Scientists use AI to produce first high-resolution map of global seagrass extent

Scientists at Arizona State University have produced the first high-resolution global map of seagrass meadows using satellite imagery and a deep-learning AI model, publishing their findings in Nature. The study mapped seagrass extent across two time periods, 2019 to 2020 and 2023 to 2024, revealing that the world's seagrass covers 148,506 square kilometres, an area larger than England.

Seagrass meadows are among the planet's most powerful natural carbon sinks, absorbing carbon dioxide 35 times faster than terrestrial forests, according to UNEP data. They also protect coastlines from storm surges, filter pollutants from water, and provide critical habitat for fish, turtles, dugongs, and countless invertebrates. Despite this, they have never been systematically mapped at global scale before, largely because they are underwater and less visually striking than coral reefs.

The AI model was trained on satellite imagery and validated through field research dives and underwater cameras contributed by collaborators worldwide. The research identified the top five countries for seagrass extent as the Bahamas, the United States, Australia, Indonesia, and Cuba, which together host roughly 70 percent of global seagrass cover. Critically, only 21 percent of seagrass meadows fall within marine protected areas.

The study documented a net loss of 5,969 square kilometres, about 4 percent of total extent, over four years, with an additional 6,220 square kilometres degraded in tropical regions. Nearly 80 percent of all seagrass loss occurred outside protected areas, highlighting a clear gap for targeted conservation action. The map gives policymakers a data-driven tool for establishing new marine protected areas and for designing climate mitigation strategies that account for seagrass carbon storage.