Tropical cyclones (TCs) have tremendous impact on coastal communities in terms of damage due to flooding and high wind velocities, as shown by the recent hurricane season of 2017. Coastal flooding due to TCs can be contributed to different types of forcing (e.g. high offshore water levels, rainfall, etc.), or multiple at the same time (i.e. compound flooding). While the impact of TCs increases, there is the need for better predictions and early warning systems (EWSs). Probabilistic forecasting by modelling different ensembles is wanted to account for uncertainties, which requires fast models. Current modelling options are static models (bathtub approach), which are fast but too simple. Furthermore there are advanced process-based models (like Delft3D, XBeach) which are accurate but too computationally expensive. The solution is the use of a semi-advanced model which solves all relevant processes, but does that in a computationally efficient way. The semi-advanced SFINCS model is developed to solve all necessary processes with computationally efficiency in mind, but is still in its development phase.

This research assesses how compound flooding due to TCs can be modelled in an accurate and computationally efficient way. Besides assessing the relevant physical processes, the implementation in a semi-advanced model is tested. In multiple tests it is shown that for conceptual situations a semi-advanced model can give accurate results within certain limits. Also it is shown in what conditions the advection term of the momentum balance needs to be solved and that a swash zone model approach with an indirect random forcing can give realistic results in 1D runup tests.

A case study of the compound flooding at Jacksonville, Florida, during Hurricane Irma (2017) shows that using a semi-advanced model, a similar accuracy compared to Delft3D can be achieved while being two orders of magnitude faster. Furthermore, a first test with wave-driven flooding in a real case study at Hernani, the Philippines, during Typhoon Haiyan (2013) shows that wave-driven processes have to be explicitly included to model all types of compound flooding. The approach to include these processes with the semi-advanced SFINCS model gives reasonable results compared to XBeach, although they can be further improved with more research. The resulting computational efficiency seems to get in the right range as needed for EWSs.