Harnessing "Microbiome Matchmaking" for Coastal Seagrass Resilience

As climate change intensifies, a dangerous feedback loop emerges: dwindling biodiversity erodes the very natural systems essential for climate stability.

Coastal and marine ecosystems, such as salt marshes and seagrass meadows, capture carbon from the air and store it in the soil beneath coastal plants, locking away these massive carbon deposits for thousands of years. This process, known as blue carbon, reduces the impact of greenhouse gases and slows global warming.  

This defense, however, is fragile.

When these ecosystems are damaged or destroyed, they release centuries of stored carbon back into the atmosphere, accelerating the warming they help prevent. Strengthening these habitats does more than just salvage the environment, it fortifies a vital line of defense for the human communities and economies that depend on a resilient, stable coastline.

Turning the tide on this ecosystem loss requires innovative intervention.

Dr. Geoffrey Zahn, an assistant professor of applied science at William & Mary, is pioneering new methods to ensure these coastal defenses survive. He was recently awarded a $369,000 grant from the Revive & Restore Climate Resilience Fund to develop an AI-powered predictive decision framework, a cutting-edge approach designed to dramatically improve the success and reliability of coastal eelgrass restoration efforts.

In addition to storing large amounts of carbon, underwater eelgrass meadows protect shorelines from erosion, improve water quality, and provide habitat for fish and shellfish. Though eelgrass is declining in many regions, restoration efforts thus far have been unpredictable.

Zahn’s methodology builds on existing techniques like microbial inoculation, where beneficial microbes are introduced to support plant resilience, and microbiome engineering, where microbial communities are deliberately designed to enhance ecosystem function.

The most novel aspect of Zahn’s strategy lies in predicting microbial compatibility before being introduced to eelgrass restoration sites.

The first step in the process involves crossing sediment microbes from one location (donor/probiotic) with sediment microbes from another location (resident/restoration site). These combinations are then introduced to a seawater table in a greenhouse where eelgrass seed germination and plant growth are monitored, with the data being used to train a graph neural network.

“In this case, hundreds of introduced microbial species are combined with an existing community of thousands more,” said Zahn. “Those interactions help determine whether restoration succeeds or fails.”  

Zahn provided a simple example: one microbe (A) might inhibit a beneficial species (B) from establishing, but that effect could change if a third microbe (C) is present that changes the interaction between A and B. The model learns these context-dependent interactions across thousands of species (meaning billions of possible interactions), then predicts the identity of the merged microbial treatment and whether it will establish and support eelgrass growth at specific restoration sites.

This approach is time and cost-effective, avoiding trial-and-error by identifying promising options before even going into the field.

The AI-predicted microbial treatments will be introduced at eelgrass restoration sites in Virginia and New Hampshire, requiring a number of partners throughout the project.

Local field restoration and monitoring in the Chesapeake Bay are anchored by William & Mary’s Batten School & VIMS, who also provide essential eelgrass expertise and greenhouse facilities.

The University of New Hampshire contributes critical expertise in coastal microbial ecology, helping to design and interpret experiments across diverse environmental gradients, while the Piscataqua Region Estuaries Partnership will coordinate the field work, site selection, and permitting required for the New Hampshire sites.

Finally, to ensure the science translates into real-world action, the Institute for Integrative Conservation and The Nature Conservancy are working to connect the team's findings directly to coastal management efforts and restoration practitioners on the ground.

Together, this collaborative team will monitor which microbial communities establish and compare them to model predictions. They will also track how eelgrass responds to the treatment — including seedling emergence, survival, and growth over time — and evaluate how closely these outcomes align with the model’s predictions.

While the project is focused on restoring coastal eelgrass, the success of this approach could have wide-ranging implications.

“Together, these comparisons tell us whether the graph neural network can both predict restoration outcomes and reliably forecast how microbial communities will develop,” Zahn explains. “That kind of predictive ability could be useful in many other settings where microbiomes matter, from agriculture to human health.”