Classification and comparison of Gliricidia provenances using near infrared reflectance spectroscopy

There is an ever-increasing need to identify new feed resources in developi ng countries and the types of forages used tend to be more complex in terms of chemical composition. Near infrared reflectance spectroscopy (NIRS) has the potential to aid the evaluation of forages and in this study was emplo yed to compare and classify Gliricidia spp. provenances. Multivariate stati stical techniques, including biplot, principal component analysis (PCA), di scrimination, hierarchical cluster and canonical variate analysis (CVA) wer e used to compare the dried foliage samples of 25 different provenances of Gliricidia spp. which were grown on one site in Honduras to avoid confoundi ng provenance effects with environmental effects or interactions, Marked in terprovenance differences were observed in the 1560-1740, 2060-2170 and 232 0-2360 nm spectral regions, particularly for provenance M23 (43/87). This p rovenance was found to be distinct in graphical plots from biplot, PCA and cluster analysis and is in fact a different species, i.e. Gliricidia macula ta. In addition, inter-provenance distances between populations representin g provenances G2, G5, H7, M10 and V17 when compared to their intra-provenan ce variation, were all found to be statistically significant, with the exce ption of that between G2 and H7. Discriminant analysis showed that of the r emaining 20 individual provenances, samples were more similar to the compos ite (V17) and multiple introduction (H7) populations than the unique popula tions (G2 and G5), NIRS combined with multivariate techniques therefore sho ws potential to classify provenances on the basis of their spectral feature s, which are a comprehensive record of sample chemistry, and aid the select ion of alternative forages. (C) 2000 Elsevier Science B.V. All rights reser ved.