Computers & Geosciences ( IF 4.2 ) Pub Date : 2021-07-24 , DOI: 10.1016/j.cageo.2021.104893 Mingxuan Dou 1 , Yandong Wang 1, 2, 3 , Yanyan Gu 1 , Shihai Dong 1 , Mengling Qiao 1 , Yuejin Deng 1
Social media data have been widely used to enrich human-centric information for situational awareness and disaster assessment. Owing to the granularity of topics detected from disaster-related contents, the effectiveness of social media in reflecting disaster losses is still limited. To address this limitation, this study developed a methodology for assessing disaster losses using social media data, which was composed of data preprocessing, fine-grained topic extraction, and quantitative damage estimation. The proposed methodology was demonstrated in a case study of Typhoons Hato & Pakhar, which caused persistent damage in southern China from August 22 to August 30, 2017. The results highlighted the capability of the proposed methodology in using fine-grained topics to assess disaster losses, e.g., the disaster losses were significantly correlated with infrastructure damage-related topics. The study provided useful insights in disaster damage assessment through the fine-grained topics in social media.