In this study, conducted for a large hydropower system in Norway, we present a novel approach that combines the use of design rain data with a catchment digital twin framework to explore the range and uncertainties associated with design flows. By dynamically coupling a well calibrated hydrological model to a one-dimensional (1D) hydraulic model that describes the hydropower installation it was possible to generate hindcasts of inflows to the reservoir, providing reliable timeseries and initial conditions for design flood modelling. Extreme value analyses for 1000-year flood (Q1000) were calculated using multiple different distributions. The probable maximum flow (PMF) was determined by applying various design rain sequences to over six decades of hindcasted initial conditions, generating near 60,000 PMF values. When compared against results from the prescribed methodology, the design floods produced by this work tended to be significantly lower than previous analyses for the hydropower system studied.