USC researchers have released a free AI tool that maps urban tree canopy using publicly available aerial photographs. The system eliminates the need for costly lidar surveys or commercial satellite imagery, giving cities a financially accessible way to target tree planting as extreme heat events become more frequent and intense.
"Trees provide a wide range of benefits, including helping reduce the health risks associated with rising temperatures in cities," said John Wilson, founding director of the Spatial Sciences Institute at the USC Dornsife College of Letters, Arts and Sciences and the study's corresponding author. "But to plant trees where they'll make the biggest difference, cities first need a clear picture of their existing tree canopy."
The tool, described in a study published in Remote Sensing, uses free aerial imagery collected every two to three years through the U.S. Department of Agriculture's National Agriculture Imagery Program (NAIP). By training a deep learning model on these images, the team built a canopy-mapping system that performs competitively with far more expensive lidar-based methods.
How the tool works
The system includes two AI models: one that identifies tree canopy cover and another that detects individual trees. The individual detection task is especially difficult because tree crowns in aerial images are small and often overlap. The model matched the accuracy of lidar-based approaches in tests, without requiring the specialized hardware or commercial satellite data that many municipalities cannot afford.
"We need fine-scale data to know where to plant new trees that provide the best return on investment," Wilson said. "Our work shows how we can use free, publicly accessible data to map tree canopy over time - data cities can use to guide planting plans at every scale, from a single street block to an entire county."
Testing across cities
The researchers first deployed the tool in Boyle Heights and City Terrace, two densely populated Los Angeles neighborhoods with historically less tree cover than wealthier areas. The model accurately mapped canopy cover and detected individual trees. To test the approach's generalizability, they applied the trained models - without retraining - to neighborhoods in San Francisco and Phoenix. Despite differences in climate and urban layout, the tool produced consistently strong results, suggesting that other communities can adopt the model rather than building one from scratch.
Next steps and open access
Wilson's team plans to pair the AI tool with freely available lidar data that captures the height and 3D structure of tree canopies. "Knowing both the height and extent of the canopy will allow us to estimate the shade trees provide today - and model how much additional shade new plantings could create," Wilson said. The researchers will begin analyzing individual street blocks, school playgrounds, and parks before scaling up to entire neighborhoods and counties.
The ArcGIS deep learning package developed through the project has been downloaded more than 12,900 times from Esri's Living Atlas platform over the past six months, according to Wilson. The research, code, and a ready-to-use ArcGIS model are freely available online, enabling municipalities without in-house machine learning expertise to use the tool. This release reflects a broader trend in AI for Science & Research where open-source tools accelerate public benefit. The tool can also help cities make data-driven decisions about green infrastructure - a practical example of AI for Government.
Why this matters for IT and development professionals
The tool is built entirely on free, publicly available data and a downloadable deep learning model, so developers can adapt it for local geospatial projects without paying for lidar surveys or commercial satellite imagery. The ArcGIS integration means GIS teams can incorporate it into existing workflows, and the model's transferability across cities reduces the need for custom training data. For IT and development teams in municipal government or urban planning, this stack represents a low-cost way to add canopy analysis to climate resilience tools.
Your membership also unlocks:
