Kazakhstan Road Surface and Passability Data for 49,000 km of Arterials

Supplying AI-derived surface and width attributes for approximately 49,000 km of arterial roads in Kazakhstan, produced by HeiGIT using 2020 and 2024 PlanetScope satellite imagery. It supplements OpenStreetMap (OSM) data by filling surface tag gaps for 24.6% of the network with predictions that achieve 89.2% accuracy. The data covers motorway, trunk, primary, and secondary road classes and their associated links.

Use Cases

• Calculating logistical accessibility for heavy vehicles using the 'Road Width Class' and 'Humanitarian Passability Index (HPI)'
• Monitoring national infrastructure growth by identifying segments in 'Road Surface Change' that transitioned from unpaved to paved
• Improving rural navigation routing by utilizing 'DL Surface Prediction' to identify paved vs. unpaved segments missing from OSM

Strengths

• 89.2% accuracy for AI-derived surface models compared to 64.7% for standard OSM tags
• Covers 49,000 km of the national arterial road backbone
• Includes multi-temporal analysis (2020 vs 2024) to track infrastructure development

Limitations

• Analysis is restricted to arterial roads and excludes local or residential street networks
• Road width is provided as categorical classes (1-3) rather than continuous meter measurements
• Surface predictions include an 'Unknown' category where imagery quality was insufficient for classification

Provenance

Source

Heidelberg Institute for Geoinformation Technology (HeiGIT)

Collection Method

Deep learning models applied to PlanetScope satellite imagery (3m resolution) integrated with OpenStreetMap vector data.

Time Range

2020-2024

Freshness

Last updated March 2026; utilizes satellite imagery from 2020 and 2024.

Geography

Kazakhstan