Patterns in the species-environment relationship depend on both scale and choice of response variables

Publication Date

2004

Journal or Book Title

Oikos

Abstract

Multi-scale investigations of species–environment relationships are an important tool in ecological research. The scale at which independent and dependent variables are measured, and how they are coded for analysis, can strongly influence the relationships that are discovered. However, little is known about how the coding of the dependent variable set influences community-level analyses. In this study, we used canonical correspondence analysis to quantify species–environment relationships between environmental factors collected at three spatial scales and the structure of a forest bird community in the Oregon coast range. The main question in our analysis was how coding the bird data as abundance versus presence/absence affected the nature and strength of observed relationships. As we expected, the structure of the bird community was better described overall using abundance data than it was using presence/absence data. However, individual species and life-history groups appeared to exhibit different species–environment relationships in abundance versus presence/absence data. In particular, common species with a high frequency of occurrence among sample points exhibited a stronger ‘abundance’ signature, whereas uncommon species with a low frequency of occurrence exhibited a stronger ‘presence/absence’ signature. In addition, the apparent importance of plot-level factors in explaining the variation in the bird community was greater for abundance data, whereas patch and landscape factors were more important in the presence/absence data. Thus, conclusions about the relative importance of factors at different scales is largely contingent on the way in which the species-response data are coded for analysis. For communities as a whole, and for individual species within them, the strength and nature of species–environment relationships can differ dramatically between analyses using presence/absence versus abundance data.

DOI

https://doi.org/10.1111/j.0030-1299.2004.12524.x

Volume

105

Issue

1

Pages

117-124

This document is currently not available here.

Share

COinS