Remote sensing and machine learning solution for early detection of onion pink root disease

Overview

Onion pink root disease poses a persistent and costly challenge to onion production in the Treasure Valley of Idaho and Eastern Oregon. Traditional visual diagnosis is unreliable because the foliar symptoms closely resemble those caused by drought or nutrient deficiencies, often requiring plants to be uprooted for confirmation. This solution leverages hyperspectral remote sensing combined with machine learning to detect disease-specific stress signatures in onion canopies, enabling earlier and more accurate identification of pink root infection. By distinguishing disease stress from other environmental stresses, growers can implement precision disease management strategies, potentially reducing unnecessary soil fumigation, minimizing yield losses, and increasing farm revenue.

Technical specifications

Key features:

• Hand-held hyperspectral sensor for leaf-level reflectance measurement to isolate stress-specific signatures
• Airborne hyperspectral imaging for field-scale monitoring and cross-validation against ground-level data
• Machine learning algorithms trained to distinguish pink root disease from confounding stresses such as drought and nutrient deficiency
• Soil pixel removal methodology to separate meaningful canopy variation from background interference
• Field validation at the Parma Research and Extension Center in Idaho under controlled stress regimes
• Dissemination pathway through Extension networks reaching ag communities across 42 counties

Research objectives:

• Quantify the effect of pink root stress on onion leaf reflectance using hand-held hyperspectral sensors
• Develop methods to remove soil-dominated pixels from hyperspectral imagery to assess meaningful canopy variation
• Cross-validate airborne and hand-held hyperspectral reflectance spectra for scalable field deployment
• Translate findings into actionable recommendations for onion growers and agricultural stakeholders

Technology readiness level

This technology is currently at an early-to-mid stage of development. Preliminary results have confirmed that remote sensing can capture reflectance variation in onion plants under different stress regimes, though initial findings indicate that plant stunting (a general indicator) rather than discreet foliar variation dominated the signal. The research team is refining their approach by moving to leaf-level measurements and removing soil interference to isolate disease-specific signatures. Future validation includes a controlled onion experiment at the Parma R&E Center and field-scale airborne hyperspectral comparison. The solution is not yet commercially deployed but is progressing toward practical field application through ongoing university-led research.