The paper describes a probabilistic method of modelling the Flood Water Level in a dam reservoir. Its purpose is to add a probability dimension to hydraulic loading computations. Unlike the standard engineering approach, it incorporates two innovative aspects. First, it uses a probabilistic hydrological method (SHYPRE method) based on simulation techniques, for generating multiple hydrological scenarios. Second, it can address the variability in reservoir level at the time of the occurrence of the flood event. It generates a random set of Flood Water Levels from a random selection of flood hydrographs and Headwater Levels. Statistical analysis of the sample of reservoir levels obtained over a given period allows us to assess a probability distribution for Flood Water Levels. The method was tested on three applications representing dams having different operating rules and hydromorphological properties. The lessons learnt with our method as compared with the standard engineering approach are discussed from the viewpoint of inclusion in the modelling of the hydraulic loading on the dam and the assessment of the hydraulic safety of the dams.