Hurricane Florence puts IST research to the test

Jessica Hallman
October 30, 2018

UNIVERSITY PARK, Pa. – In the days surrounding the landfall of Hurricane Florence, which pummeled the Carolinas in September, a team of Penn State researchers was called upon to turn their knowledge into action.

The team had previously partnered with Charleston County Consolidated 911 Center in South Carolina to develop a tool that would integrate social media posts into its emergency response operations. When Hurricane Florence was heading toward Charleston, South Carolina, the research group deployed a remote digital support team to display filtered tweets from the county. Sorting through that data, they set up a process to deliver urgent and actionable messages directly to call center dispatchers that could improve response operations.

“It’s another wave of data collection and of finding information about the community that dispatchers might not otherwise have,” said Shane Halse, a doctoral candidate in the College of Information Sciences and Technology and member of the research team.

According to Halse, the team started by defining the area they wanted to search, which was Charleston County. Then, they created a list of hurricane-related keywords to include in their search.

As the storm made landfall, 10 doctoral students and faculty ran software to analyze filtered tweets and to call attention to any that needed immediate response. The remote team worked from State College, Pennsylvania, and in Albi, France, where the research team has an exchange program with an engineering school, the École des Mines d'Albi.

“It’s another wave of data collection and of finding information about the community that dispatchers might not otherwise have." 

— Shane Halse, IST doctoral candidate, Penn State

Hurricane Florence ultimately navigated north, sparing Charleston County from the worst of its devastation and therefore minimizing the number of urgent and actionable social media posts in that area. Yet it still provided an opportunity for the research team to test their work and to make adjustments for the future.

“While remote digital support teams assist emergency and crisis management organizations around the world, the procedures for coordinating the efforts of remote teams of digital volunteers and 911 dispatchers in affected areas remains less established,” said Rob Grace, doctoral student in the College of IST and member of the research team. “Our work with the Charleston County Consolidated 911 Center during Hurricane Florence saw the development of procedures that will make coordinating the work of digital volunteers and 911 dispatchers more seamless during the next crisis.”

Hurricane Florence filtered tweets

A screen shot of the filtered tweets displayed by the College of Information Sciences and Technology remote digital support team, which highlighted urgent and actionable social media posts from Charleston County, South Carolina, in the days surrounding Hurricane Florence's landfall in September 2018.

IMAGE: Shane Halse

“We showed that we can set up a system that can be used to provide at least initial social media data to people who are responding to emergencies,” added Halse. 

According to Halse, the team has gathered all social media data surrounding Hurricane Florence for future analysis and planning. And, with the hurricane’s trajectory north, the team’s attention was turned to an issue occurring in North Carolina — people tweeting about overflowing hog waste from flooded farms in the region.

“We weren’t collecting live data for that particular area, but we can collect the historical messages on Twitter,” said Halse.

This data will help the team understand how to navigate a biohazard through social media by determining which keywords to look for, which filters to improve and the impact to the community.

“These are all research questions that we may come up with based on this data,” said Halse. “When we analyze it, I’m very interested to see the impact.”

While the research team hopes to fully implement the technology in the 911 call center in Charleston County, and someday across the country, they plan to focus on what they refer to as “layer two” crises, which are less intense than a major hurricane or terrorist attack but more significant than a fender bender. Examples of layer two crises include a multi-car pileup or a bank robbery, said Halse.

“It makes it a little trickier to find data, but that’s where we can provide the most value,” he said.

As the team continues to advance their system, they emphasize that it is simply an additional data source that could help 911 call centers more quickly and more accurately respond to emergencies. Halse cited the aftermath of Hurricane Harvey in 2017, which left call centers in the Houston area overloaded.

“You can’t call 911 and leave a message, but you can post to Twitter and leave a message,” he said. “If you have five percent of a phone battery left and get put on hold by 911, you may be better served by using what’s left of your phone battery to post to social media.”

“The goal of this research is to supplement and augment the information already flowing through a 911 center,” concluded Andrea Tapia, associate professor in IST and the director of the 3C Informatics Lab and this research effort. “In the case of Hurricane Florence, we played the role of a remote crisis communications center. In the future we’d like these tools to sit seamlessly inside the largest 911 centers, or in the case of smaller 911 centers, continue remote support in times of crisis.”

Last Updated October 31, 2018