K-State Research Takes Flight to Combat Woody Encroachment with Precision Mapping
As you traverse the Konza Prairie this summer, you might notice something unusual soaring above. It’s not a bird or a plane, but rather K-State’s cutting-edge research at work.
Kansas State University (K-State) researchers are employing aerial data in their mission to better understand and manage the swift spread of woody plants across the Great Plains. This phenomenon, known as woody encroachment, involves the transformation of open grasslands into areas dominated by shrubs and trees. It has significant impacts on biodiversity, livestock forage, water resources, and even wildfire risk.
“Woody encroachment is a pattern occurring in grasslands worldwide,” explained Zak Ratajczak, an assistant professor of biology at K-State. “Areas once home to grasses, wildflowers, and other herbaceous species are witnessing a rapid increase in shrubs and trees. While not all changes are negative, these new arrivals often drastically alter grassy ecosystems.”
A recent study, led by K-State master’s student Brynn Noble and Ratajczak, was published in the open-access journal Remote Sensing. It offers a cost-effective, open-access approach to detecting woody encroachment across landscapes as small as six-by-six feet. The system combines aerial imagery from federal programs with ground-based data collected by K-State researchers.
“We can identify whether an area is a grassland, shrub, tree, or evergreen tree with approximately 97% accuracy,” Ratajczak stated. “This means that when we label that pixel as a specific type of vegetation, we’re correct 97% of the time.”
This impressive accuracy is achieved through a blend of resources: consistent, high-resolution aerial data courtesy of the USDA and National Science Foundation; open-source machine learning computational power; and extensive training datasets gathered during fieldwork at the Konza Prairie Biological Station. Much of this work is carried out by undergraduate and graduate students.
The high-resolution aerial data is obtained using LiDAR and multispectral imagery from aircraft equipped with advanced sensors. These sensors provide detailed, three-dimensional views of the landscape, capturing both the structure and composition of vegetation to monitor changes such as woody encroachment with exceptional precision.
Ratajczak noted that many people are unknowingly familiar with these machine learning models. “If you’ve ever been asked to click all the tiles showing a crosswalk or bus in a series of photos, then you’ve participated in image-based machine learning,” he said. “Computers can learn to identify crucial features in on-the-ground or aerial photos, but to do so accurately, they need plenty of training samples to discern patterns worth picking and features to disregard. For instance, early project versions had the computer mistakenly identifying deep shadows behind trees as miniature evergreens.”
This accuracy requires thousands of samples, a task K-State students and the Konza Prairie are well-suited to handle. The Konza, a long-term ecological research site managed by K-State and The Nature Conservancy, provided the diverse vegetation necessary to teach the machines accurately.
“In this relatively small area of about 10 square miles, we have many major vegetation types found across eastern Kansas,” Ratajczak said. “When training a computer to identify a variety of vegetation types, this diversity is invaluable.”
Researchers, including students, can promptly validate or correct model predictions based on firsthand landscape knowledge. Students gather training data for the models, gaining experience in Geographic Information System (GIS) and computer coding in the process.
One of these students is Brynn Noble, co-lead of the research study. Her contributions, alongside Ratajczak, are poised to help develop this research into a regional, potentially statewide, mapping tool.
“One catalyst for expanding this work beyond our Kansas site is the federal and state-funded aerial campaigns that gather hard-to-obtain data and make it available for free. The work wouldn’t have been possible without data from the NSF and the USDA,” Ratajczak noted.
Moving beyond classification, this research allows researchers to estimate how many cattle or bison an area can sustain long-term, adjusting calculations for habitats and supporting K-State’s bison herd decisions.
Other K-State teams use woody vegetation maps to assess fire risks, habitat selection by birds and small mammals, and even tick-borne disease exposure. “It’s been an unexpected journey. We’re contributing to research on stream flow and bird diversity,” Ratajczak shared. “A study led by Katy Silber and Alice Boyle showed our data helped reveal how minimal woody cover reduced grasshopper sparrows’ habitat usage.”
As woody encroachment affects various species, Ratajczak hopes to expand the project’s reach through a website or app for regional early-warning systems. The goal is to collaborate with landowners to gather more training data and refine models further.
“If our models err, we want to know to make updates. Long-term, we aim to use satellite data for larger area coverage and modeling past and future projections,” Ratajczak explained.
For now, this technology, built from open-source tools and federally funded data, is already impacting grassland conservation. “This is real-world skill development,” Ratajczak concluded. “It’s advancing our comprehension of one of the major environmental challenges in grasslands.”
