Epsilon Open Archive

Abstract

The thesis presents a rapid and non-destructive method for characterizing the genetic, physiological and technical qualities of both temperate and tropical tree species on single seed basis. It is based on ‘cross fertilization’ of near infrared technology and multivariate analysis. The result demonstrated that seed sources, mothers and fathers of Pinus sylvestris could be identified using near infrared spectroscopy (NIRS) with 100%, 93% and 71% classification accuracies, respectively. NIRS was employed to detect internal insect infestation in Cordia africana and a 100% classification of sound and insect infested seeds were achieved on the basis of insect cuticular components and moisture difference between the two fractions. An extension of this study was performed on Picea abies seeds differing in origin and year of collection. Detection of infested and uninfested seeds with NIRS was found insensitive to subtle differences in proteins, lipids and moisture provided that between-seed lot spectral variability is removed a priori with appropriate spectral pretreatment technique or the calibration model takes into account this natural variability. Sound and insect-damaged seeds of Albizia schimperiana were also successfully separated based on differences in relative water content. The moisture gradient between sound and insect-damaged seeds was intentionally created by soaking both fractions in water at room temperature for a specified time owing to the fact that the hard and impermeable seed coat of intact seeds does not allow the diffusion of water. The application of NIRS for the discrimination of viable and empty seeds of Pinus patula was evaluated and the two fractions were discriminated with 100% accuracy on the basis of divergence in lipid and protein contents. A further study was made to simultaneously discriminate filled, empty and insect-infested seeds of three Larix species. The result demonstrated a 100% recognition of infested and empty seeds while the recognition rates of filled seeds ranged between 90% and 100%; the highest being for Larix sukaczewii followed by Larix decidua and Larix gmelinii, respectively. In seed vigour analysis, it was also possible to distinguish between vigorous and aged seeds with 100% classification accuracy. The results reported in this thesis demonstrate the capability of NIRS combined with multivariate analysis as a tool for rapid and non-destructive analysis of several seed quality attributes. As establishment of new forest plantations shows an increasing trend globally, NIRS will play a pivotal role in upgrading seed lot quality through sorting of unproductive seeds, and hence facilitating single seed sowing for containerised seedling production in nurseries and/or direct sowing out in the field. Therefore, continued emphasis should be given towards developing simple, cost-effective and automated sorting system in the future.