Monitoring Photosynthetic Pigments of Shade-Grown Tea from Hyperspectral Reflectance

Abstract

The highest-quality green tea is cultivated using shading treatments in Japan; however, shading can lead to early mortalities of tea due to excessive environmental stress. The allocation of photosynthetic pigments – chlorophyll a and b, and carotenoids – could be a good indicator for evaluating production or environmental stress in plants; thus, developing an in situ method to monitor photosynthetic pigments is useful for agricultural management. To assess the accuracy of the estimation of photosynthetic pigment contents with existing supervised learning models, 4 different approaches were compared including random forests, kernel-based extreme learning machine (KELM), deep belief nets and support vector machine. Overall, KELM had the highest performance, with a root mean square error of 1.95 ± 0.36 µg cm−2, 1.08 ± 0.11 µg cm−2 and 0.68 ± 0.10 µg cm−2 for estimating chlorophyll a, b and carotenoid contents, respectively.