USA Road Surface Data: 111 Million KM of Paved and Unpaved Classifications

HeiGIT produced this geospatial dataset in 2026 covering 111.7 million km of road networks across the United States. It integrates OpenStreetMap (OSM) attributes with deep learning predictions from Mapillary imagery to classify road segments as paved or unpaved.

Use Cases

• Infrastructure assessment using pred_class and smoothness to identify roads requiring paving
• Logistics planning by filtering highway types and width for heavy vehicle routing
• Socioeconomic analysis by comparing urban classification with road surface quality in combined_surface_osm_priority

Strengths

• Covers 111.7 million km of road segments
• Adds 5.08 million km of surface data via Mapillary-derived AI predictions
• Includes specific urban/rural classification using GHSU and AFRICAPOLIS layers

Limitations

• 65% of the road network lacks surface information
• AI predictions are dependent on the availability of Mapillary street-level imagery
• Potential for model error in the pred_class field compared to ground-truth OSM tags

Provenance

Source

HeiGIT (Heidelberg Institute for Geoinformation Technology)

Collection Method

Hybrid approach combining OpenStreetMap data with deep learning predictions from Mapillary street-level imagery.

Freshness

Last updated March 2026.

Geography

United States of America