Skip to main content
ARS Home » Research » Publications at this Location » Publication #134774
Title: MULTIVARIATE ANALYSIS OF NEAR-INFRARED (NIR) REFLECTANCE SPECTRA OF DRIED, GROUND CONIFER NEEDLES FROM DIFFERENT ENVIRONMENTS

Author
item RICHARDSON, ANDREW - YALE UNIVERSITY
item GREGOIRE, TIMOTHY - YALE UNIVERSITY
item REEVES III, JAMES

Submitted to: Annals of Botany
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 8/27/2003
Publication Date: 1/1/2004
Citation: Richardson, A.D., Gregoire, T.G., Reeves III, J.B. 2004. Multivariate analysis of near-infrared (nir) reflectance spectra of dried, ground conifer needles from different environments. Annals Of Botany.

Interpretive Summary: Near infrared reflectance (NIR) spectroscopy uses light to determine the composition of materials such as plants. When combined with statistical processes (discriminant analysis) NIR can be used to separate materials into groups based on their composition, i.e. into material of high and low protein. We applied various methods of discriminant analysis (DA) to the reflectance spectra of dried, ground spruce and fir foliage from different growth environments (two crown positions and three elevations). Results suggested significant biochemical differences between species and crown positions. However cluster analysis (groups similar materials together) did not produce satisfactory groupings. DA results showed that samples could be classified on the basis of just a few wavelengths (specific light bands), but the discriminating wavelengths could not be positively linked to known absorption features of biochemical components. Results suggested that there were only small spectral differences among samples from different elevations. In contrast, differences between species were much larger, in agreement with our chemical analysis of foliar nutrients and fiber content, which we demonstrate can be predicted on the basis of NIR. These results suggest the potential applicability of NIR as a useful tool for investigating the degree of biochemical similarity (or dissimilarity) among diverse samples of plant material.

Technical Abstract: Near infrared reflectance (NIR) of plant material is determined primarily by biochemical and structural components. We applied principal components (PCA), cluster (CA), and discriminant analysis (DA) to the reflectance spectra of dried, ground spruce and fir foliage from different growth environments (two crown positions and three elevations). PCA indicated the presence of underlying spectral patterns, and suggested significant biochemical differences between species and crown positions. CA, whether applied to the principal component scores from PCA, or the raw spectral data, did not produce satisfactory groupings. DA results showed that samples could be classified on the basis of just a few wavelengths, but the discriminating wavelengths could not be positively linked to known absorption features of biochemical components. Results suggested that there were only small spectral differences among samples from different elevations. In contrast, differences between species were much larger, in agreement with our chemical analysis of foliar nutrients and fiber content, which we demonstrate can be predicted on the basis of NIR. These results suggest the potential applicability of NIR as a useful tool for investigating the degree of biochemical similarity (or dissimilarity) among diverse samples of plant material. Similar methods might be applied to the rapid screening of other organic samples.