11Global Climatic Influence on Cenozoic Land Mammal Faunas

Chronofaunas and Turnover Pulses

Vertebrate paleontologists refer to persistent faunas of essentially uniform taxonomic composition and stable diversity as chronofaunas following Olson (1952). In ecological terms, chronofaunas are thought to represent stable coadapted sets of species, and their detailed histories (under various perturbations) shed light on the theory of how communities or community complexes are structured (Olson, 1983; Webb, 1987). Five chronofaunas account for most of the stable intervals of the North American Cenozoic. They are the Eocene chronofauna (Krause and Maas, 1990); the White River chronofauna (Emry, 1981; Krishtalka et al., 1987); and in the Miocene, the Runningwater, the Sheep Creek, and the Clarendonian chronofaunas (Webb, 1983a; Tedford et al., 1987). Only in the extreme phases of the icehouse world, during the Pliocene and the Pleistocene, did the stability and persistence of mammalian chronofaunas breakdown.

These stable chronofaunas are followed by relatively brief episodes of rapid faunal turnover. Such rapid turnover episodes (RTE) have been recognized and studied during the past decade, following the establishment of relatively precise geochronometry. Several students of mammalian faunas have documented that even during an RTE the correlation between numbers of first appearances and last appearances is maintained, indicating an equilibrium process in which faunal interactions maintain a balance between gains and losses (Gingerich, 1984; Webb, 1989; Stucky, 1990). Gingerich (1984) proposed that the high peak of land mammal extinctions in the late Pleistocene of North America was an equilibrium response compensating for the high number of immigrations in the Late Pliocene and early Pleistocene; the 2-m.y. time lag tended to conceal the correlation. Others have explained RTE patterns in terms of climatic forcing. Webb (1983b) distinguished two types of rapid turnover episodes, the criterion being whether extinction waves or immigration waves peaked first, and suggested that the former (E-type RTE) were probably caused by climatic changes. He also showed that during the Neogene, six land mammal extinction episodes (including the two largest in North America) were correlated with positive oxygen isotope excursions in

the marine record and thus with global climatic episodes (Webb, 1984). When chronofaunal stability prevails, member taxa generally experience stabilizing selection. Only when climatic change perturbs the terrestrial ecosystem sufficiently does it produce an RTE.

Wing and Tiffney (1987) described similar patterns in floristic stability and turnover. They found that "the chief mechanism for drastic reorganization in community structure" was "severe climatic change." A similar view was formally proposed as the turnover-pulse hypothesis by Vrba (1985a, p. 232) in the following terms:

Speciation does not occur unless forced (initiated) by changes in the physical environment. Similarly, forcing by the physical environment is required to produce extinctions and most migration events. Thus, most lineage turnover in the history of life has occurred in pulses, nearly (geologically) synchronous across diverse phylogenies, and in synchrony with changes in the physical environment.

This view underlies our analysis of immigration episodes in land mammal faunas. We attempt to test such terrestrial turnover patterns by comparing them with the isotopic record of Cenozoic climate change, which is derived independently from marine sediments.

Importance of Immigrants

In the following analysis we take the number of immigrant genera in each land mammal age as a key to faunal turnover episodes. Immigrants (first appearances) are recognized on the basis of two criteria: (1) the absence of that genus or closely related genera in prior North American records and (2) the presence of related forms (sister group) earlier in another continental fauna. Mammalian biostratigraphers have increasingly focused on immigrant taxa as the best diagnostic basis for defining land mammal ages, as well as informal subages or zones (Woodburne, 1987). Thus, for both its environmental and its stratigraphic significance, the record of immigrant mammals in North America has been well documented.

Figure 11.1 identifies major episodes of mammalian immigrations into North America during the Tertiary; we designate as first-order immigration episodes those that number nine or more genera. Second-order episodes are those with five to seven genera. These data are more fully explained in the following sections on results. Biostratigraphy and chronostratigraphy follow Woodburne (1987). These data will continue to be refined, by addition of immigrant genera, by extension of their known ranges into earlier intervals, and by refinements in the geochronological framework.

Three land bridges controlled the access of Cenozoic land mammals to North America. In the latest Paleocene and early Eocene the North Atlantic Thulean route between Europe and North America provided the basis for a strong faunal resemblance between these two continents. The next major immigration episode into North America took place in the Late Eocene. The mammals of this episode may have come across the Bering Strait, but more probably they again tracked the Thulean land bridge, which did not founder to great depth until the Oligocene (Rowley and Lottes, 1988). Subsequent episodes of faunal dispersal from the Old World must have crossed via the Bering land bridge. The very high latitude of this route (at the North Pole in the Cretaceous) placed increasingly severe ecological constraints on the immigrants that traversed it. Even as Beringian access shifted to somewhat lower latitudes during the Neogene, climatic regimes became more severe, so that Beringia continued to function as a narrow ecological filter. During the Pliocene (Blancan land mammal age) the Bering land bridge provided passage for temperate woodland species, such as an extinct deer (Bretzia) and an extinct panda (Parailurus), to the Pacific Northwest (Tedford and Gustafson, 1974). By the Pleistocene, however, the Bering filtered out all but steppe and steppe-tundra species, with the exception of humans who by then had developed highly specialized cultural adaptations to life in Beringian steppe-tundra.

A third route for intercontinental access to North America came from the south via the Panama (isthmian) land bridge. By Late Pliocene time it facilitated immigrations from South America into North America, mainly representing species adapted to subtropical savanna (Stehli and Webb, 1985).

In the following section we review the environmental history of mammalian faunas in North America, with an indication of the predominant biome during each chronofauna. We also briefly discuss the immigration (and faunal turnover) episodes that punctuate the chronofaunal succession. A major unresolved problem stems from the different durations of the land mammal ages in current use. The Duchesnean, for example, may span as much as 4 m.y. and the preceding Uintan as long as 6 m.y. Although it might be appropriate to treat immigrations as a rate by making the assumption that immigration events were uniformly distributed through a mammal age, in the present context we make no a priori assumptions about how immigrations may have been distributed within a mammal age subdivision, placing them simply at the stratigraphic level advocated by the best available empirical studies. We follow with the caveat that long intervals are more likely to produce artificially high immigration numbers, and note the need for continued refinement of fossiliferous local sections, especially in the mid-Cenozoic. Before we examine the possible correlations between immigration episodes and global climatic shifts, we briefly review the history of changing biomes in North America.

Recognition of Land Mammal Biomes

Several decades before the evolution of horses was documented in western North America, Kowalevsky (1873), working in Europe, recognized that hypsodonty in hipparionine horses reflected the spread of grasslands in the Miocene. Such reconstructions of environmental history from fossil mammal assemblages rest on two fundamental premises. The first is that the fossil record adequately samples the kinds and numbers of land mammal genera that lived during successive stages of the Cenozoic. The second premise is that samples of once-living mammals, through their morphologies, distributions, and numbers, and also through their sedimentary and taphonomic contexts, bear evidence of their environments. Such relationships are well established between living mammals and present-day environments; for example, the diversity and structure of African ungulate faunas reflect the vegetative cover and rainfall in which they range (Coe et al., 1976; Sinclair and Norton-Griffiths, 1979; Vrba, 1985b; McNaughton and Georgiadis, 1986). Webb (1984) and Janis (1984) independently analyzed the distribution of body size and hypsodonty in fossil ungulate faunas from North America to develop convincing analogues with modern ungulate faunas in Africa. More generally, Andrews et al. (1979) empirically determined the patterns of body size and other gross features that characterize modern mammalian faunas from various terrestrial biomes regardless of the continent on which they occur.

One particular method of faunal inference that has been applied with considerable success in recent years is called the cenogram. A cenogram displays the whole spectrum of estimated body sizes for all species from a relatively complete faunal sample, for comparison with other such cenograms representing known biomes. Legendre (1986) convincingly used this method to show that the land mammal cenogram for the Late Eocene Phosphorites of Quercy in France had the same pattern as a modern equatorial rain forest fauna, notably a predominance of small-bodied (arboreal) forms, and Gingerich (1989) applied the same method with excellent results to still earlier Eocene (Wasatchian) faunas in Wyoming. Other examples are reviewed in Behrensmeyer et al. (1992).

Actualistic studies of taphonomic processes in the deposition of mammal remains provide the critical link between modern and fossil ecosystems (Behrensmeyer and Hill, 1980; Damuth, 1982; Behrensmeyer et al., 1992). Thus, a legitimate train of logic and experiment connects the comparison of modern ecosystems with depositional settings in the recent past. It should be stressed that such relationships must be interpreted cautiously, and that data from ancient mammalian faunas apply most reasonably at the biome level in well-studied parts of the fossil record.

Another largely independent approach involves the study of ecomorphology. For example, the teeth of mammals are among their most important adaptations, and many features of their energetic and often highly specialized life styles depend to a great extent on their masticatory capabilities. The relationship between tooth morphology and ecology is particularly clear among large herbivores that must adaptively reconcile their long lives and high metabolic activities with relatively poor-quality protein resources (often further complicated by high fiber and high toxin contents). Fortunately, the fossil record of herbivore teeth is excellent and provides a clear history in diverse mammalian taxa of progressive change in their capacity to process increasingly coarse fodder. The dental adaptations of various mammalian herbivores are cited frequently in the following discussion as evidence of the prevailing conditions in North American environments. Such studies were well summarized by Scott (1937).

The most obvious such dental adaptation that many herbivore groups develop is hypsodonty (i.e., high-crowned teeth). A hypsodont animal may be defined as one in which several of its teeth (typically the molars but not uncommonly a whole battery of cheek teeth) have the crown height greater than one dimension of the crown wear surface (length, width, or the average of the two). If one imagines a cubic tooth in an unworn condition it would have a hypsodonty index of one and would be on the threshold of attaining hypsodonty. In modern ungulates and rodents it is not uncommon to find teeth with crown heights many times greater than their crown surface dimension, and some groups develop hypselodonty (i.e., ever-growing teeth).

The adaptations of large mammalian herbivores also contribute strongly to shaping their environments. For example, large mixed herds of grazers facilitate the food resources of one another and also help maintain optimum savanna settings, sometimes by grazing and sometimes by fostering fires (Sinclair and Norton-Griffiths, 1979; McNaughton et al., 1988; Owen-Smith, 1988).

With these interpretive tools in hand we scan the history of land mammal faunas and associated floras to envision the major succession of biomes in North America.

Paleocene Chronofauna: Tropical Forest

The earliest mammal faunas of the North American Cenozoic consisted of diverse small to medium-sized arboreal and scansorial forms. Most were frugivorous, omnivorous, or insectivorous; specialized carnivores and herbivores of larger body sizes came somewhat later. Four orders, Multituberculata, Insectivora, Primates, and Condylarthra, held over from the Cretaceous, comprise almost the entire Paleocene fauna. Analysis of virtually complete skeletal remains of the multituberculate genus Ptilodus demonstrated that its arboreal adaptations converged closely on those of a tree squirrel (Jenkins and Krause, 1983). The Condylarthra doubled and redoubled their generic numbers every few million years, and soon accounted for several distinct families including some moderately large herbivores (Van Valen, 1978). The distant forerunners of Carnivora (sometimes placed in their own extinct order, Creodonta) appeared by mid-Paleocene. By Late Paleocene, three relatively rare orders of larger mammals immigrated from Asia: the Pantodonta were partly amphibious molluscivores; the Taeniodonta were precociously hypsodont herbivores; and the Dinocerata were the first horned, browsing, herd animals of the Cenozoic. The Torrejonian mammal age included seven immigrant genera if one combines Torrejonian stages 2 and 3, as registered by Stucky (1990).

Predominant environments of the Paleocene land mammals were multistratal evergreen forests and cypress swamps extending north to at least 70°N latitude (Wolfe, 1985; Wing and Tiffney, 1987). During the Tiffanian (ca. 63 Ma), cooler climates—evidenced by increased percentage of deciduous trees and decreased diversity—produced "dramatically lower species richness and evenness" (Rose, 1981, p. 386), presumably because there were less reliable vegetational resources on a year-round basis. Krause and Maas (1990) introduced an enlarged data set on this interval that may weaken this conclusion.

The cooler climate and lower mammal diversity persisted into the Clarkforkian. Small mammals that had previously predominated became relatively rare, while larger forms such as phenacodont condylarths and carnivorous mammals increased in abundance. More importantly, the beginning of the Clarkforkian registers a first-order immigration episode, including Asiatic origins of the extinct order Tillodontia, the modern order Rodentia, and the pantodont family Coryphodontidae (Rose, 1981; Krause and Maas, 1990). Nine immigrant genera reached North America, if one combines Stucky's (1990) records for units 2 and 3. Although the Clarkforkian records the earliest first-order immigration episode in the history of North American land mammals, it is soon superseded by an even larger episode.

Eocene Chronofauna: Subtropical Forest

The wave of immigrations that began in the Clarkforkian intensified in the earliest Eocene or Wasatchian land mammal age. Rose (1981, p. 379) commented as follows:

The most striking aspect of the Wasatchian assemblages is their domination by several taxa that were unknown or exceedingly rare before the Wasatchian. Most of these taxa appeared for the first time in North America at or near the beginning of the Wasatchian and rapidly became the most common members of the fauna. . . . They include the first known members of the orders Perissodactyla (Hyracotherium) and Artiodactyla (Diacodexis), the primate families Adapidae (Pelycodus) and Omomyidae, and the creodont family Hyaenodontidae.

Land mammal immigrants of the early Wasatchian (see Figure 11.3 and Table 11.2) constitute the largest immigration cohort in the North American record. In Wasatchian 1 alone Stucky (1990) records 10 immigrant genera. Krause and Maas (1990, p. 90) more cautiously acknowledge that "many of the early Wasatchian first appearances appear to be immigrants from other continents." Mammals evidently crossed the North Atlantic freely in both directions via the Thulean bridge. According to Savage and Russell (1983), at least 50% of mammalian genera in the Sparnacian of Europe were shared with the Rocky Mountain region, by far the closest degree of transatlantic resemblance that occurred at any time during the Cenozoic.

Figure 11.3

Land mammal immigrations in the Miocene of North America, including first-order (triangles) as well as second- and third-order (rectangles) episodes, juxtaposed with δ¹⁸O positive excursions as numbered by Miller et al. (1991) based on detailed (more...)

TABLE 11.2

First-Order Immigration Episodes of North American Land Mammal Genera.

The early stages of the Wasatchian are well constrained by stratigraphic studies in the Bighorn Basin of Wyoming. There the base of the earliest Wastachian lies within the reversed magnetic interval below stratigraphic anomaly 24. For this reason, its age lies between 56 and 57 Ma (Butler et al., 1991).

Climatic conditions inferred from faunas and floras of this time suggest that warm equable conditions had returned during the Early Eocene (Rose, 1981; Wolfe, 1985). Primates and other arboreal mammals reach peak richness and abundance by the Middle Eocene, declining thereafter (Stucky, 1990). Rodents and Perissodactyla diversify profoundly, each accounting for about 20% of known mid-Eocene species (Savage and Russell, 1983). Even within the Arctic Circle on Ellesmere Island, diverse arboreal, frugivorous mammals, such as prosimian primates and dermopterans (frugivores distantly related to Asiatic flying foxes), persist along with a subtropical flora (West and Dawson, 1978).

White River Chronofauna: Woodland Savanna

In the continental record for North America a major faunal and environmental shift took place within the Late Eocene (late Duchesnean). Chronometric control on this episode needs further refinement, but it lies between 37 and 42 Ma, and so on an interim basis we call it 40 Ma. For a full discussion of the problem, see Krishtalka et al. (1987).

In the Middle Eocene the first indications of seasonal aridity appeared in the Rocky Mountain region. The great lakes of the Green River region, such as Lake Gosiute, formed extensive evaporites and were encroached on by deeply oxidized redbed deposits. By the Late Eocene, woodland savanna had become the predominant biome in the midcontinent (Webb, 1977; Wing and Tiffney, 1987). The dramatic shift from subtropical forest to predominantly woodland savanna in the Rocky Mountain region soon led to major faunal changes. It is unfortunate that this important turnover interval is not well documented by continuously fossiliferous sedimentary sequences.

The Duchesnean land mammal age is marked by a major faunal turnover episode, including the last appearances of such archaic groups as Condylarthra, Tillodontia, and Dinocerata. More important are the arrivals of the first eubrontothere (Duchesneodus) and several more modern taxa evidently dispersed from Asia (Emry, 1981; Krishtalka et al., 1987). Stucky (1990) recognizes a total of nine genera in Duchesnean 1 and 2. After the Duchesnean the number of browsing herbivore genera rose from 8 to about 40 (Stucky, 1990), and the species numbers of all herbivores cited in Savage and Russell (1983) rose from less than 40 to about 90 during the Eocene-Oligocene transition and remained at this level throughout the Oligocene (Webb, 1989). Emry (1981) labeled this persistent mammalian fauna of the Late Eocene and Oligocene the "White River chronofauna."

The Duchesnean immigration was preceded by a number of other immigration events apparently spread through the Uintan. Unfortunately this interval is about 6 m.y. long, and particular first appearance data are not tightly correlated. We arbitrarily place a second-order episode near the mid-Uintan. The following modern mammal families appear in North America during the Late Eocene: soricid insectivores; sciurid, castorid, cricetid, and heteromyid rodents; leporid lagomorphs; canid and mustelid carnivores; and ungulates-camelids, tayassuids, and rhinocerotids. Several of the newly appearing groups can be shown to have entered North America from Asia, among them the rabbits (Mytonolagus), the amynodont rhinocerotids, and most of the selenodont artiodactyls including camelids, hypertragulids, leptomerycids, and oreodonts (Webb, 1977; Webb and Taylor, 1980; Emry, 1981).

Most of the Late Eocene immigrant herbivores are characterized by adaptations for masticating coarse fodder. At least some members of the following groups developed hypsodont dentitions during the Late Eocene: taeniodonts, leporids, castorids, eomyids, rhinocerotids, hypertragulids, oromerycids, and "oreodonts" (Webb, 1977). For example, Wood (1980, p. 38) characterized the cheek teeth of the eomyid genus Paradjidaumo as ". . . more hypsodont and progressively more lophodont than in Adjidaumo." Also, although most oromerycids are brachydont, Prothero (1986, p. 461) observed that in the new genus Montanatylopus "the molars are much more hypsodont than in any other oromerycid." Thus it is fair to recognize the Late Eocene and Oligocene White River chronofauna as the first in North America to sustain a substantial diversity of hypsodont herbivorous mammals.

The larger mammalian herbivores may be divided into two broadly distinct habitat groups: one group lived primarily along watercourses; the other lived mainly on the interfluves. The flat-skulled leptaucheniine "oreodonts" and the trunk-bearing amynodont rhinos were short-legged, semiamphibious forms that cropped lush vegetation in or near stream courses (Scott, 1937; Wall, 1982). On the other hand, most selenodont artiodactyls, as well as the horse Mesohippus, had relatively long slender limbs and ranged widely in open habitats. Several studies of faunal facies in the White River deposits, reviewed in Webb (1977), provided direct statistical evidence that selenodont artiodactyls and rabbits occurred predominantly in open-country or upland habitats.

Several of the White River ungulates, notably Leptomeryx, ranged together in herds (Clark et al., 1967). The adaptive relationships among social behavior, body size, and feeding mode, developed by Estes (1974), Jarman (1974), and others on the basis of the modern African ungulate fauna, suggest that the appropriate comparison for leptomerycids and other selenodonts is with moderate-sized herding forms such as the gazelles (Jarman's category C). Such forms are mixed feeders, relying on grasses only in their most nutritious new-growth stages and shifting to browsing in the dry season (Janis, 1982).

A third adaptive zone, that of the small burrowing herbivore (rhizovore) and insectivore, is extensively occupied during the White River chronofauna. This implies extensive development of well-drained soils supporting shrubby vegetation. At the same time the diversity of arboreal mammal genera declines (Webb, 1977; Stucky, 1990).

With such mammalian evidence in mind, one may look for other indications that savannas were opening the landscape of the Late Eocene and Oligocene. Hutchison (1982) recognized the severe impact of increasing aridity and seasonality on the aquatic reptile fauna during that interval in the Rocky Mountain region, and in Early Oligocene floras of North America, notably the Florissant in Colorado, the dramatic decrease in the percentage of entire-margined leaves indicates approximately a 10°C drop to about 12.5°C mean annual temperature (MacGinitie, 1962; Wolfe, 1985). Retallack's (1983) pedological studies of White River sediments provide a fascinating look at local paleosols underlying various habitats, and hint at an overall trend toward increasingly grassy and shrubby environments following the Early Oligocene climatic deterioration.

In attempting to understand ecosystems of the Eocene-Oligocene, one may ask which plants the abundant browsing and mixed-feeding herbivores might have eaten. Since grass is still rare in the Eocene-Oligocene, it may be necessary to postulate an emphasis on woody shrubs such as Sarcobatus, Ephedra, and Equisetum. Huber (1982) postulates that just such herb-dominated savannas precede an abundance of true grass savannas in Tertiary floristic history. On the other hand, grass was present and may be underrepresented due to taphonomic conditions in paleofloral sites. Grasses were established in the mid-Eocene of Australia (Truswell and Harris, 1982, cited in Wolfe, 1985) and Europe (Litke, 1968). On present evidence, Wolfe (1985, p. 364) recognized ''. . . a distinctive Eocene savanna (i.e., shrubs and widely spaced trees but no grass on the interfluves)." The extent of grasses within this chaparral and woodland savanna setting remains in doubt.

The quantum jump in mixed-feeding herbivores noted for the White River chronofauna probably helped expand the extent of woodland savanna. The new fauna and flora probably formed a positive feedback loop: large herbivores preferred feeding in more open woodlands; in turn, expansion of open formations facilitated evolution of mixed-feeding herbivores. Among the subtle changes within the history of that chronofauna ("chronofaunal creep") was a decline in the rich array of browsers (e.g., loss of titanotheres after the Chadronian), and an increase in the body sizes and numbers of shrub eaters (quasi-grazers). Such changes indicate an increasing percentage of open-country patches, and tend to be corroborated by the limited botanical and pedological data.

Runningwater Chronofauna: Transitional Savanna

The next major immigration episode in the record of North American mammals takes place in the Early Miocene. Miocene immigration patterns in North American mammals were thoroughly revised by Tedford et al. (1987). Their analysis recognizes six immigrant genera in the latest Arikareean and 14 genera in the earliest Hemingfordian (20 Ma). Native members characteristic of the Runningwater chronofauna of the late Arikareean and early Hemingfordian include flat-incisored beavers (notably Palaeocastor with its deep corkscrew-shaped burrows), hypsodont oreodonts such as Merychyus and Merycochoerus, oxydactyline and protolabine camels (especially Michenia), and the mixed-feeding horse, Parahippus. Several immigrant groups also constitute a vital and characteristic part of the Runningwater chronofauna, notably the primitive bear, Phoberocyon; such genera of amphicyonids (an extinct family also known as "bear-dogs") as Cynelos and Amphicyon; raccoons including Edaphocyon and Amphictis; and several mustelid carnivores including Leptarctus. The first representatives of three ruminant families arrived from Asia at this time: Blastomeryx represents the Moschidae (family of the living Chinese musk deer); Paracosoryx is the scion of the autochthonous North American family Antilocapridae; and Barbouromeryx establishes the presence of the giraffe-like family Palaeomerycidae. Each of these immigrant ruminant groups evolves and diversifies in the course of the Miocene.

The paleoecology of Arikaree assemblages in the midcontinent is thoroughly analyzed by Hunt (1990). Earlier in the Arikareean, widespread eolian distribution of tuffaceous sediments in a seasonally arid climate played an important role in preserving rich samples of mammals in burrows and other nontransported settings. Both the mammalian taxa and the sedimentological evidence indicate ". . . well vegetated stream border communities, spatially separated by open interchannel reaches occupied by grassland, or possibly low shrub savanna or open savanna-parkland environments" (Hunt, 1990, p. 107).

Sheep Creek Chronofauna: Park Savanna

The trend toward more open country habitats continued from the Early into the Middle Miocene at least in the midcontinent. Another substantial wave of immigrants from Asia altered the makeup of the North American mammal fauna and spurred the course of its evolution. It is awkward from a nomenclatural view that this turnover episode falls within the Hemingfordian land mammal age and is not very far removed in time from the previous turnover episode. Tedford et al. (1987) comment as follows:

It is important to note that an important faunal discontinuity exists within the Hemingfordian, signaling an abrupt shift in the Great Plains from the chronofauna characteristic of the late Arikareean and early Hemingfordian to one typifying the late Hemingfordian and early Barstovian . . . . The differences between the faunas of the Box Butte through Sheep Creek interval and the chronofaunally related younger faunas here referred to the Barstovian (Observation Quarry through Lower Snake Creek) can be ascribed mainly to anagenetic change in persistent lineages, making it necessary to distinguish these ages at the specific rather than the generic level.

Major advances within the autochthonous fauna of Hemingfordian age indicate continued evolution toward more open-country adaptations. The best example is the transition from Parahippus, a browsing horse, to Merychippus, an early grazer (Hulbert and MacFadden, 1991). Similar trends are suggested by the increased hypsodonty and quantum increase in diversity of pronghorn antelopes.

The mid-Hemingfordian wave of immigrants from Asia, like the preceding episode, represents a broad ecological spectrum. Among the diverse rodents are flying squirrels of Old World (petauristine) type, eomyid rodents, and the very significant cricetid rodent, Copemys. New carnivores include weasel-like and otter-like mustelids, while the first true cats in the New World are represented by the genus Pseudaelurus (see Table 11.2). Immigration of four very large mammals had a major impact on the Middle Miocene ecosystem: these were two rhinocerotids (short-legged, aquatic Teleoceras and long-legged, terrestrial Aphelops) and also two proboscideans (the browsing mammutid Miomastodon and the probable mixed-feeding gomphotheriid, Gomphotherium). It is now generally supposed that such "megaherbivores" played a formative role in modifying the landscape, as do elephants in modern savanna ecosystems (Owen-Smith, 1988). Detailed evidence on this point, however, has not been developed in the Miocene of North America. The mid-Hemingfordian immigration episode took place about 18 to 17 Ma and was the last major episode until the latest Miocene, a span of about 12 m.y. (Tedford et al., 1989).

Clarendonian Chronofauna: Grassland Savanna

Land mammal diversity in North America reached its zenith during the Barstovian mammal age (Webb, 1989). Savage and Russell (1983) recognized 16 families with 60 genera and 141 nominal species of land mammals in the Barstovian. The next highest numbers occurred in the Clarendonian (the next mammal age), with 55 genera and 117 species. These mammal ages are thought to indicate a savanna optimum in North America, with a rich mosaic of trees, shrubs, and grasses supporting an extraordinary variety of large and small, grazing and browsing, ungulates. It is not uncommon during this savanna acme to collect in a single site 20 genera of ungulates of which half are Equidae (Webb, 1983a; Voorhies, 1990). Savage and Russell (1983, p. 300) noted that during this interval "the mammalian fauna . . . appears singularly homogeneous throughout its geographic range." The Clarendonian chronofauna spanned three mammal ages, an interval of more than 10 m.y. (Tedford et al., 1987).

The array of ungulates in the Late Miocene of North America is comparable in many respects to that living in Africa today. The resemblance is purely convergent: the two faunas have no genera in common, and members of the few shared families play substantially different roles. For example, the body form and paleoecological context of Teleoceras, the most distinctive rhinocerotid perossodactyl in the Clarendonian chronofauna, indicate a niche analogous to that of Hippopotamus (an artiodactyl) in Africa. Voorhies and Thomasson (1979) confirmed by direct evidence from the hyoid (throat) bones that Teleoceras consumed grasses and, from the biocenosis in pond deposits at Poison Ivy Quarry, that herds of young and old frequented water. Likewise in North America the role of the African giraffe is played by the native giraffe-camel, Aepycamelus (Webb, 1983a). The most plausible explanation for these convergences would seem to be that the North American fauna evolved in a broadly comparable manner because of environmental conditions similar to those in Africa, namely, grassland savanna.

Webb (1983a) and Janis (1984) independently attempted comparisons between North American Miocene and African Recent ungulate faunas, with emphases on body size and height (or volume) of cheek teeth. They found similar numbers of ungulate species distributed as browsers, roughage feeders, and mixed feeders. A rough estimate of biomass distribution among these categories, based on quarry censuses from major North American Miocene sites, also gave results that were remarkably similar to actual census data from African game reserves: browsers accounted for less than 10% of all ungulate biomass, whereas roughage feeders (with high hypsodonty indices) made up some 60 to 80% of the total. In a related study, Hulbert (1982) showed that population structure in a Clarendonian species of the grazing horse Neohipparion indicates that it underwent seasonal migrations, an essential feature of the strategy of African grazing ungulates.

African ecological studies show that savanna biotas live where annual rainfall ranges between about 400 and 1000 mm and that, within this range, biomass is positively correlated with rainfall (Coe et al., 1976). Furthermore the number of ungulate species and the relative abundance of larger-bodied species interact with and are roughly correlated with primary productivity (McNaughton et al., 1988). These general observations in Africa shed light on mechanisms governing species richness and size-frequency distribution of ungulates in Miocene local faunas of midcontinental North America before and after the savanna optimum in the Barstovian. In midcontinental North America the Late Miocene ungulate diversity decline presumably tracks a decrease in mean annual rainfall.

Floristic evidence supports the faunal evidence of widespread grassland savanna in the Middle Miocene of North America and increasingly arid climate in the Late Miocene. The Tehachapi flora of Hemingfordian age in California, the Stewart Valley flora of Barstovian age in Nevada, and the Kilgore flora of Barstovian age in Nebraska each has distinctive regional features, yet together they justify the conclusion (Wolfe, 1985, p. 371) that "during the Early to Middle Miocene, savanna developed in low latitude areas that are presently dry." Studies of as many as 18 species of fossil grasses directly associated with mid-Hemphillian ungulates in the Minium Quarry in Kansas also strongly indicate a grassland savanna biome (Thomasson, 1986; Thomasson et al., 1990). In the same floras are leaves with stomata and other detailed features indicative of the subfamily Chloridoideae and of C₄ photosynthesis characteristic of tropical floras growing at an optimum temperature of 30 to 32°C (Thomasson et al., 1986). Such floristic studies suggest that the Late Miocene climatic regime in midcontinental North America resembled that of seasonally arid subtropical savannas in South and Central America today.

Early Pliocene: Spread of Steppe

About 5 Ma, at the end of the Miocene, the long-term trend toward increasing aridity in midlatitudes led to the final breakup of the extensive North American savanna biome and its replacement in midcontinental North America by extensive grasslands or true steppe. Ungulates experienced a series of decimations that spanned the Hemphillian; and the first major decline (at about 12 Ma) affected browsers more than grazers (Webb, 1983a, 1984; see also Figure 11.4). By the end of the Hemphillian the number of ungulate genera was cut to less than half of its former peak. No new chronofauna was established, but a major immigration episode coincided with the final late Hemphillian extinctions.

Figure 11.4

Decline of ungulate genera in Miocene of North America (after Webb, 1983a) compared with Miocene record of δ¹⁸O based on benthic foraminifera (after Miller et al., 1987). Note that the major Middle Miocene cooling episode coincides with the major (more...)

Floristic changes at the end of the Miocene clearly record the expansion of grasslands in midcontinental and western North America. For example, Wolfe (1985, p. 371) notes that in the Hemphillian of Kansas "the fructifications include abundant grass, and the leaf-assemblages are the low-diversity type that typically occurs along streams in unforested grassland regions." In western Wyoming, Barnosky (1984) recognizes xeric shrub floras during the latest Miocene. The Mt. Eden flora in southern California suggests the beginning of steppe conditions (Wolfe, 1985). Axelrod's (1992) floristic review shows that in the Great Basin and the midcontinent, the most arid interval of the Tertiary occurred at about 6 Ma. Renewed Cordilleran tectonics during the Miocene-Pliocene greatly increased rain shadow effects and provincialism in many regions of temperate North America.

The vast herds of ungulates themselves probably accelerated the decline of savanna and the expansion of steppe in the Pliocene. In Africa during extended droughts, elephants and/or equids devastatingly transform savanna to steppe (Sinclair, 1983; Owen-Smith, 1988). In North America, many late Hemphillian and Blancan fossil quarries (e.g., the Hagerman Horse Quarry in Idaho) yield abundant samples of only one or two grazing equids. These suggest (but do not demonstrate) scenes of massive overgrazing. A positive feedback loop between faunal and floral evolution may have accelerated the process that established steppe conditions in the Early Pliocene.

It should be noted that the Early Pliocene expansion of steppe environments did not encompass all of North America. Provincialism is far more evident than in the Miocene. In the Pacific Northwest, faunal and floral evidence clearly indicates continuation of a mesic forest biome although it interdigitated with semiarid savannas in the Great Basin (Tedford and Gustafson, 1974). The biota around the Gulf of Mexico also indicates the continuation there of savanna conditions (Webb, 1989). The remarkable resemblances between such Hemphillian faunas as the Yepomera local fauna in Chihuahua, Mexico, and the Bone Valley fauna in Florida give evidence of a broad subtropical savanna south of latitude 30°N, which was perpetuated by the persistence of summer monsoons (Webb, 1977). Thus, the Pliocene shift to steppe biomes in the midcontinent produced greater provincialism among diverse regions of North America.

Onto such a scene of regional desiccation and increased provincialism burst a new wave of Asiatic immigrants. As indicated below (Table 11.2), these immigrations appear to be concentrated in the latest Hemphillian, about 5 Ma. Key groups of land mammals that crossed the Bering at this time were vole-like rodents, both terrestrial and aquatic (Repenning, 1987); at least six carnivores, including exotic groups such as hyaenids (Chasmaporthetes) and pandas (Parailurus), as well as more familiar groups such as Lynx and Ursus; and also at least two genera of deer, which gave rise to a substantial radiation in North and South America.

Late Pliocene and Pleistocene: Further Continentality and Provincialism

The secular trend of the Cenozoic toward colder, drier, and in other ways more extreme conditions led to the environments of the later Pliocene and Pleistocene. As the first Laurentide glaciation settled over North America about 2.5 Ma, forest tundra biomes became widespread in the Arctic Circle (Funder et al. 1985; Repenning, 1985). Mountain building in the Cordillera extended tundra corridors along high-elevation routes into temperate latitudes. Rain shadows helped extend cold-steppe habitats into many basins of western North America. In the Snake River Plain (as represented by the Hagerman fauna and flora), there were still broad-leaved deciduous forests. To the north and at higher elevations, however, conifers predominated (Lundelius et al., 1987). The southern plains (as represented by the Mt. Blanco fauna and the Rita Blanca flora) were characterized by a seasonally arid climate that produced grassy scrubland and an impoverished ungulate fauna dominated by grazers. Widespread caliche deposits in Late Pliocene deposits of the Great Plains have long been recognized as indicators of seasonal aridity (Hibbard and Taylor, 1960).

In Florida and presumably the rest of the southeastern United States, vertebrate faunas of late Blancan age carry a number of "holdover taxa"—such as the peccary, Mylohyus; the protoceratid, Kyptoceras; and the three-toed horse, Cormohipparion—which were already extinct in the midcontinent. The ecological balance of these holdovers suggests persistence of seasonal savannas and extensive mesophytic forests in warmer, more productive settings than in the High Plains (Robertson, 1976; Webb, 1978; Hulbert, 1987). In general, the Late Pliocene array of biomes in North America approximated the degree of provincialism seen at present, although the extremes of freezing winters and arid deserts were absent.

In the Late Pliocene (Blancan 2) the first large wave of immigrants from South America appeared, including glyptodonts, chlamytheres, armadillos, ground sloths, capybaras, and porcupines. At the same time, another large wave of immigrants came from Asia; among them were the spectacled bear (Tremarctos) and several rodents (Synaptomys, Pliopotamys, and Mictomys) (Lundelius et al., 1987). This final immigration episode of the Tertiary coincided with expansion of the Bering bridge. It clearly correlates with the onset of Northern Hemisphere glaciation, sea-level lowering, and evidence of pre-Nebraskan till in western Iowa (Lundelius et al., 1987; Repenning, 1987).

The late Hemphillian and early Blancan immigration episodes just considered adumbrate an increasingly active series of Quaternary immigration episodes. High latitude Quaternary studies on both sides of the Beringia land bridge record trends among microtine rodents, as well as many large mammals, such as mammoths and musk oxen, toward steppe-tundra adaptations (Sher, 1974; Repenning, 1985; Herman, 1989). The complex pattern of faunal dynamics among Quaternary mammals is well summarized by Tedford et al. (1987, p. 192): "The accelerating pace of immigration from Hemphillian into Pleistocene time reflects the availability of dispersal routes to North America and increasing environmental instability, particularly at high latitudes, that provided the goad for the movement of mammals."

Mammal Turnover on Other Continents

As the land mammal successions on other continents become more continuously documented and their chronologies become more fully correlated with the absolute time scale, it is important to determine whether they respond synchronously with the immigration and rapid turnover episodes observed in the North American record. We summarize a few highlights from other continental mammal faunas in Europe, China, the Indian subcontinent, and Africa. South America and Australia were almost completely isolated during most of the Tertiary; therefore, their land mammal faunas need not be involved in this immigration analysis.

Oxygen Isotopes and Mammal Immigrations

In an influential paper, Fischer (1983) introduced the concept of two Phanerozoic supercycles in which the Earth's climate has alternated between a greenhouse state and an icehouse state. In the Cenozoic the Earth has experienced its most recent change from greenhouse to icehouse. The broad outlines of this change are well known and have been revealed by shifts in the oxygen isotope ratios extracted from tests of foraminifera in marine sediment cores. The Cenozoic record for benthic foraminifera clearly displays a secular trend to more positive isotopic values, reflecting major cooling of bottom waters in the world ocean. There is general agreement that in the Early Eocene, δ¹⁸O values were zero or negative; that these values became more positive throughout the Eocene; and that they declined sharply in latest Eocene, and again in Middle Miocene and latest Miocene.

Another independent approach to the history of Cenozoic climates derives from seismic stratigraphic studies of passive continental margins. Haq et al. (1988) provided a recent summary of the apparent eustatic curve compiled from many well-dated seismic stratigraphic studies. The Cenozoic is represented by parts A and B of the Tejas megasequence and includes seven second-order supercycles. Each of these supercycles spans about 10 Ma and terminates at a sequence boundary of major magnitude. Shackleton et al. (1988), Williams (1988), Christie-Blick et al. (1990), and Miller et al. (1991) have carefully considered the relationships between isotopic and seismic studies as they bear on the history of Cenozoic climate. Potential pitfalls for the seismic method are confounding the effects of local sedimentary load or local tectonism with more global eustatic effects. For these reasons, we focus on the isotopic record as the more reliable signal of global climatic change during the Cenozoic. It should be noted, however, that for many intervals, there is substantial correlation between the seismic and isotopic approaches.

Shackleton and Opdyke (1977) showed that during the late Pleistocene, oxygen isotope ratios covaried between tropical planktic and benthonic foraminifera. They interpreted the simultaneous changes in surface and bottom-dwelling foraminifera as evidence of growth and decay of ice sheets on the continents. Therefore, the δ¹⁸O record represents change in both ocean temperature and ice volume. The problem has been to partition these two effects. Miller et al. (1991) recently extended the covarying isotopic studies to sediments of mid-Tertiary age, recognizing key intervals of ice buildup by covariant increases in heavy oxygen ratios in Miocene and even Oligocene cores.

Prentice and Matthews (1988) have attempted to monitor Cenozoic sea-level change by analyzing oxygen isotope ratios in planktic foraminifera from equatorial regions. The efficacy of this approach depends on the assumption that equatorial sea-surface temperatures (away from upwellings) have not changed during the Cenozoic. Thus, they reason that observed isotopic changes wholly reflect waxing and waning of glacial ice. Figure 11.2 juxtaposes North American land mammal immigration episodes with the trace of Cenozoic oxygen isotope ratios based on planktic forams from equatorial regions. We discuss each of the continental first-order immigration episodes in relation to the marine record.

Figure 11.2

First- and second-order land mammal immigration episodes in the Cenozoic of North America juxtaposed with the marine record of δ¹⁸O from equatorial planktic foraminifera after Prentice and Matthews (1988). If sea-surface temperatures have not (more...)

The earliest pair of first-order immigration episodes are the Clarkforkian and early Wasatchian, straddling the Paleocene-Eocene boundary. Although they fall generally within the warmest climatic interval of the Cenozoic, they correlate with small cooling events in the isotope curve at about 59 and 56 Ma. The Clarkforkian immigration episode correlates particularly well with the abrupt cooling episode at the end of the Paleocene (59 Ma), demonstrated by Kennett and Stott (Chapter 5, this volume) in high-resolution data based on planktonic forams from the southern ocean. Miller et al. (1987) postulated a sea-level drop at this time, and Haq et al. (1988) recognized a second-order drop in sea-level (TA 222-24) in the Thanetian.

The very large Wasatchian land mammal immigration episode reflects plate tectonic effects in the North Atlantic that established a very broad, low latitude corridor across the Thulean route and thus produced extremely close faunal resemblance between North America and Europe as discussed above.

The next first-order episode occurs in the early Duchesnean (Late Eocene) at about 40 Ma (Emry, 1981; Krishtalka et al., 1987). This episode corresponds well to a number of global Late Eocene events. Miller et al. (1987) show a major oxygen isotope increase at about 40 Ma based on benthic forams, and this event is also seen in planktonic forams from equatorial cores presented by Prentice and Matthews (1988). These isotopic events correlate with the beginning of the major sea-level drop TA4 (Priabonian) of Haq et al. (1988). According to Hallam (1984) and others cited therein, this latest Eocene sea-level drop is greater than that of the Late Oligocene. The profound global climatic shift of the Late Eocene correlates with a strong increase in the Earth's thermal gradient due to cooling in the southern ocean (Kennett and Barker, 1990).

During the Neogene, land mammal immigration episodes increased markedly in North America. Because of their importance and frequency, these Miocene and Pliocene immigrations were subjected to detailed analysis by Tedford et al. (1987). In Figure 11.3, we juxtapose the full array of Neogene land mammal immigration episodes (including third-order episodes) with the δ¹⁸O excursions numbered by Miller et al. (1991). As suggested by Opdyke (1990) the episodes correlate remarkably closely with oxygen isotope events in the marine record. Only one isotope event (namely Miocene 3) fails to correlate with an immigration episode in North America.

On the other hand, several immigration episodes in the Miocene of North America fail to correlate with any of the positive isotope excursion numbered by Miller et al. (1991). One first-order immigration episode (Arikareean 2 at 20 Ma) fails this correspondence test; as noted below it does correlate with a small (unnumbered) positive isotope excursion. Blancan 1 (at 4.8 Ma) also appears to be unrequited, but in fact it may correspond with Pliocene 1 of Miller et al. (1991). Other lesser immigration episodes at 14.5, 7.0, and 6.0 Ma do not correspond to numbered isotope excursions.

The Early Miocene records the largest set of generic immigrations in the history of the North American land mammal fauna. Three land mammal dispersal episodes fall near the boundary between Arikareean and Hemingfordian: they begin with a second-order episode at 21 Ma followed by first-order episodes at 20 and 18 Ma. The episodes at 21 and 18 Ma coincide with isotope events Mi1b and Mi1c of Miller et al. (1991). The immigration episode at 20 Ma corresponds to a positive isotope excursion in both benthic and planktonic forams as identified by Prentice and Matthews (1988). According to the latter, sea-level at 20 Ma reached depths not attained again until the Pleistocene. Miller et al. (1987) show a possible erosional interval at 20-21 Ma, and Haq et al. (1988) record a first-order drop in the seismic curve between 20 and 21 Ma (TB2). On the other hand, the episode at 18 Ma does not correspond to any major sea-level drop noted in the seismic curve.

Early Miocene land mammal immigrants, including three groups of ruminants, participated in the transition to savanna biomes in midcontinental North America, accompanied by a shift to seasonally arid climates and loess deposition, as discussed above. Presumably this Early Miocene savanna transformation was driven by a fundamental change of state in the global climatic pattern, most notably intensified deep water flow with opening of the Drake Passage and concomitant development of the Antarctic Circumpolar Current (Kennett and Barker, 1990).

Further major shifts are evident in the Middle Miocene, but in North America, these do not include significant numbers of immigrants. Instead, the acme of land mammal diversity, dominated by horses and other savanna herbivores, is attained in the Barstovian. This savanna optimum on the continent correlates with the largest Middle Miocene shift in the Neogene δ¹⁸O record. The strong string of positive isotope shifts in the marine record is widely believed to represent the buildup of permanent glaciers on East Antarctica (Kennett and Barker, 1990). Haq et al. (1988) record a major sea-level drop at 16 Ma with additional drops at 14 and 12 Ma.

In Figure 11.4, we juxtapose the Neogene portion of the benthic foram isotope record after Miller et al. (1987) with the detailed record of large herbivore diversity and feeding modes in North America. The latter record, more fully discussed by Webb (1983a), shows the acme of herbivore diversity in the Barstovian, followed by a steep stepwise decline during the Clarendonian and early Hemphillian. The losses affected primarily browsing ungulates and, secondarily, grazing ungulates. Environmentally, these Middle and Late Miocene faunal changes represent the shift from savanna to steppe biomes, presumably a result of declining rainfall. A remarkably similar transition occurs concurrently in the land mammals of the Indian subcontinent: from the herbivore acme in woodland savanna environments, a stepwise decline takes place that also represents a shift toward more arid environments between 16 and 12 Ma (Barry et al., 1991). The Middle Miocene patterns on land and sea strongly support an interpretation of climatic forcing (driven by attainment of permanent glaciation in East Antarctica) of land mammal faunal evolution in temperate latitudes.

In the Late Miocene of North America, beginning with the Clarendonian/Hemphillian boundary at about 8.5 Ma, three second-order immigration episodes occur. The episode at 8 Ma coincides with isotope event Mi7 of Miller et al. (1991) and roughly approximates the drop in sea-level recognized as TB3.2 of Haq et al. (1988). Of the two further migratory episodes at roughly 7 and 6 Ma, only the latter corresponds to a reported marine event, namely that shown by Miller et al. (1991) at 5.8 Ma. Kennett and Barker (1990) suggest that the later Miocene was the interval during which West Antarctica became permanently glaciated.

The Miocene-Pliocene boundary is associated with a first-order immigration episode in land mammal faunas of both North America and Europe. This episode correlates well with a positive shift in δ¹⁸O (Hodell et al., 1986), but does not seem to correlate with any episode of sea-level lowering in the Haq et al. (1988) curve. Possibly sea-level is beginning to fluctuate at high frequency in the Late Miocene and Pliocene, with the consequence that it becomes difficult to define any particular pulse with seismic stratigraphy. On the other hand, the Messinian is well known in the Mediterranean region and in the isotopic record as a major glacioeustatic cycle (Adams et al., 1977; Kennett, 1985; Hodell et al., 1986). As emphasized by Webb (1983b), the Messinian can be tightly correlated with the mid-Hemphillian extinction and faunal turnover episode in temperate North America.

The last second-order episode of the Tertiary sweeps through land mammal faunas on most continents at 2.5 Ma in the Late Pliocene. Such major immigrations are clearly related to the onset of Northern Hemisphere glaciation recognized in the δ¹⁸O record and also indicated by glacial sediments (Ruddiman and Raymo, 1988; Shackleton et al., 1988; Raymo and Ruddiman, 1992). All of these features correlate with marked cooling and sea-level lowering indicated by the oxygen isotope record (Shackleton and Opdyke, 1977).

Generally speaking, land mammal immigration episodes in North America and other well-studied continental areas correlate with positive excursion of the oxygen isotope record from tropical marine forams. Other continental regions clearly corroborate the significant turnovers recorded at about 20, 5, and 2.5 Ma.

In some cases, however, the terrestrial turnover pattern gives no evident response to major cooling events. The two most glaring examples are the Early Oligocene (White River chronofauna) and the Middle Miocene (Clarendonian chronofauna) in North America when, despite major global changes, the mammal faunas show remarkable stability and continuity. Possibly no land bridges were available during the Early Oligocene. In the Middle Miocene, this excuse cannot serve, because a few Asiatic immigrants did trickle in and Beringia was emergent. Evidently, during these two stable chronofaunal stages the North American mammalian fauna was resilient and not open to immigrants.