In this paper, we propose precise geometrical terrain modeling for autonomous navigation of mobile ground robots. Terrain’s geometrical features, such as elevation, roughness, and slope, are essential for terrain traversability estimation to support planning safe and efficient paths through the robot’s operational environment. Elevation 2.5D maps are suitable for traversability assessment and can profit from a high vertical resolution; however, 3D map representation is needed when operating in environments with multiple terrain layers. Vertical voxelization is utilized in existing OctoMap and UFOMap at the cost of relatively low vertical resolution induced by the voxelization, which might lead to a false assessment of terrain traversability. Therefore, we propose a novel method that combines the advantages of 3D maps with a high vertical resolution of terrain modeling. The benefits of the solution are demonstrated in three scenarios: tunnel, indoor passage, and outdoor terrain, where the proposed method outperforms existing map approaches. Besides, the impact of proposed terrain modeling is demonstrated in practical planning examples for ground mobile robots.