Texas A&M Forest Service has developed an artificial intelligence model that expands access to detailed urban tree canopy maps for communities across Texas, helping cities better understand and manage their tree resources.
Developed through the agency’s Spatial Solutions department, the model uses deep learning technology, a type of artificial intelligence that learns to recognize patterns and data, and publicly available aerial imagery to identify tree canopy coverage. It achieved an average overall mapping accuracy of 94.5% when evaluated across 295 Texas municipalities, according to findings detailed in a recently published scientific paper.
Urban tree canopy refers to the layer of leaves, branches and stems of trees visible from above. Understanding where tree canopy exists and how it is distributed can help communities make informed decisions about tree planting, urban development and long-term forest management.
Trees provide important benefits in urban environments, including reducing temperatures through shade, improving air quality, managing stormwater runoff and lowering building energy demands. Accurate canopy maps help communities assess these resources and identify areas where additional tree coverage may be beneficial.
Traditionally, detailed tree canopy mapping has relied on technologies such as light detection and ranging, or LiDAR, which uses laser pulses to collect three-dimensional information about the landscape. While LiDAR can provide highly accurate tree canopy measurements, the availability of this data varies across Texas.
“One of the biggest challenges in urban tree canopy mapping is the availability of LiDAR data,” said Joe Gorman, Texas A&M Forest Service geospatial analyst. “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.”
To address this limitation, the model was trained using LiDAR-derived tree canopy data from just four cities and aerial imagery from the National Agriculture Imagery Program, a publicly available imagery
“Traditional machine learning approaches to tree canopy mapping have often reported accuracies around 70–80%,” Gorman said. “With recent advancements in deep learning, we’re seeing significant improvements in what’s possible. Achieving an average accuracy of 94.5% using training data from just four cities demonstrates the potential of these technologies to produce reliable, high-resolution canopy maps at statewide and potentially national scales.”
The model has expanded the number of communities represented in the Texas A&M Forest Service Urban Tree Canopy application, particularly in areas without LiDAR coverage, while providing opportunities to monitor canopy changes over time. It is also available as a deep learning package for Esri users to incorporate into their own applications.
The findings are detailed in Gorman’s scientific paper, “Bridging LiDAR and NAIP: A scalable deep learning framework for statewide urban tree canopy mapping,” published in the journal Urban Forestry & Urban Greening. The work was supported in part by funding from the USDA Forest Service’s State, Private and Tribal Forestry program.
To explore urban tree canopy information for Texas communities, visit the Urban Tree Canopy tool texasforestinfo.tamu.edu
Savanna Montgomery
Media Contact
Texas A&M Forest Service Communications Office
Email address:
[email protected]
Phone number:
979-458-6606
