Data Schema Reference

Core Data Fields

Group	Attribute	Type	Definition	Example
Identifiers	`id`	`string`	Unique ID composed of a Block ID¹ and sequence number.	`3ba70f9963714924-0`
	`region_id`	`string`	NUTS3 regional identifier.	`FR931`
	`city_id`	`string`	Local Administrative Unit (LAU) identifier for the city.	`FR93066`
Attributes	`type`	`category`	Binary usage type².	`residential`
	`subtype`	`category`	Detailed usage type³.	`terraced`
	`height`	`float`	Distance in meters from ground floor to top of building.	`23.9`
	`floors`	`float`	Total number of above-ground floors.	`3.5`
	`construction_year`	`integer`	Year construction finished (not renovation year).	`1963`
Geometry	`geometry`	`binary`	Footprint geometry projected in `ETRS89` (`EPSG:3035`) encoded as WKB. Units in meters.	`01030000...`

Auxiliary Data Fields

Group	Attribute	Type	Definition	Example
Confidence	`type_confidence`	`float`	Relative intersection of footprints (merged) or calibrated class probability (predicted). Range: `[0, 1]`.	`0.95`
	`subtype_confidence`	`float`	Relative intersection of footprints (merged) or calibrated class probability (predicted). Range: `[0, 1]`.	`0.64`
	`height_confidence_lower`	`float`	Min source value (merged) or lower 95% bootstrap CI (predicted).	`7.5`
	`height_confidence_upper`	`float`	Max source value (merged) or upper 95% bootstrap CI (predicted).	`10.5`
	`floors_confidence_lower`	`float`	Min source value (merged) or lower 95% bootstrap CI (predicted).	`2.7`
	`floors_confidence_upper`	`float`	Max source value (merged) or upper 95% bootstrap CI (predicted).	`3.2`
	`construction_year_confidence_lower`	`int`	Min source year (merged) or lower 95% bootstrap CI (predicted).	`1990`
	`construction_year_confidence_upper`	`int`	Max source year (merged) or upper 95% bootstrap CI (predicted).	`2000`
Sources	`geometry_source`	`category`	Origin of footprint geometry.	`gov-france`
	`type_source`	`category`	Origin of `type` attribute.	`osm`
	`subtype_source`	`category`	Origin of `subtype` attribute.	`estimated`
	`height_source`	`category`	Origin of `height` attribute.	`estimated`
	`floors_source`	`category`	Origin of `floors` attribute.	`gov-france`
	`construction_year_source`	`category`	Origin of `construction_year` attribute.	`gov-france`
Source IDs	`geometry_source_id`	`string`	Primary identifier from the source provider.	`BATIMENT000...`
	`type_source_ids`	`array`	List of IDs from source(s) contributing to the `type` attribute.	`['osm_123']`
	`subtype_source_ids`	`array`	List of IDs from source(s) contributing to the `subtype` attribute.	`['osm_123']`
	`height_source_ids`	`array`	List of IDs from source(s) contributing to the `height` attribute.	`['ign_456']`
	`floors_source_ids`	`array`	List of IDs from source(s) contributing to the `floors` attribute.	`['ign_456']`
	`construction_year_source_ids`	`array`	List of IDs from source(s) contributing to the `construction_year`.	`['ign_456']`
Source values	`subtype_raw`	`string`	The original, unmapped building use type from the source dataset.	`Einfamilienhaus`

Incorrect OSM Source IDs

The geometry_source_id and <attr>_source_ids fields currently do not contain the original OSM ID but a sequential index.

NUTS Code Discrepancies

EUBUCCO uses modified NUTS 2016 boundaries with two regional merges (DEB33 → DEB3H, UKD73 → UKD47) and one reconstructed region: UKN1 was missing from the official download but reconstructed from its NUTS 3 components.

Example

id	region_id	city_id	type	subtype	height	floors	construction_year	type_confidence	subtype_confidence	height_confidence_lower	height_confidence_upper	floors_confidence_lower	floors_confidence_upper	construction_year_confidence_lower	construction_year_confidence_upper	geometry_source	type_source	subtype_source	height_source	floors_source	construction_year_source	geometry_source_id	type_source_ids	subtype_source_ids	height_source_ids	floors_source_ids	construction_year_source_ids	subtype_raw	geometry
aec95c88db604a13-0	FRF31	FR54029	residential	detached	3.3	1.0	nan	1.0	0.99	3.2	3.5	1.0	1.0	<NA>	<NA>	gov-france	osm	estimated	estimated	osm	<NA>	BATIMENT0000000334652469	['lorraine-latest_1372503']	<NA>	<NA>	['lorraine-latest_1372503']	<NA>	Indifférencié	POLYGON ((402...))
9d1c73d618ef419e-0	FRF31	FR54037	residential	detached	5.7	2.0	nan	0.86	0.84	5.3	6.2	2.0	2.0	<NA>	<NA>	gov-france	osm	estimated	estimated	osm	<NA>	BATIMENT0000002101827291	['lorraine-latest_1661552']	<NA>	<NA>	['lorraine-latest_1661552']	<NA>	Indifférencié	POLYGON ((404...))
ce19c9e4976d46f6-2	FRF31	FR54039	non-residential	others	3.8	2.0	nan	0.84	0.84	nan	nan	2.0	2.0	<NA>	<NA>	gov-france	osm	osm	gov-france	osm	<NA>	BATIMENT0000000334989890	['lorraine-latest_805209']	['lorraine-latest_805209']	<NA>	['lorraine-latest_804619']	<NA>	Indifférencié	POLYGON ((407...))
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...

Technical Notes

Building Type Classification

We use a two-tier hierarchy to categorize building usage. The type field provides a high-level binary split, while the subtype field contains the granular class.
- Type: residential. Subtypes: detached (single-family), semi-detached (duplex), terraced (row house), and apartment (multi-family).
- Type: non-residential. Subtypes: industrial, commercial, public, agricultural, and others (e.g. garages)
The subtype_raw column preserves the original source classification string prior to this harmonization. Refer to the metadata file on building type mapping for more details.

Geometry Encoding

The geometry column is stored as Well-Known Binary (WKB). This is a compact, machine-readable format optimized for GIS tools like PostGIS, QGIS, and GeoPandas.

Attribute Sources

Attribute source categories are osm, msft, gov-<region>, estimated (where represents the specific regional authoritative dataset identifier)

Attribute Uncertainty

Ground Truth: NaN in a <attr>_confidence column indicates the attribute was provided directly by the geometry source; no merging or machine learning estimation was required.
Interpretation: For type_confidence and subtype_confidence, a value of 0.6 implies that 60% of buildings in that cohort are statistically expected to be correctly classified.
Methodology: For details on how the uncertainty of attribute estimation and merging is quantified, please refer to the Uncertainty Section.

Attribute Merging

Source Mismatch: If geometry_source and <attr>_source differ, the attribute has been merged between datasets.
Data Fusion: If <attr>_source_ids contains multiple values, the final value has been from aggregated from multiple source buildings.

Attribute Prediction

See the Prediction Evaluation section for an assessment of attribute estimation quality.

Block ID: A unique identifier for a cluster of topologically connected (touching) buildings. ↩
Binary use type categories: residential and non-residential. ↩
Detailed use type categories: industrial, commercial, public, agricultural, others, detached, semi-detached, terraced, and apartment. ↩