Eritrea Road Network: AI-Derived Surface and Passability for 3,200km

This dataset maps approximately 3,200 km of arterial roads in Eritrea using AI-derived analysis of PlanetScope satellite imagery from 2020 and 2024. Produced by HeiGIT, it provides surface classification, width estimates, and passability scores for motorway, trunk, primary, and secondary road classes. The data supplements OpenStreetMap (OSM) by filling gaps where surface tags are missing.

Use Cases

• Identifying high-risk transport routes for heavy vehicles using the Road Width Class and Humanitarian Passability Index (HPI)
• Monitoring national infrastructure development by tracking transitions from unpaved to paved surfaces via the Road Surface Change indicator
• Improving humanitarian logistics planning by utilizing DL Surface Prediction (2024) to determine road suitability for different vehicle types

Strengths

• 89.2% accuracy for road surface type prediction
• Fills surface data gaps for 60.1% of the arterial road network missing OSM tags
• Provides multi-temporal analysis comparing 2020 and 2024 road conditions

Limitations

• Excludes minor road classes such as residential and local streets
• Road width is provided as a categorical class (1-3) rather than a continuous meter measurement
• AI predictions are subject to errors in areas with specific environmental conditions or imagery artifacts

Provenance

Source

HeiGIT (Heidelberg Institute for Geoinformation Technology)

Collection Method

AI-derived inference from PlanetScope satellite imagery combined with OpenStreetMap road geometries

Time Range

2020-2024

Freshness

Based on satellite imagery from 2020 and 2024; last updated in March 2026.

Geography

Eritrea