In this study, we generate the extent of flooding in an urban environment by integrating historical hydrometeorological observations, extreme value analysis, machine learning techniques, and physically-based hydraulic/hydrodynamic models. A fully distributed hydraulic model, encompassing rainfall-runoff-inundation processes, is developed and integrated with data-driven models. We employ machine learning algorithms like Random Forest, Support Vector Machine and Artificial Neural Networks to test computationally intensive model elements. This enhances computational efficiency and predictive accuracy. Simulated inundations are compared with ground-based observations, complemented by satellite estimates and crowdsourced datasets, for model calibration and validation. Furthermore, we incorporate state-of-the-art climate models to project potential future inundation scenarios. These projected flood extents are then used to assess the risk to assets and critical infrastructure systems. The results provide valuable insights for developing sustainable floodplain management strategies in response to a changing climate.
