Abstract: Sampling strategy is very important for core collection establishment. In this study, different methods to construct a core collection of Changjiang spring sowing soybean were performed based on data of SSR（simple sequence repeat） markers and agronomic traits in order to optimize the sampling strategy for core collection. The results showed that, based on cluster analysis of SSR data, each of the three methods, i.e., random sampling within each group, sampling by genetic similarity coefficient within each group and sampling by genetic similarity coefficient, could be used to construct a soybean core collection. When SSR allelic reserving ratios were maintained at 90% and 80%, the core collections had higher genetic diversity indices of SSR alleles than if maintained at 70%. Core collections could also be constructed by agronomic and other basic data if there was no molecular data; however, SSR allelic reserving ratio might be decreased, which suggested that assessment of genetic diversity by SSR data was not always consistent with assessment by agronomic data. We found that core collections were heterogeneous, either when created with different sampling methods or with the same sampling method in different repeats, because randomicity always existed in selecting individuals. This indicated that the germplasm to form a core collection was variable, so optimal sampling strategies should be chosen to establish core collections based on different data according to practical or scientific objectives. Integrating phenotypic and genotypic data together would be better for improving the representativeness of core collections.

Key words: 2010 biodiversity target, assessment indicators, species richness, diversity of ecosystem types, completeness of vertical stratification of vegetation, endemism, extent of biological invasions