Integrating Internet media into urban flooding susceptibility assessment: A case study in China

Abstract This study demonstrates the value of the public and of media observations (i.e. ‘citizen science’) for the modeling and understanding of the predisposing factors of urban flooding as a contribution to flooding hazard assessment. The 2016 Wuhan flood in the center of China utilized observations of non-scientists' from Internet media to capture the spatial distribution of urban flooding. Various sources in online news, blogs, bbs, and governmental websites were included. Independent flood-related factors such as rainfall, wetlands degradation, elevation, land use and land cover, curvature, slope, and normalized difference vegetation index were integrated into a logistic regression model to produce a susceptibility map. The feasibility and accuracy of this model was evaluated via comparison with official inundation maps and shortcomings of previous drainage system design. The accuracy assessment by receiver operating characteristics curve analysis estimated a success rate of 95.4%. This result shows that the data from mainstream media and public observations provide accurate and comparable information on flood occurrence at the urban scale. The assessment can be used to plan and design preventative drainage strategies for decision-makers. Finally, the limitations of Internet media as a new data resource for flooding risk assessment and future directions are discussed.