Multivariate correlates of extinction proneness in a naturally fragmented landscape

Abstract

Habitat loss and fragmentation threaten a large proportion of terrestrial biodiversity, and identifying the ecological traits associated with extinction proneness is of widespread interest. We used a multivariate statistical approach to identify combinations of ecological traits that best allowed us to identify extinction-prone amphibians and reptiles in a fragmented landscape in north-eastern Bolivia. Extinction-prone amphibians were rare and did not utilize the savannah matrix separating forest islands, whereas extinction-prone reptiles were trophically specialized. Rarity and matrix aversion are among the most widely reported correlates to extinction proneness, and we argue that an increased understanding of their role as drivers of extinction processes is necessary. We suggest that the absence of reptilian vertebrate predators may exacerbate trophic cascades in habitat patches. © 2007 The Authors.