Scientists use advanced ICP spectral analysis to forecast soil nutrition with high precision, paving the way for smarter fertilizer use and greener farming.
Tokyo, Japan, 27 November 2025 – The future of soil testing may be arriving sooner than expected. A new international study led by Nakamura and a team of global researchers has revealed a promising method that could transform how farmers understand their soil, manage fertilizer use, and protect the environment. By using deep-learning technology combined with Inductively Coupled Plasma (ICP) spectral data, the researchers achieved highly accurate predictions of key soil properties, offering a faster, more affordable alternative to traditional lab-based soil diagnosis methods.
Conventional soil testing is often slow, technical, and costly, making it inaccessible for many farmers, especially in developing regions. As a result, misapplication of fertilizers leads to nutrient imbalance, financial loss, and environmental harm. The new approach could change that.
What the Researchers Did
The team analyzed 1,941 soil samples from seven countries, representing a wide variety of climates, soil types, cropping systems, and agricultural practices. Using the full ICP wavelength spectral data of ammonium acetate (NH₄OAc) soil extracts, they trained a deep-learning model to predict an extensive list of soil properties, including:
- Exchangeable bases (Ca, Mg, K, Na)
- Soil pH (water and KCl)
- Available phosphorus (Bray1-P)
- Electrical conductivity
- Exchangeable aluminum
- Total nitrogen and carbon
- Soil texture (clay and sand content)
- Cation exchange capacity
What They Found
The results were impressive. The model’s predictions matched laboratory measurements with R² values over 0.9 across most parameters, indicating extremely strong accuracy. Even the lowest prediction values remained reliably high.
This marks the first known study to successfully use the entire ICP spectral range to predict multiple soil properties at something previously thought too complex due to the high-dimensional data.
Why This Matters for Agriculture
If further developed, this approach could revolutionize soil testing by making it:
- Faster than traditional lab analysis
- More accurate in detecting nutrient needs
- More affordable, especially in regions with limited lab access
- More sustainable, by reducing fertilizer overuse and nutrient runoff
Better soil diagnosis means smarter fertilizer application, cutting waste, protecting waterways, and helping farmers improve yields.
This is especially important for developing countries, where testing costs often prevent widespread soil monitoring. With more training data and improved AI models, this method could empower millions of farmers with reliable nutrient insights for the first time.
Despite the strong results, researchers caution that more work is needed.
The model used a limited sample size, and expanding the dataset could sharply improve accuracy.
Future studies may incorporate CCD-based ICP units and convolutional neural networks (CNNs) for deeper spectral pattern recognition.
The team also plans to evaluate other soil extractants to determine whether different chemical solutions could improve prediction performance.
Broader Applications Beyond Farming
The researchers see potential far beyond soil nutrients. With further refinement, ICP-based AI predictions might help estimate:
- Soil biodiversity
- Overall soil health
- Pollution levels (heavy metals, microplastics, PFAS)
- Chemical concentrations for laboratory workflows
This opens opportunities for environmental monitoring, public health, and industrial applications.
Barriers That Must Be Overcome
For widespread adoption, major limitations must be addressed:
- ICP instruments are expensive.
- They require high-purity argon gas and high-frequency power.
- Many rural regions lack the infrastructure to operate such systems.
Reducing equipment costs and improving accessibility will be key to bringing this technology to farmers worldwide.
This study represents a major step toward faster, cheaper, and more environmentally responsible soil testing. With additional research and expanded datasets, deep-learning models powered by ICP spectral data could help transform global agriculture, making it smarter, more efficient, and more sustainable.
