The nitrogen content of a plant can be determined by laboratory analysis or estimated from a spectral measurement and a properly calibrated model. ScanSpectrum does this using light reflected in the visible and near-infrared range, known as VIS–NIR. The model links plant spectra with laboratory nitrogen determinations made using the Kjeldahl method. We built it on four years of field trials, more than 100,000 scans and around 5,000 laboratory samples.
How does ScanSpectrum estimate nitrogen content in plants?
ScanSpectrum estimates nitrogen content from the spectrum of light reflected by the plant in the visible and near-infrared (VIS–NIR) range. A calibration model converts the spectral data into a predicted nitrogen content, using relationships established from laboratory analyses.
The shape of the spectrum is influenced by pigments, tissue structure and water content, among other things. The measurement is therefore not a direct readout of the amount of nitrogen – it relies on the relationships between a plant’s optical properties and its composition.

To teach the model to recognize these relationships, we needed a dataset that was as large and varied as possible. We started with soil analysis, which we used to plan nitrogen fertilization levels. The aim was to grow plants covering a wide range of nutrition: from clearly deficient, through adequately supplied, to plants that received more nitrogen than they needed.
Why does the model need laboratory nitrogen measurements?
Laboratory results are the reference point during calibration: they let us assign the measured nitrogen content of the material to each spectrum. From these pairings, the model learns the relationships it needs to estimate nitrogen content in new samples.
To develop the ScanSpectrum model, we used determinations made with the Kjeldahl method. The plants sent to the lab were the same ones we had scanned in the field beforehand, so the spectral data and the chemical analysis results described the same material.
From thousands of such pairings, the algorithm learned which features of the spectrum are linked to changes in nitrogen content.
How did we collect the data to calibrate the nitrogen model?
We ran field trials for four years on rapeseed, wheat, barley and corn, covering different varieties, growth stages and field locations. In total, we took more than 100,000 scans and matched them with the results of around 5,000 laboratory samples.
Each trial began with soil analysis, which we used to plan different levels of nitrogen fertilization. This let us collect data from plants with different nitrogen supply and cover a wide range of nutritional conditions in the calibration.

For the first two years we worked with the Warsaw University of Life Sciences (SGGW), and for the next two we carried out the research as part of AgriFood TEF.
How is the nitrogen model calibrated?
Calibration means establishing the relationship between plant spectra and laboratory-measured nitrogen content. Developing the ScanSpectrum model involved four main stages:
- Designing trials with different levels of nitrogen fertilization.
- Recording plant spectra on different dates and at different growth stages.
- Sending the scanned plants to the lab to measure their nitrogen content with the Kjeldahl method.
- Training the model on the paired spectral data and laboratory results.
The model uses information from many wavelengths, analyzing the relationships between the shape of the spectrum and nitrogen content. Once calibrated, it can estimate nitrogen content from new scans, within the range of applications for which its performance has been verified.

What does measuring nitrogen directly in the field give you?
A ScanSpectrum measurement gives you an estimate of nitrogen content without sending every sample to a lab. The scan is taken right in the field, and the result is calculated by a model developed from spectral data and chemical analyses.
This result can support the assessment of plant nutrition and fertilization decisions. It should be interpreted together with the growth stage, growing conditions and other agronomic data — the nitrogen content of a plant alone does not determine the fertilizer dose it needs.