A team of U.S. Geological Survey scientists have developed a web service that combines seismic data about an earthquake with Tweets of surprise and angst from the popular microblogging service’s users.
The goal of the project is to improve emergency response by providing a crowdsourced window of the conditions on the ground immediately following a quake.
“Why would such a system work?” asked Paul Earle, a geologist at the USGS, at the American Geophysical Union fall meeting Monday. “Because people like to tweet after an earthquake.”
It turns out that the “Earthquake! Earthquake!” SOS that you tweet, aggregated with thousands of others, provides an excellent indication of the strength and severity of a quake. A little rumbler yields just a small spike, while a strong quake produces a huge spike in Twitter activity, as seen in the graph above.
Right now, the system is designed to automatically harvest tweets, so they can be e-mailed to would-be responders with traditional earthquake measures and plotted in Google maps. All the OMGs and BFDs can provide some qualitative color about what’s happened on the ground.
“We’ve developed a prototype system that integrates Twitter messages with our standard earthquake alerts,” Earle summarized.
It’s one of a variety of ways that some scientists are trying to use the crowd and the cloud to augment the professional tools they already have. It’s already been suggested that some laptops with accelerometers built-in may be useful as earthquake detectors. Researchers in other disciplines, like bird-tracking, have also had success using citizen scientists. In a separate session at the AGU fall meeting, another scientist described a garage seismic station that has begun to make its way into homes around the Bay Area.
The challenge presented by data gathered outside the traditional channels is that it’s noisy. What the scientists gain in breadth is partially canceled out by the lack of control they have over the incoming information. After all, Quake is also a popular videogame and Dairy Queen serves up a “brownie earthquake,” and both are likely to find their way into tweets.
“We’ve been developing filtering techniques that allow us to tell the difference between an actual earthquake and a group of people who just finished playing a videogame and got the munchies,” Earle said.
Eventually, Earle and his software engineer, Michelle Guy, would like to use geolocated tweets to rough out a “felt range” for earthquakes, but it’s not as easy as it looks. Only half of Twitter users even give a basic location like “San Francisco” and just a few percent of earthquake tweets come from GPS-enabled devices, which allow for precise measurements, Guy said.
It also stands to reason that USGSted, as the program is known on Twitter, won’t be of much help in any earthquake strong enough to knock out telecommunications systems.
Where Guy and Earle say their service may be most useful is in the window of time between when an earthquake happens and when the USGS data starts to pour in. That time period can be anything from two to 20 minutes, and during that time, all those tweets are much better than nothing, if far from perfect.
Image: Paul Earle