Spatio-temporal Structures of Geotagged Twitter Posts After Sudden Catastrophic Events
ISBN 978-85-88783-11-9
Authors
1Ye, H.; 2Clarke, K.
1UC SANTA BARBARA Email: haiyunye@geog.ucsb.edu
2UC SANTA BARBARA Email: kclarke@geog.ucsb.edu
Abstract
Geographic information in social media posts has proven useful as a sensor for disasters and can assist in emergency response. We chose to study catastrophic events of different scales to examine whether Twitter postings about events follow patterns similar to those seen in other user-contributed content relating to disasters, such as the long tail and donut pattern. We also explored potential new spatio-temporal patterns in the dataset. We applied a mixed methods approach in this research, using techniques such as visual analytics, spatial statistics as quantitative, and text data analysis as qualitative method. The case we chose to study is the Boston Marathon Bombings that took place in April 2013. About 400,000 geolocated tweets related to the event in the U.S. were extracted and analyzed. We used visual analytics tools such as word clouds and tag maps as a means of analyzing spatio-temporal social media data. Different combinations of visual variables such as color, size, and the font of the text were used to help explore patterns in tag maps. The phenomena we found, together with the knowledge we acquired about user-contributed content related to disasters may assist emergency response efforts.