Extinction thresholds for Sapotaceae due to forest cover in Atlantic Forest landscapes

Despite the advances in scientific knowledge about causes and consequences of biodiversity loss, we still lack effective ways to incorporate this knowledge in the development of conservation policies. For example, several studies suggest a nonlinear decrease of species diversity in response to habitat loss at a landscape scale, predicting a threshold at which extinction rates increase dramatically. If this is true and we can find the threshold value with a safety margin, habitat cover is easily measured using GIS technics and can be incorporated into landscape occupation planning. This is particularly true in Neotropics where much of the occupation is founded or licensed by governmental agencies. This study aimed to investigate the thresholds in the relationship between the species richness of a particular plant family in the landscape and the percentage of forest cover on that landscape. Biological material was collected in 11 surveyed landscapes of 6 x 6 km randomly sampled from a large region of Brazilian Atlantic Forest from 11 degrees 80'S to 18 degrees 49'S and 40 degrees 08'W to 21 degrees 24'W in Bahia, Brazil. The landscape forest cover ranged from 5% to 60%, and in each landscape, eight 25 x 10 m plots were randomly established, the landscape's surveyed richness considered the entire set of eight plots. In each landscape we sampled trees with a circumference at breast height (CBH) above 8 cm belonging to Sapotaceae, a tree family that is severely threatened by human activities and generally composed of large, long-lived, animal-dispersed, shade-tolerant trees typical of the interior of forests. Those characteristics make them very representative of mature forest plant species. We evaluated whether a model with a breakpoint or a threshold was best for describing the relationship between richness and forest cover in the landscape using a model selection approach. The data were fitted to a generalized linear model, a piecewise model, and a logistic curve and evaluated using the Akaike information criterion (AIC). In all surveys, a total of 284 individuals belonging to 55 morphospecies were found, and several species are known to be very important in primate diet, which are also dispersers and even pollinators of some species. The best fitted model was the logistic model followed by the piecewise model, which confirms the nonlinearity of the relationship and the existence of a breakpoint. The value of the breakpoint was estimated at near 30% forest cover, and the data set revealed a very sharp decrease in richness and abundance. The value is close to other thresholds obtained in simulated and empirical studies, predicting a threshold between 10% and 30% of habitat in the landscape. Although many studies advocate a species-specific response to habitat loss and fragmentation, Sapotaceae as a whole experienced a dramatic decrease in richness in landscapes with reduced forest cover, most likely due to a common response to deleterious effects triggered by the reduction in habitat, and defaunation. Despite the fact that several processes occur during habitat reduction, such as fragmentation, isolation, increasing edge effects, and loss of key species, we obtained a strong signal of species reduction in response to habitat loss. The extensive sampling conducted and the scale used in this study, with regard to an important and representative group of plant species, reinforces the applicability of this type of study from the standpoint of policy strategies, to promote both species conservation and agricultural activities. (C) 2013 Elsevier B.V. All rights reserved.