Skip to content

Metadata Files

EUBUCCO includes comprehensive metadata files to support data exploration, validation, analysis, and reproducibility. These files are organized into four main categories:

Category Files Description
Data Files eubucco_lat_lon.parquet Lightweight building dataset
Boundaries LAU-cities-2016.parquet, NUTS-regions-2016.parquet Administrative boundaries used
Statistics city-stats.parquet, region-stats.parquet, prediction-eval-metrics.parquet Aggregated statistics and evaluation metrics
Reference Tables building-type-harmonization.csv, input-datasets-metadata.xlsx Source dataset information and type mappings

Data Files

eubucco_lat_lon.parquet

A lightweight version of the EUBUCCO v0.2 dataset containing only building centroids (latitude/longitude coordinates) and footprint area, without full footprint geometries.

Boundary Files

LAU-cities-2016.parquet

City (LAU) administrative boundaries used in EUBUCCO v0.2, based on 2016 boundaries.

Contents: city_id, region_id and geometry for each city.

NUTS-regions-2016.parquet

Regional (NUTS 0-3) administrative boundaries used in EUBUCCO v0.2, based on 2016 boundaries.

Contents: region_id, region_name, and geometry for each NUTS region at all levels (0-3).

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.

Statistics Files

city-stats.parquet

Comprehensive building stock statistics aggregated at the city (LAU) level. This GeoDataFrame includes city geometry and provides detailed metrics for each city.

Administrative Information

  • city_id, region_id (NUTS 3), country
  • City geometry (CRS: EPSG:3035)

Building Counts by Source

  • n_gov, n_osm, n_msft — Total counts by geometry source

Data Quality Indicators

  • Ground truth counts: Buildings with attributes from the same source as geometry
  • n_gt_type, n_gt_subtype, n_gt_height, n_gt_floors, n_gt_construction_year
  • Merged attribute counts: Buildings with attributes merged from different sources
  • n_merged_type, n_merged_subtype, n_merged_height, n_merged_floors, n_merged_construction_year
  • Estimated attribute counts: Buildings with estimated attributes
  • n_estimated_type, n_estimated_subtype, n_estimated_height, n_estimated_floors

Building Type Distributions

  • Main types: Residential, non-residential (counts and areas)
  • Subtypes: Commercial, industrial, agricultural, public, others, detached, semi-detached, terraced, apartment (counts and areas)

Attribute Distributions

  • Height bins: 0-5m, 5-10m, 10-20m, >20m
  • Floor bins: 0-3, 4-6, >6 floors
  • Construction year bins: ≤1900, 1901-1970, 1971-2000, >2000
  • Footprint area bins: 0-25m², 25-100m², 100-500m², >500m²

Area Metrics

  • Total footprint area and floor area
  • Breakdowns by source (gov, osm, msft), type, and subtype

region-stats.parquet

Building stock statistics aggregated at the regional (NUTS 3) level. Contains the same metrics as city-level statistics but aggregated to NUTS 3 regions, including regional geometry.

prediction-eval-metrics.parquet

Regional (NUTS 2) evaluation metrics for building attribute estimation models. This GeoDataFrame provides comprehensive performance metrics for predicted building attributes.

Administrative Information

  • region_id (NUTS 2), country
  • Regional geometry (CRS: EPSG:3035)

Sample Sizes

  • n — Total number of buildings
  • n_gt_binary_type, n_gt_type, n_gt_residential_type, n_gt_height, n_gt_floors — Ground truth counts per attribute

Categorical Variable Metrics

For binary_type, type, and residential_type:

  • Overall classification metrics: F1 score (macro and micro), Cohen's kappa
  • Per-class F1 scores: Individual F1 scores for each building type/subtype

Continuous Variable Metrics

For height and floors:

  • Overall metrics: MAE, RMSE, R² score
  • Binned metrics: MAE and RMSE by value ranges
  • Height: 0-5m, 5-10m, 10-20m, >20m
  • Floors: 0-3, 3-6, >6

External Validation

For height:

  • Microsoft height comparison: Metrics comparing predicted height with heights from Microsoft's GlobalMLBuildingFootprints dataset

Reference Tables

building-type-harmonization.csv

Mapping table showing how building types from various source datasets are harmonized to the standardized EUBUCCO building type classification.

input-datasets-metadata.xlsx

Comprehensive metadata table providing detailed information about all source datasets used in EUBUCCO v0.2.

Contents

  • Dataset identification: Name, country, geographic coverage
  • Access information: Data owner, license, access date, download links or procedures
  • Technical details: File format, data structure, attribute availability
  • Integration information: Processing workflow and integration approach