AI uses a pigment to reveal the condition of biocrusts

For the first time, researchers have remotely measured the concentration of scytonemin, an indicator of the development and resilience of biocrusts in arid ecosystems

by Matteo Cavallito

Drones and multispectral imagery can help assess the condition and functions of biocrusts, the thin living communities found on the surface of dryland soils. This is the finding of a Penn State University study that combined reflectance data and machine-learning models to estimate levels of soil organic carbon and nitrogen and, above all, the concentrations of certain pigments. These include scytonemin, a pigment produced by certain cyanobacteria to protect themselves from ultraviolet radiation which, according to the research published in the journal Remote Sensing in Ecology and Conservation, “is a key functional indicator of biocrust stress tolerance and resilience while influencing albedo and surface temperature in drylands”.

Biocrusts are the living skin of drylands

Biocrusts are assemblages of soil particles and various organisms, including lichens, mosses, algae, fungi and bacteria. They often occupy the spaces between plants and form a sort of “living skin” across drylands. They also stabilize the soil, the authors note, limit erosion and dust emissions, and contribute to global carbon and nitrogen cycles. However, land-use intensification and climate change threaten their survival. “Although biocrusts currently cover up to 12% of Earth’s land area”, the study explains, “the synergistic contribution of land-use intensification and climate change is projected to decrease their cover by 25%–40% within 65 years”.

Climatic stress, moreover, “may concurrently induce shifts in biocrust type dominance, favoring cyanobacteria-dominated crusts over moss or lichen-dominated biocrusts and altering ecosystem functioning”.

Monitoring their cover, composition and condition is therefore essential, yet far from easy. Biocrusts are small and patchily distributed, and studying them often requires sampling and complex laboratory analyses. Remote sensing already makes it possible to detect and map them, but provides more limited information about their functional status. Chlorophyll a, one of the indicators used to assess biocrusts, provides information about their photosynthetic biomass. However, it is also sensitive to moisture, and its signal weakens during dry periods. Scytonemin, the researchers argue, could offer a more stable indicator.

Accuracy reaches 93%

The study was conducted in two semi-arid ecosystems in the southwestern United States with different soils, climates and biocrust communities: the Jornada Experimental Range in the Chihuahuan Desert in New Mexico and Green Butte on the Colorado Plateau in Utah. After acquiring ultra-high-resolution multispectral images, the researchers collected samples from the same locations and measured scytonemin, chlorophyll a and carotenoids in the laboratory. Five machine-learning models were then tested to determine whether the images could accurately estimate the concentrations measured in the laboratory.

The strongest results were obtained in New Mexico, where soil organic carbon and nitrogen were also analyzed. “Results showed that scytonemin was the pigment at New Mexico with the highest prediction accuracy (93%) under dry conditions, while chlorophyll a had the best predictive power under wet conditions”, the study explains.

This, the authors continue, “suggests that scytonemin remote detection can complement or surpass chlorophyll a, which is more dependent on soil moisture”. This is particularly promising because biocrust-dominated ecosystems can remain completely dry for up to 85% of the year. Dark biocrusts rich in cyanobacteria and lichens contained more scytonemin and reflected less visible light.

A method still to be refined

In New Mexico, the models developed using small plots of land were also applied to drone imagery covering a larger area and successfully identified differences between the various stages of biocrust development. However, these same models “did not capture the exact values of the modeled parameters compared to local measurements”.

In short, the method “shows promise for future integration with satellite imagery to expand biocrust trait mapping at broader scales, offering a valuable tool for monitoring these key soil ecosystems and their functional attributes across diverse landscapes”. Before the method can be applied on a broad scale, however, it will need to be tested in other environments and seasons. Further trials will help refine the detection system so that it can distinguish biocrusts from surrounding soil, rocks and vegetation.