Urban environments consist of multiple interconnected spatial entities whose integrated representation is essential for urban modelling and analysis. However, existing three-dimensional (3D) modelling approaches often focus on the reconstruction of individual object classes, limiting the generation of integrated urban-space models. This paper investigates the use of Constructive Solid Geometry (CSG) for the automatic generation of urban-space models through the integration of heterogeneous geospatial datasets, including airborne Light Detection and Ranging (LiDAR), OpenStreetMap (OSM), and cadastral information. The proposed approach leverages semantically structured urban entities derived from these datasets, including buildings, vegetation, transportation spaces, and water bodies, and organises them into two urban-space representations. These are subsequently combined using CSG Boolean operations to construct volumetric urban models, enabling the generation of topologically consistent representations while preserving semantic classes and functional attributes. The method is evaluated through a case study in Luxembourg. Furthermore, the generated CSG models support the derivation of urban-form indicators, demonstrating their potential for quantitative urban characterisation.