AI Makes Biochar Phosphorus Predictions More Precise

Biochar – charred plant matter incorporated into agricultural soil – is increasingly being used as an affordable means of improving soil quality and reducing fertiliser losses.

Farmers applying it usually want to make better use of the phosphorus already present in their fields.

Yet biochar’s influence on phosphorus has long been difficult to predict. It may release nutrients that are otherwise unavailable, or subtly undermine the very farmer who used it. Fresh research could at last make its effects more foreseeable.

The hidden waste

Plants are remarkably inefficient at absorbing the phosphorus supplied to them. Just a limited proportion nourishes crops within the same growing season, with estimates commonly ranging from 15 to 20 percent.

What remains does not simply disappear. Part of it attaches to iron, aluminium or calcium, becoming chemically fixed and inaccessible to the crop. The rest can wash into rivers and ponds, where it fuels algal blooms and deprives water of oxygen.

This runoff, known as eutrophication, can leave once-healthy waterways green and devoid of life. Farmers therefore face a largely unseen double loss: money spent on wasted fertiliser and environmental harm downstream that they did not intend.

Phosphorus is sourced from mined rock that required millions of years to form. Its non-renewable nature makes that waste even more consequential.

Charcoal for soil

Biochar offers one potential solution. This charcoal-like substance is produced by heating crop residues, wood or other plant material with very little oxygen. The process leaves a porous, carbon-rich black solid that can be incorporated into fields.

Biochar has an inconsistent reputation. In certain soils, it makes fixed phosphorus available to plants; in others, it keeps phosphorus beyond the reach of roots. That can sometimes reduce pollution, but it can also simply be frustrating.

This uncertainty is the central problem. Farmers considering a bag of biochar have had no dependable way to tell whether it will benefit a particular field or hinder it, what rate to apply, or which soil conditions will determine the result.

AI predictions for biochar

Yutao Peng of Sun Yat-Sen University in Shenzhen, China, and colleagues aimed to replace this uncertainty with reliable predictions.

Using 32 previous studies, the researchers compiled 534 measurements documenting how soil phosphorus changed after biochar was applied.

They fed these data into three machine-learning systems designed to identify links between biochar characteristics, soil conditions and phosphorus outcomes. The models had 19 variables to manage, far beyond what could readily be tracked manually.

One approach performed noticeably better than the others. The Random Forest model generates an answer by conducting hundreds of separate analyses of the data and averaging their findings.

When tested on data it had not previously encountered, the model achieved an R² of about 0.91, accounting for most of the variation in the way biochar affected phosphorus.

Heat shapes biochar

When the researchers examined which factor mattered most, one stood out: the temperature used to produce the biochar, known as its pyrolysis temperature, had the greatest influence on the outcome.

Biochar produced at moderate temperatures appeared structurally well balanced, probably creating sufficient fine pores and active surfaces to regulate phosphorus without excessive effects.

The model identified an optimum range. The best outcomes were concentrated at production temperatures of 460–482ºC, alongside modest application rates.

Biochar produced at hotter temperatures acted differently, reducing phosphorus availability instead of raising it. This may be useful where the concern is phosphorus reaching nearby water. In that sense, heat acts as a control dial.

Soil conditions matter

Temperature was not the only influence. The quantity of biochar applied was the second most important factor, followed by soil acidity and the total phosphorus already present in the soil.

Soil pH was found to guide the entire interaction. In acidic soils, phosphorus commonly binds to iron and aluminium, making it hard to access, while biochar has less opportunity to act.

Neutral and mildly alkaline soils offer more scope for biochar. As these metal ions become less active, the upward shift in pH caused by biochar can convert phosphorus into forms that roots can reach.

These influences did not operate in a linear way. The model revealed complex interactions between the factors that a straightforward equation would fail to capture.

This helps clarify why a system built to manage complicated, non-linear relationships surpassed conventional statistical methods.

Simpler biochar works

One result challenges a long-standing assumption in the field. Chemically modified biochar, engineered in laboratories to improve performance, has often been viewed as the superior option and assumed gold standard.

The research indicates that this may be unnecessary. Under suitable conditions, ordinary untreated biochar can equal or even outperform modified biochar in regulating phosphorus.

The model also indicated an unexpected possibility: biochar may be beneficial less because of any phosphorus it contains and more because of how it affects phosphorus already in the soil.

That alters the calculation. Avoiding chemical modification reduces both costs and environmental impacts. Biochars made from crops or wood require less energy to produce than manure-based alternatives while delivering comparable results.

New tool for farming

Before this research, pairing a particular biochar with a particular field relied largely on trial and error, leaving farmers to bear the expense of each incorrect decision. There was no dependable means of forecasting results before application.

There is now one. Before a single handful reaches the soil, a data-led model can predict whether a given biochar will make phosphorus available for nutrient-hungry crops or retain it to safeguard a watershed.

Peng described the development as shifting biochar use from guesswork towards data-guided decisions.

For farmers and their advisers, this could result in less fertiliser waste, fewer nutrients entering local waters, and a clearer basis for deciding what should be added to the soil.

The model is an initial framework, not a completed product, and its predictions will be tested by real-world fields.

Nevertheless, it directs agriculture towards a future in which an inexpensive, carbon-rich material made from plant waste is used with something nearer to precision than optimism.