#Sandy: Social Media Mapping

Hurricane Sandy was the largest Atlantic hurricane on record, devastating parts of the Caribbean as well as the US East Coast. The loss of human life, displacements of families and damage to homes and businesses on the East Coast was unprecedented. During the storm and its aftermath, social media has been critical in spreading important information and mobilizing relief efforts. Some argue that Twitter was more accessible and far-reaching than any TV network, with over 20 million Tweets posted during the height of the storm.

Thanks to the wide range of questions posted by folks on Twitter, In this post we try to examine as many as we can. We look at prominent hashtags, urls, devices used and topics shared over the course of the past week around Hurricane Sandy. We show a map of user locations as they lose power in their homes. Finally we discuss a case where misinformation spread and how the network caught on incredibly fast.

Blackout map based on Tweets (interactive version below)

General Stats

Both Twitter and Topsy reported total number of Tweets posted during the storm. Depending on which words, hashtags and time period one looks at, the total sum is different. When compared to the debates from the past couple of weeks, even though the #Sandy hashtag didn’t spike as high as the debates, it drew more sustained levels of attention, as can be measured by the total area under the curve.

Each blue spike above represents the main Twitter debate hashtag (three presidential and one vice presidential debate), while the green represents #Sandy. Notice how quickly attention given to each debate dies down, compared to hurricane Sandy, which sustains for multiple days.

If we look at other prominent hashtags that trended across Twitter during the same period, we get the following chart:

Again, this is not surprising. We see memes such as #ToMyFutureSon that spike much higher, but also die much faster compared to #Sandy. Notice the black curve to the right, representing the relief hashtag – #SandyHelp – that has become popular since November 3rd.

When we dive into the content of the tweets, we can identify the top Sandy-related hashtags: #Sandy, #HurricaneSandy, #Hurricane, #RomneyStormTips, #FrankenStorm, #StaySafe, #ThanksSandy and #FuckYouSandy. Here’s a visualization of the prominent hashtags and user mentions during the height of the storm (for visual purposes, I left out the #Sandy, #Hurricane and #HurricaneSandy hashtags):

Colors represent graph modularity, highlighting clusters in the graph. Words with the same color appear much more with each other. In this case, there’s a significant portion of the conversation focused on direct response to the storm, weather, rain and blackout.

Alternatively, if we dive into the most prominent Sandy-related hashtag on Nov. 3rd, #SandyHelp, we see the conversation wholly shaped around relief work. The Red Cross is cited often, as well as numerous celebrity accounts urging folks to donate and help.

In dark green at the center of the graph, we see hashtags such as #jerseystrong, #jerseystrong and #restoretheshore being heavily used. @Springsteen, @billyjoel and @bonjovi were some of the most central celebrities (orange cluster, bottom right).

Apps and Shared Links

When we look at apps used by Twitter users, we see the following distribution:

iPhone, web, Android and Blackberry account for the majority of tweets posted during the height of the storm. Instagram sees quite high usage – 2.3%, especially compared to the distribution of apps used during the recent presidential debate, where iPads and Tweetdeck were much more prominently used. 

The top 5 urls shared at the height of the storm (Oct 29th-30th) were:

  1. Sandy LiveCam
  2. Mashable article on identifying fake Sandy pics
  3. Old picture of soldiers guarding memorial during a storm
  4. Brazilian model posing in front of hurricane carnage (unsurprisingly, this quickly turned into a meme)
  5. Jane’s carousel submerged in water


Mapping Blackouts using Twitter

While the majority of tweets during the storm came from mobile devices, only a tiny fraction of them were geo-coded. However, many of the non-geocoded tweets were posted by users who filled out their location field. In some cases this free text field is filled with random text or jokes (“yo mama’s house” is a fairly popular user location for Twitter users), yet for over 40% of the users we can make pretty good guesses as to which city, state and country they’re located in.

Below we plotted instances of people in the East Coast Tweeting about losing power. This begins on the evening October 28th as people mostly joke about the prospect of potentially losing power. As the storm evolves, the tone turns much more serious. The darker a region on the map, the more aggregate Tweets about power loss that were seen for that region.

The potential for mapping out this kind of information in realtime is huge. Think of generating these types of maps for different scenarios – power loss, flooding, strong winds, trees falling. While basing these observations on people’s Tweets might not always bring back valid results (Someone may jokingly tweet about losing power), the aggregate, especially when compared to the norm, can be a pretty powerful signal.


One of the big stories that came out from social media coverage during the hurricane was that of the spread of misinformation. There were a number of fake or old pictures which were heavily shared. Alexis Madrigal wrote a great article on sorting the real Sandy photos from the Fakes. This tumblr blog also does a great job at documenting the real vs. fake photos shared on Twitter.

Yet the most covered event was probably @ComfortablySmug’s false Tweets about flooding in the New York Stock Exchange and ConEd workers trapped in a facility. One of the challenges when covering such events, especially as information comes in so fast, is source verification. And unfortunately, in this case, a pieces of false information managed to spread through the network, even amplified by very visible accounts such as @CNNweather and the @NewYorkPost.

When we plot the number of retweets over time (below), we can observe a couple of interesting facts.

  1. @ComfortablySmug’s misinforming tweet about the NYSE floor flooding gets substantially less retweets than the other fake photos (shark, liberty w/crazy clouds). This is not surprising, as visual content tend to be more spreadable.
  2. The actual misinforming Tweet that both @ComfortablySmug and @CNNweather sent out peak at significantly less RT’s compared to the correction posted by @BreakingNews.
  3. The corrected information was posted within an hour of the misinforming Tweet.

Even though @CNNweather’s misinformed tweet drew more retweets than their correction, there were numerous profiles such as @BreakingNews and @NYSEEuronext who quickly posted the verified information. Based on our data, we don’t know who actually saw both the fake information as well as the corrected information, the correction was both speedy and made visible by a number of prominent accounts. So while there is indeed danger in consuming false information on Twitter, there are clear strategies that can be taken to minimize the chance of this happening (see Craig Silverman’s work on verifying social media information). The more visible the account, the more responsibility it has in making sure information posted is truthful.



If you’re around the NY/NJ region and have some spare time on your hands, consider volunteering. Two informative links with updated information can be found here and here. Otherwise, please consider donating to the following causes: Rockaways, Brooklyn Community Foundation, CraigConnects or The Mayor’s Fund.

Questions? Find me on Twitter – @gilgul

4 comments to #Sandy: Social Media Mapping

Leave a Reply




You can use these HTML tags

<a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>