作　　者： ; ; ;

机构地区： 广东省昆虫研究所

出　　处： 《生物数学学报》 2008年第1期66-78,共13页

摘　　要： K-L变换可消除样品各特征间的相关性,去除信息量较少或信息虚假的坐标轴,降低特征空间的维数,用较少量的特征描述样品,是目前最为有效的数据信息压缩抽取方法.利用K-L变换,样品之间变化较小的分类单元(目或功能群等),不会作为重要特征保留下来.因此,数量上占优势、或害虫防治中有重要地位的分类单元并不一定是识别样品的重要特征.可以藉K-L变换明确构成样品变异的主要分类单元,从而对不同地区、不同季节的取样实践有一定的指导意义.用通常的多样性指数分析及多样性差异显著性分析无法解决这个问题.用K-L变换可大大压缩数据存储空间.对数据量很大的生物多样性分析有较大的应用价值.反映样品所携信息量的热带水稻田无脊椎动物目主要是半翅目,中腹足目,以及表腹亚目.其中,在旱季半翅目占主要地位,而在雨季中腹足目占主要地位.就无脊椎动物目的组成而言,它们是构成样品变异的主要类群,是取样中需要得到特别注意的目.反映样品信息的无脊椎动物功能群主要是外部植食类。植食多食性类,半水生水面爬行类,粗粒残物取食类,以及陆生爬行、跳跃或猎者类,是取样中要重点考虑的一些类群. The larger numbers of invertebrate species in rice fields do not allow fast and precise evaluations. Standardized inventory methods must be used over longer time periods to detect significant differences in space and in time, and indicator groups of rice invertebrate biodiversity should be defined. Karhunen-Loève transformation is used to analyze the effectiveness of sample recognition and dimension reduction, using the data of invertebrate biodiversity investigated in tropical irrigated rice field. The results show that there are not significant seasonal differences in relative deviations of the sample recognitions that take functional group as taxon measure, i.e., the relative deviations of sample reconstructions using Karhunen-Loève transformation are similar for different seasons when functional group is considered as taxon measure. Results of Karhunen-Loève transformation to the sampling sets, with invertebrate order and functional group as taxon measure respectively, indicate that recognitions of the sam- ples within training samples produces the least relative deviations. Dimensions of the characteristic space are largely reduced in Karhunen-Loève transformation, with inver- tebrate order and functional group as taxon measure respectively. The lower relative deviations of sample recognition show that Karhunen-Lobve transformation could reduce the dimensions of characteristic space of invertebrate biodiversity without loss significant information carried by samples. The eigenvectors generated from Karhunen-Loève transformation are the “characteristic samples” of training samples （i.e., all training samples that construct Karhunen-Loève transformation can be reconstructed from fewer characteristic samples with a given proportion of information required）. The “charac-teristic taxon” of invertebrate orders and functional groups are analyzed based on these characteristic samples （in a sense, the indicator taxon is the most important characteristic taxon）. Karhunen-Lobve transformation is u

关 键 词： 变换 水稻 无脊椎动物 编目

领　　域： [生物学] [生物学]