Integrated evolutionary optimisation framework for explicit configuration of detention ponds in urban flood risk management

Semi-distributed hydrodynamic models combined with optimisation methods are often proposed for cost-effective pond designs, but typically oversimplify pond representation, neglecting complex urban catchment dynamics. This study addresses these limitations by introducing an integrated optimisation framework that couples a fully distributed hydrodynamic flood model with a Non-dominated Sorting Genetic Algorithm (NSGA-II). The fully distributed approach explicitly captures spatio-temporal surface flow dynamics across a high-resolution computational grid for the catchment, enhancing pond location screening and its subsequent representation in hydrodynamic simulations for flood risk reduction. The optimisation process adjusts locations and sizes to minimise direct damage cost (DDC) and expected annual damage (EAD) relative to pond life cycle cost (LCC), evaluated across 100-year and composite storm scenarios. An enhanced use of standard and marginal benefit-cost ratio (BCR) metrics identifies a set of high-performing solutions from the composite Pareto front. Results indicate ponds near flood receptors are generally more effective, while the optimisation for a 60-min storm achieves significantly higher damage reduction than a 30-min event when evaluated for a 100-year storm. In the 60-min composite storm optimisation, the high-performing solution that achieves the greatest damage reduction requires only ∼37 % of the designed storage capacity for the current study, highlighting efficient resource allocation by the optimisation and solution selection framework. This study demonstrates the critical importance of fully distributed modelling integrated with evolutionary optimisation methods in accurately assessing the location and size of detention ponds, offering urban planners and decision-makers a robust, practical tool for designing cost-effective and resilient flood risk management strategies. • Integrated framework coupling a fully distributed hydrodynamic model and a genetic algorithm. • Spatio-temporal surface flows enhance pond screening and its modelling accuracy. • Use of the marginal benefit-cost ratio efficiently identifies superior pond configurations.