Classifying and Mapping Wetland Vegetation Assemblages in Coastal Louisiana with Landsat Imagery, 1985-2025
This study utilizes machine learning and cloud computing to generate a high-resolution, annual (1985–2025) geospatial dataset of coastal Louisiana wetland vegetation assemblages, achieving 78% classification accuracy to support the region's $50 billion restoration efforts.
Original paper dedicated to the public domain under CC0 1.0 (https://creativecommons.org/publicdomain/zero/1.0/). This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer
Imagine the Earth's skin as a giant, living mosaic. Some tiles are hard rock, some are deep blue ocean, but the most magical ones are the wetlands—spongy, grassy zones where land and water hold hands. These places are nature's shock absorbers; they soak up storms, filter dirty water, and provide a bustling apartment complex for fish, birds, and plants. But like any old house, they are changing. Sometimes the water rises too fast, or the soil gets too salty, and the plants that used to live there pack their bags and move to a new neighborhood. Scientists call this "vegetation shifting," and it's a big deal because if the plants change, the whole ecosystem changes. To understand these shifts, scientists need a map that doesn't just show where the grass is today, but where it was yesterday, and where it might be tomorrow. They need to see the story of the land over decades, not just a snapshot. This is where remote sensing comes in—using satellites orbiting high above to take pictures of the Earth, acting like a giant, patient eye that never blinks, watching the wetlands grow and shrink over time.
Now, picture a team of scientists in Louisiana, a place where the land is sinking and the sea is rising faster than almost anywhere else in the US. They have a massive, urgent mission: to save the coast, a project that costs billions of dollars and involves a 50-year plan. But you can't fix what you can't see clearly. In the past, they had to rely on two imperfect methods. One was sending people into the marsh with clipboards to count plants at specific spots—great for detail, but like trying to understand a whole city by only looking at three street corners. The other was flying helicopters over the marsh every few years to take photos—good for seeing the big picture, but the photos were taken so rarely (every six years) that they missed all the small, yearly changes, like a movie played at 1 frame per second.
The authors of this paper decided to build a better movie. They used a super-smart computer trick called "machine learning" (think of it as teaching a robot to recognize patterns) and combined it with a massive archive of satellite photos taken by Landsat. These satellites have been snapping 30-meter square photos of the Earth's surface every 16 days since 1985. The scientists taught their computer to look at these photos and identify 11 different types of marsh plants, plus mangroves, based on how the plants reflect light. They didn't just guess; they trained the computer using thousands of real-world plant surveys and then let it scan every single year from 1985 to 2025.
Here is what they found: They successfully created a new, high-definition map that shows exactly which type of marsh plant is growing where, every single year for 40 years. The computer got it right about 78% of the time, which is a solid score for such a complex task. The map reveals some dramatic stories. For instance, a type of grass called "wiregrass" has been disappearing rapidly, losing about 53 square kilometers every year. Meanwhile, a plant called "three-square" is actually expanding, growing by about 19 square kilometers a year. The map also shows that some areas are turning into open water at a rate of 29.1 square kilometers per year, confirming that the land is indeed vanishing into the sea.
The scientists also noticed that the plants are arranged in a very specific order, like a color gradient, based on how salty the water is. Close to the fresh river water, you find one type of grass; as you move toward the salty ocean, the plants change to different types that can handle the brine. The new map proves that this "salinity gradient" is a reliable rule for predicting where plants live. While the computer isn't perfect—it sometimes confuses two similar-looking grasses that live in slightly different salty conditions—it is far better than the old methods. This new dataset is now a public tool, a digital time machine that allows anyone to see how Louisiana's wetlands have changed over the last four decades, providing the clear, yearly pictures needed to help save the coast before it's too late.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.