Machine learning helps farmers use biochar more precisely to manage soil phosphorus

Phosphorus is essential for crop growth, but using it efficiently remains one of agriculture’s long-standing challenges. Farmers often apply large amounts of phosphorus fertilizer to maintain yields, yet only a small fraction is taken up by crops. The rest can become locked in soil or escape into waterways, contributing to nutrient pollution and eutrophication.

A new study published in Biochar shows that machine learning could help make phosphorus management more precise by predicting how pristine biochar affects soil phosphorus availability under different soil and production conditions.

Biochar, a carbon-rich material produced by heating biomass in limited oxygen, has attracted growing attention as a soil amendment. It can improve soil properties, influence nutrient cycling, and potentially reduce fertilizer losses. However, its effects on phosphorus are not always predictable. In some soils, biochar can increase plant-available phosphorus. In others, it may immobilize phosphorus or reduce leaching risks. This uncertainty makes it difficult for farmers and land managers to know which biochar to use, how much to apply, and under what soil conditions.

To address this challenge, the research team compiled a dataset of 534 samples from 32 published studies. The dataset included biochar characteristics, soil properties, and experimental conditions. The team then trained and compared three machine learning models: Random Forest, Support Vector Regression, and Artificial Neural Networks.

Among the three models, Random Forest delivered the strongest predictive performance, with a test-set R² value of 0.9107. This means the model could explain most of the variation in how biochar changed soil phosphorus availability. The model also outperformed the other approaches in prediction error, making it the most reliable tool tested in the study.

“Our goal was to move biochar application from trial-and-error toward data-guided decision-making,” said corresponding author Yutao Peng. “By using machine learning, we can better understand when biochar is likely to activate phosphorus for crop use and when it may help passivate phosphorus to reduce environmental risks.”

Feature importance and SHAP analyses showed that biochar pyrolysis temperature was the most influential factor. Moderate pyrolysis temperatures were linked to biochars with balanced porosity and surface reactivity, which can support phosphorus regulation. In contrast, biochars produced at higher temperatures tended to favor phosphorus passivation, which may help reduce the risk of excess phosphorus movement into water bodies.

Other key factors included biochar application rate, soil pH, and total soil phosphorus content. The study found that the interaction among these variables was highly nonlinear. For example, optimal phosphorus availability regulation was observed around moderate pyrolysis temperatures and moderate application rates, while soil pH influenced whether biochar enhanced or limited phosphorus availability.

Importantly, the study suggests that pristine biochar, without additional chemical modification, may achieve or even exceed the phosphorus-regulating performance of modified biochar in some contexts. This finding could lower costs and environmental burdens associated with biochar production and use.

The researchers emphasize that the model is not just a prediction tool, but also a framework for precision soil management. By matching biochar properties with soil conditions, farmers and advisors could improve fertilizer efficiency, reduce nutrient losses, and support more sustainable agricultural systems.

“Phosphorus management must balance crop productivity with environmental protection,” said corresponding author Jia Liu. “Our findings show that machine learning can help identify that balance and guide more responsible biochar use.”

The study highlights a broader shift in sustainable agriculture: combining soil science, environmental chemistry, and artificial intelligence to make nutrient management more targeted, economical, and environmentally sound.

=== 

Journal Reference: Wang, Y., Yin, J., Yang, X. et al. Achieving precise regulation of soil phosphorus availability by guiding the application of pristine biochars with machine learning techniques. Biochar 8, 101 (2026).   

https://doi.org/10.1007/s42773-026-00611-1   

=== 

About Biochar

Biochar (e-ISSN: 2524-7867) is the first journal dedicated exclusively to biochar research, spanning agronomy, environmental science, and materials science. It publishes original studies on biochar production, processing, and applications—such as bioenergy, environmental remediation, soil enhancement, climate mitigation, water treatment, and sustainability analysis. The journal serves as an innovative and professional platform for global researchers to share advances in this rapidly expanding field. 

Follow us on Facebook, X, and Bluesky.