Bird's Eye View
01 February 2011
Print Issue: February 2011
Last year in Oxford, England, at a TED Conference—a gathering where innovators share ideas—presenter David McCandless, author of Information is Beautiful, showed the audience a traditional dual-axis graph covering the 12 months of the year. It depicted a metric trending upward during springtime, then plummeting to zigzag steadily through the summer months before climbing upward toward the December holidays. The data: status changes and updates from social networking site Facebook that indicated dating breakups, from “spring cleaning” to the safest day of the year to be in a relationship: Christmas.
“There’s a titanic amount of data out there—unprecedented,” McCandless said. “And if you ask the right kind of question or you work it in the right kind of way, interesting things can emerge.”
The substance of the chart may seem trivial, but McCandless’ broader point should resonate with security professionals, whether they use Facebook to communicate with friends and loved ones or consider Twitter a bewildering, silly, or annoying concept.
In particular for those responsible for emergency response and continuity of operations, open-source social media can provide an invaluable resource for maximizing real-time situational awareness. This was demonstrated during numerous events last year, from the catastrophic earthquake in Haiti to the September wildfires in Boulder, Colorado.
The sheer scale of data, however, and questions about its veracity, pose challenges. Practitioners, programmers, and crisis mappers alike are still working to address those concerns. Meanwhile, emergency management practitioners are getting the most out of two specific social media tools: microblog Twitter and photo-sharing site Flickr.
On Twitter, registered users can post updates of up to 140 characters and can view feeds aggregating the posts of other users they “follow.”
Several factors make Twitter useful for gathering and disseminating information. First, most of Twitter’s roughly 220 million users leave their profiles open to public view online, even by nonregistered users. Second, Twitter is searchable using Boolean logic—where one searches by saying find A and B but not C; for example, “Jackson and Andrew but not Michael.”
Third, Twitter’s application programming interface (API), the means by which other software interacts with the site, is open. That allows the world’s technorati to program tools to harness its data, from user dashboards like TweetDeck to specialized search and sentiment-analysis functions.
These factors combine to make Twitter a good tool for spotting trends or getting live reports from people at the scene of an incident, and it has become a way to gather short news alerts from established, trusted users. It can also be a way to disseminate information.
Many emergency response professionals have used Twitter for quick searches of real-time “tweets” to assess rapidly unfolding situations. The results are chronological, which is better for such situations than a Google search, because Google will rank results based on the popularity of the source among other factors.
Let’s say a person writes a low-traffic blog, and during a disaster, that person blogs via a wireless device that they are “trapped in rubble and running out of water.” An emergency manager who searches Google an hour later for “(town name) and ‘trapped’ or ‘water’” may get results with a lot of very popular, day-old news stories about the disaster but not that post. If that same person tweets that they are trapped and out of water, and includes the address, an emergency manager who follows Twitter, or more likely a saved Twitter search feed on TweetDeck for “(town name) and ‘water’ or ‘trapped’” will instantly see that tweet pop up in a feed, with the option of an audible chime.
Federal Emergency Management Agency (FEMA) Administrator W. Craig Fugate, an early adopter of social media who is pushing broader Twitter use in his agency, tells Security Management that when he hears a report of a developing event, he takes out his smart phone and conducts a quick search, often using simply a place name and the nature of the reported event.
Users should be aware, however, that search results for a location may also include tweets posted outside the desired area by users whose Twitter profile location is within the area. Moreover, tweets are not filtered or verified. Anyone can post anything claiming to be anyone located anywhere.
Fugate acknowledges that there may be chaff among the search results, but there will also be wheat. He takes the example of searching during the early hours of the Boulder fire, before the common text hashtag #boulderfire emerged.
“If you’ve got the Boulder fire you could actually say, ‘I want to see every Tweet that’s within 50 miles of downtown Boulder that says fire.’ And I may get a band called Fire Dogs and see tweets on that, but I’m also going to get the wildfires in there as well,” Fugate says.
For the uninitiated, placing an “@” sign in a tweet followed by a user’s name references them and draws their attention by posting the tweet to his or her “mentions” feed. Hashtags, meaning the “#” symbol followed by a word, classifies the tweet’s topic or type of data being tweeted. Abbreviated hashtags like “#src” would precede the source attribution of information, and “#loc” the location of the poster or the matter at hand. For example, a coworker might tweet “@joestraw power outage #loc 1625 Prince St. Alexandria VA #src @ASIS_Intl” to let me know the power is out at my office.
On Twitter, users have the option of including a location for both their base profile—ostensibly their home or office location—and their individual tweets. The nature of location data is at the discretion of the user: town, address, or even 10-digit GPS grid coordinates can be used if they choose to find and include them.
Tweets can be tagged for location either within the text—as described—or in optional metadata. With Twitter’s location-based functionality activated, users must search for a municipal location name, then either select or create an actual “place” name, such as a business or public place, to tag the tweet.
One of the first major international events in which Twitter played a notable role was the protests that surrounded the 2009 elections in Iran. While the event demonstrated the medium’s power, it also exposed some of its shortcomings. It led to some refinement in searches that are still useful now.
For example, at the height of the election protests in Iran, even major media outlets credited Twitter as a primary source of information. But that event occurred before Twitter launched its location-based functions. Thus, a quick search of the hashtag #iranelection by users looking for dispatches from the streets of Iran would result in frustration. Most of the tweets with that hashtag were messages expressing support for the protesters, while many others bore spam links.
Since then, Twitter has made vast advances in filtering out spam, says Kate Starbird, who is a Ph.D. candidate in crisis informatics at the University of Colorado at Boulder and a volunteer with Empowering the Public with Information in Crisis (Project EPIC).
Even without spam, the volume of relevant tweets can be overwhelming during a major event, so finding the most useful tweets requires advanced searches. For example, last November, a search for tweets with the hashtags #haiti and #cholera generated a variety of results, among them charitable appeals, statements of sympathy, and germane, albeit self-promoting links to blog posts. The solution might start with an advanced search that blocks posts with the words “donate,” “thoughts,” “prayers,” and “post.”
Searches are further aided by free dashboard applications like TweetDeck. Within the program, users can create customized feed columns based on saved searches. Richard Barley, TweetDeck’s community manager, acknowledges, however, that the application does not handle location-based searches, because Twitter has not yet allowed that functionality, “though it is something we very much hope to be able to add soon.”
Even a refined search feed on Twitter may look of little use to the untrained eye. Familiarity, however, and the emergence of informal standard operating procedures for use of the medium can help.
It’s a bit like listening to a police scanner. As one acclimates to the lingo, the meaning becomes clear.
Laura Madison, an independent researcher who surveyed social media use by law enforcement officers for the Canadian Association of Police on Social Media, closely followed the Boulder fire online. She agrees that as you become an experienced “listener,” on Twitter, you learn to filter out the extraneous chatter and tune in to the critical information.
This process is aided by common formats and terminology, which leaders in the volunteer community have sought to expand. An effort led by Starbird, called Tweak the Tweet, seeks to establish a standard set of hashtags to communicate clearly using a minimum of characters. Hashtags in the project’s lexicon include: #trapped or #injured, followed by name and location; #need, followed by resource; #have or #offer, followed by resource; and #imok, followed by name to show “I am okay”; and #ruok, followed by name, to ask “Are you okay?”
Similarly, the American Red Cross, which recently held a summit on the emerging issue of social media in disasters, supports efforts to market specific tags, though it does not endorse any, says Wendy Harman, the organization’s director of social media. “We’re not trying to dictate that process,” she told Security Management. “The people are going to pick the best way to talk to one another, and we just want to listen to that.”
Of course, people in danger during disasters might not be aware of these agreed-upon standard tags, which is why focused, saved searches for terms like “help” or “water” are also critical.
The functionality of Flickr, a photo sharing site owned by Yahoo! Inc., is more straightforward than Twitter’s. Registered users can upload digital photographs to their account. They can also organize and tag pictures.
While sites like Kodak Gallery are more commonly used for private sharing and purchasing of prints, Flickr is widely used for public sharing; it has more than 40 million people posting to the site and 5 billion images. Captioning and tagging of the site’s publicly posted photos allows easy searching for specific content, which can prove a useful resource for first responders seeking situational awareness in an emergency.
Users can also put in location data, which can be at the level of the user’s profile and tied to individual photos, with each type of location data searchable by others who visit the Web site.
Emergency managers consider open-source, searchable photographs from sites like Flickr a reliable source. Not only is a picture worth a thousand words, it is more reliable, because while pictures can be altered, they are more difficult to fabricate than textual lies.
The newest, and perhaps the most controversial, functionality in social media is also the most valuable to event and emergency managers: geotagging using cellular location or GPS data generated by a wireless device. This goes a step beyond a user adding location-based information as discussed in the context of Twitter and Flickr postings and profiles.
Due to the clear privacy concerns, geo-tagging is always an opt-in functionality. Neither Web 2.0 sites nor wireless devices default to include it in social networking data.
Photos can be geotagged through a variety of means. Newer smartphones with cameras, like Apple iPhones, query users before they take each photograph to ask if they want to geotag it. If that feature is enabled, the GPS coordinates are then included in the picture’s metadata. That functionality is also present in some of the newer digital cameras.
As of last year, only 4 percent of adults who used the Internet also used mobile applications to share their location. But even that small percentage can be helpful in spotting trends in emergencies, says Adam Geitjey, CTO of the Tucuxi software development unit of ESi Acquisition, which markets the Web-based emergency management tool WebEOC.
“It’s just starting to be enough people to gain a critical mass of this data that trends emerge from,” Geitjey tells Security Management.
Nearly all major emergency management agencies already use geospatial information systems (GIS)—electronic maps—in command and control, with applications like WebEOC and ArcGIS software from Esri.
The Alabama Department of Homeland Security’s (Alabama DHS) Virtual Alabama is one of the country’s most advanced GIS applications for situational awareness, built to provide the state with a common operating picture. Built on the Google Earth platform, the system incorporates data ranging from critical infrastructure like utility lines to flood plains to the floor plans of public schools.
The next step is to make these electronic maps more dynamic and useful for situational awareness purposes via an overlay of data points from social media through what is called crisis mapping.
Boulder’s Project EPIC, led by University of Colorado computer science professors and staffed by volunteer researchers like Starbird is one effort to that end. In its early stages, the program doesn’t seek to establish a map format for crises but rather to develop effective software solutions for synthesizing crisis data so that the information can be imported by other users based on their needs, and overlaid on their maps, whether open-sourced or proprietary.
Perhaps the most famous crisis map was generated in the hours following last year’s earthquake in Haiti by the international volunteer Ushahidi Project. In that case, emergency radio broadcasts directed emergency victims who were stranded and without other means of communication to send a text message of their emergency needs to INFO (3646). Volunteers accessed the texts and used the data to create a crisis map in the cloud.
In addition to compiling data from texts, the 3646 project also followed Twitter. At least one woman was pulled from the rubble after tweeting her location and tagging the message “3646.”
What the Ushahidi Project lacked in Haiti was automation—instant crisis mapping of the information in the message. That was impossible because of the lack of location-based services on Twitter then, and the fact that location data based on cellular tower position is far less accurate than GPS.
The first step toward automation is mashups of separate applications. Population of any Web-based map, such as in Google Maps, with photographs from a site like Flickr is the classic example of a mashup. Generally, they require specific programming based on open-source APIs. Applications are emerging, however, that allow population of Web-based maps with data like photographs from Flickr without additional programming. On the free site iMapFlickr.com, for example, users can simply login using their Flickr/Yahoo! account, then follow cues to create a map populated with photos of their choosing, based on variables such as keywords, from “vacation” to “#boulderfire.”
FEMA has one of the first software applications capable of mapping tweets by profile location and geotags, Fugate says. The agency applies this functionality within its application of ArcGIS.
Data gathered widely from members of the public via social media pose clear concerns. In addition to concerns about the credibility of individual reports, government agencies like FEMA must consider the privacy implications of tapping social media.
Fugate says that neither he nor his agency are interested in individual messages or tweets during a response, which limits concerns about privacy and the credibility of individual posts. As Jeff Baryani of Esri says, “You don’t want to see the trees. You want to see the forest. You want to see the trends, not the individual data points.”
Fugate offers the hypothetical of an evacuation in advance of a hurricane. Mapped tweets for a region, say, the Outer Banks of North Carolina after an evacuation order, may give event managers a sense of public sentiment and may let them know that they need to change their message, he says.
“You’ll see [tweets] like, ‘I’m going to wait for the next forecast,’ or ‘I’m going to wait until morning.’ Then we would go back to our state or local officials and tell them that we don’t think the public quite got the gravity of this message, or we need to adjust our message,” Fugate says.
Many emergency managers harbor concerns that bad actors would spoof the system to divert resources or place responders in harm’s way. But officials again indicate that the volume of responses would militate against any false leads creating a problem when viewed in the context of all the data points.
Not everyone is sold on the idea of using social media for crisis maps, however. Virtual Alabama does not incorporate public Web 2.0 information and may not for some time, says Alabama DHS Director Art Faulkner. “The reason we’re not doing that is that it could overload first responders to the point where nonessential and nonpertinent behavior could be more of a hindrance than a help,” he says.
Managing the Medium
The Haiti earthquake and its aftermath made heroes of crisis-map developers like Patrick Meier and his peers in the Ushahidi Project. More importantly, their efforts showed the power and potential of virtual volunteerism. “People converge on a disaster and want to help, now they’re converging digitally,” Starbird says.
But just as with volunteers who come physically to a scene, the contributions of virtual volunteers present a mixed bag for event managers. Tery Spataro, co-founder and chief strategy officer of Boulder-area digital communications firm Orange Insights, closely monitored and analyzed the digital response to the Fourmile Canyon fire in Boulder. She found that unlike in Haiti, where one crisis map emerged as a primary resource, 12 online crisis maps were produced by volunteers addressing the fire. “It was a race to see whose map went up first,” Spataro says.
Moreover, where the Ushahidi map was used by emergency managers and relief organizations amid a void of domestic governance in Haiti, the 12 volunteer-developed Boulder maps appear to have been used primarily by members of the community, most of them not directly affected by the disaster.
Having all those competing resources made it “incredibly difficult” for officials to coordinate information leads and to control the quality of information being disseminated, Spataro notes.
Mike Chard, executive director of the Boulder Office of Emergency Management and other government stakeholders—who must exhaustively vet and verify information before releasing it—echoed the view that the multiple crisis maps complicated their jobs, especially because those maps might have data points that were not sufficiently vetted, which would then cause confusion among the general population. “People in the community were calling and saying, ‘I’m looking at it on a map, and you don’t have the information?’” he explains.
Conversely, Boulder’s crisis mappers felt the government failed to take advantage of the wealth of data aggregated in their work.
Amanda Pingel, a recent MBA graduate from Regis University in Denver, followed the Fourmile Canyon fire on Twitter but couldn’t visualize it’s spread, so she created her own map in Google Maps and opened it to public view and to editing by trusted peers. The latter started populating the map with crisis data.
“It never occurred to me that people would start putting emergency shelters on it. But I’m glad that they did,” Pingel says. “I’m surprised by how much it took off and how far it went.”
In the mappers’ view, the government could have used that type of aggregated data and shared it with people in the community who might not have known it was online or who might not have had Internet access.
The phenomenon of volunteer crisis mapping—and the online volunteer community’s persistent requests for new information—were a new experience for Barb Halpin, a public information officer (PIO) with the Boulder County Board of Commissioners.
Emergency management needs to move into the 21st century, Halpin says. “We’re dealing with a new reality, and I would say that the current external affairs structure under the Incident Command System is not working. I just took a course in ‘advanced’ PIO, and it was advanced for the 1990s: two press conferences a day and on-camera interviews,” she says.
Officials are trying to learn from the Boulder wildfire experience. After the fire, area emergency management and public affairs personnel met with area crisis mappers to share their perspectives and possibly foster better lines of communication going forward. But, Halpin noted, “The next disaster could be an entirely different group of people.”
Colorado state officials, meanwhile, intend to develop plans that incorporate social media, says Brandon Williams, a PIO with the Colorado Division of Emergency Management. “It’s something that we’re not only going to have to draw from but also start playing into,” he says.
Boulder volunteer crisis mapper Starbird agrees with that view. “Should it be the only source of information? Absolutely not,” she says. “But if it’s there, why not use the information?”