[en] A hydrogeological model was developed for Beberg with the aim of evaluating the impact of a repository (for the operational and post-closure phases) while accounting for the effects of density-driven flow. Two embedded scales were taken into account for this modelling study: a local scale at which the granitic medium was considered as a continuum and a repository scale, where the medium is fractured and therefore was regarded to be discrete. The following step-wise approach was established to model density-driven flow at both repository and local scale: (a) modelling fracture networks at the repository scale, (b) upscaling the hydraulic properties to a continuum at local scale and (c) modelling density-driven flow to evaluate repository impact at local scale. The results demonstrate the strong impact of the repository on the flow field during the phase of operation. The distribution of the salt concentration is affected by a large upcoming effect with increased relative concentration and by the presence of fracture zones carrying freshwater from the surface. The concentrations obtained for the reference case, expressed in terms of percentage with respect to the maximum (prescribed) value in the model, are as follows: ca 30% for the phase of desaturation, and ca 20% for the resaturation phase. For the reference case, the impact of repository operations appears no longer visible after a resaturation period of about 20 years after repository closure; under resaturation conditions, evidence of the operational phase has already disappeared in terms of the observed hydraulic and concentration fields. Sensitivity calculations have proven the importance of explicitly discretising repository tunnels when assessing resaturation time and maximum concentration values. Furthermore, the definition of a fixed potential as boundary condition along the model's top surface is likely to provide underestimated values for the maximum concentration and overestimated flow rates in the repository during the operational phase. Scoping calculations have shown that the results could be improved by applying free-surface boundary conditions when modelling the impact of the repository. Modelling density-driven flow at local scale with a repository under atmospheric pressure conditions is feasible using CONNECTFLOW/NAMMU. Such non-linear problems are intrinsically difficult to solve and linked with numerical difficulties. In particular, we could overcome the numerous convergence issues but it was demanding in terms of computing performance. Concerning pre-processing, this study has allowed us to improve the integration of CONNECTFLOW/NAMMU within Colenco's computing environment. It is especially noteworthy that numerical calculations with complex geometries (e.g. repository layout with tunnels) have become possible. During the phases of repository operation and post-closure, near-surface effects are likely to occur. The evaluation of their environmental impacts needs to be performed using a numerical model with specific boundary conditions (free surface type). The following recommendations are proposed regarding additional work and open issues: Assessment of the environmental impacts in relation to the phases of repository operation and post-closure. Evaluation of repository impact using a more detailed geometry for the repository layout, such as introducing the shafts and access tunnels as well as including skin effect around the tunnels. Determination of repository impact by modelling density-driven flow including the rock matrix diffusion of salt. The approach for modelling repository impact at Beberg has successfully described the assumed conditions and relevant processes. It may certainly serve as a well founded base for future modelling tasks to provide solutions to further questions