Supercharging hydrodynamic inundation models for instant flood insight

Floods are one of the most frequent and devastating natural disasters for human communities. Currently, flood response management globally commonly relies on hydrodynamic models for accurate simulation of complex flow patterns of flood events and to provide information on flood risks. However, the computational demand of hydrodynamic models means that they cannot be deployed usefully for real-time flood inundation forecasting over large domains or for situations where simulations need to be run repeatedly for planning purposes. Here we introduce a new modelling approach that supercharges hydrodynamic models for speed while maintaining high accuracy. We found that spatiotemporal patterns of flood inundation simulated using an extremely simplified (and hence superfast) hydrodynamic model can be mathematically transformed to reproduce the results from a high-resolution model. We exploited the efficacy of this transformation to provide high-resolution and accurate flood inundation predictions in a few seconds rather than the many hours required by conventional high-resolution hydrodynamic models, which represents an important practical advancement towards saving lives and protecting assets during flood emergencies. Effective flood response management relies on rapid high-resolution and high-accuracy flood inundation predictions. This study develops a low-fidelity model and upskills its predictions, greatly reducing the computational time while maintaining a high resolution and accuracy comparable with a high-fidelity model.