TY - JOUR
T1 - Determination of leaf water content by visible and near-infrared spectrometry and multivariate calibration in Miscanthus
AU - Jin, Xiaoli
AU - Shi, Chunhai
AU - Yu, Chang Yeon
AU - Yamada, Toshihiko
AU - Sacks, Erik J.
N1 - Publisher Copyright:
© 2017 Jin, Shi, Yu, Yamada and Sacks.
PY - 2017/5/19
Y1 - 2017/5/19
N2 - Leaf water content is one of the most common physiological parameters limiting efficiency of photosynthesis and biomass productivity in plants including Miscanthus. Therefore, it is of great significance to determine or predict the water content quickly and non-destructively. In this study, we explored the relationship between leaf water content and diffuse reflectance spectra in Miscanthus. Three multivariate calibrations including partial least squares (PLS), least squares support vector machine regression (LSSVR), and radial basis function (RBF) neural network (NN) were developed for the models of leaf water content determination. The non-linear models including RBF_LSSVR and RBF_NN showed higher accuracy than the PLS and Lin_LSSVR models. Moreover, 75 sensitive wavelengths were identified to be closely associated with the leaf water content in Miscanthus. The RBF_LSSVR and RBF_NN models for predicting leaf water content, based on 75 characteristic wavelengths, obtained the high determination coefficients of 0.9838 and 0.9899, respectively. The results indicated the non-linear models were more accurate than the linear models using both wavelength intervals. These results demonstrated that visible and near-infrared (VIS/NIR) spectroscopy combined with RBF_LSSVR or RBF_NN is a useful, non-destructive tool for determinations of the leaf water content in Miscanthus, and thus very helpful for development of drought-resistant varieties in Miscanthus.
AB - Leaf water content is one of the most common physiological parameters limiting efficiency of photosynthesis and biomass productivity in plants including Miscanthus. Therefore, it is of great significance to determine or predict the water content quickly and non-destructively. In this study, we explored the relationship between leaf water content and diffuse reflectance spectra in Miscanthus. Three multivariate calibrations including partial least squares (PLS), least squares support vector machine regression (LSSVR), and radial basis function (RBF) neural network (NN) were developed for the models of leaf water content determination. The non-linear models including RBF_LSSVR and RBF_NN showed higher accuracy than the PLS and Lin_LSSVR models. Moreover, 75 sensitive wavelengths were identified to be closely associated with the leaf water content in Miscanthus. The RBF_LSSVR and RBF_NN models for predicting leaf water content, based on 75 characteristic wavelengths, obtained the high determination coefficients of 0.9838 and 0.9899, respectively. The results indicated the non-linear models were more accurate than the linear models using both wavelength intervals. These results demonstrated that visible and near-infrared (VIS/NIR) spectroscopy combined with RBF_LSSVR or RBF_NN is a useful, non-destructive tool for determinations of the leaf water content in Miscanthus, and thus very helpful for development of drought-resistant varieties in Miscanthus.
KW - Drought-resistant breeding
KW - Leaf water content
KW - Miscanthus
KW - Sensitive wavelengths
KW - VIS/NIR spectroscopy
UR - http://www.scopus.com/inward/record.url?scp=85021058375&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85021058375&partnerID=8YFLogxK
U2 - 10.3389/fpls.2017.00721
DO - 10.3389/fpls.2017.00721
M3 - Article
C2 - 28579992
AN - SCOPUS:85021058375
SN - 1664-462X
VL - 8
JO - Frontiers in Plant Science
JF - Frontiers in Plant Science
M1 - 721
ER -