(This is a sample cover image for this issue. The actual cover is not yet available at this time.)
This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
Author's personal copy
Mammoth steppe: a high-productivity phenomenon
, N.S. Zimov
, A.N. Tikhonov
, F.S. Chapin III
Northeast Science Station, Paciﬁc Institute for Geography, Russian Academy of Sciences, Cherskii 678830, Russia
Zoological Institute, Russian Academy of Sciences, Saint Petersburg 199034, Russia
Institute of Arctic Biology, University of Alaska, Fairbanks, AK 99775, USA
Received 11 January 2012
Received in revised form
2 October 2012
Accepted 4 October 2012
At the last deglaciation Earth’s largest biome, mammoth-steppe, vanished. Without knowledge of the
productivity of this ecosystem, the evolution of man and the glacialeinterglacial dynamics of carbon
storage in Earth’s main carbon reservoirs cannot be fully understood. Analyzes of fossils
C dates and
reconstruction of mammoth steppe climatic envelope indicated that changing climate wasn’t a reason for
extinction of this ecosystem. We calculate, based on animal skeleton density in frozen soils of northern
Siberia, that mammoth-steppe animal biomass and plant productivity, even in these coldest and driest of
the planet’s grasslands were close to those of an African savanna. Numerous herbivores maintained
ecosystem productivity. By reducing soil moisture and permafrost temperature, accumulating carbon in
soils, and increasing the regional albedo, mammoth-steppe ampliﬁed glacialeinterglacial climate vari-
ations. The re-establishment of grassland ecosystems would slow permafrost thawing and reduce the
current warming rate. Proposed methods can be used to estimate animal density in other ecosystems.
Ó2012 Elsevier Ltd. All rights reserved.
During the Last Glacial Maximum (LGM), mammoth steppe (MS)
was Earth’s most extensive biome. It spanned from Spain to Canada
and from arctic islands to China (Adams et al., 1990;Guthrie, 1990;
Sher, 1997;Alvarez-Lao and García, 2011). Modern man evolved in
this biome (Guthrie, 1990). During interglacial warming, forests
expanded northward, and northern Siberia, Alaska, and the Yukon
were an MS refugium. Trees and shrubs would have also penetrated
to these places, but they were not the dominate vegetation cover
(Sher, 1997). However, at the beginning of the Holocene, mossy
forests and tundra displaced MS even there. Many lakes and
wetlands have appeared. It has been assumed that, during the
Holocene, in contrast to other previous interglacials cold dry
steppe-like climate switched to a warmer wetter climate that, in
turn, caused the disappearance of grasslands and their megafauna
(Guthrie, 1990;Vereshchagin and Tikhonov, 1990;Velichko and
Zelikson, 2001;Schirrmeister et al., 2002;Sher et al., 2005).
C dating has indicated that mammoths persisted on
islands until the mid-Holocene (Guthrie, 2004;Vartanyan, 2007).
Bison in Alaska and the Yukon, and horses and musk ox in northern
Siberia lived throughout the Holocene (Stephenson et al., 2001;
Sher et al., 2005).
C data and pollen records indicate that climate
warming during the deglaciation in Alaska was accompanied by an
increase in the productivity of grasslands, and in the density of
herbivores (Guthrie, 2006). Data (Guthrie, 2006) indicate that only
after the appearance of man, w12,370
C years ago, did the number
of animals decrease, and only afterward did the area of grassland
begin to decline. This can be considered as a proof for Overkill
Hypothesis (Martin, 1967).
In contrast to Climatic Hypothesis we propose an Ecosystem
Hypothesis (Zimov and Chupryninm, 1991;Zimov et al., 1995;
Zimov, 2005), which assumes that the mammoth ecosystem was
relatively insensitive to climatic variation and that numerous
animals maintained highly productive grasslands over a wide range
of climates. Under such a strong disturbance regime, mosses and
shrubs were trampled, and highly productive actively transpiring
graminoids and herbs dominated (Zimov and Chupryninm, 1991;
Zimov et al., 1995;Zimov, 2005). During the PleistoceneeHolocene
Transition (PHT) the rise in precipitation was accompanied by
increased temperature, so climatic aridity did not change
substantially. The Ecosystem Hypothesis proposes: “In some places,
such as sandy and stony ground, trees and shrubs would have
appeared. This might have caused changes in the relative propor-
tion of horses and moose. But overall, if climate was the only
controlling factor, the total grassland productivity and the number
of herbivores should have increased in the Holocene”(Zimov, 2005,
p. 798). During the deglaciation, warming created more favorable
*Corresponding author. þ7 41157 2306 6.
E-mail address: firstname.lastname@example.org (S.A. Zimov).
Contents lists available at SciVerse ScienceDirect
Quaternary Science Reviews
journal homepage: www.elsevier.com/locate/quascirev
0277-3791/$ esee front matter Ó2012 Elsevier Ltd. All rights reserved.
Quaternary Science Reviews 57 (2012) 26e45
Author's personal copy
conditions for human survival in the north (Guthrie, 2006). As
a consequence of strong hunting pressures, the density of animals
became insufﬁcient for grassland maintenance (Zimov et al., 1995).
The resulting decline in abundance of animals would have reduced
forage consumption, causing an accumulation of surface leaf litter,
insulating the soil, and reducing summer soil temperatures. This
would have initiated a cascade of other ecosystem changes,
including a decline in productivity and transpiration, wetter soils,
and lower nutrient availability. These, in turn, would have altered
the competitive balance among species, promoting the growth of
mosses and shrubs and reducing the abundance of grasses. The net
effect would be a decline in forage quantity and quality, leading to
continued decline in animal numbers (Zimov and Chupryninm,
1991 ;Zimov et al., 1995;Zimov, 2005). An important implication
of the Ecosystem Hypothesis is that the grassland ecosystem could
be regenerated in the north if one could increase the density and
diversity of animals.
MS has no present-day analog. Many assume that MS was very
unproductive and that the carbon content of the biome soils was
very low (Cwynar and Ritchie, 1980;Adams et al., 1990;
Friedlingstein et al., 1998). However, MS soils have been preserved
in the permafrost of Siberia and Alaska (locally called “yedoma”).
MS soils are the largest reservoir of organic carbon in the past and
at present. If this permafrost thaws in the future, this reservoir will
be a strong source of greenhouse gases (Walter et al., 2006;Zimov
et al., 2006;Schuur and Abbott, 2011), and modern ecosystems
covering the yedoma will be disrupted by erosion due to ice-wedge
thawing. Ice wedges constitute half of the yedoma volume. After
thawing the surface will subside by w10e30 m, and grasslands will
appear on the fertile yedoma remnants. Therefore, the possibility
exists for reintroducing herbivores and for a revival of ecosystems
similar to MS (Zimov et al., 1995) that could slow permafrost
degradation and mitigate global warming (Zimov et al., 2012).
However, to assess future possibilities, MS trophic structure and
physiology must ﬁrst be known.
2. Chronological support for the ecosystem hypothesis
Recent publication of many new
C dates of animal fossils from
Alaska and Yukon provide an opportunity to clarify the relative
chronologies of animal extinction, vegetation dynamics, and
human colonization (Guthrie, 2006). In this section we compare
these data with chronologies assumed by the Ecosystem and
Climatic Hypotheses. We presented the data in a form of probability
C data is presented as a bell-curve with a 400 yr
base). The total
C data density curve is sum of area of all the bell-
curves (Fig. 1).
Greenland ice cores show abrupt temperature and precipitation
ﬂuctuations at the PHT. At 14,650 calendar yr BP (approximately
CyrBP,Reimer et al., 2004) the average Greenland
temperature rose sharply (up to Holocene levels), and precipitation
doubled (the Bolling Warming Event, BWE). During the Younger
Dryas (11,600e12,800 calendar years ago) climatic parameters
returned to their initial state, and at the end of the Younger Dryas
temperatures again rose sharply (Severinghaus and Brook, 1999).
Similar climatic dynamics were recorded at both high and low
latitudes of the northern hemisphere (Severinghaus and Brook,
1999). Close dynamic should also be recorded for Alaska.
If the Climatic Hypothesis explained vegetation and animal
dynamics, steppe-like vegetation would have been replaced by
tundra vegetation during the BWE, and the steppe would have been
reestablished during the Younger Dryas, when the climate returned
to glacial conditions. But in fact, Guthrie (2006) data indicates
opposite. Glacial climate was not favorable for animals in Alaska
and the Yukon. During the LGM nine animal species went extinct
there (Guthrie, 2006). When the climate warmed from 15,000 to
12,400 radiocarbon years BP there was a substantial rise in abun-
dance of grasses and sedges, and animal densities increased (Fig. 1).
Pollen inﬂux of Artemisia, a drought-adapted species, rose even
more than the inﬂux of other species (Fig. 1). This suggests that
aridity did not decrease at the PHT.
Fig. 1. Upper part: generalized pollen record, density of
C dates of animals and human
evidence for Alaska and Yukon (from Guthrie, 2006). Lower part: density of
C dates of
mammoths for Europe, southern Siberia and China (Vasil’chuk et al., 1997;Kuzminet al.,
2001), the entire Siberian Arctic (Vasil’chuk et al., 1997;Kuzmin et al., 2001;Sher et al.,
2005), and the Taimyr Peninsula (Sher et al., 2005); snow accumulation for Greenland
(Severinghaus and Brook, 1999). Vertical black lines represent Younger Dryas bound-
aries based on IntCal04 (Reimer et al., 2004); the red line is a corrected (elk peak
correction) Bowling radiocarbon boundary. (For interpretation of the references to color
in this ﬁgure legend, the reader is referred to the web version of this article.)
S.A. Zimov et al. / Quaternary Science Reviews 57 (2012) 26e45 27
Author's personal copy
The BWE did not affect plant species composition but affected
animals (Fig. 1). Data on Fig. 1 indicates that strong warming and
increases of snow depth during the BWE was not accompanied by
the extinction of any species. On the contrary animal populations
grew. During glaciations the elk (Cervus) population was very small
C dates, Guthrie, 2006). This population increased
sharply by w12,600
C yr BP. The Elk peak is statistically the most
reliable (50 dates). Its right slope is very steep. The
represented by this slope has a standard deviation of 90e240
years (Guthrie, 2006). We did additional smoothing (200 yr) in
building the ﬁgures of
C date density. Thus in reality this slope is
almost vertical ethere was a sharp population rise. Reintroduction
of musk-oxen to Wrangel Island is a contemporary example of such
a rise. Their population grew 100 times in 35 years (Vartanyan,
Approximately 12,400 years ago an abrupt change occurred:
horses vanished, bison disappeared for over 300 years, and
mammoth and elk populations began to decrease sharply (Fig. 1). It
is clear that this abrupt faunal change was not triggered by vege-
tation change, because vegetation was stable during this period
(Fig. 1). The only thing corresponding with these massive extinc-
tions was the ﬁrst evidence of humans (12,370
C years before
present) (Fig. 1). It therefore appears plausible that a relatively
small number of hunters triggered the collapse of several herbivore
populations (Martin, 1967).
After the ﬁrst appearance of humans, some of the species
recovered their populations (which is in agreement with modeling
results, Alroy, 2001), but human population continued to increase
and this led to herbivores’extinction. It was only well after that that
pastures degraded and dwarf betula appeared (Fig. 1). The
appearance of moose (Alces) at 12,200 radiocarbon years BP closely
corresponds with a decrease in Salix and rise in dwarf Betula, which
is not the usual forage for the moose. One possible explanation for
this shift is that, as long as there were abundant bison, elk, and
mammoth, which actively ate willow sprouts, willow shrubs would
not have grown above the snow cover. Moose (tallest hoofed
animal) may have appeared in large numbers only when pop-
ulations of other animals had decreased, creating a new feeding
nicheetall willow shrubs.
The increase in
C dates of mammoths at the BWE can also be
observed in Europe and the Siberian north (Vasil’chuk et al., 1997;
Sher et al., 2005)(Fig. 1). During the Holocene the mammoth
population on Wrangel Island increased substantially. There were
ﬁve-fold more Holocene dates there (3730e7710 years BP range)
than Pleistocene dates (Vartanyan, 2007). This occurred despite the
fact that, in contrast to Alaska where loess accumulated throughout
the Holocene (Muhs, 2003), in Siberia sedimentation stopped
(Schirrmeister et al., 2002;Sher et al., 2005), and most of the bones
remained on the surface, where they would have been destroyed by
3. The mammoth steppe climatic envelope
The essence of the Climatic Hypothesis is as follows:
Mammoths, other extinct animals, and their pastures (i.e., the
mammoth steppe) required a certain range of climatic parameters
that deﬁned their climatic envelope. In the Holocene, climate
changed radically, and territories with climate suitable for the
mammoth steppe disappeared throughout the planet. To further
assess the Climatic Hypothesis, we deﬁne the mammoth steppe
climatic envelope, i.e., the optimal and peripheral climatic space of
the mammoth steppe.
Here we discuss two climatic characteristics eheat and mois-
ture, as aridity level is determined by these two factors. We char-
acterize moisture by annual precipitation (P) and heat by annual
radiation balance (R), which correlates closely with mean summer
temperature. We consider a two-dimensional climate space with
these two parameters (coordinates). Aridity can be assessed using
Budyko’s radiation aridity index (Budyko, 1984), which is the ratio
of R (in this case energy received by the landscape) to the energy
needed to evaporate an amount of water equal to P. If this index is
less than 1, the climate is humid; if greater than 1, the climate is
arid: with values between 1 and 2 equivalent to steppe, 2e3 semi-
desert and >3 desert environments (Budyko, 1984).
Such a two-dimensional climate space is illustrated in Fig. 2,
which shows the most important boundaries for desert, arid and
humid climates (Budyko, 1984). Beside those we show the
approximate position of three additional boundaries: the snow line
where R equals the energy needed to melt an amount of ice (snow)
equivalent to annual precipitation; the northern (altitudinal) forest
border; and the boundary of polar desert. Vertical movement in this
climate space is equivalent to latitudinal movement in real
(geographical) space. Movement to the right of this climate space is
equivalent to movement from the ocean to the continental interior.
Movement to the upper left quadrant is equivalent to moving
upward in elevation: colder with more precipitation.
On the basis of this envelope, we make 5 arguments: (1) Similar
to geographical space, in climatic space the mammoth ecosystem
should have an optimum zone and a peripheral zone where
conditions for survival are sub-optimal; (2) The mammoth steppe
existed in different climates on a huge territory for a long period of
time. Mammoths lived simultaneously in Spain, England, Mongolia,
China and arctic islands (Vasil’chuk et al., 1997;Kuzmin et al., 2001;
Sher et al., 2005;Alvarez-Lao and García, 2011). It is obvious that
climate differed between these regions, therefore, it can be
assumed that the mammoth steppe climatic area was spacious; (3)
If mammoths became extinct in the Holocene everywhere because
of climate then there is currently no place on the planet where the
climate would still be suitable for the mammoths. It then follows
Fig. 2. Climatic space with the most important landscape boundaries shown. The axis
R could also represent average summer temperatures (20 kcal/cm
corresponds to 10 C). Black dots are the meteorological stations of northeast Siberia
(see Fig. 3). hrepresents Holocene and modern horse grasslands in Siberia; Mand B
are Holocene grasslands of mammoths and Siberian bison. Brown cross-hatching lines
represent Holocene natural bison habitat of interior Alaska and northwestern Canada.
Question mark (‘?’) is the presumed locations of the mammoth steppe climatic
envelope that are consistent with the Climatic Hypothesis. The red solid line is our
estimate of the boundary of the continuous climatic envelope of mammoth steppe.
Dashed blue lines enclose areas where climatic trajectories lie, during the Last Glacial
cycle, in 3 geographical locations: London, Ulan-Bator and top of the 1700 m moun-
tains near the Oymyakon region. Today’s snow line of the Oymyakon Mountains is
situated on 2300 m. (For interpretation of the references to color in this ﬁgure legend,
the reader is referred to the web version of this article.)
S.A. Zimov et al. / Quaternary Science Reviews 57 (2012) 26e4528
Author's personal copy
from the Climatic Hypothesis that mammoths were living under
a unique combination of R and P, which are not met anywhere
today; (4) It seems that the mammoth steppe optimum must be
situated in the colder part of the steppe sector (Fig. 2), and their
range should move smoothly into temperate climate steppe; (5)
The range of climatic optimum for mammoth steppe was approx-
imately 150e300 mm of precipitation and 8e10
temperatures, which corresponds to climatic reconstructions for
Eastern Europe, where mammoth ecosystem received annually
250e300 mm of precipitation (Velichko and Zelikson, 2001).
Many believed that in today climate of the north-east Siberia is
humid, but we showed that in fact it is arid (Zimov and
Chupryninm, 1991;Zimov et al., 1995). Fig. 3 shows the meteoro-
logical stations of north-eastern Siberia, where radiation balance
observations are conducted (all circumpolar, including islands and
continental); these are also shown on Fig. 2 with black points. We
see that even though this region has abundant lakes and wetlands
its climate is arid. The most arid part is in the Pole of Cold in
Oymyakon (Budyko’s index ¼3.28; the most right and lowest black
dot on Fig. 2), and most humid is in mountains near the glacier, not
far from Oymyakon (Budyko’s index ¼0.1; the most upper and left
dot). We can use another known climatic index, in which potential
evaporation is calculated by summer temperatures, but we get the
same results ethe climate is arid (Sokolov and Konyushkov, 1998).
Around 400e500 mm of precipitation evaporates from lake
surfaces or high productive grasslands in these territories; this is
twice precipitation this region receives (Zimov and Chupryninm,
1991 ;Zimov et al., 1995). Lakes persist only because of drainage
from slowly transpiring forests and tundra. In Central Yakutia vast
steppe territories grazed by semi-wild Yakutian horses do not have
any river drainage (Pavlov, 1984). Nevertheless, proponents of the
Climatic Hypothesis suggest that the climate of northeast Siberia
and Alaska is too wet for the mammoth steppe (Guthrie, 1990;
Vereshchagin and Tikhonov, 1990;Velichko and Zelikson, 2001;
Sher et al., 2005). If correct, the climatic envelope of the mammoth
steppe would be restricted to a small area between the upland cold
deserts of Tibet and the polar deserts of the Canadian Arctic
(indicated with a red question mark in Fig. 2). The Climatic
Hypothesis assumes that this small envelope occupied a huge
territory of Eurasia and Northern America during both the LGM and
the BWE. This climate space was characterized by only w10 0 mm of
precipitation, raising questions of how the glaciers of Europe could
have developed and persisted.
The last mammoth refuge on the continent was the north of
Taimyr Peninsula, where they persisted until the Holocene (Sher
et al., 2005). If the Climatic Hypothesis holds, in the Holocene the
climate of Taymir (77
North) became too warm for the mammoth,
and the entire climatic envelope of the mammoths should be sit-
uated in even more severe conditions (polar desert?). However
mammoths lived in the south of the Iberian Peninsula (Alvarez-Lao
and García, 2011), which would suggest that summer there (37
was colder than in Taimyr today (i.e., 25e30
C colder than today).
This is not possible. Evidently, the Climatic Hypothesis as currently
stated (i.e., “colder and dryer”) is unrealizable. The mammoth
steppe could exist at 100 mm precipitation and with northern
Taimyr summer temperatures, but it’s unlikely that this repre-
sented optimal conditions.
It is possible to amend the Climatic Hypothesis by assuming that
R and P changed very quickly, and that every region had its own
subspecies of mammoth that were adapted to local conditions.
Hence mammoths wouldn’t have had time to adapt to ﬂuctuations
or migrate to more suitable regions. However this would not
explain why in Europe, Northern Siberia and America mammoths
sustained climatic jumps of BWE yet vanished under relatively
Fig. 3. Budyko’s radiation aridity index in different locations north-eastern Siberia.
S.A. Zimov et al. / Quaternary Science Reviews 57 (2012) 26e45 29
Author's personal copy
stable climate signiﬁcantly later (see Fig. 1). A better explanation is
that the mammoth steppe did not disappear due to climate.
4. Climatic envelope of mammals of the mammoth steppe
In this section we revisit the numerous radiocarbon dates for
bones of animals which used to live in the mammoth steppe
ecosystem. Many of these bones date from the Holocene and
therefore experienced a climate similar to the modern climate. By
looking on climatic maps at the modern climate (R and P) of these
locations, we can directly estimate the climate experienced by
Holocene populations of these species. These points in climate
space are shown in Fig. 2.
The lack of horses in Alaska during the Holocene has been lead
to suggestions that the modern Alaskan climate is not suitable for
them (Guthrie, 2006). However, in Siberia, where the climate is
similar, horses persisted well into the Holocene (up to 2200
years BP) in the arid far north (Lena delta and New Siberian Islands)
and in humid climates (the Taimyr) (Sher et al., 2005). Modern
semi-wild Yakutian horses extend up to the tundra region and also
on the “pole of cold”in the Oymyakon region (Fig. 2).
With bison, the situation is the reverse, with only one Holocene
bison date in northern Siberia (9300
C years BP) (Sher et al.,
2005), but extensive bison distributions in interior Alaska and the
Yukon throughout entire Holocene. They occurred along the arctic
coast as far north as Victoria Island and as far south as southern
Alaska near Anchorage (Stephenson et al., 2001). The climatic
envelope of the American bison is shown in Fig. 2 with a dotted line.
Reindeer now live on the far north and in Mongolia. In historical
times their southern boundary passed through Germany along the
steppes of Eastern Europe (Syroechkovskii, 1986). Therefore the
climatic envelope of reindeer occupies almost the entire climate
space of Fig. 2. In many regions of Siberia and America musk-oxen
have lived until historical time (Sher et al., 2005), and as a result of
recent reintroductions they currently live in Norway, Siberia, and
We have no Holocene dates for the wooly rhinoceros, but their
bones tend to occur in regions that are more arid than those of
mammoths. For example, the northwestern portion of Eurasia to
the north of 62
N including Taimyr is a humid part of the mammoth
steppe. Many mammoth remains were found there but no rhino
(Garrut and Boeskorov, 2001). On the other hand, wooly rhinos
were common in the most arid regions of mammoth steppe, south
of Central Siberia and Mongolia (Garrut and Boeskorov, 2001),
where mammoths are rare (Kuzmin et al., 2001).
Mammoths existed in the arid zone in the Holocene (Wrangel
Island, with the most recent date of 3730
C years BP) (Vartanyan,
2007), in the humid zone on the north of Western Siberia (Gydan
C years BP), in the northern Taimyr (9670
years BP) (Sher et al., 2005), in the very humid St. Paul Island in the
Bering Sea (5700
C years BP) (Guthrie, 2004;Veltre et al., 2008),
and on the coast of Gulf of Finland (9780
C years BP) (Vasil’chuk
et al., 1997). While the early Holocene climate may not have been
the same as today; it is indisputable that on Wrangel Island and on
St. Paulo Island mammoth lived in late Holocene climate.
On the climatic space (Fig. 2) these sites are situated very far
from each other. Wrangel Island is dry polar desert, and St. Paulo is
a very humid climate with no permafrost. And the fact that
mammoth lived on these two islands genetically isolated for the
long time suggest that the mammoth’s climatic envelope was very
C data indicated that even during the LGM mammoths
lived on the archipelago of New Siberia (79e80
N). They also lived
there in the beginning of Holocene (Sher et al., 2005). On Wrangel
Island mammoths lived both in the LGM and in historical time
(Vartanyan, 2007). In the present climate, July temperatures on the
islands are approximately 0
C. Today’s vegetation would not feed
a mammoth population. The essence of the paradox is that in the
current climate, the same place can be a desert (in the absence of
animals) or grassland (if animals are present). Modern vegetation of
Wrangel Island can’t be considered an analog of mid-Holocene
vegetation. In the middle of the twentieth century Wrangel Island
was inhabited only by lemmings. At that time it was estimated that
the island was capable of sustaining up to 1000e1500 reindeers
(Vartanyan, 2007). This is approximately 100 tons of zoomass,
which is equivalent only to tens of mammoths. In 1948 and 1952,
150 domestic reindeers were introduced. By 1975 their population
reached 5000. In the same year 20 musk oxen were introduced (half
of them have survived), and the shooting of reindeer began. Finally
in 1980 only 113 reindeer were left. After that hunting decreased,
and the reindeer population reached 5000e7000 (Vartanyan,
2007). Also there are now more than 1000 musk oxen on the
island (Vartanyan, 2007). In summary, during the last 50 years
ungulate biomass has exceeded intended densities by almost by an
order of magnitude. We suggest this happened because the
biomass of forage has increased as herbivores maintain their
grasslands. This is only the beginning of ecosystem succession. In
modern arctic tundra in places with high animals density,
productivity increases by an order of magnitude, and the ratio of
uneatable vegetation sharply decreases (Chernov, 1985;Zimov and
Chupryninm,1991;Zimov et al., 1995)(Fig. 4). The potential pasture
productivity on Wrangel Island is shown in Fig. 5. Several
mammoths could live on 1 km
of such pasture. Biological
productivity in the north is limited less by photosynthesis than by
nutrients (Chapin et al., 1995). Nutrient cycling is limited by slow
decomposition and nutrient release from soil organic matter. Only
in the warm stomachs of animals can this process be substantially
In summary, the animals of the mammoth steppe occupied
a broad climatic envelope. Similarly the distribution of modern
tundra does not depend on aridity but occurs in sites ranging from
50 to 1000 mm of annual precipitation. Based on the above infor-
mation, we tried to reconstruct the boundary for a continuous
climatic envelope for the mammoth steppe (Fig. 2). In geographical
areas that now lie inside the envelope, the mammoth steppe
ecosystem would not vanish because of R and P changes. The next
important question is whether animal population densities in the
mammoth steppe were high enough to prevent expansion of moss,
shrubs, and trees.
5. Density of the mammoth skeletons in the Duvanniy Yar
Animal density for MS can be estimated from the number of
bones found in permafrost. However, only a hundredth or even
a thousandth of the bones are preserved in permafrost; from most
of the skeletons, only one bone or nothing is preserved (Guthrie,
1990;Sher et al., 2005). Our studies conﬁrm this conclusion.
We reconstructed animal densities by several methods for
different places. We have done this most precisely based on data
collected at Duvanniy Yar (DY) in the Kolyma River lowland
E) (Fig. 6), the largest yedoma exposure
(Vasil’chuk et al., 2001;Zimov et al., 2006)(Figs. 7e9). The main
exposure at Duvanniy Yar is 10 km long and 35e50 m in height. We
roughly surveyed this exposure twice annually (for a total of more
than 50 visits).
The bone concentration in DY yedoma (as for most other
yedoma and loess) is small, only one bone (or bone fragment) for
each w500 m
(Fig. 10). Therefore it is complicated to ﬁnd bone on
S.A. Zimov et al. / Quaternary Science Reviews 57 (2012) 26e4530
Author's personal copy
cliffs (exposures) and in order to see it a person would need to walk
along the exposure slope for hundreds of meters.
Average annual commercial collections of mammoth tusks from
DY are about 250 kg/yr, and range from 70 to 450 kg/yr, Due to the
length of the exposure (10 km) and the rate of erosion(2.5 m/yr), the
density of tusks is equal to 10 tons/km
. Relatively intact tusks
represent w40% of commercial tusk collections. The average weight
of a tusk in this region is 25 kg (39 kg for males and 11 kg for females)
(Vereshchagin and Tikhonov, 1990). From that follows that
frequency is 4 tons of intact tusks per km
(160 intact tusks/km
we use an average thickness for the yedoma of 50 m, the frequency is
1 tusk/312,500 m
. The length of one tusk is w1.5 m. Tusks can lie in
the permafrost parallel to the exposure front. So, the average soil
thickness, from which a tusk would stick out and be visible is w1m.
Therefore, on the slope of an exposure one tusk should be found in
an area of 312,500 m
. If a man examined a 20 m tall (width) cliff
(slope), then ﬁnding one tusk would require that he walk w16 km
along the cliff. Duvanniy Yar cliffs and steep slopes usually represent
only 20% of the exposure; the rest are screes that are mostly over-
grown by tall grass. Therefore, ﬁnding tusks in situ in permafrost at
Duvanniy Yar requires that the area be investigated several times
(within an interval greater than a week) throughout the entire
exposure. The unique feature of the Duvanniy Yar is that it provides
opportunity to observe all bones from the melted yedoma. Obser-
vation of all bones is possible only on the exposed shores, where all
sediments and bones slide or crumble. Sediments are washed from
the area by waves and streams, but bones are left on the beach. If the
water level decreases rapidly after a strong storm, and if the mud
slides are not too extensive and do not cover the washed-out bones,
all the bones are well exposed on the surface of the beach.
Fig. 4. (a) Tundra vegetation close to the mouth of Kolyma river, (b) same site, meadow developed 2 years after reindeer breeders camped during summer.
S.A. Zimov et al. / Quaternary Science Reviews 57 (2012) 26e45 31
Author's personal copy
Bones are initially deposited at the location of an animal’s
death. Predators can disperse some of the bones, but not tusks and
molars of mammoths. On the large beaches at DY, mammoth
bones occur in groups several tens of meters from one another and
belong to one animal (one skeleton), making it possible to directly
calculate the density of mammoth skeletons. Accumulations of
mammoth bones rarely (randomly) coincide with the accumula-
tion of other species or of another mammoth, indicating that bone
accumulations do not result from physical processes. Such
processes would cause bones to gather from different skeletons
into one accumulation. On average, each accumulation is 8e10
bones. Fragments of tusks and (or) molars present in the most of
Unique conditions for skeletons counts occurred in 1998 on the
western part of the exposure, when mudﬂows subsided and waves
washed away silt, creating a clear bank approximately 10 m in
Fig. 6. Map of Duvanniy Yar and Pleistocene Park locations.
Fig. 5. Wrangel Island is the last mammoth refugium. The picture shows a grassland near the Ushakovskayaweather station, where the average July temperatureis 1 C. The unique
site is characterized by high animal nitrogen inputs. It is middle of September, but photosynthesis continues.
S.A. Zimov et al. / Quaternary Science Reviews 57 (2012) 26e4532
Author's personal copy
Fig. 7. Part of Duvanniy Yar, where the ice-wedge thaw is active. In June the river water level is high and the beach absent.
Fig. 8. The most of time, the Duvanniy Yar shore is covered with mud. In the autumn of 2007 only this small section was clear.
S.A. Zimov et al. / Quaternary Science Reviews 57 (2012) 26e45 33
Author's personal copy
width by 2.8 km in length (Stone, 2001). Along this bank, we found
29 mammoth bone accumulations (3e14 m across) each of which
had 3 to 16 bones. Nineteen of the accumulations had fragments
of tusks and/or molars. If each bone accumulation represents the
remains of one mammoth, the area has a skeleton frequency of
(29 skeletons/(0.01 km 2.8 km)).
Favorable conditions occurred again in 2009. Over the extent of
the whole exposure a narrow (4e6 m), well-washed erosion terrace
appeared. On this terrace, in the central part ofDY over a distance of
1.96 km, we counted 160 mammoth bones and fragments. Four
bones were single. Five accumulations contained two bones. In 26
accumulations we found 3e8, 12, and 17 bones. In 18 of the accu-
mulations we found fragments of tusks and/or molars. The average
size (diameter) of the accumulations of mammoth bones on the
beach of DY was w10 m, which is twice the width of the well-
washed terrace (4e6 m) on which we counted mammoth skele-
tons in 2009. Even if the center of accumulation rested inside the
width of the terrace, only part of the bones from the accumulation
would be on the terrace. The terrace must also contain bones from
accumulations whose center rests beneath the terrace to a distance
of up to 5 m. We found a few single bones and small accumulations,
and most of them must represent accumulations with centers sit-
uated beyond the investigated terrace.
To calculate skeleton frequency, we either needed to exclude
single bones and small accumulations, as was done for data collected
in 1998,or increase the terrace widthfrom 5 to 15 m in order to ﬁtthe
centers of all of the detected skeletons into the terrace. Using the
latter method we determined a skeleton frequency of 1170/km
(35 skeletons/(1.96 km 0.015 km)). In addition, in the same year in
the westernpart of the terrace along a 1.74-km transect we collected
156 mammoth bones e4 single bones, 6 double bone accumulations,
and 26 accumulations that contained 3e7,10, and 19 bones. In 13 of
the accumulations, we found fragments of molars and/or tusks. For
this terrace, mammoth skeleton frequency was w1380/ km
(36 skeletons/(1.74 0.015 km).
The third time favorable conditions were in August of 2011, in
the central part of the DY. The width of the washed shore was on
average 8 m. There on a transect of 1570 m we found 169 mammoth
bones and bones fragments. 5 bones were singular, 4 accumula-
tions with 2 bones, and 32 accumulations with 3e10 bones or their
fragments. In 29 accumulations were fragments of molars or tusks.
Accepting that each accumulation or singular bone is a single
skeleton (41 units) and increasing width of the shore by 10 m
(8 þ10 m) yields a skeleton frequency of 1450 per km
.OnFig. 9 we
show 4 accumulations with mammoth jaws. In 2009 on this tran-
sect we found only 1 jaw. Note that on this photo bones of other
animals also appeared.
Our estimate of skeleton densities based on 2009 and 2011
collections is higher than our estimate based on 1998 collection.
One of the reasons for that is that in 1998 small bones and frag-
ments we did not try to detect, focusing only on the big bones.
Some skeletons (especially of young mammoths) we likely missed.
In more recent years we tried to detect all bones. From the available
4 estimates (1030, 1170, 1380 and 1450/km
) we accept the most
conservative e1000 skeletons/km
. This is an average frequency of
skeletons of adult animals. As with elephants, most dead
mammoths would have been young mammoths which would often
have been smaller than bison. Their bones were soft and most
frequently nothing would have been preserved in the permafrost
even from an entire skeleton. There are two more methods for the
mammoth skeleton density estimation.
The density of commercial tusks from DY is 200 pair per km
Most of the tusks preserve only as small fragments. If we assume
that only 20% of tusks (by weight of all tusks) are preserved in
Fig. 9. August 2011, the central part of the Duvanniy Yar. In this image we show bones from all four accumulations with mammoth jaws.
S.A. Zimov et al. / Quaternary Science Reviews 57 (2012) 26e4534
Author's personal copy
permafrost, then we obtain 1000 skeletons/km
, the same density
that we calculated from mammoth skeleton density.
On average, only ten bones were preserved from a mammoth
skeleton. This was also tested by observations of skeletons thawing
out of the permafrost in situ and exposing mammoth bone accu-
mulations in permafrost using a motorized pump. Knowing the
average number of bones in an accumulation and knowing the total
number of bones in yedoma, one can estimate skeleton frequency.
This was an additional method that we used.
6. Skeleton density of ungulates and predators
Bone accumulations of smaller animals are rarely found.
Therefore for the skeleton density estimation we used the histo-
gram method, the method of probability of bone preservation, and
method of paired bones.
In 20 09, we collected 204 bison bones, 180 horse bones, and 78
reindeer bones on 0.087 km
of the well washed terrace in the west
part of DY beach. The number of single bones of bison was 80,
horses 101, and reindeer 69. Histogram for the bones in accumu-
lations for these animals and mammoth showed on Fig.11. The
examined terrace width is half the width of the averagediameter of
an accumulation; therefore the number of bones in each accumu-
lation, as recorded by this study, is smaller than the real number of
bones that probably persisted in this accumulation. Some single
bones and small accumulations from our collection probably
belonged to larger accumulations whose centers lie beyond the
width of the terrace and hit the terrace only at its edges. If we were
to count the number of bones in accumulations in this wider
concept of a terrace it would result in an increase in the number of
large accumulations on the histogram and a decrease in the
number of small accumulations and single bones ethe histogram
“bell”would move to the right. On the histogram for mammoths,
the maximum would most likely shift to a value of 10 and occur-
rences of one or two bones would almost disappear. The number of
single bones in histograms for ungulates would also decline.
Fig. 10. Bones of mammoths (upper left), reindeer (upper right), bison (lower left) and horses (lower right) collected on the shore of Kolyma river on the western part of Duvanniy
Yar exposure on the area of w1 ha. Totally there were found w1000 bones and bone fragments. From that follows that bone density on the site is 1 bone per 10 m
, or (taking into
account, that exposure height is 50 m) 1 bone per 500 m
. Bones are positioned to be consistent with their place in the skeleton.
S.A. Zimov et al. / Quaternary Science Reviews 57 (2012) 26e45 35
Author's personal copy
However, since the number of single bones for these species
substantially exceeded the number of multi-bone occurrences, only
a minor part of all of the single bones could have represented the
edges of larger accumulations. So, their distribution maximum
(maximum on the histogram) would still remain near one.
If bones occurred within a homogeneous region with a stable
climate, then the distribution of numbers of bones in an accumu-
lation would be close to the Poisson distribution e“Poisson bell”.
However, there are likely to have been periods when due to climate
that was unfavorable for preservation, on average for example only
0.3 bones per skeleton persisted, and periods with more favorable
conditions during which 2 bones persisted per skeleton. Alterna-
tively, there may have been swampy areas where several bones
were preserved from each skeleton. The actual histogram of the
distribution will in this case be determined by summing the various
bell curves, so the recorded distribution would likely have been
broader, with less pronounced maxima as a result of these sources
of variability. Nevertheless, the maxima in our histograms are very
sharp, which indicates, with a high degree of reliability, that a high
portion of all skeletons do not preserve even a single bone, and on
average not more than one bone persists from each skeleton.
Predators could spread ungulates skeletons apart, which would
increase the number of single bones. However, single bones in the
yedoma are so abundant that, even when accounting for predators,
it is unlikely that more than one bone persisted from a bison
skeleton. The horse humerus bone type has more signs of predator
teeth than other bone types (Figs. 10 and 12). Among horse and
bison bones, for this speciﬁc example only, we can reliably say
those humeri were transported away from the skeleton by preda-
tors. The horse humerus was the only bone that was not present in
accumulations, and that was found only as a single bone. Other
bone types can be found in accumulations and as singles in equal
At low temperatures, corpses without skin freeze (becoming
covered by a frozen layer) faster than the body is eaten. Therefore,
approximately 50% of all body meat in the mammoth steppe was
eaten in the frozen state, and mammoth bodies were eaten almost
entirely frozen. In the north it is difﬁcult and useless for predators
to bury bones in the ground since for nine months of the year the
ground is frozen and bones cannot be buried nor dug out of the
frozen ground. These factors decrease the ratio of single bones in
A clear maximum for single bones on the frequency distribution
allowed us to determine that, on average, from each skeleton of
a hoofed animal, only one bone or less persisted. From a random
number of bones in an accumulation (Poisson distribution), if single
bones were twice as abundant as double bones in an accumulation,
then, on the average, from one skeleton one bone should persist; if
single bones are 5 times more abundant than double, then w0.4
bones should persist on average. The histogram of the bison bone
distribution (Fig. 11) shows that the single bone numbers exceed
the number of double bones approximately two-fold. Therefore, on
average, only one bone persisted from each bison, and the skeleton
density is 22,000/km
(204 bones/(1.74 km 0.005 km)). The ratio
of single horse bones is higher than this ratio for bison, so horse
skeletons should be more abundant. Reindeer skeletons which
performed mostly by single bones, in turn, should be several times
In order to obtain a more precise estimate of hoofed animal
density, we used full bone collections gathered in 2009 and partly
in 2007 over the area of w0.01 km
. The collection is shown in
Fig. 10. Not shown are 23 woolly rhinoceros bones (that were found
in 6 accumulations), 4 musk ox bones, 3 elk bones, 2 moose bones,1
hare bone, 4 wolf bones, 2 cave lion bones, and 61 unidentiﬁed
mammoth bone fragments. Many small bone types in the collection
are absent or only represented by one or two bones. For example, in
the entire collection there is only one patella, and numerous bones
of mammoth digits are represented by only one phalanx. No caudal
vertebra from ungulates occurred in the collection. The ribs are the
abundant bones in the skeleton, but they are not well represented
in the collection. On the other hand, it is the massive bones which
dominate in the collection.
To compare the skeleton density of different species, we used
the relative probability of bones to persist (be found) depending on
bone weight. To calculate the relative preservation of bones of
Fig. 11. Histogram for the number of bones in accumulations (red) and for the number
of accumulations (blue) in the western part of the beach at Duvanniy Yar. For the
mammoth histogram we also used data from accumulations from the central part of
Duvanniy Yar. (For interpretation of the references to color in this ﬁgure legend, the
reader is referred to the web version of this article.)
S.A. Zimov et al. / Quaternary Science Reviews 57 (2012) 26e4536
Author's personal copy
various species (the probability of obtaining the bones in the
collection), we calculated the number of bones or their large frag-
ments (over 10% of boneweight) of one type (group) in a collection,
and divided it by the number of bones of this type in the entire
skeleton for the species. If the collection contained two parts of one
bone, we accounted for only one of them.
For the construction of graphs (Fig. 13), we only counted those
bones types for which at least three bones existed in the collection.
Anatomically similar bones for one species were combined
into general groups: all tarsals and carpals of mammoths, except for
the calcanea, which are presented separately; the cervical, thoracic,
and lumbar vertebrae; and the ribs. The large bones from
extremities were combined as follows: we combined all meta-
podia; we combined radius with tibia; we combined humerus with
femur; and we combined scapula with pelvis. The pelvis consists of
several fused bones. After an animal’s death, these bones quickly
disarticulated into relatively symmetrical and sturdy halves. We
assumed that the pelvis consists of two bones. Finally, most of the
points on the graph (Fig. 13) were based on several tens of bones
from the collection.
The probability that bones persist, depends not only on their
weight, but also on a bone’s shape. Long bones, such as ribs, are
heavier than many other bones, but they are thinner and are
therefore not well preserved. Therefore, to exclude the inﬂuence of
Fig. 12. All humerus bones of horses were gnawed from the upper end, as in the picture, by predators.
Fig. 13. Dependency of bone integrity on reduced weight. Bone integrity index is the ratio of the number of bones of a speciﬁc type in a collection (Fig. 10) to the number of such
bones in an entire skeleton of a species. 1 erib; 2 ecaudal vertebra; 3 emetapodial; 4 ecarpal and tarsal; 5 evertebra; 6 ecalcanea; 7 eradius and tibia; 8 emandible; 9 e
scapula and pelvis; 10 emolar; 11 ehumerus and femur; 12 ephalanx 1; 13 ephalanx 2; 14 eulna; 15 etalus; 16 ephalanx 3 (hoof). Red “þ”edata from the
C collection (Sher
et al., 2005). (For interpretation of the references to colour in this ﬁgure legend, the reader is referred to the web version of this article.)
S.A. Zimov et al. / Quaternary Science Reviews 57 (2012) 26e45 37
Author's personal copy
shape (by representing long bones as spherical) and to increase the
correlation between bone size and probability to persist, we did not
use the weight of the bone but instead used its ‘reduced weight’.If
we divide bone weight by the bone density we obtain the volume of
the bone. By dividing this calculated volume by the bone’s length,
we determined the average area of the cross-section. The square
root from this area is the average thickness of the bone. The cube of
the thickness provides a reduced volume, and multiplying this by
the bone density allowed us to determine the reduced weight. The
unit weight for all of the bones is similar, so to ease the calculations
for the reduced weight we used a simpler parameter e(weight of
bone (g)/length of the bone (cm))
. We calculated the reduced
weight based on ﬁve bones from each group (subgroup). For this we
took the largest bone, the smallest bone, and three of the most
typical bones for the collection. If a group consisted of two
subgroups (for example, humerus and femur), ten bones were
utilized, ﬁve from each subgroup. If the number of bones in the
group was less than ﬁve, we weighed all of the bones. If most of the
bones were bone fragments for which we could not reconstruct an
average bone thickness, we utilized analogous bones from other
collections; this mostly applied to large mammoth bones that were
largely disintegrated in our collection. For us it was more important
to determine the dependence of smaller mammoth bones e
calcanea and smaller, since only bones of this size had similar
reduced weights as bones of ungulates. Mammoth bones of this size
were mostly of good integrity, and it was not difﬁcult to calculate
their reduced weight, although we mention that the dependence
we obtained from this method for large mammoth bones was
similar to the range determined for small mammoth bones. Ob-
tained dependence between reduced weight and probability to
persist didn’tﬁt only the most edible bones: the vertebrae of
ungulates and the thin-walled humerus and the femur of reindeer.
We did not consider mammoth tusks and antlers of the reindeer.
A small number of small bones is associated not only with faster
rates of weathering, but also with the fact that these bones were
more likely to be eaten or broken by predators. Predators were
abundant in the mammoth steppe. Many bones from the collection
had signs of predator tooth marks (Figs. 10 and 12). Predators took
out marrow from many humerus and femur bones of bison and
horses, but large fragments of these bones persisted and were
present in the collection. The opposite occurred for the thin-walled
humerus and femur bones of reindeer. These bones are easily
gnawed into small pieces by predators and do not make it into the
collection. Mammoth vertebrae are relatively small bones, and their
reduced weight is two times less than the reduced weight of
a calcaneum. At the same time they hold a heavy load; therefore,
they are relatively strong. In the thin-legged skeleton of a reindeer,
the vertebrae are the thickest bones, and the main reservoirs of red
marrow, and are, therefore, friable (spongy). Predators easily crack
these bones, and gnawing them is not substantially more difﬁcult
than eating frozen meat. Therefore, these numerous bones (and the
back vertebrae of horses as well) are eaten almost entirely. The back
and neck vertebrae of a horse are approximately the same size and
weight. For processes of physical weathering, these vertebrae are
identical. If weathering was the only agent of destruction, then the
number of back vertebrae in the collection must have been 24:7
times higher than the number of neck vertebrae. However, in the
collection, the actual number is equal to 4:12. Back vertebrae have
a smaller probability of being preserved due to predators. In
Pleistocene Park (Zimov, 2005) wolves and wolverines gnawed the
backbones of horses frozen in the ice and their stomach contents in
less than a month.
In the past mammoth tusks were mostly collected on the bea-
ches, but during the last few years, a rise in tusk market prices
increased the collecting intensity on Duvanniy Yar and most of the
tusks were collected when they just appear on the exposure slopes
(not on the beach) so for our collection we only found small tusk
fragments from 11 mammoths. We have 46 kg of mammoth
humerus bones, in our collection. The reduced weight of tusks is
twice as high, so they should be twice as abundant as humerus
bones in the collection, w100 kg. By accounting for the fact that the
collection was gathered on a territory of w1 ha, the value is
equivalent to 10 tons/km
. The estimate corresponds to the tusk
frequency calculated from commercial collection data.
Reindeer antlers are the most numerous of bones in the
collection. However, it is difﬁcult to estimate the number of rein-
deer skeletons using these bones ereindeer shed antlers every
year, and one needs to accurately know the average life-time of this
species. A bone with an average size of 1 0.5 0.3 m is impossible
to cover with loess. Bones of this size were stepped on by animals
several times per year, so they were frequently turned. If the bones
lie on the surface and do not get covered by dust, they will weather
and disintegrate into smaller pieces. Only afterward can small
fragments be transported to the soil. All of the antler pieces in the
collection are weathered. The bones have little chance of being
preserved. Males shed antlers at the beginning of the winter, and
females at the end. From the availability of antlers in a pasture, one
can reconstruct the migration of reindeer. If a territory is utilized by
reindeer only as a summer pasture, then there should be no antlers.
Summer antlers are soft and are eaten in their entirety by preda-
tors. The fact that our collection contains the bones of males and
females, and the fact that some of the antlers are attached to the
skull could indicate that reindeer used the Duvanniy Yar territory
during most of the year.
Linear dependence was recorded for all species and holds true
for a range of bone weights from grams to kilograms (Fig. 13).
The number of bones in the collection is also proportional to the
number of dead animals. Assuming that all four graphs reﬂect the
same functional dependence on bone weight, we determined
a relative density of skeletons for four species (using the ratio
between equation coefﬁcients, Fig. 13). We determined (rounding)
that bison skeletons were 20 times more abundant than mammoth
skeletons, horse 30 times and reindeer 80 times. Since we know the
density of mammoth skeletons (1000 per km
), we can calculate
densities of other species. These estimates are close to those ob-
tained by using the average number of bones in the accumulations.
The obtained dependence (Fig. 13) can be interpreted as follows:
the number of persistent bones is approximately proportional to
the skeleton weight. A wolf skeleton is approximately two times
lighter than a reindeer skeleton. The number of reindeer bones in
the collection (excluding antlers) numbered 87, and wolf bones
numbered 4, indicating that the number of dead wolves were
approximately 11 times less than the number of reindeer (87/4/2).
Cave lions were twice as large as reindeer. Only two lion bones were
contained in the collection so their skeletons were 87 times less
abundant than the reindeer ((87/2) 2).
For rough estimate of animal density we used additionally
method of dual bones (tusks). Large fossil collections allow the
implementation of statistical methods to estimate the probability
of preserving bones. Many bones in the skeletons are paired (left
and right leg bones, tusks, etc.). The probability of preservation of
both left and right bones is the same (p). The probability that both
paired bones were preserved is equal to the probability one bone
was preserved squared (p
Since we know the proportion of the paired bones in the
), we can calculate the probability of preservation of
these bones (p). This transformation requires large enough sample
size, so that we had to combine all bones of our collection, pres-
ervation probability of which is comparable. We combined the
following large bones of legs of bison and horses: metapodial,
S.A. Zimov et al. / Quaternary Science Reviews 57 (2012) 26e4538
Author's personal copy
radius and tibia; humerus and femur; scapula and pelvis; thus
making 8 bones of horses and 8 bones of bison. The total number of
bones was 245. Among these in the accumulations of bison and
horses bones only four pairs of paired bones were found (8 bones),
thus making 3.3% of the entire collection. Which means that ratio
between probability of a pair to preserve to probability of a single
bone to preserve (p
/p), is 0.033 indicating that pis equal to 0.033.
Therefore, in order to estimate the number of dead animals in
a speciﬁc territory we need to divide the number of bones found in
a speciﬁc territory (245) by the probability that the bones are
preserved (0.033), and by the number of bones of the full skeleton
we used in the calculations (8 þ8¼16). At the end of all we get is
the density of skeletons of bison and horses 470 per ha or 47,000
. The value is close to what we have gotten in our previous
estimation (20,000 þ30,000).
This method can be used to assess the probabilities of preser-
vation of bones based on the number of paired halves of the bones.
The probability of preserving two similar-size parts of the bone is
equal. In the collection, among ungulate animal bones, many halves
of bones or fragments existed, but all of them belonged to different
bones, indicating a very low probability of small bone preservation.
Only for mammoth did we have different fragments of one bone.
7. Animal density in the mammoth steppe
Duvanniy Yar loess accumulated from 42,000 to 13,000 years BP
(there are 40
C dates from this exposure, Vasil’chuk et al., 2001).
In collections gathered from northern Siberia, w10% of all of the
mammoth bones were older than 45,000 years (Sher et al., 2005).
Therefore, we took a more conservative estimate that all bones
accumulated for w40,000 years. Average age (calculated by rings of
tusks) for a dead mammoth was 40 years (maximum age w80
years) (Vereshchagin and Tikhonov, 1990). We are making
a conservative evaluation of the density of dead mammoths 1000/
. From this we determined that the average mammoth density
was about 1/km
40 yrs/40,000 yr).
There were very few bones of young ungulate animals (Fig. 10);
they are relatively soft and therefore are eaten by predators. By
accounting for high predator pressure, and the maximum age of
modern animals, we assumed that the average longevity of mature
horses, bison, and lions was ten years, seven years for reindeer, and
ﬁve years for wolves.
Assuming that the density of the bison skeletons was 20,000 per
, horses 30,000 per km
, reindeer 80,000 per km
(11 times less than reindeer), lions 920 km
(87 times less
then reindeer), and assuming that the average weight of adult bison
was 600 kg, horse 400 kg (as recent Yakutian horse), reindeer
100 kg, lion 200 kg, wolf 50 kg, thus making the average animal
density per square kilometer of ﬁve bison (total weight of 3 tons),
7.5 horses (3 tons),15 reindeer (1.5 tons), 0.25 lions (0.05 tons), and
one wolf (0.05 tons). By adding the weight of the mammoths
(2.5 tons, Vereshchagin and Tikhonov, 1990) and the rest of the
more uncommon herbivores (0.5 tons), we calculated a total
herbivore biomass of 10.5 tons, enough to feed two wolves. All of
these estimates were averaged over a period of 40,000 yr.
Mammoth density dynamics in the end of Pleistocene and the
Holocene are illustrated on Fig. 1 (Kuzmin et al., 2001;Sher et al.,
2005;Vartanyan, 2007;Nikolskiy et al., 2009)(Fig. 14). In the
LGM, the number of mammoths was the lowest; it increased as
climate became warmer.
Are these density ﬁgures representative for the entire Siberian
During the time of yedoma accumulation, the territory of DY,
which was far away from mountains and hills, was a homogeneous
plane, and animal density in the area was likely determined by
forage availability. We explored approximately 200 river, lake, and
coastal yedoma exposures in the Siberian north. Our visual
assessment indicated that bone and tusk concentrations, in most
regions, are similar to DY eseveral bones per kilometer of transect
along the exposure slope and several tusks per 100 km. Low bone
densities are typical for sandy yedoma, which is likely related to
low pasture productivity during the Pleistocene, for example,
Aleshkinskaya Zaimka and Plakhinskii Yar on the Kolyma River, and
Sypnoi Yar on the Indigirka River. Above-average bone concentra-
tions were recorded for some exposures close to mountains. In
these areas, animals probably died more frequently in the depres-
sion along creeks. We are aware of two such exposures in the
watershed of the Filipovka River and one more in the Krasivoe
exposure on the Malii Anui River. The exposure is tens times
smaller than Duvanniy Yar, we rarely visit it. However, we found the
same number of in situ tusks as at Duvanniy Yar.
Some may assume that bone abundance in Duvanniy Yar is
related to a connection to the big river (animals die more often at
the watering place), but in the north the most difﬁcult season for
herbivores is winter and spring (highest death rates) and rivers are
frozen at that time (for more than 8 months). Additionally, the river
bed of the Kolyma was situated tothe north of Duvanniy Yar during
the Pleistocene. Up and down along the Kolyma River, there were
several exposures, and bone concentrations on the slopes and
beaches were similar to those found at Duvanniy Yar.
Commercial collection data indicates that the tusk content of the
yedoma of different regions is roughly the same (Boeskorov et al.,
2008). The number of collected tusks mostly depends on the
number of exposures in each region. Some assume that the
frequency of tusks in the New Siberian Islands (the main tusk
provider) is higher than on the continent. However, this
Fig. 14. Density of C14 dates of Siberian mammoths. “?”,“!”etiming of human arrival.
S.A. Zimov et al. / Quaternary Science Reviews 57 (2012) 26e45 39
Author's personal copy
assumption is not true. In the north, the ice content in the yedoma
is higher (up to 90%). Therefore, there are higher rates of erosion
and, as a consequence, rates of tusk appearance on the slope
(beach) are higher. Such exposures have insufﬁcient amounts of
mud to cover (hide) large tusks.
The uniqueness of DY lies not in the density of bones but in the
fact that sometimes all bones are visible. For minor rivers and lakes,
large exposures and beaches produced by active erosion are absent.
The distribution of fossils from different species on DY was very
similar to the distribution of bones obtained from w3000 bones
gathered in the entire region of northern Siberia (Sher et al., 2005).
So the species distribution on DY is typical of the Siberian north.
To additionally assess dependence of the mammoth bone
preservation on reduced weight we explored the entire list of bones
C dating (
C collection) in the entire Siberian north
(Sher et al., 2005). The majority of these dates were obtained with
a method requiring bone samples with weight over 1 kg (scintil-
lation method). Therefore this collection contained a few average
size bones and no small bones (Note that the reduced weight of ribs
is small but their overall weight is high.) For large mammoth bones
the dependence of bone integrity on reduced weight was very
similar (linear) to that obtained for Duvanniy Yar (Fig. 13). This
similarity suggests that the probability of bone preservation is
similar across yedoma lowlands. This relationship excludes only
those bones that are most “popular”among fossil collectors etusks
and mammoth mandibles.
A rough estimate of animal density can be obtained for other
northern regions. Another famous exposure of yedoma is at
Mamontovy Khayata on the Bykovskii peninsula, east of the Lena
River delta. During the joint RussianeGerman expedition to this
site about 900 fossil mammal bones were found, nearly 160 of them
within the main Mamontovy Khayata cliff.12 of them were found in
situ in the permafrost (Sher et al., 2005). The length of this exposure
is 600 m with a maximum height of 40 m. Accounting for bad
exposure (Sher et al., 2005), the total area of eroded yedoma surface
. If we assume that all this territory was surveyed ﬁve
times, then 12 bones were found on the area of 60,000 m
or 1 bone
per 5000 m
esimilar to Duvanniy Yar. Duration of sediment
accumulation there was also w40,000 yr (Sher et al., 2005).
The New Siberian Islands are the main provider of tusks. For
more than 200 years, every year, 10e12 tons of mammoth tusks
were collected for commercial purposes on the islands (Boeskorov
et al., 2008). Of this number, 6e7 tons of tusks were gathered
from the eroded coast of Bolshoi Lyakhovskii Island (73e74
overall length of the eroded shores of the island was w150 km, and
the average rate of coastal retreat was w5 m/yr (Boeskorov et al.,
2008). Therefore, the tusk content of the yedoma along this long
transect was approximately 8 tons/km
, slightly less then for DY. In
these large commercial collections from Bolshoi Lyakhovskii Island
paired tusks represent 5e7% of the total collection. As we noted
before the probability of preserving a single tusk is also equal to 5e
7%. Using this probability, the initial frequency of tusks was w115 e
. Using an average tusk weight of 25 kg we determined
a very high mammoth skeleton frequency of 230 0e3200 mammoth
. The error is due to the fact that not all paired tusks
are identiﬁed as paired. Paired tusks very rarely lay one above the
other, and are most often situatedseveral meters from each another
(the average diameter of the place where the remains of a skeleton
lie is w10 m). Therefore, at rates of yedoma erosion of w5 m/yr, only
a minor proportion (roughly one third) of all of the paired tusks
were found in the same year, and often a second tusk was found the
second or third year after the ﬁrst. Such tusks were identiﬁed and
sold as single tusks. So, the actual proportion of paired tusks in the
yedoma of New Siberian Islands should be tripled, providing us with
a probability for tusks preserved similar to that of Duvanniy Yar. The
tusks accumulated (as at DY) for less than 40,000 years (for these
C mammoth dates are known with just 6 of them older
CyrBP,Sher et al., 2005). As a consequence,
mammoth density on the island was comparable with DY. The ratio
between the bones of different species was also close to those that
we determined in DY (Sher et al., 2005). Only the ratio of musk ox
bones was larger (Sher et al., 2005).
St. Paul Island in the Bering Sea is 1300 km closer to the equator
than DY. That island was isolated due to marine transgression at the
beginning of the Holocene. Mammoths lived isolated on this small
island (100 km
) for w5 thousand years (Guthrie, 2004). The
population of large animals could only be stable if their number
was at least several hundred (Yablokov, 1987). From this fact, we
concluded that mammoth density on the island in Holocene
exceeded 2 mammoths/km
and that inbreeding could be a cause of
8. Mammoth steppe of the Wrangel Island
The area of the island is slightly larger than Lyakhovskii Island,
but 2/3 of its area is occupied by mountains. Loess sediments and
exposures, particularly for the Holocene, are relatively sparse, and
there are no active thermokarst lakes (Vartanyan, 2007). Never-
theless S. Vartanyan (2007) found there (mostly in river beds)
approximately 200 fossil remains of mammoths. Among those he
randomly chose 126 remains e59 molars, 45 tusks and 20 bones
for radiocarbon dating. Such a high ratio of tusks and molars and
absence of bone accumulations indicates a very low probability of
bone preservation on this island. Remember that on Duvanniy Yar
most reindeer bones are singular, and from 10 reindeer skeletons
only 1 bone was preserved in the permafrost. Probability of
mammoth bones preservation on Wrangel Island is less than
probability reindeer bone preservation on DY. All this indicates
relatively high mammoth densities in the valleys of this island.
Wrangel Island is situated 250 km south of Lyakhovskii Island. On
the loamy soils of lowlands, pasture productivity (and corre-
spondingly animal biomass) was probably not lower than on Lya-
In addition to mammoth fossils on this island S. Vartanyan found
30 remains of other animals, 13 of which were used for radiocarbon
dating (bones of horses, rhinoceros, bison and musk ox). One bison
fossil was dated as early Holocene (9450 BP); all other are Pleis-
tocene dates. Reindeer bones were not found (they don’t preserve),
but in deglaciation sediments reindeer excrements were found
The conclusion based on the ratio of number of bones of
different species was that the late Pleistocene faunal assemblage on
Wrangel Island was similar to that in other regions of northern
Siberia and that, in the Holocene, probably only mammoths
inhabited the island (Vartanyan, 2007). Thus in the beginning of the
Holocene MS experienced a large stress, after which only
Holocene mammoth fossil remains on Wrangel Island (108
dates) are substantially more abundant than Pleistocene remains.
The intensity of sedimentation in the Holocene was much lower
than during the Pleistocene (Vartanyan, 2007), indicating that the
higher frequency of Holocene mammoth remains is not related to
taphonomy (selective preservation). This difference in frequency of
dates suggests that Holocene mammoth frequency on Wrangel
Island was many times higher than during the Pleistocene
(Vartanyan, 2007). The frequency of mammoth dates in Pleistocene
is 16 dates/26,000 yr and in the early Holocene 11 dates/1000 yr
which is 18 times more than in Pleistocene.
During the Holocene, mammoths on Wrangel Island were
approximately 30% shorter than during the Pleistocene, and
S.A. Zimov et al. / Quaternary Science Reviews 57 (2012) 26e4540
Author's personal copy
consequently 2 times lighter (Vartanyan, 2007), resulting in 9 (18/
2) times higher biomass of Holocene mammoths. Pollen spectrum
data from Wrangel Island indicates that the character of vegetation
and productivity of pastures in the late Pleistocene and mid Holo-
cene were similar (Vartanyan, 2007). If pasture productivity is
assumed to be similar between these two periods, this suggests
that during the Pleistocene mammoths ate only 1/9 of the available
forage. If productivity during the Pleistocene was half that of the
Holocene, then 2/9 of all forage would have been consumed by
mammoths. It is probable that mammoths ate so little because the
rest was consumed by competitors (ungulate). This estimate indi-
cates that in Pleistocene on Wrangel Island, as at Duvanniy Yar,
mammoths controlled only a minor portion of the biological
cycling. However, the Holocene mammoths were the only large
herbivore species. It seems important, therefore, to discuss the
following phenomenon. The mammoths of Wrangel island were on
average smaller in Holocene than in the Pleistocene, but alongside
the small individuals, the normal-sized mammoths existed
(Vartanyan, 2007). Perhaps it was the beginning of a new sympatric
speciation event as in the classic example with Darwin ﬁnches. On
Wrangel Island the ecological niches of ungulates became vacant in
the Holocene and collared lemming (suitably called “hoofed
lemmings”in Russian) that became larger and some mammoths
that became smaller tried to ﬁll the niches of horse and bison.
The example of the Wrangel Island shows that the mammoth
steppe ecosystem can be stable even at the low diversity of the
herbivores. But can a mammoth steppe exist in the absence of
mammoths? In the African savannah, as in the mammoth steppe,
elephants consume only a modest part of the plant biomass, but
they are the keystone species limiting expansion of trees and large
bushes (Western, 1989;Western and Maitumo, 2004). At the
southern border of the mammoth steppe where the trees were
large, mammoths were perhaps the key species. On the North,
however, the ungulates were able to contain the expansion of the
forest and tundra even in the absence of mammoths.
Based on the different regions and the different methods above,
we obtained similar estimates of animal density. The estimates are
approximate. Many parameters (animal weight, shore width, etc.)
were rounded, but the accuracy of these estimates is probably
similar to the accuracy of estimates of animal densities in modern
ecosystems. Our roughest estimate was for predators, but given
that almost all of the back vertebrae of the horses were eaten and
that all of the humeral bones were broken by predator teeth we
believe that the predator population was large enough to eat
everything that could have been eaten.
We also can see large number of herbivores in the north today.
Northern Siberia is inhabited by semi-wild horses. Their biomass in
Yakutia exceeds the biomass of reindeer (Agricultural Atlas of
Yakutia, 1989). The modern density of wild and semi-wild rein-
deer in the forest and tundra of the Kolyma lowland is only 60 kg/
(1 per km
)(Agricultural Atlas of Yakutia, 1989). In contrast,
the current biomass for horses in the Aleko-Kuel region (300 km
west of DY) on the most productive low-lying meadow, is 200 times
this value (30 horses/km
), which is close to the above estimate of
10.5 tons. The same density of ungulate animals is maintained on
the grasslands of Pleistocene Park (100 km east of DY) (Fig. 15). This
experiment shows that on the North at the high density of the
ungulates the steppe can expand in the absence of mammoths.
9. Mammoth steppe physiology
Although the accuracy of any single calculation might be ques-
tioned, as with any palaeo-reconstruction, the consistency of
patterns that we obtained from many independent data sources
and approaches suggest that the following inter-related conclu-
sions are robust:
1. Plant, herbivore, and predator productivity in mammoth
steppe was close to the theoretical maximum for a northern
ecosystem. The ecosystem very efﬁciently utilized all resources.
The density of animals and their community structure was
similar to that of an African savanna. To feed animals, plants
utilized all available water
2. The vegetation was dominated by palatable high-productivity
grasses, herbs and willow shrubs (Guthrie, 1990;Sher et al.,
2005). No other vegetation could maintain 10 tons/km
3. The soils were fertile. The content of bio available phosphorus
in yedoma is an order higher than in modern soils (Zhigotsky,
1982). If soils of mammoth steppe appear on the surface due
to erosion, they are immediately overgrown by high productive
grasses and herbs (Fig. 16).
4. Winters were much longer than summers, and winter forage
was a limiting resource. Therefore, summer overgrazing was
Fig. 15. Horses, bison and musk-ox inhabit Pleistocene Park. There are also three species of deer. This represents the highest diversity this area has seen in the last 12,000 years.
S.A. Zimov et al. / Quaternary Science Reviews 57 (2012) 26e45 41
Author's personal copy
not possible. In the summer animals could eat only half of all
available forage. During winter herbivores ate all the rest
(Fig. 17), however, without damage to the grass community,
since in winter all living parts are preserved in the soil.
Therefore, all of the insulated snow cover was trampled, and
the soils cooled signiﬁcantly during winter. A change in snow
depth of w10 cm changes the temperature of the permafrost by
C(Yershov, 1998). Soils were fully trampled throughout the
year, preventing the establishment of a moss layer and the
expansion of slowly growing shrubs and trees into grasslands
(Zimov et al., 1995). We did simple calculation: if area of foot-
prints of all animals that occupy 1 km
(their total weight is
more than 10 tons) is equal to 2 m
and if walking they did one
step per second and if they walking one ﬁfth of time then area
of their footprints for one year will be 6 km
. That means that
the mammoth steppe has been trampled 6 times per year. Only
Fig. 16. Part of the Duvanniy Yar exposure.The soils (yedoma) are fertile, so even though it’s a cold northern-faced slope, highly productive grasses appear in locations of permafrost
erosion. In this part of the exposure, grasses prevent erosion through root reinforcement of the soil. In conditions of cold and dry climate with herbivores absent, thermally
insulating litter accumulates on the surface, fertility declines, and in several years grass productivity also declines. If herbivores appear on the site they maintain meadow
productivity and also decrease permafrost thawing.
Fig. 17. The territory of ‘Pleistocene Park’in spring af ter snowmelt. Ten years prior, the area was a continuous community of 2e3 m tall willow shrubs. Due to erosion and long-term
active grazing, the plot developed into a meadow with fertile soils and nutritious grasses. Herbivores therefore graze in this area several times per winter, trample down snow, and
eat all the vegetation that grows during the summer. The winter temperature of the soil surface at this site is 15e20 C colder than for grasslands without grazing. We presume that
during herbivore population peaks in spring all of the MS pastures appeared similar.
S.A. Zimov et al. / Quaternary Science Reviews 57 (2012) 26e4542
Author's personal copy
grasses and herbs could survive in the condition. The albedo of
mammoth steppe was signiﬁcantly higher than the albedo of
forest and shrub lands, especially during the snow season
(Foley et al., 1994). Litter did not accumulate in the ecosystem.
Therefore, at the beginning of the summer, until new grass had
grown bare soil surface (Fig. 8) was intensively heated by the
5. Moderate summer grazing stimulated the regrowth of grasses.
As a result, grasses had no time to ﬁnish their life cycle and
were covered with snow still having a high nutritional value.
Due to summer grazing, graminoids shifted from sexual to
vegetative reproduction, so grasses are not evident in the
pollen record (Guthrie, 1990;Sher et al., 2005). Plant species
assemblage is usually determined from pollen records that are
collected from loess strata. These records often consist of
species that are not common in grasslands (Guthrie, 1990). If on
some territory several meters of loess accumulated, it would
indicate that another territory with sparse vegetation cover
and strong winds would erode the same amount of dust, and
that all of the spores and pollen (and everything aero-
dynamically lighter than sand including insect remains) from
deﬂated areas would appear in pollen-poor grasslands and, in
the end, in loess strata. These inferences can be applied to all
mammoth steppes that formed on loess-dominated soils. In
areas with poor stony or sandy soils, the productivity of
palatable plants would be smaller and grazing and trampling
would be weaker. Therefore, slow-growing unproductive
plants were able to survive and persisted in the regional ﬂora.
Mammoth steppe biomes consist of different ecosystems:
unproductive deﬂation areas, productive grasslands on loess or
loamy soils, and tundra and forests (savanna) on poor soils.
These patterns explain the complicated composition of the
6. Due to high productivity and corresponding plant transpira-
tion, water was often a limiting resource; grass roots
competing for water penetrated the entire depth of the active
layer. This is additionally suggested by the fact that in yedoma
numerous thin grass roots are preserved (Sher et al., 2005).
Near the permafrost table soils were thawed just for few weeks
per year, and temperatures never rose substantially above zero,
therefore organic decomposition was low, and labile carbon
accumulated (Zimov et al., 2009). Therefore, the mammoth
steppe was an ecosystem with a high rate of decomposition for
aboveground biomass (in animals’stomachs), and a very low
rate of decomposition in deeper soils.
7. The similarity of herbivore density and the assemblage of
species on the northern and southern parts of the lowlands
suggest that regular massive migrations of herbivores with
regional differences in mortality were unlikely.
8. Mammoth steppe would only be stable under conditions of
very high animal densities, as they had a strong impact on the
environment. An expansion of the high-albedo psychrophilic
steppe biome, whose dry soils and permafrost accumulate
carbon (Zimov et al., 2006,2009) and do not produce methane
(Sher, 1997;Rivkina et al., 2006), would promote climate
cooling and permafrost expansion, whereas a degradation of
the ecosystem and permafrost in response to recent warming
would amplify climate warming (Zimov and Zimov, in review).
9. The north Siberian mammoth steppe was the coldest and driest
part of the biome. In other grassland ecosystems, where there
was 2e4 times more precipitation, the grasses and herbs
productivity and the animal densities were correspondingly
higher. An analysis of paleovegetation maps indicates that
during the LGM forested areas were ten times smaller than in
the Holocene, and that an area of grass-herbs dominated
ecosystems, reached 70 10
(Adams and Faure, 1998). If
similar animals to the mammoth steppe consumed all forage,
assuming average herbivore density of 20 tons/km
than in the north of Siberia), we obtain global animal biomass
value (1.4 billion tons) close to that obtained via methane
emission by herbivores (Zimov and Zimov, in review). During
the LGM, wetlands were rare. For that period no
C basal peat
initiation dates are know; in abundance they appeared only in
Holocene (Yu et al., 2010). Furthermore, during LGM methane
concentration in the atmosphere was almost half of Holocene
values (Spahni et al., 2005), and herbivores were the main
source of methane due to a total herbivore biomass exceeding
total modern biomass of humans and domestic animals
combined (Zimov and Zimov, in review). In the Holocene the
density of wild herbivores declined by an order.
10. The mammoth steppe and humans
Animals in the mammoth steppe were very numerous, and if
humans exterminated a substantial portion, then the north should
preserve the evidence of that. However, calculating the probability
of ﬁnding such evidence suggests otherwise. Assuming that on
average over each square kilometer humans killed one to two
mammoths and 10 bison. However on the same territory 1000
mammoth and 20,000 bison skeletons that died over the course of
the late Pleistocene through natural causes are also preserved. In
a collection such as presented in Fig. 10,ﬁnding bones from animals
killed by humans is unlikely.
The Yedoma plains do not have lithic resources. Therefore, it is
likely that for hunting bone tools were mostly used (Pitulko,1993).
Assuming that killing and dressing ten animals required w100
bone tools. Bone tools are “small bones”esimilar in size to bison
ribs; so the probability that they persisted in permafrost is low e
one out of hundreds (Fig. 13)eleading to the persistence, on
average, of no more than several bone artefacts per square kilo-
meter. Even specialists can overlook these artefacts among
mudﬂows and south ends of other bones. To ﬁnd one such artifact it
would be necessary to gather tens or even hundreds of collections
such as ours. Also it should be mentioned that in contrast to Alaska,
in northern Siberia, in the BWE accumulation of loess stopped (Sher
et al., 2005) and any evidence of human activity remaining on the
surface would have decayed.
Predators in the mammoth steppe used all of the herbivores
(including soft bones) and therefore they could exist in high
densities; however only few of their bones are preserved in
permafrost. In all of the collections gathered from the Siberian
north, there were only several tens of bones from wolves and lions.
Human bones are similar in size to those of wolves and lions, but
periods of active animal extinctions were 100 times shorter than
the time of yedoma accumulation. Therefore even at rational
resource consumption rates by humans the probability of ﬁnding
a mammoth hunter bone is hundreds of times smaller than the
probability of ﬁnding wolves’or lion’s bones, and human bones
should be absent in these collections.
Since 1947 on the north of Siberia scientists have discovered
three mammoth graveyards and three archaeological sites
(Nikolskiy et al., 2009). What is the density of such sites in the
permafrost on Siberian north? Annually, 20e30 tons of tusks are
collected on the yedoma plains (Boeskorov et al., 2008). The density
of tusks in the yedoma is 10 ton/km
, indicating that on average 2e
of eroded yedoma is surveyed annually. That means that
were eroded and surveyed for the last 60 years. Thus,
there are one graveyard and one archaeological site on 40e60 km
area. Appear that such sites are not unique; they are located in
permafrost on the distance of w7 km from each other. We guess
S.A. Zimov et al. / Quaternary Science Reviews 57 (2012) 26e45 43
Author's personal copy
that most likely mammoth herds died falling down from ice cliffs.
Later, ice cliffs backed away (several meters per year), and slipped
from the cliffs sediments covered the mammoth bones. In the
Berelyokh mammoth graveyard are the remains of over 160
mammoths, on Achchaghyi-Allaikha graveyard over 28 mammoths
(Nikolskiy et al., 2009). Thus, each graveyard contains in the
average tens of mammoth skeletons. That is tens of mammoths per
tens of square kilometers. That is number of mammoth skeletons
on their cemeteries corresponds their number in population.
We can’t provide deﬁnitive proof that humans drove the
mammoths extinct. However we can show that humans could do it,
both physically and mentally. We showed that the density of
artefacts in the north does not contradiction that.
Initially Homo sapiens appeared in the southern part of the
mammoth steppe and for many thousands of years was part of that
ecosystem. Animals were so abundant on the MS that a man would
not need to look for animals. In the range of site (2 km radius) man,
even on the north, could detect hundreds of big herbivores. In the
ecosystem humans could survived even without hunting. There
was enough marrow left after another predators. But it was difﬁcult
to survive in the ecosystem.
Large herbivores were dangerous. In the ecosystem predators
saw humans all time. Humans were the slowest species with the
most defenseless young. Biologically, humans are least adapted to
cold and long winters. Therefore, the survival of humans and
possibilities for expansion into the vast woodless plains of the
north were not limited by animal density, but by severe climate
conditions, the absence of natural shelters, and their level of
technology. Every new dwelling type, weapons, clothes, and ﬁre-
making techniques all contributed to increased human efﬁciency
and survival, and therefore the expansion of the human climatic
envelope. Humans learned how to build shelters, where they could
retreat and store food and animal grease for ﬁre, and thus they
became the main predator in all ecosystems. At the end of PHT
humans had already learned how to hunt all species and how to
survive in any environment. The most striking example of this is an
early Holocene archeological site on the small island of Johovo,
500 km north of the Arctic coast (76
E). Armed with bone
tools, these people lived in those extreme conditions, mostly
hunting polar bears (Pitulko, 1993), which are three times larger
than the cave lion, and ten times larger than the hunters them-
selves. If humans could regularly (i.e., with little risk) hunt the
biggest and most dangerous predators, in the most severe envi-
ronments, it means they could hunt mammoths everywhere.
Human expansion north in geographical space reﬂected climatic
changes. People were likely absent or rare in the homogeneous
northern Siberian plains in the cold epochs before the BWE. During
the BWE the climate in northern Siberia and Alaska became similar
to the glacial climate of Eastern Europe, and these territories
became more suitable for human occupation.
The BWE also sharply changed the landscape. Ice wedges
degradation led to numerous badlands. Tens thousands of ther-
mokarst lakes with steep cliffs as on Fig. 7 (Walter et al., 2006) and
canyons appeared. These changes would not have affected animals;
however they provide huge advantages for hunters. Heterogeneous
landscapes provide better opportunities for hunters to closely stalk
their prey. The landscape also became better for cliff hunting.
Permafrost degradation would have created such cliffs every kilo-
meter, such that chances for successful and safe hunting were
substantially increased. In the south for a long time humans were in
equilibrium with other animals, but during extensive migration to
the north and to America, experienced hunters met numerous
animals that were likely unafraid of people and were therefore
easier to kill. Human expansion to the northern mammoth steppe
occurred under conditions of unlimited resources. In such cases
prey are consumed irrationally. Many of killed animals were either
not used or consumed only their little part.
The highest density of
C mammoth dates in the Siberian north
is recorded for the BWE (after 12,600
C yrs BP) (Fig. 14). But it is
interesting to note that all of the dates that were found to the south
N are found either in mammoth cemeteries or in archeo-
logical sites (Kuzmin et al., 2001;Sher et al., 2005;Nikolskiy et al.,
2009). This data indicates that to the south of 73
N mammoths died
because of humans and became extinct during the BWE, while to
the north of 73
N (north of western Siberia, Taimyr and northern
islands) mammoths persisted (Kuzmin et al., 2001;Sher et al.,
2005). Humans have not reached these extremely sever territory.
Humans arrived there only after the next sharp climate warming
subsequent to the end of the Younger Dryas cold period. After that
mammoths on the continent have disappeared.
The modern climate of north-eastern Siberia, central Alaska and
Yukon Territory are inside the mammoth steppe climatic envelope
(Fig. 2), and climate change could not cause extinction of this rich
and self organized ecosystem there. Holocene climate warming
became fatal for the mammoth ecosystem, because with warming
humans penetrated the north. Our estimates are rough, but they
indicate that in northern Siberia there were enough people to cause
a decline in the herbivore population, thereby decreasing pastures
and ecosystem productivity, with the eventual extinction of
We thank: Lazarev P.A. and G.G. Boeskorov for help in bone
identiﬁcation in our collection; F.K. Shidlovsky and A. Vatagin for
statistical data on mammoth fossil remains from commercial
collections; P.A. Nikolskiy and V.V. Pitulko for data on artifacts and
discussion; Daniel C. Fisher, Julien Louys, Marina and Eugene
Potapovs for manuscript editing and valuable comments.
Appendix A. Supplementary data
Supplementary data related to this article can be found at http://
Adams, J.M., Faure, H., Faure-Denard, L., McGlade, J.M., Woodward, F.I., 1990.
Increases in terrestrial carbon storage from the Last Glacial Maximum to the
present. Nature 348, 711e714.
Adams, J.M., Faure, H., 1998. A new estimate of changing carbon storage on land
since the last glacial maximum, based on global land ecosystem reconstruction.
Glob. Plan. Change 3, 16e17.
Agricultural Atlas of Yakutia, 1989. GUGK USSR, Moscow.
Alroy, J., 2001. A multi-species overkill simulation of the end-Pleistocene mega-
faunal mass extinction. Science 292, 1893e1896.
Alvarez-Lao, D.J., García, N., 2011. Geographical distribution of Pleistocene cold-
adapted large mammal faunas in the Iberian Peninsula. Quat. Int. 233, 159e170.
Boeskorov, G.G., Kirillov, N.D., Lazarev, P.A., Testsov, V.V., 2008. Problemy Region-
al’noi Ekologii 2, 106e109 (in Russian).
Budyko, M.I., 1984. Evolutciya biosfery. Gidrometeoizdat, Leningrad.
Chapin III, F.S., Shaver, G.R., Giblin, A.E., Nadelhoffer, K.G., Laundre, J.A., 1995.
Responses of arctic tundra to experimental and observed changes in climate.
Ecology 76, 694e711.
Chernov, Yu. I., 1985. The Living Tundra. Cambridge University Press, p. 201.
Cwynar, L.C., Ritchie, J.C., 1980. Arctic steppe-tundra: a Yukon perspective. Science
Foley, J.A., Kutzbach, J.E., Coe, M.T., Levis, S., 1994. Feedbacks between climate and
boreal forests during the Holocene epoch. Nature 371, 52e54.
Friedlingstein, P., Prentice, K.C., Fung, I.Y., John, J.G., Brasseur, G.P., 1998. Carbone
biosphereeclimate interactions in the last glacial maximum climate. J. Geophys.
Res. 100, 7203e7221.
S.A. Zimov et al. / Quaternary Science Reviews 57 (2012) 26e4544
Author's personal copy
Garrut, N.V., Boeskorov, G.G., 2001. In: Rozanov, Yu.A. (Ed.), Mammoth and Its
Environment: 200 Years of Investigations. GEOS, Moscow.
Guthrie, R.D., 1990. Frozen Fauna of the Mammoth Steppe. The University of Chi-
cago Press, Chicago.
Guthrie, R.D., 2004. Radiocarbon evidence of mid-Holocene mammoths stranded
on an Alaskan Bering Sea island. Nature 426, 746e749.
Guthrie, R.D., 2006. New carbon dates link climatic change with human coloniza-
tion and Pleistocene extinctions. Nature 441, 207e209.
Kuzmin, Y.A., Orlova, L.A., Zolnikov, I.D., Igolnikov, A.E., 2001. In: Rozanov, A.Yu.
(Ed.), Mammoth and Its Environment: 200 Years of Investigations. GEOS,
Moscow, pp. 124e138.
Martin, P.S., December 1967. Pleistocene overkill. Nat. Hist., 32e38.
Muhs, D., 2003. Stratigraphy and palaeoclimatic signiﬁcance of Late Quaternary
loessepalaeosol sequences of the Last-InterglacialeGlacial cycle in central
Alaska. Quat. Sci. Rev. 22, 1947e1986.
Nikolskiy, P.A., Basilyan, A.E., Sulerzhitsky, L.D., Pitulko, V.V., 2009. Prelude to the
extinction: revision of the AchchagyieAllaikha and Berelyokh mass accumu-
lations of mammoth. Quat. Int.. http://dx.doi.org/10.1016/j.quaint.2009.10.028
Pavlov, A.V., 1984. Energoobmen v landshaftnoi sfere zemli. Nauka, Sibirskoe
Pitulko, V.V., 1993. An early Holocene site in the Siberian High Arctic. Arctic
Anthropol. 30, 13e21.
Reimer, P.J., Baillie, M.G.L., Bard, E., et al., 2004. INTCAL0 4 terrestrial radiocarbon age
calibration, 0e26 cal kyr BP. Radiocarbon 46, 1029e1058.
Rivkina, E.M., Kraev, G.N., Krivushin, K.V., Laurinavichus, K.S., Fyodorov-
Davydov, D.G., Kholodov, A.L., Shcherbakova, V.A., Gilichinsky, D.A., 2006.
Methane in permafrost of northern arctic. Earth Cryosphere 10, 23e41.
Schirrmeister, L., Siegert, C., Kuznetsova, T., Kuzmina, S., Andreev, A.A., Kienast, F.,
Meyer, H., Bobrov, A.A.,20 02. Paleoenvironmental and paleoclimatic records from
permafrost depositsin the Arctic region of northern Siberia.Quat. Int. 89, 97e118 .
Schuur, E.A.G., Abbott, B., 2011. High risk of permafrost thaw. Nature 480, 32e33.
Severinghaus, J.P., Brook, E.J., 1999. Abrupt climate change at the end of the last
glacial period inferred from trapped air in polar ice. Science 286, 930e934.
Sher, A.V., 1997. Nature restructuring in the East-Siberian Arctic at the Pleistocenee
Holocene boundary and its role in mammal extinction and emerging of modern
ecosystems. Earth Cryosphere 1 (3e11), 21e29.
Sher, A.V., Kuzmina, S.A., Kuznetsova, T.V., Sulerzhitsky, L.D., 2005. New insights
into the Weichselian Environment and climate of the east Siberian Arctic,
derived from fossil insects, plants, and mammals. Quat. Sci. Rev. 24, 533e569.
Sokolov, I.A., Konyushkov, D.E., 1998. Pochvovedenie 11, 1303e1317.
Spahni, R., Chappellaz, J., Stocker, T.F., et al., 2005. Atmospheric methane and
nitrous oxide of the Late Pleistocene from Antarctic ice cores. Science 310,
Stephenson, R.O., Gerlach, S.C., Guthrie, R.D., Harington, C.R., Mills, R.O., Hare, G.,
2001. Wood bison in late Holocene Alaska and adjacent Canada; paleontolog-
ical, archaeological and historical records. In: Gerlach, S.C., Murray, M.S. (Eds.),
People and Wildlife in Northern North America: Essays in Honor of R. Dale
Guthrie. BAR International Series 944, pp. 125e159.
Stone,R., 2001.Mammoth: the Resurrection of IceAge Giant. PerseusPublishing, p. 228.
Syroechkovskii, V.E., 1986. Severnii Olen’Agropromizdat. Moscow (in Russian).
Vartanyan, S.L., 2007. Wrangel Island at the End of Quaternary Period: Geology and
Paleogeography. Ivan Limbakh Press, St. Petersburg.
Vasil’chuk, Y.K., Punning, J.M., Vasil’chuk, A.C., 1997. Radiocarbon ages of
mammoths in northern Eurasia: implications for population development and
late Quaternary environment. Radiocarbon 39, 1e18.
Vasil’chuk, Y.K., Vasil’chuk, A.C., Rank, D., Kutchera, W., Kim, J.-C., 2001. Radiocarbon
D plots in late Pleistocene ice-wedges of the Duvanniy Yar
(lower Kolyma river, northern Yakutia). Radiocarbon 43, 503e515.
Velichko, A.A., Zelikson, E.M., 2001. In: Rozanov, Yu A. (Ed.), Mammoth and Its
Environment: 200 Years of Investigations. GEOS, Moscow.
Veltre, D.W., Yesner, D.R., Crossen, K.J., Graham, R.W., Coltrain, J.B., 2008. Patterns of
faunal extinction and paleoclimatic change from mid-Holocene mammoth and
polar bear remains, Pribilof Islands, Alaska. Quat. Res. 70 (1), 40e50.
Vereshchagin, N.K., Tikhonov, A.N., 1990. In: Shumilov, Yu.V. (Ed.), Exter’er
Mamonta. Yakutsk, Russia (in Russian).
Walter, K.M., Zimov, S.A., Chantom, J.P., Verbyla, D., Chapin III, F.S., 2006. Methane
bubbling from Siberian thaw lakes as a positive feedback to climate warming.
Nature 443, 71.
Western, D., 1989. The ecological role of elephants in Africa. Pachyderm 12, 42e45.
Western, D., Maitumo, D., 2004. Woodland loss and restoration in a Savanna Park:
a 20-year experiment. Afr. J. Ecol. 42, 111e121.
Yablokov, A.V., 1987. Populyacionnaya biologiya. Vysshaya Shkola, Moscow (in
Yershov, E.D., 1998. General Geocryology. Cambridge University Press, Cambridge.
Yu, Z., Loisel, J., Bros