Landscape-level thresholds of habitat cover for woodland-dependent birds

Theory suggests that a disproportionate loss of species occurs when total habitat cover decreases to 10-30% of the landscape. To date, little empirical evidence has been collected to test for such thresholds in habitat cover, especially at the landscape scale. Here, we present empirical data on the species richness of woodland-dependent birds collected systematically from 24 landscapes (each 100 km(2)) that sample a gradient in habitat cover from <2% to 60%. To compare the relative effects of habitat cover and habitat configuration, landscapes with similar amounts of habitat but contrasting configuration (i.e., aggregated versus dispersed) were surveyed and the richness of woodland-dependent birds collated for each landscape. The relationship between species richness, habitat cover and habitat configuration was examined using analysis of co-variance (ANCOVA), multiple linear regression and univariate non-linear modelling. There was a significant effect of habitat cover (co-variate) in the ANCOVA, but the main treatment effect of configuration was not significant. However, comparison of non-linear models indicated that the shape of the response curve of species loss with decreasing habitat cover differed between aggregated and dispersed landscapes. Species richness was significantly related to habitat cover in all analyses, explaining between 55% and 60% of the variance in regression models. Mean patch shape complexity and the extent of habitat aggregation were also significant explanatory variables, but explained less than 10% of the variance in richness of woodland birds. Biogeographic variables (range in elevation and geographic location) explained up to 14% of the variance in species richness. There was strong evidence for a threshold response in species richness: non-linear models (broken-stick, exponential, inverse) exhibiting a sharp decline in species richness in landscapes with less than 10% habitat cover provided a better fit to the observed data than linear models. To our knowledge, this is the first empirical demonstration of landscape-level thresholds in species richness. We emphasise that thresholds in species richness denote multiple species' extinction events, the end point of the process of species decline. For viable populations, habitat cover must be maintained well above the threshold level. Finally, thresholds of assemblage measures, such as species richness, potentially mask compositional changes in the avifauna community and may also conceal the loss of species with greater sensitivity to landscape change. (c) 2005 Elsevier Ltd. All rights reserved.