Africa is renowned for the current abundance and diversity of its large mammals. The aim of this study was to assess distinctions evident in the functional composition of continental large herbivore faunas during the late Pleistocene before extinctions depleted the megafauna outside Africa.
The African large herbivore fauna was compared with that formerly inhabiting South America, Australia, North America, Eurasia and tropical Asia during the late Pleistocene.
Pleistocene faunas were reconstructed from the literature in terms of their relative body size composition, grazer/browser contributions and taxonomic representations, omitting forest and island species.
Although the three southern continents were closely similar in the overall species richness of large herbivores that they supported during the late Pleistocene, South America had a predominance of very large herbivores, while most of Australia's mammalian herbivores were relatively small and those of Africa were intermediate. Africa had many more grazers, especially in the size range 100–1000 kg, than other continents. The South American pattern resembled that in North America and Eurasia, while Africa and Australia diverged in different ways.
Neither the total extent of savannas in each continent nor the morphological features enabling bovid radiation seemed adequate on their own to explain the greater richness of macrograzers in Africa. Africa is characterized by the widespread occurrence of arid/eutrophic savannas, which are unrepresented in other continents. The prevalence of savanna is partly attributable to the high elevation of interior eastern and southern Africa, associated with relatively low rainfall, and to the comparatively high soil fertility, related to volcanic influences. This promoted an abundance and diversity of medium-sized grazing ruminants unrivalled elsewhere. Indigenous grasses in South America and Australia are less well adapted to withstand severe grazing than the African grasses introduced to support livestock. The locally high abundance of African ungulates presented conditions that facilitated the adaptive transition by early hominins from plant-gatherers to meat-scavengers.
The abundance and diversity of large herbivores extant in Africa contrast starkly with those of impoverished faunas elsewhere in the world, including in other southern continents spanning a similar latitudinal range. A visitor to South America may glimpse only an occasional tapir, a couple of camelids, a few small deer, and some giant rodents perched around the edges of lakes. In Australia, mammalian herbivores are mostly variants on the kangaroo theme. What are the differences in the ecology of these southern landmasses that have brought about the enormous contrasts in the indigenous large mammals that they contain? To explore intrinsic factors influencing the continental large herbivore faunas, one must reconstruct the species assemblages that existed within each continent prior to the extinction wave in the late Pleistocene. Australia and South America, along with the northern continents, lost a large proportion of their megafauna around this time, a phenomenon in which the arrival of human hunters is believed to have played a central role (Martin & Klein, 1984; Flannery, 1994; Barnosky & Lindsey, 2010).
The earliest radiation of grazing herbivores, primarily among the Notoungulata, took place in South America c. 30 Ma, when temperature conditions cooled markedly (Webb, 1983; MacFadden, 1997, 2000). North America was the scene of the next grazing radiation during the late Miocene, c. 15 Ma, when equids were represented by 13 genera, including both grazers and browsers. During these times all grasses had the C3 photosynthesis pathway. Grazer diversity in both of these continents declined when grasslands dominated by species with C4 photosynthesis became prevalent between 8 and 5 Ma at tropical and subtropical latitudes (Cerling et al., 1997; Edwards et al., 2010). C4 grasses have thickened bundle sheath cells, a feature that reduces their digestibility and is associated with their adaptations to high temperatures coupled with low atmospheric CO2 levels (Caswell et al., 1973; Ehleringer & Monson, 1993). Although the three southern continents exhibited an almost identical species richness of large herbivores in the late Pleistocene, major differences existed in the functional composition of their large herbivore assemblages in terms of body mass distribution and trophic guilds represented. Rather than showing convergent evolution in guilds dependent on basically similar food resources, the three continents diverged in functional features of their large herbivore faunas. What environmental features underlie this distinction, and how might they have influenced the ecological adaptations of the large herbivores? Furthermore, what are the implications for why Africa served uniquely as the cradle of humankind?
The effects of differences in herbivore activity on the ecology of the tropical rain forests of South America and Africa were reviewed by Cristoffer & Peres (2003), but without consideration of the Pleistocene faunas. De Vivo & Carmignotto (2004) suggested that the shrinkage in extent of open vegetation types in the early Holocene contributed to the large mammal extinctions in South America, but this vegetation shift could have been partly a consequence of megaherbivore extinctions through other causes, as suggested by Owen-Smith (1987). Croft (2001) considered rank-ordered body size distributions of Cenozoic mammalian herbivores in South America, but drew no comparisons with other continents. I will broaden these treatments to consider the distinctions that existed among the large herbivore faunas of the three southern continents during the late Pleistocene, between them and the two northern continents, and also with tropical Asia.
Functional distinctions among modern mammalian herbivores are underlain by differences in body size, which affects metabolic rate and hence dietary nutrient requirements (Bell, 1971; Jarman, 1974; Owen-Smith, 1988). Furthermore, modifications of the dentition, mouth and gastro-intestinal tract influence the use of particular plant types or parts (Hofmann & Stewart, 1972; Owen-Smith, 1982; Hofmann, 1989). Because the body masses inferred for extinct species are unreliable, and commonly inflated, in the analysis below I will simply distinguish three broad functional categories, as follows.
‘Megaherbivores’ attain over 1000 kg in adult body mass, enabling them to transform vegetation physiognomy and habitat composition extensively across landscapes by plucking, mowing, breaking or otherwise disrupting structural plant tissues.
‘Macroherbivores’ are within the size range 100–1000 kg, meaning that they can crop a substantial proportion of annual vegetation production, especially when aggregated in large herds.
‘Mesoherbivores’ are between 10 and 100 kg in body mass, and can selectively nibble the more nutritious plant parts in localized habitats.
Within these size ranges a basic trophic distinction exists between grazers, which draw their diets largely or entirely from easily accessible and chemically undefended but relatively fibrous grasses and sedges, and browsers, which consume a varying mix of woody plant foliage and dicotyledonous herbs or forbs (Hofmann & Stewart, 1972; Owen-Smith, 1997). Mixed feeders, consuming both plant types, can be subdivided among those dependent primarily on grasses and those reliant basically on browse over the seasonal cycle. Some more omnivorous herbivores include a substantial component of fruits, roots and tubers in their diets, supplemented sometimes by small amounts of animal material. They are exemplified by various forms of pigs, primates and rodents, and in Africa also by forest-dwelling duikers.
Materials and Methods
For reconstructing the large herbivore genera and the species that existed worldwide during the late Pleistocene, Anderson's (1984) ‘bestiary’ was used as a starting point and updated from subsequent compilations. The late Pleistocene extends from about 100 to 10 ka and is represented by Rancholabrean faunas in North America and Lujanian faunas in South America. Macdonald (1984) was used as the reference for extant species, thereby avoiding the splitting of species that has tended to be promoted recently when using genetic distinctions rather than morphology. The Australian Pleistocene fauna was updated from Murray (1991) and Flannery (1994). Cione et al. (2009) was helpful for updating the South American Pleistocene fauna, but I merged some of the species they had split using other sources in order to be consistent with more conservative lists for other continents. Extinct tropical Asian species were extracted from Louys et al. (2007) and Patnaik et al. (2008). Body size estimates for extant species were derived from Owen-Smith (1988) augmented by Macdonald (1984), and those for extinct species from Smith et al. (2003).
Herbivores were grouped into the functional body size ranges proposed above, based on the maximum body mass of adult females, allowing for the tendency for surmised body masses to be exaggerated. Arboreal herbivores associated with the forest biome, including apes, monkeys and tree sloths, were excluded. For dietary contrasts among species lists, I used a simple two-class distinction between grazers and browsers plus omnivores. For modern herbivores, mixed feeders with diets consisting predominantly of grass were included among the grazers. Mixed feeders with diets made up largely or mostly of woody browse were classed as browsers. Omnivorous pigs, peccaries and porcupines were grouped along with the browsers for statistical assessments. A dilemma arose with respect to the duikers (Cephalophinae), represented by 13 species over 10 kg in body mass in Africa, which although terrestrial are associated mostly with forested habitats. Their radiation in the forest biome is unparalleled on other continents, and hence distorted comparisons based primarily on species occupying open woodland, savanna or grassland conditions. Hence I excluded the forest duikers from the African fauna when making statistical comparisons.
The dietary dependence of extinct herbivores is commonly inferred from the degree of hypsodonty exhibited by the molar dentition, indexed by the crown height relative to width (Janis, 2008). Grazers typically have high-crowned teeth, interpreted as an adaptation to deal with the high fibre contents and silica bodies typical of grasses. However, some mixed feeders dependent on low-level browse in dry and hence open habitats also exhibit highly hypsodont dentition, indicating that this adaptation is primarily to cope with tooth wear resulting from the grit layer on low-growing plants, whether shrubs or grasses (Mendoza & Palmqvist, 2007; Damuth & Janis, 2011). Hence while grazers require high-crowned teeth, with only rare exceptions, the degree of hypsodonty is not a definitive indication of a grass-dominated diet. Dietary interpretations have become augmented by assessments of the stable carbon isotope proportions in the dental enamel layer. There are clear distinctions between C4 grasses and grasses plus woody plants and forbs exhibiting mostly C3 photosynthesis, and hence in the tissues of animals consuming these plant types (Cerling & Harris, 1999). This measure is a reliable dietary indicator in tropical and subtropical latitudes where C4 grasses are overwhelmingly predominant, but becomes less meaningful at latitudes beyond about 35o (N and S) and at higher elevations, where C3 grasses become prevalent. A further obstacle arises for the South American fauna because species within the order Xenarthra lack dental enamel. Further dietary clues may come from the width of the mouth (a wide mouth suggests grazing on low-growing vegetation) or through detailed detective work on patterns of macrowear and microwear shown on the teeth (e.g. Croft & Weinstein, 2008; Townsend & Croft, 2008). For my categorization of extinct species, I had to rely on inferences drawn in publications. Species have been classified as primarily adapted for grazing if in some part of their distribution range they exhibit a diet based almost entirely on C4 grasses (from stable isotope analyses), or if they deviate from closely related species in morphological features interpreted as adaptations for grazing, such as the degree of hypsodonty or oral dimensions.
In assigning species to continents, I excluded island endemics inhabiting, for example, the West Indies and Indonesia. The Mediterranean region of North Africa was ignored because it has a largely European fauna. The contiguous landmass of Eurasia was subdivided between (1) the Palaearctic region, comprising Europe plus Siberia and northern China, and (2) tropical and subtropical Asia, from India to southern China, including the Himalayan region.
Statistical analyses comparing proportional species distributions within body mass and trophic categories among continents were carried out by log-linear analysis in Systat 11.0 (http://www.systat.com). Whether significant differences existed between continents, or between grazers and browsers, was established by eliminating the interactions of these factors from models and evaluating the resultant change in the maximum-likelihood estimate of chi-square, also termed the G-statistic. The particular interactions contributing mainly to the overall difference were established by examination of the standardized coefficients (‘lambdas’) divided by standard errors in particular rows and columns of the tabulated interactions. The resulting values approximate z-scores, so that values exceeding 2.0 can be interpreted as roughly significant at P <0.05, and values between 1.7 and 2.0 as significant at P <0.10. The data base used for these analyses is appended as Appendix S1 (see Supporting Information).
The overall faunal richness in large mammalian herbivore species supported during the late Pleistocene was similar for the three southern continents: 77 species in South America, 73 species in Africa (excluding the forest duikers), and 55 species in Australia. As might be expected from their higher latitudinal locations, the two northern continents contained fewer herbivore species: 50 species in North America, and 31 species in Eurasia. The total of 61 species recorded for tropical Asia may underestimate the full complement of species because of the paucity of fossil deposits representing the late Pleistocene there. Generic richness followed a somewhat different pattern: 48 genera in South America, 43 in Africa, 38 in North America and 34 in tropical Asia, with both Australia (24 genera) and Eurasia (24 genera) depauperate.
The body size distributions of large herbivore species in the three southern continents differed significantly (G =39.9, d.f. = 4, P ≪ 0.001). Australia exhibited a preponderance of mesoherbivores and just one megaherbivore, while most of the South American species exceeded 100 kg in body mass, including 16 species in the megaherbivore range (Fig. 1a). These three continents also differed significantly in grazer/browser representation (G =11.0, d.f. = 2, P =0.004). Africa had a much higher proportion of grazers (56% of species) than the other two continents (36% of species for South America and 27% for Australia). Africa exhibited especially an abundance of macrograzers, and South America fewer mesograzers and more megagrazers than in Africa or Australia (Fig. 1b).
When the comparison was extended to encompass all six continents, the interaction between the body mass distribution and the continent (G =46.2, d.f. = 10, P ≪ 0.001) and between the body mass distribution and the grazer/browser distribution (G =16.2, d.f. = 5, P = 0.006) both remained highly significant. A significant interaction between trophic category and body mass (G =10.1, d.f. = 2, P =0.006) indicated that relatively more grazers fell within the macro- and megaherbivore categories, while browsers tended to be concentrated relatively more in the mesoherbivore category. Like South America, the two northern continents tended to have proportionately fewer mesoherbivores and more megaherbivores than the global pattern. When South America alone was compared with the two Palaearctic continents plus tropical Asia, there was still a significant difference in the body mass distribution of herbivores (G =16.1, d.f. = 6, P =0.013), owing to the lower proportion of mesoherbivores in South America (Fig. 1c), while the grazer/browser representation was very similar in this case (G =0.15, d.f. = 3, P =0.985). Hence Africa and Australia were jointly the most discrepant continents through their greater representation of herbivores in smaller size classes. With Africa excluded, Australia differed from all of the remaining continents in having proportionately more mesoherbivores (standardized deviate = 3.45) and fewer megaherbivores (standardized deviate = −2.14), but with no difference in grazer/browser representation (standardized deviate = −0.24). With Australia excluded, Africa likewise showed comparatively more mesoherbivores (standardized deviate = 2.94), while South America had an exceptionally high proportion of megaherbivores (standardized deviate = 2.32). However, Africa supported a much higher proportion of grazers than any other continent (standardized deviate = 3.89). In both northern continents plus South America, grazers tended to predominate among the largest species (14/28 megaherbivores, and 31/78 species within the macroherbivore size range, but only 10/52 species among the mesoherbivores). A similar pattern prevailed in tropical Asia, except for the lack of very large grazers (2/7 megaherbivores, 8/21 macroherbivores and 7/33 mesoherbivores). In Africa, relatively more grazers occurred in the smaller size classes: 2/4 megaherbivores, 22/31 macroherbivores, and 17/38 mesoherbivores. In Australia, grazers also tended to be predominantly small: 3/14 species over 100 kg, and 12/40 species under 100 kg.
Among the South American species classified as grazers, the gomphotheres Cuvieronius hyodon and Stegomastodon waringi, various equids and both toxodontid species all showed quite variable C4/C3 plant proportions in different regions of their distribution ranges, indicating some dietary flexibility (Sánchez et al., 2004; MacFadden, 2005; Prado et al., 2011). Although C. hyodon exhibited relatively low-crowned dentition, carbon isotope analysis indicated that its diet in Bolivia was drawn largely from C4 grasses (MacFadden & Shockey, 1997). The even larger S. waringi likewise showed variable C3/C4 plant proportions, despite exhibiting the extreme hypsodonty typical of a grazer (Sánchez et al., 2004). Equids in the genus Hippidion showed higher C3 plant proportions in their diets than Equus spp., but nevertheless had equally high-crowned dentition, indicating latitudinal or elevational differences in where they grazed (Prado et al., 2011).
Interpreting the diets of various ground sloths, armadillo-like pampatheres and similarly armoured glyptodonts is problematic because of the lack of enamel in their ever-growing (hypseledont) molars. Fossil dung indicates that the diet of certain mylodontid ground sloths consisted mostly of grasses and sedges (Markgraf, 1985), but how much they depended on aboveground versus underground grass parts is unclear, as they were evidently adapted for digging (Bargo et al., 2000). The wide-mouthed Glossotherium robustum and Lestodon armatus seem adapted for grazing, while the narrow-mouthed Mylodon darwini and Scelidotherium leptocephalum appear to have been mixed feeders (Bargo & Vizcaino, 2008). The squat form of pampatheres and glyptodonts means that they must have been low-level feeders, but the extent to which their ever-growing molars were adapted to handle abrasive grasses versus other types of plants remains uncertain (De Iuliis et al., 2000).
In Australia, the species classified as grazers were mostly wombats (Vombatidae), the biggest attaining a body mass of about 200 kg. These wombats were clearly adapted for digging, again raising questions about whether their relatively high-crowned teeth were adapted for grazing or for handling the soil particles on other types of plants ingested. The only marsupial marginally within the megaherbivore size range, Diprotodon optatum, was a browser, like other species in the Diprotodontidae, as also were extinct giant kangaroos (body mass up to 250 kg) in the genera Procoptodon, Simosthenurus and Sthenurus.
The African herbivore fauna is distinguished particularly by the abundance of meso- and macrograzers in the family Bovidae. Furthermore, several species of equids are also grazers, while two of the four surviving megaherbivores are strict grazers: the white rhinoceros (Ceratotherium simum) and Hippopotamus amphibius. African elephant (Loxodonta africana) diets consisted largely of C4 grasses during the Pleistocene, although based on their dentition and modern diets they are generally classified as mixed feeders favouring woody browse during the dry season (Cerling et al., 1999). Africa formerly supported two additional grazing megaherbivores, the elephant Elephas recki, and a second hippopotamus Hippopotamus gorgops. They both became extinct some time before 100 ka, although there is evidence that E. recki might have persisted into the late Pleistocene in North Africa (Todd, 2005). One grazing equid (Equus capensis) and six grazing bovids (Megalotragus kattwinkeli, Pelorovis antiquus, Parmularius altidens, Damaliscus niro, Antidorcas bondi and Antidorcas australis) became extinct around the end of the Pleistocene. Hence the grazers surviving in Africa are less diverse than those extant there earlier during the Pleistocene.
Megaherbivores were represented by more species in Pleistocene Europe and North America than exist in Africa today. In Eurasia grazing mammoth (Mammuthus primigenius), woolly rhinoceros (Coelodonta antiquitatis) and Hippopotamus sp. coexisted with the browsing straight-tusked elephant Palaeoloxodon, and two browsing rhinoceroses (Dicerorhinus kirchbergensis and Dicerorhinus hemitoechus). In North America, three species of mammoth are recognized (Mammuthus primigenius, Mammuthus columbi and Mammuthus jeffersonii, and the ground sloth Glossotherium harlani is also regarded as a grazer, while the mastodont (Mammut americanum) and the enormous sloth Eremotherium mirabile were browsers. The macrograzers in North America included Bison (one extinct species, one surviving species), several equids, apparently a camelid (Camelops hesternus), and two musk oxen (Euceratherium and Bootherium spp.). Mountain sheep (Ovis spp.), also grazers, persist today. Eurasia had the macrograzers Bison, Bos and Equus spp., plus wild sheep (Ovis spp.) in the meso category, while fallow deer (Dama dama) are also partly grazers. The tropical Asian fauna still includes as grazers the Indian one-horned rhinoceros (Rhinoceros unicornis), Asian wild ass (Equus hemionus), various wild cattle (Bos and Bubalus spp.), blackbuck (Antilope cervicapra), swamp deer (Cervus duvauceli), spotted deer (Axis axis) and wild sheep. The chiru (Pantholops hodgsonii), which forms large concentrations in Tibetan uplands, is a mixed feeder, but with grasses and sedges tending to predominate in its diet (Schaller, 1998). Hippos were formerly present in India.
The three southern continents supported comparable species richnesses of large mammalian herbivores during the late Pleistocene, but were disparate in the functional distinctions exemplified in this faunal component. Australia exhibited a predominance of comparatively small species (< 100 kg), while South America had a high proportion of very large species, including many more forms over 1000 kg in body mass than exist in Africa today or were present there in the Pleistocene. In Africa, 70% of Pleistocene herbivores in the 100–1000 kg range were classified as grazers, compared with 40% in South America and less than a quarter in Australia. The body mass and grazer/browser distributions of South American herbivores more closely resembled the patterns in northern continents than those shown in the other southern continents. Hence Africa and Australia are the global outliers in functional features of their Pleistocene large herbivore faunas.
The rich diversity of grazing herbivores typifying Africa has several possible explanations, drawing on hypotheses that have been advanced to explain the higher species richness of the tropics compared with temperate latitudes:
The greater extent of grassland habitats in Africa compared with other continents offered more opportunities for isolation and hence speciation (Rosenzweig, 1995).
The radiation of grazing bovids emanated from their phylogenetic pre-adaptations in oral and digestive morphology to become specialized on different grassland structures, as suggested for cichlid fishes in African great lakes (Liem, 1974).
The relatively fertile grasslands occurring on volcanically enriched soils prevalent in Africa supported sufficiently high biomass densities of grazing ungulates to allow viable populations to fragment and hence speciate via finer niche differentiation (Vrba, 1992).
Because arboreal forms and the forest duikers have been excluded, the herbivores considered for this analysis largely occupy savanna, grassland or open woodland habitats. The savanna biome, broadly defined by a grass layer comprising species having the C4 photosynthesis pathway, occupies a greater extent in Africa than in Australia or South America (Lehmann et al., 2011). Its basic defining feature is a seasonal deficiency in rainfall, which both suppresses tree growth and promotes frequent fires, thereby maintaining an open woody plant canopy. Savanna vegetation extends into higher-rainfall regions in South America than in the other two southern continents, with tree growth becoming limited primarily by low soil fertility (Medina & Silva, 1991). In Africa and Australia, savanna occurs widely under conditions of relatively low rainfall, where soils are thus less leached of nutrients and fires less frequent. Climatic conditions of low rainfall combined with high seasonality are not represented in South America (Lehmann et al., 2011). The pampas grasslands extending from southern Brazil through Argentina are temperate and dominated by C3 grasses. Their topographic uniformity might also help to explain why they did not contribute more to the diversification of grazers in South America. Hence the more extensive savanna grasslands in Africa would certainly have provided greater scope for a diversity of grazers to originate. Moreover, the predominance of C4 grasses extends to the southern limit of the savanna biome in South Africa, with C3 grasses becoming prevalent only on the high-elevation plateau of Lesotho. Nevertheless, the distribution of open vegetation formations would have been wider in all continents during the hypothermal conditions that prevailed for most of the Pleistocene than it is today (de Vivo & Carmignotto, 2004).
The dazzling radiation of cichlid fishes in Lake Malawi and Lake Tanganyika has been ascribed to features of their jaw architecture, which enable subtle modifications for different trophic niches (Liem, 1974). Similarly, features of muzzle width and breadth and angle of insertion of the incisor arcade (Owen-Smith, 1982; Gordon & Illius, 1988), plus variants in the relative size, shape and papillation of the foregut chambers where food is processed, fermented and absorbed (Hofmann, 1989), have been invoked as the features underlying niche separation among grazing bovids (Owen-Smith, 1985; Murray & Illius, 1996). Although adaptations in these morphological features must have facilitated the bovid radiation in Africa, this explanation seems insufficient on its own to explain the continental contrasts in grazer richness. Cervids, which originated in Eurasia and then spread from North America into South America, also have ruminant digestion and similar dental structure to bovids. Why did they not radiate after they entered the New World grasslands? Alternatively, why did the camelids, which evolved in North America, not populate the grazing niche more profusely? Bovids originated in tropical Asia, but failed to develop a grazing radiation there, except among wild cattle. Clear dietary distinctions between grazing and browsing, supported by corresponding morphological adaptations, seem to be manifested only among African ungulates (Owen-Smith, 1997).
Geomorphology and soil fertility
Within Africa, reductions in both the amount and seasonality of rainfall are largely an outcome of the high elevation of its eastern and southern interior (Partridge et al., 1995). This restricts leaching and promotes bedrock influences on soils. Notable especially are regions of comparatively high soil fertility associated with volcanic influences, not only in the Rift Valley region of eastern Africa, but also in the south through widespread flood basalts and dolerite intrusions (Bell, 1982). The arid/eutrophic variant of the savanna biome, characterized by trees formerly grouped in the genus Acacia (Huntley, 1982), is widespread in Africa but has no counterpart on other continents.
In South America, soils throughout most of the Brazilian cerrados and Venezuelan llanos are intensely leached to the extent of aluminium toxicity, a reflection of high rainfall on sandy soil substrates, coupled with extensive water-logging in the llanos (Cole, 1986; Medina & Silva, 1991). Despite prevalently low rainfall, Australian soils are generally ancient and thus mostly leached of the clay minerals that retain nutrients, partly as a consequence of the low continental relief (Cole, 1986). In both Australia and South America, African grasses have been introduced to help sustain domestic cattle and sheep, because native grasses seem less capable of sustaining high grazing impacts (Fisher et al., 1996). Indigenous South American grasses are generally much less digestible than the African species introduced, probably as a consequence of the low nutrient conditions, although nitrogen contents in the grass tissues do not differ. Australian grasses, in contrast, are quite digestible but show low protein levels, perhaps as a reflection of the leaf toughening needed to withstand the predominant aridity.
In order to thrive as grazers in South America, mammalian herbivores evidently needed (1) to be very large, and hence able to tolerate less digestible forage because of their slower metabolism, (2) to have hindgut digestion able to handle fibrous forage, as exemplified by the equids, or (3) to be able to dig for underground grass parts, as the mylodont ground sloths apparently did. The only grazers with foregut fermentation are the vicuna (Vicugna vicugna), restricted to fertile valley grasslands in the high Andes, and the pampas deer (Ozotoceros bezoarticus), which selects strongly for green leaves (Jackson & Giulietti, 1988). In Australia, the grazing kangaroos found there today are likewise ‘nibblers’ on higher-quality grass parts, and are subject to massive die-offs when rains fail (Caughley et al., 1985). Grazing wombats had the ability to seek out underground plant parts when aboveground material was exhausted.
The distinctive feature of the African savannas is the diversity of grazing ruminants, adapted in various ways in their body size and oral and digestive anatomy to handle different grassland types and structures. This radiation was promoted by the vicariant splitting and joining of grasslands on the African plateaus in response to climatic oscillations, as described by Vrba (1992). Especially relevant is the biomass density that some of these species can attain, rivalling that of domestic livestock that are managed by providing water and sometimes by augmenting dry-season forage (Oesterheld et al., 1992; Fritz & Duncan, 1994). The highest biomass densities of macrograzers are exhibited in regions with volcanically enriched soils, as exemplified in the Rift Valley region of eastern Africa (Bell, 1982; Owen-Smith, 1999). Comparably high ungulate densities formerly existed in the eastern Free State region of South Africa, with black wildebeest (Connochaetes gnou) and blesbok (Damaliscus dorcas) predominating, before settlers with guns mostly eliminated these species (Pringle, 1982). This indicates that a high proportion of the grass available in these arid/eutrophic savannas is edible and digestible on a regional scale, and able to sustain such grazing pressures (McNaughton, 1984). In moist/dystrophic savannas in Africa, megaherbivores in the form of elephants make up a larger proportion of total herbivore biomass than elsewhere, and grazing bovid abundance is greatly reduced, although certain species are typically represented (Owen-Smith, 1988).
Grazing ruminants can achieve regional densities exceeding 50 animals km−2, as manifested by wildebeest (Connochaetes taurinus) in the Serengeti ecosystem (Sinclair & Arcese, 1995). Kob (Kobus kob) and lechwe (Kobus leche and Kobus megaceros; Sayer & van Lavieren, 1975; Fryxell & Sinclair, 1988) also build up very high local densities in floodplain grasslands. Zebras are widely distributed in African savannas, but do not attain such high local densities, owing to their less efficient digestion. This feature would also have limited the abundance of the equids that were the predominant macrograzers in South America during the Pleistocene. The highest densities attained by grazing red kangaroos (Macropus rufus) in Australia are around 12 animals km−2 (Caughley et al., 1980), similar to that of zebras in the Serengeti.
Another striking faunal feature distinguishes Africa from all other continents: it formed the cradle for the early transition from ape to ape-man, and later to both early and modern humans. While the presence of primate ancestors was a necessary requirement, it was not sufficient. The crucial step from woodland ape to savanna ape-man is widely believed to have rested on the ability of the early hominins to supplement their previously largely plant-based diet with animal food, most probably derived by scavenging from the carcass remains of animals killed by carnivores (Blumenschine, 1987; Klein, 1999; Sponheimer & Lee-Thorp, 1999). This supplement would have been especially important during the intensifying dry season when plant food became less readily available, around the time of the Pliocene/Pleistocene transition c. 2.5 Ma (Owen-Smith, 1999; Cerling et al., 2011). The success of this strategy required a reliable supply of such animal food, and this depended on the larger ungulates being sufficiently abundant to support the production rate of carcasses needed. Judged from their Pleistocene faunas, neither South America nor Australia, nor even the moist savannas of West and Central Africa (Fritz, 1997), could have met this need. Megaherbivores contribute little, despite their high biomass density, because of their low mortality rates and unmanageably large carcasses. The ecological features that promoted the Pliocene radiation of grazing herbivores in Africa (Vrba, 1985, 1992) provided the necessary conditions to enable the transition of a hominin lineage from ape to ape-man. This led ultimately to modern humans with the hunting skills to eliminate much of the large herbivore faunas they depended on when they later spread across the globe.
The combination of features unique to Africa that enabled this crucial evolutionary transition were: (1) the high elevation of much of interior Africa, producing rain shadows that reduce rainfall and hence leaching, and also exposing soils to bedrock influences on the eroding land surfaces; (2) the wide prevalence of volcanic influences on soil fertility, especially in the Rift Valley region of eastern Africa, but also in the south through flood basalts and dolerite intrusions; (3) the high spatial heterogeneity on account of the topographic and edaphic diversity, promoting the local specialization of grazing ruminants; and (4) the exceptionally high abundance levels that these large grazers can attain where suitable conditions prevail. It is only in Africa that primate ancestors could have undergone the crucial evolutionary transition from ape to ape-man.
This manuscript was derived from my presentation at the IV Southern Connections Conference held in Cape Town in 2004. Its publication was delayed pending sufficiently detailed information on the South American fauna and environments. I am grateful to Sergio Lambertucci for helping augment my knowledge of the literature pertaining to South America. The manuscript was substantially improved by critical comments from Sally Archibald, Mikael Fortelius and two anonymous referees.
Norman Owen-Smith is Emeritus Research Professor at the School of Animal, Plant and Environmental Sciences at the University of the Witwatersrand. His main research interests are the ecology of large mammalian herbivores and their interactions with vegetation.