A new computer model developed at Cornell University can measure soil organic carbon processes 50 times faster than previous models while reducing spatial bias, offering a proof-of-principle for how artificial intelligence can illuminate obscure biological mechanisms. The model, called the Biogeochemistry-Informed Neural Network (BINN), addresses a critical knowledge gap: Earth's soils hold roughly three-quarters of the world's terrestrial carbon, more than the atmosphere and all plants combined, yet the speed and number of processes that break down organic matter remain poorly understood.
An AI tool designed for scientific discovery
Most common AI tools, such as ChatGPT, mainly repurpose existing information or extract patterns from data. BINN goes a step further by predicting biological processes that are not yet well understood and suggesting factors that control them. The model is one of the first of its kind to advance scientific discovery in agriculture and biogeochemistry.
"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 in the 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."
Closing the soil carbon knowledge gap
Soil scientists know that plants extract carbon from carbon dioxide to grow, and when they die, organic matter decomposes into the soil. But the speed of these processes and how many steps are required to break down plant litter remain unclear. Even small adjustments to soil carbon processes can have outsized effects-in 2015, French scientists proposed the "4 per 1,000" initiative, which suggested that increasing global soil organic carbon in agricultural lands by 0.4% annually could offset all human-caused carbon emissions.
"We use AI and data to tell us quantitatively how fast and how many of these kinds of processes are required," said Haodi Xu, a doctoral student in Luo's lab and co-first author of the study. The research was a collaboration with the lab of Carla Gomes, professor of computer science in Cornell Bowers Computing and Information Science.
Faster, less biased predictions
Compared with previous models, BINN computed 50 times faster while maintaining very similar accuracy in predicting soil organic carbon quantities. The earlier models contained spatial biases, meaning they might favor data from one region over another when making predictions across the contiguous U.S. The researchers found less spatial bias with BINN. The study was published in Geoscientific Model Development.
Adaptable to other biogeochemical processes
The new model can be adapted to reveal other little-known agricultural and biogeochemical processes, such as those relating to soil respiration or carbon accumulation in forest systems. Its design allows researchers to apply it to domains where the underlying mechanisms are not fully understood, generating quantitative insights without relying on pre-existing assumptions embedded in traditional models.
Why this matters for Science and Research professionals
BINN provides a faster, less biased way to model complex biogeochemical processes, enabling researchers to generate hypotheses and gain quantitative understanding of mechanisms that have resisted traditional approaches. As tools like this emerge, the growing field of AI for Science & Research offers new avenues for discovery. Scientists can benefit from structured training programs to apply AI in their research, building the skills to integrate such models into their own work. The model's speed and reduced spatial bias also make it practical for large-scale environmental studies where computational efficiency and geographic fairness are critical.
Your membership also unlocks:
