Which way do a city's streets run? A planned grid sends most of them along two axes; a city that grew street by street spreads them across every bearing. Measured for all 10,351 urban areas on Earth, using Geoff Boeing's street-network models.
FindingThe planned grid is a North American, Chinese and Punjabi pattern; Europe is the least gridded region.
How to citeAl, Stefan. 2026. “The Grain of the World's Streets.” Formlume. https://research.stefanal.com/streets/
DataBoeing, Global Urban Street Networks (Harvard Dataverse), from © OpenStreetMap contributors; European Commission JRC (GHSL). Please also cite the original data when you use these figures.
UseFigures on white may be reused for non-commercial purposes with credit (CC BY-NC 4.0). The page design is © Stefan Al, all rights reserved; for commercial use, please get in touch.
ThemeStreets and mobility
Each rose shows how much street length runs along each compass bearing, in 10° slices. Cities are sorted from the most gridded to the least. Order runs from 0 (bearings spread evenly in every direction) to 1 (a single perfect grid).
Each dot is one urban area; brighter means a more gridded street network, and size follows population. North America is the most gridded region, but not the only one: China's planned cities and the old canal-colony towns of Pakistan's Punjab, such as Faisalabad, are among the most ordered in the world. Europe, whose cities grew street by street, is the least.
Orientation order for every urban area of 500,000 people or more, grouped by UN region. The tick marks each region's median.
Street networks and indicators: Geoff Boeing, Urban Science Beyond Samples: Up-to-Date Street Network Models and Indicators for Every Urban Area in the World, Environment and Planning B 53(8), 2026; data on the Harvard Dataverse. Models are built with OSMnx from OpenStreetMap (2025) inside the boundaries of the GHSL Urban Centre Database 2025.
Orientation order follows Boeing (2019): the entropy H of street bearings in 36 bins, rescaled as 1 − ((H − Hgrid) / (Hmax − Hgrid))², where Hmax is a perfectly uniform spread and Hgrid a single perfect grid. The roses are computed here from each city's edge list: straight-line bearings of each street segment, weighted by length and counted in both directions.
An urban area is a GHSL urban centre: continuous settlement above 1,500 people per km², not a municipality. Large centres can join several cities (Kyoto falls inside the Osaka–Kyoto–Kobe centre, for example), and the rose then describes the whole centre. This scale also explains some surprises: Manhattan and Barcelona's Eixample are textbook grids, but their urban centres also contain many differently angled grids and older street patterns, so New York and Barcelona score low overall.
OpenStreetMap coverage is uneven; small or fast-growing areas may be missing streets. Population is the GHSL 2025 resident population. Locations are the centroids of each urban centre.