An OpenStreetMap derived building classification dataset for the United States.

Area of Science:

• Geospatial science
• Urban informatics
• Data science

Background:

• Building classification data is essential for urban planning, population estimation, and emergency response.
• Readily available, comprehensive building type data for the entire United States is lacking.
• Existing datasets often lack the scale or detail required for nationwide applications.

Purpose of the Study:

• To develop a comprehensive dataset classifying residential and non-residential buildings across the entire United States.
• To provide a valuable resource for urban planners, transportation planners, and researchers.
• To improve the availability of crucial building type information for various applications.

Main Methods:

• Utilized building footprints and OpenStreetMap data for classification.
• Developed a methodology for classifying millions of building structures.
• Validated the dataset using authoritative ground truth data in select U.S. counties.

Main Results:

• Generated a dataset classifying 67,705,475 buildings nationwide.
• Achieved high precision in non-residential building classification.
• Demonstrated high recall for residential building classification.

Conclusions:

• The developed dataset significantly enhances the availability of building type information for the U.S.
• The classification methodology proves effective for large-scale geospatial data analysis.
• This resource is expected to benefit urban planning, traffic analysis, and emergency response efforts.