Hyperspectral Detection of Spectral Responses to Acute High-Irradiance Blue Light in Five Microgreen Species

High-intensity blue light is a powerful regulatory signal for plants, but its excess induces oxidative stress requiring rapid diagnosis. This study evaluated the potential for early detection of changes in the spectral characteristics of microgreen canopy cover caused by high-intensity blue light (PPFD 2500 µmol·m−2·s−1, 12 h) in five species of microgreens (Helianthus annuus, Pisum sativum, Eruca sativa, Hordeum vulgare, Raphanus sativus ‘Sango Purple’) using hyperspectral imaging (450–950 nm) and machine learning. A Random Forest model trained on 85 vegetation indices classified light stress with high accuracy (Accuracy > 93%, Kappa > 0.87, F1-score > 93%) and detected spectral changes characteristic of light stress as early as 1–3 h of exposure. SHAP analysis identified carotenoid-sensitive indices (PRI, CCI, PRICI2) and chloroplast movement and stress indices (CMI, LSIRed, LSINorm, Carter5) as the most informative predictors. It is assumed that the primary mechanism underlying early spectral changes was chloroplast avoidance rather than pigment degradation, as confirmed by the reversibility of canopy bleaching, rapid recovery of maximum quantum yield of photosystem II, and unchanged chlorophyll and carotenoid contents. Sunflower and radish were the most sensitive species to high-dose blue light, while barley was the least sensitive. LSIRed was identified as a reliable qualitative marker of light stress that does not require a control sample, simplifying its use in automated monitoring systems. These findings demonstrate the effectiveness of hyperspectral phenotyping for non-invasive, rapid diagnosis of light stress in microgreens, providing a tool for optimising lighting regimes in controlled environment agriculture.