Study shows AI, supercomputing can predict GHG emissions from farms

Accurate cost-effective results are more than 10,000 times faster, even with limited data.

September 12, 2024

4 Min Read

For the first time, a team of scientists has demonstrated it is possible to provide accurate, high-resolution predictions of carbon cycles using Artificial Intelligence (AI) and supercomputers to measure the amount of Greenhouse Gas (GHG) emissions from every individual farm at the national scale. This breakthrough is a critical first step in developing a credible measurement, monitoring, reporting, and verification of agricultural emissions. It can be used to incentivize the implementation of climate smart practices and help mitigate the impacts of climate change, which supports the White House’s national strategy highlighting the need to quantify GHG emission across sectors with a goal of net-zero emissions by 2050. The team’s research findings were recently published in Nature Communications, based on a visionary framework the team published in Earth-Science Reviews.

Funded by the Foundation for Food & Agriculture Research, FoodShot Global, U.S. Department of Energy and U.S. National Science Foundation, the study was co-led by University of Illinois Urbana-Champaign’s Agroecosystem Sustainability Center Founding Director and Blue Waters Professor Kaiyu Guan and University of Minnesota’s Professor Zhenong Jin. The team with collaborators developed the AI and supercomputer solution to quantify the related changes in GHG emissions from adopting climate mitigation practices like cover cropping and precision nitrogen fertilizer management.

“This solution is a scalable, reliable way to measure and predict on individual farm fields the agricultural carbon fluxes, crop yields and changes in soil carbon stocks and can help the industry speak uniformly about best practices for reducing farmland emissions,” said Guan. “This breakthrough could allow the agricultural and food sectors to quantify their carbon footprints in producing or sourcing raw agricultural products so that they can design strategies to reduce GHG in their supply chain and objectively assess different GHG-reducing strategies.”

The team built their predictive modeling tool using Knowledge-Guided Machine Learning (KGML), an emerging machine learning research framework proposed by a group of computer scientists. Pioneered by this team, the KGML model for Agriculture (KGML-Ag) uses the power of satellite remote sensing, computational models and state-of-the-art AI techniques to cost-effectively produce accurate results more than 10,000 times faster than traditional process-based models, even with limited data.

“Unlike traditional model-data fusion approaches, we developed KGML-Ag as a new way to bring together the power of sensing data, domain knowledge and artificial intelligence techniques,” said Jin. “AI plays a critical role in realizing our ambitious goals to quantify every field’s carbon emission.”

“Building the KGML is very challenging due to the need of data and knowledge from various domain,” said Licheng Liu, the lead author of the KGML-Ag work and a research scientist at University of Minnesota. “Fortunately, our team brings together the experts in field measurements, domain sciences, and AI techniques, allowing us to achieve this significant breakthrough.”

To compute the vast amount of information from millions of individual farms, the team is using supercomputing platforms available at the National Center for Supercomputing Applications.

Although locally tested in the Midwest, the new approach can be scaled up to national and global levels and help the industry grasp the best practices for reducing emissions. “The strength of our tool is that it is both generic and scalable, and it can be potentially applied to different agricultural systems in any country,” said Bin Peng, co-author of the study and an assistant professor at the University of Illinois Crop Sciences Department.

Peng continued: “There are many effective farming practices that reduce GHG emissions, but if everyone measures them differently, we’ll never be able to objectively understand how well these practices work. This research helps agriculture stakeholders ‘speak the same language’ about farmland greenhouse gas emissions and will foster more scientific rigor in estimating those emissions.”

The study also details how emissions and agricultural practices data can be cross-checked against economic, policy and carbon market data to find best-practice and realistic GHG mitigation solutions locally to globally – especially in economies struggling to farm in an environmentally conscious manner.

“The real beauty of our work is that it is both very generic and scalable, meaning it can be applied to virtually any agricultural system in any country to obtain reliable emissions data using our targeted procedure and techniques, which is what we are expanding to do right now” Guan concluded.

Subscribe to Our Newsletters
Feedstuffs is the news source for animal agriculture

You May Also Like