Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

Partial Least Squares - Discriminant Analysis (PLS-DA) of Miscanthus x giganteus by FT-NIR Spectroscopy

Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan

Citation:  Paper number  131596145,  2013 Kansas City, Missouri, July 21 - July 24, 2013. (doi: @2013
Authors:   Daniel A Williams, Mary-Grace C Danao, Marvin R Paulsen, Kent D Rausch, Ana B. Ibáñez, Stefan Bauer
Keywords:   Miscanthus x giganteus composition FT-NIR spectroscopy partial least squares regression linear discriminant analysis.

Abstract. The objectives of this research were to describe the variation in chemical composition of Miscanthus x giganteus and to probe the potential of using Fourier transform near infrared (FT-NIR) spectroscopy in quantitatively analyzing the composition of Miscanthus and qualitatively classifying Miscanthus. Large variations in glucan (40.7 ± 2.37%), xylan (20.6 ± 1.20%), arabinan (1.83 ± 0.36%), acetyl (2.84 ± 0.28%), lignin (20.7 ± 1.35%), ash (2.60 ± 1.64%), and extractives (5.59 ± 0.86%) content were observed for 67 samples used in the calibration set that were collected from Miscanthus bales stored under a variety of conditions (indoors, under roof, outdoors with tarp cover, and outdoors without tarp cover) for a period of 1 to 24 months after harvest and baling. The composition of samples used for validation and model testing were comparable. Partial least squares (PLS) regression models based on the FT-NIR spectra of core samples collected from bales can be used to predict glucan, xylan, lignin, and ash contents with RPD values of 4.86, 4.08, 3.74, and 1.71, respectively. These models were used with linear discriminant analysis to classify the samples based on their glucan, lignin, and ash contents. The best classification results were based on the PLS-DA lignin model, which classified the samples into three groups, with small variations with each group. While the models developed in this study were based on a small sample size (less than 100 for calibration) and the small size contributed to some of the inaccuracy and imprecision in the predictions, the approach demonstrated that FT-NIR spectra and PLS-DA modeling can be used to rapidly screen Miscanthus samples at different stages of the supply chain, including after long-term storage.

(Download PDF)    (Export to EndNotes)