中文摘要：

采用自组织特征映射网络（SOM）对松山自然保护区山地草甸群落进行了数量分类研究,并用Kruskal-Wallis检验和Tukey多重比较方法分析了草甸类型的环境因子之间差异的显著性。结合完全连接法和SOM,将松山自然保护区的山地草甸群落分为7个类型,其群落结构、物种组成等特征明显。这7个山地草甸群落主要受海拔高度、坡度、枯枝落叶层厚度和土壤深度等环境因子的影响,其差异极显著。生态学分析表明SOM是非常有效的植物群落分类方法,适合于山地草甸植被的研究。

英文摘要：

Aims Vegetation classification is an important topic in plant ecology,and many quantitative techniques for vegetation classification have been developed.The artificial neural network is a comparative new tool of data analysis.In this paper,self-organizing map （SOM） is applied to cluster analysis of mountain meadow data to determine whether SOM is suitable for classifying meadow vegetation.Methods Data for 40 quadrats,87 species,and six environment variables （elevation,slope,aspect,litter layer thickness,soil depth and soil density） from mountain meadow communities in the Sonshan Nature Reserve were analyzed.SOM was used to classify sample quadrats using importance values of species.Important findings The trained SOM classified sample quadrats into seven groups：Saussurea nivea＋ Sanguisorba officinalis＋ Bupleurum chinensis,Sanguisorba officinalis＋ Artemisia annua＋ Vicia unijuga,Polygonum divaricatum＋ Sanguisorba officinalis＋ Carex rigescens,Carex rigescens＋ Sanguisorba officinalis＋ Saussurea nivea,Carex rigescens＋ Polygonum divaricatum＋ Rosa dahurica,Polygonum bistorta＋ Carex rigescens＋ Artemisia gmelinii and Carex rigescens＋ Saussurea iodostegia＋ Polygonum bistorta.The characteristics of community structure and species composition were significant.SOM classification and the dominant species enumeration in the meadow community data reflected the effects of elevation,slope,litter layer thickness and soil depth.SOM is suitable for classifying the mountain meadow communities and could be useful for assessing ecosystem quality and meadow community variations caused by environmental disturbances.