Understanding architectural heritage requires not only documenting physical form but also interpreting the processes through which it evolves. This study approaches spatial transformation as a narrative process, treating architectural plans as sequential states within a dynamic spatial system. It focuses on a selected area of Landshut, Germany—a city largely preserved after the Second World War—providing a valuable context for tracing transformation from the nineteenth century onwards. Predominantly domestic buildings reflect layered transformations tied to broader urban and cultural shifts, particularly at ground-floor level where changing public–private relations reshaped access, circulation, and spatial interaction.
The study is based on a structured dataset derived from historical plans and archival sources provided by the municipality of Landshut. Plans were redrawn and standardised in CAD to ensure consistent spatial segmentation. Buildings were organised into front, middle, and rear blocks reflecting plot-based urban morphology, while spaces were classified according to historical labels and functional roles (e.g. circulation, distribution, habitation). Façade elements—windows, doors, and façade zones—were extracted and linked to spatial units, enabling integration of interior and exterior data. Temporal information was incorporated through building phases, allowing spatial configurations to be analysed over time.
Building on this dataset, the research develops a queryable, time-aware graph-based representation of spatial configurations. Architectural plans are translated into graphs in which spaces are modelled as nodes and relationships—such as doors, openings, vertical connections, and culturally informed visual links—are encoded as edges. Each spatial unit is described through geometric, topological, and functional attributes, enabling analysis through metrics such as centrality, connectivity, and depth.
The methodological approach is grounded in computational models for interpreting architectural and spatial qualities. Graph-based representations analyse spatial properties such as privacy gradients and public–private relationships, building on foundational theories of spatial configuration and relational structure (Hillier & Hanson, 1984) and more recent computational implementations (Ferrando, 2018), and support typological comparison through graph-based similarity measures (Chang et al., 2026). In parallel, environmental and contextual factors—such as daylight, view, and noise—have been integrated into architectural analysis through simulation-based approaches, alongside generative methods that encode spatial relationships and orientation for layout generation (Mostafavi & Khademi, 2023; Mostafavi, Khademi & Vrachliotis, 2025). Semantic modelling supports the structuring and interoperability of architectural data (Breitling et al., 2018), while data-driven approaches enable encoding of complex cultural and spatial relationships for computational interpretation (Huang et al., 2023).
Within this framework, spatial, relational, and temporal dimensions are integrated into a unified system, where façade elements are linked to corresponding spaces. This allows analysis of interior–exterior relationships and supports assessment of daylight conditions through façade openings and spatial configuration. Temporal links connect spatial entities across historical phases, enabling tracking of transformations such as subdivision, merging, and functional change, while supporting a multi-level reading of spatial continuity and architectural transformation.
The study also builds on previous machine-learning research demonstrating that meaningful spatial inference is possible under data-scarce heritage conditions in Landshut (Talaei Koochehbagh & Pourhosseiniakbarie, forthcoming), emphasising interpretable, feature-based modelling rather than data-intensive approaches.