Early warning systems are an important tool for local authorities to detect emergency situations in advance and initiate the necessary safety measure. These systems often depend on numerical models to estimate wave overtopping in the affected areas. The SWASH model has shown to deliver good results in recent overtopping studies. The To-SEAlert project has the aim of increasing the efficiency, robustness and reliability of the HIDRALERTA early warning system. This study shows a first intent to implement the SWASH model to simulate wave overtopping for the Ericeira prototype. SWASH was implemented for one breakwater profile used to simulate the overtopping discharge and evaluate the associated risk levels. It was compared to the current approach used in HIDRALERTA, which resorts to a neural network trained with a physical modeling database, NN_OVERTOPPING2. Finally, both approaches were compared with previously analyzed video images of the breakwater. The results showed that SWASH generally overestimates overtopping and is not in good agreement with the video images. NN_OVERTOPPING2 has a better agreement with the video images. A possible reason for the overestimation might be the wave direction, which cannot be included in one-dimensional simulations in SWASH.