On 9-11 February 2015, the city of Jakarta flooded both in reality as in cyberspace. While water flowed through Jakarta’s streets due to heavy rainfall, cyberspace flooded with almost 728,000 Twitter messages related to the flood in the city, peaking at 893 tweets/minute. These messages were harvested using keywords like the Bahasa Indonesian word for flood ‘banjir’. Data mining showed that 2.200 unique tweets that were sent during the flood contained information on the local water depth. Out of these tweets, 40% contained textual location information, resulting in a total of 888 unique geo-located water depth observations. In Jakarta, the flood observations from twitter, aggregated to district level, were displayed on a map in the Flood Early Warning System of Ministry of Public Works in Jakarta, offering real-time insight in both the temporal and spatial behavior of the flood.
Jakarta is a prime example of the large amount of actual information on current floods that can be found on Twitter. The city is also referred to as the twitter capital of the world with 2.4% of the World’s tweets coming from Jakarta. However, Jakarta is not the only city where observations about floods are posted on Twitter. Some examples of the total number of tweets during recent flood events are the Philippines (85.000 tweets); Pakistan (82.000 tweets); South-Korea (50.000 tweets) and Detroit (20.000 tweets).
Flood mapping based on social media
It is recognized that hazards leave a footprint on social media. Twitter data has already been used to gain information about the social impact, the temporal progression and the relative severity of a hazard. However, the Jakarta cyber flood showed that Twitter data also contains information about the physical characteristics of hazards, such as flood depth and the location of flooding. After validation, these observations in combination with a Digital Elevation Model (DEM) could be used for real-time mapping of the extent of the hazard, the population exposed, and damage caused. A proof of concept for this method was presented at the EGU General Assembly in April this year.
Disaster risk reduction
Twitter data should be considered a valuable additional source of information for disaster risk reduction. During floods, the data can be used for decision support by providing a picture of what is happening on the ground and after the waters recede, to assess damage and recreate the chronology of the flood for future planning. The challenge is to mine tweets with flood observations and to validate those tweets.
Currently, Deltares and FloodTags, two Dutch organizations, are further exploring the possible use of flood observations from Twitter for disaster risk reduction. The joint research focusses on two subjects: 1) validation of flood observations from Twitter and 2) translating these observations to flood maps.
Dirk Eilander is a Flood Risk Analyst at Department of Inland Water Systems, Deltares, Delft, Netherlands. The work of Dirk and his colleagues on cyberfloods has been truly innovative in the move towards better flood preparedness and response. You can find out more about this groundbreaking project here.
Dirk Eilander1*, Jurjen Wagemaker2, Arnejan van Loenen1
1Department of Inland Water Systems, Deltares, Delft, 2600 MH, The Netherlands.
2Floodtags, The Hague, 2516 BE, The Netherlands.