Rapid Prediction of Storm Wave Run-Up Using a Hybrid Physics-Informed Machine Learning

Storm-induced wave run-up is responsible for wave overtopping, beach erosion, and flooding. Therefore, it is crucial to simulate such events, especially during hurricanes and nor’easters. Low-fidelity phase-averaged models are often preferred and used for the prediction of wave run-up due to their computational efficiency. However, phase-resolving numerical models have shown great accuracy in predicting wave run-up with much demanding computation resources. In this study, a mapping approach based on machine learning techniques is proposed to rapidly predict high-fidelity numerical simulations given their corresponding low-fidelity results. Specifically, the proposed model maps the wave run-up from the phase-averaged surf-beat of the XBeach model to its corresponding values from the phase-resolving nonhydrostatic mode. Two artificial neural networks were trained to simulate the extreme wave run-up \((R_{2\% } )\) and wave profile, respectively. The simulation results demonstrate the excellent performance of the proposed model in predicting the wave run-up characteristics. As a result, the models are suitable for use in early warning systems, probabilistic risk assessment, and rapid prediction of wave run-up during extreme events.