Computing volumes and surface areas including party walls for the 3DBAG data set

Abstract

This document provides a description of the project “Computing volumes and surface areas including party walls for the 3DBAG dataset” which has been carried out between the 3D Geoinformation group at TU Del6, 3DGI, and RVO in the timeframe between November 2022 and October 2023.

The goal of this project is to derive parameters from the 3DBAG that are relevant for energy consumption estimation, i.e. the enclosed volume of each building, as well as the party wall areas, the exterior wall areas, the ground floor areas and the roof areas.

As the detection of the party wall (i.e. the portion of the building shell that is shared between two buildings (BAG-panden)) is the most complex task to solve, specifically for a large data set, the main goal of the project is to define, evaluate and implement a methodology to compute the area extents of party walls between adjacent buildings from the 3DBAG data set. The 3DBAG dataset was first released in March 2021. A first revised version was released in September 2021 which we used for our analysis carried out during the first part of this project. The latest (5th) version has been released in October 2023 based on which we generated the final data for this project. In this last version AHN4 has been incorporated.

3DBAG is a country-wide dataset containing all buildings in the Netherlands, modelled in multiple LoDs, and based on the international standard CityGML. According to CityGML, a building can be modelled as a single-part unique object, or as an aggregation of building parts, each one having its own geometry. Additionally, each building is a geographical feature that can have several aTributes (e.g. year of construction, number of storeys, etc.) and different geometries representing each one a specific Level of Detail (LoD). A graphical overview of the different LoDs according to CityGML v. 2.0 is given in Figure 1.

In particular, the LoD2 allows differentiating between different thematic surfaces composing the building envelope. The geometries are semantically enriched and classified into GroundSurfaces, WallSurfaces and RoofSurfaces.

In the case of LoD2, RoofSurfaces represent the main planar surface(s) of the roofs. Smaller roof structures like chimneys and dormers are generally absent if their size is too small with regard to the surveyed data used for the 3D reconstruction process (e.g. the Lidar point cloud density).

The GroundSurfaces generally correspond to the planar extents of the roof surfaces projected onto the horizontal ground but they can also correspond to footprints. WallSurfaces connect vertically the Roof- and GroundSurfaces. This means that overhanging geometries (e.g. of roofs) lead to larger GroundSurfaces as in reality. But roof overhangs can also be represented. A graphical example can be seen comparing LoD2 and LoD3 in Figure 1. Finally, in LoD2 buildings there are no openings, i.e. neither doors nor windows.

More details about CityGML v. 2.0 and the modelling rules for the buildings can be found in the technical specifications of the standard published by the Open Geospatial Consortium.