Engineering studies on coastal risks require continuous improvement of the numerical tools to represent the physical processes involved as accurately as possible. Through various studies carried out over the past twenty years, the developments and other improvements of tools or methodologies are presented. This may involve reproducing as closely as possible the crossings or ruptures of linear protection structures in order to assess and map their hazards. To do this, the interface between the maritime and land areas has been specifically developed to provide a detailed representation of the protection system and to calculate the flows transiting the structures (phenomena of overflow or sea packages due to waves). This model then makes it possible to determine the hazards resulting from the combination of a hydro-meteorological event and a scenario concerning the protection system. Several references of studies have allowed to improve the method and make it more reliable (Gironde estuary, Somme bay, Seine estuary, Charente pertuis…). Some sites are also subject to tsunami flooding risks. It is then a question of being able to model this type of event for which the source can be very distant geographically. The initial elevation at the seismic source is obtained by the Okada method and the signal is then propagated to the coast by different nested models. The results obtained make it possible to define levels to characterize floods near a protection or flood maps in coastal areas (Turkey, Haiti…). The variability of the spatial scales to be represented requires the use of different software, which must then be well understood and controlled. For some studies, the combination of these tools is mandatory and the use of computational clusters allows to subdivide the modelled domain on several processors to reduce computation times. All these techniques and methodological choices respond to a growing demand to control the hazards and the necessary adjustments to reduce them. The engineer has a duty to explain the uncertainties associated with the input data, assumptions considered and simplifications inherent on the tools used.