Brazil Road Surface and Passability Data for 355,900 km of Arterials

HeiGIT produced this dataset of AI-derived surface, width, and passability metrics for 355,900 km of arterial roads in Brazil using 2020 and 2024 PlanetScope satellite imagery. It covers motorways, trunks, primary, and secondary roads, providing surface predictions for segments where OpenStreetMap tags are missing.

Use Cases

• Evaluating logistical accessibility for heavy vehicles using 'Road Width Class' and 'Humanitarian Passability Index (HPI)'
• Monitoring national infrastructure development by analyzing 'Road Surface Change' between 2020 and 2024
• Enhancing transport models by applying 'DL Surface Prediction' to road segments where the OSM 'surface' tag is missing

Strengths

• Covers 355,900 km of arterial road segments
• Validated 89.2% accuracy for surface type predictions
• Provides temporal change indicators for road paving between 2020 and 2024

Limitations

• Restricted to arterial road classes, excluding local and residential streets
• Width data is discretized into three broad classes rather than continuous measurements
• AI-derived surface predictions carry a documented error rate of approximately 10.8%

Provenance

Source

HeiGIT (Heidelberg Institute for Geoinformation Technology)

Collection Method

AI analysis of PlanetScope satellite imagery integrated with OpenStreetMap road geometries.

Time Range

2020-2024

Freshness

Last updated March 2026; contains temporal comparisons between 2020 and 2024 imagery.

Geography

Brazil