Cornell AI Model BINN Boosts Soil Carbon Research

Cornell University researchers, led by Yiqi Luo, developed BINN (Biogeochemistry-Informed Neural Network), an AI model that advances scientific discovery in agriculture and biogeochemistry. Published in Geoscientific Model Development on July 24, 2026, BINN is 50 times more efficient than previous models and reduces spatial biases when predicting soil organic carbon levels across the contiguous United States. The model predicts poorly understood biological processes and suggests controlling factors, offering a proof-of-principle for AI-driven scientific research. The study involved collaborations with Carla Gomes' lab and contributions from researchers including Haodi Xu, Joshua Fan, Feng Tao, and Benjamin Zhai. Funding came from multiple U.S. agencies and foundations. BINN can be adapted to study other processes like soil respiration and forest carbon accumulation.