Part of the radiocarbon dating was funded by the CNRS «Projets Exploratoires Premier Soutien» project and by the CNRS ARTEMIS radiocarbon dating programme.
The authors would like to thank Alain Queffelec (PACEA) for the Raman spectroscopy analysis of the gypsum rosettes, P. Antoine and N. Limondin for the data of the PA2000 test pit and the Languevoisin log. They also thank the three reviewers for their many comments and remarks, which have greatly improved the manuscript.

1 - Introduction

1Until recently, northern France yielded very few sites attributable to the early Upper Palaeolithic. Only a few Aurignacian artefacts discovered in the fields and a handful of sites attested to human presence in the region during this period (Fagnart et al., 2013). Among the latter, the Renancourt district, located west of the city of Amiens, has been known in archaeological literature since the early 20^th century thanks to the work of Victor Commont carried out in the «ancienne briqueterie Devalois”, i.e., former Devalois brickyard (Commont, 1913).

2The scarcity of these clues has led to the assumption that the lœss region of northern France remained inhospitable during the late Middle Pleniglacial and the entire Upper Pleniglacial periods, with human groups only returning to the region on a permanent basis during the Late glacial period (Antoine et al., 2003a).

3However, over the past fifteen years or so, mainly in the context of preventive archaeology, a number of discoveries have modified this relatively simplistic view. Indeed, the excavation of new open-air Gravettian sites in lœss contexts has made it possible to study in detail the evolution of sedimentation and palaeoenvironmental conditions within a significantly improved chronological framework, notably at Havrincourt (Antoine et al., 2014) and Amiens-Renancourt 1 (Paris et al., 2013b, 2017, 2019; Moine et al., 2021). Sedimentological, palaeontological and malacological analyses have also provided details on the palaeoenvironmental conditions during the occupation phases, making it possible to distinguish between stadial and interstadial phases. The issue of population dynamics in this periglacial region during the Weichselian Pleniglacial period has therefore been raised once again (Paris, 2020, 2024).

4The site of Amiens-Renancourt 2, like the site of Amiens-Renancourt 1, is also significant in this regard. These two sites, hereafter simply referred to as Renancourt 1 and Renancourt 2, excavated under the direction of C. Paris (Inrap), were discovered in situ in lœss sequences. The Renancourt 1 site yielded remains attributed to the recent-late Gravettian period (Paris et al., 2013b, 2017, 2019, 2021) and is more recent than Renancourt 2 being closer to an earlier phase of the Gravettian period (Paris, 2015, 2020, 2024; Paris et al., 2019).

5The key to understanding population dynamics is therefore the ability to place human settlements within a precise chronological and palaeoenvironmental framework. In this regard, the Renancourt 2 site provides a lœss sequence and associated multi-proxy records that cover the end of the Middle Pleniglacial and a part of the Upper Pleniglacial. If the latter is well documented in northern France and Western Europe by numerous lœss sequences and several high-resolution records, the former is much rarely documented. During the interval immediately preceding the Weichselian Middle-Upper Pleniglacial transition, between 36 and 31 ka, the rate of lœss sedimentation in northern France (e.g., Vercelot et al., 2025) and in Western Europe in general (Bosq et al., 2023) is much lower than during the Upper Pleniglacial and the lœss sequences are much thinner (Antoine et al., 2016).

6As a result, clear alternations between pedogenic horizons and lœss units for the late Middle Pleniglacial are rarely preserved and only clearly expressed in neighbouring regions, i.e., in Belgium at Harmignies (Haesaerts & Van Vliet-Lanoë, 1973), Maisières-Canal (Haesaerts, 1974), Rocourt (Haesaerts et al., 1981; Haesaerts et al., 2011) and Remicourt (Haesaerts et al., 1997), in the Middle Rhine Valley at Remagen-Schwalbenberg, near Bonn (Schirmer, 1990, 2012; Fischer et al., 2021; Vinnepand et al., 2020, 2023) and in the Upper Rhine Valley at Nussloch, near Heidelberg (Antoine et al., 2001, 2009). Radiocarbon chronologies and correlations with Greenland climate records are available for Nussloch (Moine et al., 2017), Remagen (Frechen & Schirmer, 2011; Prud’homme et al., 2022) and partially from some Belgian sequences (Frechen et al., 2001; Jacobi et al., 2010; Haesaerts et al., 2011, 2016). Paleoenvironmental reconstructions based on molluscan analyses are available at high resolution only for Nussloch (Moine et al., 2008), at lower resolution for Remagen-Schwalbenberg (Schiermeyer, 2000) and punctually for Maisières-Canal (de Coninck, 1973). Furthermore, among these sites, only Maisières-Canal (de Heinzelin, 1973; Haesaerts & de Heinzelin, 1979) and Remagen-Schwalbenberg (App et al., 1987, 1995; Fischer et al., 2021) yielded archaeological occupations.

7In northern France, though numerous stratigraphic sections have been compiled over decades and integrated to build a synthetic pedostratigraphic record (Antoine et al., 2016), the end of the Middle Pleniglacial remains poorly detailed and represented by the upper part of the Saint-Acheul/Villiers-Adam (SAVA) soil complex and the immediately overlying gley horizon (Vercelot et al., 2025). For this period, only four sites benefit luminescence and/or radiocarbon chronologies: Villiers-Adam (Antoine et al., 2003b; Locht et al., 2003), Havrincourt (Antoine et al., 2014), Morcourt (Sambourg et al., 2025) and Haynecourt (Vercelot et al., 2025). However, whatever their topographical context, plateau or slope, their chronologies revealed low sedimentation rates and hiatuses. These characteristics partly, when not completely, hamper to correlate the stratigraphic units with Greenland stadials and interstadials as precisely as in surrounding regions. Moreover, the SAVA soil complex is in most cases entirely decarbonated due to pedogenesis, except the uppermost decimetres. Only the overlying gley horizon is well carbonated. As a result, only three high resolution molluscan records exist for these deposits at Villiers-Adam (Limondin-Lozouet & Gauthier, 2003), Havrincourt 1 (Moine, 2012) and Morcourt (Sambourg et al., 2025) but the first one is very short, the second one is biased by decarbonatation and the third has only been partially studied with a non-definitive chronology. Consequently, they can presently not help constraining correlations with Greenland climate records. In northern France, the end of the Middle Pleniglacial is thus still poorly documented from a palaeoenvironmental point of view as well as the context of the rare contemporaneous archaeological layers (Antoine et al., 2014).

8The aim of this article is therefore to present the results of chronostratigraphic and palaeoenvironmental studies carried out on the Renancourt 2 site, in particular a detailed malacological record supported by radiocarbon dates. The main archaeological elements will also be presented. These studies allow to deduce implications for the age and environmental context of the lœss sequence and the associated Gravettian occupation. Correlations with Western European stratigraphic sequences, global climate records, and other archaeological sites are then discussed.

2 – Context

2.1 – Geological and geomorphological context

9The Amiens region is located in the northern half of the Paris Basin. Its relief is generally shaped in white chalk containing flints (Coniacian and Santonian). Remnants of Clay-with-flints and Tertiary formations (clays, sands, “avellanaires” or “Sinceny” pebbles) may persist on the heights (Dupuis et al., 1972).

10Amiens and the neighbouring cities are notable for the presence of a series of alluvial terraces resulting from the gradual incision of the valleys of the Somme and its tributaries (Selle, Avre). The relief created by this fluvial activity has been covered by slope deposits and lœss, which constitute the cover sequence (Antoine et al., 2020).

11These terraces were studied as early as the 19^th century, in particular by V. Commont (1909, 1913). By mapping of these terraces, now largely hidden by urbanisation, ten stepped terraces could be distinguished (Antoine, 1990). Some of the alluvial or cover deposits have been dated, making it possible to establish a reference sequence for river development during the Pleistocene (Antoine et al., 2007, 2021).

12Between 2007 and 2011, a series of archaeological surveys were carried out over the 50 hectares of the Renancourt Development Zone (ZAC) south-west of Amiens, totalling more than 150 test pits (measuring approximately 8-12 square meters, dug using a mechanical shovel). These are in addition to 48 test pits carried out in 1996 and 1997 on the motorway slip road, known as the «Pénétrante Ouest», connecting the A16 to Renancourt (Fagnart & Coudret, 1996, 1997). A large part of this area is occupied by Pleistocene alluvial terraces (fig. 1).

13Overall, the long slope where the Renancourt ZAC is located is highly eroded with a regularised topography, without any marked breaks in slope between the terraces. Nevertheless, during the 2007 survey of the upper part of the slope, at least four terraces were identified between 53 and 32 m a.s.l. (fig. 1): the Saveuse, Renancourt, Fréville and Garenne terraces (Antoine et al., 2007). The presence of the Epinette terrace is assumed under the lœss cover bordering the Rue Haute des Champs (Duvette et al., 2007). At the bottom of the slope, a very low terrace (Etouvie formation) has also been identified. The lœss cover thickens at the bottom of the slopes, both to the south-east towards the Selle valley and to the north towards the Grâce Valley, a dry, deeply incised little valley.

14Some lithic artefacts dating back to the Middle Palaeolithic have been discovered in the Renancourt ZAC archaeological surveys, mainly in the lœss sequences. The two Gravettian sites were discovered in test pits and trenches at the north-eastern end of the area, near the confluence of the Selle, the Grâce Valley and the Somme.

2.2 – Local topography of the bedrock

15Modelling of the chalky substrate top across the entire area explored in the test pits reveals the glacis supporting the alluvial terraces and clearly shows a chalky spur separating the Selle valley from the Grâce valley (fig. 2). The Renancourt 1 and Renancourt 2 sites are located at the bottom of the hillslope, the former on the slope descending towards the Grâce valley, the second towards the Selle valley, at the foot of this chalky spur. This spur forms a natural protection favourable to the establishment and preservation of human settlements (Paris et al., 2017, 2019).

C. Font

16On the Selle Valley side, this chalky spur partly corresponds to the scarp bordering the terrace, parallel to the valley (south-west/north-east, fig. 1). At the Renancourt 2 site, the chalky substrate was encountered several times during diagnostic surveys (Paris & Coutard, 2010; Locht et al., 2011) and in the 2011 and 2012 excavations (Paris et al., 2019). This is an unstructured chalk, probably nearby to the chalk in situ, found at around 18 to 22 m above sea level, beneath the alluvial deposits. It was identified during the excavation at the northern corner of the excavation at around 25.50 m a.s.l. (section 2012-5), at the bottom of the chalky scarp bordering the terrace.

3 – Materials and methods

3.1 – Excavation method

17The Renancourt 2 site was excavated under the direction of C. Paris (Inrap) during two preventive archaeological campaigns in 2011 and 2012 over a total area of 1,800 m² (fig. 3). Test pits in the surrounding area also revealed the existence of the sedimentary unit containing the archaeological level and yielded some remains. The settlement is therefore potentially preserved over an area of approximately 5,000 m² (red dotted line in fig. 2). This unexcavated part has not been directly affected by recent buildings.

18Given the general organisation of the remains, grouped in small scattered concentrations, mechanised excavation was preferred in order to detect the denser areas to be excavated manually. A total of 217 m² were excavated by hand in five sectors. All the remains were measured in three dimensions. Thanks to the clear recognition of the pieces in the sediment, sieving was limited to very dense areas such as flint knapping waste heap.

3.2 – Sampling design and analytic choices

19The study of the chronostratigraphic and palaeoenvironmental context of human occupation at the Renancourt 2 site is based on the analysis of the succession of litho- and pedostratigraphic units identified in the field.

20Seven sections were dug during the 2011 excavation (sections 2011-1 to 2011-7) and eight during the 2012 excavation (2012-1 to 2012-8, fig. 3). They were drawn in detail at a scale of 1:20 and/or photographed with a photographic survey corrected using spatial coordinates. The location of the 2011 sections was constrained by the existence of two former brickyards that had locally truncated the lœss sequences. The 2012 sections are mainly located on the walls bordering the excavated area (fig. 3). In this paper, we selected the six most informative sections to reconstruct a synthetic record for the site taking into account lateral facies variations over the excavated area (fig. 3). In the northern part, sections 2012-5 and 2011-6 (Paris et al., 2013a) provides a long profile allowing for the topographical and geometric relationships across the slope to be perceived despite the impact of the brickyards. They are accompanied by section 2011-1, perpendicularly connected to section 2011-6, and which crossed an area richer in lithic artefacts. Section 2012-4, located in the middle of the western wall, a few metres from a flint knapping waste pile, has been defined as the reference profile because of the thickness of the lœssic sequence and the link with the occupation level. Section 2011-3 is located downslope at the north-eastern edge, in the embankment separating two former brickyards. Finally, section 2012-1 supplemented by a test pit at its foot is located in the southern edge, which is presently wetter and shows more hydromorphic sedimentary facies.

21Sedimentological, malacological and micromorpho-logical samples were taken in parallel from sections 2011-1, 2011-3, 2012-1 and 2012-4. Not all of them have been studied (selection based on scientific interest, time and resources; see tab. 1 for details). The reference profile 2012-4 was surveyed at a scale of 1:10, dated and sampled in continuous columns for high-resolution sedi-mentary, malacological and micromorphological studies (fig. 4). Priority was given to the comprehensive study of its lower half to contextualize the archaeological material found nearby.

Sylvie Coutard

Sylvie Coutard

22Due to the time-consuming character of sorting, the whole molluscan analysis of section 2012-1 was not carried out due a high risk of redundancy with results of section 2012-4. Only three molluscan samples from the human occupation horizon and bracketing units were analysed in order to characterise a suspected spatial micro-environmental gradient at the site level. Section 2012-1 was also subjected to luminescence dating. As for sections 2011-1 and 2011-3, due to difficult-to-read stratigraphy and a sampling volume of only 5 litres per sample, which was initially considered too small as twice lower than the traditional sampling volume for terrestrial mollusc analyses (Puisségur, 1987), their study was put on hold before being replaced by that of section 2012-4.

3.3 – Micromorphology

23Sediment blocks were continuously sampled for micromorphological study in three of the sections. They document the lower half of section 2012-4 (fig. 5), the whole truncated section 2011-1 (fig. 6) and section 2011-3 (fig. 7) almost entirely, but for this latter only two samples were selected to provide an overview of the two uppermost units. This set is completed by two isolated samples from 2012-4 taken in reddish bands ("225-235" and BR) intercalated in the laminated silts that constitute the upper half of the section (fig. 5). The majority of samples from the two continuous series analysed concern the unit including the Palaeolithic occupation level.

2427 slides were prepared using standard methods of drying, resin impregnation and mounting on slides with polishing to 25-30 µm at the Ghent Laboratory. Obser-vations were made using an Olympus BX51 microscope under plane-polarized light or with crossed polarizers, following the principles of Bullock et al. (1985) and Stoops et al. (2010).

25The pedo-sedimentary features observed at the microscopic scale allow the recognition of pedogenic processes, mainly using a qualitative to semi-quantitative approach (Stoops et al., 2010). The features are ranked and interpreted in terms of sedimentary and pedogenetic phases (Fedoroff & Courty, 2002; Gerasimova, 2003) and enable an approach to palaeoenvironments (climate and vegetation; e.g., Van Vliet-Lanoë, 1988).

3.4 – Sedimentary analyses

26The four types of sedimentary analyses have been performed on the same samples taken from profile 2012-4 weighing approximately 300 g.

3.4.1 – Magnetic susceptibility

27The mass magnetic susceptibility was measured using a Bartington MS 2B magnetic susceptibility meter.

28Sub-samples were dried in an oven at 45°C for three days, then crushed and sieved to remove the fraction larger than 2 mm before being weighed. For each sample, eight measurements were taken and a correction (± 1σ) was applied to eliminate the influence of any artificial bias due to the equipment during the series of measurements.

3.4.2 – Total organic carbon (TOC) content

29The total organic carbon content was measured using a Thermo Scientific FlashEA 1112 CHNS elemental analyser. Subsamples were ground and sieved to 200 µm. Portions of 20 mg were placed in small silver capsules and decalcified with 5% hydrochloric acid (HCl) to ensure the organic origin of the measured carbon content. The capsules were finally burned in the CHNS elemental analyser by flash combustion at 1800°C.

3.4.3 – Calcimetry

30The CaCO₃ content of the sediment was measured using a Bernard calcimeter. Subsamples of 500 mg were sieved to remove the fraction larger than 200 µm, then placed in a conical flask connected to a graduated tube and a bulb filled with salt-saturated water, which constituted the apparatus. Once placed in the device, 15 ml of hydrochloric acid was added and the solution was mixed with a magnetic stirrer for 3 minutes to ensure a complete reaction. Preliminary results - not presented - and observations under a binocular microscope showed the presence of several types of secondary CaCO₃ concretions_, particularly in the fraction larger than 100 µm. Below this size, primary calcium carbonates were dominant. A new set of subsamples was therefore analysed after sieving to 100 µm.

3.4.4 – Granulometry

31The particle size analysis was performed using a Beckman Coulter LS-230 laser diffraction particle size analyser. The particle size was measured on the bulk material without prior decalcification in order to preserve the primary carbonates. Subsamples of 20 mg were dispersed in 400 ml of deionised water, loaded with 0.4% sodium phosphorous pentoxide (NaP₂O₅) and stirred for 2 hours using a rotary stirrer. Subsequently, the sieving threshold, generally set at 160 µm - the upper limit for fine sand - had to be lowered to 100 µm to avoid bias related to the high secondary carbonate content above 100 µm, which is not linked to wind activity.

32The sieved sample was supplemented with 600 ml of deionised water and mixed continuously with a magnetic stirrer. 2 to 4 ml of this solution was subsampled with a pipette for analysis. Due to the small amount of solution analysed, two additional measurements were performed on every other sample as a control. Compared to the conventional sieving and Robinson pipette measurement method, the laser diffraction particle size analyser underestimates the proportion of clays due to the lamellar shape of clay minerals, which appear too large when measured perpendicular to their flat side. Consequently, the particle size class limits must be calibrated with the conventional limits obtained by the sieving/pipette method. Based on comparison tests carried out on lœss and palaeosol samples from the Havrincourt sequence located only 70 km ENE of Renancourt (Antoine et al., 2014), upper grain size class limits equivalent to those based on the sieving/pipette method (clays at 2 µm, fine silts at 20 µm and coarse silts at 61 µm) were defined at 6, 20 and 61 µm for the laser granulometer (Jamet, 2011). In this study, the fine sand fraction is therefore restricted to the 61-100 µm range. The fraction above 100 µm was dried and weighed.

33Finally, a grain size index (GSI) was calculated. This ratio describes the proportion between coarse silt (20-61 µm for the laser method applied) and fine silt plus clays (< 20 µm). In a pleniglacial lœss context, it allows us to distinguish between aeolian deposition phases characterised by massive lœss units (1 < GSI < 2) and pedogenesis phases marked by the development of gley horizons (0.5 < GSI < 1) (Antoine, et al., 2009, 2014). The Coarse silt index (CSI; Schirmer, 2016), i.e., the ratio coarse silts/fine silts has also been used to account for clay content.

3.5 – Malacology

34The samples were sieved with water on a 500 μm mesh, then dried. Due to the exceptional abundance of most samples of the sequence, only a quarter of the sieve refusals was selected for analysis according to the method of Puisségur (1976). The sieve reject from each sample is spread out on a disc and divided into four equal quarters. After removing two opposite quarters, the two remaining quarters are re-mixed and the process of dividing and removing two opposite quarters is repeated. The shells were sorted under the binocular and identified using reference guides (Ložek, 1964; Kerney et al., 1983; Horsák et al., 2013). The nomenclature of taxa is given according to Molluscabase (2026). In several samples, the material of some taxa was limited to small apices or shell fragments. As their taxonomic attributions remain unsure, they are preceded by the mention “cf.”. The method developed by Ložek (1964) and adopted by Puisségur (1976) was used to estimate the number of individuals in each taxon.

35In addition to malacological material, remains of other bioindicators were extracted from these samples, fragments of microfauna bones and teeth and sporadically occurring ostracod valves.

3.6 – Chronology

3.6.1 – Luminescence

36Optically stimulated luminescence (OSL) dating on quartz was carried out at the Luminescence Dating Laboratory in Oxford (United Kingdom) by J.-L. Schwenninger. Four samples were taken, three from section 2012-4 (fig. 5) and one from section 2012-1 (fig. 8), to date the layer containing the archaeological level and two of the overlying layers.

37The samples were collected without exposure to light using 50 x 200 mm copper tubes. These tubes were inserted horizontally into the sediment with a hammer and then sealed at both ends with adhesive tape. To determine the annual dose and water content, approximately 1.5 kg of sediment was collected in hermetically sealed bags, so as to average the composition over a radius of 30 cm around each sampling point for OSL.

3.6.2 – Radiocarbon

38Six radiocarbon (¹⁴C) dates were obtained in three laboratories on different bones from several taxa from the archaeological level (tab. 1, fig. 3). Item no. 702 (horse metatarsal) was found in an isolated area where stratigraphic distinction was difficult. Nevertheless, its state of preservation and surface are similar to the rest of the faunal corpus, including the horse. Its probable connection to the main archaeological horizon is presented in the discussion.

39To reinforce the chronology of the sedimentary sequence, particularly the units flanking the one that yielded the archaeological level, radiocarbon dating was carried out on earthworm granules (GVT) taken from four malacological samples from section 2012-4 (fig. 5). The samples were chosen to date the peak of the four increases in mollusc abundance that mark the sedimentary sequence. For each of them, 50 granules with a diameter of 0.8 to 1 mm were collected, providing between 58.8 and 69 mg of calcium carbonate. The granules were selected and prepared following the protocol established by Moine et al. (2017). The measurements were carried out at the Carbon-14 Measurement Laboratory in Saclay (France).

4 – Results

4.1 – Lithostratigraphy

40The main sections (including those from which samples were taken) are presented in detail in figs. 5 to 9 and tab. 2. The exploitation of sediments by former brickyards over a large part of the excavated area explains the absence of the upper horizons of the stratigraphic sequence, namely the decarbonated loams and the Bt horizon of the leached brown surface soil, on most of the profiles described. For ease of reading, the numbering of the units follows that of the synthetic sequence.

4.1.1 – Section 2012-4: reference profile

Sylvie Coutard

4.1.2 – Section 2011-1

4.1.3 – Section 2011-3

4.1.4 – Section 2012-1

Sylvie Coutard

4.1.5 – Sections 2011-6 and 2012-5

41Section 2012-5 is located in the continuation of section 2011-6 (figs. 3 & 9). The chalky slope is visible to the north-west.

4.2 – Synthetic sequence and geometry of deposits

42A synthetic sequence of about 9.5 m in thickness is proposed (fig. 10), based on field observations (tab. 2), stratigraphic correlations between sections and surrounding boreholes, as well as micromorphological (figs. 11 & 12) and sedimentary (fig. 13) analysis. 14 pedo-sedimentary units - named by their numbers in brackets ([#]) - have been distinguished. They are distributed in three main subsequences, which are composed from the base to the top of (i) 2 m of alluvial layers [2-3], (ii) 3 m of slope deposits [4-6] and (iii) 4 m of lœssic deposits [7-14].

4.2.1 – The alluvial subsequence

43The alluvial subsequence [2-3] has been identified in numerous test pits directly above the substratum [1] over a much larger area than the archaeological excavation. Facies variations have only been noticed for the fine alluvium [3] (figs. 8 & 10).

4.2.2 – The slope deposits subsequence

44The coarse slope deposit subsequence [4-6] covered, and sometimes eroded, the alluvial subsequence and smoothed the topography. Its elements come from the dismantling, during periods of intense frost, of chalky scarps and of one or more older alluvial sheets located upslope. In the northern corner of the excavation (fig. 3) close to the chalky scarp, section 2012-5 (fig. 9) shows several phases of boulder deposits of limited extent. Downslope to the east, test-pits 10 and 16 and sections 2011-2 to 7 exhibit a silty-sandy layer [5] separating two phases of coarse slope deposits [4 & 6]. To the south, excavation stopped at the top of the more or less consolidated deposits of unit [6], except in the test pit 2012-1 where a pocket of fine sediments is merged in a 60 cm-thick coarse deposit (fig. 8). The erosion and the reworking and downbending of the unit [5] towards the east (fig. 9) result from the successive deposition and solifluction of the subsequent much coarser unit [6]. The geometry of these deposits along the slope (fig. 9) precludes any equivalence between [5] and any units of the overlying lœssic subsequence.

4.2.3 – The lœssic subsequence

45The lœssic subsequence begins with a gleyic horizon mainly present to the south of the excavation. Its thickness is about 40 cm in section 2012-4. In the south corner of the excavated area, [7] is 70 cm-thick, at the bottom of section 2012-1 at lower elevation, i.e., around 23.8 m a.s.l. (fig. 8). Its hydromorphic character is particularly well marked there, indicating a significantly wetter area.

46Unit [8], which yielded all faunal and archaeological remains, has been recognised over the entire excavated area and beyond, with lithic industry also present (figs. 1 & 2). It is absent or very reduced in section 2012-7 a few metres southwest of section 2012-4 due to cryoinjections affecting the underlying slope deposit, and to the north in section 2012-5, where correlations with section 2012-4 are uncertain. Apparently, the thickness of [8] increases downslope from only 30-40 cm in section 2012-4 to more than 70 cm in section 2011-1 to the east, but the distinction of [8] from bracketing units can be difficult in some sectors. Farther east, outside the excavation area, has been locally eroded by the deposition of [11] or removed by brickworks activity.

47Unit [9] is a silty deposit characterised by hydromorphic and cryogenic features. About 40 cm thick in the south (fig. 8), it is now only about 20 cm thick in the section 2011-3 (fig. 7).

48Unit [10] is a silty sequence composed of three sub-units [10a, 10b, 10c], with often diffuse boundaries. This division, particularly visible in section 2012-4 (fig. 5) is not as clear in other sections. The stratified sub-unit [10c] is often the only one or the thickest one (up to 50 cm in section 2012-1; fig. 8), suggesting an unstable transition with erosional processes between more homogeneous units [9] and [11].

49Units [11-13], approximately 2 metres thick, covered the whole area and shaped the slope with a gentle incline. Contact with the underlying units is erosive and marked by a regular gravel of chalk and flint granules at the base of [11]. On top, both decarbonated lœss [12] and leached brown surface soil [13] are essentially preserved in the north-western corner of the site. Elsewhere, they have been extensively exploited as «brick clay».

4.3 – Micromorphology

4.3.1 – Sandy deposits in unit [5]

50Unit [5] is present in blocks "110-100", "100-90" and at the base of "90-80" sampled from section 2011-1. It is located stratigraphically beneath the coarse slope deposit [6] that slopes towards the south-west of the section.

51In thin sections, the sediment appears poorly sorted. A silty fraction, probably of lœss origin, is mixed with a coarse local fraction (chalk and flint gravel, sand consisting of chalk grains, etc.). A platy structure resulting from the action of freeze-thaw cycles (Van Vliet-Lanoë, 2010) is clearly visible, especially at the base of sample "100-90" (fig. 12-A) and in sample "110-100". This structure fades towards the top of the 2011-1 "100-90" thin section, where it gives way to small aggregates separated by small horizontal or curved fissures. The layer is also affected by slight hydromorphy, marked by some oxidation and blackish manganese impregnations. Carbonate features are rare. The poorly sorted nature and the clear bedding observed upward in sample "90-80" indicate that it was formed by slope dynamics. The presence of silty aggregates surrounded by a cap in the coarse beds of this sub-unit highlights the role of gelifluction. This unit is clearly part of the slope deposits sub-sequence.

Sylvie Coutard

4.3.2 – Unit [8] or Grey-brown silt with concretions

52This unit is represented in thin sections 2011-1 "100-90" to "10-0" (fig. 6) and 2012-4 "20-30" to "50-60" (fig. 5). Including the archaeological level, it has been the subject of a special characterisation effort (fig. 12-B to J).
Sediment characteristics
This is a quartz-rich lœssic silt rich in rounded micritic pellets, also containing angular sparitic grains (sedimentary origin). The matrix is clay-carbonate. Scattered chalk granules are present.
Biological activity
Two main types of features are associated with marked biological activity:
- rounded or elongated cavities, often partially filled with small, poorly defined aggregates with irregular edges, loosely assembled, posterior to the freeze-thaw microstructure;
- sub-rounded aggregates, sometimes large (up to 2.5 mm in diameter), arranged in small groups in rounded or elongated spaces that may correspond to bioturbation, prior to the freeze-thaw microstructure (fig. 12, C and J). The presence of small, discrete caps at the top of these aggregates and not all around them implies limited rotation. The large aggregates are sometimes cut or reorganised, flattened, by the action of cryoturbation.
Furthermore, in thin sections 2011-1 "20-10" and "10-0", i.e., at the top of [8] and in the overlying gleyified unit [9], the biological activity of enchytraeids is more developed, marked by small pellets accumulated in cavities. It can reflect a change in palaeoenvironmental conditions.
Finally, there are a few calcite biospheroids, or spherulites, excreted by earthworms in the section 2012-4 and to a lesser extent in 2011-1.
. Action of frost
The action of freeze-thaw cycles (cryogenic features) is marked by a distinct lamellar to lenticular microstructure that fades with depth. Its formation is undoubtedly associated with the establishment of the overlying unit [9].
. Aggregates and solifluction

53This cryogenic microstructure sometimes overlaps with a microstructure that could be described as collapsed spongy, when numerous small (less than 100 µm in diameter) ovoid aggregates are visible. This could indicate formation by solifluction over short distances.

54Carbonate features

55Carbonate pedological features (Durand et al., 2010) are abundant. Several types of secondary carbonates (fig. 12-B, D, E & G) are present:
- hypocoatings: accumulations of carbonates around channels in the sediment mass;
- infillings or root sleeves: filling of root channels by secondary carbonates;
- root pseudomorphs (calcified root cells): replace-ment of cells by sparite crystals. These are thought to be contemporary with the life of the root or its decomposi-tion (Barta, 2011), but these roots may be posterior to the sediment;
- needle calcite, linked to mycorrhizal activity;
- lœss dolls with a hollow centre (fig. 12-B).
Hypocoatings and root sleeves are very abundant in the archaeological level: hypocoatings correspond to the abundant small beige branched concretions found in sieve refusals. The origin of hypocoatings and infillings is still discussed. It could be connected to root metabolism or to carbonate precipitation due to evaporation of calcium rich-solutions (Becze-Deák, 1997; Barta, 2014).
According to Barta (2011), hypocoatings form at the same time as the sediment deposit, under the influence of vegetation.
However, it is clear here that secondary carbonate channels (hypocoatings or infillings) are posterior to the freeze-thaw structure and are therefore not at all contemporary with the deposit. They probably form by leaching and percolation under the influence of a vegetation associated with a topsoil that can be several tens of centimetres higher in the sequence.
. Gypsum
Numerous gypsum crystals were observed in the thin sections of section 2011-1 (fig. 12-F, H & I) but not in those of section 2012-4. These are rosettes or more elongated patches, sometimes reaching several millimetres in length, consisting of elongated rhombic to lenticular crystals, transparent in PPL and grey in colour in XPL (low first-order birefringence, like quartz). The identification under the microscope was confirmed by Raman spectroscopy (A. Queffelec, PACEA, Bordeaux). Such gypsum crystals have never been observed in the lœss sequences of northern France. Gypsum is a hydrated calcium sulphate (CaSO₄, 2 H₂O) and the origin of the sulphates raises questions here. Indeed, the main possible sources appear to be ruled out here: artificial origin, evaporitic deposits (like in Mediterranean or Central Europe and Central Asia lœss deposits; Becze-Deák et al., 1997; Mestdagh et al., 1999; Boixadera et al., 2015), alteration of pyrite (FeS₂) present in Paleocene glauconitic sands (e.g., alluvial eemian deposits of Waziers, northern France; Deschodt et al., 2022), organic matter linked to archaeological occupation (documented in specific confined contexts; Bergadà et al., 2015). Nevertheless, regardless of the source of the sulphur, gypsum crystals in lœssic deposits are evidence of a marked arid phase following the deposit.
Could this be a common phenomenon that has been exceptionally well preserved here due to local circumstances that have limited leaching?

4.3.3 – Unit [9]

56Unit [9] is represented in thin sections Cpe 2011-1 0-10 (fig. 6) and Cpe 2012-4 50-60 to 80-90 (fig. 5). Unit [9] is a sediment of the same type as the underlying silt with concretions. Large aggregates are only present at the base. The microstructure is lenticular to granular (Van Vliet-Lanoë, 2010), with the lenses becoming gelifluction ovoids. Caps, i.e., accumulations of fine particles, can be observed on top of the aggregates.

57A lœss doll at the base of the thin section confirms the observations made on the section. These lœss dolls seem to be associated with leaching in a gleyified horizon with hydromorphic processes. Small oxidised aggregates (fig. 12-K), rounded millimetre-sized ferruginous concretions and small ferro-manganese impregnations also indicate gleyification, probably associated with a perched water-table.

58The dip of the lenticular microstructure in section 2012-4 "60-70" is undoubtedly the result of cryoturbation in involutions of the overlying unit [9].

4.3.4 – Unit [10a, b, c] in section 2012-4

59This part of the sedimentary sequence is represented in thin sections 2012-4 "80-90" to "125-135" (fig. 5).
In sub-unit [10a], large aggregates predating the freeze-thaw structure are found, as in unit [8].
Two small hydromorphic phases associated with freeze-thaw cycles can be distinguished in thin sections, one at the top of [10a], marked by caps on the aggregates, the development of a lamellar microstructure and small oxidised ferruginous concretions scattered throughout the sediment. The other at the top of [10c], where there are some traces of freeze-thaw microstructure, associated with some ferro-manganese impregnations.

60. The "queras"
A very marked pedological feature is found in the thin sections from these layers. These are cytomorphic sparite crystals filling a small channel surrounded by a halo of orange-coloured sediment due to decarbonatation. This is a calcification-decalcification feature associated with the impregnation of root tissue frequently observed in semi-arid calcareous soils. These features have been described in detail by Herrero & Porta (1987) in Spain in gypsum soils.

61The sizes of the features observed here correspond to those described by Herrero and Porta (1987): channel
> 2 cm long and 1-2 mm wide, filled with equigranular sparite crystals (60-90 µm), surrounded by a decarbonated hypocoating, often more developed on one side, in a carbonate matrix. This is indeed a decarbonated aureole and not a channel filled with another sediment, as there are cases of rounded aggregates decarbonated on one half. These features differ from simple root pseudomorphs (calcified root cells), which are quite common in lœss (Barta, 2011) but are not associated with decarbonatation.

62Recent work in lœss from Ebro Basin support the hypothesis that queras originate from the calcification of root tips and are related to the acidification of surrounding soil for nutrient absorption. Radiocarbon dates highlight the gap between the age of queras and the age of the lœss where they are found in Alvarez et al. (2024).

4.3.5 – Unit [11], yellow laminated silt with reddish bands

63Samples 2012-4 "225-235" and "BR" focused on two reddish bands, one continuous and the other lenticular, interspersed in the laminated silts (fig. 5). The importance of reworked elements in the laminated silts is noteworthy, particularly in the reddish bands. The abundance of reworked brown-orange ferri-argilan fragments and pedorelics indicates the dismantling of a horizon of leached brown soil, located higher up the slope. There are also many chalk granules but very few flints. Biological activity is absent or almost absent, which is consistent with the virtual absence of molluscs in the samples.

4.4 – Sedimentology

64The results of the sedimentological analyses carried out on the section 2012-4 from [7] to the basis of [11] are presented here in two figures: along the stratigraphic profile (fig. 13-A) and through a principal component analysis (PCA) (fig. 13-B). The first factorial plane explains 91.7% of the total variance of the dataset. The quasi-absence of overlap between the groups of samples of the different (sub)-units supports lithostratigraphic subdivisions of the sequence. In detail, only three samples can be questioned in the PCA. The plotting of a central sample of [9] (65:70 cm deep; #20) among those of [10a] may result from a downward injection of material from [10a] in [9] given its cryoturbated character. The plotting of the basal sample of [11] (135:140 cm deep; #6) close to those of [10c] reveals similar sedimentary characteristics though a visual aspect in section (fig. 4) that led to its attribution to [11]. The ex-centring of the basal sample of [10c] (120:125 cm deep; #9) close to those of [9] likely results from reworked material at the base of [10c] given its stratified nature.

65The variation ranges of all the proxies (TOC: 0.05-0.35%; CaCO₃: 10-25%; MS: very weak, between 7 and 10 x 10^-8 m³.kg^-1; GSI: 0.5-2.5; CSI: 0.8-4.5; grain sizeclasses proportions) are typical for regional lœss deposits (e.g., Antoine et al., 2014; Sambourg et al., 2025). Their combination allows for the distinction of three sedimentary phases.

66From the base of lower gley [7l] to the top of grey-brown loam [8], the TOC, though slightly decreasing, is maximum about 0.30-0.35%, underlining the maximum incorporation of organic matter by biological activity and the densest vegetation cover, and supplies from colluvial origin, which fits with a low eolian dynamics (GSI between 0.5 and 1, CSI between 0,8 and 2). The MS signal shows an important oscillation in [7] and [8l] marked by a minimum at their interface. The MS signal seems to follow the proportions in coarse silts, witch drop from 40 to 25%. Waterlogging marks are present but not very pronounced and do not indicate water saturation that could lead to significant magnetic impoverishment, the MS signal thus appears mostly driven by inputs of magnetic particles from colluvial or aeolian origin rather than pedogenic processes. These higher proportions in coarse silts of [7l] also clearly distinguish it from [7u] and [8l] in the PCA. The high and increasing values of CaCO₃ from 17 to 26 % most probably result from lateral supplies from the outcropping chalk scarp (sediment particles and dissolved carbonates).

67From the base of grey-brown silt [8u] to the top of the upper gley [9], the sedimentary dynamics moderately changes. [9] distinguishes from [8u] by slightly higher proportions in fine and coarse silts increasing from 25 to 35% and 35 to 40% respectively, and lower proportions in clays dropping from 30 to 20%, which reveals a decline in colluvial supplies and an enhanced eolian dynamics. The grain size proxies also show short abrupt oscillations of increasing amplitude that are absent from other sedimentary proxies. Given the weakness of pedogenic processes, the variations of SM may result once again from various sedimentary inputs. The CaCO₃ decrease can result from a relative decrease in lateral supplies and a slight decarbonation in the unit 9 (resulting in lœss dolls formation).

Sylvie Coutard

68From the base of the brownish-grey loams [10] to the top, the sedimentary dynamics is mostly driven by the eolian dynamics. Coarse silts and CSI exponentially increase from 35 to 60% and from 2 to 4.5 respectively, whereas clay and fine silts decrease from 30 to 10% each. Both TOC and CaCO₃ show decreasing trends to minimum values reached in [11].

69They reflect a constant decline in vegetation cover and topsoil stability reducing the incorporation of organic matter and an increase in the rate of eolian inputs, especially from the brownish loam [10b] upward. Nevertheless, in [10] and [11], the slight MS increase can result from the incorporation of pedorelics and reworked particles as seen in thin sections. The pedogenesis was very limited or even absent, particularly during the deposition of [11].

4.5 – Malacology

4.5.1 –  Isolated samples of fine alluvium with molluscs (unit [3])

70The test pit (SP) 14 sample yielded a very fragmented and sparse malacofauna (~200 individuals) (tab. 3). The terrestrial fraction is dominated by Pupilla muscorum, Trochulus hispidus, Pupilla alpicola and Succinea putris, accompanied by a few individuals of slugs, Succinella oblonga, Vallonia costata and Columella columella. The aquatic fraction, which was very small but relatively diverse, yielded Stagnicola sp., Euglesa obtusalis, Anisus leucostoma, Galba truncatula, cf. Peregriana peregra, Valvata piscinalis and Ancylus fluviatilis, as well as a few ostracod valves. Most of these taxa live mainly in low-energy aquatic environments, except for the last two, and especially Ancylus fluviatilis, which is a flowing water taxon (small streams to large rivers) absent from stagnant water categories. Some can even be found in purely lœssic series (e.g., Mazenot, 1963; Moine et al., 2005) because temporary pools throughout the year may be sufficient for them to complete their annual life cycle, particularly G. truncatula, which is amphibious (Welter-Schultes, 2012; Horsák et al., 2013). This molluscan assemblage reflects a glacial climate, potentially stadial (low total abundance), and an environmental mosaic of dry loci covered by a continuous steppe vegetation with shrubs, and wet loci marked by the presence of shallow stagnant water bodies fed, at least in part, during floods of the river bordering the site.

71The two samples from section 2012-1 (fig. 8) yielded similar malacofaunas but much more abundant – over 1,000 individuals each – and slightly less diverse in terrestrial taxa (absence of V. costata and C. columella) and slightly different in aquatic taxa (replacement of Stagnicola sp., G. truncatula and cf. P. peregra by Ampullaceana balthica, Armiger crista and Euglesa nitida) (tab. 3). The large predominance of A. leucostoma suggests smaller, less vegetated and temporary water bodies than for the SP14 sample, but also fed, at least in part, by the river, as wind transport of the shells of the rarest taxa from a nearby bank is unlikely. The terrestrial fraction, which is very much in the majority, consists of the predominant P. muscorum, accompanied by T. hispidus, P. alpicola, slugs, S. putris, and S. oblonga. The assemblage indicates a glacial context, potentially interstadial (high total abundances), with low-humidity steppe vegetation cover dotted with rare temporary water bodies. Sample Pr2, taken from gleyified silt overlying the sand that yielded sample Pr1, differs in that it contains significantly lower proportions of P. muscorum and T. hispidus and higher proportions of slugs, S. putris, and A. leucostoma, indicating the development of water bodies.

72These two malacofaunas indicate a probable fluvial origin for the grey silts and sands deposited between the coarse alluvial layer and the overlying slope deposits, and their formation during a cold phase at the end of the Saalian or Weichselian.

4.5.2 – Continuous recording of units [7] to [11] of section 2012-4

73The number of taxa per sample, or richness, varies between 4 and 6. However, most samples include 4 to 5 terrestrial taxa, to which are added 1 to 2 aquatic taxa among the 6 identified (tab. 2; fig. 14).

74The malacofauna is largely dominated by Pupilla muscorum and includes only Succinella oblonga, slugs and Trochulus hispidus as perennial taxa, the latter being very poorly represented except at the base of the section. Pupilla alpicola is only present in [10a]. Aquatic taxa appear only sporadically and with isolated individuals. Their transport by runoff from the plateau or by wind from the nearby valley is just as possible as their presence in in situ water bodies that could exist due to the low dip of the deposits and the irregularity of the paleosurface of the soil. The composition of the terrestrial malacofauna reflects an open environment characterised by a poorly diversified and mostly discontinuous herbaceous vegetation cover.

75In detail, however, variations in the abundance and proportions of taxa highlight cyclical changes in the environment from the base to the top of the section. The number of individuals per assemblage, or abundance, is much more variable and oscillates between 5 and 2141 (tab. 3). As the molluscan fauna composition shows little changes throughout the sequence (fig. 14), malacozones have been distinguished based on total abundance variations. This parameter has indeed been demonstrated as a major criterion to identify interstadial phases in high-resolution molluscan records combined with detailed pedostratigraphy and precise chronology in Upper Weichselian lœss sequences from western Europe (Moine et al., 2008, 2011, 2021).

76Malacozones (MZ) 1, 3 and 5 thus distinguish by very high abundances and MZ2, 4 and 6-7 by low to very low total abundances. The distinction of MZ5 is also supported by the presence of Pupilla alpicola and the distinction of MZ7 from MZ6 by the appearance of Vallonia costata. When necessary, MZ have been subdivided into subzones based on remarkable taxa abundance and proportion changes (fig. 14) as detailed below. Changes in the molluscan fauna throughout section 2012-4 reflect the following environmental evolution.

77The molluscan record thus identifies three phases of milder temperatures, i.e., MZ1, MZ3 and MZ5, presenting each a particular signature.
MZ1 ([7l]) distinguishes by high abundances of the mesophilic taxon Trochulus hispidus. Found in a wide range of wet to moderately dry habitats, i.e., floodplains, marshlands, wet grasslands, shrublands and sparse forests (Welter-Schultes, 2012), it is described as "relatively thermophilic" by Kerney (1963). Its proportions rarely exceed a few percent in north-western France (Antoine, 1989; Moine et al., 2011; Moine et al., 2021; Sambourg et al., 2025), southern England (Kerney, 1971; Preece, 1990) and the Channel Islands (Rousseau & Keen, 1989) Upper Weichselian molluscan records, when present (Moine, 2014). The increasing proportions of T. hispidus (~3 to 20%) underline a development phase of the shrub component in a moderately humid environment marked by relatively low proportions of hygrophilous taxa Succinella oblonga (~10%).
MZ3 ([8]-lower half of [9]) distinguishes by the predominance of P. muscorum (80-90%), which currently lives in dry, open and sunny environments (Welter-Schultes, 2012), and the poor representation of accessory species. The proportions of P. muscorum show a cyclical evolution with an increase during MZ3a, and then a decrease, very slight in MZ3b and more rapid in MZ3c, at the benefit of S. oblonga and slugs whose proportions increase from 5 to 30 and 10% respectively. A steppe-like vegetation settles under milder conditions in MZ3a and persists during MZ3b before evolving to a sparse herbaceous vegetation as the environment becomes slightly colder and more humid in MZ3c. The total abundance shows the same cyclical evolution but MZ3b distinguishes by an intermediate decrease visible in abundance curves of the three dominant species. The absence of synchronous change in taxa proportions raises the question of an eventual anthropic rather than climatic impact as these synchronous species abundance decreases take place within the vertical dispersal zone of the archaeological material.
MZ5 ([10a]) distinguishes by high proportions (10-55%) of P. alpicola. Maximum at the basis of MZ5a, they reflect the rapid establishment of an environmental micro-mosaic dotted with marshy loci at the beginning of the warming event. Marshy loci then reduce as temperature increases during MZ5a. In MZ5b, they sharply decrease, which mostly benefit P. muscorum preferring dry habitats, concomitantly with a strong decrease in total abundance. This last decline thus results from an abrupt shift toward colder and drier conditions.

78Each of these three phases of milder temperatures are followed by phases of decreasing and colder temperatures (MZ2 (upper half of [7]), MZ4 ([9]), MZ6 ([10b-c]), which show almost the same evolution and have been subdivided into two sub-zones (a) and (b). Subzones (a) are characterised by the predominance of S. oblonga over slugs among accessory taxa and declining total abundance. In subzones (b), slugs predominate over S. oblonga and the total abundance is minimal (fig. 14). Currently, S. oblonga lives in a wide range of mesic to wet habitats, mostly with sparse vegetation or even near watercourses and tolerates flooding (Welter-Schultes, 2012). The “Slugs” group represent a large variety of habitats, but their populations are generally favoured by moist conditions and clay soil that provides shelter and hiding places (Kerney, 1971). Given the poverty of the molluscan fauna, these three MZ are characterised by a shift from a relatively humid environment affected by runoff (sub-zone a) to a drier and more stable environment (sub-zone b) occurring in a context of decreasing temperature, which is consistent with the synchronous increasing proportions of the dominant taxon P. muscorum. Two remarks however: this trend is shortened by the appearance of palustral Pupilla alpicola at the top of MZ4 and the full scheme is less clear for MZ6 due to its extremely low abundances. Its strong increase in the proportions of T. hispidus is hardly significant for the same reason and cannot be compared with that of MZ1.

79MZ7 ([11]) stands out the oscillatory scheme described for MZ1 to 6 due to its still very low total abundance. It was distinguished from MZ6 by a sharp decrease in slug proportions, an increase to 65-80% of P. mucorum proportions and the appearance of Vallonia costata, a calciphilous species characteristic of open environments (dry or wet) that tolerates light shade but is rare in woods and marshes (Welter-Schultes, 2012). Its ecological tolerances are similar to those of P. muscorum, but it prefers more stable grassland environments (Kerney, 1971). Thus, despite the persistence of a harsh climate, a continuous herbaceous vegetation cover developed and may have favoured the trapping of windborne particles. The absence of aquatic taxa, whether native or allochthonous, also suggests a drier environment.

80Aquatic taxa are actually only absent from MZ4b, MZ6b and MZ7, which are the coldest and driest phases. They are all known from lœss sequences though not common in them (e.g., Mazenot, 1963; Hollaus, 1969; Moine et al., 2005). However, in absence of pattern in their occurrences within other MZs, it cannot be confirmed whether they were blown in from the neighbouring alluvial plain or lived locally. Their rare occurrences only indicate wetter places, sparse or distant, linked to more pronounced hydrological phases, as during stadial-interstadial transitions and interstadial phases (Moine et al., 2008, 2011, 2021).

81A comparison of MZ total abundances normalized to 10 litres with total abundance changes of high resolution molluscan records from west-European Upper Pleniglacial Weichselian lœss sequences (Moine et al., 2008, 2011; Antoine et al., 2014; Moine et al., 2021) suggests a match between MZ1, 3 and 5 and interstadials, and between MZ2 and MZ4 with transitions between gley horizons and overlying lœss units, i.e., with interstadial-stadial transitions, characterised by decreases of similar magnitude. Despite similar changes, it is not clear if MZ6 corresponds to a long poorly abundant interstadial- stadial transition or to a stadial phase like MZ7 given its very low abundance.

82Consequently, the molluscan analyses revealed the record of three interstadial phases characterised by their own signature: a shrubby character for MZ1 (lower half of [7]), a sparse moderately dry to moist steppe-like herbaceous environment for MZ3 ([8]) and a mosaic environment with marshy loci for MZ5 ([10a]). Each of them is followed by an interstadial-stadial transition, i.e., MZ2 ([7u]), MZ4 (upper half of [9]-lower third of [10a]) and possibly MZ6 ([10bc]). Each is colder and drier than the previous one but all are characterised by a discontinuous herbaceous vegetation. The absence of well distinct stadial phases between MZ2 and MZ3 and MZ4 and MZ5 is noticeable. The record ends with a cold and dry stadial phase characterised by a more continuous herbaceous vegetation.

4.5.3 – Recording of units [7], [8] and [9] in the 2012 section (fig.8)

83The three samples analysed from section 2012-1 (fig. 8, tab. 3) show a composition similar to the equivalent assemblages from section 2012-4, but also some notable differences. The aquatic fraction is much larger and more diverse. The constant presence of Succinea putris in the terrestrial fraction and the predominance of Ampullaceana balthica and Euglesa nitida in the aquatic fraction highlight the proximity of a riverbank and regular water inflows, probably from the river bordering the site. However, these inputs tend to decrease from unit 7 to unit 9, as indicated by the decline in diversity and then in the proportion of aquatic taxa in these three units.

84In the sample equivalent to unit [7], the terrestrial fractions (except S. putris) indicate herbaceous vegetation cover with a similar shrub component in both loci. In the sample equivalent to unit [8], the higher proportions of T. hispidus and lower proportions of S. oblonga suggest a less discontinuous vegetation cover with a more developed shrub component than at the 2012-4 section. In the latter, the slightly higher proportions of S. oblonga and slightly lower proportions of slugs could be explained by more persistent runoff than at the 2012-1 locus, resulting in a slightly wetter but also less stable soil surface. Finally, in the equivalent sample from subunit 9 of section 2012-4, the proportions of P. muscorum are lower and those of S. oblonga and slugs higher, except for the basal sample from unit 9, where they are similar. The absence of P. alpicola suggests that, despite inputs from the watercourse, the environment appears to have been generally less moist than at the 2012-4 sample plot.

4.6 – Archeology : main characteritics of the gravettian of Renancourt 2

85Only the main features characterising the prehistoric occupation of Renancourt 2 are presented here. For more details, readers may refer to the publication of C. Paris’ thesis (Paris, 2024).

4.6.1 – Taphonomy

86The occupation appears to be fairly confined along a south-west/north-east axis with an average width of fifteen metres. This particular human settlement is in fact contained between two geological boundaries visible during the excavation. The first, to the north-west, is marked by the chalky slope and cryo-injections, which naturally limited the extent of the occupation. This scarp overlooks the area occupied by prehistoric humans by a few metres. The second boundary is located to the south-east (fig. 3, blue dotted line). On either side of this boundary, the sediments below the archaeological level are different: hydromorphy is much more pronounced towards the valley, where the underlying chalky slope deposit is not present. On the other hand, this slope deposit is evenly distributed beneath the surface occupied by prehistoric humans. It may have played a draining role, making the soil drier and more favourable for human settlement. The partial malacological study of units equivalent to [7], [8] and [9] from section 2012-1 confirms that the southern part was significantly wetter during the occupation of the site.

87Despite a vertical dispersion of the remains of around ten centimetres and some underlying cryo-injections that may have deformed certain clusters or concentrations (vertical displacements), the spatial distribution of the remains at Renancourt 2 appears to be largely unchanged. The differential distribution of certain categories of remains clearly highlights distinct areas of activity. Clusters of flint flakes also reveal the surface area occupied by Palaeolithic humans (fig. 15).

88There are relatively numerous splinters (nearly 4,000) associated with the cutting areas, which rules out migration of small elements by leaching. The numerous lithic refittings, flint knapping waste heap and preservation of faunal remains also point to an industry in primary position and a single, well-preserved level. The slight south-easterly slope and the process of wind and colluvial cover have had little or no impact on the distribution of the remains. The general organisation of the objects that have survived has therefore been little or not at all disturbed, allowing for spatial analysis.

89However, two more destructive events are worth noting. Firstly, in the northern part of the excavation, cryo-injections have developed over a few metres at the foot of the chalky scarp (fig. 3). These phenomena constitute a geological limit to the preservation of the archaeological level. However, in this sector, no artefacts were found in the few centimetres of preserved layer, which could indicate that occupation did not develop in this part of the site. The destruction of around 100 square metres of the archaeological level should also be noted in the north-eastern quarter in the former brickyard.

4.6.2 – Stone artefacts and remains of large fauna

90The tab. 4 shows the account of archaeological remains from Renancourt 2 excavations. Three preliminary observations can be made based on this table:
. the relatively small number of artefacts in proportion to the area excavated;
. the large quantity of small flakes (< 3 cm) and splinters;
. the under-representation of certain categories such as lithic cores and tools, for example, and the absence of projectile points. This deficit cannot be explained by taphonomic problems, as the level is well preserved. It therefore reflects anthropogenic activities, which is confirmed by the location of these categories of remains in the plan (clusters of cores, tools away from concentrations, clearly defined activity areas).

91Despite the small number of finished products (30 blades and 6 tools) and cores left on site (N = 4), the main debitage methods can be identified. The selected local flint blocks come indiscriminately from the two Cretaceous stages (high-quality Upper Turonian/basal Coniacian flint or lower-quality Coniacian b/c flint) present near the site. The blocks were selected based on their morphology and dimensions, which allowed for the creation of a laminar surface exceeding ten centimetres in length. The shaping out process consisted of preparing a median anterior crest (frontal installation) to initiate the debitage, while the rest of the volume appears to have been little used. The debitage is unipolar and the products are extracted using a soft stone hammerstone after faceting the striking plane. The blades thus obtained are slender and straight, with dimensions close to ten centimetres.

92The production of laminar flakes, or even short flakes, has also been observed. This appears to be minor, hasty and unproductive.
The distinctive feature of the site is the spatial segmentation of the debitage, with the input of at least one preformed block of tertiary flint, and the removal of cores and blades produced on site from the excavated area for use elsewhere. However, some were used directly as butcher’s knives, according to the traceology -by Émilie Claud (Paris, 2015).
The treatment of animal carcasses is the second activity identified. Of the 224 bone remains, half were identified by Patrick Auguste (Paris et al., 2019). Three species are represented: reindeer, horse and bison. However, the number of animals processed at the site appears to be fairly modest (tab. 5).

93The distinctive feature of the settlement is the predominance of long bones over axial or cranial skeletal elements. None of the remains have been altered by fire. The state of preservation is average, with the external surface of the bones affected by various post-depositional phenomena. No cutting marks have therefore been preserved. On the other hand, deliberate fragmentation of certain bones with a notch and/or helical fracture has been observed on about ten pieces. The purpose of this anthropogenic action was to recover the bone marrow. Although the number of bones remains is small and therefore calls for caution, the predominance of long bones suggests that part of the carcasses, in this case the limbs, were selectively transported. However, the actual consumption of food may not have taken place in situ. The excavated area may have been used for preparation (cutting meat and breaking bones) rather than consumption, as suggested by the absence of traces of fire (for cooking or smoking, for example).
The site is attributed to an early Gravettian phase (Paris et al., 2021; Paris, 2024).

4.7 – Chronology

94Of the three age series available for the Renancourt 2 lœss sequence (fig. 16), two (OSL and ¹⁴C on GVT) relate to all or part of the sedimentary sequence and one is linked to the archaeological level (¹⁴C on bones).
Distributed from [7] to [9], the ¹⁴C on GVT ages (tab. 6) show no inversion and are increasingly younger, in accordance with the order in which the deposits were formed. The nature and sedimentological characteristics of the deposits associated with this chronological range are also similar to those of Havrincourt (Antoine et al., 2014).
Regarding OSL dating (tab. 7), the two ages obtained in the uppermost unit, [11], are the youngest in this set and are also younger than the ¹⁴C age on GVT of the underlying unit [10a]. The OSL age of 28.69 ± 3.53 ka obtained in [10c] is consistent with the OSL ages of [11] and the ¹⁴C age on GVT of [10a]. In contrast, the OSL age of 25.22 ± 2.11 ka obtained in [8] is significantly younger and aberrant compared to all other OSL ages and ¹⁴C on earthworm granules, as it corresponds to the age of the thick laminated lœss layers that are regionally contemporary with stage GS-3 (Antoine et al., 2016) and locally represented by [11]. It was therefore not taken into account in establishing the chronostratigraphy of the sequence.

P. Auguste

95Finally, the ¹⁴C ages on bones from the archaeological level (tab. 8) also need to be discussed. Firstly, the proximity of the age of bone 702 to the other ages from the fauna and its clear divergence from the GVT dating of [7] confirm its connection to the occupation and therefore the stratigraphic attribution of its encasing to [8]. Furthermore, these six ages show a dispersion that contradicts the results of the taphonomic approach, which confirms the uniqueness of the level and its rapid covering by lœss deposits (fig. 16).

96Thus, compared to the two oldest ages, two are 1,000 to 1,500 years younger and two others are 2,500 and 4,500 years younger. A similar phenomenon has been observed at the neighbouring site of Renancourt 1. Some ages on bones are sometimes rejuvenated by up to 1,500 years compared to the oldest ages on charcoal, bones or earthworm granules from the same level (Moine et al., 2021). No significant artificial ageing was therefore observed. Given the proximity of the two oldest bone ages of [8] to that of earthworm granules, they are considered to be the most accurate. The two most recent bone ages were rejected and the two "intermediate" ages, to be considered as minimum (Evin, 1992), were taken into account with caution in establishing the chronostratigraphy of the sequence.

97The ages obtained place the studied sequence in the Weichselian, at the end of the Middle Pleniglacial and in the first half of the Upper Pleniglacial between approximately 35 and 24 ka. However, the gap of nearly 4,500 years between OSL 4 and OSL 2 and the particular nature of the stratigraphic record for this interval compared to those of Havrincourt (Antoine et al., 2014) and other regional sequences (Antoine et al., 2016) suggest the occurrence of a sedimentary hiatus.

5 – Pedosedimentary, chronostratigraphic ans paleoclimatic synthesis

98In order to integrate this sequence into the regional synthetic stratigraphy, a chronostratigraphic scheme is therefore proposed based on the litho/pedostratigraphic succession, the stadial or interstadial nature of the units identified using malacological assemblages and sedimentological indicators, and dating.

5.1 – Fromy the upper Saalian to the Weichselian lower Pleniglacial

5.1.1 – The alluvial subsequence (units [2] and [3])

99At the base of the sequence (fig. 10), an alluvial formation resting on the chalky substrate [1] was identified in numerous test pits, mainly towards the Selle valley (fig. 1). This alluvial formation consists of two alluvial phases of decreasing competence, i.e., a coarse formation ([2]) overlain by fine alluvium (silt, sand, [3]) deposited in a glacial context, according to the molluscan fauna. No age has been performed in these deposits. However, in the Selle Valley bordering the site, the altitude of the substrate closest to the site, which marks the maximum incision locally, is 14 m a.s.l. at the hippodrome (BRGM Underground Bank, BSS000DYUC). At Renancourt 2, the base of the alluvial sheet is located at around 21-23 m a.s.l in SP 7 - southern part of the site - and below 19 m a.s.l. (substrate not reached) in SP 2, a few dozen metres to the east of the site. The terrace is therefore +5-8 m above the maximum incision observed at the hippodrome and would thus correspond to the Alluvial Formation I or Étouvie Formation, whose relative elevation to the maximum incision is +5-6 m in the Somme valley. It is attributed to Marine Isotopic Stage 6 (Upper Saalian) (Antoine, 1990; Bahain et al., 2007), which is consistent with the climatic context indicated by the malacological data.

Sylvie Coutard

Sylvie Coutard

Sylvie Coutard

5.1.2 – The slope deposits subsequence (units [4] to [6])

100The alluvial sequence is covered by at least two phases of thick chalky coarse slope deposits ([4] and [6]) interspersed with highly hydromorphic geliflucted sands and silts ([5], fig. 9). These coarse slope deposits correspond to the geliflucted chalk flows often found between the top of the alluvial formations of the Somme terraces and their lœss cover, near the former chalky scarps (Antoine, 1990; Antoine et al., 2000, 2007). No dating has been performed in these deposits at Renancourt 2. Their deposition may have begun at the start of the glaciation, but the main phases of accumulation of chalky slope deposits could date from the Lower Pleniglacial, as in Ault (Antoine et al., 2011, 2016).

5.2 – Weichselian Middle Pleniglacial

5.2.1 – Unit [7]

101The ¹⁴C age on earthworm granules (GVT 1) of 35425-34189 cal BP indicates that the deposition of the lœss sequence begins at the end of the Weichselian Middle Pleniglacial.
The lœssic subsequence begins with a silty horizon [7] marked by hydromorphy, gelifluction and cryogenic deformation, interpreted as a gley horizon. It is mainly present in the southern part of the site, where its hydromorphic features are more clearly expressed. In section 2012-4, the lower part of [7] (-25 to 0 cm), is only preserved in a depression (fig. 5). It consists of a clayey loam (clay: ~19%) probably colluviated during an interstadial phase (MZ1; fig. 14), marked by the development of a shrubby herbaceous vegetation (high proportions in T. hispidus) and possibly an enhanced biological activity (TOC: 0.3-0.35%) (fig. 13).

102In the upper part of [7] (0 to +15 cm), the strong clay content increase (~19 to 40%) rather implies enhanced fine supplies than pedogenetic processes (fig. 13). It cannot be excluded that part of the “clay fraction” identified in sedimentary analyses corresponds in fact to the carbonates and that the high clay content is due to redistribution phenomena. Furthermore, the molluscan data reflect the loss of the shrubby component (MZ2; fig. 14) as well as a sharp drop in temperature (MZ2a) and the settlement of drier conditions (MZ2b), which highlight a shift to an interstadial-stadial transition inconsistent with increased pedogenic processes. In this case, the strong clay content increase mechanically induces a decrease in the GSI (0.75 to 0.4) that does not necessarily imply an important decrease in eolian dynamics. In absence of a clear GSI or CSI increase, it cannot be stated that full stadial conditions are reached at the top of [7]. Despite a strong vegetation decline revealed by malacofauna, the TOC content shows no clear decrease, suggesting that its signal may be mostly inherited from reworked aggregates included in [7]. Besides, synchronous increasing trends in MS and coarse silts, opposite to that of the clay content, suggest a wind-driven allochthonous supply in magnetic particles rather than their local pedogenetic origin through bacterial activity. Likewise, the sudden increase of the calcium carbonate (CaCO₃) content to maximum values (~18 to 25%) can imply either enhanced allochthonous inputs by colluvial or wind processes, from upslope chalky substrate outcrops (fig. 3) freed by the vegetation decline and eroded by frost or CaCO3 distribution by pedogenesis.

5.2.2 – Unit [8], the grey brown silt containing the archaeological level

103This brown-grey carbonate silt with chalk granules and calcareous concretions, which includes the archaeological level, is present across the entire excavated area, with the possible exception of the north-east corner at the foot of the chalky slope (fig. 3).
At the microscopic level, the sediment is rich in aggregates of all sizes, from small rounded aggregates (Ø < 100 µm), which testify from moderate solifluction, to large sub-rounded aggregates (Ø > 1 mm and up to 2.5 mm; fig. 11, fig. 12-C and K), which are characteristic of [8] and testify from marked bioturbation. In section 2011-1, owing to a more pronounced presence of enchytraeids, the upper part of [8] may be considered as an “Arctic meadow” soil type (Van Vliet-Lanoë, 2005). This distinction is however not marked in section 2012-4. The frequent presence of chalk granules also indicates some inputs from the chalky scarp. Discrete caps topping them as well as large aggregates imply their limited rotation. In section 2012-4, [8] can therefore be interpreted as a weakly pedogenised calcareous sediment of aeolian and colluvial origin, strongly bioturbated, particularly by the activity of earthworms, prior to the formation of the cryogenic structure. The cryogenic lamellar-lenticular or even granular structure that mainly affects the upper 30 cm of [8] results from the post-depositional impact of freeze-thaw cycles. As the formation of indicators of an arid phase (mainly queras and gypsum), is posterior and disconnected from the deposit and subsequent freeze-thaw structuring of [8], they will be discussed further.

104Although [8] appears almost homogeneous in the thin sections, the analyses carried out (figs. 13 & 14) reveal a polyphased evolution and therefore a more complex deposition and pedogenic dynamics than suggested by its homogeneous aspect.

105Between +15 and +35 cm, in the lower part of [8] corresponding to malacozone MZ3a, the TOC (~0.3%) and CaCO₃ (~25%) values are similar to those at the top of [7]. The clay content is relatively high but lower (~30%) than in unit [7] and fairly stable, and the GSI slightly increases to ~0.6 but remains lower than in the lower half of [7]. This paints a picture of a fairly stable environment. The strong abundance increase of the malacofauna reflects a stadial-interstadial transition context associated with a steppe-like vegetation and slightly less dry conditions at the top of MZ3a (fig. 14).
Between +35 and +45 cm, in the central part corresponding to MZ3b, the synchronous decreases in TOC (0.32 to 0.26%) and molluscan abundance suggests a decline in biological productivity possibly linked with a slight climatic deterioration. The synchronous decrease in CaCO₃ (25 to 22%) content cannot result from the same forcing and has no equivalent in other sedimentary proxies. However, this decrease marks the beginning of a decreasing trend lasting to the top of the section and opposite to a similar but increasing trend of the GSI and CSI. Assuming the mainly local origin of carbonates, this decreasing trend in CaCO₃ content could imply a new phase of declining carbonate inputs and thus a depletion of local sources resulting from the progressive covering of local chalk outcrops by lœss deposits. The similar trend in TOC content may also be significantly influenced by a change in the colluviated materials for the same reason and not mostly reflect vegetation changes as basically expected.

106Between +45 and +60 cm, in the upper part of unit [8] corresponding to malacozone MZ3c, the environment becomes more humid and the temperature begins to decline according to the molluscan fauna. Indeed, the decreasing TOC (0.26 to 0.21%) and clay (~30 to 25%) contents indicates a decline in the vegetation cover contemporaneous from an enhancement of the eolian dynamics underlined by a clear GSI increase (~0.6 to 0.85). The expansion of the lœss cover over local chalk outcrops probably explains the synchronous decrease in CaCO₃ content.

107The three ¹⁴C ages linked to the archaeological occupation give very similar intervals of 32080-33595 (bone 664), 31749-33532 (bone 439) and 31913-33768 cal BP (GVT 2). Based on the vertical projection of the lithic cluster in section 2012-4, the prehistoric occupation took place during MZ3a, or at the latest at the beginning of MZ3b. Within MZ3, it cannot be ruled out that the decline in abundance that led to the distinction of MZ3b was not the result of cooler temperatures but rather of human occupation of the area, as suggested by the significant drop in abundance contemporary to the human occupation at Renancourt 1 site (Moine et al., 2021). The gradual evolution of the composition of the malacofauna throughout MZ3 makes the occurrence of a hiatus at the MZ3a-MZ3b interface highly unlikely. The combination of the dating results and molluscan population dynamics imply correlating [8] and the human occupation with a unique interstadial. Conversely, whereas only 30 cm of sediment separated ¹⁴C ages GVT 1 and GVT 2, the interval of almost 2 ka between their median ages (tab. 6) probably implies a sedimentary hiatus within [7] at the interface between MZ1 and MZ2, marked by abrupt changes in both molluscan and sedimentary compositions, rather than at the interface between [7] and [8] where such changes are absent.

5.2.3 – Unit [9]

108The unit [9] is marked by hydromorphy and cryoturbation (large cells, involutions visible in large sections but discreet in colour), which led to interpret it as a gley. Its microstructure is lamellar to lenticular, even granular, with the presence of ovoids implying some gelifluction (fig. 11). This is the same cryogenic microstructure that propagates downward in the upper part of [8]. The absence of the large aggregates (Ø > 1 mm) characteristic of [8] probably results from enhanced waterlogging unfavourable to the concerned animals. The malacofauna of MZ4a indeed reflects an interstadial-stadial transition context characterised by a discontinuous herbaceous vegetation, a strong humidity increase and unstable soil surface conditions. Nevertheless, the TOC values remain relatively high around 0.2%, which raises again the possibility of its sedimentary inheritance. Despite a clear peak in the coarse silt proportions, the MS signal shows no synchronous increase, which may be due to magnetic mineral dissolution or transformation by waterlogging (Maher, 1998; Baumgart et al., 2013) (fig. 13). The occasional decrease from 19% to 16% in CaCO₃ associated with the presence of rare lœss dolls may be linked to slight decalcification, as is often the case in gley horizons (Antoine et al., 2001).
The continuity between molluscan faunas of MZ3c and MZ4a (fig. 14) and the closeness (< 0.8 ka) of the three ¹⁴C ages of [8] with ¹⁴C age GVT 3 (31,386-32,990 cal BP) obtained in [9] argues in favour of the correlation of [8] and [9] with the same interstadial.

5.2.4 – Unit [10a]

109This sub-unit, distinguished on the section 2012-7 and so in 2012-4, is a 20-cm-thick grey-brown silty layer with small chalk granules, slightly gleyified at the top (deferrization and ferro-manganese precipitation, slight freeze-thaw structure; fig. 11). These pedosedimentary evolution fits with the transition from colder conditions with a sparse humid herbaceous vegetation at the base (MZ4b) to an interstadial context with an herbaceous vegetation with palustral loci (MZ5) indicated by the molluscan fauna. At the base of [10a], a drop in the TOC content is contemporaneous and consistent with the very low molluscan abundance reached in MZ4b. However, its sharpness is more similar to the synchronous steady increase in coarse silts proportions and GSI that mark the installation of an enhanced eolian dynamics (figs. 13 & 14). The drop in TOC content may thus partly result from a decrease in colluviated material. In MZ5a, the synchronous positive oscillations of the TOC content (~0.13-0.18%) (fig. 13) and the total molluscan abundance (fig. 14) highlight a phase of development of the vegetation cover. Throughout MZ5a, the proportions of P. alpicola decreasing to the benefit of these of P. muscorum reveals a reduction of the palustral loci, which probably results from a general dry up trend to a marked enhancement of the eolian dynamics as indicated by increasing proportions in coarse silts (~43 to 48%) and GSI above 1 for the first time), especially at the top. The switch to new stadial conditions begins in MZ5b by a sharp decrease in the molluscan total abundance.
The ¹⁴C age GVT 4 provides an age of 30368-31190 cal BP for [10a]. The gap of about 1.2 ka between its median values and that of ¹⁴C age GVT 3 in [9] (tab. 6) combined with the molluscan fauna dynamics suggests that [10a] is contemporaneous from the following interstadial phase.

5.3 – Weichselian Upper Pleniglacial

5.3.1 – Unit [10b-10c]

110These two subzones clearly visible on section 2012-4 (fig. 5) and more broadly in sections 2012-1 and 2012-7, consists of a silt [10b] - browner and more homogeneous than [10a] - and a stratified brown silt [10c]. A thin, discrete gleyed zone - iron removal and ferro-manganese precipitation, slight freeze-thaw structure - is also present at the top of [10c] pointing to a wet and cold environment.

111The molluscan fauna very low total abundance and composition reflect a discontinuous herbaceous vegetation associated with cold but moderately humid conditions. The very low total abundances in molluscs precludes to infer a slight development of the shrubby component from the higher proportions in mesophilous species T. hispidus to explain the slight positive oscillation in TOC. A last strong shift in grain size parameters occurs in the middle of stratified [10c], suggesting the reworking of clayey material from [9] or underlying units (fig. 13). If not taken account, sedimentary parameters show an increasing trend in GSI (1 to 1.3) and SM (8.5-9 × 10^-8 m³.kg^-1) contemporaneous with a clear decrease in CaCO₃ content (~18 to 12%) and a smoother one in TOC content. They indicate a new phase of acceleration of the eolian dynamics and of extension of the lœss cover over local chalk sources. The OSL 4 age of 28.69 ± 3.53 ka obtained for [10c] in section 2012-1 (where it is thicker) shows a large gap of about 2.2 ka with the median values GVT 4. This, associated with changes in the molluscan composition and the CaCO₃ content between [10a] and [10b], would argue in favour of a hiatus, even if the large uncertainties of the OSL date make it impossible to state this with certainty. However, based on this age, this part of the record is the first attributed to the Weichselian Upper Pleniglacial.

112From a micromorphological point of view, [10b] and [10c] are the uppermost (sub)zones in which marker features of aridity have been observed. As reminder, "queras"-type root pseudomorphs (sparite crystals associated with decarbonatation of the surrounding sediment) were mainly observed in [10a-c] of the 2012-4 section (figs. 5, 11) and gypsum neoformations, i.e., rosettes, in [8] only in section 2011-1 (and occasionally in [10] of 2011-3 section) (figs. 6 & 11). In addition, [8] is also characterised by the development of carbonate hypocoatings and needle calcite, also linked to a semi-arid environment. Subsequent to the cryogenic structuring of the sediment, all these features logically correspond to one (or more) phase(s) of aridity subsequent to [9] and [10] and prior to the deposition of the laminated silts [11] from which they are absent. Given the hydromorphic character of [10c], the presence of these pedogenic features suggest the occurrence of an arid phase between [10] and [11] and absent from the stratigraphic record of Renancourt 2.

5.3.2 – Unit [11]

113This unit consists of yellow laminated silts, the base of which is a clear erosion surface marked by a gravel bed. This significant erosion phase clearly marks a break with the underlying sedimentation. However, the sedimentary characteristics of the two basal samples (+130-135 and +135-140), corresponding to MZ6b, suggests a sedimentary dynamics closer to that of [10c] than to the upper part of [11]. Nevertheless, the differences (including higher proportions in fine sands) are significant enough to rule out a simple reshuffle (fig. 13). Though still poorly abundant, the molluscan composition of the two samples from the base of [11] is consistently different from that of [10c].

114On the contrary, in the upper 40 cm of the unit [11], characterised by very fine, sometimes millimetric bedding, a much coarser grain size (coarse silts: 60%; GSI > 2) indicates an input of material linked to a strong increase in eolian dynamics. CaCO₃ and TOC reach their minimum values at 10 and 0.1% respectively, which is consistent with a change in sources and the persistence of a very low molluscan abundance, even if the higher proportions in P. muscorum and occasional presence of Vallonia costata suggests the establishment in slightly drier conditions of a more continuous herbaceous cover than previously. Given the collapse of pedological and biological activity, the increase in magnetic susceptibility probably results from enhanced wind-driven supplies of magnetic particles. Despite the onset of colder and drier conditions, the presence of beds of chalk granules and more reddish-brown clayey beds consisting of material eroded higher up the hillslope indicates the continuation of erosive water activity and occasional phases of runoff and/or gelifluction. It is possibly linked to the spring melting of the snow cover that gave rise to the so-called "niveo-aeolian" facies of these deposits (Dijkmans & Mücher, 1989). Ages at 24.18 ± 2.35 ka (OSL 2) and 22.69 ± 2.5 ka (OSL 3) show a gap of 4.5 to 6 ka with OSL 4. The gap between [11] and [10c] is therefore even more important than the gap between [10b] and [10a], which strengthens the hypothesis of a hiatus between [10c] and [11] suggested by the presence of aridity markers in underlying units.

5.3.3 – Unit [12]

115Unit [12] is a decarbonated lœss. The homogeneous calcareous lœss following [11] in regional sequences and preserved at Renancourt 1 (Moine et al., 2021), has not been observed at Renancourt 2. Probably not very thick, it has been completely decarbonated by the Lateglacial-Holocene pedogenesis.
Outside the excavation area, this Upper Pleniglacial homogeneous lœss is rare on this slope. However, the local equivalent of the Nagelbeek Tongue Horizon has been identified in a few test pits, directly beneath the illuvial (Bt) horizon of the leached brown surface soil, in the form of a well-developed pale grey tongue horizon. A thin layer of homogeneous calcareous lœss overlies it.

5.4 – Weichselian Late Glacial and Holocene

116The lœssic subsequence ends with the development of a leached brown soil ([13]) at the expense of pleniglacial lœss. The illuvial horizon (Bt horizon) of this soil and the associated decarbonated lœss ([12]) are only present outside the areas used for brickmaking.

6 – Discussion

6.1 – The Middle to Upper Pleniglacial Transition in Regional and North-West European Lœss Sequences

117In most of the sequences of northern France, the Middle Pleniglacial is characterized by a very poor pedosedimentary budget corresponding to a single (or polygenic) brown soil horizon (Saint-Acheul Soil; Antoine, 1990). Thanks to a few detailed sequences, as Villiers-Adam (Val d’Oise, France) and Havrincourt (Pas-de-Calais, France), a more complex synthetic sequence has been established (Antoine et al., 2016), with no less than four horizons at Villiers-Adam (Locht et al., 2003). This pedocomplex, the Saint-Acheul/ Villiers-Adam Soil complex (SAVA), is characterised by boreal brown soils (Bw horizons) to arctic brown soils formed during the main interstadials that characterize this period. Furthermore, at Havrincourt, the upper part of the soil complex also comprises at least two phases of soil formation whose level of development remains significantly lower than for the underlying brown soil (Antoine et al., 2014). The unit 6 (fig. 17) corresponds to a hydromorphic arctic brown to arctic meadow soil horizon (TOC: 0.4%, clay: 29%). An archaeological layer close to the top of this unit has been dated at 32900-31656 cal BP (28100 ± 180 BP; Beta-332604) and the base of this unit at 42,1 ± 2.8 ka by OSL (Goval & Hérisson, 2018). The overlying unit 5 is a tundra gley associated with an ice-wedge network. More recently, detailed studies at Haynecourt (around 10 km north of Havrincourt) have clarified the chronostratigraphy of the transition between the Weichselian Middle and Upper Pleniglacial (Vercelot et al., 2025). The authors proposed to distinguish a sub-phase A (~55-36,5 ka) characterised by the development of a complex of decarbonated boreal-arctic brown soils, and a sub-phase B (~36,5-30,5 ka) characterised by a progressive change toward a more rigorous tundra-type environment ending with the formation of a cryoturbated gley horizon associated with a large polygonal network of ice-wedge cast pseudomorphoses. The tundra gley would therefore belong to the Middle Pleniglacial B (MPG-B). At Haynecourt (Vercelot et al., 2025), the uppermost unit of SAVA soil complex, e.g., unit 7, is dated between 34166-32897 cal BP (28990 ± 200 BP; GifA24323) and 42047- 41173 cal BP (36720 ± 310 BP; GifA23523). The topping gley horizon (unit 6) yielded ECG-based radiocarbon ages of 30415-31046 cal BP (26510 ± 110 BP; GifA23528), consistent with dates obtained at Havrincourt and Morcourt (fig. 17), Antoine et al., 2014; Sambourg et al., 2025). Based on the relationship established between gley horizons and Greenland interstadials (Moine et al., 2017), this gley horizon topping the SAVA Complex has been correlated with interstadial GI-5.1 and the uppermost unit of the SAVA Complex, when distinct, to the time interval encompassing GS-8 to GI-5.2 (Vercelot et al., 2025). The thickness and the nature of this uppermost unit depend on the context and the local sedimentation rate, but it may be carbonated.

118In this context, the dates of the Renancourt 2 lœssic sub-sequence from [7] to [10a] place it in the Middle Pleniglacial B. At Renancourt 2, well-developed arctic or boreal brown soils typical of the SAVA soil complex are absent. None has been identified throughout the area explored during archaeological survey (fig. 1). Either they never developed, or more likely they preceded the deposition of the lœssic subsequence and they were completely eroded being of limited thickness. This major erosion phase could be associated with the end of a permafrost episode and a degradation event (rapid warming), as recorded at Nussloch and Havrincourt (Antoine et al., 2016).

119The Renancourt 2 section is represented by 130 cm almost integrally deposited between 34189-35425 cal BP in [7] and 31190-30368 cal BP in [10a]. Based on its age and very humid interstadial character, the slightly gleyified upper half of [10a] can be correlated with the gley horizon topping the Middle Pleniglacial B. The 110 cm of the Renancourt 2 lœssic sub-sequence for MPG-B are then thicker than the 70 cm (unit 7) at Haynecourt (Vercelot et al., 2025), 30 to 75 cm (unit 6) at Havrincourt (Antoine et al., 2014) and 60 cm (unit 3) at Villiers-Adam (Antoine et al., 2003b), 45-60 cm (unit 6) at Morcourt (Sambourg et al., 2025). These correlations highlight the exceptional thickness of the Renancourt 2 sequence at the regional scale.

120Only the sequence of Morcourt may present a comparable sedimentary budget, especially if the 25 to 40 cm of the top of the unit 7 have to be included to the 50-60 cm of unit 6 to complete the time interval of Middle Pleniglacial B. The same remark is also valid for other north-west European MPG-B sequences, with thicknesses between about 1 and 3 m for plateau sequences in Belgium (Rocourt: ~1.1 m, Remicourt: ~1.5 m, Harmignies: ~1.5 m; Maisières-Canal: ~2 m; Haesearts, 2016) and in Germany (Nussloch: ~2.5 m; Antoine et al., 2009; Remagen, ~2.7 m; Vinnepand et al., 2023), whereas those occupying a slope position at Maisières-Canal reach between 4 and 5 m in thickness. The dilated character distinguishing the three sequences of Renancourt 2, Morcourt and Maisières-Canal seems indeed related with their topographic context, i.e., all occupy a leeward slope position, or even the foot of a chalk scarp.

Sylvie Coutard

121The Renancourt 2 lœssic sub-sequence appears much more detailed than the MPG-B record of other regional sequences, with a more complex structure, evidenced by lithostratigraphy and malacofauna evolution.

122Nevertheless, it is still less complex than that of the Belgian and German sequences, which are even more diversified with two to four weak pedogenic horizons (arctic brown soils, gleys or humic soils) alternating with lœss units and topped by a thick tundra gley, in the same period. Owing to radiocarbon chronologies the contemporaneity of these gleys and arctic brown soils with interstadials has been demonstrated at Nussloch (Moine et al., 2017), Remagen-Schwalbenberg (Prud’homme et al., 2022) and for Belgian sequences (Haesaerts et al., 2016). The thickness and complexity of these sequences are due to a higher sedimentation rate than in Picardy. In this model of lœss/soil (gley) doublets, the accretion is maximum during stadials and minimum during interstadials. At Renancourt 2, the molluscan record of the MPG-B, which mostly recorded interstadials and interstadial-stadial transitions, reveals a different pedosedimentary dynamics. The accretion of the sequence during interstadials characterised by lower dust deposition rates thus implies the predominance of colluvial processes, which is confirmed by the dynamics of sedimentary proxies and some micromorphological features. More humid conditions are indeed typical of interstadials (e.g., Duprat-Oualid, 2017; Prud’homme et al., 2022) and favour colluvial processes in particular in context of discontinuous vegetation cover (AFES, 2008; Bertran, 2004), whereas stadial are systematically colder and drier. At Renancourt 1, the end of interstadial-stadial transitions are marked by a peak in slugs proportions and minimal GSI values near the top of a brownish horizon, just before lœss sedimentation resume and GSI strongly increase (Moine et al., 2021). The absence of such a trend reversal after MZ2b and MZ4b at Renancourt 2 implies the absence of record of subsequent stadials, but neither their non-deposition nor their erosion can yet be confirmed.

123Consequently, the Renancourt 2 lœssic subsequence thus constitutes a regional reference record for the end of the Middle Pleniglacial, thanks to a complete and detailed record of palaeoenvironmental changes and a robust chronology. However, the sedimentation processes mixing colluvial and aeolian deposits, as well as the behaviour of carbonates in this specific context, make this a distinct sequence in terms of sedimentological parameters. Nevertheless, a few points can be mentioned to draw comparisons. At Morcourt, the unit 6 is distinguished by high proportions (20-25%) of Trochulus hispidus similar to those of MZ1 in Renancourt, suggesting a possible contemporaneity between layer [7] in Renancourt 2 and unit 6 in Morcourt despite their different facies, i.e., hydromorphic horizon vs. arctic brown soil respectively. However, this malacological characteristic, which has only just been identified, is still too poorly documented to constitute a reliable chronological biostratigraphic marker at the regional scale. Concerning the unit 8 of Renancourt 2, the micromorphological features indicate significant biological activity, including enchytraeids activity, leading to the hypothesis that all or part of the unit constitutes an arctic meadow soil (Van Vliet, 2005), a designation currently proposed for the unit 6 of Havrincourt (Goval & Hérisson, 2018; Antoine et al., 2016), which is partially contemporary. However, the latter lacks secondary carbonates and the structuring by freeze-thaw cycles is less pronounced. Secondary carbonates, particularly needle-like calcite, are on the other hand found in abundance in unit 6a of Sourdon (Antoine, 1990), which has a similar facies. Only high-resolution proxy records and numerous ages on numerous sections will enable to depict the chronostratigraphy, the sedimentary dynamics, the soil-forming processes and the palaeoenvironmental evolution of the MPG-B and will help to establish precise correlations.

6.2 – Correlations with global climate changes

124The unique combination of a thick succession, high-resolution sedimentary and malacological records and numerous ages for the Renancourt 2 lœssic sequence allows for the proposition of a more precise correlation scheme between stratigraphic units and stadial-interstadial alternations of the Greenland ice core climate records (fig. 18) than for the contemporaneous lœss sequences of all other regional sites, lacking either a precise chronology, a sufficient resolution with a detailed chronostratigraphic succession or palaeoenvironmental records.

125[7] : The ¹⁴C age on earthworm granules (GVT) of 35425-34189 cal BP combined with the interstadial character of MZ1 supports the correlation of the lower half of [7] with interstadial GI-7 (fig. 18).
[8] : The three ¹⁴C ages (bone 664), (bone 439) and (GVT 2) obtained for the archaeological occupation yielded very similar ages of 32080-33595, 31749-33532 and 31913-33768 cal BP respectively. However, (i) the medians of the distributions of the two oldest ages are centred on stage GS-6, which is inconsistent with the very high total abundance of molluscs at the peak of MZ3a, (ii) there are less than 250 years from the older limit of the acceptable range for attribution to GI-5.2 and about 150-350 years from the younger limit of the GI-6 interstadial, (iii) [8] is characterised by high biological activity marked by the presence of numerous aggregates and channels of biological origin, some of which are posterior to the freeze-thaw structure (root channels). The combination of the dating results and molluscan population dynamics imply correlating [8] and the human occupation with a single interstadial, but the accuracy of the dates does not allow us to discriminate between interstadials GI-6 and GI-5.2 (fig. 19). However, the presence in the radiocarbon calibration curve of a plateau during the 500 years preceding the chronological interval of the GI-5.2 interstadial tends to artificially age by several hundred years the ages contemporary from GI-5.2 (fig. 19). Thus, the age probability density curves of ¹⁴C ages GVT 2 and on bone 669 show greater overlap with the GI-5.2 interval than with that of GI-6, and their median ages are included in the GI-5.2 interval. A correlation with GI-6 cannot be ruled out, but appears less likely when this parameter is taken into account.
Furthermore, there is no evidence to rule out the occurrence of granule mixing due to the introduction of older granules via the erosion of older deposits or the excretion of more recent granules by worms at depth in the absence of a clear active layer, or any other process, as highlighted by the vertical dispersion of the lithic industry from the human occupation level over a thickness of nearly 10 cm. The study of the Morcourt site has indeed revealed that in a slope context and in the absence of permafrost, ECG-based radiocarbon dating can yield ages that are significantly older or younger than those obtained from slug plates (Sambourg et al., 2025). The use of the latter could perhaps remove some or all of the remaining doubts about the chronology of the Renancourt 2 sequence, for which this material was not considered for ¹⁴C dating.
[9] : The ¹⁴C age GVT 3 (31,386-32,990 cal BP)
combined with the gradual transition between molluscan fauna of MZ3c and MZ4a supports the correlation of [9] with the end of interstadial GI-5.2.
[10a]: The ¹⁴C age GVT 4 at 30368-31190 cal BP
combined with the interstadial character of MZ5 supports the good correlation of [10a] with the GI-5.1 interstadial. The molluscan fauna of MZ5 revealed the humid character of this interstadial marked by the presence of Pupilla alpicola. Interestingly, the contemporaneous gley horizon G2b of Nussloch also distinguishes by marshy conditions marked by the predominance of the hygrophilous taxa Succinella oblonga, along with Columella columella, from open and humid mountain environments, and palustrine taxa Vertigo genesii (Moine et al., 2008). We also note that in both sites the amplitude of the total abundance peak contemporary of the weaker interstadial GI-5.1 is half that of the other interstadial maxima. However, with only two sites this resemblance might be incidental, or linked to very humid conditions unfavourable for a majority of taxa populations rather than to a weaker temperature increase.
In many northwestern European sequences, the opening at the top of this gley horizon of a decametric polygonal networks and large metric ice wedges (F-4 network in northern France: Antoine et al., 2016; Harveng soil in Belgium: Haesaerts et al., 2016) also supports its particularly humid character. As mentioned above, this gley horizon has been recently dated at Haynecourt (Vercelot at al., 2025) around 31 000 cal BP, which improves the reliability of older data, e.g., thermoluminescence (TL) ages of 29 ± 3 and 30 ± 3 ka at Savy (Aisne, France) (Locht et al., 2006). This implies that the interstadial GI-5.1 corresponds to the period of degradation of this impressive ice-wedges network that immediately precedes the deposition of a thick lœss unit contemporary with GS-5.1, and marking the beginning of the Upper Pleniglacial (Antoine et al., 2003a, 2016; Vercelot et al., 2025). Mostly present on plateaus and flat areas (Lautridou & Sommé, 1981; Antoine et al., 2016; Haesaerts et al., 2016), this typical marker horizon has not been recognised in the Renancourt area. Even if numerous parameters interact to influence the rates and magnitude of ice wedge development, this absence is probably linked to the hilly topography, which prevented its formation by the enhanced drainage of soil water. Furthermore, the Renancourt 2 sequence also lacks the subsequent 60-to-80-cm-thick lœss unit present in many regional sequences (Antoine et al., 2016), whereas it is present though reduced to a 30-cm-thick unit at Renancourt 1 (see unit 11 in Moine et al., 2021). This supports the occurrence of an erosion phase or a sedimentary hiatus between [10a] and [10b].
[10b-10c]: The OSL 4 age of 28.69 ± 3.53 ka obtained for [10c] in section 2012-1 (where it is thicker) is centred on stadial GS-4, but its uncertainty ranges from approximately 32.2 to 25.2 ka, i.e., from the middle of GI-5.2 to the middle of stadial GS-3. In this interval, the correlation of the [10b-10c] doublet with interstadial GI-5.1 is excluded since already correlated with [10a], and the very low abundance in molluscs in MZ6 excludes any correlation with interstadials GI-4 and GI-3. By comparison, the abundance of molluscs is very high for these interstadials at Renancourt 1 (Moine et al., 2021), Havrincourt (Antoine et al., 2014) and Nussloch (Moine et al., 2008). The numbers in MZ6 at Renancourt 2 are very low, without any evidence of post-depositional disappearance of shell material; despite the presence of "queras", the sediment remains indeed carbonated. The GSI values of [10b-10c] ranges between 1.25 and 1.45, which fits with the GSI interval of the lœss unit contemporary from GS-4 (~1.2-1.5) at Renancourt 1, but not for those of lœss units contemporary from GS-5.1 (~1.6-1.8) and from GS-3 (> 2). The clay content range of [10b-10c] (~15-20%) is also closer from that of the lœss unit contemporary from GS-4 (~12–16%) than from those contemporary of GS-5.1 (~10-13%) and GS-3 (~9-11%). Given the single available OSL age and the similarity of the values of the sedimentary proxies between Renancourt 1 et 2, the attribution of [10b-10c] to the stadial GS-4 constitutes at this stage the most parsimonious hypothesis. In this case, a first important hiatus encompassing stadial GS-5.1 and interstadial GI-4 would take place between units [10a] and [10b] and explain the lack of continuity between the malacofaunas of MZ5 and MZ6. The second hiatus would take place between units [10b] and [11] and encompass interstadial GI-3 and stadial GS-3 pro parte, given the absence at Renancourt 2 of the 50-60 cm of homogeneous lœss deposits overlying the gley horizon contemporary of GI-3 at Renancourt 1 (Moine et al., 2021). It is interesting to note that an important hiatus covering almost the same time interval, i.e., GI-5.1 to GS-3 pro parte has been identified in the RP1 lœss sequence of Remagen-Schwalbenberg in the lower Rhine valley (Prud’homme et al., 2022).
[11]: This laminated lœss unit yielded two luminescence ages of 22.69 ± 2.5 ka (OSL 3) and 24.18 ± 2.35 ka (OSL 2), which are not statistically different. Laminated lœss are about 1.3-m-thick in section 2012-4, and 1.2-m-thick in a studied sequence at Renancourt 1, where they cover a homogeneous lœss whose top is dated to around 26.5-26 ka. They are overlain by the Nagelbeek Tongue Horizon (Paris et al., 2017; Moine, 2021), which has been dated around 23-22 ka in Belgium (van den Haute et al., 1998; Zens et al., 2018). This succession is typical at the scale of north-western Europe (Antoine et al., 2009, 2013, 2016; Haesaerts et al., 2016).

126As a consequence, [11] has been correlated with stadial GS-3, otherwise characterised by the highest Greenland dust concentration peaks (fig. 18) and lœss sedimentation rates (Bosq et al., 2023) of the last glacial period, reaching up to 1 m/ka (Antoine et al., 2014; Moine et al., 2017). It can be noted that these laminated lœss reach up to about 3.5 m in a neighbouring survey carried out in 2000 near the Grâce Valley (location marked PA2000 in fig. 2), which is consistent with the duration of nearly 4 ka of stage GS-3. Their abundance of molluscs there decreases from a few dozen individuals in the first metre to less than ten, with the disappearance of Vallonia costata and slugs in the following 2.5 m (Limondin-Lozouet, unpublished), which matches with their molluscan characteristics at Renancourt 2.

Sylvie Coutard

6.3 – Correlations of molluscan records

127First, as indicated above, only several molluscan records of variable quality are available for comparisons for the targeted time period (Middle Pleniglacial B: 36.5-30.5 ka). A few others exist but were not taken into account as old, often single and most of the time undated, assemblages, or unpublished recent records that still need partial revisions (see Moine (2014) for a compilation). Second, until about ten years ago, the SAVA Complex was almost never sampled as though to be systematically entirely decarbonated except in a very few identified sites. This is actually not the case at least for its upper part, hence systematic checks and the evolution in the sampling strategy.

128The taxa present at Renancourt 2 are similar to those of other regional molluscan records and consistent with the position of the site in the "western"malacobiogeographical domain, i.e., the least diverse and located west of a line joining Rouen (France) and Maastricht (The Netherlands) (Moine, 2014). Nevertheless, the presence of Vallonia costata at the base of the laminated lœss
[11] indicates the proximity, or a westward shift, of this boundary during their deposition. Thus, despite a microtopographic context at the edge of a plateau and a sedimentation context favourable to high mollusc abundance, the Renancourt 2 site does not stand out by the presence of particular taxa that would provide it a status of (cryptic) refuge area for extra-regional or regionally under-represented molluscan taxa. Yet this is the case of Renancourt 1, located approximately 170 m to the northwest (fig. 1) and characterised by the occasional presence of Clausilia rugosa parvula and Punctum pygmaeum (Moine et al., 2021) only known from the Rhine Valley during the Upper Pleniglacial (Moine, 2014).

129Normalised to 10 litres, i.e., multiplied by 4, the total abundance values are in the range of those of the Lohne soil and the two overlying gley horizons of Nussloch (Moine et al., 2008) and Sinzig soils of Remagen-Schwalbenberg (Schiermeyer, 2000) in Germany but much higher than those of all regional sites at the transition between the top of the SAVA Complex and the overlying gley horizon, which hardly exceed 400 individuals, e.g. Villiers-Adam (Limondin-Lozouet & Gauthier, 2003) and Havrincourt 1 (Moine, 2012), except Morcourt (Sambourg et al., 2025) though still not as abundant. The volume of Maisières-Canal samples being unknown (de Coninck, 1973), abundance comparisons are not possible with this site. The decarbonatation of unit 6a of Havrincourt 1 from 11 to ~1% towards its base explains this difference (Antoine et al., 2014). Although set in a similar plateau context, this explanation does not apply to unit 3 of Villiers-Adam, which remains carbonated enough to ensure shell preservation despite a downward decrease of the CaCO₃ content from 18% to 11% (Antoine et al., 2003b, Locht et al., 2003). Moreover, unit 3 may also predate the Renancourt 2 record, and not be comparable, as its single luminescence age precludes establishing precise correlations with Renancourt 2. Conversely, the very high abundances recorded at Renancourt 2, and to a lesser extent at Morcourt, may result from the sheltered context of the site at the foot of a SE exposed chalky leeward slope - as yet still undocumented except partially at Morcourt (Sambourg et al., 2025) - and possibly involving a better availability of CaCO₃ that favoured the mollusc population development and ensured the shell preservation.

6.4 – Implications for the age and environment of the Gravettian occupation

130From the chronostratigraphic and palaeoenvironmental synthesis presented above, it appears that the Gravettian occupation is associated with an interstadial phase, revealed by the malacological assemblage of the MZ3a-b sub-malacozones. This relative climatic improvement is consistent with the biological activity detected in the sediment and the presence of large fauna in the hunted fauna (horse, bison and reindeer). Nevertheless, the environment remained cold and open. Furthermore, unlike the archaeological level at Renancourt 1, the positioning of Renancourt 2 in relation to the evolution of the malacofauna composition is contemporary with a maximum abundance of molluscs and therefore with the interstadial optimum. This optimum is characterised only by a slight humidification of the environment, marked by a very slight increase in hygrophilous taxa. Conditions at Renancourt 2 were therefore locally much less humid during the optimum in question, which made the site much more habitable than during the optimum of the GI-3 interstadial at the Renancourt 1 site. At the latter site, occupation took place at the end of the stadial-interstadial transition (Moine et al., 2021).

131The dynamics of sediment deposition, through at least partially colluvial processes, cast doubt on the precise chronological attribution of the Gravettian occupation level taken from unit [8]. However, despite the difficulty in proposing a definitive dating and correlation of the archaeological level with interstadial GI-6 or GI-5.2 based solely on the ¹⁴C dates, the other ¹⁴C ages available have made it possible to establish the correlation of unit 7 with GI-7 and unit 9 with GI-5.2. The available data therefore tend to correlate the archaeological level with interstadial GI-5.2.

132Furthermore, ¹⁴C dating place the Renancourt 2 site within the same chronological range as the occupation of Havrincourt N2 (Pas-de-Calais), the few remains of Languevoisin (Somme) and the end of the Middle Pleniglacial record described at Sourdon (unpublished data). In each case, the pedosedimentary horizon concerned is a brownish to grey-brown silt, interpreted rather as an arctic meadow soil, preceding the establishment of a tundra gley associated with permafrost marked on the plateaus by large polygonal networks. At Havrincourt, Languevoisin and Sourdon, this brownish to grey-brown silt is clearly the top of the Saint-Acheul/ Villiers-Adam soil complex.

133In unit 6 of Havrincourt, ¹⁴C ages (tab. 9) on bones equivalent to those of Renancourt 2 were obtained for the archaeological level HAV2-N2: 28,100 ± 180 BP, i.e., 31,656-32,900 (32,078) cal BP, and 27,020 ± 140 BP, i.e., 30,971-31,290 (31,131) cal BP (Antoine et al., 2014). The mammalian fauna, mostly associated with human occupation, is characteristic of the biome known as «mammoth steppe» (horse, reindeer, bison, woolly rhinoceros), and also indicates an open environment with probably steppe vegetation (Goval & Hérisson, 2018). The low diversity of mollusc taxa indicates a low diversity of ecological niches and vegetation (steppe to steppe-tundra). The fauna of Renancourt 2 is similar with horse, bison and reindeer. The attribution of the three sites of Havrincourt 2, Renancourt 2 and Languevoisin tends towards an early Gravettian facies, characterised among other things by the exclusive use of soft stone hammer percussion.

134In the same chronological range, the Maisières-Canal site in Belgium offers many similarities with northern France. The main occupation is preserved at the bottom of the valley in humic loam that developed during a short "medium cold" interstadial episode (Haesaerts & Heinzelin, 1979). It has been dated by several radiocarbon measurements to around 28,000 BP (Jacobi et al., 2010). P. Haesaerts calls this episode the "Maisières oscillation" and a correlation with GI-5 of the NGRIP curve is proposed (Jacobi et al., 2010). This correlation, made before the GI-5.2/GI-5.1 doublet was identified, is based on radiocarbon dating, which implies the same margins of error and inaccuracies as for Renancourt 2. However, the humic horizon that yielded the Maisières-Canal industry, i.e., MC-4, underlies the gley known in Belgium as the "Harveng Horizon", characterised by large ice wedge pseudomorphs and correlated with the GI-5.1 interstadial (Haesaerts et al., 2016). In the synthetic regional stratigraphy of northern France, the Harveng Horizon is correlated with the gley supporting the network of large F-4 cracks correlated with the GI-5.1 interstadial (Antoine et al., 2016) and marking the end of the Middle Weichselian Pleniglacial (Vercelot et al., 2025). MC-4 is also preceded by the undated humic horizon MC-3 and by the humiferous horizon MC-2 underlying MC-3, which yielded a ¹⁴C age similar to that of unit [7] (Haesaerts et al., 2016). Its correlation with the GI-7 interstadial would therefore constrain that of MC-3 with the GI-6 interstadial and consequently support that of MC-4 with the GI-5.2 interstadial. Thus, even if some caution is warranted in the absence of characteristic pieces with sharp retouching, a comparison of the two sites in northern France (Havrincourt and Renancourt 2) with Maisières-Canal can be proposed on chronological, geographical and technological criteria (Paris, 2020, 2024).

135Indeed, radiocarbon dating place the site of Renancourt 2 within a chronological range that, culturally, encompasses the end of the Aurignacian and the beginning of the Gravettian, passing through specific regional industries interspersed between these two major technocomplexes, such as the Bayacian in south-western France and the Maisierian in north-western Europe. The most convincing comparisons based on chronological, geographical and technological criteria (Paris, 2024) are with the latter tradition, whose eponymous site is Maisières-Canal in Belgium (Haesaerts & de Heinzelin, 1979). The relationship between the Maisierian and the Gravettian is still under discussion (Touzé, 2019). Whatever the case may be, this tradition is currently only recognised in north-western Europe. Renancourt 2, to which we can add level N2 of Havrincourt 2 (Goval & Hérisson, 2018; Paris, 2024) and perhaps Languevoisin-Quiquery (Paris, 2024), both also in the Somme, thus fill a geographical gap between the Maisierian sites of Belgium, the British Isles and the Paris Basin.

136Finally, the Renancourt 2 site has enabled a very detailed study of palaeoenvironmental changes within a chronostratigraphic framework supported by numerous dating results. It provides major arguments for the discussion on the discontinuity of occupation during the Pleniglacial period in northern France. The Gravettian occupation of Renancourt 2 clearly took place during an interstadial climate improvement. This is also the case for the contemporary site of Havrincourt 2 and the more recent site of Renancourt 1 (Paris, 2024). This relationship has been established by correlating the position of archaeological levels with peaks in the abundance of bio-indicators, as well as the specific characteristics of stratigraphic units. Other Upper Palaeolithic sites in northern France discovered in stratigraphic positions are no exception to this rule (Paris, 2024).

7 – Conclusion

137The Renancourt 2 lœss sequence document the phases of sedimentation and pedogenesis and their link with palaeoenvironmental conditions at the very end of the Weichselian Middle Pleniglacial (Middle Pleniglacial B) and during Upper Pleniglacial periods (fig. 20). It provides the most accurate and best dated record for the end of the Middle Pleniglacial in northern France, since, in the absence of malacological records (often decarbonated contexts), the formation of pedo-sedimentary units from this period of low aeolian sedimentation rates has been little studied and remains to be understood in greater detail.

138In the lower part of the Renancourt 2 lœssic sequence, the rate of aeolian sedimentation appears to be low and the dynamics of deposition involve colluvial processes. Pedogenesis is poorly developed and the sediments are marked by the influence of bioturbation. Aeolian supplies and sedimentation rate increase towards the top of the sequence. The evolution of malacological assemblages highlights a succession of stadial and interstadial phases. Correlations are proposed between these successions, supported by dating, and reference curves from Greenland.

139Detailed study of the stratigraphic sequence, malacofauna and dating allows us to propose a hypothesis for the correlation between the Early Gravettian occupation of Renancourt 2 and rapid climatic fluctuations. The set of indicators tends to suggest a correlation with the interstadial GI-5.2 from Greenland ice record, but the ^¹⁴C dating does not allow us to reach a definitive conclusion. In any case, the detailed study of the Renancourt 2 sequence demonstrates the link between archaeological occupation and an optimum phase of interstadial climatic improvement. This site is a key element in demonstrating the discontinuous occupation of the region during the Upper Palaeolithic, in relation to short phases of climatic improvement and the concomitant presence of vegetation allowing the expansion of large herbivores. The difference in the precise timing of this human occupation during an interstadial period compared to that of Renancourt 1, during a stadial-interstadial transition phase, suggests that the settlement of human groups and the dynamics of land occupation depend (i) on classical local parameters (topography, raw material availibility…) but also (ii) on the specific dynamics of each of the millennial climatic oscillation, which implies differences in the palaeoenvironmental evolution and context.