RESUMO
Tropical tree species are increasingly being pushed to inhabit deforested landscapes. The habitat amount hypothesis posits that, in remaining forest patches, species diversity in equal-sized samples decreases with decreasing forest cover in the surrounding landscape. We tested this prediction by taking into account three important factors that can affect species responses to forest loss. First, forest loss effects can be linear (proportional) or nonlinear, as there can be threshold values of forest loss beyond which species extirpation may be accelerated. Second, such effects are usually scale dependent and may go unnoticed if assessed at suboptimal scales. Finally, species extirpation may take decades to become evident, so the effects of forest loss can be undetected when assessing long-lived organisms, like adult old-growth forest trees. Here, we evaluated the linear and nonlinear effects of landscape forest loss across different spatial scales on site-scale abundance and diversity of old-growth forest trees, separately for four plant-life stages (seeds, saplings, juveniles, and adults) in two rainforest regions with different levels of deforestation. We expected stronger (and negative) forest loss effects on early plant-life stages, especially in the region with the highest deforestation. Surprisingly, in 13 of 16 study cases (2 responses × 4 life stages × 2 regions), null models showed higher empirical support than linear and nonlinear models at any scale. Therefore, the species richness and abundance of local tree assemblages seem to be weakly affected by landscape-scale forest loss independently of the spatial scale, life stage, and region. Yet, as expected, the predictive power of forest cover was relatively lower in the least deforested region. Our findings suggest that landscape-scale forest loss is poorly related to site-scale processes, such as seed dispersal and seedling recruitment, or, at least, such effects are too small to shape the abundance and diversity of tree assemblages within forest patches. Therefore, our findings do not support the most important prediction of the habitat amount hypothesis but imply that, on a per-area basis, a unit of habitat (forest) in a highly deforested landscape has a conservation value similar to that of a more forested one, particularly in moderately deforested rainforests.