The proper response to an earthquake? Run, scream, take cover?… no wait, Tweet!
On Tuesday, the denizens of the East Coast had exactly this choice, and they responded by flooding the interwebz with messages: startled, mundane, humorous, informational. And it happened fast. Seismic waves travel at 3-5 km/s, communication signals in fiber optic cables move at a speed of 200,000 km/s [as this XKCD cartoon brilliantly notes]. Tweets do take time to compose, but significantly less when you’re tweeting “EARTHQUAKE”!
We thought you’d like to see some of the data behind it. The visualization below replays the spread of earthquake related Tweets across North America, from the moment the epicenter hit Mineral Virginia (1:51PM) on August 23rd through its spread across the East coast and the South.
In the first couple of seconds, Tweets come from across Virginia, while at the same time a number of users were still responding to the previous day’s Colorado quake (green circles out west). One of the first people to Tweet about the quake was @b_mc817, posted a mere 30 seconds after the earthquake began, it is most likely that this user was typing as the tremor was taking place:
Here’s how it looks at the very beginning:
Some 30 seconds later, thousands of people are tweeting around the DC area. At this point in time, we’re also starting to see people outside the area currently affected by the quake, tweeting about it. The green circles represent people who have spotted an earthquake related Tweet most likely through their stream, and within 30 seconds have already posted a response!
First Tweet from Baltimore comes at 1:51:52, Jersey at 1:52:23 and New York City at 1:52:26, as Twitter user Amanda (@_ambassador) posts from Bushwick:
Here’s the interactive version of the map. Use the slider to go back and forth in time:
(viz works only on Chrome/Firefox)
Each circle represents a single Tweet, and its color, the distance between the epicenter and the location the Tweet was published. Locations are based on what people place in their public profiles. The expanding white circle is an estimation of the actual earthquake spread. For this data sample, we extracted some 90k Tweets in total that were published within the first 5 minutes of the earthquake. For more than half of them, we have fairly good estimates of location.
Humans as Sensors
While this is not the first time that Twitter has been used in realtime during an earthquake, the importance of this event lies in the fact that 1) it hit a major metropolitan area, and 2) it was relatively mild, meaning no communication lines were harmed. Within seconds of the earthquake hitting VA, we see hundreds of people across the States passing information. There’s a clear 40-50 second warning signal between the very start and the New York City region. This signal manages to reach tens of thousands of people before a minute is over, in effect, a network of human sensors that not only identifies a substantial event, but also passes on information in remarkable ways.
By plotting time of first identified earthquake tweet per unique identified location vs. distance of that location from epicenter, we can see a clear clustering form around the earthquake’s path. The points further away (above the red line) represent people responding to those experiencing the quake. The slope of the line aligns well with the speed at which the quake is estimated to have spread (around 3km/s).
Taking over the Conversation
Another important lesson that this event reinforces is just how quickly a new topic can sweep over and claim people’s attention. Within seconds, Twitter streams filled up with status updates about the earthquake. First there were the shocked responses, then informational ones, eventually evolving into humorous takes. Mainstream Media followed, covering miniscule details of the story, as long as people were willing to pay attention to it. On Twitter we see a clear and massive spike in the number of people posting content about the earthquake. In the chart below the term ‘earthquake’ drastically spikes to tens of thousands of shares per minute. It clearly dwarfs even the usual daily chatter (including words such as ‘hello’ and ‘like’).
Events like this week’s earthquake highlight the incredible speed at which information will flow, especially when there’s an event that touches so many people’s lives. From cases such as this, we learn just how volatile conversations and in general topicality can behave within social networked spaces. Whatever people had talked about previously dissipated with the onslaught of perspectives and stories from the earthquake. One cannot demand nor schedule to receive other people’s attention. This is something we care deeply about at SocialFlow and are working hard to analyze, quantify and predict.
PS – Big props to our recently graduated summer intern Andrew Lott who spent the summer visualizing our data on maps. This widget is heavily based on his summer work!
PPS – And finally, here’s a link to full screen version of the visualization (warning: heavy to load!)