A new study of Twitter activity during Hurricane Sandy describes a correlation between activity on the service around the time of the storm in 2012 and actual damage. In the study, published in Science Advances, the authors suggest social media could eventually be used to assess the distribution of damage during similar crises.
The researchers studied 9.7 million geo-located tweets, first nationally, and then focusing on two states hit hardest by the hurricane: New York and New Jersey. Using that data, the researchers were able to chart how proximity to the hurricane changed the volume of tweets, specifically those using keywords like "sandy" and "frankenstorm." "At the peak of the disaster, event-related keywords rank higher and activity increases with proximity," the researchers write.
Taking the experiment one step further, the researchers sought to analyze how Twitter activity correlated with actual damage. Using information from FEMA assistance grants and insurance claims associated with the hurricane, the team was able to determine that activity on the service was a solid predictor of damage.
The researchers point to other studies that have come to similar conclusions, including one that found a correlation between Flickr uploads with the #sandy tag and atmospheric pressure. But the relative simplicity of tweet volume could one day be a useful tool for emergency responders. "Immediately after a disaster, [officials] should focus on persistence in activity levels to assess which areas are likely to need the most assistance," the researchers write.