Article Preview
TopIntroduction
In this paper, we present the evaluation of an automated system to extract and process volunteered geographic information (VGI) from social media networks during crisis events. To this end, we conducted two case studies on forest fires in France, both with encouraging results.
The rationale behind our approach is as follows: The development and adoption of integrated mobile devices like smartphones and their connection with social networks has changed the way information is collected, distributed and used. In the past, information flowed only in one direction, from the “top” of administrations and government agencies to the “bottom” of the broad public. The main communication channels were traditional broadcasting media like newspapers, radio or television. Horizontal peer-to-peer communication was very limited due to the small reach of the available media (word-of-mouth and letters, later slightly improved through telephones and e-mail). This has changed with the emergence of new information and communication technologies, commonly referred to as the Web 2.0 (O’Reilly, 2005). The Web 2.0 services and platforms have given the public the opportunity to share freely various media through general social networks (Facebook, Google+) or more focused platforms, including text (Blogspot, Wordpress, Twitter), images (Flickr, Picasa, Panoramio), videos (YouTube, Vimeo) and maps (GoogleMaps, GeoCommons, MapBox), thus enabling every user to seek and provide information and experiences. This development has also had an impact on the vertical flow of information, with public administrations and agencies adopting the new communication channels to disseminate information to and retrieve information from the public (De Longueville, Annoni, Schade, Ostlaender, & Whitmore, 2010; Palen & Liu, 2007; Puras & Iglesias, 2009; Roche, Propeck-Zimmermann, & Mericskay, 2011).
We can expect that these trends continue in the near future. With increasing wireless Internet access and availability of satellite-based navigation systems such as the Global Positioning System (GPS) in consumer communication devices, a related trend is highly likely: That the amount of public VGI will increase manifold during the coming years.
Several studies have already investigated the impact of (geo)social media during crisis events and show the value for relief workers or coordinators, and the affected population. Examples include wild fires in the United States and France, hurricanes in the United States, the 2010 earthquake in Haiti, and floods in the United Kingdom (Al-Khudhairy, 2010; Bressler, Jennex, & Frost, 2012; De Longueville, Smith, & Luraschi, 2009; Hughes & Palen, 2009; Liu & Palen, 2010; Schade et al., 2011; Sutton, 2010).
However, these studies have also shown that until now these newly created information back channels do not yet integrate well with traditional established emergency response protocols. The two main challenges for integration of VGI into disaster response are the large volume of information, and its unknown and potentially low quality. Consequently, in a crisis management context where inaccurate or incomplete information can have dire consequences including the loss of human life, skepticism by practitioners is understandable. Although networks of volunteers have demonstrated repeatedly the ability to curate a large amount of information reliably, this approach itself faces two main challenges: Sustainability and scalability. There is no guarantee that for any given crisis event, there will be a sufficiently numerous volunteer force. Further, it remains doubtful whether the volunteer approach scales well to the expected increase of information volume.
For these reasons, we propose a semi-automatic approach to filter VGI and evaluate its quality, and thereby render it useful during any crisis event. Human supervision of such a system will remain crucial, because VGI is intrinsically heterogeneous and unstructured: It originates from different persons, using different media such as photographs, text, or video, and authors often overcome device and software limitations in imaginative and unpredictable ways.