1 Introduction

Universities have been affected by numerous disasters in the last decade (Federal Emergency Management Agency 2003). Recent research suggests that while universities are paying increasing attention to crisis preparation in the wake of high profile events such as the Virginia Tech massacre, they might not be able to handle such an event (June 2007). Some universities do have a published emergency management plan online, especially for physical disasters, but evidence suggests that many colleges and universities may not be fully prepared for a range of events (Mitroff et al. 2006). A survey of 22 universities in Canada found that less than half (45%) indicated their institution was prepared to deal with the impacts of a disaster (Friesen and Bell 2006). Similarly, a 2004 survey of the United States colleges and universities found that schools were generally only prepared for the types of events they had experienced in the past (Mitroff et al. 2006). Focusing on economic growth and expansion of facilities, universities have not considered their vulnerabilities to hazards until recently even though the cost of natural disasters to universities has been high (Yemaiel 2006). This lack of readiness can be extended to tornado preparedness, if school administrators adopt an “it can’t happen here” argument (Schultz 2007, 38).

One method many colleges and universities have recently added as part of an emergency preparedness strategy is the adoption of an emergency alert messaging system. There have been publicized examples of tests of these alert systems on college campuses, some successful, some not. Reports include a malfunction of one college’s emergency alert system after fatal shooting on campus in which some messages were never received (Young 2007), successful delivery during an earthquake in California (Leggiere 2008) as well as numerous descriptions of routine tests publicized on university websites. Because the technology has only been adopted recently by most universities, there have been few published studies about its use during an actual emergency. This paper presents a case study of a large southeastern university’s response to a tornado near-miss. The results of a survey given to the campus community describe the campus response to this event as well as the success of the emergency alert system in delivering the warning messages quickly and to a large percentage of the sample.

In January 2008, a long-lived supercell with a history of producing tornado damage moved over the Mississippi State University campus just outside of Starkville in east-central Mississippi (see Fig. 1). Mississippi State is a land grant university with a main-campus student, faculty and staff population of ~20,000. During this event, the university activated its incident command staff called the Crisis Action Team (henceforth referred to as CAT) and utilized its emergency alert messaging system called Maroon Alert to announce the suspension of classes and the need to take shelter during a tornado warning. Maroon Alert messages are sent via cell phone, instant message and email, although students and faculty must register their cell phone number with the university or download an instant messenger client to receive the messages. A campus-wide survey examined the success of the emergency notification used during this event, determines the campus-wide response to the tornado threat and provides information about the use of such communication methods in future events.

Fig. 1
figure 1

Map of tornado track (drawn in red) and a summary of watch and warning messages. Tornado times are beginning times listed on NWS local storm reports. Messages are shortened and are not directly quoted except where indicated

Full size image

The reason for conducting the survey was a perceived problem among some university administrators. This problem was a possible disconnect between desired response and actual response, including number of people who left campus during severe weather rather than seeking a safe place on campus. Some also believed that portions of the messages were not clear. This study had three specific goals, which were to

  1. 1.

    determine if and how the university community received tornado warning and class suspension messages

  2. 2.

    determine whether the following terms used in the messages were understood

    1. (a)

      Tornado watch

    2. (b)

      Tornado warning

    3. (c)

      Shelter in place

  3. 3.

    determine what people did upon receipt of these messages.

To meet these goals, an online survey questionnaire was administered to students and university employees on the main campus.

2 Event background

The event occurred on January 10, 2008, which was the second day of classes of the spring semester. Severe weather had been forecast and a tornado watch was first issued by the Storm Prediction Center (SPC) at 0745 CST for the sample area (See Fig. 1). The university CAT was activated during the morning hours and began monitoring weather conditions. Soon after noon the CAT made an announcement via its Maroon Alert emergency alert messaging system to suspend classes from 1230 to 1530 CST and instructed students to “shelter in place” during that time frame. Students, faculty and staff were also instructed to “seek shelter on the bottom floor of your building and stay away from windows and doors.”

Two tornado warnings were issued by the National Weather Service (NWS) during the time classes were suspended. These warnings were both issued for the same long-lived storm which formed over 100 miles to the southwest of campus and went on to produce an EF3 (Enhanced Fujita) tornado that badly damaged a school in Caledonia, Mississippi about 30 miles to the northeast of campus (Bryant 2008). Injuries, as well as EF3 damage were reported on each of the three tornado touchdown tracks shown in Fig. 1. The NWS warnings contained language similar to the Maroon Alert messages regarding the safest location to seek shelter and the immediacy of the required action.

After the first tornado warning was issued, the CAT implemented their plan to instruct resident-students to take shelter in hallways and safe areas within the residence halls. Sirens were sounded and a loudspeaker announcement was made from a campus building tower. Additionally, two emergency messages were sent via the Maroon Alert emergency alert messaging system announcing the proximity of the tornado to campus and instructing all to seek shelter immediately.

Rain intensified over campus during the time of the warnings, but the tornado did not touch down on campus and there were no reports of damage on campus. A final message announced that the threat of severe weather had passed and classes would resume as scheduled. Even though there were no reports of damage on campus, elsewhere the system caused wind, hail and tornado damage. Most notable in the local area was heavy damage to a school gymnasium located about 30 miles away.

3 Literature

There is a good deal of literature regarding how individuals hear tornado warnings and how they respond to them. Fewer studies examine whether individuals can define a tornado watch and warning through open-ended questions. Few, if any, academic studies have been published on response to a tornado warning on a college campus, at least since cell phones and the internet have become commonplace. As recently as 2006, Mitroff et al. (2006, 62) reported that “virtually no national research” had been conducted on colleges’ preparedness for crises. Since then, Collins et al. (2008) conducted a review of two colleges’ response to Hurricane Katrina, including damage incurred, and how the colleges planned to move forward to become more disaster resistant. Also, a special issue of International Journal of Mass Emergencies and Disasters focused on the Disaster Resistant University, but none of the articles presented a case study of the university-wide response to any particular event. Articles focused on how one university became more prepared for disasters through risk assessment (Human et al. 2006), how a department utilized its skills to assist the emergency operations center (Curtis et al. 2006), and how well prepared Canadian universities (and municipalities) are for disasters (Friesen and Bell 2006).

The current study will be situated in the literature, as it applies to all tornado warnings, even though it is different from many of the other studies about response during a tornado or the communication of the tornado threat in that the population of interest was a university community with its own system of emergency message delivery. Other differences may also exist, which make the university a unique focus of study, including a diverse population, structured hierarchy, and the role of the university to ensure the well being of the students (Friesen and Bell 2006). One expected difference was that television would not be as an important factor in delivering the tornado warning because the event occurred during the day on a weekday, when many students, faculty and staff would have been away from their televisions.

Television has been shown to be a primary source of warning information in several previous studies (Schmidlin et al. 2009; Sherman-Morris 2005; Hammer and Schmidlin 2002; Paul et al. 2003; Legates and Biddle 1999). Among residents of Moore, Oklahoma who had been through tornadoes in both 1999 and 2003, fewer survey respondents heard the tornado warnings on television and based their actions on television in 2003 than in 1999 (Comstock and Mallonee 2005). The difference may have been in part due to time of day. The 2003 tornado struck at 5 p.m., when fewer respondents reported being at home. Although only marginally higher than television, hearing the siren was more important than any other method in 2003 in motivating individuals to take action (Comstock and Mallonee 2005). Sirens have been cited repeatedly as primary or confirmatory warning devices (Paul et al. 2003; Balluz et al. 2000; Legates and Biddle 1999).

The percentage of residents of Moore, OK who received a tornado warning via phone or pager did not increase from 1999 to 2003, and that percentage was not very large when compared to more traditional methods such as television and sirens (Comstock and Mallonee 2005). The percentage hearing the warning over the phone (not specified whether land-line or cell phone) was similar in Schmidlin et al. (2009) at 9%. Although internet usage, including email and instant messaging surpassed television viewing as a favorite pastime (Hein and McClellan 2005), there are no clear data on the influence of text or instant messaging for distributing weather warning information. AccuWeather and other private weather companies offer specific weather content for mobile phones, and many local television stations are providing more weather content specifically for mobile devices (Whitney 2006) lending support to the idea that this is a method of disseminating weather information that is on the rise. It was expected that one of the university’s emergency alert delivery methods (text message via cell phone, instant message or email) would be the most significant way people received the tornado messages, in addition to tornado sirens simply because most of the sample would have been at school and away from televisions during the event. Many commonalities should exist between this sample of the university community and other samples with a broader focus. Education level may play a role as it has in other studies (e.g., Blanchard-Boehm and Cook 2004; De Man and Simpson-Housley 1987), although the education level of the university community is somewhat higher that the general population. Having a high school diploma has been shown to make one more likely to take shelter during a tornado (Schmidlin et al. 2009; Balluz et al. 2000) or to prepare for future events (Blanchard-Boehm and Cook 2004). Also, of prime importance is whether people heard and understood the warnings (Doswell et al. 1999). Knowing the difference between a watch and warning may predict shelter-taking during a tornado warning (Liu et al. 1996). A number of studies report whether respondents know or understand the difference between a watch and a warning, but few explain how this knowledge was measured. There is a difference between being able to pick the correct answer from among a set of choices and providing a correct definition. Balluz et al. (2000) and Liu et al. (1996) do not make it clear how this variable was measured. It was not evident in Legates and Biddle (1999) how they judged knowledge, but personal communication with one of the authors revealed it was through open-ended questions (Biddle 2008, personal communication). The percentage of respondents who understood a watch and a warning in Legates and Biddle (1999) was quite high (90%). An even higher percentage knew the difference between a watch and a warning (96%) in Balluz et al. (2000). More respondents (21–30%) could not correctly define a watch and a warning in Mitchem (2003), and a common mistake was to switch the two. When asked the open-ended question “what is the difference between a watch and a warning?” only 58% of survey respondents from three states could correctly do this (Powell and O’Hair 2008). In that study, there was a significant positive influence of both education and age on whether one could correctly answer this question. Besides determining through open-ended questions whether respondents can correctly define a tornado watch and warning, the current study also sought the same information on the phrase “shelter in place,” which was used in university emergency alert messages and was later criticized by members of the university community for being confusing. Finally, this study sought to determine what respondents did upon hearing a series of warning messages. It was slightly different from previous studies, however, due to the nature of the messages that were delivered. The first message did alert the university community to the possibility of severe weather later in the day. It was followed closely by a class suspension message and then two emergency messages warning of a tornado near campus. These messages were separate from the watch statement issued in the morning by the SPC and two warnings issued by the NWS.

There are some inconsistencies in the percentage of respondents who take shelter following a tornado warning. The majority of respondents in most studies do appear to take shelter. Ninety percent sought a safer location upon hearing the warning in Paul et al. (2003). About half of another study fled their homes altogether, and most of those who remained in their homes took shelter somewhere inside (Hammer and Schmidlin 2002). In another study, 46% of respondents did not seek shelter (Mitchem 2003). Approximately three quarters of respondents reported being in a safe location when two separate tornadoes struck Moore, OK (Comstock and Mallonee 2005). However, it is also common for people to wait for confirmation of risk before doing anything. A plurality of respondents to a survey about the Oklahoma City tornado waited more than 30 min before seeking shelter (Hammer and Schmidlin 2002). This delay may be explained in part by the tendency for people to seek other sources to confirm a warning (Mileti and Darlington 1997). In Hammer and Schmidlin (2002) respondents confirmed the tornado risk via multiple sources, but especially by telephone.

Differences between employees and students are likely to exist in actual response to the tornado alert messages. Students should be more likely to leave the campus because classes were suspended, while employees would not have been released. The fact that classes were only suspended temporarily and not canceled for the rest of the day may also have an impact on student behavior.

4 Methodology

A survey was created and placed on a university website. A link to the survey was placed on the university’s sign-on portal and all faculty, staff and students located on the main campus were emailed a request to participate in the survey. The chance to win one of three gift cards worth either $50 or $100 was offered as an incentive. One week later, the participants were sent another email as a reminder to take the survey. The survey consisted of 30 questions, mostly measuring how respondents heard about the severe weather and what they did immediately following receiving messages about the severe weather. Additional questions asked about use of weather information, satisfaction with the university warning communication, knowledge of the proper actions to take and about how well respondents thought others reacted to the severe weather.

5 Results and discussion

5.1 Sample characteristics

Despite offering an incentive and sending a reminder email, response rates were average to low—though comparable to other studies. Of the 3,818 employees who were sent an email asking them to take the survey, 970 employees completed one. Of the 14,634 students who were emailed, 1,959 returned the survey, but 8 had to be discarded because 5 respondents reported being 17 when asked their age, and 3 others were duplicates. This represents an overall response rate of 15.8%, but the response rate was higher for employees than students (25.4% of employees vs. 13.3% of students).

A majority of both groups was female (62.9% of employees and 56.7% of students). The average age of the students was 22.4, while the average age of the employees was 43.2 (see Table 1). Over half of the employees had advanced degrees compared to the most frequent student response of ‘some college,’ both of which might be expected. The percentage of employees who identified themselves as teaching faculty (33.6%) was similar to the percentage listed in the university’s institutional research statistics. The majority of the sample (85%) was informed of the possibility of severe weather before being provided immediate cues such as when it began to rain or when they heard the siren (see Table 2). Additional characteristics regarding use of information are discussed in the following paragraphs.

Table 1 Sample characteristics
Full size table
Table 2 Initial information about the possibility of severe weather
Full size table

5.2 Information sources

Employees were initially more informed about the possibility of severe weather. The first question asked respondents when they became aware that severe weather was possible (Table 2). The largest percentage of employees (40%) found out before the tornado watch was even issued, while most students heard about the possibility of severe weather from the university alert message (47.5%). Roughly 8% of both samples did not know about the possibility of severe weather until they heard the siren, and less than 1% never knew about it. Cell phones were very important in disseminating this and other information during this event. The highest frequency of both employee and student respondents reported receiving the messages that classes were suspended due to the possibility of severe weather, and that a tornado was nearing campus through cell phones (Table 3). Regarding other communication methods, instant messaging was preferred and used more by employees than students. Students heard about the tornado warning from local television more than any other source (38.4%), but this source was ranked fourth for employees (18.7%). More important sources of the tornado warning for employees were a weather text messaging service on a cell phone, the internet and communication with another person, which was ranked the highest for employees (35.9%). Less than 6% of employees and only 3% of students received the NWS tornado warnings through NOAA weather radio.

Table 3 Information sources used by students and employees during the tornado event
Full size table

Talking to others was important in several other situations besides providing information about the tornado warning. Approximately 20% of the respondents heard the first message about classes being suspended from another person (Table 3). It was also cited as the second-most effective influence on respondents’ decision to take shelter (26%) after hearing the tornado siren (27.2%, not in table). Approximately 65% of the respondents also reported talking to others about the warnings after they heard them (Not in Table). Personal communication was also an important way of letting people know the severe weather had passed, and was similar to the percentage who heard this via cell phone (~30% each, not in Table). The results of this survey indicate that the majority of respondents (66.2%) received the emergency alert message suspending classes and encouraging the campus community to shelter in place within 15 min after it was issued (Not in Table).

Respondents were also asked which method would be the best for communicating emergency information in the future. Students overwhelmingly chose cell phone messaging, while this was chosen by only a small majority of employees (Table 4). Almost 20% more employees than students preferred instant messaging. More employees were also signed up for this service than students (52 vs. 8% of students, not in Table). Differences in the use of the instant messaging program may be explained by the fact that many employees have the program installed as part of their network software, whereas students must download the program in order to receive the instant messages. More importantly, employees may be more likely to be near a computer to receive these messages.

Table 4 How respondents would like to be informed of a tornado or other emergency in the future
Full size table

When asked to choose the best way to provide warning information while they are teaching a class, just over half (52%) of teaching faculty said that text messaging via cell phone was the best way to reach them (Table 4). Approximately a third (35.6%) chose ‘other’—i.e., not via cell phone, email or instant message. Of the respondents who wrote comments following this question, just under half freely suggested that some designated person should deliver the warning information to the classrooms, further emphasizing the importance of interpersonal communication. More respondents suggested this than those who suggested sirens or a PA system. Several respondents raised the concern that having a cell phone with them or turned on during class sends the wrong message to students regarding the university cell phone policy. A few respondents stated that they do not bring a phone to class or even own one.

Students and employees were similar in the number who used weather media in the 24 h before the tornado, but a greater percentage of employees used local television (56.2%) and local radio (17.7%) than students (Fig. 2). Students were more likely to have used The Weather Channel (45%) than employees (36.7%). This use of local media may have explained why more employees than students knew about the possibility of severe weather before receiving the first emergency alert message. Communication with friends, relatives and co-workers was the highest ranked source of weather information for students (47.5%) and the second-highest for employees (42%).

Fig. 2
figure 2

Sources of weather information used in the 24 h leading up to the tornado

Full size image

5.3 Knowledge and understanding

Both students and employees rated themselves well (on a 5-point scale) regarding their knowledge of what actions to take during the severe weather event. Employees rated their knowledge higher on average than students (4.48 vs. 4.30), a significant difference (p = .000). Employees also rated highly the actions taken by their co-workers and also felt they understood their own responsibilities during the severe weather. As a check on whether employees knew what to do if they were teaching during a warning, they were given choices regarding what action they would take. Of the 312 employees who responded, 92.3% stated that they would “stop teaching and instruct [their] class to go to a particular place in the building” if they were teaching during a tornado warning. This choice was selected before four other specified responses including “nothing—continue with class,” “stop teaching and dismiss class,” “continue teaching until someone came to tell you where to go” and “I don’t know.” It was difficult to design this question without one response that stood out as most socially desirable.

Besides the questions asking students and faculty to self-evaluate their response and the response of others in the university community, both groups were also asked to define three terms through free response questions. These terms were tornado watch, tornado warning and shelter in place. Each of these definitions was graded on a scale of 0–5, where 0 was an inappropriate or irrelevant response and 5 was a correct and complete response. Shelter in place was the phrase with the highest score (4.16 out of 5), indicating respondents had the greatest knowledge of what that phrase meant. Tornado watch was scored the lowest (3.39). The average score for tornado warning was 3.79. It is interesting that shelter in place was the phrase with the highest score because the phrase was criticized as being confusing. Possible explanations are discussed below. The average score for each term was significantly higher (at p = .000) for employees than students. Regression analysis indicated that status as a student or employee is the only variable that can significantly explain (p = .000) the variance in scores for all three terms. Age (p = .04) and how frequently one uses weather information (p = .000) significantly predicted tornado watch. Age also significantly predicted shelter in place (p = .05). When the three variables were grouped together in one model, the variance explained ranged only from 1.4% for shelter in place to 4.1% for tornado watch.

The amount of variance explained is quite small to be of much practical significance, but a random sample of 10% of the respondents demonstrated significant correlations among several of the variables. Defining tornado watch correctly was related to age, status and use of weather information. Tornado warning was related to age and status, but defining shelter in place correctly was no longer related to any of these variables. As might be expected from the sample demographics, age and status were highly correlated with each other, where employees were older than students.

The most common score was 4 out of 5 for the definition of the term tornado watch. The most common definition was a statement that conditions were favorable for tornadoes. Some definitions included a similar statement as well as that individuals should be prepared to take shelter or be alert for a warning to be issued. Such correct and complete statements were scored the highest. Another common response was less specific about the likelihood of tornadoes. For example, some responses included “a tornado can occur,” and “tornadoes might happen.” Without the ability to follow-up with the respondent on this type of response, it is difficult to determine whether they understand the meaning of a tornado watch from these definitions. Nearly 10% of the definitions could be described as confusing a tornado watch and warning. The following definition of tornado watch, “a tornado was spotted close by and everyone needs to be aware and take cover if necessary” is an example.

Approximately 80% of respondents could correctly define a tornado warning. Another 10% of definitions were vague, underestimated the threat or included both correct and incorrect information. One vague response stated, “it’s possible that a tornado may come through your town.” Another response mentioned a “high risk of tornadoes.” While these responses are not necessarily incorrect, they are vague. The frequent use of the word “possible” to describe tornadoes in both watch and warning definitions is interesting. Besides the possibility that the respondents simply did not choose the best wording in their definitions, it also illustrates the potentially confusing nature of the word “possible” with regard to tornadoes.

The term “shelter in place” was criticized as being misunderstood, but even among students, the most frequent score for the definition of shelter in place was 5 out of 5 (42.3%). Some students were indignant about being asked such an easy question. One individual’s response did support initial anecdotal evidence that the term was misunderstood however. This student response stated:

I at first thought ‘shelter in place’ refered [sic] to a shelter that the university had established for those who lived off campus. Now I have learned that the word shelter was really meant to be a verb advising us to take shelter where we were.

Another student response expressed similar confusion:

This was confusing to me—specifically the ‘in place’ phrase. To me it implied to not leave your office or classroom. This was my understanding sense it did not say to move to a safe area of your building. My inclination was to litereally [sic] remain in the ‘place’ where I received the message. Others felt it just meant not to go outside. Better wording is needed for future messages.

Employee responses, while for the most part correct, expressed similar confusion:

I took it to mean stay where you are and not leave the building. But others in the area said it meant to go take cover in the basement. This phrasing was a little confusing.

You shouldn’t leave the building you are in. This seems ambiguous if you are in the car (like I was). Was it better for me to stop at the nearest place, or get to where I was going? Campus radio wasn’t very helpful about where the tornado had been spotted.

The wording was ambiguous: It seemed to me the wording was ‘Shelters are in place’ (i.e., have been established), not to seek shelter at your current location. This belief seemed to be shared by those I was with.

As the first example indicated, some of the respondents learned the correct definition of shelter in place after the event was over. This is one possible explanation why the score for this term’s definition was the highest. A person hearing the term shelter in place for the first time may have been more likely to seek information about what it meant. Most people have heard the terms tornado watch and tornado warning before and may as a result feel confident that they know what the terms mean (even if their understanding is not correct). Another possible reason why its score was higher than tornado watch or tornado warning is that the two were scored by slightly different criteria. While a score of 5 points was assigned for a complete and correct definition for all three terms, less information was required for a definition of shelter in place to be considered correct and complete than for either of the other terms, which have widely available accepted definitions.

Some of the definitions also shed light on the decision-making process. For instance, one response defined shelter in place as, “to get somewhere and stay. Not to be riding around on the streets.” This raises the question of how long a commute is acceptable to “get somewhere.” Would a 10 min drive be appropriate as long as a person stayed there for the remainder of the 3 h? Other responses indicated that respondents made a judgment on how far they needed or wanted to go and how soon they believed the tornado would arrive. One student believed it meant to “go back to a dorm, apartment, house or stay in the building you were in if you live too far away to drive there quickly.” Another student explained a choice he made to leave to pick up a child in daycare, even though his response indicated he understood the definition. Two responses, first by a student and secondly by an employee illustrate other trade-offs:

This means stay on campus in a safe place, but knowing what it means, and finding shelter where you are comfortable are two different things. If a tornado was literally outside, then I would have stayed, but my first instinct when the conditions were not hazardous yet, in my judgment, I knew I could make it 3 miles to my father’s office where I was advised to go by my father.

To move from our offices to a secure portion of the building. However, when you leave your office you no long have access to [instant] messages…. If we had not had a computer in the shelter area we would not have known what was being communicated in real time. Because of the interior sheltering location, most cell phones don’t receive a signal. Thus, in the shelter, we were cut off from any text message alerts.

5.4 Response

Due to some concern over the number of people who left campus during the severe weather, respondents were asked what they did upon receipt of the warnings. While about three quarters took shelter, 11.1% of students reported driving somewhere else after hearing the tornado warnings, and 25.2% reported leaving campus upon hearing that classes were suspended. Of greater concern, 40.3% of students and almost a quarter of employees reported going outside to look at the weather. Of the 449 respondents who left or tried to leave campus, 60.4% did not have a class or scheduled meeting later that day. The most frequent reason for not taking shelter was that one did not believe the tornado was a serious threat. Surprisingly, almost 20% of employees (who chose not to take shelter during the tornado) said they did not take shelter because they did not feel they had a safe place to go. Another common response to the warning was additional information seeking for both students (34.3%) and employees (38.9%) online or on television (43.9% of students and 22.8% of employees).

To help explain why a respondent chose to take shelter or not upon hearing the warning, logistic regression was performed with predicting who took shelter the dependent variable. The analysis was conducted for each sample separately and grouped together. For students, age and sex (being female) improved the prediction of who took shelter over the constant alone. Other research has also shown a link between age, gender and behavior in hazardous weather. For example, Drobot et al. (2007) found younger respondents were more likely to drive across a flooded road, and reported males generally exhibit more risky flood-related behavior although gender was not a significant predictor in that study. No other group of variables added to the model was statistically justified but using weather information daily or more frequently was also significant (along with age and sex) at each step. When the same analysis was conducted on the employee sample, being female was significant (p = .001) in each step. The only addition that contributed significantly was the variable having heard the tornado siren in the final step (p = .002). In the model with both samples combined, the status variable (employee or student) was added. This was a significant predictor (p = .000) of taking shelter where employees were about two times more likely to take shelter than students. Being aware of the possibility of severe weather at or before the time the watch was issued and being able to define the three terms discussed above were not significant in any of the models. To account for the possibility of sample size affecting statistical significance, a random sample of 10% of the respondents was selected. A similar logistic regression on this reduced sample indicated status, age and sex were still significant in predicting likelihood of taking shelter (at p < .01) although hearing the siren was not.

6 Conclusions and recommendations

The emergency alert message system used during this event appeared to have been successful. Less than 1% of the respondents reported having never heard of the Maroon Alert emergency alert system and 85% of respondents heard about the possibility of severe weather through the first emergency alert message or earlier. The 20% of respondents who heard about the suspension of classes from another person shows the continued importance of interpersonal communication as well. Interpersonal communication also accounted for how a large percentage of respondents heard about the NWS tornado warning. Also, the fact that local television was the most frequently identified source of information about the tornado warning for students shows that this traditionally relied-on method of risk communication is still important to this generation. The survey was also important because it shows how individuals heard about a weather emergency while at work. The nature of communication with co-workers, friends and relatives at other places of employment would be an interesting topic for future study.

The survey also showed that the respondents were able to define successfully the three terms used in the warning communication, tornado watch, tornado warning and shelter in place. The lowest scores were associated with definitions of tornado watch. Part of this can be explained by the ambiguity of the word “possible,” which was very commonly used in definitions of both watch and warning. It raises the question whether such ambiguous words are too common also in communication about warnings.

Apart from the high number of respondents who went outside to look at the weather, respondents generally took appropriate action upon hearing the warning. While almost a quarter of students did report trying to leave campus despite being encouraged to shelter in place, 73.1% of students and 81.4% of faculty reported taking shelter in one’s home or building. Some of the responses highlighted above indicate that some students also believed it was possible to both shelter in place and try to leave campus when the safe place was their house or apartment. Future messages would benefit from more explicit language regarding where students and employees should take shelter. That is, rather than saying shelter in place, it would have been clearer to instruct students to stay where they were if this location met the specified safety characteristics. Female respondents were more likely to take shelter than males, and employees were more likely than students. Students who were more frequent users of weather information were also more likely to take shelter than those who used weather information less than once per day. Similar to other studies on warning communication, respondents in this study also sought to confirm the warning information upon receiving it.

It is possible to make several recommendations for other universities based on this study and others. These include:

  1. 1.

    When buildings are not equipped with public address systems, an interpersonal method for delivering warning information may be necessary to reach faculty in classrooms without cell phones.

  2. 2.

    Provide teaching assistants (and even new faculty) with information regarding emergency safety procedures during orientations or with new employee information packets.

  3. 3.

    Account for the possibility that students may attempt to leave campus if classes are canceled or suspended.

  4. 4.

    If students are expected to remain on campus during the threat of severe weather, a message should be clear and explicit about what they should do and when.

The first recommendation was based on the number of faculty who wrote responses regarding how to get in touch with them during severe weather if they were teaching a class. A number of faculty stated that the best way to do this would be via specific interpersonal communication. Given the relative importance of interpersonal communication, colleges may wish to consider if there is a feasible and safe way to communicate emergency information to the classrooms to account for faculty who do not take phones to class or have access to instant messaging. Few buildings on this college campus have public address systems (although the university has since added a new campus-wide address system), and some respondents reported not being able to hear the siren. Also, since the desired reaction is not the same for all emergencies, it may be beneficial to have a method of delivering a specific hazard message that does not rely on cell phone or internet access. Classes were already dismissed by the time the tornado necessitated taking shelter in this event. Other situations (e.g., toxic chemical release, armed intruder) might provide less lead time but still require immediate action.

The second recommendation was to help some students with roles of authority become more knowledgeable about severe weather. A few students reported that they were from a place that does not experience severe weather and therefore did not know what to do. It might prove important to provide additional information to graduate students who are more likely to be from out of state and especially those who will have teaching responsibilities. Resident assistants could also benefit from additional information. The majority of teaching faculty who responded to the survey knew the correct action to take during severe weather, but it is likely that not all employees would know the correct action for every event.

Some respondents reported leaving campus upon hearing that classes were suspended. Based on this, classes should be canceled only if time will allow a large number of people to leave campus prior to the event. A post-storm assessment following the Super Tuesday outbreak of 2008 suggested that K-12 students should be kept on campus during severe weather because school buildings are often better constructed than homes in the rural South (Hayes 2009). This was described as the safer option only when a school has a well defined and practiced emergency plan. Whether students are kept on campus or dismissed, it is always recommended for a warning message to be clear about what students and employees should do in an emergency.

Finally, it should be noted that the student response rate to this survey was lowered than desired, even though a large number of students were surveyed. Care should be taken in applying these results to all students or all colleges. It is possible that the low response rate may have contributed to the percentage of students who understood the terms used in the warning communication, although other studies have typically reported high understanding of the terms tornado watch and tornado warning. The percentage of respondents who took shelter during the warning is also consistent with other studies. Many of the other findings such as the importance of local television and interpersonal communication as well as the influence of gender and age on behavior have been documented in other studies as well. There is no reason to expect this sample is biased. It is possible, however, that schools located in other areas where tornadoes are either more or less frequent may respond differently in a similar event.