Three-dimensional (3D) city models support geometry-sensitive urban analyses related to energy demand, material stocks, and urban climate. While modelling decisions at the scale of individual buildings have been widely examined&mdash;particularly in relation to Levels of Detail (LOD)&mdash;the geometric specification of intermediate spatial units such as urban blocks remains weakly defined. In practice, block-level representations are often derived through implicit aggregation of building geometries, without explicitly assessing the generalisation strategies involved or their analytical implications.</p> <p>This study systematically evaluates alternative 3D urban block modelling variants derived from identical building-based cadastral data. Three 3D generalisation strategies are considered: footprint simplification, height simplification, and semantic grouping of envelope or fa&ccedil;ade elements. Their combinations generate 19 distinct configurations applied to a morphologically heterogeneous urban block in Zaragoza (Spain).</p> <p>For each variant, total built volume, exposed envelope area (surface), and geometric complexity (polygon count) are computed and compared. Results show that height simplification primarily affects volumetric estimates, whereas semantic grouping strongly reduces exposed surface area and geometric complexity without affecting volume. Footprint simplification exhibits intermediate effects across indicators.</p> <p>These findings demonstrate that block-level generalisation strategies significantly influence analytical outcomes and cannot be treated as neutral aggregation steps. The study provides a structured and reproducible framework for evaluating 3D urban block modelling variants in urban-scale analyses.