A computer model drawing on satellite and climate data could give scientists an early warning of coastal marsh decline.
Using the model, scientists detected a decline in underground plant biomass across much of Georgia’s coastal marshes between 2014 and 2023. Critically, this loss occurred even though the marsh grasses appeared green and thriving at the surface.
The findings, published last month in Proceedings of the National Academy of Sciences of the United States of America, could help land managers identify targets for restoration before more severe damage takes hold.
Roots of Concern
Marshes “are not only economically but culturally and recreationally important places for the people who both live along the coast and visit the coast.”
Marshes “are not only economically but culturally and recreationally important places for the people who both live along the coast and visit the coast,” said study coauthor Kyle Runion, a landscape ecologist at the University of Georgia. They help control flooding, sequester carbon, and provide space for hunting, fishing, and wildlife spotting.
But rapid sea level rise has threatened coastal marsh grasses, as higher waters and more frequent flooding inundate the soil and choke oxygen supply at the roots. In a healthy ecosystem, underground plant biomass staves off erosion and adds organic matter that eventually decomposes into more soil, boosting the marsh’s resilience to sea level rise, so declining root systems can be an early sign of trouble in marshlands.
Marshlands can appear healthy even as their roots are dying off, said Bernard Wood, a wetland ecologist at the Coastal Protection and Restoration Authority of Louisiana who was not involved in the study.
A trip into the marsh itself tells a different story, however. “You could just pick up this huge clump of grass with one hand, and it barely has anything holding it to the ground,” Wood said.
Sea level rise can threaten the roots of smooth cordgrass, even as the leafy part of the plant can appear healthy. The exposed roots of smooth cordgrass are seen here at a marsh edge along the Folly River in Georgia. Credit:
Kyle Runion/Colorado State University
BERM and Biomass
To understand how Georgia’s marshes are responding to changing conditions, researchers developed and tested the Belowground Ecosystem Resilience Model (BERM) in 2021. BERM draws from satellite and climate data to estimate the belowground biomass of Spartina alterniflora, or smooth cordgrass, in coastal areas.
In the 2021 study, the team collected information on environmental conditions in Georgia salt marshes from Landsat 8, Daymet climate summaries, and other publicly available datasets. They built a machine learning model that could predict belowground biomass and trained it on field data from four marsh sites. Researchers found that elevation, vapor pressure, and flooding frequency and depth were some of the most important variables in predicting root biomass.
How a salt marsh looks on the surface isn’t necessarily an indicator of how it’s truly faring.
In the new study, Runion and his colleagues applied the model to estimate changes in S. alterniflora root biomass over nearly 700 square kilometers of Georgia coast between 2014 and 2023.
During that time, belowground biomass decreased about 1% per year on average, the team found. About 72% of the salt marsh area saw declines in underground plant mass. At the same time, aboveground biomass—the visible part of the marsh grass—increased over most of the study area.
The disparity between biomass above and below could occur because aboveground biomass is less sensitive to flooding than root systems. Or the increase might be temporary, as flooding initially delivers nutrients but eventually drowns the plant. In either case, how a salt marsh looks on the surface isn’t necessarily an indicator of how it’s truly faring.
Tool for Conservation
Early-warning signs of marsh decline provided by the model could be crucial for conservation. “Once [marsh] loss occurs, that can be irreversible,” Runion said. “By getting a sign of deterioration before loss happens, that’s when we can intervene and much more easily do something about this.”
Mapping which areas of the marsh are most vulnerable could also combat the tendency to see marshes as either “doomed” or “not doomed” and target conservation efforts to the areas most in need, said Denise Reed, a coastal geomorphologist at the University of New Orleans who was not involved in the study. Though belowground biomass is declining on average, some areas of the coast are experiencing less change than others.
“There are some complex patterns going on—probably something that it would be great to understand a little bit better,” Reed said. But “this idea of being able to detect areas which are in worse condition versus areas that are in better condition from the soil’s perspective is really helpful.”
For now, BERM can predict belowground biomass only in Georgia marshes. Other regions have different plant species and flooding dynamics that could alter the relationships BERM relies on. But with additional calibration data from other salt marshes, the team could make the model more widely applicable, Runion said.
“We are looking to expand this sort of modeling framework to include different species along the Gulf and East Coast,” Runion said.
—Skyler Ware (@skylerdware), Science Writer
Citation: Ware, S. (2025), Machine learning model flags early, invisible signs of marsh decline,
Eos, 106, https://doi.org/10.1029/2025EO250253. Published on 17 July 2025.
Text © 2025. The authors.
CC BY-NC-ND 3.0Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.