. In an attempt to provide a cogent framework to tie together the many attributes of several risks, Slovic (1987) developed a psychometric paradigm including the creation of a “hazard taxonomy” that is used to understand and predict responses to risk. This effort resulted in the generation of a risk factor map (Fig. 1) that depicts risk perception as a function of the degree to which a hazard is unknown or dreaded. Risks that map in the upper-right quadrant of the figure tend to be those for which public interventions are demanded, and where expert knowledge may be trusted for the purposes of risk reduction (Slovic 1987). Temporal qualities of risks, however, have only translated onto the factor map based on the delay time between an event and the occurrence of a risk (e.g., the time between exposure to a radiation source and the development of cancer), rather than time until impact, affecting risk perception in a short (<1 h) time frame.

Fig. 1.

Simplified version of Slovic’s (1987) risk factor map, depicting risk perception as a function of the degree to which a hazard is “unknown” or “dreaded.” The lower left quadrant depicts those hazards that are clearly understood, common everyday risks (e.g., elevators). The upper left depicts those that are less known and less risky (e.g., caffeine). The lower right depicts those whose risks are more known and more dreaded (e.g., nuclear weapons). The upper right depicts those that are not well known and more dreaded (e.g., tornadoes); people demand public intervention for these types of hazards. The lighter region represents the hypothesized zone in which the authors suspect tornadoes to fall.

Fig. 1.

Simplified version of Slovic’s (1987) risk factor map, depicting risk perception as a function of the degree to which a hazard is “unknown” or “dreaded.” The lower left quadrant depicts those hazards that are clearly understood, common everyday risks (e.g., elevators). The upper left depicts those that are less known and less risky (e.g., caffeine). The lower right depicts those whose risks are more known and more dreaded (e.g., nuclear weapons). The upper right depicts those that are not well known and more dreaded (e.g., tornadoes); people demand public intervention for these types of hazards. The lighter region represents the hypothesized zone in which the authors suspect tornadoes to fall.

With respect to a hazard warning lead time, the temporal aspects may have very different properties than merely increasing uncertainty due to delay of impact (as with cancer risk). In fact, a variety of competing effects are possible. An increase in lead time may increase the certainty individuals feel in their ability to respond, and make an event seem more observable, providing the opportunity for better risk management. Increasing lead time may then effectively move an individual from the upper right corner of the map back to the lower left, where they feel more certain and in control of the risk (an effective intervention). However, increasing lead time will also increase the spatial uncertainty associated with the impact of an event to an individual, and may make the situation seem much more dire (a longer lead time may imply that a greater amount of time may be required for individuals to seek safety). These effects may cap the longer lead time’s impact on bolstering individual self-efficacy during the event. Thus, a preferred lead time can potentially be framed as the equilibrium between these competing effects. Additionally, given the short-fuse nature of the hazard, there is some upper limit to lead time beyond which being warned would add no value due to an array of potential factors: loss of attention/interest in the risk, increase in opportunity cost for responding, and increased uncertainty regarding the existence or nature of the event.

The balance between these effects may be mediated by the knowledge the public has of tornado risk, because knowledge will dictate how observable and known the risk will seem to be. Consider lead time as the provision of a public good for the purpose of reducing risk, whereby dread is minimized at a stated lead-time preference point (moving left along the x axis of Slovic’s diagram) and uncertainty is minimized (moving down along the y axis), promoting the ability of individuals to respond. If an individual has less knowledge about tornado hazard risks before an event occurs, their starting point may be much more uncertain than someone who is knowledgeable regarding their risk, which may carry implications for a demand for lead time. As such, it is important to contextualize hazard knowledge alongside lead-time preferences.

3. Methodology

a. Survey

The survey included 34 questions and was composed of two parts, addressing two questions related by risk perception theory: 1) How accurate are the general public’s perceptions of weather risks? 2) What is the preferred tornado warning lead time for the general public? The public’s knowledge of general weather risks was tested by asking them to choose the correct response for fatality rates of five severe weather events (hurricanes, flooding, heat, tornadoes, and lightning). In this way, the position of tornadoes within the “risk ladder” could be identified. The focus then shifted to assessing the public’s knowledge of tornado risks, using a nine-question schedule. From there, respondents answered true/false questions pertaining to common tornado myths.

The second part of the survey elicited tornado-warning lead-time preferences for multiple scenarios. Respondents were asked to write down (i) the number of minutes they felt was appropriate for a tornado warning to be issued for the absolute minimum time required to get to shelter, (ii) the time needed to get necessary belongings and get to shelter, and (iii) the desired tornado warning lead time. The survey ended with a question asking if there can actually be too much lead time. The demographic and background component of the survey included questions about age, gender, ethnicity, the state in which they reside, number of children, prior experience being in a tornado, and if they had a designated tornado shelter.

b. Survey population and distribution methods

The survey was administered to visitors touring the National Weather Center (NWC) in Norman, Oklahoma. It was taken by 320 people at the start of their tour of the building. Tours usually lasted about an hour and were given one time per day, three to five times per week. The average size of the tour group was 15. Minors were not allowed to take the survey. The first 136 surveys were collected during the summer of 2009, while the remaining 184 were collected in the fall of 2009. This population sample included a broader demographic than in previous studies (e.g., Ewald and Guyer 2002). The participants represented all age groups over 18 years old with a wide range of education levels and residing in 34 states. They came from many different professional backgrounds and were not part of a specific institution. One main limitation of this study is that the results may be biased given that the survey was taken by people who may have some interest or knowledge of weather, as they were taking a tour of the NWC. Sampling groups outside of the NWC would contextualize the reliability of our results.

c. Data analysis and quality control

Simple descriptive statistics such as response rates, average or median responses, and difference testing were used to analyze the data. A handful of surveys were returned incomplete. Thus, the sample size changed from question to question, depending on which questions were skipped. The free response questions addressing the public’s actions and behaviors when given a 15-min versus 1-h lead time were the questions left blank the most often (about 22% of the sample did not respond to this section).

Another set of questions may have been misinterpreted by some participants. When asked about the minimum tornado warning lead time to take shelter, or the minimum lead time to gather any necessary belongs and take shelter, about 11% of the sample reported a higher lead time on the former question than the latter, which is illogical.

4. Results

a. Survey demographics and background

Of the 320 participants, 54% were female and 46% were male. About 92% were Caucasian. About 11% were between 18 and 25 years of age, 10% between 26 and 35, 16% between 36 and 45, 20% between 46 and 55, 14% between 56 and 65, and 30% over 65 years of age (a bias toward those older than 65 is acknowledged). Respondents were asked if they had an action plan if a tornado were to strike, as well as if they had a designated tornado shelter (which can be defined as any location the individual safely locates themselves during a tornado warning). While 70% of respondents stated that they have a tornado action plan, only approximately 53% of respondents said that they have a designated tornado shelter (Fig. 2).

Fig. 2.

Percent of respondents who answered yes/no to having a tornado action plan and a designated tornado shelter.

Fig. 2.

Percent of respondents who answered yes/no to having a tornado action plan and a designated tornado shelter.

b. General weather risk perceptions

Each respondent was asked to choose the number of people killed each year by tornadoes, lightning, flooding, heat, and hurricanes (Fig. 3). The “correct” responses were defined as the 10-yr average between 1998 and 2007 in the United States (NWS 2010). The public overestimated the number of deaths caused by lightning each year in the United States. Nearly 61% of respondents overestimated the total number of lightning fatalities, while the true fatality rate is on average 44 per year (23% of respondents chose this range). The public perceived the relative fatality risk of tornadoes, flooding, and heat accurately overall, choosing ranges that corresponded to correct fatality rates of 62, 74, and 170, respectively. In this study, respondents were able to select a range rather than provide their own estimate, which may have made the correct answer much easier to guess than in earlier studies. The public perceived hurricanes as being the least fatal, even though hurricanes produced the second most fatalities per year for the period considered after heat waves with an average of 117. The hurricane fatality rate response is skewed toward a low average response, which would be correct if outliers such as Hurricane Katrina were removed from the average. Approximately 66% of the respondents estimated the fatalities to be 100 or less each year.

Fig. 3.

Distribution of the number of people who chose each range of fatalities caused each year in the United States by (a) lightning, (b) tornadoes, (c) heat, (d) flooding, and (e) hurricanes (NWS 2010). Correct category for the period between 1998 and 2007 lightly shaded; number represents actual fatalities.

Fig. 3.

Distribution of the number of people who chose each range of fatalities caused each year in the United States by (a) lightning, (b) tornadoes, (c) heat, (d) flooding, and (e) hurricanes (NWS 2010). Correct category for the period between 1998 and 2007 lightly shaded; number represents actual fatalities.

The fatality responses were also analyzed by ranking how the public perceived the relative danger of each hazard, with 1 being the most fatal and 5 the least fatal. The rankings according to each age group are listed in Table 1, and the rankings according to educational level are listed in Table 2. The 18–25 and 26–35 age groups ranked the hazard fatality rates most accurately relative to the other age groups. Those older than 35 on average ranked hurricanes as being one of the least deadly events. If the participants based their rankings on fatality rates for a period of more than 10 years, the older groups may have actually ranked them more accurately given recent events such as Hurricane Katrina, which may have highly skewed the average 10-yr hurricane fatality rate.

Table 1.

Table of ranking of perceived fatalities (with 1 being viewed as most fatal) for five different severe weather events according to the different age groups. Accurate rankings were defined based on data from the National Weather Service (NWS 2010).

Table of ranking of perceived fatalities (with 1 being viewed as most fatal) for five different severe weather events according to the different age groups. Accurate rankings were defined based on data from the National Weather Service (NWS 2010).
Table of ranking of perceived fatalities (with 1 being viewed as most fatal) for five different severe weather events according to the different age groups. Accurate rankings were defined based on data from the National Weather Service (NWS 2010).
Table 2.

Table of ranking of perceived fatalities (with 1 being viewed as most fatal) for five different severe weather events according to highest level of education. Accurate rankings were defined based on data from the National Weather Service (NWS 2010).

Table of ranking of perceived fatalities (with 1 being viewed as most fatal) for five different severe weather events according to highest level of education. Accurate rankings were defined based on data from the National Weather Service (NWS 2010).
Table of ranking of perceived fatalities (with 1 being viewed as most fatal) for five different severe weather events according to highest level of education. Accurate rankings were defined based on data from the National Weather Service (NWS 2010).

When looking at the breakdown for education levels, it is clear once again that hurricanes were considerably underestimated across all educational categories. All age and educational categories reported heat as being the deadliest weather event, which is correct. The responses for tornadoes, lightning, and flooding, on the other hand, ranged across all possible values from 1 to 5.

c. Tornado knowledge

Participants were asked three questions pertaining to tornadoes. The first question asked them where they would take shelter in the case of a tornado, assuming that they were at home when the warning was issued. About 97% of respondents stated that they would take shelter in either a tornado shelter/basement or in an interior room of their house.

Participants were also asked about the number of tornadoes that hit the United States each year. Answers were sorted by the region in which each participant resided; regions were defined as depicted in Fig. 4. Those from the Midwest reported the closest average response to the correct response (~1000–1500 tornadoes, depending on the year), followed by the Southeast, while the Northeast was the farthest from the actual value. A possible reason for why the Midwest and Southeast reported most accurately could be due to the fact that individuals in this region experience tornadoes more frequently and with relatively high visibility due to terrain than those in some of the other regions. Although it appeared that all five regions underestimated the number of tornadoes that hit the United States each year, the lack of significance of the differences between responses (p > 0.05) prohibits any conclusions.

Fig. 4.

This map illustrates the five regions used for this study: Northeast, Southeast, Midwest, Southwest, and West. The dots depict the 34 states represented by the survey population.

Fig. 4.

This map illustrates the five regions used for this study: Northeast, Southeast, Midwest, Southwest, and West. The dots depict the 34 states represented by the survey population.

The next question tested the public’s knowledge of tornado seasons. They were asked to circle up to three months during which they felt tornadoes are most likely to occur where they live. The probabilities of their responses (again, calculated according to region; Fig. 4) were compared to the actual probability that a tornado would occur during each month in each region (Brooks 2010), and both were graphed on the same axis in order to assess how accurate the public’s responses were (Fig. 5). The maximum probabilities of having one or more days with a tornado within 25 miles of a point sometime during a month were used when calculating the probability that a tornado would occur in each region (Brooks 2010). With that said, the actual and stated probabilities cannot be directly related given that the respondents were not trying to estimate the likelihood of a tornado being within 25 miles of a point, but were only asked for the months with the greatest risk in their state; we simply compared the distributions for overall correspondence and thus the absolute percentages are not important. In general, the Northeast and Southwest responded most accurately overall, with the sample population choosing the three months during which there is the highest probability of a tornado actually occurring in each region. However, in general the participants in all of the regions performed well.

Fig. 5.

The actual tornado probabilities (maximum probabilities) for each month within each region compared with the percent of respondents that chose each month: (a) Northwest, (b) Southeast, (c) West, (d) Midwest, and (e) Southwest.

Fig. 5.

The actual tornado probabilities (maximum probabilities) for each month within each region compared with the percent of respondents that chose each month: (a) Northwest, (b) Southeast, (c) West, (d) Midwest, and (e) Southwest.

d. Common tornado myths

The survey included nine true/false questions pertaining to common tornado myths (see the  appendix for a complete list of the myths). The average correct response for all respondents was about 78%, or approximately seven out of the nine questions correct. Twelve percent of respondents answered every myth question correctly. The lowest score was from one participant who answered two out of the nine correctly. The average percentage of participants who answered a particular question correctly varied greatly, from 45% to 99% (Fig. 6). The most common believed myth was that the southwest corner of the basement is the best part of the basement in which to take shelter. Approximately 55% stated that this was true, which is not correct. On the other hand, there were three myths that nearly everyone answered correctly: it is possible for tornadoes to strike the same location twice, it is possible for tornadoes to cross water, and cities are not necessarily safe from tornadoes.

Fig. 6.

Percentages correct for each true/false tornado myth. The entire list of questions can be found in the  appendix.

Fig. 6.

Percentages correct for each true/false tornado myth. The entire list of questions can be found in the  appendix.

There were no statistically significant differences in correct responses between age groups or education levels for any of the myths. This can be attributed to the small sample size that resulted when the participants were divided demographically, limiting the possibility of finding statistical significance.

e. Tornado warning lead times

Tornado warning lead times were the focus of the second part of the survey. This section informs the second research question: What is the preferred tornado warning lead time for the general public? Participants were asked to write down the number of minutes that they felt was appropriate for three different situations. The average number of minutes stated for the minimum time needed to just take shelter was 10.2 min. The average number of minutes stated for the minimum time needed to gather any necessary belongings and take shelter was 14.4 min. The average time stated for the preferred tornado warning lead time was 34.3 min.

The differences in minutes stated for each of the situations varied by region and by age. Figure 7 shows that the participants from all five regions stated a similar average amount of time (in min) needed to take shelter, and to gather belongs and take shelter. However, the preferred lead time fluctuated depending on the region. On average, those from the Northeast and West preferred a desired lead time of approximately 41 min, which is about nine minutes more than the average desired lead time for those in the Southeast, the region where respondents stated the lowest preferred lead time. The other regions had lead-time preferences between those two extremes. Thus, location may be an influential factor in the public’s stated desired warning time. Additionally, desired lead time may be causally related to relative exposure to or experience with tornado warnings. As there are regional differences to tornado climatology, this relationship would show up as a difference in warning lead-time preferences.

Fig. 7.

Minimum time needed to get shelter, gather belongs and get shelter, and desired tornado-warning lead time according to region (p < 0.05).

Fig. 7.

Minimum time needed to get shelter, gather belongs and get shelter, and desired tornado-warning lead time according to region (p < 0.05).

Age also played an important determinant in warning lead times. Overall, the preferred lead time decreased with age. Figure 8 illustrates a decrease in the amount of time (in min) stated as the preferred lead time, with a difference of around 15 min between the youngest and oldest age groups. The fact that the preferred lead time was lower for the oldest group differs from the findings of Ewald and Guyer (2002). A possible explanation for this could be that older people feel more vulnerable to severe weather events, and thus are more prepared for when the event actually occurs.

Fig. 8.

Amount of time (in min) stated by the public (according to age group) for the preferred tornado lead time (p < 0.05 for 18–25 and 65+ age groups).

Fig. 8.

Amount of time (in min) stated by the public (according to age group) for the preferred tornado lead time (p < 0.05 for 18–25 and 65+ age groups).

The preferred lead-time preferences were also compared according to gender, prior tornado experience, having a designated shelter and tornado action plan, and number of children. Females reported a preferred lead time of nearly 5 minutes more than males, although this proved to be not statistically significant (Fig. 9). The participants who had either experienced a tornado themselves or had family or friends who had experienced a tornado reported lower preferred lead times compared to those who either witnessed one from afar or had no tornado experience at all (Fig. 9). One conclusion that may be drawn from these results is that those who have had more personal tornado experience have the knowledge base to accurately assess how much time they need to respond to tornadoes more so than those with less tornado experience, and are thus more prepared when one strikes again. Interestingly, 62 respondents stated that they experienced a tornado themselves, which is unlikely given that that would mean nearly 20% of the survey participants personally experienced a tornado. The participants may have considered experiencing a tornado warning to be the same as experiencing a tornado. Survey participants with no designated tornado shelter and no tornado action plan reported preferred lead times substantially higher than those who had a designated tornado shelter and tornado action plan (p < 0.01) (Fig. 9). One possible reason for this could be that having a designated shelter and action plan reduces the confusion and fear during an actual tornado warning, requiring fewer minutes to gather belongings and take shelter. Additionally, the actions required by those with no shelter may be more time-consuming than those who have a shelter. The final variable evaluated, whether number of children affects preferred lead time, resulted in no statistically significant conclusions.

Fig. 9.

Amount of preferred lead time reported by tornado experience levels (p < 0.01), whether or not they had a designated shelter (p < 0.01), and whether or not they had a tornado action plan (p = 0.01).

Fig. 9.

Amount of preferred lead time reported by tornado experience levels (p < 0.01), whether or not they had a designated shelter (p < 0.01), and whether or not they had a tornado action plan (p = 0.01).

f. Actions taken with a longer lead time

The participants were asked two free response questions pertaining to how they would act/behave given both a 15-min warning lead time and a 1-h warning lead time. Their responses were coded into seven categories, along with the percentage of participants who stated that category in their response (Table 3). A respondent often stated more than one action, and therefore the sum of percentages for each column total more than 100%. It is important to note how the percentages change from the 15-min lead time to the 1-h lead time. If given a 1-h lead time, approximately four times more people would flee, six times more people would gather/secure belongings, three times more people would obtain further information regarding the storm, and one-sixth fewer people would take shelter.

Table 3.

The percentage of respondents who said they would take the actions described in each of the seven categories for both the 15-min and 1-h lead-times.

The percentage of respondents who said they would take the actions described in each of the seven categories for both the 15-min and 1-h lead-times.
The percentage of respondents who said they would take the actions described in each of the seven categories for both the 15-min and 1-h lead-times.

A typical response for actions given a 15-min lead time was: pack belongings and then take shelter. A typical response for actions given a 1-h lead time was: pack belongings, gather family, listen to the radio/watch TV, and then take shelter. There were also several very interesting responses, some of which include questioning the possibility of a 1-h lead time, waiting to take action until the storm was really close, using Twitter, going about business as if a normal day, and opening the windows.

The results of the free response questions were also broken down according to three categories: gender, number of children, and age. Although no significant differences were found according to gender, there are some notable differences regarding the number of children someone has as well as their age. Those with no children were less likely to flee, obtain more information on the situation, or gather/call family and friends. In fact, if given a 1-h lead time, about 6% of respondents with no children would obtain more information, compared to approximately 15% for those with children. In addition, those with no children reported the lowest percentage to gather/call family (approximately 11%), which is logical given the decreased responsibility of having no children.

Respondents aged 26 to 35 stated taking shelter (93%) and gathering family (30%) as the two most important steps to take during a 15-min warning, which is logical given that this age group has a high likelihood of having children (about 58%). These percentages towered above the percentages of the other age groups, especially those under 25 and over 56, of whom only three percent reported gathering family. With the 1-h lead time, respondents aged 26 to 35 reported gathering more information about the situation (about 26% reported this action) and fleeing (about 26%). This represents the actions of a typical family: necessities first, then extras second, if time permits. Those under age 25, on the other hand, represent the typical young person. On the whole, these participants stated taking shelter as a lesser priority, while packing personal belongings become a more likely substitute (about 35%, while less than 20% for all other age groups). This age group was also more than twice as likely as any other age group to flee when a tornado warning is issued, which once again represents a typical response of someone with less experience. When given a 1-h lead time, those 65 and older were more likely than any other age group to take shelter (approximately 78%). Thus, demographics as well as personal situations, such as having a family, significantly influence how one behaves during a tornado warning, whether that be a 15-min or 1-h warning time.

A follow-up question to the free response questions asked if a longer tornado warning lead time would make the tornado situation any more or less life threatening. Over 44% of the respondents stated that a 1-h lead time would make the situation less life threatening, while approximately 47% reported no change, and about 8% stated it would make the situation more life threatening (Fig. 10).

Fig. 10.

Percentage of respondents who said a longer tornado warning lead time would make the situation more life threatening, less life threatening, or the amount of perceived danger would not change.

Fig. 10.

Percentage of respondents who said a longer tornado warning lead time would make the situation more life threatening, less life threatening, or the amount of perceived danger would not change.

g. Too much lead time

The survey participants were asked whether or not they feel that there can be too much lead time. Over 60% of participants stated there cannot be too much lead time, while the remaining 40% felt as if there can be too much lead time (Fig. 11). Of that 40%, nearly 60% of them reported that between one and two hours is too much warning time. These results are comparable to a study conducted by Ewald and Guyer (2002), which found that 37% of respondents thought that there could be too much warning time. A possible explanation for why a longer lead time could be too much might be because the individual feels the added time is inconveniencing the rhythm of their day. This extra time would not be as valuable if the individual waits until the last minute to seek shelter anyway.

Fig. 11.

Percentages of respondents who felt that there can/cannot be too much warning time. Of the nearly 40% who felt that there could be too much lead time, the amount of time they felt is too much is given along with the corresponding percentages.

Fig. 11.

Percentages of respondents who felt that there can/cannot be too much warning time. Of the nearly 40% who felt that there could be too much lead time, the amount of time they felt is too much is given along with the corresponding percentages.

5. Conclusions

Two of the central questions posed by this research included 1) identifying the public’s perception of weather risks and 2) revealing the preferred tornado warning lead time for the general public. Overall, the public had a relatively accurate perception of weather risks. They knew for the most part the appropriate value for annual fatality rates for most of the weather hazard categories. However, they underestimated the hurricane fatalities and overestimated the lightning fatalities. Neither age nor education level was a vital determinant in general weather event knowledge, with the exception that the younger age groups may have perceived the relative danger of the events slightly more accurately. When it came to tornado knowledge, the participants correctly perceived the fatality risks associated with tornadoes but they underestimated the average number of tornadoes per year. They had an excellent understanding of the months during which most tornadoes occur. The participants performed well on the tornado true/false questions, with an average of 7 out of 9 correct. It would be interesting to find any significant correlations between age or gender and responses to the tornado myth questions; however, this could only be done with a much larger sample.

When asked about the minimum time to gather belongings and take shelter, the average response was about 14.4 min. This estimate is significant given that it is comparable to the current lead time today. The preferred lead time was found to be 34.3 min. This indicates that the general public may not need the one to two hours that warn-on-forecast could provide. In fact, when asked if there can be too much lead time, nearly 40% of the respondents stated yes, with approximately 23% of respondents saying that 1–2 h is too much lead time. These results are comparable to results from other studies (Ewald and Guyer 2002). Thus, not only may some of the public not desire a longer lead time, but Simmons and Sutter (2008) also showed that a longer lead time may not necessarily improve fatality statistics. Therefore, a longer lead time may not be of the best interest to the general public, for reasons that could only be found by looking at this topic more in depth in a future study.

It is important to note that when asked about actions taken during warnings, fewer people would immediately gather their family and seek shelter when given a 1-h lead time compared to a 15-min lead-time. Further research is required in order to conclude whether this would actually be more dangerous than taking immediate shelter. More people also stated that they would flee/drive away from the area if given more time, which, according to Hammer and Schmidlin (2002) and Golden and Adams (2000), might be a safer response than staying at home in certain situations. Similarly, over 44% of the respondents stated that they would feel as if the situation was less life threatening if given a 1-h lead time compared to a 15-min lead time. Once again, it would take further research to figure out if there is a correlation between people taking fewer precautions, since they feel the situation is less life threatening, and fatality rates.

In conclusion, the research questions addressed in this study have an important connection beyond the scope of merely assessing personal risk and preferred lead time. Understanding how an individual perceives the risk from a severe weather event and responds to current warnings helps mold a framework for how lead times can be changed in the near future. If a longer lead time is to become standard practice in the future, then an entirely new paradigm for warning response would be imminent and data collected on current lead-time actions would become less useful. The authors recognize that this study is purely a preference-based study; however, it does provide a foundation for understanding actions during tornado warnings on which to build. The possible future tornado warning lead time of one to two hours that warn-on-forecast would provide may not be necessary for the general public, and may actually be more inconvenient to some individuals. The social perspective of warn-on-forecast is a key and evolving research area that will better define the very specific and critical needs posed by individuals.

Acknowledgments

The corresponding author would like to thank her mentors for all of their support, as well as Daphne LaDue for giving her the opportunity to participate in the Research Experience for Undergraduates Program at the National Weather Center in Norman, Oklahoma. This work is supported by the National Science Foundation under Grant ATM-0648566 and by the Engineering Research Centers Program of the National Science Foundation under NSF Award 0313747. Any opinions, findings, conclusions, or recommendations expressed in this material are those of the authors and do not necessarily reflect those of the National Science Foundation.

APPENDIX

True/False Tornado Myth Questions

The correct responses according to the authors are provided in parentheses.

  1. If a tornado is coming toward your house, you should open the windows. (F)

  2. The southwest corner of a basement is the safest location during passage of a tornado. (F)

  3. The northeast corner of a basement is the safest location during passage of a tornado. (F)

  4. Tornadoes, like lightning, never strike the same place twice. (F)

  5. Tornadoes can cross water. (T)

  6. If you’re driving, you should take shelter under a bridge during a tornado. (F)

  7. Areas near mountains are safe from tornadoes. (F)

  8. Areas near populated cities are safe from tornadoes. (F)

  9. Mobile home parks are more likely to be hit by a tornado. (F)

REFERENCES

REFERENCES
Ashley
,
W.
,
2007
:
Spatial and temporal analysis of tornado fatalities in the United States: 1880–2005
.
Wea. Forecasting
,
22
,
1214
1228
.
Balluz
,
L.
,
L.
Schieve
,
T.
Holmes
,
S.
Kiezak
, and
J.
Malilay
,
2000
:
Predictors for people’s response to a tornado warning, Arkansas, March 1, 1997
.
Disasters
,
24
,
71
77
.
Bieringer
,
P.
, and
P. S.
Ray
,
1996
:
A comparison of tornado warning lead times with and without NEXRAD Doppler radar
.
Wea. Forecasting
,
11
,
47
52
.
Brooks
,
H. E.
, cited
2010
:
Severe thunderstorm climatology
.
[Available online at http://www.nssl.noaa.gov/hazard/.]
Brooks
,
H. E.
,
C. A.
Doswell
III
, and
M. P.
Kay
,
2003
:
Climatological estimates of local daily tornado probability for the United States
.
Wea. Forecasting
,
18
,
626
640
.
Donner
,
W. R.
,
2007
:
The political ecology of disaster: An analysis of factors influencing U.S. tornado fatalities and injuries, 1998–2000
.
Demography
,
44
,
669
685
.
Doswell
,
C. A.
, cited
1999
:
Are warning lead times the most important issue in tornado events?
WeatherZine, Center for Science and Technology Policy Research, Boulder, CO, No. 17. [Available online at http://sciencepolicy.colorado.edu/zine/archives/1-29/17/editorial.html.]
Doswell
,
C. A.
,
A. R.
Moller
, and
H. E.
Brooks
,
1999
:
Storm spotting and public awareness since the first tornado forecasts of 1948
.
Wea. Forecasting
,
14
,
544
557
.
Erickson
,
S.
, and
H.
Brooks
,
2006
:
Lead time and time under tornado warnings: 1986–2004
.
Extended Abstracts, 23rd Conf. on Severe Local Storms, St. Louis, MO, Amer. Meteor. Soc., 11.5. [Available online at http://ams.confex.com/ams/23SLS/techprogram/paper_115194.htm.]
Ewald
,
R.
, and
G. L.
Guyer
,
2002
:
The ideal lead time for tornado warnings—A look from the customer’s perspective
.
Extended Abstracts, 21st Conf. on Severe Local Storms, San Antonio TX, Amer. Meteor. Soc., 7.4. [Available online at http://ams.confex.com/ams/SLS_WAF_NWP/techprogram/paper_47592.htm.]
Golden
,
J. H.
, and
C. R.
Adams
,
2000
:
The tornado problem: Forecast, warning, and response
.
Nat. Hazards Rev.
,
1
(
2
),
107
118
.
Hammer
,
B.
, and
T. W.
Schmidlin
,
2002
:
Response to warnings during the 3 May 1999 Oklahoma City tornado: Reasons and relative injury rates
.
Wea. Forecasting
,
17
,
577
581
.
Kahneman
,
D.
, and
A.
Tversky
,
1979
:
Prospect theory: An analysis of decision under risk
.
Econometrica
,
47
,
263
291
.
NWS
, cited
2010
:
Weather fatality, injury, and damage statistics
.
Riad
,
J. K.
,
F. H.
Norris
, and
R. B.
Ruback
,
1999
:
Predicting evacuation in two major disasters: Risk perception, social influence, and access to resources
.
J. Appl. Soc. Psychol.
,
29
,
918
934
.
Schmidt
,
M.
, cited
2004
:
Investigating risk perception: A short introduction
.
Simmons
,
K. M.
, and
D.
Sutter
,
2006
:
Improvements in tornado warnings and tornado casualties
.
Int. J. Mass Emerg. Disasters
,
23
,
351
369
.
Simmons
,
K. M.
, and
D.
Sutter
,
2008
:
Tornado warnings, lead times, and tornado casualties: An empirical investigation
.
Wea. Forecasting
,
23
,
246
258
.
Sims
,
J. H.
, and
D. D.
Baumann
,
1972
:
The tornado threat: Coping styles of the north and south
.
Science
,
176
,
1386
1392
.
Slovic
,
P.
,
1987
:
Perception of risk
.
Science
,
236
,
280
285
.
Sorensen
,
J. H.
,
2000
:
Hazard warning systems: Review of 20 years of progress
.
Nat. Hazards Rev.
,
1
,
119
125
.
Stensrud
,
D. J.
, and
Coauthors
,
2009
:
Convective-scale warn-on-forecast: A vision for 2020
.
Bull. Amer. Meteor. Soc.
,
90
,
1487
1499
.

Footnotes

+

Current affiliation: Department of Geography and Environmental Sustainability, University of Oklahoma, Norman, Oklahoma.

&

Current affiliation: Oklahoma Climatological Survey, Norman, Oklahoma.

This article is included in the Tornado Warning, Preparedness, and Impacts Special Collection.

1

The aim of the study is to approximate a random sample of the U.S. public, therefore when this paper discusses the “public,” it refers to individuals, not businesses or groups.