Hyperspectral imaging, combined with advanced data analytics, could enable earlier detection of potato cyst nematode infestations before visible crop damage appears, according to new research published in Plant Phenomics.

The study, released on 14 October 2025 and distributed via Newswise by the Chinese Academy of Sciences, addresses one of the most persistent challenges in potato production: identifying nematode pressure at an early stage, when conventional scouting and soil diagnostics often fail to detect the problem.

Potato cyst nematodes are classified as quarantine pests in many regions due to their capacity to cause substantial yield losses. Early infections typically show few or no visible symptoms above ground, meaning infestations can go unnoticed for several growing seasons. Current monitoring approaches rely heavily on soil sampling and laboratory analysis, methods that are labour-intensive and impractical for large-scale or frequent screening.

The research team, led by Uroš Žibrat of the Agricultural Institute of Slovenia, investigated whether hyperspectral imaging could detect subtle physiological changes in potato plants caused by nematode infection, and whether these signals could be distinguished from those induced by drought stress. Hyperspectral imaging captures reflectance data across hundreds of narrow spectral bands, including visible, near-infrared and short-wave infrared wavelengths, making it sensitive to changes in plant water status, pigments and biochemical composition.

In a controlled greenhouse experiment, potato plants were inoculated with two species of potato cyst nematodes at different population levels and grown under both well-watered and water-deficient conditions. The researchers combined nematode reproduction data, plant morphology and physiological measurements with hyperspectral imagery collected at several growth stages. Explainable machine-learning techniques were then applied to assess how well different stress factors could be identified and to determine which spectral regions contributed most to classification performance.

Biological assessments confirmed successful nematode establishment under all watering regimes, while most visible plant traits remained largely unaffected. Only limited and temporary reductions in plant height were observed under specific combined stress conditions. This lack of obvious symptoms underscores the difficulty of early detection using conventional field observations.

Analysis of the hyperspectral data showed that plant growth stage exerted the strongest influence on spectral variation, overshadowing both nematode and drought stress signals. When machine-learning models were trained to classify stress factors, drought stress was identified with high accuracy, while nematode infection and species differentiation were detected with moderate reliability. Performance declined further when models attempted to separate multiple stress factors simultaneously.

Physiological measurements supported these findings. Water deficit consistently reduced photosynthesis, photosystem II efficiency and stomatal conductance, whereas nematode effects on these parameters were generally subtle and often not statistically significant. Explainable modelling highlighted the importance of water- and biochemistry-related spectral bands, particularly in the near- and short-wave infrared regions, helping to explain why drought signals dominated the classification results.

Despite these limitations, the authors conclude that hyperspectral imaging can reveal “hidden” nematode stress before visible symptoms develop. From an industry perspective, such capabilities could support earlier and more targeted interventions, helping growers and regulators identify infestation hotspots, improve pest management strategies and potentially reduce unnecessary chemical treatments.

While the technology is not yet a standalone solution for distinguishing overlapping stress factors under all conditions, the study demonstrates a clear step forward in non-invasive crop monitoring and highlights areas where further refinement and field-scale validation will be required.