Impact of climate change on the transition of
Neanderthals to modern humans in Europe
, Virgil Dr ˘
, Bogdan P. Onac
, Sergey Assonov
, Vasile Ersek
, Dirk L. Hoffmann
and Daniel Veres
Institute of Geologie and Mineralogy, University of Cologne, 50674 Cologne, Germany;
a Institute of Speleology, Romanian Academy, 010986
School of Geosciences, University of South Florida, Tampa, FL 33620;
a Institute of Speleology, Romanian Academy,
400006 Cluj-Napoca, Romania;
Terrestrial Environment Laboratory, Environmental Laboratories, Department of Nuclear Applications, International Atomic
Energy Agency, 1400 Vienna, Austria;
Department of Geography and Environmental Sciences, Northumbria University, Newcastle upon Tyne NE1 8ST,
United Kingdom; and
Department of Human Evolution, Max Planck Institute for Evolutionary Anthropology, 04103 Leipzig, Germany
Edited by Richard G. Klein, Stanford University, Stanford, CA, and approved July 30, 2018 (received for review May 19, 2018)
Two speleothem stable isotope records from East-Central Europe
demonstrate that Greenland Stadial 12 (GS12) and GS10—at 44.3–
43.3 and 40.8–40.2 ka—were prominent intervals of cold and arid
conditions. GS12, GS11, and GS10 are coeval with a regional pat-
tern of culturally (near-)sterile layers within Europe’s diachronous
archeologic transition from Neanderthals to modern human Auri-
gnacian. Sterile layers coeval with GS12 precede the Aurignacian
throughout the middle and upper Danube region. In some records
from the northern Iberian Peninsula, such layers are coeval with
GS11 and separate the Châtelperronian from the Aurignacian.
Sterile layers preceding the Aurignacian in the remaining Châtel-
perronian domain are coeval with GS10 and the previously
reported 40.0- to 40.8-ka cal BP [calendar years before present
(1950)] time range of Neanderthals’disappearance from most of
Europe. This suggests that ecologic stress during stadial expansion
of steppe landscape caused a diachronous pattern of depopulation
of Neanderthals, which facilitated repopulation by modern humans
who appear to have been better adapted to this environment. Con-
secutive depopulation–repopulation cycles during severe stadials of
the middle pleniglacial may principally explain the repeated replace-
ment of Europe’s population and its genetic composition.
millennial-scale climate cycles
Middle—Upper Paleolithic transition
The replacement of Neanderthals by modern humans is
recorded across Europe in a diachronous and culturally complex
succession of distinct stone tool assemblages from the Middle–
Upper Paleolithic transition (MUPT) roughly between 48 and 36 ka
cal BP [calendar years before present (1950)] (1, 2). The succession
often includes a regionally distinct “transitional”assemblage of local
origin, or an intrusive Initial Upper Paleolithic assemblage between
the Neanderthal Mousterian and modern human Aurignacian. The
oldest anatomically modern human remains from Europe—found in
East-Central Europe and radiocarbon dated to 40.6–38.6 ka cal BP
(68% probability) toward the time of Neanderthals’disappearance
from most of continental Europe—carry genetic evidence for species
interbreeding four to six generations earlier (3–6). This individual,
however, represents a population that did not contribute to the
genome of modern humans present in glacial Europe after the
MUPT (7), and the archeologic record provides no site with in-
dication of local coexistence. Within a few millennia after the
MUPT, at least two other genetically distinct modern human pop-
ulations came to subsequently dominate Middle Pleniglacial Europe.
During the entire interval, northern hemispheric climate went
through several millennial-scale Dansgaard–Oeschger (DO) cold
cycles (8, 9). A causality between climate change, the archeologic
succession, and modern humans’genetic makeup has been tenta-
tively suggested but not demonstrated (1, 2, 7). Below, we present
the climatic history of continental Europe during the MUPT and
derive the impact of climate change on MUPT demography, which
may have led to the apparent repetitive genome turnover reported
for Europe’s human population during the middle pleniglacial.
The MUPT spans five DO cycles approximately between
Greenland Interstadial 12 (GI12) and Greenland Stadial 8 (GS8)
(9) for which climate change over continental Europe is poorly
constrained. The paleoclimatic and environmental context is
known with sufficient resolution and age-control only along the
continent’s western and southern fringe. In the Aegean and
Black Sea region, records of sea surface temperature (10),
coastal ice-rafted detritus (IRDc) (11), pollen assemblages (12–
15), and stable isotopes in speleothems (16) suggest a DO-type
response without the clear prominence of ice-rafting intervals
(Heinrich stadials) seen in the Atlantic domain (17). Forest was
generally more abundant in Europe during interstadials, while
steppe landscape advanced during stadials (12). A taiga and
tundra shrub/forest landscape covered the eastern European
plains, with some loess deposition east of the Carpathians (12,
18, 19). The middle and lower Danube Plain was a steppe
landscape with continuous loess deposition (20). A temperate
open forest in the mountains of the southern Balkan passed into
a xerophytic steppe toward the Aegean Sea (12, 13, 15). Boreal
forest with birch and pine trees was present at 50° N in Western
Europe (Eifel maar lakes) but began to degrade after GI12,
∼44.5 ka cal BP (21). Two sparsely dated records from the
Western Carpathians (Safarka, Jablunka) suggest a dense taiga
forest landscape (14). For the upper Danube Plain, pollen
(Füramoos) and loess/paleosol profiles (Willendorf II, Nussloch)
A causality between millennial-scale climate cycles and the
replacement of Neanderthals by modern humans in Europe has
tentatively been suggested. However, that replacement was
diachronous and occurred over several such cycles. A poorly
constrained continental paleoclimate framework has hindered
identification of any inherent causality. Speleothems from the
Carpathians reveal that, between 44,000 and 40,000 years ago,
a sequence of stadials with severely cold and arid conditions
caused successive regional Neanderthal depopulation intervals
across Europe and facilitated staggered repopulation by mod-
ern humans. Repetitive depopulation–repopulation cycles may
have facilitated multiple genetic turnover in Europe between
44,000 and 34,000 years ago.
Author contributions: M.S., V.D., and B.P.O. designed research; M.S. and V.D. performed
research; V.D., B.P.O., S.A., and D.L.H. analyzed data; and M.S., V.D., B.P.O., V.E., and D.V.
wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
This open access article is distributed under Creative Commons Attribution-NonCommercial-
NoDeriv atives L icense 4.0 (CC BY-N C-ND).
To whom correspondence should be addressed. Email: email@example.com.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
www.pnas.org/cgi/doi/10.1073/pnas.1808647115 PNAS Latest Articles
AND PLANETARY SCIENCES
suggest a cold steppe environment with few conifers during in-
terstadials, and a tundra landscape with cryosoil formation dur-
ing stadials (12, 22, 23). During GIs, a temperate forest-steppe
prevailed west (marine records) and south of the Alps (Monticchio,
Castiglione, and Lagaccione) (12). Dust deposition in the Eifel
maar lakes and speleothem carbon isotopes from Villars Cave
suggest increasing aridity in Western Europe across the MUPT
The MUPT Speleothem Paleoclimate Record of East-Central
Here, we present two speleothem records from East-Central
Europe (Romania). Stalagmite POM1 is from Ascuns˘
(AC), South Carpathians, 50 km north of the Danube Valley at
1,050-m altitude (SI Appendix, Fig. S1). Stalagmite 1152 is from
¸oare Cave (TC), East Carpathians, at 950-m altitude, 2.5°
north of the AC site. We dated the speleothems by U-Th and
measured stable carbon and oxygen isotopes continuously at
decadal resolution (SI Appendix, Fig. S2 and Table S1).
The AC speleothem δ
C record reproduces the DO pacing
of the Greenland ice core record during the MUPT (Fig. 1). The
O record reproduces some aspects of the Greenland record.
The climatic meaning of the AC δ
C data may be constrained
within context of other records from the region, particularly the
Black Sea (Fig. 1). The coherency of the AC δ
C record with
southern Black Sea IRDc (11) implies that speleothem δ
responded to changes in the length of the sea ice and winter frost
season. Chemically, speleothem δ
C depends on the proportion of
C aqueous CO
derived from microbial respiration of soil
organic matter above the cave (26). Extended frost and shorter
plant growth seasons reduces the supply of fresh soil organic matter
(SOM) and increases the proportion of CO
from old SOM in drip
water, which has up to 8‰higher δ
C than fresh SOM under
comparable conditions (27). Other potential controls are insuffi-
cient to explain the 6‰amplitude of DO cycles in the speleothem.
For example, the effect of a seasonally variable CO
from drip water may add a kinetic fractionation effect on the δ
of precipitating speleothem calcite (28). However, recent moni-
toring showed that interannual variability in δ
C of precipitating
calciteinACislessthan1.5‰(29). Similarity between the AC δ
O records is limited (SI Appendix,Fig.S3)—another
qualitative argument against major kinetic isotope fractionation.
Also, a compositional change in SOM δ
C can be ruled out as
variability in Central European paleosol profiles does not exceed
1‰during the MUPT (30). Finally, variable moisture availability
may have influenced soil formation above the cave and speleothem
C. The Danube loess record suggests such an interval of en-
hanced moisture availability after GI8 (20). The AC-speleothem
O record does indeed suggest some change to that effect be-
tween GS8 and GI7, but the overall pacing across the MUPT re-
sembles that of southeastern Mediterranean records and shows
little coherency with DO cycles (see below). In general, soil for-
mation intervals recorded by low values in AC-δ
C are coeval with
the range of ages obtained for paleosols in the upper Danube
Willendorf II profile (Fig. 2). This suggests that the entire Danube
region was inside the same climate zone and responded coherently
to temperature change during DO cycles of the MUPT.
Little variance is observed in the East Carpathian TC δ
record (SI Appendix, Fig. S3). Here, pyrite weathering and en-
hanced limestone dissolution in the host rock obscures any po-
tential influence on δ
C from SOM (SI Appendix). That record
is not considered in this study.
In general, speleothem δ
O reflects a combination of calcite
precipitation temperature—the annual average temperature in-
side the cave—and factors controlling δ
O in drip water. Within
a few days, water and dissolved CO
equilibrate isotopically in
the aquifer above the cave, and drip water composition reflects
regional hydroclimatology, that is, moisture source, rain-out
history, local rain-out temperature, and the annual distribution
of rainfall (26, 31). The AC δ
O signal’s overall amplitude is
∼4‰. Such a large change is beyond reasonable temperature
variability (28) and suggests a dominant influence of changes in
hydrology on AC δ
O. The dissimilarity of AC δ
O to the co-
C and Greenland temperature records (SI Appendix,
Fig. S3) renders a temperature control of the AC δ
unlikely. Hydroclimatic dominance on speleothem δ
O is typical
for the entire Eastern Mediterranean domain’s Holocene record
including at AC, which reflects significant variability in rainfall
seasonality and moisture source proportions (31). The similar
pacing between AC δ
O and southeastern Mediterranean δ
record (SI Appendix, Fig. S4) may indicate that the underlying
synoptic-scale mechanism was also active during MIS3.
The TC speleothem δ
O record has an amplitude of ∼1‰
and a different pacing compared with AC δ
O(SI Appendix, Fig.
S3). This suggests that the East Carpathians were in a different
hydroclimatic regime during the MUPT. What defines the East
Carpathian regime may be investigated by exploring the partial
coherency between TC δ
O and the other paleoclimate record.
After ∼60 ka, the δ
O record shows low-amplitude variability
with a few pronounced minima of ∼0.5‰that are coeval with
some of the Greenland stadials (Fig. 1). These include GS3-H2,
GS5-H3, and GS13-H5. Other minima just after 34 ka and at
Fig. 1. (A) Greenland: North Greenland Ice Core Project (NGRIP) tempera-
ture (8). (B) South Carpathians: δ
C of stalagmite POM1 from AC. (C)
Southern Black Sea: coastal IRD abundance in core M72-5 (11). (D)East
O of stalagmite 1152, TC. (E) Northern Black Sea: TEX
summer sea surface temperatures (33). The gray bars indicate Heinrich sta-
dials. A map with locations of records is available in SI Appendix, Fig. S1.
www.pnas.org/cgi/doi/10.1073/pnas.1808647115 Staubwasser et al.
41 ka correspond to GS7 and GS10 (Fig. 1). The latter lasted
from 40.66 ±0.47 ka
,P=95%)—directly dated—to an in-
terpolated end date of ∼39.70 ±0.43 ka (32). Although there
may be a few centuries of overlap with GS9-H4, the measured
age interval of 40.7–39.8 ka
,±0.47 to ±0.3 uncertainty,
matches GS10 (40.80–40.16 ka
) better than GS9-H4 (from
39.90 to 38.22 ka
) (9). A poorly resolved relative δ
maximum at 39.6 ±0.4 ka
and a subsequent broad relative
O minimum in the TC record appear to be coeval with GI9
and GS9-H4, which apparently had a lesser impact on Central
European climate. The TC δ
O record thus responded to stadial
cooling in the Atlantic but with a regionally controlled ampli-
tude. Similarity with the northern Black Sea paleotemperature
archive suggests this response is temperature driven (Fig. 1). For
two reasons, TC δ
O likely records the local summer temper-
ature signal. First, the northern Black Sea record shows that
summer sea surface temperature between 40 and 20 ka cal BP
dropped by ∼2 °C during stadials from a MIS3 average of 4–5°C
(33). Assuming published global empirical temperature rela-
tionships for cave calcite—Δδ
/ΔT=−0.18 (28)—and for
rainwater in midlatitudes—where Δδ
O amplitude in the TC speleothem would corre-
spond to a comparable 1.0–1.5 °C temperature change. Second,
palynologic, geomorphologic, and geochemical studies support
this interpretation as follows. During the last glacial, the low-
lands north of the East Carpathians were situated inside the
tundra biome, where field evidence of ice wedges indicates
permafrost conditions (35–37). Deep winter frost or even dis-
continuous permafrost is a likely scenario for the >1,000-m al-
titude of the ground above TC in the East Carpathians. At 20–40 ka,
the average ground temperature of the northern Black Sea
drainage basin—comprising the Danube, Dniestr, Dniepr, and
Don Lowlands—was ∼4 °C (38). Common adiabatic gradients of
6–10 °C per km of altitude imply that annual average ground
temperatures above TC were lower than −2 °C. This suggests an
alpine near-permafrost environment. Permafrost conditions may
have been continuous during extreme stadials, thus preventing
groundwater recharge to the cave and interrupting speleothem
growth. During GS12—the coldest of all stadials in the Green-
land record—the TC speleothem shows a hiatus coeval with the
most pronounced δ
C maximum in the AC speleothem further
south. In the Willendorf II loess profile, this interval correlates
with tundra gley horizon C9 with evidence of permafrost (22)
(Fig. 2). C9 is constrained in time by paleosols C8-3 above and
D1 underneath with dating ranges similar to GI11 and GI12,
respectively. The Willendorf II profile is located at the same
latitude as the TC speleothem, but only at 230-m altitude.
Continuous permafrost conditions at 950-m altitude in the East
Carpathians were thus likely during GS12. Seasonally permeable
frost or discontinuous permafrost at other times would restrict
speleothem growth to the summer season. Unlike the hydro-
logically dominated AC δ
O record, TC δ
O-only records re-
gional summer temperature change.
Paleoclimatic Context of the Middle Pleniglacial in Europe
A reduction of forest and expansion of steppe biomes occurred
during all stadials (12–15, 21), but climate records between the
Atlantic and the Black Sea show a spatially heterogeneous response
to DO cycles (Fig. 2). GS13-H5 is apparent in the TC speleothem
O record but unlike in the Atlantic domain, its amplitude is
small compared with subsequent stadials. Southern Black Sea
IRDc data (Fig. 1) indicate less sea ice than during subsequent
stadials despite apparent colder annual sea surface temperature
(Fig. 2). This suggests a different seasonality with less severe win-
ters. Thus, extreme cold was unlikely in East-Central Europe. The
extreme aridity apparent in the Aegean region (13) may not have
been as severe in the Balkan and Black Sea region (15, 16). GS12,
from 43.4–44.3 ka
or 43.3–44.0 ka
(AC), is a very
Fig. 2. (A) Greenland: NGRIP temperature (8). (B) Southern Black Sea: TEX
sea temperature core M72-5 (10). (C) South Carpathians: AC stalagmite
C. (D) Upper Danube: Willendorf loess/paleosol profile, paleosol
ages, age probability density functions (age-pdf), and stratigraphically con-
strained loess and gley horizons C7-2, C7-3, and C9 (22). (E) East Carpathians:
TC stalagmite 1152 δ
O. (F) Eifel maar lakes, ELSA dust stack. (G) Western
Massive Central: Villars Cave stalagmites δ
C(25).(H) Iberian Atlantic margin:
sea surface temperature core MD95-2042 (17). Numbers indicate
Greenland stadials at the end of the respective DO cycle.A map with locations
of records is available in SI Appendix,Fig.S1.
Staubwasser et al. PNAS Latest Articles
AND PLANETARY SCIENCES
prominent stadial in Greenland and Central Europe (Fig. 2), where
permafrost changed the upper Danube lowland cold steppe into a
tundra (see above). Malacological data from Willendorf II and the
Eifel maar dust record suggests dryer conditions also in Western
and Central Europe (22, 24) (Fig. 2). A cooling step likely occurred
over Western Europe, supported by evidence of deep frost in the
loess/paleosol record of northern France at ∼43–44 ka cal BP (39).
However, this stadial is less prominent in the Atlantic record (17).
GS11 is present in all continental records, but not as pronounced as
GS12. A common cooling trend superimposed over DO cycles 12–9
is apparent in all records (Fig. 2). GS10—from 40.1–40.9 ka
and 39.7–40.7 ka
(TC)—is a very prominent stadial over
continental Europe, coeval with another cooling step in Western
Europe (Fig. 2), but also a reduction in moisture and reduced
speleothem growth (25). The first occurrence of loess in the
Willendorf II profile (C7-2) around that time (22) and another
strong dust peak in the Eifel maar record suggest stronger aridity
than during GS12. Although present in all continental records, this
stadial is unremarkable in the Atlantic and in the Black Sea (Fig.
2). However, annual sea surface temperature and winter coastal
ice abundance in the Black Sea show conflicting results for GS10,
which could again indicate a seasonality change. GS9-H4 is a long-
lasting stadial that was less prominent in the Central European
record but coeval with significant cooling and aridity in Western
Europe (Fig. 2). Over the Atlantic and the Black Sea, GS9-H4 is a
much more significant stadial. Subsequent GI8 is as warm as GI12
in Central Europe and the Black Sea region, with soil formation in
the Danube region and East Carpathians (19, 20, 22), but colder in
Western Europe and the Atlantic. Dust deposition in the Eifel
maar records was also high during GI8, unlike GI13. The causes to
this regionally heterogeneous response to DO climate cycles re-
main uncertain, but a variable influence of the Siberian high has
been suggested (20).
Following the MUPT, GS8 from 36.6 to 35.5 ka
from 35.5 to 34.7 ka
are difficult to separate from each
other in the speleothem records. The Black Sea record shows
severe cooling during GS8 (Fig. 2). The AC-δ
O record suggests
a significant fluctuation of the regional hydrologic condition
during GS8 and GI7 in agreement with extended soil formation
in the Danube Valley and the East Carpathians (19, 22). Higher
moisture availability may have masked the temperature signal in
C. After GI7, a long-term cooling trend is superimposed
on DO variability throughout Europe (Fig. 2), in many places
coincident with a long drying trend (20, 25).
The Relationship Between Demography and Climate Change
During GS12 and GS10
GS12 and GS10 stand out as the most significant stadials in
Central and Western Europe during the MUPT with severe cold,
aridity, retreat of woodland, and expansion of steppe biomes.
The consequence for human populations present is reflected in
regional population decline (40) and in hunted game species,
which in Western Europe changed from bovine dominated to
reindeer dominated (41). Climatic and environmental con-
straints may not per se prove causality between the archeologic
succession, the rapid replacement of humans’genome, and cli-
mate change, respectively. However, they do allow testing such a
scenario for chronological feasibility: The biome changes during
GS12 and 10 likely forced the population in open woodland
habitats throughout Europe to adapt their subsistence strategy
or habitat track their preferred biome to survive. Where biomes
changed significantly, adaptation may not have been possible.
During GS12 and 10, the permafrost boundary encroached the
upper Danube region and East Carpathians. Both speleothems
presented here suggest a rapid onset of stadial cooling and aridity
in continental Europe within a few decades (Fig. 2). Depopulation
would have been the consequence of failure to adapt or migrate
away in time.
A number of chronologically well-constrained archeologic
records allow for testing directly the feasibility of a regional
depopulation scenario during severe stadials. Culturally (nearly)
sterile layers have been reported—directly
C dated or con-
C-dated cultural layers above and below—in some
regions and mark an archeological hiatus of several centuries
duration. While such layers in some places represent short-term
sedimentologic events without apparent discontinuity of habita-
tion (42), at many sites their time span is multicentennial. These
layers mostly relate chronologically to the time of GS12 and
GS10, and suggest regionally widespread depopulation for many
centuries between two subsequent cultures. (Fig. 3) (SI Appendix,
Table S2). During GS12, this is the case for sites from the upper
and middle Danube region. At Geissenklösterle, near-sterile
(geologic) layer 17 separates Mousterian and Aurignacian as-
semblages (43) and is directly dated to 44.4–42.7 and 43.5–
42.0 ka cal BP by two samples, with a Bayesian model age of
42.9–42.1 ka cal BP (44). At Sesselfelsgrotte, sterile layer F is
directly constrained by Mousterian layer G1, 45.8–44.3 ka cal BP,
and the late Mousterian layer E3, 42.2–40.5 ka cal BP, both
recalibrated (4, 45). At Willendorf II, sterile layer C9 and sub-
sequent layer C8-3 containing Aurignacian artifacts are con-
strained between two dated paleosols D1 and C8-2 (Fig. 2) (22).
Mousterian Neanderthal presence in the middle Danube region
has been redated to until ∼45 ka cal BP, just before GS12 (46).
Two sites with Aurignacian finds, at Keilberg-Kirche (47), 43.6–
41.7 ka cal BP recalibrated (4), and at Pes-kὅCave, 43.8–41.2 ka
cal BP (48), immediately postdate GS12. These fit the scenario
of a cultural hiatus but lack local constraint by older assemblages.
All of the above suggest widespread Mousterian Neanderthal
Fig. 3. The temporal pattern of Greenland stadials (red), prominent events
in the Central European speleothem record (blue), and culturally (near-)
sterile layers in archeologic records of Western and Central Europe (black).
For stadials and speleothems, the bar shows the duration. For archeologic
layers, the bar shows the68% age interval (ka, cal BP) defined by the youngest
date of the preceding layer and the oldest date of the succeeding layer. For
details on archeologic radiocarbon chronologies, see SI Appendix,TableS2.A
map with the locations of records is available in SI Appendix,Fig.S1.
www.pnas.org/cgi/doi/10.1073/pnas.1808647115 Staubwasser et al.
depopulation of the upper and middle Danube Valley and—
except for the tributary Altmühl Valley (49)—repopulation by
Aurignacian after GS12.
Circumstantial evidence points to a similar impact of GS12 in
northern Italy, where two sites, Riparo Mochi on the Ligurian
Mediterranean coast—layer H—and at Grotta di Fumane in the
Venetian Pre-Alps—layer A7—contain sterile layers probably
coeval with GS12. However, reliable dating is only available for
the base of the subsequent archeologic layer (50, 51) and age
uncertainty is large (Fig. 3). A muted response to GS12 may be
inferred for the site of Grotte Mandrin, Rhone Valley, where
sterile layers separate strata C, B3, and B2 with post-Neronian-II
artifacts (50). Layers B3–B1 of this latest regional Mousterian
expression were dated in the range between 42.7 and 45.0 ka BP
(6). These layers may indicate environmental change but do not
appear to reflect longer breaks in the site’s occupation. No such
layers associated with GS12 are reported from archeologic sites
further west. This suggests that the environmental impact of
GS12 may have been less severe in Western Europe.
Sterile or near-sterile layers, or cultural hiatuses in the
archeologic succession also occur around the time of GS10 (Fig.
3). This time is coeval with Neanderthals’disappearance from
Western Europe (6). Sites include the Grotte-du-Renne—the
low-density artifact layer VIII of intermittent Châtelperronian
occupation (52); Les Cottés—sterile unit 5 (53); and Saint Césaire—
layer Ejo inf (6). At Willendorf II, dated paleosols with lithic
artifacts in C8 and C7 broadly constrain the sterile gley and loess
horizons (C7-3 and C7-2) at that time (22) (Fig. 2). Subsequent
archeologic layers are generally Aurignacian and document
modern human expansion or reoccupation during GI9 and GI8
but under more arid conditions in Europe (Fig. 2).
A few archeologic profiles from around the Pyrenees—at
Labeko Koba, Abric Romani, and possibly at Isturitz, but not at
L’Arbreda (54–56)—contain sterile layers coeval with GS11 (Fig.
3). These represent a regional cultural hiatus between the
Châtelperronian and the (Proto-)Aurignacian. However, apart
from a long-term cooling and drying trend over Western Europe,
an unambiguously dated climatic cold event cannot be detected
in the present speleothem records for that time (Fig. 2).
A general case for likely adaptation of modern humans in
response to climate change during the MUPT has already been
made (48, 57, 58). Increasing cold and aridity around the onset of
GS12 (Fig. 2) appears to mark transitions in the archeologic
sequence attributable to Neanderthals that suggest some adap-
tation as well. GS12 is coeval with the transition from Mouste-
rian to archaic Uluzzian, and then to evolved Uluzzian in
northern Italy (59), and from Mousterian to Châtelperronian in
Western Europe (53). However, late Neanderthals may have had
a less diverse diet than modern humans (60). In open grasslands,
Neanderthals’exclusive diet was meat from terrestrial animals,
whereas modern human Aurignacian also exploited plant and
aquatic foods (60, 61). The frequently observed cultural hiatuses,
however, do not suggest a direct competitive displacement of
Neanderthals, but rather a higher vulnerability to rapid envi-
ronmental change and ecologic stress in the open landscape
during cold and arid GS12–GS10. While Neanderthals did not
survive GS12 in most of the Danube steppe and tundra, modern
humans may have been more capable to adapt and habitat track
the expanding steppe in Central Europe. GS10 likely caused a
repeat of that process in Western Europe.
A depopulation scenario was suggested for GS13-H5 as a
trigger for the first intrusion of modern humans into Europe,
represented by the Bohunician stone tool assemblage in Moravia
(Central Europe) and further east (13, 62). Because a potential
cultural hiatus is difficult to prove in the archeologic record of
Moravia at given accuracy and precision of applied radiometric
dating methods (63), this scenario may not currently be tested.
There is also not as strong a climatic evidence in support.
Cooling likely was less severe in East-Central Europe compared
with subsequent stadials (see above). Severe aridity was apparent
between the northern Aegean coastal region and the Levant (13,
64) but not over the Balkan (15), the Adriatic and the Black Sea
regions (16, 65), or Central-East Europe.
The Moravian record suggests contemporaneity of modern
humans—the Bohunician—and Neanderthals—the Szeletian—
at close geographic proximity during GI12 and perhaps later but
with some chronologic uncertainty (63, 66). During GI11-10, the
situation is comparable for the Aurignacian and Mousterian on
the upper Danube—at Keilberg-Kirche and Sesselfelsgrotte (see
above)—and for Aurignacian and Szeletian on the middle
Danube—at Pes-kὅCave and Szeleta Cave (42, 67) (SI Appen-
dix, Fig. S1). Here, however, the interbreeding between species
only six generations before the lifetime of the Oase Cave modern
human specimen in that time range (5) principally confirms true
contemporaneity along the northern fringe of the steppe land-
scape in the Danube Valley. Nonetheless, in the subsequent ∼6
millennia after GS10, the modern human genome in Europe was
replaced twice with a different genetic lineage harboring a much
older Neanderthal ancestry (7). One modern European genetic
branch found inside Goyet Cave (Belgium), was attributed to the
late Aurignacian and dated to ∼35 ka cal BP—coeval with GI7.
This interval of modern human repopulation may have followed
the extended cold interval from GS10 to GS9-H4, over 4,000 y
long and only interrupted briefly by the 250-y-long GI9 (Fig. 2).
A different genetic branch then dominated Europe after ∼34 ka cal
BP—coeval to the end of the severe and 1,000-y-long GS7 (Fig.
2). This lineage was attributed to the Gravettian (7), who repo-
pulated Europe during yet another time of low population
density (68). The discussion above lays out a general scenario of
depopulation–repopulation cycles associated with steppe land-
scape expansion following extreme or long stadials. The com-
parable timing of stadials and population changes seen in the
archeologic and genetic record suggests that millennial-scale
climate cycles may have been the pacesetter for Europe’s de-
mographic history during the Middle Pleniglacial.
ACKNOWLEDGMENTS. This research was supported by Deutsche For-
schungsgemeinschaft Funding (SFB 806, TP B2) (to M.S.). V.D. acknowledges
support by theEuropean Social Fund, Sectoral Operational Programme Human
Resources Development, Contract POSDRU 6/1.5/S/3—“Doctoral Studies:
Through Science Towards Society,”PCE-2016-0179 Grant CARPATHEMS, and
IFA-CEA C4-08 (FREem). Part of the isotopic analysis were funded by the Uni-
versity of South Florida via an internal grant (to B.P.O.).
1. Hoffecker JF (2009) Out of Africa: Modern human origins special feature:
The spread of modern humans in Europe. Proc Natl Acad Sci USA 106:16040–
2. Hublin J-J (2015) The modern human colonization of western Eurasia: When and
where? Quat Sci Rev 118:194–210.
3. Trinkaus E, et al. (2003) An early modern human from the Pes
¸tera cu Oase, Romania.
Proc Natl Acad Sci USA 100:11231–11236.
4. Reimer PJ, et al. (2013) Intcal13 and Marine13 radiocarbon age calibration curves 0–
50,000 years cal Bp. Radiocarbon 55:1869–1887.
5. Fu Q, et al. (2015) An early modern human from Romania with a recent Neanderthal
ancestor. Nature 524:216–219.
6. Higham T, et al. (2014) The timing and spatiotemporal patterning of Neanderthal
disappearance. Nature 512:306–309.
7. Fu Q, et al. (2016) The genetic history of ice age Europe. Nature 534:200–205.
8. Kindler P, et al. (2014) Temperature reconstruction from 10 to 120 kyr b2k from the
NGRIP ice core. Clim Past 10:887–902.
9. Rasmussen SO, et al. (2014) A stratigraphic framework for abrupt climatic changes
during the last glacial period based on three synchronized Greenland ice-core records:
Refining and extending the INTIMATE event stratigraphy. Quat Sci Rev 106:14–28.
10. Wegwerth A, et al. (2015) Black Sea temperature response to glacial millennial-scale
climate variability: Black Sea temperature and DO cycles. Geophys Res Lett 42:
11. Nowaczyk NR, et al. (2012) Dynamics of the Laschamp geomagnetic excursion from
Black Sea sediments. Earth Planet Sci Lett 351-352:54–69.
12. Fletcher WJ, et al. (2010) Millennial-scale variability during the last glacial in vege-
tation records from Europe. Quat Sci Rev 29:2839–2864.
Staubwasser et al. PNAS Latest Articles
AND PLANETARY SCIENCES
13. Müller UC, et al. (2011) The role of climate in the spread of modern humans into
Europe. Quat Sci Rev 30:273–279.
14. Feurdean A, et al. (2014) Climate variability and associated vegetation response
throughout Central and Eastern Europe (CEE) between 60 and 8 ka. Quat Sci Rev 106:
15. Panagiotopoulos K, et al. (2014) Climate variability over the last 92 ka in SW Balkans
from analysis of sediments from Lake Prespa. Clim Past 10:643–660.
16. Fleitmann D, et al. (2009) Timing and climatic impact of Greenland interstadials re-
corded in stalagmites from northern Turkey. Geophys Res Lett 36:L19707.
17. Darfeuil S, et al. (2017) Sea surface temperature reconstructions over the last 70kyr
off Portugal: Biomarker data and regional modeling. Paleoceanography 31:40–65.
18. Bokhorst MP, et al. (2011) Atmospheric circulation patterns in central and eastern
Europe during the Weichselian Pleniglacial inferred from loess grain-size records.
Quat Int 234:62–74.
19. Haesaerts P, et al. (2010) Charcoal and wood remains for radiocarbon dating upper
Pleistocene loess sequences in eastern Europe and central Siberia. Palaeogeogr
Palaeoclimatol Palaeoecol 291:106–127.
20. Obreht I, et al. (2017) Shift of large-scale atmospheric systems over Europe during late
MIS 3 and implications for modern human dispersal. Sci Rep 7:5848.
21. Sirocko F, et al. (2016) The ELSA-vegetation-stack: Reconstruction of landscape evo-
lution zones (LEZ) from laminated Eifel maar sediments of the last 60,000 years.
Global Planet Change 142:108–135.
22. Nigst PR, et al. (2014) Early modern human settlement of Europe north of the Alps
occurred 43,500 years ago in a cold steppe-type environment. Proc Natl Acad Sci USA
23. Moine O, et al. (2017) The impact of last glacial climate variability in west-European
loess revealed by radiocarbon dating of fossil earthworm granules. Proc Natl Acad Sci
24. Sirocko F, et al. (2013) Multi-proxy dating of Holocene maar lakes and Pleistocene dry
maar sediments in the Eifel, Germany. Quat Sci Rev 62:56–76.
25. Genty D, et al. (2010) Isotopic characterization of rapid climatic events during OIS3
and OIS4 in Villars Cave stalagmites (SW-France) and correlation with Atlantic and
Mediterranean pollen records. Quat Sci Rev 29:2799–2820.
26. Dreybrodt W, Scholz D (2011) Climatic dependence of stable carbon and oxygen
isotope signals recorded in speleothems: From soil water to speleothem calcite.
Geochim Cosmochim Acta 75:734–752.
27. Wang G, Jia Y, Li W (2015) Effects of environmental and biotic factors on carbon
isotopic fractionation during decomposition of soil organic matter. Sci Rep 5:11043.
28. Tremaine DM, Froelich PN, Wang Y (2011) Speleothem calcite farmed in situ: Modern
calibration of δ
O and δ
C paleoclimate proxies in a continuously-monitored natural
cave system. Geochim Cosmochim Acta 75:4929–4950.
¸in V, et al. (2017) Transfer of environmental signals from the surface to the
underground at Ascuns˘
a Cave, Romania. Hydrol Earth Syst Sci 21:5357–5373.
30. Hatté C, et al. (2013) Excursions to C4 vegetation recorded in the upper Pleistocene
loess of surduk (Northern Serbia): An organic isotope geochemistry study. Clim Past 9:
¸in V, et al. (2014) Constraining Holocene hydrological changes in the Carpathian–
Balkan region using speleothem
O and pollen-based temperature reconstructions.
Clim Past 10:1363–1380.
32. Scholz D, Hoffmann DL (2011) StalAge–An algorithm designed for construction of
speleothem age models. Quat Geochronol 6:369–382.
33. Ménot G, Bard E (2012) A precise search for drastic temperature shifts of the past
40,000 years in southeastern Europe: Abrupt SST event in southeastern Europe.
34. Rozanski K, Araguas-Araguas L, Gonfiantini R (1993) Isotopic patterns in modern
global precipitation. Climate Change in Continental Isotope Records, Geophysical
Monograph, eds Swart PK, et al. (American Geophysical Union, Washington, DC), pp
35. Huntley B, et al. (2003) European vegetation during marine oxygen isotope Stage-3.
Quat Res 59:195–212.
36. van Huissteden K, Vandenberghe J, Pollard D (2003) Palaeotemperature reconstruc-
tions of the European permafrost zone during marine oxygen isotope stage 3 com-
pared with climate model results. J Quat Sci 18:453–464.
37. Vandenberghe J, et al. (2014) The last permafrost maximum (LPM) map of the
northern hemisphere: Permafrost extent and mean annual air temperatures, 25–
17 ka BP. Boreas 43:652–666.
38. Sanchi L, Menot G, Bard E (2014) Insights into continental temperatures in the
northwestern Black Sea area during the last glacial period using branched tetraether
lipids. Quat Sci Rev 84:98–108.
39. Antoine P, et al. (2016) Upper Pleistocene loess-palaeosol records from northern
France in the european context: Environmental background and dating of the middle
palaeolithic. Quat Int 411:4–24.
40. Morin E (2008) Evidence for declines in human population densities during the early
Upper Paleolithic in western Europe. Proc Natl Acad Sci USA 105:48–53.
41. Discamps E, Jaubert J, Bachellerie F (2011) Human choices and environmental con-
straints: Deciphering the variability of large game procurement: From mousterian to
Aurignacian times (MIS 5-3) in southwestern France. Quat Sci Rev 30:2755–2775.
42. Vandevelde S, Brochier JÉ, Petit C, Slimak L (2017) Establishment of occupation
chronicles in Grotte Mandrin using sooted concretions: Rethinking the middle to
upper paleolithic transition. J Hum Evol 112:70–78.
43. Conard NJ, Bolus M (2003) Radiocarbon dating the appearance of modern humans
and timing of cultural innovations in Europe: New results and new challenges. J Hum
44. Higham T, et al. (2012) Τesting models for the beginnings of the Aurignacian and the
advent of figurative art and music: The radiocarbon chronology of Geißenklösterle.
J Hum Evol 62:664–676.
45. Böhner U (2008) Sesselfelsgrotte IV—Die Schicht E3 der Sesselfelsgrotte und die
Funde am Abri Schulerloch (Franz Steiner Verlag Stuttgart, Stuttgart, Germany).
46. Devièse T, et al. (2017) Direct dating of Neanderthal remains from the site of Vindija
cave and implications for the middle to upper paleolithic transition. Proc Natl Acad Sci
47. Uthmeyer T (1996) Ein bemerkenswert frühes Inventar des Aurignacien von der
Freilandfundstelle “Keilberg-Kirche”bei Regensburg. Archaeologisches Korrespondenzblatt
48. Davies W, et al. (2015) Evaluating the transitional mosaic: Frameworks of change
from Neanderthals to Homo sapiens in eastern Europe. Quat Sci Rev 118:211–242.
49. Richter J (2016) Leave at the height of the party: A critical review of the middle pa-
leolithic in western central Europe from its beginnings to its rapid decline. Quat Int
50. Douka K, Grimaldi S, Boschian G, del Lucchese A, Higham TF (2012) A new chro-
nostratigraphic framework for the upper palaeolithic of Riparo Mochi (Italy). J Hum
51. Higham T, et al. (2009) Problems with radiocarbon dating the middle to upper pa-
laeolithic transition in Italy. Quat Sci Rev 28:1257–1267.
52. Hublin J-J, et al. (2012) Radiocarbon dates from the Grotte du Renne and Saint-
Césaire support a Neandertal origin for the Châtelperronian. Proc Natl Acad Sci USA
53. Talamo S, et al. (2012) A radiocarbon chronology for the complete middle to upper
Palaeolithic transitional sequence of Les Cottés (France). J Archaeol Sci 39:175–183.
54. Barshay-Szmidt C, et al. (2018) Radiocarbon dating the Aurignacian sequence at
Isturitz (France): Implications for the timing and development of the proto-
aurignacian and early Aurignacian in western Europe. J Archaeol Sci Rep 17:809–838.
55. Wood RE, et al. (2014) The chronology of the earliest upper palaeolithic in northern
Iberia: New insights from L’Arbreda, Labeko Koba and La Viña. J Hum Evol 69:91–109.
56. Camps M, Higham T (2012) Chronology of the Middle to Upper Palaeolithic transition
at Abric Romaní, Catalunya. J Hum Evol 62:89–103.
57. Banks WE, d’Errico F, Zilhão J (2013) Human-climate interaction during the early
upper paleolithic: Testing the hypothesis of an adaptive shift between the proto-
Aurignacian and the early Aurignacian. J Hum Evol 64:39–55.
58. Riel-Salvatore J, Popescu G, Barton CM (2008) Standing at the gates of Europe: Hu-
man behavior and biogeography in the southern Carpathians during the late Pleis-
tocene. J Anthropol Archaeol 27:399–417.
59. Douka K, et al. (2014) On the chronology of the Uluzzian. J Hum Evol 68:1–13.
60. Richards MP, Trinkaus E (2009) Out of Africa: Modern human origins special feature:
Isotopic evidence for the diets of European Neanderthals and early modern humans.
Proc Natl Acad Sci USA 106:16034–16039.
61. El Zaatari S, Grine FE, Ungar PS, Hublin J-J (2016) Neandertal versus modern human
dietary responses to climatic fluctuations. PLoS One 11:e0153277.
62. Hoffecker JF (2011) The early upper Paleolithic of eastern Europe reconsidered. Evol
63. Skrdla P (2017) Middle to Upper Paleolithic transition in Moravia: New sites, new
dates, new ideas. Quat Int 450:116–125.
64. Torfstein A, Goldstein SL, Stein M, Enzel Y (2013) Impacts of abrupt climate changes in
the levant from last glacial dead sea levels. Quat Sci Rev 69:1–7.
65. Tzedakis PC, et al. (2004) Ecological thresholds and patterns of millennial-scale cli-
mate variability: The response of vegetation in Greece during the last glacial period.
66. Nigst PR (2014) First modern human occupation of Europe: The middle Danube region
asacasestudy.Living in the Landscape: Essays in Honour of Graeme Barker, McDonald
Institute Monographs, eds Boyle K, Rabett RJ, Hunt CO (McDonald Institute for
Archaeological Research, University of Cambridge, Cambridge, UK), pp 35–47.
67. Hauck TC, et al. (2016) Neanderthals or early modern humans? A revised C-14 chro-
nology and geoarcheological study of the Szeletian sequence in Szeleta Cave
(Kom. Borsod-Abauj-Zemplen) in Hungary. Archaeologisches Korrespondenzblatt 46:
68. Bicho N, Cascalheira J, Gonçalves C (2017) Early Upper Paleolithic colonization across
Europe: Time and mode of the Gravettian diffusion. PLoS One 12:e0178506.
www.pnas.org/cgi/doi/10.1073/pnas.1808647115 Staubwasser et al.
Supplementary Information for
The impact of climate change on the transition of Neanderthals to modern
humans in Europe
Michael Staubwasser, Virgil Drăgușin, Bogdan P. Onac, Sergey Assonov, Vasile Ersek,
Dirk L. Hoffmann, Daniel Veres
This PDF file includes:
Figs. S1 to S4
Tables S1 to S2
References for SI reference citations
Other supplementary materials for this manuscript include the following:
Supplementary Information Text
Materials and Methods.
Figure S1 shows the location of paleoclimate and environmental records used for
comparison, in addition to archaeologic sites discussed in the text. The two speleothems
presented in this study are from two sites in the East and South Carpathians. Tăușoare Cave (TC)
is situated in the Rodnei Mountains of the East Carpathians, northern Romania (47º 26’ N, 24º 31’
E), at an altitude of 950 m. The cave has a surveyed length of over 8.5 km and is 329 m deep and
hosts an abundance of gypsum crusts and flowers along with the presence of various mixed cation
sulfates (1, 2). Sulfate derives from weathering of pyrite present in the calcite and bituminous
sediments. The process provides significant acidity for calcite dissolution in this cave. Thus, the
isotopic composition of the host-rock calcite rather than soil CO2 dominates the δ13C record of TC
speleothems. Stalagmite 1152 was retrieved from the “Dining Room” at about 625 m from and
about 190 m below the entrance. It is 27 cm long and is made of dark-brown dense calcite (Fig.
S2). The growth axis of stalagmite T-1152 is stable for the first 23 cm and changes direction for
the last 4 cm. Ascunsă Cave (AC) is situated in the Mehedinți Mountains, South Carpathians,
southwestern Romania, at an altitude of 1050 m (45º 00’ N, 22º 36’ E). The cave is 400 m long
and over 200 m deep, formed mostly on the contact between Barremian-Aptian limestone and
Turonian-Senonian wildflysch (melange) (3). Stalagmite POM1 was retrieved from the “White
Chamber”, at about 80 m from the entrance. It is ~30 cm long and composed of alternations of
dense dark calcite and lighter, less dense calcite (Fig. S2). The growth axis presents several slight
For U-Th dating, samples were drilled parallel to growth layers. U and Th were analyzed
by MC-ICP-MS (Neptune+, Thermo) after adding a 229Th – 236U double spike and subsequent
ion-exchange chromatographic separation of U and Th from the sample matrix (4) (Tab. S1). A
detrital Th-correction was applied assuming the average crustal 232Th/238U activity ratio of 1,250
± 0.625. Detrital 230Th correction amounts to within 50 – 350 years for the AC-speleothem, and
10 – 260 years for the TC-speleothem. The final chronology was modeled using the STALAGE
software package (5). Results can be found in Dataset S1.
For stable C and O isotope ratios samples were drilled continuously at 250 µm (TC) and
at 300 µm (AC) resolution and analyzed by gas-IRMS (Thermo 253, and Thermo Delta+) at the
University of Cologne and (TC) and the University of South Florida (AC). Results can be found
in Dataset S1. The commonly applied ‘Hendy-Test’ to exclude kinetic domination of isotope
fractionation was not performed. Both speleothems show relatively low growth rates and poor
resolution of growth layering at the 250 - 300 µm sampling resolution level. The criterion of the
Hendy-Test is to look for covariance of δ13C and δ18O within the same growth layer, which may
hint at potential kinetically dominated isotope fractionation as a result e.g. of rapid degassing or
rapid calcite precipitation. Because the two speleothems generally show little covariance between
δ13C and δ18O overall, the criterion to meet the Hendy-Test is essentially fulfilled. In addition,
cave monitoring water isotope data (δD, δ18O) suggests that oxygen isotopes fractionate at
equilibrium between drip water and calcite, and carbon isotopes may at the worst contain a
kinetic fractionation of 1.0 - 1.5 ‰ (6).
Regarding the association of the TC-stalagmite’s pronounced low δ18O interval between
~ 40.7 and 39.7 ka with GS-10 (between 40.8 and 40.2 ka in the GICC05 chronology) (5), as
opposed to an association with GS-9 – H4 (between 39.9 and 38.2 ka in the GICC05 chronology),
it is worthwhile to assess potential chronologic inaccuracy due to the correction for detrital Th.
The onset of the low δ18O has been dated directly to 40.66 ± 0.47 ka (40.92 ± 0.47 ka without
detrital Th correction) (Tab. S1). An activity ratio for 230Th/232Th of 101 suggests a small detrital
contribution – amounting to an age correction of ~ 260 years. However, detrital 230Th just
amounts to ~0.5% of the total measured 230Th activity based on the average crustal Th/U activity
ratio (1.20) for a detrital contamination. To bring the resulting 40.7 ka onset age in line with the
39.9 ka BP onset age of GS-9 – H4 would require a low detrital Th/U ratio of less than 0.4, well
outside the lower boundary for crustal Th/U of 0.625 and unusual for the given setting.
The Ascunsă Cave δ18O record
Stable O-isotopes combine information on temperature and a number of hydrologic
aspects that may be constrained by data comparison in regional and temporal context. Throughout
the Holocene, the amplitude of δ18O variability in the western Mediterranean, the Alps, and
northern Romania – including TC – was significantly lower (~ 0.5 ‰) than in the eastern
Mediterranean and southern Romania – including AC (7). An even larger amplitude difference
between the AC-δ18O and the TC-δ18O existed during MIS-3. (Fig. S3). This cannot be explained
by a different temperature regime alone (see main article) but must reflect a significant
hydroclimatic contribution to the AC speleothem δ18O records. The pacing of δ18O is very
different between AC and TC, and between AC and Greenland temperatures (8, 9). Thus, it is
unlikely, that the two speleothems were inside the same hydrologic regime but record latitudinal
differences in rain-out history from the same source. However, there is a similar structure
between AC-δ18O record and eastern Mediterranean marine and speleothem records (Fig. S4)
(10). This pattern-similarity between the AC-δ18O record and the eastern Mediterranean is
observed only when the AC-stalagmite's δ18O record is plotted with an inverted scale relative to
the speleothem record from the Levant (Fig. S4). Numeric climate simulations of the last glacial
(11) demonstrate the underlying causality for this apparent inverse correlation between the East
Carpathians and the Eastern Mediterranean. All simulated modes of atmospheric circulation
patterns that contributed to the average glacial precipitation over European and the Mediterranean
show seasonal precipitation anomalies of opposite sign between the eastern Mediterranean and
east-central Europe. This synoptic-scale relationship is corroborated by a similar relationship
between the speleothems of AC and Karaca Cave (12), eastern Turkey.
Interstadial GI-7 stands out in the AC-δ18O record with an anomalous isotope event (Fig.
S3), a minimum more than 1 ‰ lower than all other features in the record. It is also a prominent
feature in the speleothem and marine records from the eastern Mediterranean (Fig. S4). This
could suggest that supply and influence of Mediterranean moisture in the region of the South
Carpathians was highly variable between interstadials. Although the loess record of the Danube
Valley appears to confirm this observation within given chronologic uncertainty (13), there is,
however, no apparent explanation available as to the cause of this moist interval during GI-7.
Fig. S1. Map of the southern and central Europe with rivers (blue: the Danube), coastlines for sea
level 40 m and 60 m lower than at present, and sites discussed in the main text. Geologic records
(dots): (1) Tăușoare Cave; (2) Ascunsă Cave; (3) core MD04-2790; (4) core M72/5-25-GC1; (5)
Willendorf-II paleosol/loess profile; (6) Eifel Maar Lakes; (7) Villars Cave; (8) core MD95-2042.
Archeologic sites: (a1) Sesselfelsgrotte; (a2) Keilberg-Kirche; (a3) Geissenklösterle; (a4) Riparo
Mochi; (a5) Willendorf-II; (a6) Grotta di Fumane; (a7) Grotte Mandrin; (a8) Pes-kὅ Cave; (a9)
Szeleta Cave; (a10) Vindija Cave; (b1) Labeko Koba; (b2) Abric Romani; (b3) Isturitz; (c1)
Grotte du Renne; (c2) Les Cottés; (c3) Saint Césaire; (d1) Oase Cave; (d2) Goyet Cave. The map
was generated with the GeoMapApp software, http://www.geomapapp.org (14).
Fig. S2. U-Th age models and photographs of a) stalagmite 1152, Tăușoare Cave, and b)
stalagmite POM1, Ascunsă Cave.
Fig. S3. A: Greenland NGRIP ice core temperatures on the GICC05 time scale with numbered
interstadials (8). B: δ13C of stalagmite 1152 from Tăușoare Cave. C: δ13C of stalagmite POM1
from Ascunsă Cave. D: δ18O of stalagmite 1152 from Tăușoare Cave. E: δ18O of stalagmite
POM1 from Ascunsă Cave.
Fig. S4. A: Greenland NGRIP ice core temperatures on the GICC05 time scale with numbered
interstadials (8). B: Eastern Mediterranean and Levantine records (10): δ18O data of planktonic
foraminifera (G. ruber) from eastern Mediterranean marine sediment core MD9501, and δ18O data
of a stalagmite from Soreq cave (Israel). C: δ18O data of stalagmite POM1 from Ascunsă Cave.
Note that the scale for the two stalagmites is inverted with respect to each other.
Table S1. U-Th data for stalagmites POM1, Ascunsă Cave, and 1152, Tăușoare Cave. Corrected values include detrital Th correction.
age (ka) ± co rrected
age (ka) ±
1152 top 0,50 1019,5 6,6 3,442 0,034 6,31 0,15 0,0070 0,0002 1,4296 0,0024 0,53 0,01 0,47 0,04 1,4305 0,0024
1152 / XIII 16,25 963,1 34,4 3,054 0,467 97,29 1,28 0 ,1009 0,0038 1 ,4464 0,0033 7,87 0,30 7,81 0,30 1,4567 0,0034
1153 / XII 37,38 1304,6 50,9 2,486 0,125 235,37 2,60 0,1468 0,0013 1 ,5133 0,0035 11,08 0,10 11,04 0,11 1,5298 0,0036
1152 / XI 38,88 1231,0 40,9 1,438 0,054 404,15 5 ,16 0,1545 0,0016 1,5394 0,0039 1 1,48 0,13 11 ,46 0,13 1,5573 0,0040
1152 / X 78,50 2632,7 73,8 1,672 0,062 1069,73 15,75 0,2223 0,0019 1,7095 0 ,0036 15,07 0,14 15,06 0,14 1,7405 0,0037
1152 / IX 81,50 952,4 28,6 2,950 0,101 306,65 2 ,91 0,3108 0,0027 1,8252 0,0044 2 0,10 0,19 20 ,06 0,20 1,8740 0,0045
1152 / VIII 97,63 1854,4 70,8 3,301 0,259 578,76 7,41 0,3371 0,0028 1 ,7996 0,0060 22,31 0,22 22,28 0,22 1,8520 0,0062
1152 / VII 114,88 1674,3 61,1 1 ,603 0,552 1061,09 7,72 0,3323 0 ,0045 1,7126 0,0029 2 3,21 0,35 23 ,20 0,35 1,7610 0,0031
1152 / VI 135,88 942,6 33,0 1,656 0,279 583,06 4,91 0,3352 0,0034 1 ,6309 0,0050 24,75 0,29 24,72 0,29 1,6769 0,0052
1152 XXII 149,63 1086 ,5 53,1 1,051 0,054 1161,09 11,40 0,3674 0 ,0025 1,6158 0,0034 27 ,71 0,22 27,69 0,22 1,6661 0,0036
1152 / V 155,00 1397,5 39,1 1,318 0,040 1151,27 11,89 0,3552 0,0024 1,5133 0,0037 28,77 0 ,24 28,75 0,24 1,5568 0,0039
1152 /IV 174,63 931,3 30,0 2,383 0,085 4 72,40 4,68 0 ,3955 0,0031 1,5418 0,0034 31,80 0,30 31,75 0 ,30 1,5931 0,0037
1152 XVII 179,88 802,2 16 ,3 1,246 0,047 802,42 22,17 0,4078 0,0022 1,5125 0 ,0031 33,68 0,22 33,65 0,22 1,5639 0,0033
1152 XVI 187,88 612,2 13,9 3,442 0,122 243 ,85 6,17 0,4486 0 ,0023 1,5034 0,0033 37,91 0,25 37,80 0,25 1,5610 0,0035
1152 / III 193,63 517,7 21,8 7,404 0,315 101,37 1,00 0,4744 0,0041 1,4903 0 ,0049 40,92 0,45 40,66 0,47 1,5521 0,0054
1152 XXI 196,25 506,4 12,8 3,987 0,106 188 ,71 1,28 0,4862 0 ,0029 1,5296 0,0031 40,81 0,31 40,67 0,31 1,5953 0,0034
1152 XV 201,25 566,7 14,1 3,243 0,093 269 ,37 3,24 0,5044 0 ,0031 1,4390 0,0033 46,02 0,37 45,91 0,37 1,5006 0,0036
1152 XIV 207,25 639,8 12,8 2,372 0,052 433 ,22 3,25 0,5255 0 ,0023 1,4452 0,0030 48,12 0,29 48,05 0,29 1,5105 0,0033
1152 / II 215 ,00 570,5 16,6 3,212 0,100 292,59 2,68 0,5389 0 ,0039 1,4293 0,0040 50 ,36 0,48 50,25 0,49 1,4955 0,0044
1152 / I 234,88 1182 ,4 38 ,7 6,035 0,229 360,50 2,83 0,6021 0,0035 1,4059 0,0034 59,21 0,49 59,11 0,49 1,4803 0,0038
1152/ base 264,00 622,8 3,5 5,433 0,033 225,32 0,50 0 ,6432 0 ,0023 1,3694 0,0027 66,98 0,37 66,81 0,38 1,4472 0,0031
POM 1 / top 32 ,5 0,2 0 ,576 0,005 22,93 0,51 0,1331 0,0031 1,2610 0,0040 12,14 0,30 11,74 0,36 1,2710 0,0041 outside measured profile
POM 1 / III 29,67 38,1 0,2 1,923 0,017 20,95 0,18 0,3460 0,0028 1 ,2746 0,0032 34,18 0,34 33,07 0,61 1,3055 0,0040
POM1/X 94,33 45,2 0,3 0,164 0,005 292,85 3,83 0,3472 0,0040 1,1799 0,0044 37,69 0,54 37,61 0 ,54 1,2003 0,0048
POM 1 / V 119,67 84,8 0 ,4 0,620 0,006 140,69 1,06 0,3366 0,0021 1,1306 0,0027 38,32 0,31 38,14 0,32 1,1457 0,0030
POM1/VII 153,33 37,3 0,2 0,313 0,003 160,76 1 ,64 0,4406 0,0038 1,3985 0,0047 40,53 0,45 40,37 0,46 1,4477 0,0051
POM1/XI 175,33 27,7 0,1 0,079 0,002 473,33 5,60 0,4390 0,0042 1,3776 0,0048 41,13 0,50 41,07 0,50 1,4244 0,0052
POM 1 / II 226,67 27,7 0,2 0,386 0,004 104,84 1,52 0,4777 0,0063 1 ,4225 0,0040 43,73 0,71 43,46 0,72 1,4794 0,0045
POM 1 / I 228,33 20,9 0 ,1 0,396 0,004 80,09 1,01 0,4955 0,0056 1,4684 0,0050 4 3,91 0,63 43 ,56 0,64 1,5324 0,0056
POM 1 XXII 267,00 24,3 0,1 0 ,239 0,003 1 49,77 1,29 0 ,4825 0,0035 1,3771 0,0031 46,11 0,43 45,92 0 ,43 1,4304 0,0034
POM 1 XXIII 289,00 26,4 0,1 0 ,236 0,006 1 62,37 1,71 0 ,4752 0,0042 1,3742 0,0036 45,38 0,51 45,20 0 ,51 1,4262 0,0040
POM 1 /base 33,2 0 ,2 29,626 0 ,162 2,47 0 ,02 0,7209 0,0054 1,3700 0,0036 78,29 0,88 59,06 7,83 1,5708 0,0819 discarded from age model
Table S2. Published 14C ages of archaeologic sequences included in Figure 3 and discussed in the text.
position / layer
sample material conventional
14C age, ka
ka, cal BP (
calibration (c) /
Bayesian model (m);
age ka, cal
BP (p =
ka, cal BP
(p = 68%)
cal BP (p =
age ka, cal
BP (p =
Rennes at Arcy-
34.81 ± 0.21
age 40.7 - 41.4 (m)
age 41.2 - 41.6 (m)
35.50 ± 0.22
Saint Cesaire c2
ejop sup / ejo inf
age 39.7 - 41.1 (m) sterile (ejo inf)
36.00 ± 0.70
summary ejo inf
Les Cottes c3
35.15 ± 0.28
age 39.5 - 40.3 (m)
sterile (layer 5)
age 40.8 - 41.6 (m)
36.23 ± 0.21
model 40.0 - 40.8 40.8 40.0 16
Labeko Koba b1
36.85 ± 0.80
VIII / VII
age 41.4 - 42.0 (m)
36.55 ± 0.75
38.40 ± 0.90
lower / IX
boundary 41.9 - 42.4 (m)
37.40 ± 0.80
upper 42.2 42.2 41.6 41.6
Table S2 continued. Published 14C ages of archaeologic sequences included in Figure 3 and discussed in the text.
position / layer
sample materia l conventional
14C age, ka
ka, cal BP (
calibration (c) /
Bayesian model (m);
age ka, cal
BP (p =
ka, cal BP (p
ka, cal BP (p
age ka, cal
BP (p =
original authors' artifact
Abric Romani b2
36.09 ± 0.23
AR3/AR6 av. U-series
39.06 ± 0.35
AR3/AR6 42.8 42.8 41.7 41.7
37.1 ± 1.1
41.84 ± 1.10
39.95 ± 0.92
41.37 ± 1.06
from GrN-20302 &
41.39 ± 0.58 44.3 - 45.4 (r)
Keilberg a2 layer 2, base
37.50 ± 1.45
37.50 ± 1.25
38.6 ± 1.20
Riparo Mochi a4
37.4 ± 1.3
transition H/G range
age 41.6 - 42.8 (m) sterile
transition I/H range
age 41.8 - 44.0 (m) Mousterian
average of 9 OxA
& GrA dates charcoal 41.7 - 42-9 (c)
average of 9 OxA
& GrA dates charcoal 43.9 - 46.6 (c) IUP
1. Onac BP, Drăguşin V, Papiu F, Theodorescu C-T (2019) Rodna Mountains: Izvorul Tăusoarelor Cave
(Pestera de la Izvorul Tăusoarelor). Cave and Karst Systems of Romania, eds Ponta GML, Onac BP
(Springer International, Cham), pp 47–56.
2. Onac BP, White WB, Viehmann I (2001) Leonite [K2Mg(SO4)2·4H2O], konyaite [Na2Mg(SO4)2·5H2O]
and syngenite [K2Ca(SO4)2·H2O] fromTausoare Cave, Rodnei Mts, Romania. Min Mag 65(1):103-109.
3. Drăguşin V, Vlaicu M, Isverceanu E (2019) Mehedinti Mountains: The Cave from Mohilii Creek
(Ascunsă Cave). Cave and Karst Systems of Romania, eds Ponta GML, Onac BP (Springer
International, Cham), pp 171–173.
4. Hoffmann D, et al. (2007) Procedures for accurate U and Th isotope measurements by high precision
MC-ICPMS. Int J Mass Spec 264(2-3): 97-109.
5. Scholz D, Hoffmann DL (2011) StalAge – An algorithm designed for construction of speleothem age
models. Quat Geochronol 6(3–4):369–382.
6. Dragusin V, et al. (2017) Transfer of environmental signals from the surface to the underground at
Ascunsă Cave, Romania. Hydrol Earth Syst Sci 21(10):5357–5373.
7. Drăguşin V, et al. (2014) Constraining Holocene hydrological changes in the Carpathian–Balkan region
using speleothem δ18O and pollen-based temperature reconstructions. Clim Past 10(4):1363–1380.
8. Rasmussen SO, et al. (2014) A stratigraphic framework for abrupt climatic changes during the Last
Glacial period based on three synchronized Greenland ice-core records: refining and extending the
INTIMATE event stratigraphy. Quat Sci Rev 106:14–28.
9. Kindler P, et al. (2014) Temperature reconstruction from 10 to 120 kyr b2k from the NGRIP ice core.
Clim Past 10(2):887–902.
10. Almogi-Labin A, et al. (2009) Climatic variability during the last ~90ka of the southern and northern
Levantine Basin as evident from marine records and speleothems. Quat Sci Rev 28:2882-2896.
11. Hofer D, et al. (2012) Simulated winter circulation types in the North Atlantic and European region for
preindustrial and glacial conditions: Glacial Circulation Types. Geophys Res Lett 39(15): DOI
12. Rowe PJ et al. (2012) Speleothem isotopic evidence of winter rainfall variability in northeast Turkey
between 77 and 6 ka. Quat Sci Rev 45:60-72.
13. Obreht I, et al. (2017) Shift of large-scale atmospheric systems over Europe during late MIS 3 and
implications for Modern Human dispersal. Sci Rep 7:5848.
14. Ryan, W.B.F, et al (2009) Global Multi-Resolution Topography synthesis, Geochem. Geophys.
Geosyst., 10, Q03014, doi:10.1029/2008GC002332.
15. Hublin J-J, et al. (2012) Radiocarbon dates from the Grotte du Renne and Saint-Césaire support a
Neanderthal origin for the Châtelperronian. Proc Natl Acad Sci U S A 109(46):18743–18748.
16. Higham T, et al. (2014) The timing and spatiotemporal patterning of Neanderthal disappearance.
17. Talamo S, et al. (2012) A radiocarbon chronology for the complete Middle to Upper Palaeolithic
transitional sequence of Les Cottés (France). J Archaeol Sci 39(1):175–183.
18. Wood RE, et al. (2014) The chronology of the earliest Upper Palaeolithic in northern Iberia: New
insights from L’Arbreda, Labeko Koba and La Vina. J Hum Evol 69:91–109.
19. Camps M, Higham T (2012) Chronology of the Middle to Upper Palaeolithic transition at Abric
Romani, Catalunya. J Hum Evol 62(1):89–103.
20. Böhner U (2008) Sesselfelsgrotte IV - Die Schicht E3 der Sesselfelsgrotte und die Funde am Abri
Schulerloch (Franz Steiner Verlag Stuttgart, Germany).
21. Uthmeyer T (1996) Ein bemerkenswert frühes Inventar des Aurignacien von der Freilandfundstelle
“Keilberg-Kirche” bei Regensburg. Archaeologisches Korrespondenzblatt 26:233–248.
22. Higham T, et al. (2012) Τesting models for the beginnings of the Aurignacian and the advent of
figurative art and music: The radiocarbon chronology of Geißenklösterle. J Hum Evol 62(6):664–676.
23. Douka K, et al. (2012) A new chronostratigraphic framework for the Upper Palaeolithic of Riparo
Mochi (Italy). J Hum Evol 62(2):286–299.
24. Nigst PR, et al. (2014) Early modern human settlement of Europe north of the Alps occurred 43,500
years ago in a cold steppe-type environment. Proc Natl Acad Sci 111(40):14394–14399.