The reliability of damage models employed in the estimate of potential cost from natural disasters is influenced by many factors, among them some of the most important are a reliable spatial representation of the exposed value and a damage function which has been calibrated for the specific region. However, many countries still lack proper datasets for this purpose and often relies on functions created elsewhere. The objective of this research is to develop a model suited for the Italian context. To do so, a 20m national land use dataset is created by combining available spatial data. The land use allows to project disaggregated census data such demography and added value by dasymetric mapping. Then, empirical damage records from past flood events are employed to calibrate a new damage function for residential structures by the same statistical approach recently employed in Australia.