Sustainable urban development relies on green infrastructure that supports both environmental health and social equity. In the U.S., unequal investment has led to persistent disparities in access to urban greenspace. As global organizations increasingly highlight the importance of green spaces for urban livability, there is a growing need for fine-scale, consistent assessments of vegetation. However, existing remote sensing approaches often miss fragmented or low-height vegetation, especially in dense urban areas.

In this study, we used a combination of aerial imagery, tree height models, and land cover data, integrated through a weakly supervised deep learning model, to generate the first nationwide, sub-meter-resolution map of urban vegetation across 254 U.S. cities. The map identifies vegetation by height class (trees, shrubs, grass) and captures details like canopy cover and seasonal patterns.

By linking this high-resolution vegetation map with neighborhood-level socio-economic data, we found clear patterns of inequality: neighborhoods with lower economic investment consistently had less tall vegetation, particularly trees. These patterns held even after accounting for climate and region, revealing systemic environmental inequities.

Our work offers a detailed and scalable tool for assessing urban vegetation, with implications for environmental justice and policy planning worldwide.