Natural phenomena in mountains put people and assets at risk. Risk level is often described as a combination of hazard and vulnerability. Hazard relates to the intensity and frequency of phenomena while vulnerability concerns damages and values assessment. In the debris-flows context, numerical models are used to assess height, speed and extent of flow. One important issue is to consider the influence of input data imperfection on simulation results. This article presents the “Hybrid” approach that propagates the uncertainty through numerical simulation models and considers the different aspects of information imperfection, especially its imprecision (lack of information, inaccuracy of measure…). This new method generalizes, under some restrictive conditions, the usual Monte Carlo method, by using probability, possibility and belief function theories, used as tools for coding sets of probability densities. An example of results (quantile of deposition heights, threshold exceedance probabilities) of a numerical simulation of debris-flows is proposed: they show the influence of data imperfection including those resulting from expert assessments on the simulation results.