中文摘要：

为了解植物群落在多尺度下的空间变异规律及其对空间尺度的依赖,以川西南山地阔叶混交林为对象,在有代表性的地段设立100m×100m样地,采用传统的罗盘仪对树体的相对空间位置进行定位,运用主轴邻距法（principal coordinates of neighbor matrices,PCNM）对群落空间结构的多尺度（100m内）特征进行了研究。结果表明,群落均匀度指数、空隙度和开度、林分胸高断面积和林分密度在大尺度上表现出显著的空间结构。在不同的尺度下,林分密度、叶面积指数和林下地表直接辐射都对均匀度指数有影响,林分密度、林下地表直接辐射、生物量和空隙度都对土壤有机质含量有显著影响,群落结构与环境因子都表现出明显的空间结构。PCNM分析可获取样本间空间关系在不同尺度上的分解向量,与群落结构显著相关的PCNM因子即是群落或生境在该尺度上所表现出的显著的空间结构。典型相关分析结果表明群落结构因子和光因子相关性显著,两者都与PCNM变量极显著相关且相互影响,表明群落结构和光因子都表现出显著的空间结构。因此,利用PCNM对群落结构进行空间分析有助于理解群落空间异质性对尺度的依赖。

英文摘要：

To investigate the spatial-dependence of heterogeneity at multiple scales for a community, we se- lected a representative plot of 100 m× 100 m in the mountainous evergreen and deciduous mixed broad-leaved forest in Sichuan, southwest China （102°50′E, 30°02′N）. The location of every tree was mapped by compass, and all-scale analysis of spatial structure of forest community was conducted by the method of PCNM （principal coordinates of neighbor matrices）. The results showed that Pielou evenness index, gap-fraction, openness, stand basal area, and stand density were influenced by spatial structure of the community at a broad scale. At the same time Pielou evenness index was impacted significantly by stand density, direct radiation, and leaf area index at all scales, while soil organic matter contents were markedly influenced by stand density, biomass, direct radiation, and gap-fraction at each scale. It could be concluded that community structure and environmental factors are markedly influenced by spatial sampling procedures. Our results highlighted that PCNM analysis could achieve a spectral decomposition vector of the spatial rela- tionships among sampling sites, and that the significant PCNM variables could be directly interpreted in terms of spatial scales, or including variation decomposition with respect to spatial and environmental com- ponents. Canonical correlation analysis also indicated that forest community structure variables were corre- lated significantly with light factors, and both were interacted with each other and correlated well with PCNM variables. Therefore, the method of PCNM could help to understand the scale-dependence of heterogeneity at the community level.