“I Felt That”: USGS Leverages Earthquake Tweets

Earthquakes are the perfect Twitter phenomenon. You feel something weird, you want to share it and confirm it with the world. That’s what happened yesterday in the San Francisco Bay Area, when we experienced a 4.1 quake centered near Milpitas.

Tweets about the Jan. 7 earthquake near Mipitas, Calif.

It’s such a canonical example that Twitter CEO Evan Williams used it in a recent blog post about Twitter improving its location integration. “As a dramatic example, twittering ‘Earthquake!’ alone is not as informative as ‘Earthquake!’ coupled with your current location.”

The forward-thinking scientists at the U.S. Geological Survey want to figure out what they can do with this user-contributed data, and to that end have started the Twitter Earthquake Detection Program using funding from the American Recovery and Reinvestment Act (found via Ecopolitology via The Business Insider). They look for earthquake-related keywords and map them to locations, filtering out tweets about Dairy Queen’s Earthquake brownie desserts, for example.

Yesterday’s 4.1 quake produced 296 quake-related tweets per minute, which is comparatively very high — but then, the Bay Area is also an epicenter of Twitter use.

Clearly, the USGS already has tools for predicting and measuring earthquakes, and Twitter isn’t going to improve them (unless perhaps they start monitoring tweets about dogs acting funny). Twitter is most helpful in less populated areas, according to USGS Seismologist Paul Earle, which often also have fewer measurement instruments, and data about earthquakes takes longer for scientists to verify and publish. The data will also improve dramatically as Twitter users start sharing their location more specifically through tools like Twitter’s geo-tagging API, as opposed to just their home city in their profile.

That’s when this project could be really useful — collecting precisely located first-person accounts (and potentially even photos and videos as well, especially for larger quakes).