Texas A&M Forest Service says it has built an artificial intelligence model that can map the tree canopy of Texas cities from ordinary aerial photographs, a job that until now has leaned on expensive laser surveys many towns cannot afford.
The agency, headquartered in College Station, announced the work Monday. The model was developed in its Spatial Solutions department and uses deep learning, a form of AI that learns to pick out patterns, to find tree cover in photos taken for the National Agriculture Imagery Program, a federal effort that is free to the public and generally refreshed every two years.
How accurate it is
Tested across 295 Texas municipalities, the model averaged 94.5% overall mapping accuracy, according to a paper by agency geospatial analyst Joe Gorman published in the journal Urban Forestry & Urban Greening. Gorman said older machine learning methods for canopy mapping have often landed in the 70% to 80% range.
The model learned from LiDAR-based canopy data covering only four cities. LiDAR, short for light detection and ranging, fires laser pulses from aircraft to build a three-dimensional picture of the ground and is considered highly accurate, but coverage is uneven across Texas.
"Many communities across Texas need accurate, high-resolution canopy maps to support planning and urban forest management, but don't necessarily have the resources to acquire updated LiDAR data on a regular basis," Gorman said in the agency's announcement.





