Looking for ways to predict response to hurricane evacuation orders

Hurricanes, too, can be quite unpredictable, as evidenced by the ever-changing “cone of uncertainty” included in forecasts as storm systems approach.

But expansive review of data gathered in prior evacuations reveals patterns that can be analyzed and incorporated in regional models, based on mathematical predictions and controls, to strengthen the reliability of predictions in future storms.

That’s what the two new papers show. Both draw on data from the eastern part of hurricane-throttled North Carolina, where the Outer Banks and other coastal areas have seen more than their share of evacuation orders.

One of the papers, published by Environmental Hazards, looks at demographic data to see which factors influence the decisions of various groups and their likelihood of evacuating when mandatory orders are issued versus voluntary orders.

“Social and environmental cues influence behavior,” said Sarah DeYoung, postdoctoral researcher at DRC who was the lead author of that paper and has just accepted a tenure-track position at the University of Georgia.

“But we understand, too, the temporal nature of that,” Davidson said. “It’s not like people make a decision on Day 1 and follow through with that. They see what happens and change their minds, too.”

The study, based on survey data collected in 2011 through phone interviews with North Carolina residents in Wilmington, Raleigh, Jacksonville and the Outer Banks, looked at respondents’ “threshold for evacuation” – whether they had a high threshold and were less likely to evacuate or a low threshold and more likely to evacuate.

Those lines moved a bit, depending on whether the storm discussed was a higher or lower category of strength and whether the evacuation order was mandatory or voluntary.

But in general, DeYoung said, white respondents had a higher threshold than non-white respondents, a finding that was particularly interesting given that other studies in the United States suggest that non-whites evacuated later.

“This was really notable for us,” said Wachtendorf, associate professor of sociology and the lead social scientist on the research. “Is it that minority segments of the community are willing to leave but don’t always have sufficient resources to do so? Is it because, as other research suggests, they have less trust in officials and, particularly after what happened after Hurricane Katrina, they believe they can’t rely on officials if they stay? It really points to an area where more research is needed.”

Respondents who had ignored previous evacuation warnings were also more likely to ignore an order in the future.

And DeYoung said most people saw wind as more dangerous than water, but in reality it is the storm surge and flooding that causes more deaths. Most hurricane-related deaths occur in areas where people decided not to evacuate. Wachtendorf said this could lead people to dismiss the threat of lower category storms, with relatively lower wind speeds, despite the threat flooding can pose.

One recommendation is to increase public awareness of the risk associated with drowning and flooding versus the probability of death caused by wind damage.

Research in progress points to other important factors in the decision-making process, including concern about traffic jams, caring for pets and livestock, and fear of crime in public shelters.

“It’s a moving target,” Davidson said. “There are challenges in science and challenges in engineering. But understanding people’s behavior is one of the most challenging parts.”

“That’s one reason why the interdisciplinary approach is so valuable,” said Wachtendorf.

The second paper, published on Science Direct in the journal Transportation Research Part A: Policy and Practice and authored by Kecheng Xu, a graduate student in Cornell University’s Department of Civil and Environmental Engineering, and Cornell professor Linda Nozick, describes new models that estimate the number of evacuees in specific evacuation zones and predicted accurately what individual households would do about 70 percent of the time. Accuracy improves as data are aggregated regionally.

The work by the civil engineers on the project used data collected by the social scientists to inform many of the assumptions for their models.

Having reliable models puts the power of the knowledge into useful form for planners and helps them shape effective, efficient evacuation plans that could save lives in the future.

— Read more in Kecheng Xu et al., “Hurricane evacuation demand models with a focus on use for prediction in future events,” Transportation Research Part A: Policy and Practice 87 (May 2016): 90-101 (DOI: 10.1016/j.tra.2016.02.012)