Tropical deforestation is a key contributor to species extinction and climate change, yet the extent of tropical forests and their rate of destruction and degradation through fragmentation remain poorly known. Madagascar's forests are among the most biologically rich and unique in the world but, in spite of longstanding concern about their destruction, past estimates of forest cover and deforestation have varied widely. Analysis of aerial photographs (c. 1953) and Landsat images (c. 1973, c. 1990 and c. 2000) indicates that forest cover decreased by almost 40% from the 1950s to c. 2000, with a reduction in 'core forest' > 1 km from a non-forest edge of almost 80%. This forest destruction and degradation threaten thousands of species with extinction. Country-wide coverage of highresolution validated forest cover and deforestation data enables the precise monitoring of trends in habitat extent and fragmentation critical for assessment of species' conservation status.

T he Convention on Biological Diversity (CBD) sets the policy framework for biodiversity conservation and sustainable use through the commitments of 195 countries and the European Union. The Strategic Plan for Biodiversity 2011-2020 included Aichi Biodiversity Target 12, which set the goal for 2020 of preventing the extinction of known threatened species and improving and sustaining their conservation status. Despite government commitments and successful efforts for certain species 1 , the overall extinction risk continues to increase, and widespread implementation shortfalls will prevent Target 12 from being met 2 . A new global framework with revised goals and targets is currently being negotiated, which places the stabilization and restoration of species' populations as an outcome goal for 2030, as a stepping stone towards the CBD's 2050 Vision 3,4 .

The spatial scale of conservation necessary to avoid species extinctions is one of the most vigorous debates in conservation biology. One approach holds that protecting sites should be the primary level for action on the ground, the other that conservation action targeting broader seascapes and landscapes is more important. We address this debate systematically by assessing the appropriate spatial scales of conservation for all 4,239 threatened mammals, birds, tortoises and turtles, and amphibians. We find that, in the short-to medium term, 20% of these species are dependent on conservation at single sites, 62% on multiple sites, 18% on both sites and sea-or landscape-scale efforts, and <1% on broadscale actions alone (where sites are variably sized units that are actually or could potentially be managed for conservation, and "broad scale" refers to seaor landscape-scale and is determined by the needs of the species in question). Calls for broad-scale conservation action have generally focused on terrestrial birds and mammals, and we confirm that a fifth and a tenth of these, respectively, require conservation action at the landscape scale. However, we also find that two-fifths of threatened freshwater turtles and one-fifth of threatened amphibians depend on broad-scale conservation action to address changes in freshwater processes. Furthermore, the overwhelming majority of threatened marine mammals, birds, and turtles require urgent conservation action at the seascape scale. Our key conclusion is that neither site-scale nor broad-scale approaches alone can prevent mass extinction. Although site protection should remain the cornerstone for almost all threatened species, we demonstrate that a substantial proportion and unexpected diversity of threatened species will be lost in the absence of urgent conservation interventions at the sea-or landscape scale.

No abstract