USBF: U.S. Building Footprints

USBF is a building-footprint dataset for the 48 contiguous U.S. states and the District of Columbia, containing approximately 150 million buildings. It combines GlobalBuildingAtlas (GBA) footprints and inherited building heights with selective footprint refinements supported by aerial RGB imagery and normalized digital surface models (nDSM).

The data are distributed as geographic GeoParquet tiles. Each feature includes its footprint and available source, height and edit attributes, allowing users to distinguish retained footprints from recorded geometry changes.

Coverage and tile layout

Data files are in the USBF_v1 data directory. A filename such as USBF_x-00236_y+00067.parquet identifies grid indices x = -236, y = 67. The nominal tile bounds are:

For this example, the bounds are 118.0°W–117.5°W and 33.5°N–34.0°N, including part of Orange County, California. Each building is assigned to one tile using its original ownership coordinates. A footprint can extend beyond the nominal tile bounds. Use the actual spatial bounds in tile_index.parquet when selecting tiles for an area.

Files

Attribute table

The following are the principal attributes. Field names and types for the complete table are in schema.json. Fields that were not recorded for a feature can be null; a null flag is distinct from false.

The nDSM coverage fields describe input availability, separately from the inherited height value. A feature may have a GBA height even when its nDSM coverage is missing. Upstream missing-height sentinels, where present, are retained. Area additions and removals are calculated in the CONUS Albers equal-area coordinate system, EPSG:5070.

Recorded modification categories

Example attribute records

The rows below are actual records from USBF_x-00236_y+00067.parquet in Orange County, California. Heights and areas are rounded only for display. This example shows the attribute table; field availability can vary by feature.

Read or download an area

The tile index can be read first to identify relevant files. There is no need to download the entire country. The following Python example lists tiles intersecting a small Orange County area, then reads one selected tile and filters its footprints to the area.

The spatial filter selects intersecting footprints and retains their full geometry. Tile files can also be downloaded through the Source Cooperative file browser and opened in GIS software supporting GeoParquet.

Sources, attribution and citation

GBA supplies the base footprints, source identifiers and building heights. Its upstream footprint attribution is retained in source and id. Footprint refinements use aerial RGB imagery and elevation-supported evidence, including NAIP imagery and USGS 3DEP-derived nDSM where available.

Consult footprint_license, height_license and LICENSES.md for the applicable component terms. Footprint and height components can have different licenses.

For citation, identify USBF: U.S. Building Footprints, the dataset URL, the selected tiles and the access date. Also acknowledge the upstream dataset: GlobalBuildingAtlas, Earth System Science Data 17, 6647–6668 (2025), doi:10.5194/essd-17-6647-2025.