Natural phenomena in mountains such as rockfalls cause severe damage to exposed population and assets. Numerical modelling is widely used to assess the hazard level (combination of phenomenon’s intensity and frequency). This paper describes how innovative methods and tools are used to assess both thematic (related to the characteristics of the rockfalls) and spatial (related to the local topography) data uncertainties considering rockfall propagation. Those uncertainties are propagated in rockfall simulation models using classical Monte Carlo (probabilistic) and hybrid (possibilistic) approaches. Uncertainties related to the altimetric information, through Digital Elevation Model (DEM), are modelled by random fields using the newModTer software, that can produce terrain simulations. This paper compares both uncertainty propagation approaches, taking into account spatialized uncertain variables, and proposes a sensitivity analysis describing the contribution ofDEMvariability on the global uncertainty.