Strategies for Developing Core Collections of Safflower (Carthamus tinctorius L.) Germplasm – Part III. Obtaining Diversity Groups Based on an Information Measure

Evaluation and passport data collected from a germplasm collection of 3250 safflower (Carthamus tinctorius L.) accessions were used to obtain diversity groups based on three different methods. In the first method, 30 diversity groups were obtained by using multivariate cluster analysis followed by further classification into geographical origin and plant types as proposed by Suresh and Balakrishnan (2001). In the second method 13 diversity groups were obtained based on the geographical origin of the accessions. In the third method, the evaluation data were used to compute an information measure (designated as the Length of Encoded Attribute Values or Length of Encoded Attribute Values (LEAV) of the accessions) and this measure was used to divide the whole collection into fourteen diversity groups. Estimates of phenotypic diversity of core samples of size ranging from 5% to 20% of the whole collection were obtained through stratified random sampling from each set of diversity groups obtained by these three methods. The sampling variance of the pooled Shannon Diversity Index (SDI) of 28 descriptors was compared among the three methods of grouping of the accessions. It was found that grouping of the accessions based on the proposed information measure (LEAV) resulted in the least sampling variance. The LEAV index was also used to obtain core samples of required size by ranking the accessions as per the magnitude of this index. It compared better than the core samples obtained through the Principal Component Score method proposed by Noirot et al (1996) in terms of diversity of several qualitative descriptors.