常用景观指数的因子分析和筛选方法研究

Study on Factor Analysis and Selection of Common Landscape Metrics

本文以42幅云南省一平浪林场卫星遥感图为数据源,采用相关分析、因子分析和敏感性分析等方法,对13个常用的景观指数进行了分类和筛选方法的研究。研究结果显示大多数指数间呈现极高的相关性,通过因子分析提取出了累积贡献率达82.03%的3个公因子,遂将13个景观指数分为了3大类,并利用敏感性系数筛选出了4个具有良好灵敏度、能充分反映生态学意义的代表性景观指数,即斑块个数、平均最近距离、面积加权的平均形状指数和散布与并列指数。本方法有效地解决了景观分类和评价中指数繁多冗余的问题。

Based on data from 42 remote sensing images of Yipinglang Forest Farm in Yunnan Province, 13 common landscape metrics were classified and selected by using the methods of correlation coefficient analysis, multivariate factor analysis and sensitivity analysis. The results indicated that most of them had significant correlationship, and three factors were identified, which can explain about 82.03% of the variation in the 13 metrics. And the 13 landscape metrics were classified into three groups. Four representative landscape metrics which can perform highly- sensitive and reflect eco-system well were selected by sensitivity index, which are number of patches（ NP）, euclidean nearest neighbor distance（MNN）, area-weighted mean shape index（MSI）, interspersion ＆ juxtaposition index（IJI）. This method can solve the problem of metrics redundancy efficiently in landscape classification and evaluation.