Researchers from Heidelberg University and HeiGIT have released the first global, multi-temporal road surface dataset, mapping pavedness, width and passability for 9.2 million km of arterial roads in 2020 and 2024 with 89.2 per cent accuracy.
Using geospatial AI and PlanetScope imagery, the open dataset reveals global infrastructure disparities and supports planning, logistics, climate resilience and humanitarian operations.
Accurate, high-resolution, and up-to-date global data on road surface conditions is a critical logistical asset and a fundamental prerequisite for monitoring economic progress, targeting infrastructure investment, and assessing climate vulnerability.
To address this, researchers of Heidelberg University and HeiGIT (Heidelberg Institute of Geoinformation Technology) developed the first global, multi-temporal global road surface dataset detailing road pavedness and width and passability.
Using geospatial AI and PlanetScope satellite imagery, researchers mapped 95.5 per cent of road surface type and width across 9.2 million km of critical arterial roads worldwide for 2020 and 2024, achieving an overall accuracy of 89.2 per cent. This provides a much more complete and up-to-date picture of the physical state of global road infrastructure.
The dataset also captures changes in road conditions over time, providing valuable information for humanitarian applications and serving as an important indicator for assessing socioeconomic development, particularly in data-scarce regions.
The analysis published in Nature Communications on September 1, 2026, demonstrated the dataset’s utility for research, logistics operations, and decision-making. The dataset is publicly available via a platform operated by the United Nations Office for the Coordination of Humanitarian Affairs.
The most significant increases were concentrated in lower- and lower-middle-income regions across South America, Africa, parts of South Asia, and Southeast Asia, visually pinpointing global hotspots of rapid road development. While the overarching trend was positive, localised areas of apparent decrease in pavedness were also present.
High- and high-middle-income regions such as Europe and Central Asia and North America show near-complete paved networks (97.4 per cent and 96.9 per cent, respectively).
In stark contrast, Sub-Saharan Africa averages only 63.1 per cent pavedness, with a “significantly larger variance indicating high intra-regional heterogeneity”.
The data further revealed that the primary driver of this disparity is rural infrastructure; while urban road networks are largely paved across all regions (93 per cent), rural pavedness in Sub-Saharan Africa (61.4 per cent) lags dramatically behind that of Europe and Central Asia (97.2 per cent).
A central finding of the study is that, when analysed at high resolution, the rural road network serves as a sensitive indicator of a nation’s stage of development. The study quantifies the pronounced urban-rural divide and its strong correlation with socioeconomic status.
Even in high-income regions, rural networks were less paved. This disparity was more pronounced in low- and lower-middle-income countries. Notably, a belt of low pavedness was evident across Central Africa and parts of South America.
The strength of the multi-scale framework is most evident at the local scale, where it reveals the underlying drivers and real-world consequences of these broader infrastructure patterns.
Two case studies illustrate this in distinct contexts: in Ghana, the data reveal how fragmented governance structures manifest as deep-seated inequities in urban road quality, while in Pakistan, the Humanitarian Passability Matrix provides actionable intelligence to assess climate vulnerability and strengthen humanitarian response logistics. The ability to bridge global dynamics to local realities is a key contribution of the work, the study said.
“By looking at how road surfaces change over time, we show how infrastructure development and investment can be tracked across different timescales — offering a high-resolution proxy for economic activity (when compared to traditional proxies such as nightime lights) that directly captures the physical infrastructure underpinning resilience and development,” said Sukanya Randhawa, who leads the “GeoAI for Good” projects at HeiGIT and the lead author.
And this opens the door to many downstream applications across development, mobility, climate adaptation, humanitarian response, infrastructure planning, and equity, Randhawa added.
The study cautioned that the findings should not be interpreted as a universal call for paving all roads, but as a framework for understanding strategic infrastructure trade-offs. While paved roads are associated with higher development indices, unpaved roads continue to play an essential role in many regions due to their cost-effectiveness and, in some contexts, lower environmental footprint.
“The dataset should therefore be viewed as a decision-support tool to identify where surface upgrades may yield the greatest gains in resilience and connectivity, while recognizing the continued value of well-maintained unpaved infrastructure,” the study said.
The study also underscored a critical insight for the big data era: while crowdsourced platforms like OpenStreetMap (OSM) OSM are foundational, their attribute data often fails to keep pace with real-world infrastructure change, especially in global regions without very active OSM communities.
The study’s human-validated assessment revealed that OSM surface tags are frequently outdated, achieving only 26 per cent global average accuracy on unpaved roads. Rather than replacing community mapping, the AI-driven approach serves as a powerful complement—offering a scalable mechanism to systematically enrich and update these globally important datasets.