Skip to main content

An ecoregion-based approach to restoring the world’s intact large mammal assemblages

Cite this dataset

Vynne, Carly et al. (2022). An ecoregion-based approach to restoring the world’s intact large mammal assemblages [Dataset]. Dryad.


Assemblages of large mammal species play a disproportionate role in the structure and composition of natural habitats. Loss of these assemblages destabilizes natural systems, while their recovery can restore ecological integrity. Here we take an ecoregion-based approach to identify landscapes that retain their historically present large mammal assemblages, and map ecoregions where reintroduction of 1–3 species could restore intact assemblages. Intact mammal assemblages occur across more than one-third of the 730 terrestrial ecoregions where large mammals were historically present, and 22% of these ecoregions retain complete assemblages across >20% of the ecoregion area. Twenty species, if reintroduced or allowed to recolonize through improved connectivity, can trigger restoration of complete assemblages over 54% of the terrestrial realm (11,116,000 km2). Each of these species have at least two large, intact habitat areas (>10,000 km2) in a given ecoregion. Timely integration of recovery efforts for large mammals strengthens area-based targets being considered under the Convention on Biological Diversity.


The IUCN Red List of Threatened Species, along with data on body mass, was used to generate a species list of 298 large extant terrestrial mammals (IUCN, 2019; Jones et al., 2009; Myers et al., 2020; Smith et al., 2003; Table S1). ‘Large mammals’ were defined as those exceeding 15 kg maximum recorded body mass, a threshold that allows the inclusion of key predators, their prey, and other large herbivores, and that is consistent with other studies (Ferreira et al. 2020; Ripple et al., 2014; Salom-Peréz et al. 2021; Wolf et al., 2018). 

To identify intact and near-intact large mammal assemblages, data from the IUCN Red List (IUCN, 2019) were used for current species ranges and data from Faurby & Svenning (2015) for historical ranges. Faurby & Svenning (2015) modeled mammal species ranges as the ranges would have been today in the absence of human influence. AD 1500 was chosen as the cutoff for historical ranges, following the rationale provided by Morrison et al. (2007). Briefly, this marks a globally synchronous period after which there were the most profound anthropogenic changes to Earth’s terrestrial area; it is the same demarcation used by the IUCN Red List as the cutoff for examining ‘recent’ extinctions. Moreover, all except six of the large mammal species (>15 kg) present in AD 1500 are still extant and have opportunities for in situ conservation. The six extinct species (EX) were removed from the analysis since there is no opportunity for their restoration; one species listed as extinct in the wild (EW; Elapharus davidianus) was retained.

Intact and near-intact large mammal assemblages were identified by first converting current species range polygons into ~100 km² rasters to match the historical species range data. To ensure small ranges were not missed, any grid cells that overlapped the polygon were included. These raster layers of current and historical species ranges were then downscaled and refined to 10 km² using the Land Cover product published by the European Space Agency Climate Change Initiative (ESA CCI) (Bontemps et al. 2013). To perform this step, the ESA CCI landcover was first resampled to 10 km² and each landcover class was linked to habitat preferences from the IUCN Red List (IUCN, 2019). For example, the ESA CCI landcover class ‘Tree cover, broadleaved, evergreen, closed to open (>15%)’ was linked to the IUCN habitat preference of ‘Forest’. The individual 100 km² current and historical species range rasters were then downscaled by resampling them to 10 km² and removing grid cells that did not match the habitat preferences linked to the matching 10km² landcover data. This process removes unsuitable habitat form each species’ range, limiting errors of commission. Historical range maps were also downscaled using current landcover to ensure areas no longer suitable for a species were not included as potential restoration areas. Thus, the spatial scale at which decisions were made about presence or absence of a particular large mammal species was at a fine-grained scale of 10 km2. Only later in the analysis was the ecoregion boundary coverage overlayed on the relevant grid cells.  

To identify areas where a species is no longer present but might have occurred in the absence of human influence, each species' current refined range was then subtracted from its refined historical range. These areas of loss for each species were then combined, yielding a raster of the number of species missing per grid cell. Using this raster, places having all species present were identified as intact large mammal assemblage areas. Those with 1–3 missing species were identified as near-intact large mammal assemblages as. This range was chosen because the objective was to take a pragmatic approach to identify places for restoration to a complete assemblage. This decision serves as a reasonable starting place for operationally defining the term ‘near-intact’. The rationale wa if restoration of the assemblage will require targeted, species-based reintroductions, major effort required will be needed for each species. Conversely, areas with more than three species missing are more likely to be degraded, isolated, or have significant hunting pressures and make near-term restoration less feasible. Grid cell output was summarized within ecoregion boundaries (Dinerstein et al., 2017; Table S2).