In most cases, three-dimensional groundwater models are quite small. Predominantly, the local three-dimensional flow is limited to a small area, but requires a fine discretisation into the third dimension too. This causes large numbers of nodes. Hence, the problem is to find correct boundary conditions which do not change during the simulation. Two-dimensional models can be enlarged until some natural boundary such as a watershed divide or a receiving water course is reached. For three-dimensional models, the number of nodes would grow quite fast. Otherwise, oversized horizontal elements in 3D models cause numerical problems. The best way of solving the problem is represented by a large 2D model combined with a partial 3D model.

The partial 3D mesh is generated automatically, and is equivalent to the complete three-dimensional model. In addition, the definition of one or more closed boundaries for the 3D area by node numbers (3DRA) is required. The program identifies the elements inside the boundaries and projects them into the 3rd dimension.

 

Example of a horizontal model with a three dimensional part

 

Transition zone between the 2D model and the 3D area