Evaluating urban flood risk using hybrid method of TOPSIS and machine learning

With the growth of cities, urban flooding has increasingly become an issue for regional and national governments. The destructive effects of floods are magnified in cities. Accurate models of urban flood susceptibility are required to mitigate this hazard mitigation and build resilience in cities. In this paper, we evaluate flood riskin Jiroft city, Iran, using a combination of machine learning and decision-making methods. Flood hazard maps were created using three state-of-the-art machine learning methods (support vector machine, random forest, and boosted regression tree). The metadata supporting our analysis comprises 218 flood inundation points and a variety of derived factors: slope aspect, elevation, slope angle, rainfall, distance to streets, distance to rivers, land use/land cover, distance to urban drainages, urban drainage density, and curve number. We then employed the TOPSIS decision-making tool for urban flood vulnerability analysis, which is based on socio-economic factors such as building density, population density, building history, and socio-economic conditions. Finally, we derived an urban flood risk map for Jiroft based on flood hazard and vulnerability maps. Of the three models tested, the random forest model yielded the most accurate map. The results indicate that urban drainage density and distance to urban drainages are the most important factors in urban flood hazard modeling. As might be expected, areas with a high or very high population density are most vulnerable to flooding. These results show that flood risk mapping provide insights for priority planning in flood risk management, especially in areas with limited hydrological data.