Developing a flood risk assessment model with genetic algorithm-based weights

To reduce flood risk efficiently within constrained disaster prevention budgets, governments employ economic analyses and qualitative flood risk assessments. However, conventional methods, such as entropy weight and the Analytic Hierarchy Process method, have limitations in terms of the accuracy of the flood risk index. Here, we overcome these limitations by applying a genetic algorithm (GA) - an optimization method mimicking a natural selection process and biological genetic evolution. We developed a new flood risk index by using GA to calculate weights to indicators associated with four items (Hazard, Exposure, Vulnerability, and Capacity) for 161 Korean cities and counties from 2016 to 2021. The indicators (number of buildings, farmland area, dependent population, etc.) for the Exposure and Vulnerability items were reflected in the evaluation only for damaged targets directly exposed to flood risk, using grid cells of indicators overlaid on the flood risk map. Our GA-based method aimed to optimize each indicator's weights to minimize errors between damage rankings and flood risk index rankings. Results show that our method reduced errors by 21.42 % during 2016-2021, outperforming traditional methods. Therefore, it is easy to identify municipalities that lack disaster prevention capabilities and are vulnerable to flood risk by comparing flood risk indices under the same conditions, such as maximum rainfall index. Our proposed method could better aid local government in decision-making for flood risk mitigation by allocating constrained budgets efficiently.