Frozen Fauna of the Mammoth Steppe: The Story of Blue Babe

by Guthrie, R. Dale

  Fig. 9.4. Joe Lake herbaceous pollen. Unlike Squirrel Lake and Kaiyak Lake, the deep core at Joe Lake (Anderson 1988) has disparate sedimentation rates. Maximum sedimentation occurred in the Boutellier Interval and the Holocene. One of the six or so samples from the short Duvanny Yar core segment had little pollen, giving the graph a large bite at 16,000 yr B.P. However, the general pattern from the other five samples shows no dramatic pollen reduction during the Duvanny Yar.

  Ritchie shows that annual pollen influx in the Manitoba sites ranges around 2,000 grains per cm2 per year. Although his comparison with such a grassland is not quite appropriate, we can, for lack of other data, dissect this total as we did his Holocene northern totals and eliminate trees as well as, in this case, domestic cereals and groups comprising introduced European weeds (certainly one cannot justify including these pollen producers in the comparison). We can then focus on the influx values of wild grasses, sedges, and sage, which are most relevant for our comparisons. Surprisingly, the Manitoba sites are roughly comparable to the Alaskan fossil sites. When one looks only at these three dominant taxa from grasslands, total production at each location is not far from 400 grains per year. These are plotted in figure 9.5. I am not saying, of course, that Pleistocene Alaska was a tallgrass prairie or that it was as productive as a tallgrass prairie, only that influx data are very blunt and often misleading tools.

  To determine whether these modern samples are really representative of influx volumes of preagriculture Holocene grasslands we must look at another core, north of these modern collecting sites. Ritchie’s (1969) pollen profile from Riding Mountain, Manitoba, is informative about the meaning of pollen influx in grasslands. He interprets the changes from glacial spruce-dominated boreal forest, to grassland, to aspen parkland (for Ritchie this zone is rather enigmatic), to pine-birch boreal forests of today. Going bottom to top, the pollen influx in grains per year for all graminoids averages about 450, 450, 250, and 400, respectively, for these four different environments (fig. 9.6). These graminoid influx amounts are about the same as those from glacial portions of Alaskan pollen cores summarized in the same figure. Remember that pollen being sampled from the grasslands of Manitoba are east of (and hence downwind in the Pacific Air Mass) one of the greatest grasslands on earth. If pollen from the general region is being broadly sampled, one would expect sage and grass pollen to be overrepresented in these cores rather than underrepresented.

  Fig. 9.5. Riding Mountain Manitoba, herbaceous pollen. To illustrate the limited usefulness of pollen influx data to our understanding of Mammoth Steppe vegetation, I used data from Ritchie’s (1969) Riding Mountain core in Manitoba to plot influx of grass, sedge, and sage pollen. This site is downwind from the largest wild grassland in North America, and yet influx values for these wild herb taxa are modest. Ritchie’s reconstruction of the dominant physiognomy (right) on the basis of the arboreal pollen is not readily apparent from influx of herb taxa. I conclude from this that influx data is a poor paleoecological tool in the far north.

  Fig. 9.6. The meaning of influx data from the Mammoth Steppe. This comparison of herb pollen influx at five Holocene grassland sites in Manitoba (Ritchie 1969) with Alaskan full glacial sites (Duvanny Yar Interval) shows that the patterns of pollen influx are not very different. This does not mean herbaceous vegetation was similar; it suggests, however, that we should use influx data with great caution.

  Borrowing Ritchie’s assumption that pollen influx is an indication of plant biomass, one could say that in Alaska plant biomass for graminoids and Artemisia per year during peak glacial was about the same as it now is on the mid- and tallgrass prairies of Manitoba. Although my purpose in these exercises is to show that one cannot take the details of influx values too seriously in reconstructing vegetation patterns, they do allow us to reject the remainder of Proposition A.

  Actually, even a cursory review of the variations and idiosyncrasies of pollen influx figures is enough to make one cautious about paleoecological reconstruction on the basis of influx alone. Variations of influx values from within and between different sites, even when looking at the same taxa, range from the hundreds to the thousands (Waddington 1969; Davis 1983), with probably little difference in biomass or percentage of ground cover over large segments of the profile. Even on a long-term scale, taxa do not seem to have fixed reproductive efforts. But whatever factors enter the equation, we can say that influx data from Beringia do not require a polar desert. On the contrary, herb influx figures and mammalian fossils are quite consistent with the concept of a Mammoth Steppe, a rangeland suitable for large-mammal grazers.

  As a vertebrate biologist, I do not purport to use these pollen influx data as the sole basis to argue for a grassland environment. The vertebrate fossils do that well enough. Rather I wish to show that existing pollen data for Alaska cannot credibly be used against such a concept. These herb influx totals for Pleistocene Alaska in figure 9.6 are around 400 per year. And we should remember that many forbs are insect pollinated and remain invisible in the fossil pollen record. The closed sward of alpine and low-land tundras today does not produce more pollen than found in Alaskan Pleistocene lake sediments, yet the standing biomass of these areas exceeds most arid steppes. If today we were to change the vegetation of Alaskan tundra to more xeric dominants, which winter cure with more nutrients, make them more widespread, and leave them exposed without much snow cover, they would be a productive rangeland for northern wild mammals of the mammoth fauna. I do not have to say “would,” because mammalian fossils illustrate that Mammoth Steppe grazers flourished here.

  Proposition B

  The glacial environment, as reconstructed from pollen data, was so harsh that it could not possibly have supported the “mammoth fauna”; this is borne out by the lack of large-mammal fossils dating from full glacial (Duvanny Yar) times.

  For nearly a century miners have been removing the silt overburden from hundreds of placer deposits in upland interior Alaska. Seldom have these efforts failed to produce large-mammal fossils. Some mines unearth a few bones; at many the fossils are measured in tons. Most of these fossils do not end up in museums, but are kept in Alaskan homes and sheds. Others leave the north and are scattered across the country, taken home as souvenirs. Many are sold on the curio market. No areas of unglaciated Alaska have not produced fossils of Pleistocene large mammals. The concentration of their distribution is almost completely correlated with mining activity. I gather from talking to Soviet paleontologists that this is true for Siberia as well. There are literally hundreds of Pleistocene large-mammal fossil localities in Beringia, and these invariably produce the same grazers, particularly bison, horses, and mammoths.

  This is not the entire story, however. Not only are these fossil grassland mammals found in a “coarse grain” scattered across such a large area, but normally the dominant species occur together locally. The common Pleistocene large-mammal species are found in virtually all of the terrestrial Pleistocene paleontological sites. This is true whether one is in the high hill country (Guthrie 1968) or the lowlands (Bonnichsen 1979; Harington 1978). Species composition does not always retain the same proportions, for example, bison and sheep tend to increase in the uplands (Guthrie 1968). In some areas, a species drops out (such as the woolly rhino, which has not been found in Alaska) or changes in a geoclinal gradient so that a niche is occupied by another related species (red deer, Cervus elaphus, in the west and wapiti, Cervus canadensis, in the east), but for the most part the mammalian community retains its character. This is especially true for mammoths, bison, and horses. Their coexistence in individual sites suggests that these key species of the Mammoth Steppe were not mutually exclusive in spatial distributions, or chronologically exclusive, but were indeed adapted to coexist on a very local scale.

  The impressive diversity of this large-mammal grazing community suggests that the Mammoth Steppe was (1) a more productive rangeland for large mammals than the natural vegetation in those areas today, (2) a vegetation pattern
dominated by grasses, and (3) an environment without an appropriate modern analogue in the north (Matthews 1979; Guthrie 1982). However, Schweger and Habgood (1976), Ritchie and Cwynar (1982), and Colinvaux and West (1984) have argued that most of these mammals did not exist together, that the apparent high diversity is artificial, produced from a compilation of species from deposits that cover a long time span. Colinvaux and West, for example, argue that musk-oxen and bison probably did not coexist at all. Instead, musk-oxen took the place of bison during the interstades and interglacials. In European and Asian portions of the Mammoth Steppe there are many localities, both paleontological and archaeological, with these two species together, and in fact a wide assortment of these species is present in a given horizon in each site. However, the nature of deposition in Alaska and the Yukon Territory means that most bones do not occur in high concentrations in a typical “fauna,” as one finds elsewhere in cave deposits. There are some exceptions, like Blue Fish Cave (Cinq-Mars 1979), with concentrations of bones from around 13,000 to 15,000 years ago, Loess Den (Guthrie 1988) at around 24,000 years ago, and Dixon’s (1984) Porcupine Cave at around 21,000 years ago. These concentrations, like late Pleistocene deposits in Europe and Asia, do contain the characteristic species of the mammoth fauna (see synthesis in Hopkins et al. 1982).

  Aside from the concentrations of bones with specific dates mentioned above, most of the thousands of Pleistocene bones now in collections occurred singly and have been dated in a haphazard manner for specific occasions and purposes. However, a number of dates have accumulated (Matthews 1982; Morlan and Cinq-Mars 1982). These dates, and others not published in those reviews, clearly show that the main members of the mammoth fauna were in Pleistocene eastern Beringia during the entire period covered by radiocarbon-dating and beyond. This is contrary to Ritchie’s (1984) criticism of my original statement (Guthrie 1968) that virtually all of these fossil mammals date from late Wisconsinan. About 80% to 90% of these dates fall in the finite radiocarbon range of less than 40,000 years (although some dates in the high 30,000’s and greater may actually have been brought into the finite range by contamination).

  In Siberia and in Old Crow, Yukon Territory, the scatter of dates is most dense during the interstadial (around the 30,000’s) and late glacial (14,000–12,000). This has prompted Ritchie and Cwynar (1982), Colinvaux and West (1984), and others to argue that the full glacial was unoccupied by large mammals or that the few here were “itinerants” from farther south. But the Siberian and Old Crow dates cluster during the two warm episodes because of major dating biases. In the Yukon Territory the dates are mainly from the once well-funded Old Crow early-man archaeological surveys. Full glacial dates from fossil mammals in the Old Crow lowlands cannot exist, because sediments of that age are all lacustrine. Proglacial lakes flooded the area during the full glacials (28,000–12,000), and hence large-mammal fossils do not occur in these sediments. In Siberia most radiocarbon dates are taken in association with archaeological sites, and there are few, if any, archaeological sites in the far north during the peak glacial. Furthermore, Soviets prefer to date archaeological sites (and other bone accumulations) with wood, and there was little or no wood in the Asian far north during the peak glacial times. Little wonder that there are so few dates from full glacial age on mammals from Siberia and the Old Crow.

  There are, however, other means to test the proposition that large mammals were not in the far north during the Duvanny Yar full glacial. Since we cannot ever determine even approximate densities, except by some relative comparison or, in extreme cases, by absence, in a fairly sampled chronological scatter of dates the real question is whether there are fossils dating from the full glacial. Since there are a number of large mammal fossils of full glacial age, we can reject the latter part of Proposition B outright. A more useful question would be whether there are dates of this mammoth fauna during the full glacial, and whether they scatter without bimodial cluster during warm wet periods. We could fairly test chronological distribution of fossil mammals by examining dates from Alaska and the Yukon Territory, away from the proglacial lakes of Old Crow Flats. This would test Proposition B and the Mammoth Steppe concept.

  We can use bison for our first example. I collected the dates in table 9.1 from a brief survey of Alaskan Pleistocene bison.

  The existence of glacial-aged (14,000–25,000 B.P.) bison fossils allows us to reject Proposition B immediately. Furthermore, we can respond to the question of nonrandom clustering of dates outside of the peak glacial. The probability of such clustering is beyond the normal acceptable level at p = .028. There is a continuous scatter through the full glacial, contrary to the proposition of Ritchie and Cwynar (1982), Ritchie (1984), and Colinvaux and West (1984) that large mammals were confined to, or even concentrated in, the interstades and interglacials. It is apparent from these data that there are no striking absences of bison from Alaska during the last full glacial (Duvanny Yar Interval), from about 25,000 to 14,000 years ago.

  Horses (Equus sp.) in Alaska and the Yukon Territory (minus Old Crow) show much the same pattern as bison (see table 9.2).

  Equid samples used for radiocarbon dating were submitted precisely because most had no stratigraphic context; thus the dates can be considered as random samples from late Pleistocene sediments. The equid dates, like those of bison, do not show a marked bimodal curve on either side of the full glacial. The probability that these dates cluster outside of the 14,000–25,000 B.P. full glacial span is unacceptably low at p = .018.

  Table 9.1 Chronological Distribution of Alaskan Pleistocene Bison

  Table 9.2 Chronological Distribution of the Horse

  Although thousands of mammoth bones have been found in Alaska, few have been dated. Dated mammoth bones show no tendency for a thinner spread during the full glacial (Duvanny Yar Interval). In fact, when we look at all the mammoth dates from Alaska, including those from the uplands of the Northwest Territory and Yukon Territory, the dates still do not show any modality away from the full glacial (see table 9.3).

  As with the dates from bison and horse bones, one could conclude that mammoths were present in eastern Beringia throughout the Pleistocene covered by radiocarbon range. Sample size of dated mammoth bones is not sufficient to do a valid probability test, but one can see from the scatter that there is no hint of clustering outside the full glacial. The dates show that mammoths existed in eastern Beringia during the full glacial, again allowing us to reject Proposition B. And as I concluded from a comparison of several thousands of bones found in different late Pleistocene sediments (Guthrie 1968), these three large grazers account for the largest proportion of mammalian fossils.

  Table 9.3 Chronological Distribution of the Mammoth

  Proposition B proposes that the region must have been barren tundra or polar desert, and as such large grazers could not have been present during glacials or were at least not present contemporaneously. This perception is voiced by Colinvaux and West (1984): “in early interglacial times . . . game herds harried by large felids and canids may have existed in a complex landscape . . . but the Beringia of land bridge (full glacial) times was an unproductive place” (p. 12); Ritchie and Cwynar (1982) state: “The large and diverse ungulate populations probably were present during Pleistocene interstadials more than 30,000 years ago . . . rather than during the time of the herb zone, 30,000 to 14,000 years ago” (p. 113). The radiocarbon dates above clearly do not support this perception.

  The logic of Colinvaux and West’s (1984) comment that the scarcity of radiocarbon dates on large mammal bones (of one date per thousand years) show that few animals were around is peculiar; this is comparable to saying that the few pollen cores studied from Alaska mean that there are few lakes. It is as obviously false to conclude from the small number of dated fossils that few bones of Pleistocene mammals have been found in Alaska. In fact, the radiocarbon-dated bones are a tiny sample of the tens of thousands of mammal bones removed from natural cut banks and mining exposures. Interior Alas
ka is extraordinarily rich in late Pleistocene vertebrate fossils; bones from the Fairbanks area are scattered in museums around the world and even predominate in some museums. Several floors of Pleistocene mammal fossils in the Frick Wing of the American Museum of Natural History are dominated by Alaskan material. It is true that hundreds of dates for each species would give us more information on a variety of paleoecological topics. Paleontologists working in Beringia can certainly be criticized for not having enough dates on fossil mammals, but the dates listed above, while few, are probably more numerous than for individual large-mammal species and for individual localities from other regions.

  There are fewer dates on other species of the mammoth fauna. In some instances this is because the species are comparatively rare and dating would involve sacrificing the whole specimen. Museums are reluctant, for example, to donate Pleistocene lion or saiga specimens for dating. Species that did not become extinct are also poorly dated, as the chronology of their disappearance has not been an issue. Finally, fossils in most other places are found in concentrated sinkhole or cave deposits, where one radiocarbon date can be assigned to a number of fossils in the assemblage. Such concentrated faunal assemblages are uncommon in the north; usually each fossil must be individually dated.

  There are some dates on species other than bison, horse, and mammoth. Colinvaux and West (1984) and Ritchie (1984) have argued that these other members of the mammoth fauna must have existed during interglacials or interstadials, that they could not have lived together during glacials. Yet published dates exist showing that in addition to horse, bison, and mammoth, many other species were living in Beringia during the full glacial (14,000–25,000): lion (Felis), musk-oxen (Ovibos), bonnet-horned musk-oxen (Symbos), camel (Camelops), sheep (Ovis), woolly rhinoceros (Coelodonta), and caribou (Rangifer) (Matthews 1982; Dixon 1984; Vereshchagin and Baryshnikov 1984). Proposition B ignores this evidence.