Hong Kong building function and energy typology dataset for urban energy modeling
2026-08-12
High-density cities require building-level spatial information to characterize urban form, functional composition, and the spatial distribution of energy-relevant building activities. In Hong Kong, official energy statistics are not spatially explicit at the building level, building footprint records lack the functional semantics needed for energy-related spatial analysis, and conventional single-label classification performs poorly for vertically mixed-use buildings. This study presents the Hong Kong Building Function and Energy Typology Dataset (HK-BFETD), a building-level spatial typology inventory aligned with the energy end-use taxonomy of the Hong Kong Electrical and Mechanical Services Department. The dataset was constructed on the Lands Department footprint foundation by integrating Buildings Department records, OpenStreetMap and Overture volunteered geographic information, statutory Outline Zoning Plan data, and geometric attributes through a five-stage hybrid workflow combining rule-based parsing, semantic arbitration, multimodal geometric calibration, and probabilistic inference. The resulting dataset covers 341,153 building footprints with 100% geometric coverage across the territory. Official-record-based deterministic semantic coverage, obtained by matching Buildings Department records to Lands Department footprints, accounts for 77.70% of the estimated floor-area-capacity proxy represented by the Lands Department footprint inventory. The final calibrated inventory comprises 251,901 residential, 49,129 commercial, 31,395 mixed-use, 3,355 industrial, and 5,373 non-assessed structures. Technical validation, provenance fields, and probability attributes support source-aware and uncertainty-aware use of the records. HK-BFETD provides a traceable spatial framework for characterizing Hong Kong’s building stock, representing vertical mixed-use composition, and supporting benchmark-informed urban energy, carbon, climate, and spatial analyses.