Content uploaded by Tiffanie Fourcade
Author content
All content in this area was uploaded by Tiffanie Fourcade on Sep 18, 2024
Content may be subject to copyright.
Research Article
Environmental changes in SW France during the Middle to Upper
Paleolithic transition from the pollen analysis of an eastern North
Atlantic deep-sea core
Tiffanie Fourcade1,2* , María Fernanda Sánchez Goñi2,3, Christelle Lahaye1, Linda Rossignol2and Anne Philippe4
1
Archéosciences Bordeaux, UMR 6034, Université Bordeaux Montaigne, CNRS, Maison de l’Archéologie, Esplanade des Antilles, 33600 Pessac, France;
2
Environnements et Paléoenvironnements Océaniques et Continentaux (EPOC), UMR CNRS 5805, Université de Bordeaux, 33600 Pessac, France;
3
École Pratique des
Hautes Études, EPHE PSL University, Paris, France and
4
Laboratoire de Mathématiques Jean Leray, Nantes Université, 44322, Nantes, France
Abstract
Evaluating synchronies between climate and cultural changes is a prerequisite for addressing the possible effect of environmental changes on
human populations. Searching for synchronies during the Middle-Upper Paleolithic transition (ca. 48–36 ka) is hampered by the limits of
radiocarbon dating techniques and the large chronological uncertainties affecting the archaeological and paleoclimatic records, as well by
their low temporal resolution. Here, we present a high-resolution, pollen-based vegetation record from the Bay of Biscay, sea surface tem-
perature changes, additional
14
C ages, and a new IRSL date on the fine-sediment fraction of Heinrich Stadial (HS) 6. The IRSL measure-
ments give an age of ca. 54.0 ± 3.4 ka. The paleoclimatic results reveal a succession of rapid climatic changes during the Middle-Upper
Paleolithic transition in SW France (i.e. D-O 12–8 and two distinct climatic phases during HS 4). Comparison of the new paleoclimatic
record with chronologically well-constrained regional archaeological changes shows that no synchronies exist between cultural transitions
and environmental changes. The disappearance of Neanderthals and the arrival of Homo sapiens in SW France encompassed a long-term
forest opening, suggesting that Homo sapiens may have progressively replaced Neanderthals from D-O 10 to HS 4 through competition for
the same ecological niches.
Keywords: Dansgaard-Oeschger cycles, Heinrich events,
14
C dating, Vegetation, Bayesian age modeling, Luminescence dating,
Neanderthal, Homo sapiens
(Received 17 May 2021; accepted 5 April 2022)
INTRODUCTION
The role of climate change as a driver of cultural changes is a
recurrent topic in the current scientific literature. Some authors
explain the origin of technical or demographic changes during
the Mousterian, the Middle-Upper Paleolithic transition or the
Neolithic through environmental changes (Richerson et al.,
2005; Berger and Guilaine, 2009; Borrell et al., 2015; Defleur
et al., 2020). Others postulate that expansions and contractions
of eco-cultural niches have been caused by climate and environ-
mental changes (Banks et al., 2008,2013; Vignoles et al., 2020).
Still others advocate that cultural changes are independent of
environmental changes (Pétillon et al., 2016).
Of particular importance is the possible effect of environmen-
tal changes for late Neanderthal (Homo neanderthalensis) adapta-
tion and its disappearance, which occurred during the Middle and
Upper Paleolithic transition (ca. 50–40 ka, Greenbaum et al.,
2019). Several important events took place during this period,
such as the end of different Mousterian lithic techno-complexes
(LTCs); the presence of a transitional one, the so-called and still
debated Châtelperronian in SW France (Higham et al., 2010;
Hublin et al., 2012; Gravina et al., 2018); and the arrival of
Homo sapiens in Western Europe, who brought the Aurignacian
culture. Although debated, Châtelperronian industry is associated
with Neanderthal remains and Mousterian elements (Hublin
et al., 2012; Gravina et al., 2018). This LTC is characterized by
Upper Paleolithic features: curved backed blades, end-scrapers,
and bladelets, but also ornaments, pigments, and bone industries
(e.g., d’Errico et al., 2003; Dayet et al., 2014; Ruebens et al., 2015).
The Aurignacian, and more broadly the emergence of Upper
Paleolithic industries, is considered as a clear rupture with the
Middle Paleolithic (Mellars, 2004) and has been subdivided in
three techno-complexes—Proto-Aurignacian, Early Aurignacian,
and Aurignacian—with only partial geographical overlap.
The Proto-Aurignacian has been defined and located on the
western Mediterranean rim within northern Italy (Bartolomei
et al., 1994), the Basque country and the French Pyrenees
(Laplace, 1966; Normand and Turq, 2005), SE France (Bazile,
1977; Onoratini, 1986,2006), and Catalonia (Maroto et al.,
1996). It is characterized by the production of small rectilinear
flakes, larger pointed, convex flakes, and large rectilinear flakes.
*Corresponding author email address: tiffanie.fourcade@u-bordeaux-montaigne.fr
Cite this article: Fourcade T, Sánchez Goñi MF, Lahaye C, Rossignol L, Philippe A
(2022). Environmental changes in SW France during the Middle to Upper Paleolithic
transition from the pollen analysis of an eastern North Atlantic deep-sea core.
Quaternary Research 1–18. https://doi.org/10.1017/qua.2022.21
© University of Washington. Published by Cambridge University Press, 2022
Quaternary Research (2022), 1–18
doi:10.1017/qua.2022.21
https://doi.org/10.1017/qua.2022.21 Published online by Cambridge University Press
Bladelets can be arranged in Dufour bladelets or Dufour subtype.
Perforated shell ornaments have been found (Taborin, 1993).
Following the Proto-Aurignacian is the Early Aurignacian,
whose sites containing this LTC are located from the Atlantic to
the Near East. It is characterized by the presence of recurrent
characters, such as carinated scrapers, blades with lateral
retouches, and split-base points (Bon, 2002). The production of
blades and flakes is made from distinct chains of operation
depending on the activity. Blades, which are wide and thick, are
produced by soft percussion on unipolar nuclei and are intended
for domestic use (Tartar et al., 2005).
Determining whether these technological variabilities are asso-
ciated with environmental and climatic changes, requires reliable
and robust chronologies to estimate the appearance and duration
of each LTC. The increasing use of chronological modeling using
Bayesian statistics in archaeological sciences (Higham et al., 2010;
Discamps et al., 2011; Banks et al., 2013) aims to fill this gap for
SW France during MIS 5–3. Many studies on the biostratigraphy
and chronology of archaeological sites in SW Europe have been
carried out (Guibert et al., 2008; Vieillevigne et al., 2008; Discamps
et al., 2011; Jaubert, 2011; Jaubert et al., 2011; Discamps and
Royer, 2017) to investigate the temporal variability and spatial
variabilities of these LTCs during the Middle to Upper Paleolithic
transition.
These technological changes occurred within the middle part
of MIS 3. This time interval was marked by millennial to centen-
nial climate changes, a succession of warming and cooling events
originally detected in Greenland ice cores, the Dansgaard-
Oeschger (D-O) cycles (Dansgaard et al., 1984,1993), and in
the North Atlantic (Bond and Lotti, 1995). The period between
50 ka and 36 ka includes D-O cycles 12–8 and the large iceberg
discharge event called Heinrich event (HE) 4. Each cycle starts
with a D-O warming event, followed by a progressive cooling,
forming the Greenland Interstadial (GI) warm phase, and a sub-
sequent cold phase called Greenland Stadial (GS) (Rasmussen
et al., 2014). HE 4 was associated with the Heinrich Stadial
(HS) 4 cold phase, which lasted ca. 2,000 years (Sanchez Goñi
and Harrison, 2010). Deep-sea and terrestrial pollen records
and speleothem archives from Europe and its margin show that
D-O cycles and HEs have strongly affected European ecosystems
(Fletcher et al., 2010) and, in particular, those of western France
(Genty et al., 2003; Sánchez Goñi et al., 2008, 2013; Discamps
et al., 2011). Marine palynology allows the reconstruction of
long and continuous regional environmental sequences (Ning
and Dupont, 1997; Moss and Kershaw, 2007; Oliveira et al.,
2014). The comparison between pollen and other marine proxies,
some of which are suitable for dating, provides good chronologies
for terrestrial and marine paleoenvironmental and climate
changes (Sánchez Goñi et al., 1999).
Some authors have postulated that Neanderthal disappearance
was caused by a volcanic eruption (Golovanova et al., 2010)or
abrupt cooling (Finlayson and Carrión, 2007). Other authors
have proposed a competition between the two human groups in
Western Europe (d’Errico and Sánchez Goñi, 2003; Sepulchre
et al., 2007). The hypothesis of competition has been corroborated
by an eco-cultural modeling approach (Banks et al., 2008) and
more recently by a numerical model of interspecific competition
including the “culture level”of a species as a variable that interacts
with population size (Gilpin et al., 2016). Further, a new spatially
resolved numerical hominin dispersal model that simulates the
migration and interaction of H.sapiens and Neanderthals during
the rapid D-O events shows that these climatic events were not
the major cause of the disappearance of Neanderthals. A realistic
disappearance of Neanderthals requires the choice of H.sapiens as
a more effective population in exploiting scarce glacial food
resources as compared to Neanderthals (Timmermann, 2020).
Climate variability could have played a role in the dietary
behavior of hunter-gatherer groups. Hodgkins et al. (2016)
proposed that during part of the last ice age, MIS 4 and 3 (ca.
73–40 ka), treatments of carcasses (cut and percussion marks)
byNeanderthalsatthePechdel’AzéIVandRocdeMarsal
(Dordogne) sites were more frequent during cold than warm cli-
mates. The cold climates would be associated with nutritional
stress, as Neanderthals intensified their efforts to search for calo-
ries. These studies show that changes in the strategies of subsis-
tence, and thus in their technical adaptation, would have been
conditioned by the characteristics of the ecosystems in which
they lived. However, it remains difficult to disentangle the role
of climatic variations and deliberate cultural choices in their sub-
sistence strategies (Discamps et al., 2011).
The possible effect of climate change on late Neanderthal tech-
nical adaptations and their replacement by H.sapiens is therefore
still an open question, and can only be addressed if a synchronic-
ity is found between climatic events and biological and technolog-
ical changes. SW France is one of the best regions for tackling this
question due to its richness of dated archaeological sites and the
availability of deep-sea pollen and speleothem-based vegetation
and climatic records. However, correlating environmental and
archaeological records is a complicated task due to their chrono-
logical uncertainties.
Marine archives can be dated by numerical methods such
as tephrochronology, magnetic events,
14
C, and OSL (e.g., Kuehl
et al., 1996; Stokes et al., 2003; Waelbroeck et al., 2019).
However, some marine records have not been dated using these
methods yet and their chronology is based on the synchronization
of the δ
18
O of planktonic foraminifera, SST, or pollen profiles
with the δ
18
O ice core record. Records younger than ca. 45 ka
are most commonly based on
14
C dating, but regional differences
in radiocarbon quantities between marine and terrestrial organ-
isms have been demonstrated, and particularly for the reservoir
effect affecting marine records (Monge Soares, 1993; Bard et al.,
1994). This effect remains a major concern in the radiocarbon
community, because it introduces an additional source of error
that is often difficult to quantify accurately and requires a correc-
tion (Stuiver and Braziunas, 1993; Reimer and Reimer, 2001).
Luminescence dating, which can avoid the problem of C reser-
voirs and age calibration, has been applied to many deep-sea
cores in different regions (e.g., Pacific, Arctic, Baltic, and
Atlantic oceans), but not in the Bay of Biscay (e.g., Wintle and
Huntley, 1979,1980; Olley et al., 2004; Armitage and Pinder,
2017).
The aim of our study is threefold: (1) document at higher tem-
poral resolution the environmental and climatic changes in SW
France during the Middle-Upper Paleolithic transition (ca. 50–
40 ka) and improve its chronology; (2) create a new well-
constrained chronology for the LTCs in SW Europe (i.e.,
Châtelperronian, Proto-Aurignacian, and Early Aurignacian);
and (3) compare the paleoenvironmental changes with the new
chronologically constrained succession of these LTCs. To meet
these aims, we increased the sampling resolution of MD04-2845
deep-sea core (Sánchez Goñi et al., 2008) to reach a 300–
400-year resolution between samples and improve the age
model by adding new absolute control points using radiocarbon
and IRSL dating techniques (Thomsen et al., 2008; Thiel et al.,
2 T. Fourcade et al.
https://doi.org/10.1017/qua.2022.21 Published online by Cambridge University Press
2011a,b; Buylaert et al., 2012; Kars et al., 2012; Lowick et al.,
2012). MD04-2845 core retrieved at 45°N in the Bay of Biscay
(Northeastern Atlantic) contains pollen grains and fine-grained
sediments coming mainly from the hydrographic basins of south-
western France and transported by rivers that have theirs sources
in the Massif Central and Pyrenees.
PRESENT-DAY ENVIRONMENTAL SETTING
Oceanographic setting and sediment supply
The Bay of Biscay (48°N–43°N) is a gulf of the northeast Atlantic
Ocean, limited geographically to the north by the northern Biscay
continental margin and to the south by the Iberian-Cantabrian
margin (Fig. 1). The main surface current in the Bay of Biscay
is the European Slope Current (ESC), flowing northward as far
as the Armorican and Celtic coasts (Pingree and Cann, 1990).
In winter, the ESC reaches its maximum intensity by the intrusion
of the strong Iberian Poleward Current (IPC) flowing along the
Iberian margin. This current, which brings warm waters into
the southeastern part of the Bay of Biscay, is known as the
Navidad current and leads to the appearance of a thermohaline
front along the shelf (Castaing et al., 1999). This current branches
off in the Gulf and generates a cyclonic cell circulation at 46°N,
6.5°W (Colas, 2003).
Three submarine canyons are present in the Bay of Biscay:
Cap-Ferret, Cap-Breton, and Torrelavega canyons. The sediment
supply preserved in the marine core comes mainly from rivers
in France (Jouanneau et al., 1999). Five main rivers contribute
to this sediment supply: the Vilaine, Charente, Adour, Gironde,
and Loire, with the latter two contributing the most (Lapierre,
1967; Jouanneau et al., 1999). The sediments of the Adour,
which drains the western Pyrenees, indirectly feed the Cap-
Breton Canyon (Brocheray et al., 2014; Mazières et al., 2014),
although the head of the canyon became disconnected from the
river in AD 1310 (Klingebiel and Legigan, 1978). The Garonne
watershed feeds the Cap-Ferret canyon (Brocheray, 2015). The
Gironde estuary is the most important contributor—at least
60% of its sediment reaches the Bay of Biscay (Castaing, 1981).
Some of the suspended matter discharged from the Gironde estu-
ary and the Charente is diverted northwards by river flows and
density currents where it contributes to the formation of mudflats
(Allen and Castaing, 1977; Castaing, 1981; Froidefond et al., 1996;
Jouanneau et al., 1999). Half of the small watersheds of the
Cantabrian margin are connected to small straight canyons, lead-
ing to the Cap-Breton Canyon, while others feed into a network
of canyons, converging to form the Torrelavega Canyon
(Brocheray, 2015). These studies show that terrestrial fine-grained
(<60 μm in diameter) sediment (Weber et al., 1991) containing
pollen grains in the Bay of Biscay is mainly dominated by input
from the Loire and Garonne river basins.
Climate and vegetation
The climate of southwestern Europe is controlled by atmospheric
perturbations from the west (e.g., Feser et al., 2015). In southeast-
ern Bay of Biscay, winds are variable through the year, but show
seasonal patterns with a northwesterly direction in spring and
summer and a southwesterly position in autumn and winter
(Lavin et al., 2006). The prevailing climate in the Bay of Biscay
is controlled by the NAO (North Atlantic Oscillation), which is
defined as the atmospheric pressure gradient between the
Azores high and the Icelandic low. Depending on its positive or
negative mode, the position and intensity of the westerly winds
change, bringing more or less precipitation to western Europe
(Hurrell, 1995) A positive NAO leads to a higher winter storm
activity over the Atlantic, warm and wet winters in northern
Europe, and dry winters in southern Europe. On the contrary, a
negative NAO leads to weaker winter storms crossing on a
more west-to-east pathway, bringing moist air into southern
Europe and cold air to northern Europe (e.g., Visbeck et al.,
2001). The climate of southwestern Europe, and in particular
southwestern France, from which the pollen grains come, is
humid for much of the year with annual precipitation in the
order of 500–1000 mm and temperatures in winter between 0–
8°C and in summer between 15–22°C (Serryn, 1994).
This temperate oceanic climate allows development of the
deciduous temperate Atlantic forest in western Europe (Polunin
and Walter, 1985). This forest nowadays is composed of oaks
(Quercus) in the lower elevations, associated with birches (Betula)
on acidic soils or hornbeams (Carpinus)onbasicsoils.Quercus
is found associated with beech (Fagus sylvatica) in the higher eleva-
tions, where rainfall is higher (900–1500 mm) and average annual
temperatures vary between 8–10°C. In the Massif Central, the dom-
inant conifers, Pinus (Pinus sylvestris), spruce (Picea alba), and fir
(Abies alba) colonize altitudes >600 m (Ozenda, 1982). In the
Pyrenees, Fagus sylvatica and, locally, Abies or Pinus sylvestris,
occupy the montane level from 900 m above sea level, while hooked
pine (Pinus uncinata) and rhododendrons (Ericaceae) colonize the
subalpine level (Ozenda, 1982).
MATERIAL AND METHODS
Deep-sea core MD04-2845
The MD04-2845 core (45°21’N, 5°13’W, 4175 m water depth)
(Fig. 1) was taken from the Dôme Gascogne, during the
ALIENOR cruise, with the oceanographic vessel Marion
Dufresne equipped with a Calypso piston corer (Turon et al.,
2004). The marine sedimentary core is located ∼350 km from
the French coast and influenced by the Bottom Water (BW) flow-
ing at >1500 m, which is composed of the cold Northeast Atlantic
Deep Water (NADW) and the Antarctic Bottom Water (ABW).
This core is mostly composed of hemipelagic clayey mud sedi-
ments with scattered silty strata, with carbonate contents ranging
from 10–65% and an organic carbon content of <1% (Daniau
et al., 2009). This core showed a well-preserved, continuous sedi-
mentary sequence and was found not to be affected by turbidites.
Chronology
The initial chronology of core MD04-2845 covering the last 140
ka was constructed from 17 AMS
14
C dates (Sánchez Goñi
et al., 2008; Daniau et al., 2009) and 11 isotopic events presented
in the ACER database (Sánchez Goñi et al., 2017). The new
MD04-2845 core chronology is a revised version with the new
addition of three radiocarbon and one IRSL dates (Table 1).
Radiocarbon dating. Three AMS
14
C dates were obtained at Beta
Analytic Inc (USA) on samples of monospecific planktonic fora-
minifera—Neogloboquadrina pachyderma (s)—from the levels of
maximum abundance of this species, at depths 1164, 1191, and
1226 cm, corresponding to HS 4. Dating required ∼10 g of fora-
minifera. The reservoir effect in the Bay of Biscay was calculated
from 10 stations, located from Brittany to the Arcachon basin
Middle‐to‐Upper Paleolithic environmental changes, SW France 3
https://doi.org/10.1017/qua.2022.21 Published online by Cambridge University Press
(Broecker and Olson, 1961; Mangerud et al., 2006; Tisnérat-
Laborde et al., 2010), and recorded online in the Marine
Reservoir Database (Reimer and Reimer, 2001). The calculated
reservoir age is 383 ± 53 years.
Luminescence dating. The analytical strategy for dating the
MD04-2845 core was to select layers containing ice rafted detritus
(IRD)—coarse sediments coming from the melting of massive ice-
berg discharges in the North Atlantic from the fragmentation of
the Laurentide (HS) or from the European ice sheets. We assumed
as a working hypothesis that either the IRD was exposed to the
sun just before covering by the ice cap, or that the luminescence
signal had been reset by the shearing of ice sheets (Bateman et al.,
2012,2018). The ice breakup allowed for iceberg discharges,
Figure 1. Partial European map showing the location of the MD04-2845 deep-sea core (black star) and the other cores discussed in the text: Iberian deep-sea cores
(white stars) MD99-2331 (Naughton et al., 2009) and MD95-2039 (Roucoux et al., 2005), Greenland ice core (NGRIP, black square) (Rasmussen et al., 2014), and
Villars Cave speleothem (purple star) (Genty et al., 2010). Major western rivers (blue lines) and Middle-Upper Paleolithic transition archaeological sites available
in the database (black diamonds) are located on the map. The main oceanic currents and their names are represented in orange and red arrows (modified from
Mary et al., 2017).
Table 1.
14
C, IRSL, and biostratigraphic ages with their respective uncertainties and depths used in the Bayesian depth-age model. The calendar ages of D-O events
(D-O 10–17) are based on the tuning between increase in Atlantic forest from MD04-2845 deep-sea core and rapid warming events at the start of GIs, that have an
estimated age and *uncertainties (Wolff et al., 2010; Rasmussen et al., 2014).
Laboratory ID/event Depth (cm)
14
C age (ka BP)
Error
(± ka BP) Calendar age (ka)
Error*
(± ka) References
SacA-002977 1048 29.87 0.39 Sánchez Goñi et al., 2008
SacA-002978 1060 30.67 0.42 Sánchez Goñi et al., 2008
SacA-002976 1078 31.35 0.46 Sánchez Goñi et al., 2008
Beta-491854 1164 33.92 0.22 This study
Beta-491855 1191 33.39 0.20 This study
Beta-491856 1226 34.14 0.24 This study
D-O 10 1260 41.41 0.82 Sánchez Goñi et al., 2008
D-O 11 1290 43.29 0.87 Sánchez Goñi et al., 2008
D-O 12 1335 46.81 0.96 Sánchez Goñi et al., 2008
D-O 14 1450 54.17 1.15 Sánchez Goñi et al., 2008
D-O 17 1510 59.39 1.29 Sánchez Goñi et al., 2008
BDX24931 1535–1545 53.60 3.4 This study
4 T. Fourcade et al.
https://doi.org/10.1017/qua.2022.21 Published online by Cambridge University Press
estimated to last from 50–1,500 years (Roche et al., 2004; Ziemen
et al., 2019), that produced additional sediment supply and larger
grain sizes, thus optimizing the quantity and quality of the sam-
pled material. The MD04-2845 core was stored at the EPOC lab-
oratory (Université de Bordeaux) in a refrigerated and dark
environment, ideal conditions for luminescence dating. Due to
the inherent constraints of the method (exposure to light), we
worked on the archive part, whose surface was only exposed to
light during core cutting. We sampled three levels at 936–
944 cm, 1076–1081 cm, and 1535–1545 cm depths, containing
IRD and the first and third ones corresponding to HS 3 and
HS 6, respectively, and the second one to a very low-IRD layer
(Supplementary Fig. 1). The samples were collected according
to protocols described by Armitage and Pinder (2017) and
Nelson et al. (2019). The samples were sieved, but only the sample
corresponding to depths of 1535–1545 cm provided enough
grains for dating. Due to the small amount of available material,
the 41–60 μm grain size fraction was selected because it was the
most abundant in proportion. The grains were successively treated
with HCl (10%) and then H
2
O
2
(30%) for 24 hours to remove
carbonates and organic matter, respectively. They were then
treated with 10% HF for 10 minutes to clean the grain surfaces
and finally treated with 10% HCl to remove all fluorides eventually
created during the precedent step. After rinsing and drying, the pol-
ymineral fraction was mounted on stainless steel cups previously
sprayed with silicone oil using a 1 mm mask. The pIR-IR
290
lumi-
nescence signals were measured at the Archeosciences Bordeaux lab-
oratory (Univ. Bordeaux Montaigne) using a Freiberg Instruments
Lexsyg SMART reader with an internal beta
90
Sr/
90
Y source deliver-
ing a dose rate of 0.171 ± 0.004 Gy/s (Risø calibration quartz batch
113, Hansen et al., 2015) to the polymineral fraction at the time of
measurement. The fraction was stimulated with near-infrared
diodes emitting at 850 nm. The IRSL signal was detected in the
blue-violet region with an H7360-02 Photon Counting Head in
the UV/Vis region (410 nm) through a combination of optical
filters (Schott BG3, 3 mm + Semrock 414/46 BrightLine HC
interference filter) placed in front of a Hamamatsu H7360-02
photomultiplier tube (PMT).The equivalent dose (D
e
) was deter-
mined using the IRSL single-aliquot regenerative dose (SAR) pro-
tocol (Murray and Wintle, 2003) adapted for the pIR-IR
290
protocol (Supplementary Table 1), with a stimulation temperature
of 50°C followed by a high temperature measurement at 290°C
(pIR-IR
290
) (Thiel et al., 2011a). The data analysis was performed
with Analyst software (Duller, 2015). For each aliquot (n = 20),
pIR-IR
290
measurements passed all acceptance criteria: the recy-
cling ratio averages 1.01 ± 0.03, within 5%, the recuperation ratios
were also <5%, and the maximum paleodose error is <10%.
pIR-IR
290
curves are provided for the sample (Fig. 2). D
e
value
was calculated using the CAM (Central Age Model) (Galbraith
et al., 1999), an arithmetic average and the Average Dose Model
(Guérin et al., 2017).The D
e
results are similar in all three cases
(Table 2) and equal within uncertainties.
The external alpha, beta, and gamma dose rates received by
feldspar grains were deduced from high-purity low-background
BEGe gamma spectrometry measurements (Guibert and
Schvoerer, 1991). No significant disequilibrium in U-series was
detected (Table 3). The internal dose rate of the feldspars was
derived from internal K contents, assumed to be 12.5 ± 0.5%
(Huntley and Baril, 1997). An a-value of 0.08 ± 0.02 was assumed
(Rees-Jones, 1995). The most important issue is the water content,
which could lead to a significant underestimation or overestima-
tion of the age obtained (Aitken, 1998). There exists a large
variability in water content values considered for marine sedi-
ments in the previous studies (Supplementary Table 2). A value
of 40% water content was measured and confirmed by further
measurements (33–43%) of other samples in the core. An uncer-
tainty value of 10% was assigned to cover all realistic uncertain-
ties. Considering the depth of the sample, we also estimated
that the cosmic dose received was negligible (Prescott and
Hutton, 1994; Supplementary Table 2).
Bayesian age-depth model. A new depth-age curve was developed
using all available radiocarbon ages, the paleodosimetric age
obtained in the present study, and five isotopic events (Table 1).
The model, named BaCON (Blaauw and Christen, 2011), is a
Bayesian age modeling in sedimentary sequences that requires
mainly prior information about sedimentation rates, which is dif-
ficult to obtain for long sedimentary sequences. This type of model
(i.e., BaCON) does not handle sudden variations in sedimentation
rates, which are found during periods of deglaciation and ice rafted
debris deposition (Sánchez Goñi et al., 2017). On the contrary, our
Bayesian modeling used in Archaeological Sciences does not con-
sider sedimentation rate as a prior for Bayesian analysis embedded
in the chronological model (Lanos and Philippe, 2018). For this
reason, we used the ChronoModel v. 2.0 software (Lanos and
Dufresne, 2019) to construct the most reliable chronological
model for core MD04-2845. The prior information included in
the model is a stratigraphic order according to the depth of
dated samples (see Lanos and Philippe, 2018, for a description
of the chronological model). The calibration step for radiocarbon
ages was performed using Marine20 (Heaton et al., 2020). The
posterior distribution of collection dates/ages is approximated
using samples simulated by Markov chain Monte Carlo
(MCMC) algorithms. Then, the MCMC samples from the joint
posterior distribution are analyzed in the ArchaeoPhases
R-package v. 1.4.5 (Philippe and Vibet, 2020).
Firstly, we represent the 95% credible interval for each dated
sample of our collection (Fig. 3). The credible interval is calculated
from the posterior distribution of each dated sample. This is the
shortest interval that contains the date of sample with 95% poste-
rior probability (i.e., there is a 95% probability that the unknown
date of sample falls within this interval). Then, we estimate the age-
depth curve from this sequence of ages and their depth. The curve
is estimated using the classical local regression (LOESS), which is
applied to express the age as a function of depth. The estimate
of the curve depends on the collection of ages, which are unknown,
but their posterior distribution is provided by the chronological
model. Thanks to the MCMC sample, we can easily estimate the
posterior distribution of the depth-age curve at each value of the
depth. Therefore, we can predict the age of undated levels. For
each depth value, we summarize the posterior distribution of age
by its median value and 68% and 95% credible intervals.
Pollen study
Fifty-five new samples were analyzed between 1134–1317 cm
depth, corresponding to a resolution of 300–400 years. The
extraction protocol consists of sieving about 2–5 cc of sediment
at 150 μm to separate the lower fraction containing pollen and
the upper fraction composed mainly of foraminifera. A known
concentration of an exotic spore, Lycopodium, was added to the
sediment at the beginning of the treatment to calculate the total
sporo-pollen concentration and that of each taxon (Stockmarr,
1971). This sediment was chemically attacked (cold HCl at 10%,
25%, and 75%, and cold HF at 45% and 75%) to remove
Middle‐to‐Upper Paleolithic environmental changes, SW France 5
https://doi.org/10.1017/qua.2022.21 Published online by Cambridge University Press
carbonates and silica. The pollen residue was then filtered through
a nylon filter of 10 μm and mounted on a microscope slide in a
bi-distilled glycerin medium, which allows the mobility of pollen
grains and their identification in polar and equatorial view. For
each sample, at least 20 taxa and a main sum of >100 pollen,
excluding Pinus, were counted using a Zeiss Axioscope optical
microscope at magnifications ×400 and ×1000 (immersion oil).
Pinus is over-represented in marine sediments (Heusser and
Balsam, 1977) and including it in the pollen main sum would
mask variations in the percentages of other pollen taxa. The
sum of pollen grains excluding Pinus ranges from 59–151
(including pine, from 184–888 grains), with only six samples
having pollen sums between 59–100 pollen grains. Pollen
percentages were calculated on the total pollen counted, excluding
Pinus, aquatic plants, spores, and undetermined pollen grains.
The different pollen taxa were grouped by their ecological
affinity into five main ecological groups (Fig. 4): (1) Atlantic
deciduous forest, composed mainly of deciduous Quercus-type,
Alnus,Corylus,Carpinus, and Fagus; (2) boreal forest, formed
by Abies and Picea; (3) semi-desert plants formed by
Amaranthaceae, Ephedra distachya, and E.fragilis types; (4)
heats and heathers of the family Ericaceae, including the species
Calluna; and (5) Central European steppe, composed mainly of
Artemisia, Cyperaceae, and Poaceae.
Zonation of the pollen diagram has been carried out by clus-
ters or hierarchical groupings, constrained by a matrix of
Euclidean distance between each sample (CONISS) (Grimm,
1987). This analysis was performed in the RStudio v. 1.2.5019
environment using the chclust program in the Rioja 0.9-21 pack-
age (Juggins, 2019).The number of significant pollen zones was
determined using optimal partitioning with minimal
sum-of-squares and broken-stick method using vegan v. 2.5-7
R-package (Oksanen et al., 2020).
Principal Components Analysis (PCA) was applied to the pol-
len percentages to reduce the dimensionality for detecting
climatic and environmental fluctuations. Of the 125 western
European taxa (excluding Pinus, spores, and unidentified pollen
grains), 72 taxa were retained to create the PCA without standard-
ization. However, a matrix reduction was applied to select only
major taxa with pollen percentages >6% in at least 5 samples
(Fig. 5). Prior to the analysis, a Hellinger transformation was
used (Legendre and Gallagher, 2001) to normalize the variance
of the different taxa and make it therefore more suitable for
Euclidean-based ordination methods, such as PCA. The
Hellinger transformation was made using the R-vegan package
(Oksanen et al., 2020). Then, the values according the dimension
scores were extracted for each sample. Another PCA taking into
account the pollen zones, was made to determine the environ-
mental significance of these pollen zones. The analysis was carried
out in the RStudio environment with three packages—factoMineR
v. 2.3 (Husson et al., 2017), factoextra v. 1.0.7 (Kassambara and
Mundt, 2020), and paleoMAS v. 2.0.1 (Correa-Metrio et al.,
2012)—to extract and visualize the results of the multivariate data.
Analysis of foraminifera assemblages and SST quantitative
reconstruction
Foraminifera assemblages of the MD04-2845 deep-sea core have
been published previously (Sánchez Goñi et al., 2008). Here, we
present data for the three new samples (levels 1164, 1191, and
1226 cm), which were dated by
14
C. The assemblages were ana-
lyzed in the >150 μm fraction of the same sample used for pollen
analysis. Between 362–382 foraminifera were counted in these
three samples. Quantitative values of seasonal and annual sea
surface temperatures (SST) from planktonic foraminifera assem-
blages were reconstructed. This reconstruction used a paleoeco-
logical reconstruction program developed at the EPOC
laboratory, based on modern analogues previously applied to
this core (Sánchez Goñi et al., 2013). This reconstruction relies
on an extended modern database using North Atlantic and
Mediterranean samples (1007 points).
Figure 2. Decay curves and dose-response of the sample BDX24931 using pIR-IR
290
signal. On the left side of the figure, the green lines delimit the background
noise, which is subtracted from the signal. The red lines on the right side of the figure are the graphical representation of how an equivalent dose (De) is calculated.
Table 2. Equivalent doses (D
e
) obtained with the Average Dose Model (ADM) and Central Age Model (CAM), from pIRIR
290
measurements. The overdispersion (OD)
values were determined with the Central Age Model. The age integrated in the age-depth model was determinate with the Central Model Age.
Sample De (Gy) CAM De (Gy) ADM De (Gy) Arithmetic average OD (%) CAM Age (ka) CAM Age (ka) ADM
BDX24931 (GdG18 OSL 1) 156 ± 3 156 ± 4 156 ± 13 8 ± 2 53.6 ± 3.4 53.6 ± 4.0
6 T. Fourcade et al.
https://doi.org/10.1017/qua.2022.21 Published online by Cambridge University Press
Middle and Early Upper Paleolithic lithic techno-complexes
We have created a database initially containing 32 sites and ∼300
previously published ages. To this database, we applied a series of
qualitative methodological, taphonomic, and sampling criteria,
from 0–3 following Guibert et al. (2008), to select the most rele-
vant ages (Excel file in Supplementary Information). The ages
with index 3 are the most reliable and were integrated in the
Bayesian models to reconstruct the temporal range of each LTC
in each archaeological site. A model for each site was carried
out with ChronoModel v. 2.0.18 (Lanos and Dufresne, 2019), tak-
ing into account the archaeostratigraphy between the different
dated levels and using the most recent IntCal20 calibration
curve (Reimer et al., 2020). As for the age model of core
MD04-2845, ChronoModel has the advantage of integrating mul-
tiple dated ages for the same event/sample and it is able to deal
statistically with the outliers. Then the events were organized in
phases (i.e., archaeological levels associated with a LTCs), only
constrained stratigraphically within each site. We have chosen
not to impose a chronological succession of these three LTCs,
because although they could be stratified in the same site, they
are not necessarily all contemporaneous from Aquitaine to
northern Spain. Ages of archaeological levels below and above
the targeted levels (i.e., Châtelperronian, Proto-Aurignacian, and
Early Aurignacian) served as boundaries. However, the Early
Aurignacian corpus needs further improvement, because only
the ages corresponding to the Early Aurignacian levels of the
sites delivering first Proto-Aurignacian and/or Châtelperronian
layers have been integrated. The prior information included in
the model is a stratigraphic order according to archaeological
sequence. The posterior distributions and the HPD regions at
68.2% and 95.5% were approximated for each collection events
and phases (Supplementary Information). We performed the
age modeling for thirteen sites. However, among them, two
sites provided only one age to include in the presentation and dis-
cussion of the results (Supplementary File 2).
RESULTS
Dating the Bay of Biscay core: results and improvements
The new
14
C dates range from 33.9 ± 0.2 to 34.1 ± 0.2
14
CyrBP
(Table 4). The time interval corresponding to HS 4 in several
North Atlantic cores (Elliot et al., 1998,2001,2002) is estimated
between 33.9 ± 0.7 and 34.9 ± 1.1
14
C yr BP, while farther south
on the Iberian margin it is dated between 33.7–34.7
14
CyrBP
(Naughton et al., 2009) (Supplementary Figure 2,
Supplementary Table 3). Chronological uncertainties for our
AMS
14
C dates range from 390–460 years (2σerror), and those
based on the marine isotopic events have uncertainties between
817–1287 years (Wolff et al., 2010; Sánchez Goñi et al., 2017).
The new
14
C ages appear to be statistically indistinguishable in
terms of uncertainties (Table 4, Supplementary Fig. 2). These
ages do not yield a consistent series of increasing age with increas-
ing depth, perhaps due to variations in the marine age reservoir.
Marine reservoir age simulations have highlighted variations in
global mean marine reservoir ages of several hundreds of years,
especially between ca. 42–38 ka close to the Laschamps (ca.
42.9–41.5 ka, Lascu et al., 2016) geomagnetic excursion (Butzin
et al., 2020; Heaton et al., 2020).
The pIR-IR
290
age obtained is 53.6 ± 3.4 ka (Table 2). This age,
if considered with its uncertainty at 1 sigma, is younger than the
other ages from other records for the HS 6, which is estimated
Table 3. Summary of water content, calculated equivalente dose, dose rates, paleodose, and age obtained. K, U, and Th contents were determined by high-resolution and low-background BEGe gamma spectrometry.
Sample
Depth
(cm)
Water content
(%)
Equivalente
dose Dose rate (Gy/a)
D
e
(Gy) CAM K (%)
U(
238
U)
(ppm)
U(
236
Ra)
(ppm) Th (ppm) alpha
Beta
(ext + int) gamma total
BDX24931 1535–1545 40 ± 10 156 ± 3 2.37 ± 0.04 2.96 ± 0.04 2.71 ± 0.15 10.35 ± 0.12 0.19 ±0.03 1.81 ± 0.04 0.95 ± 0.08 2.91 ± 0.05
Middle‐to‐Upper Paleolithic environmental changes, SW France 7
https://doi.org/10.1017/qua.2022.21 Published online by Cambridge University Press
between 64.0–60.0 ka based on the combination of
14
C and isoto-
pic event stratigraphy (Sánchez Goñi et al., 2013). It is also youn-
ger compared to the age given by the Villars Cave speleothem
(63.3–60.9 ± 0.8 ka; Genty et al., 2010). Moreover, GS 18, which
would be the counterpart of HS 6 in Greenland, is dated at ca.
63.8–59.4 ka (Rasmussen et al., 2014). However, in our record,
there is no evidence that the grains were not completely bleached.
HS 6 is a climatic phase encompassing a large input of IRD, which
may have resulted in reworking, but we need to further investigate
this hypothesis and increase the number of dated ages. This
exploratory approach of dating marine sediments by lumines-
cence and the preliminary pIR-IR
290
age seems promising,
which obviously needs to be confirmed with other luminescence
testing and the dating of younger HSs (HS 1–4) whose ages can be
compared with those obtained with
14
C in order to validate the
analytical strategy adopted.
The age-depth curve constructed with ArchaeoPhases, accord-
ing to the hierarchical Bayesian model in ChronoModel, gives
Figure 3. Depth-age model with Bayesian statistics using
ChronoModel software and ArchaeoPhases package.
Horizontal colored and thick lines represent the ages used
to create the model and sample (Table 1); labels are repre-
sented on the right. The other colored lines, which cut the
ages, are the calculated median age (pink) with two esti-
mated probabilities at 68% (between the two purple lines)
and 95% (green lines).
Figure 4. Pollen diagram of the MD04-2845 deep-sea core between 1335–1130 cm depth. From left to right: selected taxa, ecological groups (Atlantic forest, boreal
forest, semi-desert plants, heathlands), and pollen zones, based on the clustering analysis and optical partitioning.
8 T. Fourcade et al.
https://doi.org/10.1017/qua.2022.21 Published online by Cambridge University Press
information on sedimentation rate based on ages and uncertainties
(Fig. 3). Although the
14
C and IRSL ages are younger, all the ages
are slightly older in the depth-age model. This effect comes from
the a priori stratigraphic constraint, which adjusts the ages by
reducing uncertainties according to the stratigraphy of the sedi-
mentary sequence. The sedimentation rate of the core is relatively
constant from 70–40 ka, but it changes later on. The two dates
around 1200 cm correspond to HS 4, which is marked by a differ-
ent and rapid sedimentary process, such as an IRD deposition.
Vegetation and climate changes in SW France and its margin
Cluster analysis and broken-stick method applied to the pollen
assemblage identified five pollen zones (MD04-1 to MD04-5),
ranging from 1350–1130 cm deep (Fig. 4), ca. 49.8 ± 2.0 to 35.6
± 0.8 ka (Fig. 6).
Zones MD04-1 and MD04-5 are characterized by an increase
in the percentages of Atlantic forest taxa. The first zone (ca. 49.8–
47.3 ka) is marked by a higher increase in temperate forest (29%)
and, in particular, Betula (17%), Hippophae (3%), and deciduous
Quercus (5%). Zone MD04-5 (ca. 36.7–35.6 ka) suggests a
renewed forest expansion composed of Betula, Cupressaceae,
deciduous Quercus,Alnus,Corylus,Carpinus, and Fagus, reflect-
ing a warmer and wetter climate. Zone MD04-2 (ca. 47.3–40.9
ka) is characterized by steppe expansion, consisting of
Artemisia, Amaranthaceae, Cyperaceae, and Poaceae, but also
by forest expansion, as shown by increases of Betula, deciduous
Quercus,Alnus,Corylus, and Carpinus. The boreal forest is
represented in zones MD04-3 (ca. 40.9–38.3 ka), which is marked
by the highest percentage of Abies in the entire sequence (4%),
reflecting a cooling. This zone is also marked by a significant
increase in humidity and a cold climate, as reflected by percent-
ages of Calluna (e.g., 22% at ca. 38.7 ± 0.8 ka, 1209 cm; 23% at
ca. 39.4 ± 0.9 ka, 1220 cm). On the contrary, MD04-4 (ca. 38.3–
36.7 ka) is characterized by an increase in steppe taxa (32% at
ca. 36.9 ± 0.9 ka, 1170 cm), especially Artemisia (19%), Poaceae
(30%), and Cyperaceae (23%), along with the presence of boreal
forest (2% of Picea), indicating a colder and drier climate. The
beginning of the zone is marked by an increase in the percentages
of Calluna species (38% at ca. 37.7 ± 0.8/0.81 ka, 1190 cm) reflect-
ing a cold and humid climate. In this zone, E.fragilis is virtually
absent. Within this zone, the percentages of Pinus decrease
slightly around 38.2 ± 0.8 ka (1200 cm).
The results of the PCA explain 45% of the variance. The PCA
identifies a first component (Fig. 5), that explains 25.9% of the var-
iance characterized by deciduous Quercus and Betula with positive
scores (warm), and herbaceous and shrub taxa such as Ericaceaea,
Calluna, and Cyperaceae (cold) with negative scores. The second
dimension, which explains 19% of the variance, is characterized
by the positive scores of forest taxa, such as deciduous Quercus
and Betula, and Ericaceae, which is a moist-loving taxon, while
Artemisia, Poaceae, and Cyperaceae fall in the negative scores. In
other words, variations of PC1 scores are used in this study as a
warm/cold index and PC2 variations are used as a dry/wet and
oceanic/continental proxy (Fig. 6E). Therefore, zones MD04-1
and MD04-5 represent relatively warm and wet climate, zones
Table 4.
14
C dates of the three new samples and the other
14
C ages used in this study and their calibration with Calib v.8.2.
ID_laboratory
Depth
(cm) Material pMC δ
13
C(‰) AMS
14
C age (
14
C ka BP)
Error
(
14
C ka BP)
Age Calib
(2σ–cal BP)
SacA-002977 1048 G.bulloides 2.47 ± 0.12 −1.7 29.87 0.39 32.389–30.493
SacA-002978 1060 N.pachyderma (s) 2.19 ± 0.12 2.6 30.67 0.42 33.253–31.518
SacA-002976 1078 G.bulloides 2.02 ± 0.12 0.6 31.35 0.46 33.985–32.182
Beta - 491854 1164 N.pachyderma (s) 1.47 ± 0.04 0.3 33.92 0.22 36.173–34.731
Beta - 491855 1191 N.pachyderma (s) 1.57 ± 0.04 0.2 33.39 0.22 35.220–34.215
Beta - 491856 1226 N.pachyderma (s) 1.43 ± 0.04 0.5 34.14 0.23 36.557–34.982
Figure 5. Principal Components Analysis (PCA) representing (a) pollen zones-based confidence ellipses, and (b) the 9 major taxa, with pollen percentages esti-
mated to be >6% for at least 5 samples.
Middle‐to‐Upper Paleolithic environmental changes, SW France 9
https://doi.org/10.1017/qua.2022.21 Published online by Cambridge University Press
Figure 6. Oceanic and continental climatic multiproxies in the core, discussed from the bottom to the top of the figure. (F, G) Pollen percentages of Atlantic forest
(degraded green), boreal forest (green), semi-desert plants (orange), Calluna (blue); obliquity curve (Laskar et al., 2004) is represented with the temperate and
boreal forest (G). (E) PC1 (purple) and PC2 (red) scores are represented in two different forms: purple/red bar charts and purple/red dotted curves. Dim1
(=PC1): the negative values correspond to dryness and positive to cold environments and Dim2 (=PC2) with positive scores to warm and negative scores to
humid taxa (D), the annual sea surface temperature of Bay of Biscay (SST) derived from the polar foraminiferan Neogloboquadrina pachyderma (s) percentages,
and quantity of IRD (C). The percentage scale of N.pachyderma (s) is reversed with respect to IRD. Villars stalagmites δ
13
C(B) Vil27 (gray) and Vil9 (red) and both of
their chronological uncertainties are represented (only some uncertainties are shown; all uncertainties are available in Genty et al., 2003). (A) NorthGRIP δ
18
O curve
(Rasmussen et al., 2014) and the uncertainties for the GIs (Wolff et al., 2010). Dotted rectangle (C–G) represents HS 4; dashed line indicates the separation between
the two climatic phases within HS 4.
10 T. Fourcade et al.
https://doi.org/10.1017/qua.2022.21 Published online by Cambridge University Press
MD04-3 and MD04-4 colder and drier than the previous ones,
and MD04-2 represents a relatively cold and wet climate. The
warm and humid terrestrial phases are associated with the highest
SST in the Bay of Biscay (Fig. 6D). The cold and dry terrestrial
phases (Fig. 6F) are synchronous with cold SST, while the cold
and humid phase is associated with SST oscillations.
Clustering analysis recognizes 14 pollen zones. However, opti-
mal partitioning that gives the statistical significance of the zones
has only detected 5 pollen zones and, therefore, D-O variability is
not recorded from pollen percentages. However, in pollen zone
MD04-2, two small increases in forest percentages are associated
with two large SST increases that correspond with D-O 11 and
D-O 10. Therefore, this variability, which was undetected by the
optimal partitioning, seems to be real.
The chronology of the Middle to Upper Paleolithic
technocomplexes
The ensemble of age models performed for each of the 11 sites in
SW Europe (Fig. 7D, SI) show that the Châtelperronian, repre-
sented in this study by five sites mostly from the Aquitaine
basin (i.e., Bordes-Fitte, Les Cottés, La Quina aval, and la
Ferrassie) and Labeko-Koba (Wood et al., 2014) in the Spanish
Basque country (Fig. 7D) spans from ca. 44.5 ka to ca. 40.1 ka.
The Châtelperronian at la Quina aval and La Ferrassie could be
the oldest (ca. 44 ka). The youngest ending time of this LTC
would be, taking all uncertainties into account, ca. 40. 3 ka at
Les Cottés and ca. 40.1 ka at La Ferrassie.
The Proto-Aurignacian, the first LTC attributed to the Upper
Paleolithic and Homo sapiens, is found at Les Cottés, Isturitz
(Barshay-Szmidt et al., 2018) and Gatzarria (Barshay-Szmidt
et al., 2012) in the French Basque country, and at Labeko-Koba,
Covalejos, and Cobrante in the Cantabrian of Spain
(Marín-Arroyo et al., 2018). It could have developed between ca.
44.2–41.4 ka at Isturitz and between 42.3–40.4 ka at Gatzarria. In
northern Spain, it appeared in Covalejos between ca. 41.9–40.0
ka, in Labeko-Koba ca. 41.3–40.8 ka, ca. 41.3–39.1 ka in
Cobrante, and in El Cuco between ca. 41.1–39.0 ka. In SE
France, the Proto-Aurignacian, which is represented in our
model only by Esquicho-Grapaou (Barshay-Szmidt et al., 2020),
spans between ca. 42.9–38.0 ka. The youngest Proto-Aurignacian
starting time is found at the Les Cottés site, which is dated between
ca. 40.6–39.6 ka. The Proto-Aurignacian would be present in
Western Europe first in northern Spain, then in SE France and
the Aquitaine basin. However, this hypothesis is based on only a
few well-dated sites from SW Europe.
The Proto-Aurignacian chronologically overlaps with the
Châtelperronian. In SW France, taking into account the oldest
and the most recent age from the Les Cottés and La Ferrassie
sites, the overlap spans ca. 500 years. However, according to the
techno-cultural attribution at the Bordes-Fitte site, which has
the oldest presence of an Aurignacian occupation, this overlap
would be ca. 1,100 years. Farther south, in the Basque country
and Cantabrian region, the Châtelperronian overlaps with the
Proto-Aurignacian for ca. 1,800 years, if we accept that the chro-
nology of Isturitz is sufficiently reliable.
The Early Aurignacian is stratigraphically above the
Proto-Aurignacian. It is represented by six sites located in SW
France, the French and Spanish Basque country, and northern
Spain. The Early Aurignacian is the oldest north of the Aquitaine
basin at the Bordes-Fitte site, 41.2–40.1 ka, and in the Basque coun-
try at Isturitz, ca. 40.8–40.0 ka. It could have developed at
Labeko-Koba between ca. 40.5–39.7 ka, and at Covalejos ca.
40.3–39.0 ka. In southwestern France, it appeared at Gatzarria
between ca. 39.9–38.3 ka and ca. 39.4–38.8 ka at Les Cottés.
DISCUSSION
Climatic and environmental changes in SW France from GI
12–8 (ca. 50–36 ka)
The phases marked by the increase of Atlantic forest (Fig. 6G)are
associated with the SST warming in the Bay of Biscay (Fig. 6C).
Conversely, the phases dominated by semi-desert plants
(Fig. 6F) are synchronous with cold SST. The chronologies of
the four temperate phases punctuating the period between HS 5
and HS 4, does not correspond with the chronologies of GI 12–
8, defined according to the Greenland δ
18
O record (Fig. 6A)
(Rasmussen et al., 2014). These SW European warming events
show a difference of ca. 1,000 years from our age-depth model,
compared to the onset of these GIs. This difference falls, however,
within the uncertainties of the Greenland age model (800–1,300
years, Table 1) and the uncertainties of radiocarbon ages of this
period. Atlantic forest pollen percentages indicate a
progressive long-term decrease in the forest cover from GI
12–8, paralleling the decrease in obliquity (Fig. 6G), suggesting
a warmer GI 12 (28%), compared to the other GIs. The high
values of δ
13
C from two stalagmites in the Villars Cave
(Genty et al., 2010) further indicate an increase in precipitation
during GI 12 (Fig. 6A, B).
During HS 4, the SST was strongly imprinted by N.pachy-
derma (s) percentages, which decreased a few centuries before
the increase of IRD (Fig. 6C). SST in the Bay of Biscay probably
cooled contemporaneously with the first IRD discharges in the
more northwestern regions. Therefore, and following the previous
work of Sánchez Goñi et al. (2000), we define the time period of
HS 4 in the Bay of Biscay between ca. 40.2–36.5 ka by using col-
lectively the decrease of SST and the increases of N.pachyderma
(s) and IRD. The age of HS 4, which is estimated between 40.2–
38.3 cal BP (Sanchez Goñi and Harrison, 2010), is based on the
synthesis of
14
C ages from North Atlantic deep-sea cores made
by Elliot et al. (2002). The timing of the end of HS 4 given by
the Bayesian age-depth model in core MD04-2845 using the
new
14
C age is ca. 1,500 years younger than that based on the
North Atlantic
14
C. However, taking into account all the uncer-
tainties, our new age-model does not fundamentally contradict
the traditional chronology of HS 4.
In SW France, HS 4 is composed of two climatic phases
(Fig. 6). This subdivision into two phases was detected previ-
ously in a core from the northwestern Iberian margin
(MD99-2331, 42°9’N, 9°69’W; Naughton et al., 2009). Our
first phase (ca. 40.2–37.5 ka) is marked by the strongest iceberg
discharges and is contemporaneouswiththedecreaseofthe
Atlantic forest and maximum percentages of Calluna.This
genus is a moisture and light-demanding plant whose develop-
ment is favored by forest contraction (Naughton et al., 2009),
reflecting an increase in humidity. Sea surface temperatures
drop by ∼7 ± 5°C. A slowdown in the growth of the Villars
Cave speleothems is recorded during HS 4, also indicating a
drastic decrease in precipitation in this region (Genty et al.,
2003). Our second phase of HS 4 (ca. 37.5–36.4 ka) is synchro-
nous with a moderate amount of IRD compared to the first
phase, but the maximum of N.pachyderma (s) maintained low
SSTs, between 0–2±2°C in the Bayof Biscay.
Middle‐to‐Upper Paleolithic environmental changes, SW France 11
https://doi.org/10.1017/qua.2022.21 Published online by Cambridge University Press
Interestingly, within the first phase, SST and N.pachyderma (s)
indicate a slight warming of 2 ± 3°C towards 38.2 ± 0.9 ka, which
is associated with a small decrease in IRD. The percentages of
N.pachyderma (s) fall to a value of 86.4% and SST increases by
∼4 ± 3°C. Another deep-sea core off the Iberian Peninsula
(MD95-2039, 40°34’N, 10°20’W) shows a slight warming associ-
ated with an increase in deciduous Quercus (Roucoux et al.,
2005), which could indicate regional warming in the middle of
HS 4 between 40–45°N. In the Bay of Biscay, the percentage of
deciduous Quercus (1%) remains low, but the Atlantic forest
reaches almost 9% due to a 3% increase in Alnus,2%in
Corylus, and 1% in Betula and Carpinus.
HS 4 is thus divided into two different climatic and environ-
mental phases in the eastern North Atlantic region between 40°
N and 45°N: a first phase, associated with the maximum amount
of IRD and marked by extreme cooling and wet conditions; and a
second phase, characterized by a drier and colder climate showing
a warming trend. However, unlike the cores from the Iberian mar-
gin, where the first phase is considered to be the coldest one, the
two phases in the Bay of Biscay are relatively similar in terms of
forest cover and oceanic temperatures throughout HS 4.
Climatic and environmental changes: triggers for
technological adaptations?
The studies of SW France archaeological sites played a major role
in the definition of both Middle and Upper Paleolithic cultures.
The oldest appearance of the Châtelperronian at la Quina aval
and La Ferrassie (ca. 44 ka) could be contemporaneous with GS
11 (Fig. 7). The presence of Châtelperronian at La Ferrassie
would be the earliest presence in SW France, well before the
Châtelperronian from Arcy-sur-Cure (Talamo et al., 2020).
Farther north, at Bordes-Fitte rockshelter and Les Cottés Cave,
it might have begun between 42.9–41.5 ka, corresponding perhaps
Figure 7. Comparison between environmental/climatic data and archaeological LTCs. (A) Bay of Biscay SST. (B, C) Pollen percentages of Atlantic forest ( green),
boreal forest (dark green), semi-desert steppe (orange), and Calluna species. (D) Representation of Middle to Upper Paleolithic transition LTCs. Modeled time ranges
at 68.2% for Châtelperronian (green), Proto-Aurignacian (dark orange), and Early Aurignacian (yellow). The gray horizontal rectangle represents the only site of SE
France. Dotted rectangle represents HS 4; dashed line indicates the separation between the two climatic phases within HS 4.
12 T. Fourcade et al.
https://doi.org/10.1017/qua.2022.21 Published online by Cambridge University Press
to the end of GS 11. At Labeko-Koba, the most southern site, the
Châtelperronian is dated between 43.0–41.6 ka and encompasses
the end of GS 11. Depending on the region, the Châtelperronian
developed since the end of GS 11 or during GS 10. However, in
SW France and northern Spain, the Châtelperronian does not
appear beyond the beginning of HS 4. The last age of the
Châtelperronian, which is associated with the last Neanderthals,
indicates that their disappearance would have occurred in this
region at the same time as the start of HS4.
The Proto-Aurignacian could have appeared in SW and SE
France, as well as in northern Spain, around the GS 10/GS 9 tran-
sition, which is marked by the expansion of an open forest with
steppe-like elements. Shao et al. (2021), using a global climate
model, developed a human-existence model combining climate
data with archaeological sites to reconstruct patterns of
Aurignacian dispersal. The earliest Aurignacian dispersal in
Europe started before 45 ka (Hublin et al., 2020). The recent dis-
covery in Mandrin Cave (Rhone valley) of a tooth that belonged
to H.sapiens, has highlighted its presence in Western Europe
before 50 ka (Slimak et al., 2022).
In Proto-Aurignacian levels, reindeer dominates the faunal
assemblages in Charentes and Périgord, while horses and
Bovidae are dominant in the Pyrenees (Discamps et al., 2010;
Soressi et al., 2010; Barshay-Szmidt et al., 2018). Around GS 9,
steppe and, to a lesser extent, boreal forest could explain the
strong presence of reindeer; bison, and horses in both regions.
Progressive reduction of forest cover from GS 9 to HS 4 probably
also had an effect on modern human migration (Badino et al.,
2020). So far, our age modeling suggests that the appearance of
Homo sapiens would have occurred first in SW France, with
migration later in SE France, arriving in northern Spain at ca.
41.8 ka. However, Talamo et al. (2020), using another age-
modeling approach (IntCal 13, Reimer et al., 2013, and OxCal,
Ramsey, 2009) obtained an age for Isturitz that is younger than
our model, and, in this case, Homo sapiens would have arrived
in SE France before before farther west. Their scenario is consis-
tent with the recent dating of the Proto-Aurignacian at Mandrin
Cave, the beginning of which is dated between 43.3–42.2 ka
(Slimak et al., 2022).
The first occurrences of the Early Aurignacian happened
during GS 9 and the onset of HS 4 in Gatzarria and Les Cottés.
The Early Aurignacian persisted during the beginning of HS
4. The faunal record also indicates a dominance of open and
cold environmental faunal species in the Early Aurignacian
(Discamps et al., 2010). Interestingly, within this LTC, a geograph-
ical difference also seems to emerge between the north and south,
which are dominated by reindeer and by horses and Bovidae,
respectively (Barshay-Szmidt et al., 2018). Unfortunately, no
pollen data exists documenting a difference in the vegetation
between the north and the south of SW France and northern
Spain. The transition between Proto-Aurignacian and Early
Aurignacian would overlap GS 9 and the first phase of HS 4.
The overlap between Châtelperronian and Proto-Aurignacian
could be ca. 3,300 years between SW France and northern
Spain. This overlap suggests a coexistence of several millennia
between Neanderthal and modern groups (Marín-Arroyo et al.,
2018), and interbreeding between Neanderthals and ancestors of
non-African modern humans (Green et al., 2010). In recent
years, genetic studies show the coexistence of Neanderthals and
Homo sapiens by admixture from Neanderthals into ancestors
of present populations in several regions of the world (Green
et al., 2010; Prüfer et al., 2014,2021; Fu et al., 2015; Bokelmann
et al., 2019; Bergström et al., 2020) either through a single
(Sankararaman et al., 2012; Bergström et al., 2020) or multiple
episodes of gene flow (Prüfer et al., 2014; Vernot and Akey,
2015; Villanea and Schraiber, 2019; Hublin et al., 2020).
Western Europe experienced strong climate changes, which
affected the environments and the food resources of
Neanderthals and modern humans. So far, the archaeological
records show that the transition between each of the LTCs
between 44–36 ka encompassed several warming and cooling
events, and that the same LTC is not synchronous throughout
SW France and northern Spain. However, the late Neanderthal
LTC seems to have developed in a moderately forested landscape,
while modern humans developed in successively more open
environments.
Comparison of both paleoclimatic and archaeological records
aiming to detect potential synchronies is a complex process due
to the age resolution and typo-technological definitions.
Therefore, the chronology of the Middle-Upper Paleolithic transi-
tion is still not conclusive and needs to be improved by enlarging
the database and pursuing the chronological modeling approach
using Bayesian statistics.
Our study further suggests that the disappearance of
Neanderthals does not seem to be directly related to climate-
driven environmental changes, although the Châtelperronian
ended in several sites before HS 4 onset. However, uncertainty
in determining the age of HS 4 and the apparent young age of
this event in this marine record reveal that there is an uncertainty
making it difficult to assume a relationship between the end of the
Châtelperronian and HS 4 onset. In SW France and the
Cantabrian regions, the late Neanderthals survived several warm-
ing and cooling events, but just disappeared after the first LTCs
associated with modern humans are recorded. Furthermore, the
progressive increase of open environments from ca. 50–40 ka
would have been favorable to expansion of modern humans,
who could have been well adapted to the steppe. They could
have competed with the late Neanderthals for the same ecological
niches, causing their regional disappearance. Our new data are in
line with previous modeling studies showing that Neanderthal
disappearance can only be achieved when modern humans are
chosen in the model as being more adapted in the exploitation
of food resources and hunting technology compared to
Neanderthals (Sepulchre et al., 2007; Banks et al., 2008;
Columbu et al., 2020; Timmermann, 2020).
CONCLUSIONS
The high resolution pollen study of the MD04-2845 deep-sea core
retrieved from the Bay of Biscay precisely identified the effect of
climate changes from D-O 12–8 cycles and HS 4 in SW France.
HS 4, defined on the basis of increases in IRD and N.pachyderma
(s), is characterized by two distinct climatic phases: a first wet and
cold phase and a second drier and colder phase. In the long term,
a progressive decrease in the Atlantic forest and concomitant
expansion of open environments is recorded from 50–40 ka.
From a chronological point of view, IRSL dating results on the
MD04-2845 deep-sea core seems promising, and future sample
dating should confirm its application for sedimentary sequences
older than the limits of radiocarbon dating. In addition,
Bayesian statistics for both paleoclimatic and archaeological data
from the westernmost part of Europe allowed us to improve the
identification of potential synchronies. This critical work on an
archaeological database and chronological modeling could be
Middle‐to‐Upper Paleolithic environmental changes, SW France 13
https://doi.org/10.1017/qua.2022.21 Published online by Cambridge University Press
applied to Mousterian (older) LTCs to better characterize their
variability over the Middle Paleolithic.
This comparison shows that changes in LTCs during the
Middle to Upper Paleolithic transition do not correspond with
punctual vegetation and climate changes. In contrast, progressive
opening of the regional landscape seems to have provided the
context for the replacement of Neanderthals by modern humans,
which lasted several millennia during which potential interbreed-
ing and cultural changes occurred. Finally, our data suggest that
climate changes did not directly cause the disappearance of
Neanderthal. That disappearance was probably the result of com-
petition with Homo sapiens for the same ecological niches.
Supplementary Material. The supplementary material for this article can
be found at https://doi.org/10.1017/qua.2022.21
Acknowledgments. This work was initially funded by the New Aquitaine
Region NATCH scientific project (Neanderthalenses aquitanensis:
Territoires, Chronologie, Humanité, co-dir. J.Ph. Faivre, C. Lahaye,
B. Maureille), and continued during a doctoral contract n°12-18 (Ecole
Doctorale Montaigne Humanités), and by financial support from the French
Research National Agency under the Investissements d’Avenir Program
(ANR-10-LABX-52).
14
C dating was carried out with the support of ERC
Advanced Grant TRACSYMBOLS no. 249587. We thank the members of
UMR 5805 EPOC and UMR 6034 Archeosciences Bordeaux for their technical
assistance (L. Devaux, M. Georget, M.-H. Castera, O. Thier, and I. Billy;
J. Faure). Thanks to D. Genty and S. Salonen for their help and interesting dis-
cussions. We thank the three anonymous referees for their insightful com-
ments to revise this paper.
REFERENCES
Aitken, M.J., 1998. An Introduction to Optical Dating: The Dating of
Quaternary Sediments by the Use of Photon-stimulated Luminescence.
Oxford University Press, Oxford, New York.
Allen, G., Castaing, P., 1977. Carte de répartition des sédiments superficiels
sur le plateau continental du Golfe de Gascogne. Bulletin de l’Institut de
Géologie du Bassin d’Aquitaine 23, 255–260.
Armitage, S.J., Pinder, R.C., 2017. Testing the applicability of optically stim-
ulated luminescence dating to Ocean Drilling Program cores. Quaternary
Geochronology 39, 124–130.
Badino, F., Pini, R., Ravazzi, C., Margaritora, D., Arrighi, S., Bortolini, E.,
Figus, et al., 2020. An overview of Alpine and Mediterranean paleogeography,
terrestrial ecosystems and climate history during MIS 3 with focus on the
Middle to Upper Palaeolithic transition. Quaternary International 551,7–28.
Banks, W.E., d’Errico, F., Peterson, A.T., Kageyama, M., Sima, A.,
Sánchez-Goñi, M.-F., 2008. Neanderthal extinction by competitive exclu-
sion. PLoS ONE 3, e3972. https://doi.org/10.1371/journal.pone.0003972.
Banks, W.E., 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. Journal of
Human Evolution 64,39–55.
Bard, E., Arnold, M., Mangerud, J., Paterne, M., Labeyrie, L., Duprat, J.,
Mélières, M.-A., Sønstegaard, E., Duplessy, J.-C., 1994. The North
Atlantic atmosphere-sea surface
14
C gradient during the Younger Dryas cli-
matic event. Earth and Planetary Science Letters 126, 275–287.
Barshay-Szmidt, C.C., Eizenberg, L., Deschamps, M., 2012. Radiocarbon
(AMS) dating the Classic Aurignacian, Proto-Aurignacian and Vasconian
Mousterian at Gatzarria Cave (Pyrénées-Atlantiques, France). PALEO
Revue d’Archéologie Préhistorique 23,11–38.
Barshay-Szmidt, C., Normand, C., Flas, D., Soulier, M.-C., 2018.
Radiocarbon dating the Aurignacian sequence at Isturitz (France):
Implications for the timing and development of the Protoaurignacian and
Early Aurignacian in western Europe. Journal of Archaeological Science:
Reports 17, 809–838.
Barshay-Szmidt, C., Bazile, F., Brugal, J.-P., 2020. First AMS
14
C dates on the
Protoaurignacian in Mediterranean France: The site of Esquicho-Grapaou
(Russan-Ste-Anastasie, Gard). Journal of Archaeological Science: Reports
33, 102474. https://doi.org/10.1016/j.jasrep.2017.09.003.
Bartolomei, G., Broglio, A., Cassoli, P.F., Castelletti, L., Cattani, L.,
Cremaschi, M., Giacobini, G., et al., 1994. La Grotte de Fumane. Un
site aurignacien au pied des Alpes. Preistoria Alpina 28, 131–179.
Bateman, M.D., Swift, D.A., Piotrowski, J.A., Rhodes, E.J., Damsgaard, A.,
2018. Can glacial shearing of sediment reset the signal used for lumines-
cence dating? Geomorphology 306,90–101.
Bateman, M.D., Swift, D.A., Piotrowski, J.A., Sanderson, D.C.W., 2012.
Investigating the effects of glacial shearing of sediment on luminescence.
Quaternary Geochronology 10, 230–236.
Bazile, F., 1977. Nouvelles données sur le Paléolithique Supérieur ancien en
Languedoc Oriental. 20th Congrés Préhistorique de France, Provence,
1974, Châteauneuf-les-Martigues. SPF, Paris, pp. 24–28.
Berger, J.-F., Guilaine, J., 2009. The 8200 cal BP abrupt environmental
change and the Neolithic transition: a Mediterranean perspective.
Quaternary International 200,31–49.
Bergström, A., McCarthy, S.A., Hui, R., Almarri, M.A., Ayub, Q., Danecek,
P., Chen, Y., et al., 2020. Insights into human genetic variation and popu-
lation history from 929 diverse genomes. Science 367, eaay5012. https://doi.
org/10.1126/science.aay5012.
Blaauw, M., Christen, J.A., 2011. Flexible paleoclimate age-depth models
using an autoregressive gamma process. Bayesian Analysis 6, 457–474.
Bokelmann, L., Hajdinjak, M., Peyrégne, S., Brace, S., Essel, E., Filippo, C.
de, Glocke, I., et al., 2019. A genetic analysis of the Gibraltar Neanderthals.
PNAS 116, 15610–15615.
Bon, F., 2002. A brief overview of Aurignacian cultures in the context of the
industries of the transition from the Middle to the Upper Paleolithic. In:
Bar-Yosef, O., Zilhao, J. (Eds.), Towards a Definition of the Aurignacian.
Trabalhos de Arqueologia 45, Lisboa, American School of Prehistoric
Research/Instituto Português de Arqueologia, pp. 133–144.
Bond, G., Lotti, R., 1995. Iceberg discharges into the North Atlantic on mil-
lennial timescales during the last glaciation. Science 267, 1005–1010.
Borrell, F., Junno, A., Barceló, J.A., 2015. Synchronous environmental and
cultural change in the emergence of agricultural economies 10,000 years
ago in the Levant. PLoS ONE 10, e0134810. https://doi.org/10.1371/jour-
nal.pone.0134810.
Brocheray, S., 2015. Transferts et Accumulations sur les Marges du Golfe de
Gascogne: Architecture, Fonctionnement et Contrôles. These de doctorat,
Océanographie, Université de Bordeaux, France.
Brocheray, S., Cremer, M., Zaragosi, S., Schmidt, S., Eynaud, F., Rossignol,
L., Gillet, H., 2014. 2000 years of frequent turbidite activity in the
Capbreton Canyon (Bay of Biscay). Marine Geology 347, 136–152.
Broecker, W.S., Olson, E.A., 1961. Lamont radiocarbon measurements VIII.
Radiocarbon 3, 176–204.
Butzin, M., Heaton, T.J., Köhler, P., Lohmann, G., 2020. A short note on
marine reservoir age simulations used in IntCal20. Radiocarbon 62, 865–
871.
Buylaert, J.-P., Jain, M., Murray, A.S., Thomsen, K.J., Thiel, C., Sohbati, R.,
2012. A robust feldspar luminescence dating method for Middle and Late
Pleistocene sediments. Boreas 41, 435–451.
Castaing, P., 1981. Le Transfert à l’Océan des Suspensions Estuariennes: Cas de
la Gironde. Institut de géologie du Bassin d’Aquitaine, Bordeaux 1,
Bordeaux, 530 pp.
Castaing, P., Froidefond, J.-M., Lazure, P., Weber, O., Prud’homme, R.,
Jouanneau, J.-M., 1999. Relationship between hydrology and seasonal dis-
tribution of suspended sediments on the continental shelf of the Bay of
Biscay. Deep-Sea Research II 46, 1979–2001.
Colas, F., 2003. Circulation et Dispersion Lagrangiennes en Atlantique
Nord-est. Ph.D. dissertation, Université de Bretagne Occidentale, Brest,
253 pp.
Columbu, A., Chiarini, V., Spötl, C., Benazzi, S., Hellstrom, J., Cheng, H.,
De Waele, J., 2020. Speleothem record attests to stable environmental con-
ditions during Neanderthal–modern human turnover in southern Italy.
Nature Ecology & Evolution 4, 1188–1195.
14 T. Fourcade et al.
https://doi.org/10.1017/qua.2022.21 Published online by Cambridge University Press
Correa-Metrio, A., Urrego, D.H., Cabrera, K.R., Bush, M.B., 2012.
paleoMAS. Paleoecological Analysis, R package version 2.0-1. Vienna, The
R Project for Statistical Computing, available at <http://CRAN.R-project.
org/package=paleoMAS>
Daniau, A.-L., Sánchez Goñi, M.F., Duprat, J., 2009. Last glacial fire regime
variability in western France inferred from microcharcoal preserved in core
MD04-2845, Bay of Biscay. Quaternary Research 71, 385–396.
Dansgaard, W., Johnsen, S., Clausen, H., Dahljensen, D., Gundestrup, N.,
Hammer, C., Oeschger, H., 1984. Evidence for general instability of past
climate from a 250-kyr ice-core record. In: Hansen, J.E., Takahashi, T.
(Eds.), Climate Processes and Climate Sensitivity. American Geophysical