A new computer model called the Biogeochemistry-Informed Neural Network (BINN) measures soil organic carbon processes 50 times more efficiently than previous models, according to a study published in the journal Geoscientific Model Development. The Cornell University-led research is one of the first AI tools to advance scientific discovery in agriculture and biogeochemistry.
Most common AI tools, such as ChatGPT, repurpose existing information or extract patterns from data. BINN goes further: it predicts biological processes that are not yet well understood and suggests the factors that control them.
How BINN works
Soil scientists know the mechanisms by which soils acquire organic carbon. Plants extract and sequester carbon from carbon dioxide to grow, and when those plants die, organic matter from stems, leaves and roots decomposes into smaller bits. What remains unclear is the speed of these processes and how many are required to break down the litter.
"BINN is very easy to use and can be democratized among the scientific community in various disciplines," said Yiqi Luo, the Liberty Hyde Bailey Professor in the School of Integrative Plant Science at Cornell's College of Agriculture and Life Sciences and a senior author of the study. "This is one of the first tools of this type that can promote scientific research with AI."
Haodi Xu, a doctoral student in Luo's lab, is co-first author. The research also involved the lab of Carla Gomes, professor of computer science at Cornell Bowers Computing and Information Science.
Why soil carbon matters
The Earth's soils hold roughly three-quarters of the world's terrestrial carbon - more carbon than the atmosphere and all the world's plants combined. Small adjustments to soil carbon processes can have oversized downstream effects.
In 2015, French soil scientists proposed the "4 per 1,000" initiative, arguing in theory that if humans increased global soil organic carbon in agricultural lands by 0.4% annually, it would improve soil health and offset all human-caused carbon emissions.
Performance compared to previous models
BINN computed 50 times faster than earlier models, and its predictions of soil organic carbon quantities were similar in accuracy. Previous models contained spatial biases - when making predictions across the contiguous U.S., they might favor data from one region over another. The researchers found less spatial bias with BINN.
"We use AI and data to tell us quantitatively how fast and how many of these kinds of processes are required," Xu said.
BINN can be adapted to other little-known agricultural and biogeochemical processes, such as soil respiration or carbon accumulation in forest systems. The work shows how AI can support scientific discovery in fields like AI for Science & Research, moving beyond pattern recognition to hypothesis generation.
Why this matters for science and research professionals
BINN demonstrates that AI can identify unknown variables in biological systems and quantify them, not just classify data or generate text. For researchers in biogeochemistry, ecology, or related fields, the model offers a practical template for using AI to test assumptions about processes that are difficult to observe directly. Those interested in applying similar methods to their own work can start with an AI Learning Path for Research Scientists.
Your membership also unlocks:
