A correlation analysis on landscape metrics
景观指数之间的相关分析

Numerous landscape metrics have been developed during last three decades to describe and quantify complex landscape patterns and link landscape patterns and ecological processes.This generates a great deal correlated and redundant information if all of the indices were calculated.Thus,it is necessary to identify the correlated landscape metrics and to find out the relationships among them.In our study,the 1999 Landsat imageries of Liaoning Province of China were classified into 16 cover types,the entire study area was clipped into 78 landscape maps according to the county or city boundaries.Thirty nine landscape metrics were calculated for each map,including area,perimeter and density,shape,diversity and contagion.F-Test showed that the correlation coefficient greater than 0.28 is significant at p=0.01 in while the empirical study showed that the correlations between indices exist when correlation coefficient is greater than 0.75. Thus,we considered strong correlations between the landscape indices exist only when the correlation coefficient is greater than 0.75.The result showed that total area is the most basic index within landscape metrics.It correlated with many landscape metrics such as total edge and number of patches in the landscape.Shape metrics are more independent than other groups.Only fractal dimension(log) and mean shape index correlated significantly.Many diversity metrics correlated among themselves,especially the Shannon and Simpson indices.However,Shannon indices are more sensitive to the minor cover type and Simpson indices are more sensitive to common cover types in landscape.Furthermore,Simpson indices are more sensitive to patch density,edge density and average patch area than Shannon indices.Diversity metrics use the proportions of cover types and number of classes to describe the landscape diversity,and contagion metrics use the adjacency matrix to describe the landscape edge diversity,therefore diversity indices are often correlated with contagion indices.Diversity and contagion indices show different aspects of heterogeneity of the study landscape. When an index correlates with another index,the coefficient of variation does not correlate with each other.For example,mean shape index correlates with mean fractal dimension,but their coefficients of variation are independent.Once two basic metrics are correlate with each other,the index consists of two of them are not correlated with the two basic metrics significantly.For example, the total area and number of patches correlate each other,but the patch density does not correlate with them strongly.Finally,the correlation coefficients among landscape metrics were influenced by the landscape patterns,spatial scales,classification systems,mathematical equations,units and ecological meanings. The more common influencing elements above mentioned are in two landscape metrics,the higher probability of strong correlation are among them.