Abstract: Ramie (Boehmeria nivea L. Gaud) is an important bast fiber crop. Its fiber, one of the best bast fibers, is widely used in textile industry and plays an indispensable and irreplaceable role in the national economy. Core collection is a representative subset of entire germplasm collection, consisting of introduced accessions with minimum genetic redundancy and retained most initial collection. To develop the method of establishing ramie core collection, based on 25 agrinomic traits data from 790 ramie germplasm of national ramie germplasm nursery, using software provided by professor Xu of Zhejiang university, we sampled a serials of subsets at 20% proportion, respectively, by six hierarchical clustering methods (single linkage, complete linkage, median method, centroid method, ward’s, unweighted pair-group method with arithmetic means), three sampling methods (stepwise clustering with random sampling strategy and preferred sampling strategy, stepwise clustering with deviation sampling strategy and preferred sampling strategy, random sampling strategy), three genetic distances in disordered qualitative traits (simple matching, jaccard, nei and li), and three genetic distances in ordered qualitative traits and quantitative characteristics (mahalanobis distance, euclidean distance, principal component). The genetic variation among subsets was compared by evaluating the I (index of genetic diversity), RPR (ratio of phenotype retained) of qualitative traits and MD (mean difference percentage), VD (variance difference percentage), CR (coincidence difference percentage), VR (variable rate of coefficient of variation) of quantitative traits. The results showed that different sampling methods have different impacts on qualitative traits and quantitative traits. For the largest genetic diversity of qualitative traits, it is effective to choose stepwise clustering with random sampling strategy and preferred sampling strategy. For the largest genetic diversity of quantitative traits, it is effective to choose stepwise clustering with deviation sampling strategy and preferred sampling strategy. The ramie core collection is the best one constructed by centroid method, single linkage clustering way under stepwise clustering with random sampling strategy and preferred sampling strategy and constructed by ward’s clustering way under stepwise clustering with deviation sampling strategy and preferred sampling strategy. Core collection in ramie was not related with genetic distances of qualitative traits, whereas the core collection of ramie constructed by euclidean distance of quantitative traits was the best.

Key words: Ramie, Core collection, Sampling method, Clustering method, Genetic distance