Global Roads Inventory Project (GRIP) - PBL 2018

The Global Roads Inventory Project (GRIP) consists of a global vector dataset and derived road density raster datasets. The raster datasets contain the meters of road per km2 at a 5 arcminute resolution (~8x8km at the equator) and are available for the total amount (all roads equal) and also per road type: highways (1), primary roads (2), secondary roads (3), tertiary roads (4) and local/urban roads (5).
For GRIP we gathered, harmonized and integrated nearly 60 geospatial datasets on road infrastructure into a global roads dataset. The resulting dataset covers 222 countries and includes over 21 million km of roads, which is 2 to 3 times the total length in the currently best available country-based global roads datasets.
For more information see Meijer et al (2018): https://doi.org/10.1088/1748-9326/aabd42