Evaluation of variability in a fog grass (Holcus spp.) gene pool : a thesis presented in partial fulfilment of the requirement for the Master of Agricultural Science, Massey University, Palmerston North, New Zealand
The utimate aim of a breeder's working collection is to utilize the genetic variation for breeding new cultivars. Before this variation can be utilised, it is necessary to obtain the description of attribute from the collection. These can either be obtained from the records of genetic resources (base collection or active collection) or obtained directly from the working collection itself. This will resulted in a huge amount of data. To bo of any value, this information need to be classified systematically, and the classification need high degree of objectivity, especially for species of no a priori knowledge. A Working collection of 160 Yorkshire Fog seed populations, from all over New Zealand, were planted out in Massey University. This formed the gene pool. During Summer 1975, 11 agronomic andi morphologic characters were scored in a semi-quantitative scale. This yield approximately 42,000 data records. These were then systematically reduced to 550 by a series of multivariate analysis techniques. The procedures of Multivariate Analysis of Variance, Multiple Discriminant Analysis and Cluster Analysis were reviewed and their computer programmes were developed. The clustering behaviours of seven agglomerative polythetic strategies were studied and compared, using the full set of characters. Most of the results concurred with studies carried out by other workers The Minimum Increment Sum of Squares strategy was found to be most suitable for this analysis. A probabilistic decision method was devised to decide objectively, the truncating point for clustering. For all set of data, the studies did not reveal any ecotypes and hence did not agree with the ecoclinal trends hypotheses (of Yorkshire Fog in New Zealand) of Jacquee.The approaches of both studies (of Yorkshire Fog in New Zealand) of Jacques. The approaches of both studies (that of Jacques and the present one) were reviewed critically and a more appropriate approach was suggested for future ecological study. Preliminary results revealed that there were a few promising groups showing agronomic desirable characters. They were promising breeding materials for future lines selection. Of all the characters studied, flowering date and clump erectness were found to be the most discriminating characters amongst groups, and the most dominant characters in clustering. These implied that selection should be beneficial, if they had moderate high predictive heretability.