Using Twitter to locate floods
PhD researcher Jens de Bruijn (IVM-VU) uses Twitter to determine the location of floods. The new approach can help first-responders react quicker in case of emergency. Earlier this month, he tested the system for use by the Red Cross in the Philippines.
The method depends on a new algorithm that allows to identify the geographic locations of natural disasters mentioned in tweets. The detection method works in real time and can complement satellite data and other sources of information. The advantage is that the information comes from people who are actually at the disaster site, who communicate what is happening at that exact moment.
Determining the geographic origin of tweets is not as easy as it sounds. Twitter does not by default attach GPS coordinates to user’s posts. Only 1 in 50 tweets have coordinate information. To solve this problem, Jens de Bruijn (PhD researcher at IVM-VU) and his colleagues set out to find locations referred to in tweets. Because many place names have multiple occurrences worldwide, the team analyses secondary information attached to the tweets, such as the user’s time zone and hometown, and, if available, GPS coordinates to find the actual location referred to.
What makes the method powerful is that it does not look at individual tweets, but groups tweets together based on their timing and the locations they mention. It is then possible to determine the location of the whole group of tweets using the secondary information. The algorithm was applied on a database of more than 58 million historical flood-related tweets published in 12 different languages.
De Bruijn works with FloodTags to integrate the new system in real-time in the FloodTags dashboard. With them, De Bruijn visited the Red Cross in the Philippines to test the method at the Philippine Red Cross operations center. After showing the dashboard, the staff of the operations center, within seconds they found a new flood and could immediately dispatch a response team. The team is also in touch with the World Bank and other organisations to further develop the system.
Using their flood management expertise, further research will be done to detect new flood events based on geoparsed tweets. The team also has a paper under review in Natural Hazards and Earth System Sciences.
Jens de Bruijn, de Moel, H., Jongman, B., Wagemaker, J., and Aerts, J. C. J. H.. TAGGS: Grouping Tweets to Improve Global Geotagging for Disaster Response. In: Nat. Hazards Earth Syst. Sci. Discuss., https://doi.org/10.5194/nhess-2017-203, in review, 2017.
Magazine for water professionals H20 published an interview with Jens de Bruijn: