Comprehensive metrics tracking what matters for thriving, sustainable communities, Part of https://walkthru.earth/hormones-cities
Global Building Atlas LOD1 (2.75B buildings) aggregated to H3 hexagonal cells with urban density metrics. Six H3 resolutions (3–8), one file per resolution, sorted by h3_index. Available in two layouts: v2 (recommended, BIGINT h3_index, 11 columns) and v1 (legacy, VARCHAR h3_index, 15 columns with geometry/lat/lon/area_km2).
Total: 53,632,093 cells per version. Compression: ZSTD level 3. Row groups: 1,000,000 rows.
The raw Global Building Atlas LOD1 tiles are also available as a zoom-level pyramid:
Columns dropped from v1 (geometry, lat, lon, area_km2) are derivable from h3_index via the DuckDB h3 extension:
Sample values (v2, res 5, most built-up cells):
h3.latlng_to_cell()GROUP BY h3_index, SUM() in DuckDBh3_index. v2 stores BIGINT h3_index (no geometry); v1 adds native GEOMETRY via DuckDB spatialAll resolutions produce consistent world totals (~2.75 billion buildings).
v2 has no geometry column. Coordinates, hex strings, and cell areas are derived at query time from the BIGINT h3_index via the DuckDB h3 community extension. This keeps files smaller and avoids redundant stored columns.
v1 (legacy) retains the geometry column using the native Parquet 2.11+ GEOMETRY logical type with GeoParquet 1.0 file-level metadata for backwards compatibility (GEOPARQUET_VERSION 'BOTH'). DuckDB 1.5+ writes per-row-group bounding box statistics automatically. Supported by: DuckDB 1.5+, Apache Arrow (Rust), Apache Iceberg, GDAL 3.12+.
GlobalBuildingAtlas LoD1 — 2.75 billion building footprints with LOD1 heights, distributed as ~hundreds of 5°x5° Parquet tiles (210 GB total). POLYGON geometry with height attribute in meters (-999 = no data). Combines Google Open Buildings (1.62B), OpenStreetMap (490M), Microsoft Building Footprints (432M), TUM deep learning from Planet Labs imagery (135M), and 3D Global Footprints (68M).
Zhu, X. X., Chen, S., Zhang, F., Shi, Y., & Wang, Y. (2025). GlobalBuildingAtlas: An Open Global and Complete Dataset of Building Polygons, Heights and LoD1 3D Models. Technical University of Munich. doi:10.14459/2025mp1782307
Parquet conversion hosted on Source Cooperative:
GlobalBuildingAtlas LoD1 (Parquet Format). source.coop/tge-labs/globalbuildingatlas-lod1
This dataset is licensed under CC BY 4.0 by walkthru.earth. The source GlobalBuildingAtlas is by Zhu et al. (TU Munich), hosted as Parquet by TGE Labs on Source Cooperative.
volume_density_m3_per_km2 | FLOAT | Built volume per km² (m³/km²) |
avg_footprint_m2 | FLOAT | Mean building footprint area (m²) |