THE USE OF GEOSTATISTICS FOR SAMPLING A CORE COLLECTION OF PERENNIAL RYEGRASS POPULATIONS

The concept of core collections as developed by Brown (1989a) would be very useful in optimizing conservation strategies of natural populati on of outbreeding grasses. The aim of a core is to represent, in a sub sample of manageable size, as much as possible of the genetic variatio n from a large collection. In the case of natural populations of ryegr ass, different methods of stratified sampling with one level of classi fication have been compared: 1. Random sampling. 2. Clustering based o n agronomic traits. 3. Clustering based on the administrative region o f origin. 4. Clustering based on agronomic traits with geographic cont iguity constraint.This last method is based on geostatistics analysis, which allows to study the spatial pattern of genetic diversity. Ryegr ass populations show for many agronomic traits a spatial structure of range 120 km, which could be attributed to the isolation-by-distance p henomenon. This range was used as a constraint for clustering populati ons based on multisite evaluation data. The results show that, in a sp ecies like perennial ryegrass, a random sample of 5% of the accessions maintains 86% of the diversity. Core samples of 10% as recommended by Brown (1989b) enhance the representation to more than 90%. The use of stratified sampling methods is always more efficient than random samp ling. The clustering based on geostatistics gives the best results wit h 92% of the variation being maintained in a 5% core collection.