EIIP Virtual Forum Presentation — March 22, 2006

The Graniteville South Carolina Chlorine Spill
A Study of Evacuation Behavior

Jerry T. Mitchell, Ph.D.
Director, Center of Excellence for Geographic Education
University of South Carolina

Avagene Moore
EIIP Moderator

The following version of the transcript has been edited for easier reading and comprehension. A raw, unedited transcript is available from our archives. See our home page at http://www.emforum.org

[Welcome / Introduction]

Avagene Moore: Welcome to the EIIP Virtual Forum! Amy Sebring, my partner/associate, and I are pleased you could join us today!

Today's topic is the Natural Hazards Center Quick Response Report #178, "Evacuation Behavior in Response to the Graniteville, South Carolina, Chlorine Spill." If you have not already read the background materials, which include our speaker’s bio please check it out after the session at http://www.emforum.org/vforum/060322.htm.

It is a pleasure to welcome Dr. Jerry T. Mitchell to the EIIP Virtual Forum. Dr. Mitchell currently serves as the Director of the Center of Excellence for Geographic Education at the University of South Carolina in Columbia, South Carolina. Dr. Mitchell co-authored this study and has previously authored more than twenty peer-reviewed articles and book chapters in the field of hazards. His work has appeared in the journals Natural Hazards Review, Environmental Hazards, Social Science Quarterly, and the Annals of the Association of American Geographers.

Jerry, we are pleased to have you with us today. I now turn the floor to you.


Jerry Mitchell: Hello. I am glad to be here and look forward to hearing your comments and questions today. The research I am going to discuss today, "Evacuation Behavior in Response to the Graniteville, South Carolina Chlorine Spill, was supported by the Natural Hazards Information and Applications Center, Quick Response Grant Program.

A copy of this report is located at http://www.colorado.edu/hazards/qr/qr178/qr178.html.

Situation reports for the incident may be found here: Norfolk Southern Graniteville Derailment, US EPA, http://www.epaosc.org/doc_list.asp?site_id=A4GY.

First I will give you background on the incident. At 2:39 am on the morning of January 6, 2005, a Norfolk Southern freight train traveling from Macon, Georgia towards Columbia, South Carolina missed a switch and crashed into a parked locomotive in the town of Graniteville, South Carolina. Graniteville is located in western Aiken County approximately 10 miles from the Georgia/South Carolina border.

The collision ruptured one of the train’s tank cars, carrying 90 tons of chlorine, sending a plume of chlorine gas across the northern portion of town. Four other tankers carrying chlorine, liquid sodium hydroxide, and liquid cresol derailed but did not rupture. Nine people died as a result of injuries suffered from chlorine inhalation; more than 550 people sought medical assistance.

A mandatory evacuation order was given at 2:30 pm to residents within a one-mile radius of the crash site. County officials estimated 5,400 evacuated that day and returned home between 1-2 weeks later. The physical, psychological, and financial toll endured will remain with Graniteville area residents for some time to come.

The following slide shows the timeline of events in this incident. Please click on the link and it will come up in your browser. I will give you time to read / view the slide before going on with my formal remarks. Please come back to the chat screen after viewing the slide.

[Slide: Timeline]

The Graniteville experience provides an opportunity to assess preparedness for and responses to hazardous materials spills. The primary purpose of this study was to examine the evacuation behavior of residents. The lessons learned from the experiences of the affected residents will help improve preparedness efforts in the future. Additionally, this research will assist in our understanding of how people respond to extreme events and those factors that influence evacuation decision-making.

Research Questions:
There were three primary research questions that provided the focus for this Quick Response project.

1) Were there disparities in the timing and execution of the evacuation order based on location (within the one-mile evacuation zone), the demographic characteristics of residents, or the risk?

2) Was there an evacuation shadow produced beyond the mandated one-mile evacuation area?

3) How did the media coverage of the event amplify the perception of risk and/or evacuation behavior and how do local, regional, and national news media frame the event (normal accident v. catastrophe)?

In the information that follows, I will primarily discuss the results of questions 1 and 2, and hope that an interesting discussion might be had about the evacuation shadow in particular. Some very interesting results (especially for a geographer) regarding communication were uncovered.

To better understand how Graniteville area residents responded to the spill, two mail survey instruments were created. The two surveys measured the behavioral differences between two distinct groups: 1) those faced with a mandatory evacuation and, 2) those who evacuated without an expressed directive to do so.

The first survey was sent to all households within the one-mile evacuation zone. A second survey was sent to residents within the 1-2 mile curfew zone, the "shadow evacuation zone." A shadow evacuation refers to the group of people evacuating that were not required to do so. Copies of these surveys can be viewed through the QR report website mentioned earlier.

Survey recipients were identified from 1) a GIS-based boundary file that articulated the geographic extent of all the parcels in the county; 2) an electronic Tax Assessor table that provided owner information, including both property and mailing addresses; and 3) an electronic Emergency 911 table that supplied address data of the residents.

In short, over 2,700 surveys were mailed to Graniteville area residents (a loose definition since the town is not incorporated with definitive boundaries). A total of 574 usable surveys were returned. In general the survey questions asked about the timing of received messages, the decision to leave or stay, the source of the warning message, and respondent demographics.

General Findings:
The majority of respondents were in the area and at home when the derailment occurred. According to our timeline, the reverse 911 system was used to notify 3,600 homes with a shelter in place message around 6:30 am on Thursday January 6, 2005. This was the first use of the system since its installation a year earlier. More than three-fourths of the respondents (76.8%) received a message to stay in their home or work area until the danger passed (all sources combined).

At noon on Thursday, January 6, 2005, South Carolina Governor Mark Sanford declared a state of emergency for Aiken County. At 2:30 pm an evacuation order was given for a 1-mile buffer surrounding the crash site asking all residents within the affected zone to evacuate by 6 pm. Within a 2-mile radius of the site, a curfew (lasting from 6 pm to 7 am) was put into effect.

Eighty-three percent of the respondents in the one-mile zone reported receiving a message to evacuate. While the official evacuation order was targeted to residents in the one-mile zone, respondents within the 1-2 mile zone also thought the mandatory evacuation order was directed to them as well with 54% saying they received a message to evacuate.

Seventy-six percent of our total respondents evacuated in response to the accident. As expected, this percentage was much higher in the mandatory one-mile buffer (98.4%), than in the 1-2 mile zone (59.0% report evacuating). For the majority of our respondents, the chlorine spill was their first experience with a mandatory evacuation order. Thirty percent evacuated with a pet.

Sheltering Options:
Most of our respondents sought shelter with relatives (52%). There are few differences between those in the mandatory evacuation zone and those outside it. The one exception is the use of public shelters—those in the mandated zone were more likely to use them than those who did not reside there. Only 6% of our respondents used the public shelters.

To assess their reluctance to use such mass care facilities, the respondents were asked their reasons why they did not go to a public shelter. The majority replied that other family members offered to take them in (32%). The other answers were dislike for shelters (25.5%), not told about shelters (12.0%), pets not allowed (10.3%), didn’t know the location or couldn’t find it (6.4%).

Accident Impacts:
One report stated that 529 people had been treated in local hospitals. Approximately 18.6% of our respondents sought medical attention, but there is a wide discrepancy between those within the mandated zone and those outside it.

Respondents within the one-mile mandatory evacuation zone reported that 28.7% of them sought medical attention compared to 7% of the respondents who lived in the 1-2 mile buffer. Three quarters of our respondents within the mandatory evacuation zone (75.5%) reported having their home inspected, compared on only 15.2% of those respondents living within the 1-2 mile zone, upon their return.

Detailed Findings:
The two locations – the one-mile evacuation zone and the 1-2 zone – were compared for differences regarding the timing of first learning of the event, notification to stay or leave, evacuation timing and the return date of the respondent if they evacuated.

Respondents in the one-mile zone learned of the accident earlier than respondents in the 1-2 mile zone. Respondents in the one-mile zone evacuated earlier than respondents in the 1-2 mile zone and returned home later. These findings indicate generally (and not surprisingly) that those living closest to the crash site learned of the accident sooner (witnessed it or were told of it), received messages to shelter in place earlier, and in fact, evacuated earlier as well. These same respondents also returned to their homes later as those neighborhoods closest to the crash site were ‘re-opened’ last.

A variety of demographic characteristics were available for investigation. Regarding the time of evacuation, it should be noted that the mean evacuation times do not differ significantly by family size, the presence of special needs residents, level of education, or gender. Furthermore, of these demographic factors, only the level of education appears to have influenced the return date: more educated respondents returned home earlier on average.

Race and ethnicity was also explored. Collectively, white respondents reported receiving the shelter in place message slightly earlier than black respondents, but this appears to be related to faulty recall. There were, however, significant differences between the groups on the timing of their return. White respondents indicated that they returned home earlier than African-American respondents. Several of the census blocks closest to the crash site contain higher numbers of African-American residents. As such, these neighborhoods would have ‘re-opened’ later, thus explaining the later return of these respondents as compared to white respondents.

Evacuation Shadow:
An over-reaction to an evacuation order generates the so-called "evacuation shadow." There is ample evidence to suggest that this occurred in Graniteville. Evacuation shadows have been identified in events ranging from the Three Mile nuclear incident in Pennsylvania (1979) to Hurricane Andrew (1992).

An optimistic view of the evacuation shadow would observe that with more evacuees fewer casualties would result. While possibly true, the evacuation shadow presents a number of harmful consequences. For example, higher than expected traffic volume will slow evacuation times.

Ninety-three percent (93%) of the respondents living within the 1-2 mile zone from the crash site indicated that they were in the Graniteville area at the time of the accident. A very large percentage of these – 59% – indicated that they evacuated the immediate area.

While a dusk to dawn curfew was in effect for the first three days, no evacuation order was in place for the 1-2 mile radius buffer. Why then did these people leave the area? The top answers were: mandatory order (20%), threat to family was high (20%), situation was getting worse (14%), and advice of family/friend (11%).

Given that there was no mandatory order for their area, why is that cited as a frequent motivator to leave? It is highly probable that residents did not know where they were in relation to the accident site and its accompanying mandatory evacuation order. Without a map or extensive local knowledge, residents could not determine where the crash site was located and how far their home was from that site. A sample response from the 1-2 mile zone indicates that this indeed was the case:

"When the evacuation automated call came we were not sure if we were in the one mile. It took way too long to find out where we were to the radius of the 1 mile. Basically we were waiting (packed) for someone to knock on the door & say go! We measured a map and figured we were just outside the edge by a few hundred feet."

Based on this research, we have concluded that:

1. Evacuation disparities did exist based upon location. Residents within the one-mile zone evacuated earlier and returned home later. This result was expected.

2. Notification disparities were also seen. These systems errors must be investigated. Additionally the problem of reaching residents without telephone land lines remains, especially given the ubiquitous nature of cell phones.

3. Considerable confusion existed regarding one’s location within the 1-mile evacuation zone. Residents did not know if the evacuation message was applicable to them personally. Future evacuation notices must include not only the general evacuation area but specific landmarks or street names to indicate precise boundaries.

4. There were a few differences in the timing and execution of the evacuation order in terms of household demographics. Most of these disparities can be related to location.

5. A substantial shadow evacuation did occur. This suggests an area of opportunity to sharpen risk communication strategies.

These findings are but a few of many other lessons that might be derived from these survey responses. Hopefully the Graniteville experience can lead us toward proactive measures that mitigate the effects of hazardous material spills when they happen again. That summarizes the main issues, so I will be happy to take questions now.

Avagene Moore: Thank you, Jerry, for your insight into this report. I am sure there are several questions for you from our audience.

[Audience Questions & Answers]

Burt Wallrich: Did you do any research about the experiences of seniors and people with disabilities, e.g. Were they more or less likely to evacuate? Did they report particular problems? Were they more or less likely to be among the casualties?

Jerry Mitchell: Good question. We did research on age and did not find evacuation decisions to be related to that factor. The casualties were evenly spread among age groups - fatalities that is. I would have to investigate hospital visits to know about injuries; however, we have not done that.

Amy Sebring: From the timeline there appears to be a 4 hour delay in getting the first ring down shelter-in-place message out. Do you know why it took so long? Especially since the EAS message went out an hour earlier?

Jerry Mitchell: We do not know why. Reports had gone out over TV and radio. The same problem occurred later with the evacuation message. The decision to evacuate was around 2 PM, but the phone message did not go out until 4:30 PM.

Bob Armstrong: Did the survey also include information regarding how far residents evacuated to? If they just went across town to a family or friends house, this is understandable, even if they were not in the mandatory evacuation zone. Most people subscribe to the "better safe than sorry" theory.

Jerry Mitchell: No we did not ask distance info, but in the free response section, people did tell us where they went. Most indicated that they stayed in the general area. I'd say within about 10 miles generally.

Burt Wallrich: With 574 returns out of 2,700 surveys mailed out the question arises of how the results might have been skewed by a mail survey, as opposed to other methods that might have been used such as going door-to-door or telephoning a sample?

Jerry Mitchell: We were surprised by the 20% turnout. We figured people might want to vent or relate their experience. Given the funds we had available, the mail survey was the method that we went after and not without difficulty. Many surveys were returned due to issues with post office boxes and rural routes. We had to hand delivery 500 plus to a post office and put them in the P.O. boxes just to get them to people.

Amy Sebring: Do you know whether their ring down system was integrated with GIS? That is, could they draw the radius on a map to activate the system?

Jerry Mitchell: No the system was separate. Aiken County GIS provided the buffering distances separately. It appears that they used zip codes for the ring down which obviously has a different geography than a round buffer. The same problem would occur with local exchanges. You will hit some people out of the zone, and maybe miss others within it. This was their first use of the system, so trial/error issues are understandable.

Rich Vandame: Was there a particular evacuation notification method that was the most effective?

Jerry Mitchell: People report different methods: there was door-to-door notification, TV, and radio. People seemed disappointed with the TV coverage. They did not know the local geography, and could not tell people where to go. The phone messages were also devoid of geography and thus confused a lot of people. That topic is an interesting one, if people are interested in pursuing it.

Avagene Moore: You mentioned venting. Did your research cover how the evacuees were provided with ongoing updates during the evacuation period? If so, how?

Jerry Mitchell: There were no updates officially. People were asked to tune to TV or radio.

Jonathan Haber: I missed the first part, so I'm sorry if this was already covered. Did the evacuation message reach out to businesses, guests in the area (maybe those in hotels). If so, how did people who weren't residents understand what and how to evacuate?

Jerry Mitchell: The phone message dialed up exchanges in the local area. Presumably businesses were also called. Given that this happened at 2:40 AM, most people were at home. This is a rural area with no significant hotels, etc.

Burt Wallrich: I don't want to hammer at this, but didn't you hear of anyone who found it hard or impossible to evacuate because of a physical disability, or a hearing or language problem that interfered with getting the word, or lack of transportation?

Jerry Mitchell: No reports were made in the survey related to special needs, or language [being a problem overall. Several stated that they had special needs, but few indicated that this impacted what they did]. People mainly complained about the lack of info in the messages to decide whether to stay or go

Amy Sebring: Your study focused on evacuation behavior, but did you have any survey questions pertaining to the shelter-in-place message? Whether residents understood what that meant and complied?

Jerry Mitchell: We asked people if they received that message, what time, etc. Again, people were concerned about the lack of info. I have a copy of the shelter in place message here:

"This is an emergency message. Press any number key to hear the message. Please remain indoors. Turn off heating and air systems. Please tune to your local television stations and radio stations for more information."

This message could be better crafted. First, there is no indication about the nature of the threat. Sent roughly three hours after the train wreck, it is quite possible that residents were not even aware that an accident had occurred. Residents are told to turn off their heating systems, but not why.

Avagene Moore: Jerry, it seems to me your study and this experience would prompt many lessons learned for this community. Are you aware of any after action changes in their planning?

Jerry Mitchell: Yes, there has been discussion about prevention, namely with the railroads.

This section is used to fast trains, which now move more slowly. Other communities are reviewing their hazmat plans. For example, at USC we have dorms near a rail line but had not planned for a spill for the campus. More attention is being paid to the messages sent as well. There needs to be some geographic specificity to get people to respond. For example, this might be a better shelter message:

"This is an emergency message. Press any number key to hear the message. A large chemical spill has occurred at Avondale Mills in Graniteville. Do not leave your home. Remain indoors and turn off heating and air systems to avoid injury. An update to this message will occur at ‘x’ am. Please tune to your local television stations and radio stations for more information."

This slightly longer message helps the listener understand what has happened and why they need to take the requested action. The type of accident is indicated along with the name of a prominent landmark. A time element also assures the resident of further updates if other sources of information are unavailable.

Amy Sebring: Am I correct in assuming that the shelter-in-place message remained in force until the evacuation order was communicated - at least theoretically?

Jerry Mitchell: Yes, that is true - theoretically. Many people did move about.

Lori Wieber: Your responses indicated 30% of evacuees had a pet. Did any of them include remarks regarding animal issues in their comments?

Jerry Mitchell: Yes, good question. Most took a dog, instead of the hard to herd cat. Many did not look out for shelters because of the pet issue. The local authorities did put together a pet rescue plan and held onto a few pets until the owners returned. Several pets did die after they were returned to their owners.

Amy Sebring: I know it is beyond the scope of your study but was this community, being located in the western part of the state, generally inexperienced with previous disasters and emergencies? Do you think this is a factor?

Jerry Mitchell: We asked if they had evacuated before and for nearly all this was the first evacuation. Their inland location meant that a previous hurricane evacuation, for example, would not have applied to them.

Amy Sebring: Anything we have not covered Jerry that you would like to add?

Jerry Mitchell: Yes, and this is the geographer in me, the striking lack of 'space' in the messages really puzzled me. No mention of where the event happened. Shelter names were mentioned but not how to get there, and some paths would take you to a shelter via the plume!

Donald Pinegar: Were first responders involved with the evacuation or did the conditions preclude their entry?

Jerry Mitchell: First responders did help with some of the door-to-door notification on the outskirts of the spill area.

Amy Sebring: Do you have any follow up research planned on this topic? If not this incident per se, other evacuation-related studies?

Jerry Mitchell: I have a paper under review right now about the message content and it would be useful to understand how to bridge the different geographies. For example, there is a zone for zip code notification, telephone exchange notification, a buffer zone (that may be round or plume-shaped), and the political jurisdictions. Lots of overlapping geographies to pull together successfully.

Amy Sebring: I also understand there has been legislation as a result of this incident, requiring railroads to notify communities of the hazardous materials?

Jerry Mitchell: I am not sure where that is currently. I know that various communities have been discussing the re-routing of those trains around urban areas, but that simply displaces the risk.


Lori Wieber: Thank you Dr. Mitchell. Was English the only language used in this area?

Jerry Mitchell: We are only aware of English, but there is a growing Hispanic migrant farm population here in SC, so some people speaking Spanish only may have been missed.


Avagene Moore: Thank you, Jerry! We greatly appreciate your effort and time on our behalf.

Please stand by a moment while we make some quick announcements.

If you are not currently on our mailing list, and would like to get program announcements and notices of transcript availability, please see the Subscribe link on our home page. If interested in receiving announcements of the EIIP Virtual Forum sessions, please note that subscribing for the EIIP Mail List is much easier now. See our homepage http://www.emforum.org and subscribe today.

We have one new Partner to announce today - the Disability Preparedness Center (DPC) http://www.disabilitypreparedness.org/. The EIIP Point of Contact is Carl T Cameron, PhD, President. We welcome the DPC to the EIIP!

If you are interested in becoming an EIIP Partner, please see the "Partnership for You" link on the EIIP Virtual Forum homepage http://www.emforum.org .

Again, the transcript of today's session will be posted later today and you will be able to access it from our home page. An announcement will also be sent to our Mail Lists when the transcript is available.

Thanks to everyone for participating today. We appreciate you, the audience! Before you go, please help me show our appreciation to Dr. Jerry Mitchell for a fine job. The EIIP Virtual Forum is adjourned! Thank you, Jerry! Great job!