How social media can be used as a source of Volunteered Geographic Information for disaster risk management?
Nov 11th, 2014 by j0p
By Felix Bühler and Jan Sichau
Introduction
Nowadays social media plays an important role in disaster risk management. Information from platforms like Twitter can be searched, extracted, and used for gathering a huge amount of information in a very short time and furthermore these information are linked to a specific position, from where the information was published.
Therefore incidents like natural hazards can be threatened in a very fast way, as well as other incidents like counter terrorism, pandemics and other emergencies. With such an amount of information Twitter offers, there has to be a way how this mass of data is going to be managed. In this specific case, the system of the CCC, the Australian Government Crisis Coordination Centre, is described as an example of how a system is built up.
Crisis Coordination
The aim of the crisis coordination is to “deliver the right information to the right people in the right format in the right place at the right time” (Cameron, M., Et Al. 2012). In order to reach this goal, the Endsley model of situation awareness plays an important role. For this huge data coming up, there is no possible way to manage it manually. A computer is needed in order to process the steps.
The first step is the situation awareness. It is split up in three levels: First of all, Perception. This means accumulate data based on information such as position, keywords, reposts, etc. In the second level (Comprehension) the classified data needs to be understood. For example the difference between a current earthquake and tweets from a famous football match. The last level (Projection) is going to determine whether there will be more upcoming data of this specific case or not. In the second step it is important to decide whether this accumulated information is important for the CCC or not. Known keywords or regions where natural hazards are going to happen can influence this decision making. The third and last step is the action on how this incident is going to be managed and how this action will be performed.
By processing this model by computers, five problems occur, which need to be solved, since a computer can’t think and therefore can’t decide by themselves which incidents are important ones.
Problems
1. Automatically find unexpected incidents or information, such as uncommon tweets
2. Cluster these information without getting the awareness of the specific incident
3. Understand and value these clusters and as a result getting the awareness of the specific incident
4. Observe the incident. This means, getting information about the location, the duration and issues coming up with the incidents
5. Analyse the incident chronological. What happened before, during and after the incident?
How these problems can be solved
The CCC has its own system to determine these problems, called ESA-AWTM, the “Emergency Situation Awareness – Automated Web Text Mining”. This architecture has several engines to solve the problems stated above. For the first problem, it executes a burst detection method, which means collecting same keywords, same location or other similar attributes and even visualizes them in an alert monitor. With this visualization the second problem can be solved through a clustering taking place. The watch officer, meaning someone who supervises the visualizations, now sees the different burst messages in a condensed way and not only the single messages. He can therefore draw a conclusion about the potential incident. While clustering a set of information, there still might be more or less useless information shown in the visualization, which is weighted in the solution for the third problem. Information about the strength or the more accurate position about an incident is more important than for example a personal opinion about the event. The ESA-AWTM solves to problem four by managing the incident. This means merge or split information-cluster, tracking them or even delete useless information. For example the Fukushima nuclear disaster resulting from the Tōhoku earthquake. For the last problem a Historical Alert Viewer is implemented. This means, that every incident is stored in a database which can be accessed, even offline. It can take therefore an important role by comparing current incidents with past ones and draw a conclusion about the strength and dimension in the past.
With this system some problems occur. For example legal aspects as Twitter’s Developers Rules of the Road which lead to an architecture based completely on the intranet (company-own network).
The system of Situation Awareness from Twitter does not replace older methods for crisis management, but is a huge improvement in the early awareness of an incident.
Discussion questions:
- Can this data be manipulated and is this a serious risk?
- How can the system be improved in extracting different information?
- How can this system be implemented in areas with limited access to social media
Sources
Cameron, M. A., Power, R., Robinson, B., & Yin, J. (2012). Emergency situation awareness from twitter for crisis management. In Proceedings of the 21st international conference companion on World Wide Web – WWW ’12 Companion (p. 695). New York, New York, USA: ACM Press.
Yin, J., Lampert, A., Cameron, M., Robinson, B., & Power, R. (2012). Using social media to enhance emergency situation awareness. IEEE Intelligent Systems, 27(6), 52-59
Thank you for your post. It is really well structrued and I could understand clearly the topic. Neverthelesse I have a question.
As I concluded from the mandatory reading, social media as a source of VGI for DRM is nowadays in practice not really important. It is an option for the future to improve DRM. Do you think the same? Because you wrote “[...] but is a huge improvement in the early awareness of an incident”.
If you do not think like I do, you might convince me with an applied example, if you want :-P
Also I am not really sure, what to think about the system ESA-AWTM. For me it sounds a little bit like a data leech, which scares me. But maybe I am just of the old school…
Hey Dirk,
I’m not quite sure, if i understood your question properly. Social Media is not the only source for VGI. But it is still important, maybe not that important like “real” volunteered information (like we discussed at the weekend, about active and passive collected data).
I don’t think it’s an option, I think it’s mandatory. At the end of the blogpost, I wrote that Twitter is important for the early awareness of an incident. It’s not written in one of those suggested readings, but I remember from I think the Taifun Hayan, that the first information about the earthquake at all came from twitter tweets. So therefore it’s important for areas with no (or a bad) early-warning systems, since the information is available only some seconds after the incidents.
To draw it up: Social Media is mostly a passive source and therefore maybe not the best one, but it’s a very fast way for quick up-to-date information.
I hope your question is answered. If not, let me know :)
The CCC/ESA-AWTM is quite concerned about the privacy issue. But as you can read,the are some limitations made yet, like the forbidden republishing of gathered tweets and the censoring of names. And it only takes information, which is available in public, so information everybody can search for.