Social media, as an emerging source of data, has become a valuable data source for disaster prevention and management with its vast database. This study takes the “7·20” extreme rainstorm disaster in Zhengzhou as an example, extracts social media information related to the flooding disaster on social media platforms, constructs a dataset of the rainstorm and flooding disaster, and analyzes its temporal evolution and public sentiment change trends. Secondly, the K-means text clustering method based on TF-IDF was used to extract geographic location information related to waterlogging in social media data, and the resilience matrix was used as a framework to construct a resilience assessment model for urban infrastructure systems in combination with the Pressure-State-Response model. The results show that there was a sharp increase in the discussion of microblog topics about flooding information during heavy rainfall, and there was a 4-h lag between the peak time of topic discussion and the peak time of rainfall intensity. The geographical locations affected by waterlogging extracted based on microblogs can cover 87.5 % of the officially announced waterlogging points. In the resilience assessment of infrastructure systems, the transportation system and drainage system of Zhengzhou performed poorly in response to this rainstorm disaster, whereas the power system and communication system had relatively stronger resilience. This study provides an effective solution to help identify disaster events and promote disaster risk management.
Read full abstract