The person in charge of the archaeological operation, V. Duphil, and the Inrap scientific and technical director of Aquitaine, L. Detrain, are thanked for their help. The study was financed by Inrap and a project led by W. Banks (UMR PACEA, CNRS-University of Bordeaux), whom we thank. M.F. Sanchez-Goñi and an anonymous reviewer are greatly acknowledged for their detailed and constructive comments on the manuscript.

* corresponding author

1 - Introduction

1While pollen sequences documenting the Last Glacial (e.g., Woillard, 1978; De Beaulieu and Reille, 1984, 1992a, 1992b; Guiter et al., 2003; Fletcher et al., 2010; Sirocko et al., 2016; Duprat-Oualid et al., 2017 ; Kern et al., 2022) remain scarce in western Europe, they are especially rare and discontinuous in Aquitaine, southwest France (Paquereau and Schoeller, 1959; Paquereau, 1961; Diot, 1999). As a result, vegetation contemporary with the Last Glacial remains poorly documented there. The closest available data are mainly from the Massif Central (Reille and De Beaulieu, 1988, 1990; De Beaulieu and Reille, 1992b), the Pyrenees (Olfield, 1968; Andrieu-Ponel et al., 1988; Jalut et al., 1992; 1998; González-Sampériz et al., 2006) and the Loire and Garonne basins recorded in the marine sediments of Bay of Biscay (Sanchez-Goñi et al., 2008, Fletcher et al., 2010). These records provide only an imperfect picture of lowland vegetation in Aquitaine, either because of the altitude (Massif Central and Pyrenees) or because the pollen inputs come from a wide geographical area (Bay of Biscay).

2The Pleistocene peat levels discovered during preventive archaeological work offer the opportunity to document in detail, albeit during short time windows, the glacial palaeoenvironments of the region. Although discontinuous, these records provide information on the vegetation mosaic characterizing a region and may allow the identification of potential refuge areas for temperate trees. Due to poor drainage, the Landes plateau has yielded numerous occurrences of peat intercalated in the aeolian ‘Sable des Landes’ Formation. Some of these levels have been the subject of pollen analysis and numerical dating, which made it possible to relate them to interstadials of the first half of Marine Isotopic Stage (MIS) 3 (i.e., 55-40 ka; Cestas-Les Pins de Jarry, Bertran et al., 2009, 2020; Barp-Pot aux Pins, Sitzia et al., 2015) (fig. 1) or to the Middle Pleistocene (Hourtin, Sitzia et al., 2015, north of the coversands not shown in fig. 1). The discovery of a peat level about 15 cm thick at Saint-Vincent-de-Paul in the southern margin of the ‘Sable des Landes’ provides an opportunity to document another late MIS 3 interstadial and is the focus of this paper. The presence of wood fragments gives this peat an exceptional character.

2 - Geomorphological setting

3The study site is located in the Adour valley on the right bank, approximately 11 km northeast of Dax (southwest France) (fig. 1). The 1:250,000 digital geologic map of Aquitaine (Bourbon, 2019) indicates that the bedrock is a high alluvial terrace of the Adour River (Fv, Lower to Middle Pleistocene). Aeolian sands several meters thick, which are an integral part of the ‘Sable des Landes’ Formation (Sitzia et al., 2017), overlie the alluvial deposits. The geotechnical drillings carried out as part of the development of the site did not reach the alluvium at a depth of 3 m. According to borehole BSS002EAWU located approximately 3.3 km to the southwest and stored in the Banque du Sous-Sol of the Bureau des Recherches Géologiques et Minières (http://infoterre.brgm.fr/​), the sands would reach 5.5 m in thickness in the area. The aeolian cover has yielded two ages by Optically Stimulated Luminescence (OSL), one related to the Middle Pleistocene (~160 ka), the other to the Upper Pleistocene (~26 ka) in the Pontonx quarry 3 km to the southeast (Bertran et al., 2020).

4The topography visible from the 5 m DEM of the Institut Géographique National (RGEALTI® 5 m) shows that the area is covered by ridges 1 to 2 m high and oriented WNW-ESE, which correspond to the arms of elongated parabolic dunes (fig. 2A). These degraded dunes have been dated to the Younger Dryas (~12 ka) at various points in the Landes, notably at Castets and Saint-Geours-de-Maremne (Bertran et al., 2020), and mark the ultimate evolution of Pleistocene coversands. The relief was then dissected by a network of gullies connected to tributaries of the Adour River.

3 - Methods

5A peat block was collected in trench CM1-T10 (fig. 3) for a multidisciplinary study including pollen and spores, non-pollen palynomorphs (NPPs), charred particles and woody macroremains. For the palynological study (pollen and NPPs), all samples were processed following the standard method developed by Faegri and Iversen (1975), which consists of chemical treatments with 10% NaOH, 37% HCl and 48% HF. Then a separation with a heavy liquid (ZnCl2, density=2) has been carried out. No acetolysis was made in order to recognize any contamination by exogenous pollen stocks. Finally, the rest of the sediment was sieved using 10 µm and 200 µm mesh sieves.

6The preparation involved a volume of 4 cm³ of sediment. To estimate the richness in palynomorphs, the absolute pollen concentrations (number of pollen grains and spores per cm³ of sediment) were calculated according to the method of Stockmarr (1971). Palynomorphs were mounted in a mobile medium (glycerine) and identified at 500x and 1000x magnification. Identification of spores and pollen material was made using determination keys (Faegri and Iversen, 1975; Moore et al., 1991; Beug, 2004), photographic atlases (Reille, 1992, 1995, 1998) and the CReAAH pollen reference collection. The names of taxa followed by “type” refer to the nomenclature of “The Northwest European pollen flora” for Asteraceae, Gentianaceae, Polygonaceae, Sparganiaceae families, and to the nomenclature of Moore, Webb and Collinson for Plumbaginaceae family (Punt, 1975; Punt and Nienhuis, 1976; van Leeuwen et al., 1988; Moore et al., 1991; Punt and Hoen, 2009). Determination of non-pollen microfossils was carried out with the atlas published in 1998 based on the work of van Geel (van Hoeve and Hendrikse, 1998), supplemented by the synthesis of Miola (2012) and the theses of Jouffroy-Bapicot (2010) and Cugny (2011). The diagram was built on a basic sum from which sedges (Cyperaceae), spores and indeterminate grains were excluded. NPPs are expressed as a percentage of the pollen sum. Microscopic charred particles (10-200 µm) observed in pollen sample residues were also counted on five lines at 500x magnification on each pollen slides. The concentration of charred particles in the sediment was obtained using an adaptation of the volumetric method (Cour, 1974). This method was chosen because the morphological features of charred particles differ from Lycopodium spores, which are spherical. The concentration of charred particles is expressed as the number of charred particles per cm³ of fresh sediment.

7For the xylological study, macroremains recovery involved water sieving of the sediment on a column of 250 µm, 500 µm and 2 mm sieves. The volume of sieved sediment ranged from 100 to 400 cm³, depending on the available sampling volume (tab. 1). Wood analysis was performed on the remains larger than 2 mm. A number of 40 to 50 individuals were analyzed in the majority of samples.

8The minimum number of fragments to be analyzed per sample was determined from the cumulative frequencies of taxa as a function of the number of individuals, the optimal number being reached when the frequencies of taxa are stabilized. Prior to taxonomic identification, macroscopic analysis of the wood fragments was carried out under a binocular microscope, i.e. length, width, thickness, presence of bark, pith, and stem starts associated or not with leaf scars, degree of ring curvature according to three classes (strong, intermediate and weak). The taxonomic identification was carried out with a transmitted light microscope, (x200 and x500) using xylological atlases (Schweingruber, 1978, 1990). For the genus Erica, the arrangement of leaf scars and their number, observed under binocular microscope, was a criterion for the specific determination, using the flora of Rameau et al. (1989). The taxonomic identification of the root wood could not be realized as root wood does not have a uniform anatomical structure and cannot be identified except for rare cases (Schweingruber, 1978).

4 – Results

4.1 - Stratigraphy

9The trenches carried out during the archaeological survey allowed us to observe the sedimentary formations over a thickness of 3 m. The stratigraphy is relatively homogeneous on the land parcel. It is schematically composed of the following units, from the top to the bottom (fig. 3):

10(1) A Holocene podzol, about 1 m thick. The upper horizons have been mixed by agricultural work (black Ap horizon, 0.4 to 0.5 m thick). The underlying horizons (Bh, Bs) are poorly cemented by Fe oxides.

11(2) Light brown sands becoming light grey at the base, with dominant subhorizontal bedding. Levels with deformed or slightly oblique bedding are present. This unit corresponds to aeolian sandsheet deposits (coversands).

12Two trenches (CM1-T10 and CM2-T9) revealed a peaty level beneath the sandy cover between 2.2 and 2.4 m deep (fig. 2B). This level was described in detail and sampled in trench CM1-T10 (fig. 3). It shows at the base (fig. 4) alternating peaty and sandy beds (2 cm thick); then brown peat with abundant plant debris (4 cm) (fig. 5), humified black peat with plant debris (3 cm), and finally, at the top, alternating sand and black peaty beds (5 cm). The base of the sequence consists of coarse subhorizontally bedded brown sands of alluvial origin.

13The limited spatial extent of the peat suggests that it formed in a gully incised in aeolian deposits. Examination of the 5 m DEM (fig. 2) indicates that the area is intersected by several small tributaries of the Adour River, notably the Pont-de-Paul creek that flows 400 m west of the parcel. These shallowly incised tributaries (2 to 3 m) form a roughly dendritic network, converging on a single, deeper talweg at the edge of the present Adour plain. The hypothesis of an ancient gully incised into the sands during an interstadial period, then largely filled by aeolian deposits during the Last Glacial Maximum can therefore be proposed to account for the localized presence of peat accumulation. In trench CM1-T10, two generations of aeolian sandy fill separated by a small incipient podzol can be distinguished. The more recent generation of sand is likely related to the emplacement of the Younger Dryas dunes.

4.2 - Chronology

14Two samples were dated by AMS-14C at the CIRAM laboratory (Martillac). A first large wood fragment (4 cm long, 1 cm wide) was taken from the median brown peat level. The age obtained (CIRAM-AV85) is 26,731 ± 101 BP, i.e. 31,136 to 30,844 cal BP (2s, Intcal20). Another smaller willow fragment from the basal sand and peat beds was also dated. The age (CIRAM-4395) is 28,066 ± 106 BP, i.e. 32,821 to 31,655 cal BP. The peat can therefore be correlated with the GI-5.1 or GI-5.2 interstadials of the Greenland ice core record (Rasmussen et al., 2014), i.e., in the latter part of MIS 3. Insofar as the study of pollen assemblages (see infra) shows that the peat recorded only one phase of climate amelioration and no discontinuity in sedimentation is identifiable within the level, a correlation with GI-5.1 is favored here. According to this hypothesis, the small willow fragment taken from the base of the peat could correspond to slightly older material reworked at the bottom of the gully.

4.3 - Palynology

4.3.1 - Pollen and NPP

15Pollen sedimentation conditions appear favorable, ensuring reliable and significant data from a palaeoecological point of view (Reille, 1990; Leroyer, 1997), as shown by high counts and absolute concentrations, good preservation of pollen stocks, and satisfactory taxonomic diversity (tab. 2, fig. 6). This has made it possible to draw up a floristic list of 44 taxa, to which can be added 26 NPPs (fig. 6). The pollen assemblages (tab. 3, fig. 6) allows us to identify three local pollen zones (LPZ), with the first two composed of only one sample each.

16Characterized by the scarcity of tree pollen, the first local pollen zone (LPZ 1, sample at 15 cm from the top of the peat) covers the base of the organic level (peaty sand). The environment appears very open with the local presence of a reed bed-megaphorbia or a sedge meadow (Poaceae, Cyperaceae, Rosaceae, Ranunculaceae, Thalictrum), formations that thrives on waterlogged soils with little oxygen (Fare et al., 2001). The absence of aquatics indicates that there was no perennial water body (Dutartre and Rebillard, 2015). In addition to pollen assemblages, the presence of this vegetation is corroborated by the predominance of Clasterosporium caricinum (HdV-126) and Byssothecium circinans (HdV-16), microfossils that are respectively associated with Carex (Cyperaceae) and Poaceae but are absent from very wet areas (van Geel, 1978; van Geel and Middeldorp, 1988; van Geel and Aptroot, 2006; Yeloff et al., 2007). A few rare trees were present, mainly willows (Salix) whose occurrence in situ is confirmed by woody debris (fig. 7) but also some rare hygrophilous shrubs (Myrica, Ericaceae, Calluna). In the surroundings, the drier areas hosted a low steppe vegetation, dominated by sagebrush (Artemisia) accompanied by other heliophilous herbaceous species requiring well-drained soils. In light of the low pollen count (12 %), the local presence of pine appears unlikely (Huntley and Birks, 1983). Its record is more likely to be the result of distant contributions in an open environment because of high pollen production and its capacity for long-distance dispersal (Heim, 1970). Similarly, the sporadic nature and low occurrence (< 0.5 %) of temperate hardwoods (Fagus, Carpinus, Corylus) also argue for distant contributions even though the maintenance of beech in southwestern France during the Late Pleistocene is plausible based on palaeogenetic data (de Lafontaine et al., 2014). Although current pollen studies show that percentages of Fagus below 2 % may indicate the local presence of isolated trees (Huntley and Birks 1983), the persistence of beech during MIS 3 near St. Vincent de Paul cannot be assumed from our data, since it is only recorded in the basal peat sample, which also provides the greatest amount of pine pollen from distant sources.

17The base of the brown peat (LPZ 2, sample at 10 cm) shows a profound local environmental change. The wetland was colonized by birch trees (Betula), accompanied by willows and rarer Ericaceae and marsh heather (Calluna). The development of trees created a filter to distant inputs of pine pollen, which supports the idea that pine was not established in the immediate vicinity. Similarly, the perception of steppe vegetation (Artemisia, Asteraceae, Chenopodiaceae) decreased, but it is not possible to assess whether it really declined. Although more constrained, a reed bed remained in the site. It evolved somewhat with the appearance of meadowsweet (Filipendula) and bur-reed (Sparganium emersum type) at the side of Cyperaceae and Poaceae. Bur-reed indicates, in the absence of true aquatic plants, soils that are often wet but without perennial water body (Aniotsbehere, 1999). This interpretation is corroborated by the microfossils HdV-128 and HdV-187d, the latter infatuated with Phragmites and Carex, both of which indicate shallow, eutrophic water (Pals et al., 1980; van Geel et al., 1983). However, the appearance of copepod spermatophores could suggest the at least temporary occurrence of open water (van Geel, 1978). These small crustaceans can also be related to the establishment of sphagnum mosses (Mighall et al., 2006), reported by their spores and the presence of Tilletia sphagni (HdV-27), which is dependent on them (van Geel, 1978).

18The third pollen (LPZ 3, samples at 7.5, 5 and 2 cm) zone shows the development of a sphagnum bog, which is attested, in addition to the rise of Sphagnum spores, by Tilletia sphagni (HdV-27). Other microfossils typical of sphagnum bogs are present, e.g. Assulina muscorum (HdV-32a) and Callidina angusticolis (HdV-37). The former has broad ecological requirements, but has a high relative abundance in drier stages of bogs (Payne et al., 2012). The evolution of vegetation appears to have been gradual with, initially (LPZ 3a), the retreat of birch trees to the benefit of willows, with also some Ericaceae and bog myrtle (Myrica). Then (LPZ 3b and c), the site was colonized by Myrica, a shrub specific to acidic and humid soils, especially peat bogs (Bournerias, 1979). Birch trees (LPZ 3b) and then willows (LPZ 3c) disappeared almost completely, but a few Ericaceae and Calluna remained. The development of a denser local vegetation led to the non-perception of pines. On the other hand, the persistence of a reed bed is testified by pollen of Poaceae and Cyperaceae but also by the presence of microfossils that are attached to them such as Clasterosporium caricinum (HdV-126) and sometimes Glomus cf. fasciculatum (HdV-207) (van Geel et al., 1989, 2003).

4.3.2 - Palynofacies and charred particles

19The palynofacies consists mainly of fresh non-woody plant debris, i.e. rootlets of Cyperaceae, tissues of aerial parts of Monocotyledons, indeterminate herbaceous tissues, and more rarely Bryophyte and Sphagnum tissues. Charred particles are rarer, especially in the two lower samples (15 and 10 cm), where their concentrations are less than 1500 debris/cm³. However, the charred particles increase abruptly from 7.5 cm where they reach their maximum concentration (117,365/cm³) and then decrease to around 85,000/cm³. They include remains of woody and herbaceous nature, of which some correspond to tissues of aerial parts of Monocotyledons. Their size is small, mostly less than 100 µm, but some herbaceous remains can reach up to 420 µm in length and 140 µm in width. The increase in the concentration of charred particles is synchronous with the densification of local woodlands, corroborating their association with a higher biomass characteristic of wetter interstadials (Daniau et al., 2009, 2010).

4.4 - Macroremains

4.4.1 - Representativeness of data

20While a minimum of 40-50 individuals was analyzed in most samples, the optimum number was not systematically reached (i.e. for sample 0-4 cm). Wood fragment concentrations increased from the base to the top (tab. 1), reaching up to 243 fragments per 100 cm3 of sediment in the top sample. Only five taxa were identified, i.e. birch (Betula pendula/pubescens), willow (Salix), marsh heather (Erica cf. tetralix), heather (Erica) and Ericaceae (tab. 3). The high rate of indeterminate debris results from the abundance of root fragments, which do not allow identification except occasionally for birch (Schweingruber, 1978, 1990).

21Preservation of woody material is variable across layers and taxa. For example, Erica cf. tetralix and Ericaceae fragments are generally well preserved in the four samples where they were identified (tab. 4). The condition of willow fragments is variable, but mostly quite poor, while birch fragments are poorly preserved in the three lower samples (7.5-10, 10-15 and 15-17 cm). The willow and birch fragments have a compressed cross section. The fragments are small with a length ranging from 0.06 to 5.1 cm (mean 1.3 cm) and a width from 0.05 to 2 cm (mean 0.25 cm). In fact, the different species show slightly different sizes (tab. 4). When the cross-section is complete, the diameter could be calculated. It never exceed 0.5 cm and is mostly smaller than 0.25 cm. The largest diameters correspond mainly to indeterminate remains, to which are added one fragment of birch and one of heather. The woody material is, therefore, of small size. When identifiable, the curvature of the rings is always strong, except for one individual with intermediate curvature, pointing to small branches.

4.4.2 - Evolution of the macroremain assemblages

22The general diagram (fig. 8) shows the predominance of willow (Salix) ahead of birch (Betula) in the three lower samples (7.5 to 17 cm), followed by indeterminate remains (4 and 5 cm) and, finally, heather (Erica) at the top. Salix has its maximum representation at 10-15 cm then decreases somewhat (7.5-10 cm) and then disappears completely from 7.5 cm. Its evolution as woody fragments roughly follows that documented by the pollen assemblages, which shows the retreat of willow in the upper levels (fig. 6). However, the relative abundance of Betula in the basal peaty sand beds raises questions, as it is not attested in pollen assemblages. Since this absence cannot be attributed to a deficit in pollen production and dispersal (Heim, 1970), it is likely that the birch fragments correspond to roots, which are identifiable for this species (Schweingruber, 1978, 1990). They would, therefore, derive from the birch stand that later established in the site.

23Erica appears in the 7.5-10 cm sample. It is the only taxon recorded in the two overlying samples (0-5 cm) that yielded mainly indeterminate woody debris (> 90 %). Erica becomes largely predominant at the top, especially if we accept that the Ericaceae fragments refer to this genus. The woody debris thus document an environment somewhat different from that depicted by palynology, as it includes no remains of Myrica but almost exclusively of Erica cf. tetralix. This difference is likely due to two factors, the pollination mode of Ericaceae, which are always under-represented in pollen assemblages (Heim, 1970) even if in areas where Erica sp are locally abundant, pollen values > 25 % can be frequent (Huntley and Birks, 1983), and the very local information provided by the macroremains. It can be assumed that only marsh heather was growing at the sampled site, while bog myrtle was located a little further away, but probably in the immediate vicinity as Myrica usually has a low pollen representation, even where it is abundant and growing luxuriantly (Huntley and Birks, 1983).

5 - Discussion

24As mentioned above, continental records documenting the Last Glacial vegetation remain scarce in Western Europe. Vegetation is mainly documented by a few long continental sequences geographically distant from southwest France, supplemented by fragmentary records from northwest Europe coming from organic levels intercalated in glacial deposits (Fletcher et al., 2010). The pollen analyses carried out on the peaty levels discovered during preventive archaeological work in the Landes region are closely related to the latter. Although discontinuous, these records have the advantage of being close to each other and of having been deposited in a similar geomorphological context, i.e. Pleistocene coversands. Also dated to MIS 3, they allow for comparison with the Saint-Vincent-de-Paul record.

25Two sites near the town of Cestas (fig. 1) have been analyzed and dated. In both sites, the peat levels are cryoturbated and developed in interdune depressions in connection with the rise of the groundwater table. At Barp-Pot aux Pins, the peat was OSL-dated between 50 ± 3 ka and 55 ± 3 ka and correlated with GI-14 (Sitzia et al., 2012, 2015). Pollen assemblages are dominated by Pinus, ahead of Ericaceae, Poaceae and Myrica. The internal evolution of the peat level attests to a rise in woody species (Pinus, Myrica, Ericaceae) at the expense of Poaceae. The peat sample analyzed at Cestas-Les Pins de Jarry was ¹⁴C-dated on bulk sediment to 23,550 ± 160 BP (27,330-28,020 cal BP) and 22,390 ± 120 BP (26,140-26,670 cal BP) (Bertran et al., 2009). Nevertheless, an OSL date on aeolian sand made 10 cm above the peat subsequently provided an age of 40.2 ± 2.7 ka (Bertran et al., 2020), which strongly suggests that the ¹⁴C ages were rejuvenated probably due to illuvial organic matter derived from the superficial podzol. Therefore, this level could correspond to a relatively old MIS 3 interstadial (GI-9 to GI-14). The single pollen assemblage obtained shows a pattern quite similar to the last phase seen at Saint-Vincent-de-Paul, and is dominated by Poaceae, Myrica, Ericaceae, Pinus and Betula. If the distribution of the different taxa is not completely similar (Ericaceae, Pinus, Betula and Poaceae are better perceived there than Myrica), it nevertheless refers to the same type of environment.

26Overall, the MIS 3 peat levels described in the Landes show the development of localized woodlands during various interstadials, mainly composed of cold-tolerant species typical of the boreal forest (Salix, Betula) associated with hygrophilous plants such as Myrica and Ericaceae. Pinus, more or less abundantly represented in pollen assemblages, is probably present only at great distance from the sites (cf. Cheddadi et al., 2006). It is only known in the region as macroremains (charcoal) during the Alleröd interstadial in palaeosols preserved beneath Younger Dryas dunes (Bertran et al., 2009; Sitzia et al., 2015). Better representation of Pinus during GI-14 than GI-5, which suggests greater regional abundance, is in line with the Bay of Biscay record (Sanchez-Goñi et al., 2008; Fletcher et al., 2010). The absence of thermophilous tree taxa such as Quercus and Fagus is also noteworthy. The single occurrence of a Fagus pollen grain in the basal peat sample cannot lend support to the debate about the presence of beech stands in the Landes during MIS 3 proposed from genetic studies (de Lafontaine et al., 2014).

27The interstadial vegetation dynamics recorded in southwest France do not differ significantly from those seen elsewhere in lowland western Europe, with the development of a few trees (Salix, Betula, Juniperus) in the context of an open vegetation dominated by Poaceae and Cyperaceae (Guiter et al. 2003; Fletcher et al., 2010; Kern et al., 2022). Comparison with the long French continental records (La Grande Pile, Les Echets, Velay maars, Sost, Barbazan, Lourdes) is not always easy because of the low resolution and poor chronological control that limit correlations to the various MIS 3 interstadials. The Velay sequences (Lac du Bouchet, Ribains), geographically closest but located at much higher elevation, reflect punctual extensions of Pinus and Picea into an open Betula and Salix boreal forest during the Upper Pleniglacial, which were attributed to long-distance transport or to older pollen stocks reworked by erosion (Reille and de Beaulieu 1988, 1990; de Beaulieu and Reille, 1992a). In the French Pyrenees, the Barbazan, Sost and Lourdes pollen sequences (Andrieu-Ponel et al., 1988; Jalut et al., 1992; Jalut et al., 1998) attest to the occurrence of Picea up to ca. 26 ka cal BP together with few mesothermophilous trees (mainly Quercus and Corylus). By contrast, the El Portelet sequence in the drier Spanish Pyrenees only records the presence of Pinus and Juniperus, while the few pollen grains from mesothermophilous trees were thought to reflect long-distance transport (Gonzalez-Sampeniz et al., 2006). At Bergsee (southwest Germany), sharp increases in Betula and Pinus are recorded, accompanied by Juniperus that evoke the development of open boreal forests (Duprat-Oualid et al., 2017). Pine played a notable role there, which has already been observed at La Grande Pile during the “Grand Bois Complex” correlated to the end of MIS 3 (de Beaulieu and Reille, 1992a). Recent reanalysis of the Füramoos sequence (southern Germany) allows better correlation with the late MIS 3 interstadials, which were marked by rapid tree extensions including Betula, Pinus and sporadically Picea. However, vegetation remained dominated by Cyperaceae and Poaceae grasslands and sagebrush steppe (Kern et al., 2022).

28Finally, late MIS 3 pollen assemblages from core MD04-2845, located offshore but not far from the Landes, indicate a development of deciduous oaks (Quercus deciduous type) together with boreal forest (Pinus, Abies, Picea, Betula) during the interstadials (Sanchez-Goñi et al., 2008; Fletcher et al., 2010), in sharp contrast to the Landes records. This obviously reflects a regional signal that includes refuge areas in the valleys of the Pyrenees where thermophilous trees were able to survive during the Glacial, and possibly in the Massif Central (although thermophilous trees are not recorded in the available pollen sequences). These also clearly demonstrate the specific character of the Landes vegetation, largely related to the acidic nature of the soils developed on aeolian sand.

29The development of peat in the Landes region, although rarely preserved presumably due to deflation, strongly suggests that the aeolian dynamics was only intermittently active during the Last Glacial, or at least that sand transport was spatially restricted to a few favorable areas during the interstadials. Concurrently, overland flow and gully incision were active. The gullies, later plugged by new inputs of aeolian sand, provide favorable contexts for the preservation of peat. The discovery of peat levels in a similar context at Saint-Paul-les-Dax, dated to 28,230 ± 140 BP (32,480 - 31,530 cal BP), i.e. GI-5.2 (Béague, 2014), and at Pontonx, dated to 32,410 ± 450 BP (37,800 - 35,300 cal BP), i.e. GI-8 (Bertran, unpublished data), allows us to anticipate the possibility that other interstadials may be documented in the future thanks to preventive archaeological work in the region.

6 - Conclusion

30The peat level discovered during preventive archaeological work at Saint-Vincent-de-Paul in the Landes region (southwestern France) documents the vegetation of interstadial GI-5.1, ca. 31 ka cal BP (or possibly GI-5.2). This level was preserved at the bottom of a gully incised in aeolian sands. During the early peat formation, a reed bed associated with willows developed in an open sagebrush steppe landscape. The gully was then colonized by birch trees, accompanied by willows, Myrica, Calluna and Erica cf. tetralix. Sphagnum peat finally formed. Comparison with other analyses of MIS 3 peat in the region shows that patches of boreal forest developed locally repeatedly during the interstadials. It shows as well that aeolian activity was discontinuous and probably restricted to spatially limited favorable areas during these periods. The sporadic discovery of other peat levels in similar contexts also dated to MIS 3 suggests the possibility of documenting other interstadials in the future if careful attention is paid to these finds.