DEVELOPMENT OF A CORE COLLECTION FOR PERENNIAL MEDICAGO PLANT INTRODUCTIONS

The size and heterogeneity of world germplasm collections often have h indered their evaluation and utilization. Most of the diversity of ger mplasm collections could be represented by core collections (or core s ubsets), which are representative samples of the genetic diversity of a crop species and its relatives. Individual accessions not included i n the core are retained as reserve subsets. The objectives of this res earch were to evaluate selection procedures for designating a core col lection of plant introduction accessions (P.I.s) and to designate a U. S. P.I. core collection for alfalfa (Medicago sativa L. sensu late). A bout 1100 perennial Medicago P.I.s collected from 47 countries were cl assified into 18 geographical groups based on passport data from the G ermplasm Resources Information Network (GRIN) System. A data file was developed for 50 agronomic, forage quality, root and crown morphology, pest resistance, and stress tolerance traits. Eight methods for devel oping a 200-entry core were compared by nonparametric procedures with the original P.I. collection. The methods included multivariate proced ures, random and/or directed selection of entries, and a totally rando mly selected core. The sign test was used to compare the means and the variances for 38 traits having the most complete data sets among the cores developed by the eight methods. None of the methods changed the trait means; but nearly all methods, with the exception of the randoml y selected core, significantly increased the variance for most traits. Two methods were the best for designating a core collection. One meth od combined cluster analysis based on principal components within each geographical group with random selection of entries within each clust er (Method 3), the other method was direct selection of entries within each geographical group (Method 8). Method 3 needs a complete data se t and extensive computer facilities. Method 8 can utilize incomplete d ata sets and does not require computer facilities. Method 8 was chosen for designating the core collection for perennial Medicago spp. P.I.s because it retained the greatest variability for all traits. This 200 -entry core was one of the first designated for any crop in the GRIN S ystem.