AI Maps “Ghost Forests” From Maine to South Carolina

Driving along the Atlantic coast — past North Carolina’s Outer Banks or Virginia’s Eastern Shore — you may notice stands of pale, lifeless trees rising from wetlands. Some call them ghost forests, and while their haunting appearance is easy to spot, the scale of the phenomenon has been largely invisible until now.

A new study led by University of Virginia environmental scientist Xi Yang and his fourth-year Ph.D. student Henry Yeung has mapped these ghost forests in unprecedented detail, revealing nearly 6 million dead trees across low-lying areas from Maine to South Carolina. The research, published in Nature Sustainability and recently featured in Science, uses artificial intelligence and satellite data to pinpoint individual dead trees and identify the environmental forces behind their decline.

“A lot of people notice these ghost forests, but we don’t know a lot about how they’re formed,” said Yang, an associate professor in UVA’s Department of Environmental Sciences. “Our goal was to figure out the true extent of ghost forest formation and the potential drivers.”

AI and Satellite Data Illuminate a Hidden Landscape

Previous studies relied on local fieldwork — scientists wading into marshes to measure dead trees one plot at a time. However, Yang’s team used satellite remote sensing and AI-powered image analysis to create a more complete picture.

The project began with seed funding from UVA’s Environmental Institute through its Water Futures Initiative, which helped Yang and collaborators at Duke University explore where ghost forests were emerging. With additional funding from NASA, the team was able to take the project even further.

Using UVA’s Rivanna supercomputing system, Yeung trained a deep-learning model to detect not just forests but individual dead trees with the help of high-resolution satellite imagery. 

“We were able to find not just patches but every single dead tree,” Yang said. “That's a big improvement compared with all the traditional approaches used by the scientific community in the past.”

Yeung explained that AI made the analysis possible on a scale that’s never been achieved before. 

“When you’re working with satellite images, there are so many variations — lighting, texture, color,” he said. “AI can learn those patterns and decide when it’s confident that what it’s seeing really is a dead tree. At this spatial scale, it wouldn’t have been possible without AI.”

The findings revealed that saltwater intrusion and sea level rise are transforming coastal ecosystems far beyond where scientists once thought ghost forests were confined. 

“Most people assumed this was mostly a North Carolina or Virginia problem,” Yang said. “But we found dead trees all the way into New England, even in rocky coastal areas where you wouldn’t expect them.”

By analyzing tree mortality against environmental variables, the team confirmed that elevation and salinity are the strongest drivers, transforming healthy treescapes into arboreal wastelands. 

“Our analysis showed that salinity is a really important factor, and that’s never been demonstrated before at such a large scale,” Yeung said.

The research also uncovered another unexpected pattern: roads built to drain farmland can sometimes slow the advance of saltwater and reduce tree loss nearby, though not without trade-offs. 

“Roads reduce the number of dead trees by about 40 percent on average,” Yeung said, but he added that they can also accelerate flooding elsewhere.

Implications for Coastal Communities and Future Research

For Yang and Yeung, the implications of this work extend far beyond the science. Like a canary in a coal mine, ghost forests are an ecological warning sign, but they also foreshadow economic and cultural losses for coastal communities.

“When we drove through some of those areas, we noticed that there are still new houses being built right next to ghost forests,” Yang said. “That’s a clear sign those places aren’t going to have fresh groundwater for long.”

He added that flooding is already threatening local landmarks. “We saw cemeteries that are constantly flooded,” he said. “Those sites mean a lot to people, and they’re disappearing.”

The researchers hope the data can help local planners, policy makers and insurance companies anticipate and adapt to these changes. 

Ghost forests, Yeung said, are “a very visible indicator of saltwater intrusion,” which also affects agriculture and timber across the coast.

Looking ahead, the UVA team plans to expand its analysis to other U.S. coastlines, from the Gulf of Mexico to the Great Lakes and Pacific. They also hope to trace when ghost forests formed—whether the damage is accelerating with rising seas.

The project also reflects the role of UVA’s Environmental Institute in supporting research that connects data-driven science to environmental decision-making. 

“Research like this embodies what the Environmental Institute is here to do — support innovative science that informs how communities respond to environmental change,” said Karen McGlathery, the institute’s director. “It’s exciting to see that kind of impact reach from Virginia’s coastlines to the national stage.”

For Yang, the project’s success is also about training the next generation of environmental scientists. “The postdoc, Elliot White Jr., who worked with me on the pilot project, is now a professor at Stanford, and Henry is preparing to take the next step in his career,” he said. “That’s something I’m really proud of — seeing students and young scientists build on this work and carry it forward.”