Rarefaction, Relative Abundance, and Diversity of Avian Communities

The common practice of expressing community structure in terms of indices of diversity and evenness involves a serious loss of information.Differences attributable to the accumulation of species with increasing area are ignored, differences in the density of individuals are often masked by other factors, and many combinations of species richness and relative abundance can produce the same value of the index.As an alternative we suggest (1) comparing species richness by standardizing samples either to equal numbers of individuals or to the number of individuals expected on equal-sized plots, and (2) expressing the relative abundance of species as a graph of their relative abundances arranged in a decreasing array.We present an analysis of bird census data based on the proposed methods, and we include comparisons with applications of four indices commonly used in ecology, the Shannon-Weaver index of diversity, the J' evenness index, the inverse of Simpson's measure of concentration, and Hill's evenness index.For 37 Breeding Bird Censuses taken in various terrestrial habitats across the United States and Canada, the proposed methods reveal some very general relationships about the organization of bird communities in different habitats.Equal-sized areas of mature deciduous forest and secondgrowth habitats may be equally species rich (14-24 species with •> 1 breeding territory per 6 ha); the density of individuals (territorial pairs) is generally higher in deciduous forest habitats, and the relative abundance of bird species shows more dominance (less evenness) in the deciduous forest.Mixed coniferous-deciduous forests and dense young deciduous forests have fewer species than mature eastern deciduous forests or second-growth habitats (9-16 and 7-10 species per 6 ha, respectively), although the density of individuals is approximately equal to that in second-growth habitats.Coniferous forests are species-poor (5-8 per 6 ha), and the density of territorial pairs is low (8-12 per 6 ha compared with 40-70 in deciduous forests).Although the proposed methods require assumptions that need to be evaluated carefully, we are optimistic that they will have other useful applications in the analysis of arian communities.