Drone cameras can measure pigments indicating surface health of drylands soils

UNIVERSITY PARK, Pa. — Drylands soils are being degraded by human activities and climate change, threatening the thin living layers covering their surface, called biocrusts, according to the U.S. Geological Survey. Composed of cyanobacteria, lichens, mosses, algae, fungi and bacteria, biocrusts play major roles in storing carbon, cycling nitrogen, preventing erosion and supporting ecosystems. To better understand the status of these complex biological communities, a team led by researchers at Penn State developed a novel approach for assessing the health and function of biocrusts.

In a study recently published in Remote Sensing in Ecology and Conservation, the team described a method using cameras on drones and multispectral imagery — the analysis of image data across multiple wavelength bands of the electromagnetic spectrum, including invisible radiation like near-infrared, ultraviolet and thermal infrared — to measure pigments created by microorganisms in the soil surface that reveal how healthy or developed the biocrust is.

Even though biocrusts cover only about 12% of Earth's land, they have a disproportionately large ecological impact, according to team leader and senior author Estelle Couradeau, assistant professor of soils and environmental microbiology in the Penn State College of Agricultural Sciences.

“Climate change and increasing land use are expected to reduce biocrust cover by 25% to 40% within about 65 years,” she said. “Not only will there be fewer biocrusts, but their composition may change. Since different biocrust types perform ecosystem functions differently, this could change how dryland ecosystems work. So, we need inexpensive and effective monitoring methods.”

Monitoring biocrusts is difficult because they are tiny, patchy and hard to identify, Couradeau explained. Studying them usually requires field sampling, laboratory work and expert identification, all of which are expensive, slow and impossible over huge areas. That’s why the team focused on developing a remote sensing approach with drones.

In the study, the researchers measured scytonemin, a pigment made by cyanobacteria that acts as a sunscreen to protect itself against intense UV radiation and indicates mature, well-developed biocrusts; chlorophyll a, the main photosynthetic pigment that indicates living biomass; and carotenoid, a pigment involved in photosynthesis and protection from light stress. The researchers also measured soil organic carbon — an indicator of stored carbon — and nitrogen — an essential nutrient in arid and semi-arid regions.

The research was conducted at two different semi-arid ecosystems in the U.S. Southwest: the Jornada Experimental Range within the Chihuahuan Desert of southern New Mexico, and the Green Butte site within the Colorado Plateau Desert, north of Moab, Utah.

At each location, the researchers used a drone to take extremely high-resolution multispectral photographs — less than a half inch of land surface per pixel. Using machine learning — a type of artificial intelligence — the researchers predicted levels of pigments, carbon and nitrogen in the soil from the images. They compared those predictions to lab measurements of soil samples collected from the photographed plots.

The method worked well in New Mexico, where the machine learning models predicted the scytonemin pigment with 93% accuracy, noted study first author Raúl Román, who was a Maria Skłodowska Curie postdoctoral fellow in the Department of Ecosystem Science and Management at Penn State when the research was conducted and is now a postdoctoral fellow in environmental sciences at the University of Almería in Spain. The prediction for soil organic carbon was even better, at 95%. The prediction for nitrogen was reasonably accurate at 80%.

The method didn't work as well in Utah. Overall prediction accuracy dropped below 60%. Román theorized that this was caused by scytonemin saturation, meaning many biocrusts already contained very high amounts of scytonemin. He explained that, as a result, the camera couldn't distinguish between “high” and “very high” levels — similar to how an overexposed photo loses detail.

“The main goal of this study was to evaluate multispectral imagery as a tool to monitor indicators of biocrust cover and function, along with their sensitivity to soil moisture to pave the way toward the development of remote-sensing-based biocrust-health indicators,” Román said. “This study provides the first demonstration that multispectral remote sensing data can be used to quantify the UV-protective pigment scytonemin, the most abundant pigment in biocrusts.”

Remote sensing of biocrusts — now from drones, later perhaps from aircraft and satellites — can monitor large areas much faster than field surveys, Couradeau said, explaining that this study’s findings suggest that this monitoring could be even more precise with multispectral imagery that can differentiate between healthy and declining biocrusts.

“Current remote sensing mainly tells us where biocrusts are, but it does not tell us how healthy or functional they are,” Couradeau said. “Knowing about their presence alone isn't enough. We also need to know: Are they actively photosynthesizing? Are they stressed? Are they storing carbon? Are they functioning normally?”

Contributing to the research were: Ryan Trexler, Intercollege Graduate Degree Program in Ecology, Huck Institutes of the Life Sciences, Penn State; Tong Qiu, assistant professor of ecology, Duke University; Fernando Maestre, professor of environmental science, King Abdullah University of Science and Technology, Saudi Arabia; Elizabeth La Rue, assistant professor in biological sciences, University of Texas, El Paso; Sergio Vargas Zesati, postdoctoral research fellow, University of Texas, El Paso; Anthony Schaefer, graduate student, New Mexico State University; and Nicole Pietrasiak, associate professor of sustainability in arid lands, University of Nevada.

The work conducted by Penn State researchers was supported by the U.S. Department of Agriculture’s National Institute of Food and Agriculture and Hatch Appropriations under project number PEN04949 and accession number 7006508. Full funding details are available in the paper. This content is solely the responsibility of the authors and does not necessarily represent the views of the funders.