When a series of EF5 tornadoes, the most powerful on the scale, hit Alabama and areas of surrounding states, houses were torn apart, their contents scattered by the winds. Almost all the photographs, diplomas, magazines, and objects were lost, but a few were found thanks to a collective effort organized through a Facebook page created by Patty Bullion, a resident of Lester, Alabama, population 111.
"I got on Facebook right after the storm," Bullion told ABC News about the page's creation. "A friend of mine who lives down the road posted that it was raining pictures -- falling out of the sky."
"A friend of mine who lives down the road posted that it was raining pictures."
The page she made, called "Pictures and Documents found after the April 27, 2011 Tornadoes," began with items she found in her own yard, but expanded as more people heard about the page and contributed belongings they'd found. Within a year, more than 100,000 people had "liked" the page and 1,700 items were returned to their owners through the simple matchmaking of the project.
This attracted the attention of John Knox, a weather and climate scientist at the University of Georgia. He'd studied meteorology at the University of Wisconsin-Madison, where Charles Anderson had done a pathbreaking study on the debris fallout from the Barneveld Wisconsin tornado, and was familiar with the work of John Snow at the University of Oklahoma, who extended the study of debris through aggregating historical newspaper accounts. Both efforts suffered from the same defect: it was hard to build a large enough dataset to offset the low precision of many reports. In the past, it was simply logistically, practically difficult to find a lot of people who had both lost and found items.
That is, until Bullion created her Facebook page, and through word-of-mouth, people across the region made it into the hub for returning items to their owners. Knox knew a novel dataset when he saw one, and he contacted Bullion, who allowed his students to access her Facebook account. They painstakingly took the postings and turned them into structured data that they could study. Out of respect for tornado victims, Knox decided against contacting people who'd lost items, sacrificing some data and precision. He called his decision-making process "data mining with a heart."
With that limitation in place, they set about figuring out which objects had defined beginning and endpoints. They were aided by the fact that many of the towns in which people lost and found items were geographically small, so they could circumscribe both poles of the trajectory easily. Still, they had to throw out 800 objects for which they could not ascertain decent geo-data.