采用经典样地法，在云南鸡足山的山桂花（Osmanthus delavayi Franch.）和野桂花〔O. yunnanensis （Franch.） P. S. Green〕群落分别设置5和7个样地，对2个群落的物种组成进行调查；采用自组织特征映射网络（SOM）对2个群落的48个样方进行分类，对群落类型及其与环境因子之间的关系进行分析，并采用多元回归树对分类结果进行检验。结果表明：2个群落分布植物68科114属167种，其中，蕨类植物5科8属10种，裸子植物2科2属2种，被子植物61科104属155种，单种科所占比例达39.7％。山桂花和野桂花在各自群落的乔木层和灌木层中的重要值均较高，伴生种多属壳斗科（Fagaceae）和杜鹃花科（Ericaceae）。物种排序结果显示：高海拔样地以山桂花为优势种，主要伴生种为川滇高山栎（Quercus aquifolioides Rehd. et Wils.）、绒毛杜鹃（Rhododendron pachytrichum Franch.）、火绒草〔Leontopodium leontopodioides （Willd.） Beauv.〕、亮叶杜鹃（Rhododendron vernicosum Franch.）和银叶杜鹃（Rhododendron argyrophyllum Franch.）；低海拔样地以野桂花、滇青冈（Cyclobalanopsis glaucoides Schott.）和华南铁角蕨（Asplenium austrochinense Ching）为优势种，主要伴生种为头状四照花（Cornus capitata Wall.）、大果冬青（Ilex macrocarpa Oliv.）、亮叶杜鹃、银叶杜鹃和黄水枝（Tiarella polyphylla D. Don）。SOM将48个样方划分为6个群落类型，6个群落类型在SOM的拓扑映射图上明显分开且界限明显；海拔、坡度和坡向是影响山桂花群落分布的主要因子，海拔、土壤含水量和坡位是影响野桂花群落分布的主要因子，其中海拔是影响群落类型和分布的最主要因子。多元回归树的分类结果与SOM一致。上述研究结果显示：SOM应用于群落研究可以摆脱传统分类方法的二维限制，揭示群落物种和植物分布与环境因子间的关系，本研究结果为山桂花和野桂花2种中国特有香花植物的群落管理和种质资源保护提供理论参考。

Taking classic plot method， 5 and 7 plots were set up in Osmanthus delavayi Franch. and O. yunnanensis （Franch.） P. S. Green communities in Jizu Mountain of Yunnan Province， respectively， species composition of two communities was surveyed； 48 quadrats of two communities were classified by self-organizing feature map （SOM）， community types and their relationships with environmental factors were analyzed， and classification result was tested by multiple regression trees. The results show that there are 167 species in 114 genera belonging to 68 families in two communities， in which， there are 10 species of ferns in 8 genera belonging to 5 families， 2 species of gymnosperms in 2 genera belonging to 2 families， and 155 species of angiosperms in 104 genera belonging to 61 families. The percentage of single species families is 39.7％. Important values of O. delavayi and O. yunnanensis are both high in arbor and shrub layers of their own communities， and their accompanying species are mostly belonging to Fagaceae and Ericaceae. The result of species ordination shows that the dominant species in plots at high elevation is O. delavayi， and the main accompanying species are Quercus aquifolioides Rehd. et Wils.， Rhododendron pachytrichum Franch.， Leontopodium leontopodioides （Willd.） Beauv.， Rhododendron vernicosum Franch.， and Rhododendron argyrophyllum Franch.； the dominant species in plots at low elevation are O. yunnanensis， Cyclobalanopsis glaucoides Schott.， and Asplenium austrochinense Ching， and the main accompanying species are Cornus capitata Wall.， Ilex macrocarpa Oliv.， R. vernicosum， R. argyrophyllum, and Tiarella polyphylla D. Don. 48 quadrats are divided into 6 community types by SOM, 6 community types have their own regions, and their boundaries on SOM topology map are obvious. Elevation， slope， and aspect are the main factors affecting the distribution of O. delavayi community， while elevation， soil moisture content， and slope position are the main factors affecting the distribution of O. yunnanensis community， in which elevation is the most important factor affecting type and distribution of community. The classification result of multivariate regression trees is consistent with that of SOM. It is suggested that SOM applied for community research can get rid of two-dimensional limitation of traditional classification methods， and reveal the relationships of community species and plant distribution with environmental factors， and these findings provide a theoretical reference for community management and germplasm resources protection of two Chinese endemic fragrant species of O. delavayi and O. yunnanensis.