← All research
Global Streetscapes · 6.8 million street photos · ESA WorldCover

Green at Eye Level

Satellites measure tree canopy from above. Street photos measure what someone walking sees: the share of the view that is leaves, hedges and grass. The two do not always agree. Trees behind walls and in backyards count from the sky; street trees, climbers and verges count from the pavement.

FindingAcross 375 cities, street-level greenery and satellite canopy correlate at r = 0.77, but some cities, such as New York, look far greener from the pavement than their canopy suggests.

How to citeAl, Stefan. 2026. “Green at Eye Level.” Formlume. https://research.stefanal.com/street-green/

DataGlobal Streetscapes (Hou et al., NUS Urban Analytics Lab, CC BY-SA 4.0); © ESA WorldCover project 2021, contains modified Copernicus Sentinel data (2021) processed by the ESA WorldCover consortium. Please also cite the original data when you use these figures.

UseFigures on white may be reused, including commercially, with credit and under the same licence (CC BY-SA 4.0), because the street-level data are ShareAlike. The page design is © Stefan Al, all rights reserved.

ThemeHeat and climate · Streets and mobility

Every city

Satellite tree cover within 30 m of the photographed spots (across) against the green view from those spots (up). Above the line, a city looks greener from the street than its canopy suggests. Click a dot to map it.

City
0%20%40%60%+

Greener from the street

Cities whose street view is greenest relative to their canopy (among cities over 250,000 people).

Greener from the sky

Cities whose canopy hides most from the street (among cities over 250,000 people).

About the data

Street. Global Streetscapes (Hou et al. 2024, NUS Urban Analytics Lab; CC BY-SA 4.0): 10 million crowdsourced Mapillary and KartaView photos from 688 cities, each segmented into object classes. The green view index is the share of the image classed as vegetation. Only ordinary (non-panoramic) photos are used, thinned to one value per 30 m cell (the median of its photos) so that long photo sequences do not dominate; cities with fewer than 300 cells are left out.

Sky. ESA WorldCover 2021 at 10 m, via Microsoft's Planetary Computer: the share of tree cover within about 30 m of each cell. "Line" on the chart is the least-squares fit across cities.

Reading it. Crowdsourced photos follow the people who take them — main roads, tourist areas, cyclists — and seasons: a winter photo of a leafy street reads as bare. City figures describe the photographed streets, not the whole city.

Tools. Python (PyArrow and fsspec for the street-photo tables, Rasterio and pystac for WorldCover) and D3.js.