Apex predators are threatened globally, and their local extinctions are often driven by failures in sustaining prey acquisition under contexts of severe prey scarcity. The harpy eagle Harpia harpyja is Earth’s largest eagle and the apex aerial predator of Amazonian forests, but no previous study has examined the impact of forest loss on their feeding ecology. We monitored 16 active harpy eagle nests embedded within landscapes that had experienced 0 to 85% of forest loss, and identified 306 captured prey items. Harpy eagles could not switch to open-habitat prey in deforested habitats, and retained a diet based on canopy vertebrates even in deforested landscapes. Feeding rates decreased with forest loss, with three fledged individuals dying of starvation in landscapes that succumbed to 50–70% deforestation. Because landscapes deforested by > 70% supported no nests, and eaglets could not be provisioned to independence within landscapes > 50% forest loss, we established a 50% forest cover threshold for the reproductive viability of harpy eagle pairs. Our scaling-up estimate indicates that 35% of the entire 428,800-km2 Amazonian ‘Arc of Deforestation’ study region cannot support breeding harpy eagle populations. Our results suggest that restoring harpy eagle population viability within highly fragmented forest landscapes critically depends on decisive forest conservation action.

It is commonly accepted that species should move toward higher elevations and latitudes to track shifting isotherms as climate warms. However, temperature might not be the only limiting factor determining species distribution. Species might move to opposite directions to track changes in other climatic variables. Here, we used an extensive occurrence data set and an ensemble modelling approach to model the climatic niche and to predict the distribution of the seven baobab species (genus Adansonia) present in Madagascar. Using climatic projections from three global circulation models, we predicted species' future distribution and extinction risk for 2055 and 2085 under two representative concentration pathways (RCPs) and two dispersal scenarios. We disentangled the role of each climatic variable in explaining species range shift looking at relative variable importance and future climatic anomalies. Four baobab species (Adansonia rubrostipa, Adansonia madagascariensis, Adansonia perrieri¸ and Adansonia suarezensis) could experience a severe range contraction in the future (>70% for year 2085 under RCP 8.5, assuming a zero-dispersal hypothesis). For three out of the four threatened species, range contraction was mainly explained by an increase in temperature seasonality, especially in the North of Madagascar, where they are currently distributed. In tropical regions, where species are commonly adapted to low seasonality, we found that temperature seasonality will generally increase. It is, thus, very likely that many species in the tropics will be forced to move equatorward to avoid an increase in temperature seasonality. Yet, several ecological (e.g., equatorial limit, or unsuitable deforested habitat) or geographical barriers (absence of lands) could prevent species to move equatorward, thus increasing the extinction risk of many tropical species, like endemic baobab species in Madagascar.

A central tenet in ecology states that biodiversity distribution on Earth is neither uniform nor random (Soria-Carrasco & Castresana, 2012).Variation in latitude and elevation is widely recognized biogeographic gradients that profoundly influence multilevel patterns of species diversity world-wide (Willig & Bloch, 2006). Latitudinal influences on species diversity range from taxonomic components to functional traits and phylogeny (Buckley et al., 2010). Species diversity increase from polar to tropical regions (Hillebrand, 2004;Willig et al., 2003).To explain the high levels of tropical diversity, three main hypotheses have been posed: (a) the 'Cradle hypothesis': the premise that there