Probabilistic modeling framework for flood risk assessment: A case study of Poldokhtar city

Abstract

Study region: Poldokhtar City is located on the bank of the Kashkan river in Iran. Study focus: This study presents a probabilistic modeling framework for flood risk assessment using Monte Carlo simulations. We developed a Machine Learning (ML)-based flood depth prediction model for Poldokhtar city using the Least Squares Support Vector Machine. HEC-RAS was utilized for 2D flood modeling, and its performance was evaluated by comparing simulated and remote sensing-derived flood extent maps. The simulated results were used to develop a surrogate ML-based model that predicts flood depth maps. Finally, we used this model to estimate the flood risk of Poldokhtar city as a combination of hazard, exposure, and vulnerability for 10000 flood scenarios. New hydrological insights for the region: The mean annual flood damage of the city based on the proposed framework is US$ 1177034, which is about three times lower than that calculated using the simplified method used in the classical risk analysis (US$ 3455400). Buildings near the floodwalls of the river and in the southwestern parts of the city have higher mean flood losses than those in other areas. Risk index frequencies of buildings reveal the most at-risk zones in the city, where there have been building constructions without considering flood hazards. The proposed framework would be of use to stakeholders such as policymakers to develop effective flood management strategies

Keywords

• Risk analysis
• Monte Carlo simulation
• Machine learning
• Least squares support vector machine
• Two-dimensional hydrodynamic modeling
• Flood modeling