Nondestructive Seed Number Prediction of Citrus Cultivar ‘Koyama New Summer’ Using Near-infrared Spectroscopy

• Published in: Engeigaku kenkyuu Vol. 22; no. 1; pp. 89 - 97
• Main Authors: Hamabe, Naoya, Baba, Akiko, Sono, Yuga, Ikegaya, Atsushi, Ohba, Seiji, Taneishi, Motohiro, Baba, Fujio, Noda, Katsuji
• Format: Journal Article
• Language: Japanese; English
• Published: Kyoto-City THE JAPANESE SOCIETY FOR HORTICULTURAL SCIENCE 2023; Japan Science and Technology Agency
• Subjects: Accuracy; Calibration; calibration curve; Cultivars; Evaluation; Fruits; Infrared spectra; Infrared spectroscopy; Line spectra; Near infrared radiation; PLS regression analysis; Predictions; seed weight; seedless fruit; Seeds; Spectrum analysis; Summer; Weight
• ISSN: 1347-2658; 1880-3571
• DOI: 10.2503/hrj.22.89

We used near-infrared spectroscopy to predict the number of seeds of ‘Koyama New Summer’, a bud mutation of Izu ‘Hyuganatsu’ (Citrus tamurana hort. Ex. Tanaka), and validated its prediction accuracy. The calibration lines for soluble solid content, number of developed seeds, number of total seeds, developed seed weight, and total seed weight prediction were generated from the 730–970-nm spectra after the second derivative to evaluate the accuracy of the calibration lines. The accuracy of the calibration lines was assessed to be high for four measurement items, excluding the developed seed weight. Although the calibration sample had few fruits near the boundary value of five or six seeds, the percentages of correct and incorrect answers were calculated from the predicted and measured values, with five or fewer total seeds as fruits with few seeds and six or more seeds as fruits with many seeds, and the percentage of correct answers was as high as 92.0%. The potential for incorrect prediction of fruit with a high seed count as fruit with a low seed count would lead to complaints from consumers, but this incorrect prediction did not occur. The accuracy of the calibration line constructed using the spectra obtained by irradiating five fruit sites with near-infrared light was rated high. There were a few incorrect answers, but they were considered unlikely to be a problem. The calibration lines were prepared and evaluated using fruit from three collection years. The calibration lines were evaluated to be highly accurate when using the 2020 production year with the highest number of total seeds and the 2018 production year with the lowest number of total seeds as the calibration samples. The results suggest that the seed number prediction can be improved using a variety of samples from seedless to many-seeded. Conclusively, it was shown that near-infrared spectroscopy could be used as a selection technique to ensure a stable supply of ‘Koyama New Summer’ as seedless or containing few seeds.