Predicting Urban Waterlogging Risks by Regression Models and Internet Open-Data Sources

Abstract

In the context of climate change and rapid urbanization, urban waterlogging risks due to rainstorms are becoming more frequent and serious in developing countries. One of the most important means of solving this problem lies in elucidating the roles played by the spatial factors of urban surfaces that cause urban waterlogging, as well as in predicting urban waterlogging risks. We applied a regression model in ArcGIS with internet open-data sources to predict the probabilities of urban waterlogging risks in Hanoi, Vietnam, during the period 2012–2018 by considering six spatial factors of urban surfaces: population density (POP-Dens), road density (Road-Dens), distances from water bodies (DW-Dist), impervious surface percentage (ISP), normalized difference vegetation index (NDVI), and digital elevation model (DEM). The results show that the frequency of urban waterlogging occurrences is positively related to the first four factors but negatively related to NDVI, and DEM is not an important explanatory factor in the study area. The model achieved a good modeling effect and was able to explain the urban waterlogging risk with a confidence level of 67.6%. These results represent an important analytic step for urban development strategic planners in optimizing the spatial factors of urban surfaces to prevent and control urban waterlogging.