Urban flood susceptibility mapping based on social media data in Chengdu city, China

Increase in urban flood hazards has become a major threat to cities, causing considerable losses of life and in the economy. To improve pre-disaster strategies and to mitigate potential losses, it is important to make urban flood susceptibility assessments and to carry out spatiotemporal analyses. In this study, we used standard deviation ellipse (SDE) to analyze the spatial pattern of urban floods and find the area of interest (AOI) based upon related social media data that were collected in Chengdu city, China. We used the social media data as the response variable and selected 10 urban flood-influencing factors as independent variables. We estimated the suscepti-bility model using the Naive Bayes (NB) method. The results show that the urban flood events are concentrated in the northeast-central part of Chengdu city, especially around the city center. Results of the susceptibility model were checked by the Receiver Operating Characteristic (ROC) curve, showing that the area under the curve (AUC) was equal to 0.8299. This validation result confirmed that the susceptibility model can predict urban flood with a satisfactory accuracy. The urban flood susceptibility map in the city center area provides a realistic reference for flood monitoring and early warning.