Oil spill in the sea has a strong environmental and economic impact, despite the common belief about catastrophically events (for example the Exxon Valdez in Alaska or the Deepwater Horizon in the Gulf of Mexico), most of the oil spills are characterized by small pollutant release, less than 7 tons [1] and occur in coastal areas, especially inside harbours. These oil spills are related, for example, to tank collision or to system failure during bunkering procedure. Usually, oil tends to spread at the sea surface as a very thin layer and to cover very large area, its dispersion is mainly driven by gravity, wind drag and sea currents. Even a small oil spill represents a potential enormous damage, especially in harbours and closed basin [2]. Therefore, it becomes crucial to assess the most likely patterns of pollutant dispersion and accordingly detect the most vulnerable areas, to develop targeted policies and hazard management plans. In this context, the Augusta (Sicily, Italy) harbour case has been studied, within the Isyport project (www.isyport.com). From a large-scale circulation model, several meteorological scenarios have been selected, then, through a nesting procedure, the hydrodynamics of Augusta harbour has been studied with a fine grid. Eventually, different pollutant dispersions have been considered, varying the source points and the chemical characteristics. The preliminary results show that the time evolution of the pollutant sources and its interaction with the coastline and the harbour structures is of the order of hours, therefore, mitigation protocols, to be effective, requires an a-priori deep knowledge of the pollutant dynamics. This type of analysis poses a computational cost issue, a very large number of numerical simulations need to be run to obtain a complete risk map definition. In this sense, taking advantage of simplify models, based on a smaller database, could be strongly beneficial.