Efficient supplemental lighting control is crucial for optimizing crop productivity and energy use in controlled environment agriculture (CEA). While environmental factors such as temperature and carbon dioxide (CO2) concentration significantly influence photosynthesis, current lighting control strategies rely solely on ambient sunlight levels. To address this limitation, a chlorophyll fluorescence (CF)-based biofeedback system has been proposed to dynamically adjust LED light intensities based on real-time photosynthetic responses. However, frequent CF measurements using pulse-amplitude modulated (PAM) fluorometers can induce severe photoinhibition due to repetitive saturating light pulses, limiting long-term application. This study explores an alternative approach by developing a machine learning model to estimate the quantum yield of photosystem II (ΦPSII) from environmental parameters, eliminating the need for the fluorometer and continuous physical measurements. Four-week-old green and red lettuce (Lactuca sativa) cultivars (‘Casey’ and ‘Cherokee’) were grown in a greenhouse for a month, where ΦPSII was measured every 15 minutes using a fluorometer (Monitoring-PAM; Heinz Walz, Effeltrich, Germany) alongside environmental data, including extended photosynthetically active radiation, temperature, CO₂ concentration, and vapor pressure deficit. A linear regression model was developed to estimate ΦPSII, generating cultivar-specific equations that were integrated into the biofeedback system for LED control. The estimated ΦPSII values exhibited a strong correlation with the measured data, allowing the biofeedback system to optimize lighting without the risk of photoinhibition associated with frequent PAM fluorometer measurements. This approach enabled dynamic light adjustment based on environmental conditions and lettuce cultivar, with the regulated light levels closely aligning with direct measurements. These findings highlight the potential of integrating predictive models into the biofeedback-controlled lighting systems, offering a cost-effective and non-invasive alternative to direct CF measurements for precision lighting management in CEA.
