Large mammals are fundamental elements in many ecosystems. Large carnivores frequently shape the number, distribution, and behavior of their prey (Berger et al. 2001b; Sinclair et al. 2003; Terborgh 1988; Terborgh et al. 2001). Large herbivores function as ecological engineers by changing the structure and species composition of surrounding vegetation (Dinerstein 2003; Owen-Smith 1988). Furthermore, both sets of mammals profoundly influence the environment beyond direct species interactions, such as through cascading trophic effects (Berger et al. 2001a; Côté et al. 2004; Crooks and Soulé 1999).

Today, the ranges of individual species of large mammals have been reduced greatly because of human activities, primarily through habitat alteration and direct exploitation or persecution (Ceballos and Ehrlich 2002; Sechrest 2003). Large species are particularly prone to local extirpation because they are differentially hunted for the burgeoning trade in wild meat, controlled as competitors, or otherwise persecuted (Allen et al. 1999; Cardillo et al. 2004; Milner-Gulland et al. 2003; Orians et al. 1997). Large mammals also are sensitive to habitat fragmentation that isolates populations (Woodroffe and Ginsberg 1998). Indeed, a full 39% of these species are considered threatened with extinction compared with 24% for mammals as a whole (the World Conservation Union—IUCN 2005a). Substantial range contractions also have occurred among species whose global conservation status is assessed as Least Concern, such as the wolf (Canis lupus—IUCN 2005a). The result is that there are few regions that retain their full complement of native large mammals. Our objectives were to indicate where “historical” human impacts have occurred, and more importantly to show where the remaining intact large mammal assemblages are found, how they are distributed, and their level of protection.

To identify areas still retaining large mammals, we compared current ranges of the largest 263 terrestrial mammal species (body mass >20 kg; Appendix I) with their distributions in AD 1500 (see “Materials and Methods” for an explanation of our rationale behind this body mass threshold and historical baseline).

Materials and Methods

Mammal taxonomy was based on Wilson and Reeder (1993, 2005), modified in collaboration with IUCN Species Survival Commission (SSC) Specialist Groups. Body size information came from Smith et al. (2003) and Nowak (1999).

A 20-kg body mass was used as the threshold to define a large mammal species because it represents the mass at which carnivores typically switch from invertebrates to larger prey (Carbone et al. 1999). Although other cutoffs have been proposed, and this threshold is perhaps arbitrary with respect to orders of mammals other than Carnivora, this is the sole proposed cutoff based on physiology. We repeated the analyses using a 40-kg threshold, as suggested by Martin and Steadman (1999) and Roberts et al. (2001), and found results to be virtually identical. The lower 20-kg threshold added 70 additional species such as the thylacine (Thylacinus cynocephalus) for Australia, although this species occupied only a small portion of the continent circa 1500. At least 35% of the species in our study (using the 20-kg threshold) have experienced serious contractions (>50%) in range (IUCN 2005a; MacPhee and Flemming 1999), and even the species with small range losses typically have experienced localized population extirpations and reduced abundance.

To determine the starting point or baseline for our analysis, we evaluated 4 themes or rationales for selecting a historical cutoff. The spread of anatomically modern humans was a catastrophic event for many prehistoric large mammals, and there is evidence that humans were complicit in the extinction of many species soon after colonization (Martin and Steadman 1999; but see Grayson and Meltzer 2003). Nonetheless, using different historical cutoffs for each continent has 2 major drawbacks. Accurate data on large mammal distributions reaching back 10,000–70,000 years (corresponding to the respective times when modern humans arrived on each continent) for even a large proportion of the 262 large mammals are lacking, and are complicated by natural range shifts due to climate change, competitors, prey, and other factors. In addition, precisely because of the catastrophic nature of the human impacts at these times, many species were pushed to extinction, leaving few options for current conservation and restoration of those species.

Alternatively, we could use the time of the advent of settled agriculture (∼10,000 years ago), which began the era of large-scale conversion of natural habitats for human use. Unfortunately, tracking the spread of settled agriculture across the planet continent by continent creates significant complications. For example, the Gangetic plain, home to one of the most extensive intact large mammal assemblages in the Indomalayan realm, was only settled by agriculture in AD 1400, yet a deadly strain of malaria kept much of the Terai zone sparsely settled until the early 1950s when the disease was eradicated (Dinerstein 2003). A similar situation existed for diseases in Africa until relatively recently.

Another possible cutoff is the great increase in absolute human population numbers that occurred after World War II at approximately AD 1950–1960. Although most of this growth occurred in developing countries, it is certain that human impacts on the planet have increased significantly since this time. Yet our large mammal species maps do not have the temporal resolution that would allow us to analyze the change before and after AD 1950–1960, and one would expect time lags between the human population explosion and large mammal ranges, which would differ significantly by species.

Finally, exploration by Europeans occurred in earnest in the AD 1400s, but colonization began to increase significantly after approximately AD 1500 and the industrial revolution followed approximately 200 years later. The spread of Europeans and the subsequent industrial revolution mark the start of the most profound anthropogenic changes to the planet. As stated above, our objectives were to suggest where “historical” human impacts have occurred, and to show where the remaining intact large mammal assemblages are found. The 1st objective forced us to map accurate historical large mammal range distributions—and this type of information was restricted to the recent historical period. Both the IUCN Red List (IUCN 2005a) and the Committee on Recently Extinct Organisms (2007) mammal subgroup also use the year AD 1500 as their cutoff for examining “recent” extinctions. Only 7 large mammals have become extinct since AD 1500, providing opportunities for active conservation of the remaining species.

Historical ranges of individual species were gathered from over 500 published and unpublished sources, including IUCN/SSC Action Plans followed by expert consultation. These ranges are the best approximation for the time period and in some cases were reinforced with historical accounts, although in many instances such maps were necessarily reliant on extrapolations based on habitat preferences.

The data on current ranges were gathered as part of the Global Mammal Assessment. The Global Mammal Assessment is in the process of assessing the conservation status of all mammal species. This work is being carried out with extensive collaboration with experts, especially through the existing IUCN/SSC Specialist Groups for mammals. A global land cover classification (Hansen et al. 1998) was digitally applied to all species range maps to remove converted or inappropriate habitat. Individual species maps can be provided (both current and historical) upon request.

The historical range maps and current range maps differ in precision. In our analysis, we were wary of 2 types of errors: species incorrectly identified as present in an intact area in current range maps; and areas that were disqualified from being considered intact because the existing data indicated that 1 or more historically present species was extirpated from the area, but in actuality they were never present there, for instance because of ecological reasons such as inappropriate habitat. The result of such inprecise historical range maps would be to incorrectly disqualify areas from being identified as an intact assemblage. This danger, nonetheless, is reduced for 2 reasons. First, a number of the species that suffered the greatest range contractions are habitat generalists that, in all probability, occupied most of the mapped extent of occurrence; thus, the risk of falsely disqualifying possible assemblages is minimized. Examples include tigers (Panthera tigris), elk (Cervus elaphus), American bison (Bison bison), leopards (Panthera pardus), lions (Panthera leo), and wolves (Canis lupus). Second, we actively sought intact assemblage areas that might have been overlooked because of imprecision in the historical maps, and we consulted extensively with regional experts for evaluation of potential problematic areas.

Range contractions were located and quantified by removing each species' currently known extent of occurrence from its historical range. The areas with intact mammal assemblages were initially mapped as those that were not part of a range contraction for any species. These areas were then subjected to further scrutiny by evaluating the presence of protected areas, proximity to human settlement and agriculture, and – most important – by further consultation with regional experts to ascertain that such intact assemblages were valid.

Despite these efforts there are likely to be errors in the historical maps that affect our estimates of range contraction. Nonetheless, our main emphasis was on locating those areas that still contain a full complement of historic large mammal assemblages, and we are confident that the identified sites are accurate. Thus, the results of our analysis are most robust where they have conservation implications.

Data on level of protection of areas with intact assemblages of large mammals were developed by overlaying the United Nations Environment Programme—World Conservation Monitoring Centre's World Database on Protected Areas (UNEP-WCMC 2005) with the intact assemblage polygons. “Poorly protected” was defined as ≤25% coverage by IUCN I–VI protected areas, “partially protected” indicates between 25% and 75% coverage, and “largely protected” indicates ≥75% coverage.

It would be desirable to quantify the percentage loss of large mammals from those areas without intact large mammal assemblages. Nonetheless, at present the quality of the data does not permit comprehensive estimations of all large mammal range losses at fine scales (sensu Ceballos and Ehrlich 2002). We hope to pursue the larger goal of quantifying losses comprehensively as information concerning current species ranges improves in the coming years.

With few exceptions, our analysis is restricted to polygons larger than 100 km². A number of intact assemblages are made up of more than 1 polygon (especially in island groups such as Arctic Canada or the Philippines).

All the methods followed the guidelines approved by the American Society of Mammalogists (Gannon et al. 2007).

Results

Intact large mammal assemblages occur in 108 distinct areas. The smallest intact assemblage identified is 24-km² Bawean Island in Indonesia. More than 97% of individual polygons are larger than 100 km² and 83% are larger than 500 km². Siberia is the largest area at 6,961,155 km². These areas include 6 extensive wilderness regions (an arctic–northern–eastern Canadian complex, Amazon–Orinoco basins, west-central Africa and the Congo Basin, Siberia, central Australia, and the Himalayas; Fig. 1). Together, the wilderness complexes constitute 82% of the land area retaining assemblages of large mammals. The large portions of Australia supporting a full assemblage represent a unique case. Three native large kangaroos (Macropus fuliginosus, Macropus giganteus, and Macropus rufus) have expanded their ranges with the spread of the livestock industry, including the clearing of land, extirpation of prehistorically introduced dingoes (Canis lupus dingo), and water provision intended for stock (Calaby and Grigg 1989). Because the extinct thylacine (or Tasmanian wolf [T. cynocephalus]) had a restricted range on continental Australia in 1500 (approximately the Flinders Range—Paddle 2000), the loss of this large carnivorous mammal did not exclude the majority of the continent. Paradoxically, more mammals have become extinct in Australia in historical times than any other continent despite the continued presence of a few large-bodied species (Cardillo and Bromham 2001; IUCN 2005a). The extirpated mammals were small-bodied, mostly 0.035–5.5 kg in size (Cardillo and Bromham 2001).

The other 99 sites are inhospitable (e.g., Novaya Zemlya), have naturally impoverished large mammal faunas (e.g., Pacific coast of South America), or are under intensive conservation management (e.g., Kruger National Park, South Africa; Yellowstone National Park, United States; Fig. 1). Altogether, the 108 intact large mammal sites represent approximately 21% of the area formerly occupied by large mammals (Table 1). We say “approximately” because of the imprecision of the historic mammal range maps relative to the intact large mammal areas. Among the biogeographic realms, the proportion of land area retaining intact assemblages varies from 68% in Australasia to only 1% in Indomalaya.

Twelve percent of the total area retaining large mammal assemblages are formally protected (IUCN I–VI—UNEP-WCMC 2005). This percentage is equivalent to the global total of 12% (IUCN 2005b). The degree of protection (IUCN I–VI—UNEP-WCMC 2005) varies markedly among sites in different biogeographic realms, from 9% in the Palearctic to 44% in Indomalaya. The overall percentage with full protection for biodiversity (IUCN I–IV—UNEP-WCMC 2005) is only 8%, and ranges from 6% in the Palearctic to 35% in Indomalaya (Table 2). On an individual basis, just 25% of the intact areas are largely covered (>75%) by protected areas of any type (Table 1). Of course, the presence of protected areas does not guarantee actual protection.

Sites vary greatly in the number of large mammal species they support; for example, the highest are at Hwange and Serengeti-Mara sites in Africa (30 species in each), whereas lower numbers are found in northern Eurasia and Siberia (7 species). Five species-richness classes depict the distribution of intact large mammal diversity around the planet (Fig. 1). Overall, 10 sites in sub-Saharan Africa and 1 site in the Palearctic realm each conserve more than 25 species (Fig. 1). Nearly all of the sites with large numbers of species receive some formal protection (Table 1), and the most species-rich sites are generally largely protected. Full species lists for each site are available in Appendix II.

Twenty species with the largest absolute range contractions eliminated large areas of the planet from inclusion as areas with complete mammal faunas (Table 3). Examples include: American bison (B. bison), wolf (C. lupus), and cougar (Puma concolor) in North America; jaguar (Panthera onca) in South America; lion (P. leo) in a broad swath of North Africa and the Near East; African elephant (Loxodonta africana), giraffe (Giraffa camelopardalis), and African wild dog (Lycaon pictus) in Africa; and horses (Equus caballus) in Eurasia. Combined range contraction of the 20 formerly widespread species represents 72% of the total range once occupied by large mammals.

Range contraction showed some differences by functional groups. Megaherbivores, defined as plant feeders >1,000 kg in body mass (n = 14), which play the most conspicuous role as landscape engineers (Dinerstein 2003; Owen-Smith 1988) have had much greater average range contractions (88% versus 32%; analysis of variance [ANOVA], F = 31.94, d.f. = 1, 226, P < 0.0001) than smaller herbivores (n = 214). The 7 largest obligate carnivores, species that would be expected to exert the most powerful top-down predator effects on prey, had a slightly but not significantly greater average range contraction than other large Carnivora (n = 24; 55% versus 44%; ANOVA, F = 0.64, d.f. = 1, 29, P = 0.43).

Discussion

Intact faunas represent another ecologically based measurement of human impact (Imhoff et al. 2004; Sanderson et al. 2002; Vitousek et al. 1997). They overlap portions of the “Wildest 10%” of the terrestrial Earth described in a recent analysis (“Human Footprint”—Sanderson et al. 2002). Yet, even at a coarse scale, there are some notable differences between our analysis and that of Sanderson et al. (2002). The total area of the planet that still retains large mammal assemblages (27 million km²) is 1.2 times greater than the total area of the “Wildest 10%,” but overlaps only 48% of the “Wildest 10%.” Substantial portions of the Nearctic, Neotropical, and Palearctic regions are sufficiently remote and undisturbed to qualify for inclusion as wilderness, but are missing 1 or more large mammals. Conversely, areas in the Congo Basin, the Amazon Basin, Australia, and portions of Siberia that are not among the “Wildest 10%” still retain their native large mammals despite human activities.

These mismatches are partly explained by historical relationships between humans and large mammals. Although habitat loss is the most important factor in range contractions generally, some species are affected primarily by human persecution. Nonetheless, even large carnivores can persist at relatively high human densities. Linnell et al. (2001) showed that carnivores increased after the introduction of favorable legislation, and that there is no clear relationship between human densities and current carnivore distributions.

The presence of a large mammal species does not imply that population densities today are comparable to what existed in AD 1500 or that the populations are even viable. Furthermore, human-induced mammal extinctions before this time resulted in altered ecosystems throughout the world, particularly in North America, Eurasia (MacPhee and Flemming 1999), and Australia (Cardillo and Bromham 2001; IUCN 2005b), although imprecise knowledge of former species ranges precludes analysis at deeper time periods. Many species no longer play the same ecological roles as before (Soulé et al. 2003), although in some instances extirpation of 1 species may be functionally mitigated by the continued presence of another with a similar niche (Ives and Cardinale 2004).

Areas that contain complete large mammal assemblages merit conservation attention because only 8% of the land area that still retains complete assemblages of large mammals is well protected. Thus, there is a strong need for creation of new reserves in unprotected areas and enhanced efforts to prevent poaching and habitat degradation within current reserves. Further analysis of these areas is required to determine the density of large mammals present and what other, smaller species may be missing. In general, areas retaining a full complement of large mammals are more likely to be ecologically functional than those that are missing 1 or more large mammal species, and the (temporary) loss of other taxa will often matter less to the recovery of an ecological system. Intact large mammal assemblages should be preferentially included in regional conservation portfolios, all else being equal. Modern reserve design methods can incorporate a wide variety of data layers, and we propose that the results of this analysis be another layer to be considered. The weight of these data will depend on the goals of the organizations and agencies involved in the conservation planning. Already, large international conservation organizations have used this layer to prioritize their global actions. Additionally, our analysis reveals that there are 2 general types of intact large mammal assemblages around the world—remote and inhospitable or small and intensively managed—it is critical to make sure that the latter receive adequate support for long-term conservation.

Finally, reintroductions of large mammals to their former range are possible and have been shown to have dramatic positive ecological effects, a prime example being the return of wolves to parts of North America (Ripple and Beschta 2003). To secure and expand areas with a full roster of native megafauna would seem to be at least as important as (and perhaps complementary to) proposed Pleistocene refaunation projects using large mammal surrogates from other continents (Donlan et al. 2005).