The USA Structures Dataset provides a comprehensive geospatial database of over 125 million building structures across the U.S. and its territories, capturing every structure larger than 450 square feet. Developed by Oak Ridge National Laboratory (ORNL) through the integration of advanced geospatial science, high-performance computing, and AI, this dataset processes over 1.1 petabytes of imagery. It serves as a valuable resource across various fields, supporting tasks from urban planning to emergency response with detailed location-based data. Its precise, accessible structure facilitates cross-agency collaboration, making it an essential tool for federal, local, state, tribal, and territorial organizations. The dataset accelerates tasks like damage assessments and recovery operations, aiding communities in recovering from disasters efficiently.
Special thank you to Mark Litwintschik and his original post that outlined the initial ETL process.
Access to this dataset is available in GeoParquet, Wherobots Havasu (Iceberg), and PMTiles.
You can access the GeoParquet data directly as partitioned GeoParquet files. These are partitioned by the primary occupancy which includes:
These files are located in /geoparquet.
Example code
Havasu is a spatial table format built on top of Apache Iceberg. WherobotsDB can save its in-memory geometry and raster tables to the Havasu table format and store them on AWS S3, or any other cloud file systems or object stores.
Havasu supports ACID transactions, schema evolution, and time travel on spatial data including geometries and rasters. Havasu stores spatial data in Apache Parquet format on cloud storage, which is very cost effective and highly decoupled with computation.
To use these files first copy the files to an S3 bucket of your choosing. You can get access to a bucket by signing up for a Wherobots account here.
Once you have done that you can add a new spatial catalog using these instructions.
If you are using your own S3 bucket please refer to these instructions.
You can use the PMTiles URL in any application or tool that supports PMTiles such as Felt, leafmap, MapLibre, and the PM Tiles Viewer