Confounding Variables and Independent Observations in Comparative Analyses of Food Webs

Abstract

Habitat differences in variables describing food web structure were examined using data from two food web compilations, taking into account differences between webs in species richness and using independent data points only. Log—log regressions of connectance, linkage density, and food chain length on species number were calculated. In most cases these relationships were significant. The residuals from the regression lines were used to analyze differences between habitats in connectance, linkage density, and food chain length. In one compilation of insect—dominated webs, aquatic habitats had comparatively high connectance and linkage density when species number had been accounted for. In the other compilation, marine pelagic webs had higher relative connectance, relative linkage density, and mean chain length. In addition, desert webs had higher relative connectance. Several of the results when taking species richness into account do not agree with previous studies. Therefore it is concluded that it is important in examinations of food web patterns to account for web species number and to use independent observations as far as possible.