In this paper two concepts for modelling 3D topography are introduced. The first concept is a very pragmatic approach of 3D modelling, trying to model as much as possible in (less complicated) 2.5D and use 3D modelling only in exceptional cases. The idea is to use a constrained TIN in 2.5D and place 3D TENs on top or below this surface. As both data structures use the same simplexes (nodes, edges, triangles) this integration should be very well possible. At a conceptual level this approach seems suitable, but at design level serious problems occur. To overcome these a rigid approach is developed, modelling all features in a 3D TEN, including the air above and earth beneath these topographic features. This model is stored and maintained within a spatial database. Despite its more advanced concept, it is shown that this approach offers huge advantages compared to the initial pragmatic approach.