In a study published Friday in the journal Science Advances, researchers detailed how 52 million geographically pinpointed tweets they gathered from before, during and after the hurricane offered telling insight into where it ultimately wreaked the most havoc. In essence, the scientists determined that the areas that experienced the most notable spike in Twitter activity were associated with areas where residents filed the most insurance claims and received the most individual assistance from Federal Emergency Management Agency grants.
“The signal is not perfect, but it does exist. There’s a moderate, positive relationship between the activity on Twitter and the dollar amount of relief funds that go into a certain area,” said Manuel Cebrian, a co-author of the study and a senior scientist at Australia’s Commonwealth Scientific and Industrial Research Organization, or CSIRO. “Can you predict exactly how much this town or that town is going to get [in aid]? You cannot. But you can predict how much it will get in relation to another town.”
Cebrian and a group of colleagues amassed more than 52 million tweets from roughly 14 million people located along the path of Hurricane Sandy, which caused massive flooding and widespread property damage in communities along the East Coast, particularly in New Jersey and New York. The tweets were sent between Oct. 15 and Nov. 12, 2012, covering the time before the storm made landfall until well after it departed. Researchers did not analyze the content of individual tweets but rather searched them for a collection of keywords, such as “recovery,” “gas,” “FEMA” and “hurricane.” While the data was “very noisy,” Cebrian said, researchers noticed that the places with the highest per-capita number of tweets correlated with the spots hit hardest by the storm.
Pascal Van Hentenryck, another co-author and a professor of engineering at the University of Michigan, said that while relying on Twitter to predict damage during disasters might not be “ready for prime time,” the study offers a promising glimpse of what might be possible in the future. The effort is among the latest in a growing number of projects in recent years aimed at finding ways to better predict where natural disasters are going to strike and where the worst damage will be.
“It’s hard to predict how [things] are going to be damaged. You have to send crews to evaluate,” he said. “Any kind of additional information that tells you more is going to speed up that process. … With Twitter, there’s a situational awareness. People will tell you more if they are there. It’s eyes on the ground.”
Indeed, researchers in recent years have looked for ways to harness the power of social media to better predict and respond to calamities. They have studied whether Twitter might help in tracking everything from the flu to outbreaks of foodborne illness.
A group of researchers in 2013 found that Flickr could be an effective tool in tracking natural disasters in real time after they documented the many photos that users posted of Hurricane Sandy as it pounded the Eastern Seaboard. Photos tagged with words such as “sandy” or “hurricane” peaked as the unprecedented storm was making landfall. Even the U.S. Geological Survey has begun to use Twitter to supplement its own vast seismological data, according to a 2015 blog post, because the social media service can help quickly identify any earthquake large enough to be felt by humans.