12 June 2024

In the USA, a predictive model shortens soil test timing


Researchers at the University of Arkansas have developed a predictive model of soil structure and organic matter content that halves the overall testing time

by Matteo Cavallito

Soil tests are notoriously important in directing strategic choices in cultivation. Carrying them out, however, can be time-consuming and complicated and therefore incompatible with farmers’ needs. So how can we solve this problem? One possible answer came recently from the Agricultural Experiment Station at the University of Arkansas in the United States.

Here, a team of researchers led by Gerson Drescher, professor of soil fertility at the same institute, has developed a new forecasting model capable, the authors note, of adding crucial information about the soil.

Soil analyses

“Standard soil testing evaluates plant-available nutrient content and soil pH,” explains an article published by the University of Arkansas. “However, these properties are also affected by soil texture and organic matter in the soil, which require additional expensive and time-consuming tests.” Drescher’s idea is therefore to evaluate these indicators by replacing traditional tests with a new prediction model developed from the results of standard tests already used in sample analysis.

The prediction model, in this way, would be able to halve the time required for the overall  analysis.

The test for organic matter is determined by a method known as “loss on ignition”, which requires weighing the samples, subjecting them to high temperatures to burn off the organic matter, and weighing them again to calculate the change.  Particle size, on the other hand, is measured with a device called a hydrometer, which determines the size of soil particles.

The study

To adjust and validate the prediction models, Drescher used nutrient and pH data from samples sent to the University of Arkansas laboratory. “Two datasets were used to calibrate clay and sand and soil organic matter prediction models using simple and multiple regression,” the study states.

In particular, “Clay and sand prediction models presented comparable accuracy when validated on a new dataset.”

In general, the sand and clay analyses proved accurate for medium- and fine-grained soils. But they showed limited capability for those characterised by a coarse structure. Over time, the researchers explained, the models can be improved with new samples for later use elsewhere in areas with similar soils.

The crucial importance of testing

Focus on testing increased significantly at the beginning of the decade when the price of fertilisers experienced a sharp rise. The need to limit their use implicitly highlighted the importance of gathering key information through testing. So as to plan a more rational and widespread use of inputs.

In land surveys, pH data, in particular, is a crucial indicator because the level of acidity is a critical factor in nutrient availability. A correct assessment of acidity therefore makes fertility promotion strategies more efficient.