This study was initiated and supported by the LITAQ project (« Du Pléistocène à l’Anthropocène : connaître les mécanismes passés d’évolution des populations -végétales, animales, humaines- et des milieux pour prédire les réponses futures. L’exemple du LITtoral AQuitain », inter-Cluster of Excellence COTE and LaScArBx - ANR-10-LABX-45), funded by the Agence Nationale de la Recherche under the Programme Investissement d’Avenir (reference ANR-10-IDEX-03-02). We thank all the colleagues who contributed to the LITAQ project (see the team at http://www.geocean.net/wikilitaq/doku.php?id=start), Ludovic Deveaux for the pollen preparation at EPOC, and Peter Steier (from the VERA Labor, Universität Wien, Austria) for his patience and great reactivity with the LITAQ ¹⁴C AMS dating. We are grateful to Olivier Moine (editor of Quaternaire in charge of this special issue), to Agnès Baltzer and to an anonymous reviewer, for their constructive remarks on the manuscript. We thank Christopher Sutcliffe for his careful review of the language.

1 - Introduction

1The Atlantic coast of south-west France (Aquitaine) has a typical linear sandy shoreline shaped by the highly dynamic neritic environment of the southern Bay of Biscay (Durrieu de Madron et al., 1999; Herbert et al., 2011; Charria et al., 2013; Ba & Senechal, 2013; Castelle et al., 2015). This open sandy coast is interrupted only by the mouths of a few major estuaries (i.e. Gironde, Adour) and lagoons (e.g. Arcachon). Clay relict deposits observed beside these continental distributaries are dated from the Pleistocene and Holocene; they attest to highly divergent past sedimentary contexts before the establishment of the present-day sand dune system (e.g. Allen et al., 1974). Such typical clay formations can be found beneath the modern dunes on the northern beaches of the Médoc peninsula, south of Soulac-sur-Mer (fig. 1). They are related to the palaeo-mouth of the Gironde, which is divided into several channels before merging and stabilizing next to the Tertiary limestones of the northern, right bank of the estuary (Fabre, 1939; Buffault, 1942; Dubreuihl, 1971; Allen et al., 1974; Howa, 1985, 1987; Legigan et al., 1986; Diot & Tastet, 1995; O’Brien & Jones, 2003). A high concentration of archaeological remains is recorded in those formations in the vicinity of Soulac-sur-Mer. They attest to dense human occupation extending back in time to at least the Mesolithic (Buffault, 1897; Welsch, 1917; Caralp & Vigneaux, 1960; Roussot-Larroque, 1979; Frugier et al., 1982; Verdin et al., 2019). Accordingly this zone has been the subject of numerous studies since the late nineteenth century mainly aimed at establishing the stratigraphy of the pre- and proto-historic human settlements. The studies have often involved sporadic archaeological rescues and discoveries under the constraints of permanent offensive marine erosion.

2As part of the InterLabEx COTE / LASCARBX 2013–2015 LITAQ project, the area has been re-investigated with a particular focus on the Lède du Gurp (LDG, fig. 1) archaeological site (Roussot-Larroque & Villes, 1988; Roussot-Larroque, 2007, and references therein). The site has frontal dune outcrops at the upper limit of the intertidal beach, with stratified accumulations of sandy clays and peats covered by modern aeolian sands (Roussot-Larroque & Villes, 1988; Tastet, 1999). After the year 2000, studies of the site were halted and it was left to the sea, which has since destroyed and wasted large amounts of material. The 2014 winter storms (Castelle et al., 2015), which coincided with the beginning of the LITAQ field operations, isolated the archaeological site on the foreshore (see figs. 1 & 2 in Verdin et al., 2019). This permanent shore erosion modifies the LDG site year after year opening up transient, fortuitous but precious observational windows in the modern dune.

3In order to retrace the palaeoenvironmental and palaeogeographic evolution of this coastal ecosystem during the Holocene, new samples of continuous and undisturbed sections measuring several metres have been collected from on and around the LDG site (tab. 1, fig. 1). These new sedimentological records have been studied for their lithostratigraphy (using classic and X-ray imaging coupled with spectrocolorimetry, X-ray fluorescence and grain-size analyses together with radiocarbon dating) and integrated with existing archaeological data, including new data produced as part of the LITAQ project and its related excavation in 2014 (Verdin et al., 2019).

4Here we present results obtained for three of the four LDG sequences (G2, 3 and 4, figs. 1 & 2) in order to define a comprehensive framework of depositional history that is consistent with the well-known Flandrian (last deglaciation) transgression.

2 - Study site

5The LDG study site is located on shore in the northern part of the Médoc peninsula. It is bounded to the north-east by the Gironde estuary and to the west by the Atlantic Ocean (fig. 1). More specifically, the site forms a natural boundary between the localities of Soulac-sur-Mer and Grayan-et-l’Hôpital (fig. 1A). This area is geomorphologically identified as a cape. The change in the direction of the coastline induces a divergence in the longshore drift (one branch flooding northward and the other southward) which explains the high rates of erosion along this portion (mean value of 10 m/year, e.g. Howa, 1987; Aubié & Tastet, 2000; Castelle et al., 2015, see also Stéphan et al., 2019).

2.1 - Geological setting

6The underlying geological structures are poorly expressed in the Médoc landscape. Only low reliefs are observed and they are related to Quaternary alluvial and wind-blown cover, i.e. terraces in the inner estuary border and dunes along the coast, which mask the Tertiary substratum and its tectonic heritage (Fabre, 1939; Caralp & Vigneaux, 1960; Dubreuilh & Marionnaud, 1972; Bosq et al., 2019). The region is characterized by a monoclinal structure that dips smoothly to the south-west; however, the few outcrops and borehole data obtained from this area show a WSW-ENE synclinal orientation emerging in Cenozoic levels (Daguin, 1948; Howa, 1987; Courrèges, 1997).

7The northern part of the Médoc peninsula is characterized by late Quaternary deposits, occupying nine-tenths of the total area (Dubreuilh, 1971). As summarized by Coquillas et al. (2007), pre‑Holocene vs modern (Holocene included) deposits can be schematically distinguished as follows: (a) the ante-Holocene substrate comprises Eocene and Oligocene formations with potentially some submarine Miocene outcrops (Londeix, personal communication). These Tertiary formations characterize the buttress of the region and are generally capped by Pleistocene gravel terraces of fluvial origin (Fabre, 1939; Daguin, 1948; Dubreuilh, 1971). These limestones are marked by soft deformation and by karstic depressions (Dubreuilh & Marionnaud, 1972; Courrèges, 1997); (b) the Holocene formations consist mostly of sedimentary units extending laterally to the Gironde estuary or to the Atlantic coast, with a clear west-east zonal gradient ranging from massive sand dunes to muddy estuarine marsh deposits (e.g. Dubreuilh & Marionnaud, 1972; see also Bosq et al., 2019 and Stéphan et al., 2019). Marshes occupy substrate depressions: they are divided into two distinct morphological sub-units separated by a shelly-sandy ridge known as the “Cordon de Richard”. This natural barrier can be assimilated to a tempestite deposit coeval with the Iron Age (formed around 3 to 1.5 ka BP: Dubreuilh & Marionnaud, 1972; Allen et al., 1974; Diot & Tastet, 1995; Pontee et al., 1998; Massé et al., 2001). Its formation seems to have been coeval with storm events identified along the Atlantic coasts of northern Europe (Sorrel et al., 2012; Van Vliet-Lanoë et al., 2014). The “Cordon de Richard” isolated internal marshes, known locally as “palus” (the oldest natural formations), from external marshes known locally as “mattes” (which were artificially disconnected from active estuarine intertidal mudflats by Dutch polders built in the seventeenth century).

8At least two generations of Holocene dunes extend westward along the Atlantic coast. They were formed by significant inland migration of sand related to cold climate anomalies of the last millennia (Clarke et al., 2002; Stéphan et al., 2019). However, dune migration was stopped by the development of vegetation, especially during the medieval warm anomaly, and later on, in the early nineteenth century by artificial pine plantations (e.g. Tastet, 1998). At the inner foot of the modern dunes, waterlogged depressions known locally as “lette” or “lède” (typological name also identifying the present study site) are commonly observed.

2.2 - Archaeological context

9The Soulac-sur-Mer area (fig. 1) offers the possibility of directly combining palaeo-environmental (this work) and palaeogeographic reconstructions (Bosq et al., 2019; Stéphan et al., 2019; Bertrand et al., 2019) with archaeological observations (Verdin et al., 2019) and so reconstructing the local settlement history in detail. This area is well known for frequent - and largely fortuitous - discoveries of archaeological remains after erosional episodes (Dulignon-Desgranges, 1878; Buffault, 1897, 1942; Roussot-Larroque & Villes, 1988; Roussot-Larroque, 2005, 2010, and references therein). These remains are indicative of recurrent human settlement since the mid-Mesolithic at least. The LDG site has yielded flint nodules, pottery, scrapers and various chips, collected during several archaeological campaigns from 1972 (led by G. Frugier, e.g. Frugier, 1982) to 1993 (led by J. Roussot-Larroque, e.g. Roussot-Larroque, 2005). Those works helped to identify and define multiple instances of successive protohistorical settlement ranging from the Sauveterrian up to the Iron Age (i.e. roughly from 10 ka up to 2 ka BP, see Verdin et al., 2019 for details).

10Following the harsh winter storms in 2014, the site was isolated as a spur measuring about 150 m² advancing on the foreshore and separated from the dune front (fig. 1b). Excavations were conducted in October 2014 (led by F. Verdin, see Verdin et al., 2019), profoundly changing the architecture of the LDG mound which was thus subject to gradual lateral collapse under the double action of marine erosion and desiccation (see fig. 2 in Verdin et al., 2019). This rapid evolution allowed the exhumation of three complete wood traps (first observed during a field trip in February 2015, Verdin et al., 2019). These spectacular discoveries again attest to the strategic position of the LDG site for the understanding of history of local human settlement.

3 - Materials and methods

11As part of the InterLabEX LITAQ project, a large set of complementary techniques and approaches were used during field campaigns and in the laboratory. This multiproxy and multidisciplinary approach was designed to provide a better understanding of the structure of the LDG deposits in relation to environmental and climatic changes within the context of successive human settlements (consider the whole special issue which collates key results obtained throughout the 2013-2015 LITAQ project).

3.1 - Field operations

12Two main sampling campaigns were conducted as part of the LITAQ project on the LDG site. The first was carried out in March 2014 during a common campaign in which two sequences were retrieved (one core, i.e. LDG-G1 and one slab-composite sequence LDG-G2, tab. 1, fig. 1). The second campaign was conducted during the LDG excavation programme in October 2014, with the recovery of two additional sequences (slab-composite sequences LDG-G3 and G4 respectively) which completed the onsite circum-sampling (tab. 1, figs. 1 & 2). The sampling polarity for the LDG-Gx sequences was different due to the tidal context (from top to bottom for G2; from bottom to top for G3 and G4) but the protocol was the same.

13Thanks to the LITAQ project, these operations were associated with field surveys in collaboration with several teams including sedimentologists, geographers, archaeologists and (palaeo)environmentalists (UMR Ausonius – Bordeaux Montaigne Univ.; UMR PACEA and BIOGECO from Bordeaux Univ.; UMR PRODIG – Paris Diderot Univ.; UMR LETG from Brest and Caen Univ., see other articles in this special issue).

14Discrete sampling of wood, seeds, palaeosols and of sediments and outcrops from the dune escarpment were also performed all along the shore of Soulac-sur-Mer for datings and stratigraphic correlations (Stéphan et al., 2019; Bosq et al., 2019). They were coupled with topographic measurements which were compiled using geomatic tools (ArcGIS) in a regional geographic information system (part of this information can be seen on http://www.climcor-cyber.fr/​cores/​).

3.2 - Sedimentary analysis

15Sedimentological records were processed in the week following their sampling using the set of non-destructive devices at the EPOC laboratory (“Core Drill Sedimentary Analysis” PACS platform, http://www.epoc.u-bordeaux.fr/​index.php?lang=fr&page=pf_carottes). SCOPIX X-ray imaging (Migeon et al., 1999), photography, spectrocolorimetry and major/minor element quantification by X-ray fluorescence scans were performed on the sections (core or slabs) before sub-sampling for additional analyses. Some of these techniques are described below.

3.2.1 - Spectrocolorimetry

16Sediment colour and brightness measured on fresh (humid) sediment slabs is a good proxy for characterizing the nature of the deposits and helps to define major transitions. Light/dark patterns primarily reflect carbonate/clay/organic enrichments frequently forced by climate and/or sedimentary source changes (Chapman & Shackleton, 1998; Vidal et al., 1998; Balsam et al., 1999; Debret et al., 2011). The Konica Minolta CM2000d spectrometer is a portable device designed to measure SCI (Spectral Component Included) and SCE (Spectral Component Excluded) values of surface reflectance, i.e. the percentage of energy reflected (optical lightness). The latter is based on colour analysis, which is the human eye’s perception of radiation reflected in the visible range of the electromagnetic spectrum between 400 and 700 nm. From this spectrum it is possible to obtain several parameters: the total reflectance (L*) and variable chromaticity (a* and b*). This system overcomes many drawbacks with the Munsell system by specifying enlightenment and quantifying the description of the colours (e.g. Debret et al., 2011). L* ranges from 0 (black) to 100 (white). The parameter a* detects the colour changes from red (60) to green (-60) and b* measures the yellow (60) to blue colour (-60). Measurements were made for each of the Litaq-Gs sequences at a resolution varying from every 0.5 cm to 1 cm, but only the G3 sequence provided a continuous exploitable record (problem of surface heterogeneity for G2 and G4).

3.2.2 - X-ray fluorescence data

17Non-destructive sediment core scanners incorporating X-fluorescence (XRF) spectrometry measure down-core major (and some minor) element variations at high resolution (Richter et al., 2006; Croudace & Rothwell, 2015). The AVAATECH XRF core scanner used at EPOC produces semi‑quantitative results and thus gives a valuable estimation of the concentration of key lithogenic and biogenic elements (Al, Si, S, Cl, K, Ca, Rh, Ti, V, Cr, Mn, Fe, Zn, Br, Rb, Sr, Zr) reflecting the variability of the chemical composition of the sedimentary sequence over time.

18Each of the sampled sections was measured at an acceleration voltage of 10 kV for the light elements and 30 kV for heavy elements at 2 mm resolution. It is important to note that this technique requires a flat surface for analysis. In our case, we were challenged by the brittle rheology of peat. To overcome this problem, part of the LDG-G4 sections were indurated (see below, section 3.2.3.), and then measured again by XRF, in order to obtain a continuous record.

19XRF data were used as ratios since element intensities are measured on a volumetric basis and are thus sensitive to variations in grain size (Croudace & Rothwell, 2015). We focused on a few ratios which are all indicative of a specific genetic signature (Richter et al., 2006; Croudace & Rothwell, 2015). They are: (1) the Sr/Si, Ca/Si and Ti/Ca ratios: those ratios in a peaty and muddy sedimentary context can be used to differentiate marine/biogenic versus lithogenic/siliciclastic inputs (Peterson et al., 2000), and especially the Ti/Ca ratio, which reflects the proportion of sands related to aeolian advection in coastal contexts; (2) the S/Ca ratio and Br/Al ratio are related to diagenetic processes in anoxic conditions, and often trace the residual organic matter content in the sediment. Ca represents the carbonate content and biophilic halogen bromine (Br) is used as a tracer of organic matter in sediments (Caley et al., 2013); and finally (3) the (Al+Si+Ti+K+Rb)/(Br+Ca) ratio. For this latter ratio, Al, Si, Ti, K and Rb are related to the siliciclastic components of the sediment and vary with the terrigenous fraction abundance (Jansen et al., 1998). In sediments, the coarse sediment fractions are generally enriched in Ti, while Al is mostly associated with fine-particle clay minerals. The (Al+Si+Ti+K+Rb)/(Br+Ca) ratio is thus supposed to differentiate organic rich layers, and so possibly to trace and quantify the degree of peatification of the soils with low values corresponding to a peatland/bog environment whereas high values reflect more open conditions such as estuarine internal marshes.

3.2.3 - G4 induration

20Due to the rheology of the peat, obtaining a smooth surface was impossible on some samples, although it was a preconditioning step for XRF measurements. Consequently, in order to generate a continuous XRF record, an induration procedure developed at EPOC (Zaragosi et al., 2006; see also http://www.epoc.u-bordeaux.fr/​index.php?lang=fr&page=eq_sedimentologie7) was applied to the G4 sequence, before rescanning the slides and generating new XRF measurements. Previous tests performed using this procedure showed that no elemental changes were due to the resin addition.

3.2.4 - Microscopy analyses of the lithic and biogenic fractions

21To better circumscribe sources and the history of the LDG deposits, core and slab sections were subsampled at a different resolution according to the lithological changes (minimum step of 1 cm), except for the G2 sequence which was kept as an archive. Samples were then wet-sieved at 150 µm, dried and weighed, before being stored for further microscopic analyses. These analyses focused on the inventory of lithic/versus biogenic particles (including morphoscopic characterization) and made under a WILD Heerbrugg binocular microscope (http://www.wild-heerbrugg.com/microscopes.htm) at x16 to x64 magnifications. The <150 µm fraction was kept for palynological preparations (http://www.epoc.u-bordeaux.fr/&#8203;index.php?lang=fr&page=eq_paleo_protocoles) and analyses. These analyses were performed on the LDG-G3 archive by the “Pollen & Cores” company (Célia Beaudoin, ongoing study). For this work, only the AP/NAP (Arboreal/Non-Arboreal pollen) ratio will be shown. The detailed palynological results will be used for a review paper.

3.2.5 - Grain-size analysis

22These destructive analyses were undertaken on the LDG-G3 and G4 sections only (the complete G2 sequence was kept as an archive). Subsamples of bulk sediment were taken every 2 cm and analysed using a MALVERN Mastersizer S. This technique is based on illuminating the material with a laser light. The amount of light deflected and the size of the deflection angle are used to accurately measure the particle size (Loizeau et al., 1994). The range of measurable diameters is from 0.05 to 900 µm with an option of up to 3.5 mm (see MALVERN instruments at http://www.malvern.com/​fr/​products/​measurement-type/​particle-size/​). Samples were processed without any chemical pre-treatment except on some samples that were exceptionally rich in wood fragments, which were, however, previously sieved at 1115 µm to remove long wood pieces to avoid bias in the results. Considering the peaty facies of the samples, removing the organic or the very minor carbonate fractions, will have artificially modified the results towards a characterization of the siliceous fraction only. Results derived from the bulk samples are presented as a map of the relative percentages of the different grain-size fractions along the sequences. Some grain-size parameters were additionally calculated: median (D50) and 90^th percentiles (D90). They were compared with the Large Lithic Grain concentrations (weight of >150 μm grains per gramme of dry sediment).

3.3 - Datings and stratigraphic correlations

23Ten samples were sent to the VERA (University of Vienna, Austria) laboratory for AMS ¹⁴C dating (tab. 2). Large wood elements sampled for palaeogenetic analyses (BIOGECO) at the base of the LDG site (n=6) were used to infer the basal date formation of the peat deposits. Those wood samples were also the support for molecular investigations (Wagner et al., 2018). They were supplemented by additional discrete samples (n=4) from within the G4 archive where seeds were picked from the >150 µm fraction. Additional datings were obtained from the front dune outcrops (tree trunks and oak acorn caps). Dates were converted using CALIB7.02. software with the IntCal13 Radiocarbon Age Calibration Curve (Reimer et al., 2013; see tab. 2 for details). Additional stratigraphic indices were obtained from archaeological remains found directly within the LDG mound (see Verdin et al., 2019).

4 - Description of the lithological facies

24Three sedimentological units (U1 to U3) can be schematically defined along the LDG- sequence (fig. 2), with a basal unit (U1) mainly composed of clayed-peats with numerous large wood fragments, which are especially thick at the bottom. Most of them are clearly flattened and the largest have burnt barks. Datings obtained on six of these trunks have provided a mean age of 9.9 ka cal BP (tab. 2). These well preserved wood fragments were used for molecular analyses (see Wagner et al., 2018 for further details). The basal unit is ~1.5 m thick (it extends from at least 1 m to ~2.5/3 m NGF), depending on the side of the mound which is sampled. It is overlain by a second unit (U2), about 1 m thick (from at least ~2.5/3 m to ~4 m NGF) made up of grey organic sandy-clays containing many pottery and lithic fragments (see Verdin et al., 2019, for its cultural attribution). U2 is capped by a thinner Unit (U3, nearly 0.5 m thick) identified as brown peats which have preserved many organic (leaves, branches, seeds, insect wings) and archaeological remains (see Verdin et al., 2019). This topmost U3 unit can be assimilated to the last palaeosoil development before the modern dune sands were deposited. The LDG site has sizeable dips all converging towards the centre of the mound.

4.1 - Non destructive analysis (Imaging and XRF) results

25The imaging together with spectrocolorimetry data also support the identification of three main units (fig. 2, right), showing lower brightness indexes for the basal (U1) and top-most (U3) peats whereas the intermediate clayed unit (U2) exhibits higher brightness indices.

26XRF data obtained from the three sampled sections (G2 taken from the eastern side of the site, G3 from the north, and G4 from the south, fig. 3) reveal coherent patterns, in spite of lateral variations in the thickness and the occurrence of discrete centimetre-sized sub-units which give rise to amplitude fluctuation in the signals.

27As seen in the XRF selected ratio (same colour code for the Gx sequences, fig. 3), the basal peat unit is characterized by the dominance of Ca, Sr and Br elements whereas the second half upper unit records the dominance of Ti, Al and Si. A very sharp level of transition (indicated by a brown band on fig. 3) shows a high sulphur (S) content. Close to ~2.5 m NGF, this pluri-centimetric sub-unit is visible on the outcrop, especially after intense desiccation periods. Accumulations of small wood fragments together with crystallized sulphur minerals can be easily seen in the field. For all the LDG-Gx samples, the topmost unit U3 was not correctly sampled due to the high degree of desiccation and cracks. The few centimetres collected in LDG-G2 and G3 reveal high S, Br and Ca values as seen also within the bottom peat U1. Some discrete sandy and/or organic intervals are identified within the main units thanks to XRF data, with large excursions in elementary ratios, especially for Br/Al, Sr/Si, Ti/Ca, Ca/Si ratios (see for instance the large amplitude excursions in the values of the Ca/ Si ratio in fig. 3 for each LDG-Gx). The sandiest levels (also identified on the basis of grain-size data, see fig. 4) are marked by high Fe/Al and Ti/Al ratios (data not shown here but available on request from the corresponding author).

28The selected XRF ratios highlight distinctive geochemical signatures within the Gx sequences (fig. 3):

29- the S/Ca ratio indicates a transition level at around ~2.5 m NGF. One dating obtained in G4 places this level at 6.25 ka BP (4.3 ka BCE);

30- before this level, low values of the (Al + Si + Ti + K + Rb)/(Br + Ca) are recorded whereas they increase in the upper U2 and U3 units. In G3 and G2, the Ti/Ca ratio identifies the clay-rich part of the sequence whereas this trend is interrupted in G4. It is worth noting that one of the “wood traps” was found next to G4, above the sulphur level;

31- the Br/Al and Sr/Si ratios show high (and spiky) values mainly in Unit 1 and are near to zero in the uppermost layers, except when entering the peaty Unit 3;

32- the (Al + Si + Ti + K + Rb)/(Br + Ca) ratio exhibits a highly discriminant pattern throughout the sequence and also clearly indicates the sulphur-rich “central” layers.

4.2 - Destructive analysis results

33LDG-G3 and LDG-G4 sub-samples were used for morphoscopic observations of the large (>150 µm) fraction and for grain-size analyses (see methods above). The figure 4 compiles those data together with an analysis of the biogenic fraction: from pollen with the AP/NAP ratio in G3 and for plant remains for G4.

34The large (>150 µm) fraction varies sharply within the 3 metres covered by the LDG-G3 and G4 sequences, mostly reflecting various amounts of organic (vegetation mainly) and sandy remains (see below). This large fraction represents a mean of 10 % of the total raw sediment in G4 and G3 (fig. 4), with a noticeable peak observed in G3 around 200 cm at the top of unit U2 before U3 (i.e. between 3.5 and 4 m NGF, not sampled in G4, probably because of discrete lateral changes within the mound). The median (D50) varies between 10-60 µm (mean of 42 µm in G3 and 35 µm in G4) and the percentile (D90) between 65-1150 µm (mean 295 µm in G3 and 275 µm in G4) similarly in the two retrieved records.

35Grain-size data highlight a distinct change between the basal peats and the overlying clays with the sulphur-rich layer (around ~2.5 m NGF) as a transitional mid-level. A sharp change therefore occurred in the sedimentation at that time. The different units are well identified with well-sorted sediment within the bottom unit U1 with a principal modal distribution of particles between 4 and 5 Φ (63-250 µm) (see the coloured mapping of the grain-size data) in a topmost unit where the modal distribution of sediments is larger between 5 and 10 Φ, in G3 and G4 as well.

36The data compiled from visual particle analyses on the large fraction (>150 µm) in G4 are shown as: the roundness of quartz grains, the ratio between biogenic vs terrigeneous remains (expressed as a percentage, fig. 4) and some observations of plant remains. This work reveals the dominance of quartz with some iron-coated grains, together with abundant and well preserved organic remains: plant debris (seeds, wood, fibres) from aquatic species mainly and with some arthropod remains (insects, wing covers). Heavy minerals and mica are rare. Based on plant remains, the basal unit U1 is interpretated as a peat layer. This portion of the sequence exhibits weakly or moderately abundant rounded sand grains at the base whereas their concentration increases at the top (see blue curve in fig. 4). The occurrence of filamentous plant debris, wood and seeds is observed throughout the unit and the biogenic fraction could reach 80 % before the sulphur level. Over the first 75 cm, we note the predominance of a mixture of angular and rounded quartz grains, possibly due to the reworking of fluvial or estuarine sands whereas the uppermost part of the mound is marked by a clear predominance of rounded quartz grains shaped by wind erosion (Cailleux & Tricart, 1959). The topmost unit U2 is characterized by more plastic and compact clayed sediments and includes rounded quartz grains exclusively. Seeds and woody debris are less frequent in this unit. The NAP/AP ratio obtained from the pollen analysis of G3 confirms a more open landscape after the sulphur level transition.

5 - Discussion

5.1 - Towards a synthetic view of the LDG stratigraphy

37Our study proposes a reappraisal of the sedimentology of the Lède du Gurp outcrop. Sedimentological descriptions of this site were previously documented in only a few archaeological studies (Roussot-Larroque & Villes, 1988; Guibert et al., 1996; Rousssot-Larroque, 2010). Here, coupled with archaeological investigations (Verdin et al., 2019), the new data generated by our study represent a true update and provide the opportunity to redefine the sedimentological context accompanying nearly the last 10 millennia and the successive human settlements. It is also the first time that ¹⁴C AMS datings combined with imaging and XRF measurements on continuous sections, together with high-resolution analyses, have been used at this location. Many of the previous studies used data obtained in the early 1970s by Dubreuihl (1971) for which stratigraphical elements were extrapolated from one site to another along the coast between Soulac-sur-Mer and Hourtin, leading to some misinterpretations especially of the supposed Pleistocene surrounding units (e.g. grey-blue sands and palaeosoils at the foot of the modern dune). However, archaeological studies have provided precious chrono-cultural elements, supported by a few ¹⁴C ages (Roussot-Larroque, 2005). These ages are robust enough to constitute a reference time framework which will form the backbone of the following discussion and support our work (fig. 5). The following discussion tries to reconcile previous observations with the new lithostratigraphic observations. Our aim is to provide the basis for a reference Holocene sequence for this part of the Médoc peninsula.

38This attempt is illustrated in figure 5, in which we have chosen to summarize the lithology of the LDG site by the (Al + Si + Ti + K + Rb)/(Br + Ca) and S/Ca ratios (as obtained from the three sedimentological records LDG-G2-3-4). Those ratios provide a good illustration of changes related to the content of organic matter and clays trapped in what was probably a pond. They may also reflect part of the local human practices (see Verdin et al., 2019). According to the successive studies of Roussot-Larroque & Villes (1988), Marambat & Roussot-Larroque (1989), Marrambat (1992) and Roussot-Larroque (2005, 2007, 2010), several levels are distinguishable in the LDG sequence. Even if our results argue for a three-step history of infilling and thus simplify the recovered records, discrete events are accessible thanks to the high resolution of our XRF measurements (every 0.2 cm) and destructive sub-samplings (every 1 cm). However, at present we do not have complete datings for the three sequences with which to further interpret the detected events but we observe a quite coherent succession of deposits. Taking into account the synthesis and chronological references in Roussot-Larroque & Villes (1988), the LDG site includes:

39- a thick peaty layer dated from the Mesolithic to the Early Neolithic at the bottom of the deposit (from at least 1 m to ~2.5 m NGF);

40- this layer is overlain by an early Neolithic to Middle Bronze unit (layers 8 to 4b, Marambat & Roussot-Larroque, 1989) characterized by a clayey sediment intercalated with discrete sandy layers and numerous archaeological remains;

41- the site ends with a peat cap (layer 4a) transitional between the Middle Bronze Age and the Iron Age.

42It is worth noting that sizeable lateral variations sometimes occur within the mound (mainly due to centripetal dips) and that the outcrop previously described by Roussot-Larroque & Villes (1988) may thus have included some additional (or missing layers). The following section tentatively reconstructs the environmental conditions that may have prevailed during the accumulation of the LDG sequence.

5.2 - Palaeoenvironmental evolution of the Lède du Grup during the Holocene

43At the end of the last ice age, the warmer Holocene climate and the associated boreal ice-sheet melting produced a rapid eustatic sea-level rise with estimated flooding rates of up to 1 meter per century (10 mm/yr) between 10000 and 6000 years ago (early Holocene, Flandrian transgression, see fig. 6C). Those transgression rates decreased significantly around 6000 years BP when reaching roughly its current level (Gensous, 1971; Ters, 1973; Paepe et al., 1973; Diot & Tastet, 1995; Lambeck, 1997; Allard et al., 2008; Ward et al., 2016). The shoreline consequently moved from - 80 to - 40 m between 13000 and 9000 years BP (Gensous, 1971) and the sea invaded palaeovalleys along the western boundary of France (Houthuys et al., 1993), Gironde included.

44Due to the low topography of the Médoc area, three main arms were formed at the Gironde estuary mouth (Gensous, 1971; Allen et al., 1974): one, to the north that extended along the line of the modern main channel; a second, between Soulac-sur-Mer and Le Verdon-sur-Mer and a third to the south, between the beaches at Le Gurp and Montalivet (fig. 6, see also Stéphan et al., 2019). If some contributions (e.g. Dulignon-Desgranges, 1878) report the existence of a stream still flowing in the vicinity of the Le Gurp area (not accurately located), robust sedimentary evidence argues for an active channel near Soulac-sur-Mer up to the fifth century of the Common Era (Pétorin 1988; Stéphan et al., 2019). The extension of these palaeovalleys has also been recognized seaward (e.g. Froidefond, 1984; Lericolais et al., 1998).

45Along the Soulac-sur-Mer coast, many authors (for a review, see Stéphan et al. 2019) have tried to link the occurrence of clayed facies or palaeosoils beyond the modern dune to deglacial flooding contexts: they were related to post-glacial infillings attributed to the Neolithic/Flandrian period for the topmost ones (e.g. Fabre, 1939; Roussot-Larroque, 1984; Pétorin, 1988), and to pre-Würmian transgression for the deepest (e.g. Dubreuilh, 1971; Bosq et al., 2019).

46The Gironde estuary output volume progressively decreased with the Flandrian flooding, promoting the development of swamps in the floodplain (Allen et al., 1970; Pontee et al., 1998; Diot et al., 1999). The intense peatification which characterizes the clayed unit at the LDG site is remarkable. The presence of tangled tree trunks at the base testifies to the development of local forest cover in a riparian and/or swampland context (Marambat & Roussot-Larroque, 1989), as is also supported by the AP/NAP ratio. The new dates obtained for the large basal pieces of wood (oak mainly, those located at a mean altitude of 1 m NGF) reveal a mean age of nearly 10 ka BP (fig. 5), which coincides with the site flooding (probably linked to changes in the phreatic level, Roussot-Larroque, 2007) and the further development of peats, stopped and sealed later by the establishment of the modern dune. According to the classification of peat ponds (e.g. Manneville et al., 2006), the site was probably linked to a fluviogenetic context as it was probably perennially flooded and had a positive hydrological budget. According to previous authors, the dune formation (three successive generations) seems to have taken place in a context of climatic deterioration associated with unidirectional winds remobilizing sandy deposits, especially those of the earlier continental system known as the Sables des Landes (Pontee et al., 1998; Tastet & Pontee, 1998; Diot, 1999). In between these two extreme stages (pond vs dune), a multi-millennial sedimentation occurred, concentrating both natural (pollen as seen with the NAP/AP curve, seeds, wood, plants, insect remains) and human artefacts (pottery, bones, manufactured woods, pebbles) in a water-saturated and organic-rich context. Several sublayers are distinguishable within the three main units on the basis of XRF (fig. 4). According to Roussot-Larroque (2005), many of them, and especially those capped by thin sandy layers, were marked at their surface by animal tracks. For that author, it was there, in the shelter of the dunes, between river and marsh that Meso-Neolithic populations settled.

47Our new datings and observations (see also Verdin et al., 2019) show that denser human occupation probably began around 6 ka BP (fig. 5). This date is contemporaneous with the development of proximal tidal marshes on the left bank of the Gironde estuary (La Perge and Reysson marshes; Diot et al., 1999; 2002; Stéphan et al., 2019) also accompanying the slowing of the sea-level rise (fig. 6C).

48From 6 ka BP to the onset of the dune invasion, sedimentation dominated by clay deposits characterizes the entire LDG depression (figs. 2-5). The general organization of the deposits within this median unit U2 (from ~2.5 to 4 m NGF) corresponds to sediment accumulation in a quiet environment such as a marsh or wetland. This is also what was deduced by Roussot-Larroque (2007), especially for the Middle Bronze period coinciding with a high level of water with compact clay deposits. Within those deposits, an obvious and marked intensification of human activity, with traces of livestock, agriculture and metallurgy, associated with typical Middle Bronze Médoc artefacts (pan, axe, crucible slag, ceramic pelletizing) was detected by that author. A rapid transition in sedimentation thus occurred when entering the Iron Age (approximately next to the top of the Gx sequence, potentially U3 unit), with the development of a 30 cm-thick (to 50 cm-thick according to our observations) peat layer especially rich in well-preserved plant remains. This latter probably coincided with wetter climatic conditions that allowed the local development of palaeosoils and forests just before (and during, for some discrete episodes) the formation of the modern dune (Clavé, 2001; Roussot-Larroque, 2005). It is noteworthy that archaeological investigations identified no indication of the final bronze industry, a gap attributed to a decline in agriculture also observed on a wider scale across France (Roussot-Larroque, 2007).

49To explain the palaeogeographic conditions that promoted the establishment of the Holocene sequence of the LDG, several hypotheses could be considered:

50(1) The first hypothesis implies an infilling of an abandoned meandering channel or arm (Gé et al., 1995). This echoes the context of a fluviogenetic peat. However, the absence of marked sedimentary features (current ripples, stratification, etc.) in the most basal part of the site seems to exclude this hypothesis. Although the existence of the channel at the Le Gurp location is supported by several previous studies (Gensous, 1971; Allen et al., 1974), its centre line lay south of the LDG site (see fig. 6b).

51(2) The second hypothesis calls for the collapse of a sinkhole (Bosq, 2015; Bosq et al., 2019); this assumption is based on observations by Courèges (1997) and is reinforced by the existence of depressions due to karstification processes in the low-Médoc (e.g. Polje of Reysson, near Saint-Germain-d’Esteuil). Several elements significantly support this assumption, especially the inward dip of the peat deposits (compared to an amphitheatre by Julia Roussot-Larroque) and the now isolated and circular morphology of the LDG site on the coast;

52(3) A third alternative and final hypothesis, potentially reconciling all the observations and hypotheses can also be proposed by referring to an alas-type thermokarst structure (well known as thermokarst lakes, e.g. van Everdingen, 2005). This latter is especially pertinent when considering the glacial and post-glacial evolution of the area and deserves to be further developed. Thermokarst lakes are round depressions at present commonly observed in boreal latitudes (Tonolio et al., 2009; Groose et al., 2013), close to fluvial distributaries (e.g. Mackenzie delta, River Lena, Lake Albanal where they coexist with bogs: Cailleux, 1971), but not only. Their formation is mainly derived from loading phenomena during the melting of permafrost. In several Western European countries many periglacial structures and forms are due in reality to these thermokarsts (Dylik, 1964). These cavities are produced (or at least begin) by melting corner slot networks (ice wedges) (Cailleux, 1971; Costard & Kargel, 1995; Groose et al., 2013). Such evidence of periglacial coverage and of cryoturbation figures were noted by Dubreuilh and Marionnaud (1972) and Tastet (1999) on Pleistocene age sands that underlay/surrounded the Holocene deposits of the LDG site. We also observed such artefacts at the dune foot during our field campaigns (as did Bosq et al., 2019). Based on those observations, we can assume the existence of peri-glacial processes within the soil formation type at the LDG site during the late Pleistocene/Holocene transition. Heinrich stadials, which register the collapse of the boreal ice-sheets and especially the later HS1 (recorded for 3 ka from 18 to 15 ka, e.g. Sanchez-Goñi & Harrison, 2010; Eynaud et al., 2009), are worth considering in this context. HS1 was actually marked by the migration of the polar front as far south as latitude 42°N across western Europe with pan North-Atlantic sea-surface freezing conditions recorded at that time (e.g. Bard et al., 2000; Eynaud et al., 2009, 2012). Such marked oceanic cooling could have implied the development of discontinuous permafrost on-land and especially at coastal sites. A derived alas structure is therefore a good candidate for explaining the round LDG site geometry, the circular structure later became a trap where peaty clayed (and sometimes more sandy) sedimentation accumulated through the Holocene.

6 - Conclusions

53This study reports new sedimentological information that provides the basis for new chronostratigraphic interpretations of the Lède du Gurp (LDG) archaeological site within the framework of Holocene palaeoenvironmental and palaeogeographical changes. Especially based on high resolution elemental data obtained by X-ray fluorescence scanning, grain-size and morphoscopic analyses and new datings, our observations provide evidence of a depositional history of at least three main units in a peaty-clay-dominant context. Units of at maximum 3 m (for those visible on the shore) are piled in a circular depression, which could be the heritage of a (thermo)karst structure. Invaluable archives of both natural landscape evolution and of anthropic-related changes are trapped in the sediment, with especially a marked transition which occurred around 6.2 ka BP (4.3 ka BCE) when the sea-level rise decelerated. It is also during that time that increasing evidence of human settlement has been identified.

54Together with other contributions to this special issue, this work, within the framework of the LITAQ (2013-2015) project, helps to refine previous archaeological investigations and reactivates palaeoenvironmental research in the area of Soulac-sur-Mer after 15 years of interruption. However, additional analyses and regional correlations are still needed. The observations and conclusions of this study are only preliminary results, and supplementary datings and a detailed event-sequence analysis still have to be conducted to better constrain the fine stratigraphy and the chronological evolution of the deposits with regard to the human settlement of the site under changing Holocene environments.