April 17, 2013

Mapping the Boston Marathon Bombing

The tragedy in Boston this week shook us all. Several of us have strong ties to the area and the randomness and sheer viciousness of the event is stunning.

We noted that many people felt similarly and many took immediately to social media (such as Twitter) to participate in a larger discussion. Some used it to assure loved ones that they were OK while cell phone service was spotty. Some used social media to spread misinformation for personal gain or to make a political point. So too did the Boston Police and Fire Departments rely on social media to get a better idea of what actually happened. But the focus on social media's role in responding to the bombings neglected the intensely geographic element of such user-generated content as individuals and society tries to make sense of it all.  Thus, in an effort to document the diffusion of spatial awareness of the tragedy we offer the following analysis.

Using DOLLY, we collected all geotagged tweets in North America referencing "Boston" from March 1, 2013 through April 15, 2013. We've divided the data from the last month and a half into three separate temporal snapshots: from March 1 to March 31, from April 1 to April 15 at 2:45pm and, finally, April 15 from 2:45pm to 11:59pm, roughly the time following the first explosion on Boylston Street near the finish line of the race. While the visual differences in the maps below may be somewhat subtle, the data behind them is anything but.

For the entire month of March 2013, there were a total of 48,622 geotagged tweets with reference to "Boston", of which 44,221 had exact lat/lon coordinates. Of the 48K+ tweets, nearly half (23,895) of them were within Boston's city limits [1]. A fairly similar pattern was evident in tweets in the first half of this month, with 24,991 tweets total (23,151 had lat/lon coordinates attached) and 12,206 in Boston. These general trends were evident in earlier data as well, especially with respect to the pattern of roughly half of the references to the city being located within it.

References to "Boston" in the Continental USA, March 1 to April 15 [2]

But in the time since yesterday's bombing, tweeting activity about Boston has both intensified and dispersed. After 2:45pm EST on the 15th, there were 52,339 tweets in our dataset -- that is, several thousand more tweets in roughly nine-and-a-quarter hours than there usually are in an entire month, indicating an expected spike in overall activity as a result of the news coverage. But of these, a lower percentage (83.6%, as compared to 90.9% for March and 92.6% for the first half of April) were geotagged with exact lat/lon coordinates. And, perhaps most interestingly, an incredibly small number of these tweets originated within Boston.

Whereas roughly half of the tweets about Boston originated there in the earlier time frames, only 3% of tweets were located within the city following the bombings. All of this remains in stark contrast to the numbers from last year's Boston Marathon, where there were only 775 total mentions of the city in geotagged tweets from North America, with 333 (again, close to half) within the city. So not only was there a considerably smaller amount of geotagged tweeting, but so too did it remain concentrated largely within the city.

References to "Boston" in the Greater Boston Area, March 1 to April 15

In addition to the overall intensification of discussion about Boston in the wake of the bombing, there are a couple of distinct spatial patterns at play here. First, yesterday's tragic events led to discussion of Boston on Twitter to become much more spatially diffuse around the country. This is likely the result of a combination of things: people within the city tweeting less due to concerns for their own safety, people within the city not feeling it necessary to include "Boston" in all of their topically-relevant tweets, and a heightened interest nationwide in what is just the latest in a long string of violence in recent months.

But second, discussion of the city within the city is also more spatially dispersed. While the time frames prior to the bombing demonstrate a massive concentration of tweets in Downtown and the Back Bay -- the areas in closest proximity to the bombings, as well as some of the more densely populated during daytime hours -- tweeting activity after the bombings shows less focus on these areas and a more random spatial distribution throughout the greater Boston area, though these areas maintain the highest concentrations.

This analysis shows how established spatial patterns of place-based social media activity can be disrupted by extraordinary circumstances, such as a terrorist attack, as well as the importance of looking at how such spatial patterns change over time [3]. While there remains more one could do with this data -- including a focus on tweeting activity within particular spaces of the city near the bombing or looking beyond particular keyword searches, or using social network analysis to understand the spatial and temporal diffusion of the tragic news -- these maps and statistics provides an initial look at how tragedies such as these and the outpouring of emotions about them result in shifting geographies of social media activity.

[1] The greater Boston area -- including Cambridge, Somerville, Brookline, Newton, etc. -- were excluded from these counts for reasons of convenience.
[2] Note that both maps are in reverse chronological order, with the post-bombing time frame shown at the first in each series.
[3] There are also some important and potentially anomalous patterns relative to some of our earlier findings but this awaits further study.

1 comment:

  1. This is great stuff!!

    If interested, we would love to talk with you.


    Nigel Jacob
    Co-Chair, Mayor's Office of New Urban Mechanics, Boston


Note: only a member of this blog may post a comment.