Thailand Road Surface and Passability: 79,300 km of AI-Derived Metrics

Approximately 79,300 km of Thailand's arterial road network are mapped with AI-derived surface and width attributes produced by HeiGIT from 2020 and 2024 PlanetScope imagery. This resource provides surface predictions for 41,200 km of roads that currently lack surface tags in OpenStreetMap.

Use Cases

• Tracking infrastructure development by identifying segments that transitioned from unpaved to paved using the Road Surface Change (2020-2024) indicator
• Planning disaster response logistics by filtering for high-capacity routes using the Road Width Class and Humanitarian Passability Index (HPI)
• Improving OpenStreetMap data quality by comparing DL Surface Prediction values against existing surface tags to identify potential errors

Strengths

• 89.2% accuracy for AI-derived surface predictions
• Covers 79,300 km of the national arterial road network
• Includes multi-temporal data from 2020 and 2024 for change detection

Limitations

• Excludes tertiary, residential, and unclassified road types
• AI model performance may vary in areas with dense canopy cover or extreme terrain
• Humanitarian Passability Index is a static logistical model and does not account for real-time obstructions like flooding

Provenance

Source

Heidelberg Institute for Geoinformation Technology (HeiGIT)

Collection Method

Deep learning classification of PlanetScope satellite imagery integrated with OpenStreetMap road geometries.

Time Range

2020-2024

Freshness

Contains data up to 2024; last updated in early 2026.

Geography

Thailand