It is hard to know how much carbon is in the soil at large spatial scales. However, knowing about soil organic carbon at global, national or regional scales could help researchers estimate crop productivity, soil health and even global carbon cycles.
University of Illinois SMARTFARM team members collect soil samples. The SMARTFARM project, funded by the U.S. Department of Energy, focuses on pioneering technology to quantify field-scale carbon credits. Image Credit: University of Illinois.
Classically, scientists gather soil samples in the field and transport them back to the lab, where they examine the material to establish its constituents. But that is time-intensive, laborious, expensive and only offers insights on particular locations.
In a new study, University of Illinois scientists demonstrate new machine-learning techniques based on laboratory soil hyperspectral data that could deliver equally accurate approximations of soil organic carbon. Their study offers the groundwork to use airborne and satellite hyperspectral sensing to track surface soil organic carbon spanning large areas.
Soil organic carbon is a very important component for soil health, as well as for cropland productivity. We did a comprehensive evaluation of machine learning algorithms with a very intensive national soil laboratory spectral database to quantify soil organic carbon.
Sheng Wang, Study Lead Author and Research Assistant Professor, Agroecosystem Sustainability Center (ASC), Department of Natural Resources and Environmental Sciences (NRES), University of Illinois
Wang and his collaborators made use of a public soil spectral library from the USDA Natural Resources Conservation Service comprising over 37,500 field-collected records and characterizing all soil types around the U.S.
Like every material, soil reflects light in distinctive spectral bands which researchers can deduce to establish the chemical composition.
Spectra are data-rich fingerprints of soil properties; we're talking thousands of points for each sample. You can get carbon content by scanning an unknown sample and applying a statistical method that’s been used for decades, but here, we tried to screen across pretty much every potential modeling method. We knew some of these models worked, but the novelty is the scale and that we tried the full gamut of machine learning algorithms.
Andrew Margenot, Study Co-Author and Assistant Professor, Department of Crop Sciences, University of Illinois
Kaiyu Guan, principal investigator, ASC founding director, and associate professor at NRES, says, “This work established the foundation for using hyperspectral and multispectral remote sensing technology to measure soil carbon properties at the soil surface level. This could enable scaling to possibly everywhere.”
After choosing a suitable algorithm based on the soil library, the scientists tested it with simulated airborne and spaceborne hyperspectral data. As projected, their model attributed for the “noise” integral in surface spectral imagery, delivering a very precise and large-scale view of soil organic carbon.
NASA and other institutions have new or forthcoming hyperspectral satellite missions, and it's very exciting to know we will be ready to leverage new AI technology to predict important soil properties with spectral data coming back from these missions.
Sheng Wang, Study Lead Author and Research Assistant Professor, Agroecosystem Sustainability Center (ASC), Department of Natural Resources and Environmental Sciences (NRES), University of Illinois
Chenhui Zhang, an undergraduate student studying computer science at Illinois, also worked on the project as part of an internship with the National Center for Supercomputing Applications’ Students Pushing Innovation (SPIN) program.
Hyperspectral data can provide rich information on soil properties. Recent advances in machine learning saved us from the nuisance of constructing hand-crafted features while providing high predictive performance for soil carbon. As a leading university in computer sciences and agriculture, U of I gives a great opportunity to explore interdisciplinary sciences on AI and agriculture. I feel really excited about that.
Chenhui Zhang, Undergraduate Student of Computer Science, University of Illinois
The study was detailed in an article titled, “Using soil library hyperspectral reflectance and machine learning to predict soil organic carbon: Assessing potential of airborne and spaceborne optical soil sensing,” which was published in Remote Sensing of Environment. The authors include Sheng Wang, Kaiyu Guan, Chenhui Zhang, DoKyoung Lee, Andrew Margenot, Yufeng Ge, Jian Peng, Wang Zhou, Qu Zhou, and Yizhi Huang.
The study received support from the U.S. Department of Energy's Advanced Research Projects Agency-Energy (ARPA-E) SMARTFARM and SYMFONI projects, Illinois Discovery Partners Institute (DPI), Institute for Sustainability, Energy, and Environment (iSEE), and College of Agricultural, Consumer, and Environmental Sciences Future Interdisciplinary Research Explorations (ACES FIRE), Center for Digital Agriculture (CDA-NCSA), the University of Illinois at Urbana-Champaign.
This study was also partly funded by the USDA National Institute of Food and Agriculture (NIFA) Artificial Intelligence for Future Agricultural Resilience, Management, and Sustainability grant.
The Departments of Natural Resources and Environmental Sciences and Crop Sciences are in the College of Agricultural, Consumer and Environmental Sciences (ACES) at the University of Illinois Urbana-Champaign.
The Agroecosystem Sustainability Center is cooperatively established by the Institute for Sustainability, Energy, and Environment (iSEE), the College of ACES, and the Office of the Vice-Chancellor for Research and Innovation at the University of Illinois Urbana-Champaign.
Journal Reference:
Wang, S., et al. (2022) Using soil library hyperspectral reflectance and machine learning to predict soil organic carbon: Assessing potential of airborne and spaceborne optical soil sensing,. Remote Sensing of Environment. doi.org/10.1016/j.rse.2022.112914.