Posted on February 27, 2023
According to a recent report in the peer-reviewed journal Science of the Total Environment, the Chesapeake Bay region could leverage artificial intelligence (AI) to track wetlands more accurately. Nonprofit Chesapeake Conservancy has developed an “AI deep learning” model to map wetlands with 94% accuracy.
Tracking wetlands has long been a challenge, owing to the difficulty of defining them and their varied appearance from above. However, AI technology, using thousands of data points fed into a machine-learning model, can identify and produce better maps of wetlands. It could even identify properties where wetland restoration might be a good fit.
Wetlands play a crucial role in maintaining water quality, supporting wildlife, reducing erosion, storm damage, and flooding, and storing more carbon than rainforests, keeping gases that contribute to climate change out of the atmosphere.
Despite their importance, the United States has lost an estimated 36.5% of its wetlands since 1900. The Chesapeake Bay Watershed Agreement aims to create or reestablish 85,000 acres of wetlands and enhance an additional 150,000 acres of degraded wetlands by 2025. But the state-federal partnership was only about 19% of the way to that goal by 2021.
The AI approach to tracking wetlands can be updated to account for environmental changes. It can be used to direct field technicians, who often need to spend time on the ground confirming the presence of wetland species. The report noted that the model could also influence policy at the local, state, and national levels to help conserve wetlands.
The Chesapeake Conservancy team trained the computer model using data from Minnesota, Delaware, and New York, which have a variety of wetland types. They tested the model by examining an area of Nebraska with an outdated wetlands analysis.
Joel Dunn, president and CEO of Chesapeake Conservancy, hopes the project will eventually influence local, state, and Baywide scale and beyond policy. A larger data layer covering an entire region could cost about $450,000, Dunn said, adding that the dawn of a new era is the application of AI and machine learning to some of the most challenging conservation problems of our time.
