A global assessment of the drivers of threatened terrestrial species richness

High numbers of threatened species might be expected to occur where overall species richness is also high; however, this explains only a proportion of the global variation in threatened species richness. Understanding why many areas have more or fewer threatened species than would be expected given their species richness, and whether that is consistent across taxa, is essential for identifying global conservation priorities. Here, we show that, after controlling for species richness, environmental factors, such as temperature and insularity, are typically more important than human impacts for explaining spatial variation in global threatened species richness. Human impacts, nevertheless, have an important role, with relationships varying between vertebrate groups and zoogeographic regions. Understanding this variation provides a framework for establishing global conservation priorities, identifying those regions where species are inherently more vulnerable to the effects of threatening human processes, and forecasting how threatened species might be distributed in a changing world.

stability since the last interglacial period (125,000 years ago) to explain threatened species richness.
Palaeo-climate data were made available by the Bristol Research Initiative for the Global Environment (BRIDGE, http://www.bridge.bris.ac.uk/). Comprehensive details of the model used to derive the palaeo-climate data are available elsewhere 13,14 . Data on precipitation and temperature were sampled at 4,000-year intervals. For each temporal transition, the Euclidean distance was calculated between z-transformed temperature and precipitation in bivariate space. The mean Euclidean distance was used as a measure of long-term climate stability in a grid cell, with smaller values indicating more stable climates 15 .

Human Influence
Area of anthropogenic land use: Using the land cover data derived from GlobCover version 2.3 (2009, http://due.esrin.esa.int/page_globcover.php) we summed the area of each grid cell that was classified as intensively used by humans. These intensively used land covers included areas of cropland of varying intensity and urbanised areas with artificial surfaces.
Human Influence Index: As the occurrence of intensively used lands can manifest in areas with minimal human settlement 16,17 , we also included the Human Influence Index (HII, V2, http://sedac.ciesin.columbia.edu/wildareas/) as an additional measure of human impact. These data are available at a resolution of 1 km 2 and are derived from nine global data layers that cover the period between 1995 and 2004. These data layers cover a range of anthropogenic pressures, both direct and indirect, including human population density, distributions of roads, railways and navigable rivers, and night-time lighting.
Area of protected land: The total area of land receiving some form of protection from transformation was obtained from the World Database on Protected Areas (WDPA: https://www.protectedplanet.net/). We summed the area of land classified as strict nature reserve (Ia), wilderness areas (Ib), national park (II), natural monument or feature (III), habitat or species management area (IV), protected landscape where the interaction of people and nature has produced an area of distinct character with significant ecological and cultural value (V), and areas managed for sustainable use of natural resources (VI). Our expectation was that grid cells with a large area of protected land would support more threatened species than those grids with little to no protection from land transformation, as species could seek refuge in these relatively less disturbed areas 18 .

Short-term land cover change:
Habitat loss is widely regarded as the primary driver of biodiversity loss 19 , with recent changes in the extent of anthropogenic land use linked with changes in species' extinction risk 20 . To quantify the extent of recent changes in land cover, we obtained data on global land cover from the European Space Agency Climate Change Initiative (ESA CCI, https://www.esalandcover-cci.org/?q=node/1). These data are available at a spatial resolution of 300 m and consist of annual maps of global land cover between 1992 and 2015. We used these maps to calculate the percentage change in the area of land cover classes associated with anthropogenic land uses between 1992 and 2015 for each of our 0.5° grid cells.
Long-term land cover change: Over the past three centuries, human activities have significantly altered global landscapes, primarily through the conversion of primary habitats to agriculture 21 . If we are to appreciate fully the potential effects of human influences on threatened species richness, we need to consider the magnitude of these past changes. For this we obtained data on geographically explicit changes in croplands between 1700 and 1992. These data were derived from models that use remotely sensed land cover classification data alongside contemporary and historical cropland inventory data to reconstruct historic cropland distributions 21 . These data were available at a 5 min resolution (approximately 10km). We used them to calculate the percentage change in cropland area between 1700 and 1992 for each of our 0.5° grid cells.
Invasive alien species: Invasive alien species are considered one of the greatest threats to biodiversity, second only to habitat loss and fragmentation 22 . For our index of invasive alien species pressures, we obtained data that use information from the Global Invasive Species Database (GSID) and the Centre for Agriculture and Bioscience International's Invasive Species Compendium (CABI ISC) to estimate the number of invasive alien species (IAS) per country 23 . These data include the majority of recorded taxa, including plants, arthropods, amphibians, reptiles, birds and mammals.
We intersected these national measures of IAS with the equal area 0.5° grid. For grid cells that intersected more than one country we calculated the area-weighted average IAS across all intersected countries. Variables are grouped into broader classes, which are indicated by the capital letters on the side of the variable names: Environmental (E) and Human Impact (H) covariates. Note these models do not included total species richness as a covariate. Bars are shaded according to modelled response: threatened species richness (purple) and total species richness (yellow). Variables are ordered from top to bottom first by group, and then by their importance in the global model of threatened vertebrate species richness. The line across each box indicates the median and the box boundaries indicate the interquartile range (IQR). Whiskers identify extreme data points that are not more than 1.5 times the IQR on both sides; the dots are more extreme outliers. Source data are provided as a Source Data file. Figure 10: Partial residual plots from global models of threatened amphibian, reptile, bird, and mammal species richness. Lines indicate the mean partial relationship between variables and threatened species richness from across 10 random forest models. The x-axis is limited to the central 90% of a variable's range. Shaded areas indicate the standard deviation around the mean partial relationship. Colours of both lines and shaded areas indicate taxonomic group: green = all taxa, yellow= amphibians, blue = birds, orange = mammals, and pink = reptiles. To aid comparison, responses have been scaled to have a mean of one and a standard deviation of zero. Source data are provided as a Source Data file. Figure 11: Partial residual plots from global models of total amphibian, reptile, bird, and mammal species richness. Lines indicate the mean partial relationship between variables and threatened species richness from across 10 random forest models. The x-axis is limited to the central 90% of a variable's range. Shaded areas indicate the standard deviation around the mean partial relationship. Colours of both lines and shaded areas indicate taxonomic group: green = all taxa, yellow= amphibians, blue = birds, orange = mammals, and pink = reptiles. To aid comparison, responses have been scaled to have a mean of one and a standard deviation of zero. Source data are provided as a Source Data file.