Shaking and Tweeting: The USGS Twitter Earthquake Detection Program

Marisa Lubeck: Welcome. And thanks for tuning in to this episode of CoreCast. I’m Marisa Lubeck. The US Geological Survey is investigating the use of Twitter, a popular micro-blogging tool, to collect and analyze citizen accounts of earthquakes from around the world. I’m here with USGS scientists Paul Earle and Michelle Guy. Both of whom have been working on this Twitter Earthquake Detection program, better known as TED. Welcome, Paul and Michelle.

How can Twitter, a social-blogging type tool that’s commonly used to connect friends, celebrities, fans be useful for earthquake response? What’s the connection?

Paul Earle: Well, many people use Twitter to tell others what’s going on around them. So after an earthquake, they often rapidly report that an earthquake has occurred and describe what they’ve experienced. For felt earthquakes in populated regions, Twitter reports often precede the USGS’s publically-released, scientifically-verified earthquake alerts.

01:00

By collecting these tweets, which are open for anyone to search and analyze, it is possible to obtain first-hand accounts, albeit short, of what people experienced during an earthquake.

Marisa Lubeck: What can Twitter provide that your other earthquake response products like ShakeMap and Did You Feel It can’t provide?

Michelle Guy: Well the basic difference is speed versus accuracy. In densely instrumented regions, like California, USGS locations and magnitudes are produced within two to three minutes of an event. This time increases up to 20 minutes for earthquakes in sparsely instrumented regions around the world. So comparatively, people start tweeting about an event within seconds after it occurs.

And analyzing the tweets provides an early indication of what people experience before the quantitative information provided by products like ShakeMap and Did You Feel It becomes available. So in a nutshell, Twitter augments, but it certainly does not replace current earthquake products.

02:08

Marisa Lubeck: How are you currently using the tweets for this project?

Paul Earle: Well we’ve developed a prototype system that continuously gathers the tweets containing the word “earthquake” and it’s equivalents in other languages. This database is queried after the USGS or contributing network to the Advanced National Seismic System detects an earthquake. If we find an increase in earthquake tweets near the epicenter, they are attached to the earthquake alert.

This Twitter-enhanced alert contains a summary of the cities where the tweets originated and the text of the tweets. We also create an interactive map of the tweets. This system enables us to receive short personal accounts from people who experience the shaking at about the same time we pre-release what have only received the location and magnitude.

Marisa Lubeck: So Twitter seems to be a potentially beneficial tool for earthquake response. But what sort of benefit do citizens get from tweeting about earthquakes?

03:01

Michelle Guy: So after experiencing an earthquake, people will often search Twitter to see what others experienced. And for small quakes, people are often simply trying to just confirm, "Was that an earthquake or was that a really large truck driving by?". But for larger earthquakes, people might search Twitter to get an idea of how many people or how large of an area was impacted. And I think there’s some amount of comfort in knowing you are not alone.

Marisa Lubeck: Are there any drawbacks to this system or kinks that still need to be worked out?

Michelle Guy: Well one drawback is that not all Twitter users provide locations. Or they may provide only a city such as Los Angeles as their location. So this information is very vague compared to a tweet that comes from a GPS-enabled mobile device which would provide a latitude-longitude Twitter accurate within city blocks.

04:02

Another related kink is also trying to determine if a user location is only the word “Ontario”. We have to determine if that refers to Ontario, California or Ontario, Canada. These very indiscernible user locations end up making the tweets not usable in searches for related events.

Another kink that we have in the system is out-of-context tweets. So for example, there’s often tweeting about enjoying Dairy Queen’s Earthquake brownie desserts or maybe even a good game of Quake. Once we are aware of these kinds of trends, then we can try to filter out tweets that might, for example have both the words “earthquake” and “brownie “in them.

Marisa Lubeck: What is your future plan for the project?

Paul Earle: We’re currently focusing on two areas. First we are determining the best way to broadcast USGS scientifically-confirmed earthquake alerts via Twitter. This is not as easy as it seems because the USGS publishes the location and magnitude for about 50 earthquakes a day. And a typical Twitter user will be overwhelmed by this many tweets.

05:15

So we are developing a way to sift through the earthquakes and present only the ones to the public that they might be interested in. And second, we are investigating if we can actually detect and characterize earthquakes directly by monitoring the earthquake data. For earthquakes in sparsely instrumented regions, these detections could provide an initial heads up that an earthquake may have occurred.

We have shown that this can work in specific cases but since Twitter only allows users to search the last seven days of data, doing a comprehensive world study will take some time. To complete the study, we have begun gathering and storing earthquake tweets from around the world. The public should know that the USGS rapidly detects and reports accurate earthquake locations and magnitudes for all widely-felt earthquakes in the United States and earthquakes that may have caused widespread damage anywhere in the world.

06:01

So a Twitter-based system will not help us uncover missed events missing these criteria. But what such a system may do is help fill the information gap between the initial shock and the distribution of quantitative and confirmed information.

Marisa Lubeck: How can citizens get involved in this system? And where can they turn for more information?

Paul Earle: Well, citizens are already involved in this system. Without their tweets, we would have no system. So they can continue to contribute by rapidly reporting their experiences through Twitter. People also can follow the progress of this specific project by following USGS TED on Twitter.

One difficulty Twitter users often have is determining which information sources are authoritative and which are not. To help users to determine which sites are officially USGS sites, the USGS has developed a website that lists all our social media contributions. This can be looked at to see what the USGS is currently producing in the world of social media.

06:57

Marisa Lubeck: And the URL for that website is www.USGS.gov/socialmedia. Thanks to Paul and Michelle for speaking with us today. This podcast is a product of the US Geological Survey Department of the Interior. I’m Marisa Lubeck. Thanks for tuning in.

Related information:

• @USGSted account on Twitter
• DOI Recovery information on the USGSted project