Geometric comparison of 3D city models for daylight simulations

Abstract: As cities keep densifying, essential resources such as daylight access become increasingly restricted. Therefore, prioritizing energy efficiency and enhancing the energy performance of our buildings is required for obtaining a sustainable urban environment. The challenges of implementing these ideas have triggered efforts to make cities smarter, for example, by using 3D city models for solar energy and daylight access simulations. While 3D model daylight simulations can help carry out planning duties, planners have been hesitant to employ them due to the difficulty of data integration and the time required to prepare input data and set up the models. In this thesis, three different 3D city models are geometrically compared and utilized to estimate daylight access in buildings. The reference model is Lund municipality’s official 3D city model. The two models compared to the reference model are VGI3D, created by researchers from NTNU, Norway, and DTCC, created by a research team at Chalmers University, Sweden. The comparisons are based on spatial accuracy, geometry, level of detail, and spatial resolution. The daylight metric used is obstruction angle, which is calculated to determine how much skylight reaches the interior of a building or apartment. In an attempt to make daylight simulations more accessible, an open-source GIS tool was designed, implemented, and published to automate the calculation of obstruction angles using 3D window information and a digital surface model. The obstruction angles were affected the most by differences in geometry and the level of detail of surrounding buildings, usually due to differences in height. Spatial accuracy and spatial resolution did not seem to influence the results much. The comparison results showed that the accuracy of the 3D city models varied. VGI3D’s results were, for the most part, relatively accurate. However, it did have one outlier in both geometry and spatial accuracy. Chalmers’ results were very accurate for geometry and spatial accuracy, with two exceptions related to the height of the buildings caused by uncertainties in the input data.