Skip to content

Data Format

Data Schema

The core dataset structure is organized into five functional groups:

  • Identifiers: Unique building id and geographic identifies: region_id (NUTS3) and city_id (LAU).
  • Attributes: Building characteristics (type, subtype, height, floors, construction_year).
  • Confidence: Uncertainty of ML-based estimation and cross-source merging of attributes.
  • Sources: Provenance tracking for every attribute (i.e. osm, msft, gov-*, estimated).
  • Geometry: ETRS89 (EPSG:3035) footprints stored in Well-Known Binary (WKB) format.

File Format

The EUBUCCO dataset is distributed in multiple file formats:

  • Parquet (.parquet)
  • GeoPackage (.gpkg)
  • Shapefile (.shp)

Partitioning

The dataset uses Hive-style partitioning by NUTS2 regions and multi-level sorting (source type, NUTS3, city ID) to optimize data access. This enables partition pruning for efficient regional filtering and spatial locality for bounding box queries.

Metadata Files

In addition to the core building data, metadata information with respect to the source datasets, data retrieval dates, and attribute harmonization (e.g. building type mapping and aggregation) are provided as additional files.

Uncertainty

For attributes that were estimated using machine learning techniques or merged from OpenStreetMap data, we provide an uncertainty quantification.