MSU research introduces novel method to improve prediction of cropland nitrous oxide emissions
A team of Michigan State University (MSU) researchers has developed a groundbreaking machine learning system capable of predicting nitrous oxide emissions from U.S. croplands with unprecedented accuracy, a finding with valuable implications for national greenhouse gas (GHG) accounting and mitigation.
The study was published in the journal Proceedings of the U.S. National Academy of Sciences.
Capturing nitrous oxide emissions
Nitrous oxide is a GHG emitted in agricultural operations primarily through the use of nitrogen fertilizers.
Accurately predicting emissions has eluded scientists due to the complex interplay of weather, soil conditions and crop management practices which influence the microbes responsible for producing the gas.
MSU’s new research changes this.
Led by former MSU Graduate Student Prateek Sharma and Bruno Basso, a John A. Hannah distinguished professor in MSU’s Department of Earth and Environmental Sciences and the W.K. Kellogg Biological Station (KBS), the team developed a hybrid modeling system which combines machine learning and ecosystem models to capture daily nitrous oxide emissions.
G. Philip Robertson, university distinguished professor at KBS in the Department of Plant, Soil and Microbial Sciences, co-led the research.
Professor Michael Murillo in the Department of Computational Mathematics, Science and Engineering also contributed.
Additionally, the research team included Aditya Manuraj, Neville Millar, Tommaso Tadiello, Mukta Sharma and Mathieu Delandmeter of the Department of Earth and Environmental Sciences.
Achieving prediction accuracy
The modeling system leveraged more than 12,000 nitrous oxide measurements collected across 17 sites in the Midwest and Great Plains, spanning six cropping systems and 35 management practices – one of the most comprehensive datasets of its kind.
Whereas conventional single-model approaches for estimating nitrous oxide emissions struggle to achieve 20 percent prediction accuracy, Basso said accuracy for the new ensemble system exceeded 80 percent.
“One of the limiting factors of current predictive models is they rely on outdated national GHG emission inventories and often need to be calibrated to a specific site,” said Basso. “With this effort, we’ve moved past these limitations to provide management-specific predictions for crucial combinations of cropping systems, soils, management practices and weather conditions.”
“We’re hopeful this approach can lead to field-specific emission mitigation strategies, as well as much-needed updates to estimates of GHG emissions from agriculture,” Basso added.
The project was supported by the Great Lakes Bioenergy Research Center, the U.S. Department of Energy Office of Science, the National Science Foundation Long-Term Ecological Research Program at KBS, the U.S. Department of Agriculture’s (USDA) National Institute of Food and Agriculture, the USDA Long-Term Agroecosystem Research Program at KBS, the CERCA-Foundation for Food and Agriculture Research Project, Climate Trace, the Soil Inventory Project and MSU AgBioResearch.
Originally formed in 1888 as the Michigan Agricultural Experiment Station, MSU AgBioResearch oversees numerous on-campus research facilities, as well as 15 outlying centers throughout Michigan, where scientists are conducting leading-edge research on a variety of topics from health and agriculture to food systems and natural resources. To learn more, visit agbioresearch.msu.edu. This article was originally published by MSU AgBioResearch Communications on March 5.
