Iceland Road Network: AI-Derived Surface and Passability for 9,000 km

HeiGIT generated this geospatial dataset covering approximately 9,000 km of arterial roads in Iceland using AI analysis of PlanetScope satellite imagery from 2020 and 2024. It provides deep-learning predictions for road surface, width, and humanitarian passability for motorway, trunk, primary, and secondary road classes.

Use Cases

• Evaluating logistical constraints for heavy vehicles using Road Width Class and Humanitarian Passability Index (HPI)
• Identifying infrastructure upgrades by filtering for Road Surface Change from unpaved to paved
• Enhancing OpenStreetMap data by filling missing surface tags with DL Surface Prediction values

Strengths

• 89.2% accuracy for road surface model
• Covers 9,000 km of arterial roads
• Multi-temporal data for 2020 and 2024
• Standardized HPI scoring

Limitations

• Geographic focus restricted to arterial road classes
• AI-derived predictions may contain errors compared to ground truth
• Small total unpaved road sample size of 0.0001 million km

Provenance

Source

Heidelberg Institute for Geoinformation Technology (HeiGIT)

Collection Method

Deep learning analysis of PlanetScope satellite imagery combined with OpenStreetMap road geometries

Time Range

2020-2024

Freshness

Last updated March 2026; contains imagery-derived data from 2020 and 2024.

Geography

Iceland