Using Imagery and GeoAI to Guide Urban Tree Canopy Restoration in Chattanooga, TN

In Chattanooga, Tennessee, urban planning and conservation shape a community where forests, mountains, and lakes are part of everyday life. The city is also North America’s first National Park City, a designation that recognizes its commitment to connecting urban life with nature. But as the city has grown, that balance has come under pressure. Since 1984, Chattanooga has lost nearly half of its urban tree canopy, with the greatest declines concentrated in areas where residents are already more vulnerable. 

To address this challenge, city leaders are using imagery and geospatial artificial intelligence (AI) to better understand where tree canopy has been lost and where restoration efforts will have the greatest impact. At the center of this work is the Center for Applied Geospatial Data Science (CAGDS) at the University of Tennessee at Chattanooga. CAGDS is a team of geographic information system (GIS) researchers and students who provide geospatial expertise to local leadership. 

CAGDS GIS director Charlie Mix says that Chattanooga has been described as a place where “you can’t tell where the city starts and where it ends from the surrounding mountains and forests and landscape.” 

“Wherever you go in the city, you feel a very tangible city-nature connection,” says Alex Garretson, GIS technician at CAGDS.  

That connection—and the need to preserve it—has shaped the team’s approach to conservation-focused GIS work across the region. 

In 2021, the Lyndhurst Foundation, a local philanthropic foundation, commissioned the CAGDS team to conduct a tree canopy assessment to measure the extent and impact of the city’s urban forest. “In 2021, the city hired a new urban forester…and he had no data. He had no idea about the extent of our urban forest,” says Mix. This analysis revealed that Chattanooga’s tree canopy had declined by 48 percent since 1984 and was unevenly distributed across neighborhoods.

The findings helped catalyze the Chattanooga Tree Project, a coalition effort that secured $6 million from the US Forest Service to plant 5,000 trees. This effort included CAGDS, the City of Chattanooga, and other nonprofit groups. But identifying where change had occurred was only the first step. To ensure those trees would have the greatest impact, the coalition partners needed a way to identify priority areas across the city at a much finer scale—and to do it quickly.

To guide the tree planting initiative, the CAGDS team conducted its analysis in ArcGIS Pro and used ArcGIS Image Analyst to extract insights from multiple types of imagery data. During this process, CAGDS met regularly with city leaders and Chattanooga Tree Project staff to collect feedback on its analysis and the mapping tools it was creating. From there, CAGDS published the datasets as web maps, dashboards, and geodatabases, as well as on the city’s open data portal, ChattaData.

Reflecting on her experience working with the imagery tools in ArcGIS, Nyssa Hunt, assistant GIS director for CAGDS, said that the platform’s evolving image analysis capabilities have enabled her team to complete the imagery analysis in a single environment. 

“Over time, the more that was built into ArcGIS Pro it was like ‘ok, this is honestly becoming our one stop shop.’ It’s amazing what we’re able to do in remote sensing now.”  

The project team started its analysis by seeking to identify tree canopy loss at a neighborhood level. To do this, it created a custom deep learning model to classify trees in urban areas visible via high-resolution aerial imagery of Chattanooga. The team trained the model using US Department of Agriculture National Agriculture Imagery Program (USDA NAIP) satellite imagery from 2014 and 2023 and achieved a 97 percent accuracy rate in identifying tree canopy changes over time. The model outputs were used to create a comparison across the years, showing a three percent loss in tree canopy (more than 2,800 acres) from 2014 to 2023. 

Given the scale and level of detail required, analyzing canopy change across the entire city would have been difficult to achieve with manual methods alone, making this approach a practical way to generate results within a usable time frame. 

“The results were really surprising. In my computer, I can run that deep learning model, and it will classify imagery from NAIP for the whole city in two hours, essentially doing in a day what took months of work,” explained Mix. 

Next, the team compared the percentage of tree canopy coverage between different neighborhoods in the city to identify the areas with the greatest need for tree planting—a step that revealed that the greatest loss of tree canopy existed in downtown areas. While this discovery confirmed the team’s expectations, having imagery data to ground target areas for tree planting prompted action by the Chattanooga Tree Project team. 

With its coverage analysis complete, the CAGDS team took its work a step further by estimating the total number of trees in the city using an aerial lidar (light detection and ranging) dataset. While the team encountered challenges mapping individual trees with the lidar data due to the variation in tree species and density, it found success using a pretrained deep learning model from Esri.

“Esri has a deep learning package to classify Lidar point clouds, which is awesome and worked really well. It gave a really good visualization of the structure of our [tree] canopy,” said Mix. 

This work allowed the team to estimate that the city had 5.3 million trees, revealing where most were located and where best to allocate new trees. 

“If you look at the Chattanooga Tree Project, bottom line, AI has already planted over 2,200 trees in Chattanooga,” said Mimi White, graduate research assistant GIS analyst at the University of Tennessee at Chattanooga.

With the analysis of the city’s tree canopy complete, the CAGDS team needed to make its findings actionable in order to accomplish the Chattanooga Tree Project’s goal of prioritizing tree planting efforts across the city. And the unique advantage of conducting image analysis in ArcGIS is that imagery becomes connected to rich contextual layers that help with decision-making.

To guide the distribution of tree planting work, the CAGDS team combined its imagery data with NASA’s Landsat Level-2 Surface Temperature collection and census and demographic data. From these layers, the team produced a heat risk index map to visualize the areas of the city with the greatest need for tree canopy to help local leadership prioritize its tree planting efforts. 

The map identified priority areas based on neighborhoods with less canopy and residents facing higher risk due to factors like vulnerable age groups, income levels, vehicle ownership, and access to air conditioning. The CAGDS team was able to pull these demographic variables from ArcGIS Living Atlas using GeoEnrichment.

The effect of this product was that it organized the CAGDS team’s insights and turned them into decision-ready intelligence for local leadership to direct its efforts and resources. 

“I think that’s what’s exciting for us: It’s not just a pretty map or a publication…it actually has some action tied to it,” said Mix. 

Beyond supporting planning decisions, the project has also created opportunities for students and the broader community to engage directly with Chattanooga’s efforts to restore its urban forest. Members of the Chattanooga Tree Project are now using the heat risk index map, a community ArcGIS StoryMaps story, and additional GIS products developed for the project to build local leadership support and community engagement for its tree planting efforts. 

“For this project to be approachable for students, and also for the public to be able to take part in this, it helps us to feel like we’re part of something that’s bigger than ourselves,” shared Hunt. “We’re part of our environment and we can have an impact on improving it with our tree canopy. It’s empowering to work on this project.” 

The CAGDS team used imagery as a strategic data source to support its geospatial AI workflows, enabling it to build models that scaled its analysis across the entire city. This approach allowed the team to generate results quickly while making effective use of available resources. 

The CAGDS team continues to support the Chattanooga Tree Project with GIS expertise helping translate analysis into action that supports long-term canopy restoration and the city’s connection to its natural environment.