Morphostratigraphy and Dating of Last Glacial Loess–Palaeosol Sequences in Northwestern Europe: New Results from the Track of the Seine-Nord Europe Canal Project (Northern France)

Abstract

The Hermies-Ruyaulcourt site (Pas-de-Calais), investigated within the “Canal Seine-Nord Europe” project, provides an exceptional record of pedosedimentary dynamics throughout the last interglacial-glacial cycle (Eemian–Weichselian). Eight stratigraphic trenches, correlated along 350 m, reveal several pedosedimentary units strongly influenced by local topography. This study combines sedimentological and micromorphological analyses with optically stimulated luminescence (OSL) dating. For OSL ages, a correction of the water content calculation protocol was developed, accounting for long-term moisture variations during burial. Nine OSL ages from humic horizons of the Early Glacial (MIS 5d-5a) and colluvial deposits of the Lower Pleniglacial (MIS 4) represent the first robust chronological dataset for these periods in northern France. Their internal consistency and agreement with existing thermoluminescence ages on burnt flints support their reliability. Moreover, geomorphological analysis highlights intense erosional phases which are interpreted as rapid permafrost destabilisation events linked to the melting of large ice-wedge networks around 60–55 ka and 30 ka (thermokarst erosion gullies). These investigations thus enable the chronology of the loess–palaeosols and the link with associated climatic events to be refined. This leads to a spatio-temporal model describing the evolution of Last Glacial environments in Western Europe, providing a robust reference for studying the Neanderthal occupation of the area.

1. Introduction

The loess–palaeosol sequences of the last interglacial–glacial cycle (Eemian–Weichselian) display a rich and complex stratigraphy, including numerous marker horizons that can be correlated across Western Europe [1,2,3,4,5,6,7,8]. In addition, the cyclostratigraphic analysis [9] and the dating of successive depositional units and interstratified palaeosols have demonstrated a climatic control expressed at the millennial scale of Dansgaard–Oeschger stadial–interstadial cycles [2,10,11,12,13,14,15,16,17]. Within this framework, the Weichselian loess–palaeosol sequences of northern France have been subdivided into four main chrono-climatic phases [3,18,19,20,21]: the Early Glacial (EG, ~112–70 ka), the Lower Pleniglacial (LPG, ~70–60 ka), the Middle Pleniglacial (MPG, ~60–30 ka), and the Upper Pleniglacial (UPG, ~30–15 ka), the main characteristics of which are summarised below.

Following the major erosional crisis that marks the end of the Eemian interglacial during the first pronounced cooling around 112 ka (Marine Isotope Stage MIS 5d/Greenland stadial GS-25), the Early Glacial (EG) represents a long transitional period (~40 ka) between the temperate conditions of the Eemian interglacial and the cold and more arid environments of the Weichselian Pleniglacial. In northern France, it is represented by a soil complex documented at several key sites such as Bettencourt-Saint-Ouen and Saint-Sauflieu (Figure 1). This complex of humic soils and colluvial deposits is particularly well preserved at the footslopes and in sediment traps linked to dissolution of the chalk bedrock, such as dolines where they can reach up to three metres in thickness [20,21,22]. Such trapping favoured the formation of cumulic or upbuilding soils, characterised by the gradual development of horizons under conditions of slow and continuous colluviation, enabling a detailed record of successive pedogenetic phases. This process also ensured the good preservation of Palaeolithic occupation layers and the spatial distribution of artefacts attributed to Neanderthal populations [22,23,24,25,26,27]. The sites of Hermies “Le Champ Bruquette” [23] and Bettencourt-Saint-Ouen [22] are outstanding examples. By contrast, fauna, plant remains (wood), and more generally organic material are rarely preserved in these contexts due to the relative acidity of the soils, which is unfavourable to perishable matter [28]. Molluscan assemblages are also rarely preserved because of the near-systematic decalcification of the deposits [20,29]. As for pollen, its record is particularly problematic in this region, as most levels are either barren or dominated by the most resistant taxa [22,29,30,31].

Only the site of Saint-Sauflieu yielded a representative pollen diagram produced by A.-V. Munaut, which is consistent with the pedostratigraphic record [30]. At this site the base of the Early Glacial humic complex (grey forest soil SS-1) is associated with boreal forest vegetation dominated by pine and birch, while its upper part (steppe soils SS-2, SS-3a, and SS-3b) reflects more open conditions, characteristic of birch steppe environments. However, because the main soils of the EG complex developed on colluvial deposits, luminescence dating has often produced inversions and significant under- or over-estimations [20,32,33]. Such inversions are common in colluvial soils because these deposits consist of reworked and incompletely bleached sediments, often transported as aggregates, and may mix materials with contrasting dose rates and moisture histories—processes known to generate scattered, under- or over-estimated luminescence ages.

Figure 1. Extent of Upper Pleistocene loess in Western Europe and limit of continuous permafrost during the Last Glacial Maximum (ca. 25–17 ka) [34,35], with the location of reference sites cited in this work (blue = France, orange = Belgium, grey = Netherlands, green = Germany). Map modified after [8].

Figure 1. Extent of Upper Pleistocene loess in Western Europe and limit of continuous permafrost during the Last Glacial Maximum (ca. 25–17 ka) [34,35], with the location of reference sites cited in this work (blue = France, orange = Belgium, grey = Netherlands, green = Germany). Map modified after [8].

The transition to the Weichselian Pleniglacial is marked by an erosional discontinuity at the top of the EG humic soil complex. This event, which announces the onset of the Lower Pleniglacial, is first expressed by the deposition of homogeneous, non-calcareous loess—the earliest evidence of intense aeolian activity and a major reduction in vegetation cover. This is overlain by stratified colluvium reworking the underlying soils, known as the “Hermies laminated colluvial deposits”, and then by a hydromorphic brown silty horizon at the top, referred to as the “Havrincourt brown silt” [3,21]. Finally, the first calcareous loess typical of the Lower Pleniglacial appears at about 65–60 ka.

The Middle Pleniglacial (MPG) is then characterised by a marked slowdown in loess sedimentation and the development of several pedogenetic horizons over almost 30 ka (ca. 60–30 ka) [3], giving rise to the Saint-Acheul/Villiers-Adam soil complex [20]. In contrast, from ~30.5 ka [14], the Upper Pleniglacial (UPG) is marked by a rapid and widespread increase in loess sedimentation rates across Western Europe, resulting in thick loess deposits alternating with tundra gley horizons [11,36,37,38]. A similar timing has recently been demonstrated for northern France using radiocarbon dating on earthworm calcite granules [38,39], previously tested and validated at Nussloch [14]

The synthesis of horizons defining these chrono-climatic phases highlights the fact that palaeoenvironmental and geochronological data for the Early Glacial and Lower Pleniglacial remain scarce and still represent a major challenge in the study of northern French loess sequences. In this context, rescue excavations carried out as part of the Canal Seine-Nord Europe (CSNE) project (https://www.canal-seine-nord-europe.fr/en/, accessed on 3 December 2025) provide a unique opportunity to access new sections and exceptionally well-preserved archaeological levels. Among the recently investigated sites, the Hermies area occupies a central place, as illustrated by the history of research conducted in this sector [23,40,41,42,43].

2. Hermies-Ruyaulcourt: From the First Palaeolithic Discoveries to Recent Research

The first Palaeolithic discoveries and stratigraphic descriptions in the Hermies–Ruyaulcourt area (Pas-de-Calais) date back to the early 20th century, during the construction works of the Canal du Nord. On that occasion, three main Mousterian sites were identified in the Bertincourt valley and at the locality of Bertincourt, where numerous flint artefacts (Levallois cores and flakes) were discovered by A. Salomon (1911–1913) [44,45,46,47,48]. These first observations already demonstrated the high archaeological potential of the area. Much later, between 1986 and 2003, investigations carried out in the same sector by L. Vallin and B. Masson confirmed the importance of these sites, revealing archaeological levels preserved within Weichselian loess and palaeosols horizons. These sites have since become key references for the Middle Palaeolithic of northwestern Europe [23,40,49,50,51]. In addition, their stratigraphic study made a major contribution to the establishment of the regional pedostratigraphic framework, in particular with the definition of the marker horizon known as the “Hermies laminated colluvial deposits” [3,21].

Since 2008, test-pits and preventive excavations carried out by INRAP along the initial route of the Canal Seine–Nord Europe (CSNE) have confirmed the relevance of this sector for reconstructing Weichselian palaeoenvironments [21,24,41,52]. In this context, the new track of the CSNE defined at the end of 2020 was the subject of a large-scale diagnostic operation, involving deep test-pits (10 m in depth) to assess the presence, extent, and preservation of Palaeolithic remains [42,43]. In the Hermies-Ruyaulcourt area (Pas-de-Calais), these investigations confirmed the presence of sedimentary horizons attributable to the last glacial period, in particular the Weichselian Early Glacial, as well as at least three major Palaeolithic sites that are currently under excavation (Figure 2A).

This study is based on stratigraphic data collected during a targeted investigation along the present Canal du Nord, conducted as part of a collaboration between CNRS (Centre National de la Recherche Scientifique), the CSNE company (Canal Seine-Nord Europe), and INRAP (Institut National de Recherches Archéologiques Préventives). The work focused on the north-facing slope, oriented east–west, bordering the Hermies valley, currently occupied by the Canal du Nord between Hermies and Ruyaulcourt (Figure 2A). The topographic surface gently descends eastwards, in relation to the development of an adjacent dry valley oriented southwest–northeast. Eight stepped trenches (T1–T8), spaced approximately 50 m apart, were opened along this slope (Figure 2B,C). The Upper Cretaceous chalk bedrock (Senonian) was reached at depths of 6–7 m. Three of these sections showing the most complete record of the Last Glacial were selected for sediment sampling. (T1, T3, and T6, Figure 3A). These investigations were conducted in parallel with a campaign of deep test pits (n = 52) carried out as part of the preventive archaeological evaluation of the area known as “Paléo 3” (Figure 2A) [42].

3. Methods

3.1. Stratigraphy and Sampling

A detailed stratigraphic survey at 1:10 scale (Figure 4) was carried out for each of the eight trenches. Correlation of the identified pedosedimentary units allowed the reconstruction of a 350 m long stratigraphic transect (Figure 5). To facilitate comparison between profiles, a correspondence table (

Table 1. Correspondence between field-based stratigraphic units and the synthetic units defined in Figure 5. This table provides the equivalence between the units observed and numbered individually in each profile (T1 to T8) and the synthetic pedosedimentary units (0–10) used in Figure 5. It serves as a guide to harmonise the nomenclature across figures and to facilitate comparison between profiles.

		Units— Figure 5	Units—T1	Units—T2	Units—T3	Units—T4	Units—T5	Units—T6	Units—T7	Units—T8
Holocene		0, 1	0, 1	0, 1	0	0	0, 1	0, 1	0, 1	0, 1
	Upper Plen.	2								2–8
Weichselian	Middle Plen.	3	2–5	2, 3	1	1		2	2–9
	Lower Plen.	4	6, 7	4			2	3
	Early Glacial	5, 6	8–13	5, 6	2	2	3, 4	4–6
Eemian		7		7	3
Saalian		8		8, 9	4–11	3–8	5–7
Chalky deposit		9	14	10	12	9, 10	8	7	10, 11
Chalk		10	S		S		S	S

) links the field-based units of each trench to the synthetic units defined in Figure 5. A total of 84 samples were collected from section T1 as a 420 cm-thick continuous column sliced into 5 cm samples (Figure 3C and Figure 6) for sedimentological analyses including grain-size distribution, magnetic susceptibility, and total organic carbon (TOC).

In addition, ten undisturbed blocks were extracted for thin-section preparation and micromorphological observations: eight from section T1 and two from section T6 (Figure 6). Finally, in order to establish a precise chronological framework, eleven samples were collected for optically stimulated luminescence (OSL) dating using steel tubes: seven from section T1, two from section T3, and two from section T6 (Figure 6).

3.2. Sedimentology

3.2.1. Grain-Size Distribution

Grain-size analyses were conducted at the Laboratoire de Géographie Physique (LGP) using a Malvern Panalytical laser granulometer (Mastersizer 3000; Malvern Panalytical Ltd., Malvern, UK). Samples were prepared by dispersion in a 5‰ sodium hexametaphosphate solution, agitated for 12 h, followed by sieving at 160 µm to remove coarse particles, following the protocol described in Antoine et al. [2,21]. The grain-size classes used in this study were defined by comparison with a set of reference samples analysed using the conventional pipette-sieving method. Two grain-size indices were applied: the Grain-size Ratio Index (IGR; 20–61 µm/<20 µm; [2]) and the U-ratio (20–61 µm/4–20 µm; [53]). High values of these indices indicate a dominance of coarse silt, interpreted as a signal of strong aeolian activity under cold and arid conditions [11,54,55].

3.2.2. Magnetic Susceptibility

Magnetic susceptibility measurements were performed at the Laboratoire de Géographie Physique (LGP) with a Bartington MS3 system equipped with an MSB probe (Bartington Instruments Ltd., Witney, Oxfordshire, UK. Software: BartSoft MS3 v.5.3), generating a magnetic field of 200 A·m⁻¹ and operating between 465 and 4650 Hz. Each measurement was repeated two to three times to ensure reliability. Low-frequency magnetic susceptibility (χlf) was investigated. In loess deposits, χlf values are generally low because of the dominance of quartz and carbonates. In contrast, palaeosols typically show higher values, reflecting pedogenic enrichment in maghemite and in situ formation of ultrafine magnetite [56,57,58].

3.2.3. Total Organic Carbon (TOC)

Total organic carbon was measured at the Laboratoire de Géographie Physique (LGP) using a ThermoScientific Flash 2000 CHNS elemental analyser (Thermo Fisher Scientific Inc., Waltham, MA, USA). Samples were first dried at 45 °C and finely ground, then precisely weighed before being sealed in capsules. Organic carbon was oxidised at high temperature in the presence of oxygen, releasing CO₂, whose concentration was measured by infrared detection. As the samples did not contain carbonates, no pre-treatment with acid was required.

3.3. OSL Dating

Eleven samples were collected in the field (Figure 6) for dating using the optically stimulated luminescence (OSL) method: seven from section T1 (T1-OSL1 to T1-OSL7), two from section T3 (T3-OSL1 and T3-OSL2), and two from section T6 (T6-OSL1 and T6-OSL2). Samples were taken with steel tubes (35 mm in diameter, 140 mm in length) hammered horizontally into the sediment with a non-rebound hammer. The tube ends were immediately sealed with opaque adhesive tape to prevent any exposure to light that could partially reset the luminescence signal.

Three of these samples (T1-OSL1, T1-OSL2, and T1-OSL3) were analysed at the Re.S.Artes laboratory (Emmanuel Vartanian, Bordeaux, France), while the remaining eight were processed at the Archéosciences Bordeaux laboratory, University Bordeaux Montaigne (UBM, Pessac, France), by the first author, under the supervision of Maïlys Richard. In both laboratories, OSL analyses were performed on the quartz fraction, following the Single Aliquot Regenerative dose (SAR) protocol ([59]; see Supplementary Materials). Luminescence signals were measured on multi-grain aliquots using green and/or blue optical stimulation. The aliquots were exposed to a sequence of regenerative doses interspersed with test doses, systematically applied to normalise the signal (

Table 2. SAR OSL protocol applied to quartz at Archéosciences Bordeaux.

Step	Treatment
1	Dose (for n = 0)
2	Preheat (240 °C for 10 s)
3 (Lx)	Green stimulation (125 °C for 100 s)
4	Test dose (24 Gy)
5	Cutheat (200 °C for 0 s)
6 (Tx)	Green stimulation (125 °C for 100 s)
7	Return to step 1

and

Table 3. OSL SAR protocol applied to quartz by Re.S.Artes.

Step	Treatment
1	Dose (for n = 0)
2	Preheat (240 °C, 5 °C for 10 s)
3 (Lx)	Blue stimulation (125 °C for 40 s)
4	Test dose (7 Gy)
5	Cutheat (160 °C for 10 s)
6 (Tx)	Blue stimulation (125 °C for 40 s)
7	Return to step 1

). In this study, we focused on dating quartz using OSL because its signal bleaches more efficiently than the one from the feldspars (especially for the pIRIR protocols), and it does not suffer from fading. A Bayesian age–depth model was then constructed using the software Chronomodel v3.2.7 [60].

3.3.1. Equivalent Dose Determination

Protocol Used at Archéosciences Bordeaux

The sediments were prepared and measured in controlled conditions under subdued red light, following standard procedures for 20–40 µm quartz grains [59,61,62]. We used a silicon-based stamp that produced a ~1 mm-diameter spot of adhesive on each disc that was covered with quartz grains. Luminescence data were treated using Analyst v. 4.57 [62].

Equivalent doses (D_e) were determined using the Single Aliquot Regenerative dose (SAR) protocol (Table 2) on a Freiberg Lexsyg Research TL/OSL reader (Freiberg Instruments GmbH, Freiberg, Germany) [63], equipped with a calibrated β source delivering 0.120 Gy/s for the 20–40 µm fraction. Green-light stimulation at 125 °C was used throughout the SAR sequence. A preheat plateau test combined with a dose recovery test (DRT) was conducted on sample T1-OSL6, using preheat temperatures of 240 °C and 260 °C (step 2 of Table 2). The recovered/given dose ratios (1.01 ± 1% at 240 °C and 0.92 ± 5% at 260 °C) validated the use of 240 °C for all subsequent measurements. An infrared stimulation test, inserted between steps 2 and 3 of the SAR protocol, confirmed the absence of feldspar contamination. Multigrain aliquots were selected using the following criteria: a recycling ratio within 10% of unity, a recuperation < 5% of the natural signal, a test-dose error < 10%, and a test-dose signal > 3σ above background.

Eight to ten discs per sample were analysed using the SAR protocol and used to construct a global growth curve (GGC). The L_n/T_n of 20 additional aliquots was measured per sample and interpolated on the global growth curve [64] (Figures S1–S3, Table 1, Table 2 and Table 3). The Central Age Model (CAM; [65]) was computed on the whole set of D_e (28 to 30 values per sample).

Protocol Used at the Re.S.Artes Laboratory

Samples T1-OSL1 to T1-OSL3 were prepared and measured in controlled conditions under subdued red light, following standard quartz purification procedures [59,61], using the 80–125 µm fraction. Equivalent doses (D_e) were determined using the SAR protocol described in Table 3, on a Risø TL/OSL DA-15 reader (DTU—Danish Technical University, Roskilde, Denmark) equipped with blue diodes (470 ± 40 nm) and a calibrated β source delivering 0.0688 Gy/s.

A dose recovery test (DRT) was carried out on the three samples (T1-OSL1, T1-OSL2, T1-OSL3), using a fixed preheat of 240 °C (step 2 in Table 3). The measured-to-given dose ratios obtained (1.02 ± 0.10, 0.98 ± 0.08, and 0.95 ± 0.16) validate the use of a 240 °C preheat. Multigrain aliquots were retained only when satisfying the standard acceptance criteria (recycling ratios within 10% of unity, recuperation < 5% of the natural signal, test-dose errors < 10%, and test-dose signals > 3σ above background). No infrared signal was detected, confirming the absence of feldspar contamination.

In total, eighteen discs were analysed per sample. For each aliquot, a full SAR dose–response curve was constructed by applying a sequence of regenerative doses interspersed with test doses, and the natural-to-test dose luminescence ratio (L_n/T_n) was interpolated directly on the individual growth curve to obtain D_e (Figure S4, Table S4). All accepted aliquots (18 per sample) were first used to calculate CAM ages. For the minimum ages, the lowest D_e values were retained (4 aliquots for T1-OSL1 and T1-OSL3, and 3 aliquots for T1-OSL2). This selection was based on a combination of a χ² outlier test and visual inspection of the D_e histogram, which consistently identified a small cluster of lowest-dose aliquots interpreted as the best-bleached population. The choice of proposing minimum ages is detailed in Section 4.4.2.

3.3.2. Annual Dose Determination

Field (In Situ) Gamma Spectrometry

For samples T1-OSL1, T1-OSL2, T1-OSL3, T3-OSL1, and T3-OSL2, in situ gamma spectrometry measurements were performed using a portable gamma-ray multichannel analyser connected to a NaI(Tl) detector (Mirion Technologies (Canberra), Meriden, CT, USA), inserted into 30 cm deep auger holes (J.-J. Bahain, MNHN, Paris). The data were processed following the “threshold” technique [66]. These measurements provided the gamma dose rate to which the quartz grains were exposed, corresponding to the radioactive decay of natural isotopes from the uranium (²³⁵U and ²³⁸U) and thorium (²³²Th) decay chains, as well as potassium (⁴⁰K).

Laboratory Gamma Spectrometry and Dose-Rate Calculation

For samples T1-OSL4 to T1-OSL7, and T6-OSL1 and T6-OSL2, laboratory gamma spectrometry was carried out to determine U, Th, and K concentrations. The dose rate was estimated from sediments collected around the OSL tubes, sealed in hermetic gamma boxes (60 cm³), and measured by high-resolution gamma spectrometry (BEGe) (Ortec, Oak Ridge, TN, USA) after radon re-equilibration. Dose rate calculations were based on the conversion factors of Guérin et al. [67], beta attenuation coefficients of Guérin et al. [68], alpha attenuation factors of Brennan et al. [69], a s-a value of 4.5 ± 20%, and the cosmic dose rate estimation was calculated after Prescott and Hutton [70]. Ages are reported at 1σ. Water content corrections were applied using values discussed in the following section (Section 3.3.3).

3.3.3. Modelling of Moisture Content

According to the protocol generally adopted, the moisture values used to calculate OSL ages are those measured in the field or in the laboratory. These values therefore reflect the present-day moisture of the sample, which does not necessarily represent its entire hydrological history. In this study, to improve the reliability and the accuracy of the dating, the long-term history of moisture content was modelled over the whole burial period in order to approximate a more geologically realistic value. For this purpose, precipitation data from the Nussloch reference sequence (Germany) [71] were used to estimate indicative mean moisture values for each major climatic phase of the Last Glacial in Western Europe: Early Glacial, Lower Pleniglacial, Middle Pleniglacial, Upper Pleniglacial, and Lateglacial. During the Last Glacial, regional precipitation differences between Nussloch and Hermies were far less pronounced than today: both sites were subject to a continental climate, and the difference in distance from the coasts was strongly reduced due to lower sea level (dried out Channel and North Sea basins). It therefore seems reasonable to use the trends observed at Nussloch as a reference for northern France during the Last Glacial. However, in the present-day (Holocene), Hermies is much more affected by oceanic influence than Nussloch. Holocene precipitation and moisture values were therefore derived from current regional data available for northern France [72,73]. The dataset used here corresponds to reconstructed annual precipitation at Nussloch, estimated from δ¹³C [71]. For each major climatic phase of the Last Glacial, a weighted average was calculated to obtain a representative value of mean precipitation (mm/yr) per phase.

These mean precipitation values (mm/yr) were then converted into moisture content (%) using a simple proportionality relationship, taking as reference current values measured in northern France loess sequence (Normandy): mean precipitation of 727 mm/yr [73] for a moisture content of 19% [72]. This approach therefore relies on the assumption of a simplified linear relationship between precipitation and soil moisture. The resulting moisture values were integrated as a time-weighted average according to the duration of each climatic phase, and the final values were applied to the corresponding samples based on their stratigraphic attribution.

4. Results

4.1. Stratigraphy and Morphostratigraphy

4.1.1. Description of Units

The description of the stratigraphic units from profiles T1, T3, and T6, which were the subject of sedimentological and chronological analyses, is presented in

Table 4. Description of pedosedimentary horizons of T1, T3, and T6.

	Profile T1		Profile T3		Profile T6
Unit No.	Description	Unit No.	Description	Unit No.	Description
0	Clayey greyish−brown silt with a granular structure.	0	Clayey greyish−brown silt with a granular structure.	0	Clayey greyish−brown silt with a granular structure.
1	Brown clayey silt with numerous recent roots.	1	Lightly clayey beige silt, structureless.	1	Reddish−brown clayey silt.
2	Dark beige silts with doublets, including grey−white centimetric layers, and small Fe−Mn concretions.	2.a	Stratified silt of unit 1 with grey−brown humic material. Bioturbations, with a horizontal silt layer at −16 cm.	2	Lighter, more massive, greyish silt with lower clay content.
3	Grey−beige hydromorphic horizon with flowed structure and scattered orange oxidized streaks. The base rises towards the right−hand side of the profile.	2.b	Brownish humic horizon with burrows filled with yellowish−beige silt.	3	Laminated silts with grey, brown, beige, and white layers within a dark beige loess matrix. Visible lenses.
4	Beige silt, very dense and compact. Structureless, with a few roots. Presence of a grey centimetric band at ~−1.56 m depth, possibly a relict gley horizon.	3	Reddish−orange silt with structure, containing humic fragments at the top.	4.a	Brown−grey structured silt with horizontal siltans extending over 5 cm.
5	Beige−brown clayey silt with Fe−Mn concretions, mainly visible at the base.	4	Mixed silty horizon.	4.b	Greyish silt with brown horizontal and rightward−oblique laminae, disappearing towards the base.
6.a	Alternation of brown−beige and grey−white silts. The beds follow the concave shape of unit 6b, though to a lesser extent: they dip by 5–10 cm per metre (local deformation).	5	Beige and white silts with doublets, poorly defined.	4.c	Bleached horizon with pellets and oblique bands.
6.b	Alternation of beige and white silts with ferric illuviation features/stains at the base. At the base, the beds are horizontal, while at the top they dip towards the center with amorphous deposits and elongated blocks. Evidence of unconformities, cross−bedding, faults, and local deformations.	6	Silts with doublets, very diffuse silt, with Fe−Mn microconcretions.	5	Brownish silty horizon with silt pellets at the top. Brown to dark−brown clayey silt with structure.
7	Light beige−yellow silt, compact and dense, structureless. Presence of ferric stains and Fe−Mn microconcretions. Two thin black bands at the base.	7	Beige silts with well−defined doublets, Fe−Mn microconcretions and ferric oxides. Doublets become finer towards the base.	6	Lighter clayey silt (siltan?), orange−coloured.
8.a	Top of the humic horizon: dark−brown silt with closely spaced black beds, silt veins and ferric illuviation features. The top surface dips to the left.	8	Light−brown slightly clayey silt, with very diffuse silt at the base. Fe−Mn microconcretions.	7	Flint gravel within the matrix of unit 6.
8.b	Brown−greyish silt.	9	Yellow−brown silt with abundant diffuse silt. Homogeneous matrix, slightly structured.	S	Chalk bedrock at 4.30 m.
8.c	Base of the humic horizon: brownish−grey silt with millimetric alternations. Apparent grain−size grading. Presence of burrows and pellets filled with material from units 9 and 10.	10	Brown to brownish silt, Fe−Mn microconcretions mainly at the base. A few grey bands, possibly corresponding to ghost gley horizons.
9	Silty horizon, mixed with the underlying unit at its base and with the overlying unit at its top. Presence of burrows filled with humic pellets from unit 10.	11	Yellow silt with frost−deformed beds; large Fe−Mn concretions at base.
10.a	Compact orange−brown fine sandy silt, with small burrow casts at the top and larger burrows.	12.a	Flint gravel in a loessic matrix, with few Fe−Mn features.
10.b	Transition phase: lighter orange−brown silt. Structured white silt coating, with small burrow casts.	12.b	Dark−brown clayey matrix with Fe−Mn accumulation.
11.a	Homogeneous brown humic silt with root traces and a few pebbles.	S	Chalk bedrock, 70 cm below the gravel layer.
11.b	Brownish−brown slightly clayey silt with prismatic structures and large Fe−Mn concretions.
12	Mixed silty horizon with Fe−Mn concretions.
13	Light−brown sandy silt with polyhedral structure, reddish−brown coatings, and pores filled with white silt. Bioturbation and Fe−Mn concretions.
14	Heterometric flint gravel in a silty−clayey matrix. Fe−Mn concretions. Flint material: frost−shattered, knapped, worn, and patinated. Towards the chalk bedrock, less evidence of wear and more rounded shapes.
S	Chalk bedrock at 7 m.

. Four knapped flints (debitage flakes) were discovered in situ: two in profile T1 (unit 12 and the top of unit 11), one in profile T3 (top of the gravel bed, unit 12), and one in profile T6 (top of unit 5). The description of the units from the other profiles is presented in

Table 5. Description of pedosedimentary horizons of T2, T4, T5, T7, and T8.

	Profile T2
Unit No.	Description
0	Topsoil.
1	Brownish-orange clayey silt.
2	Beige silt with some oxides and hydromorphic features at the top.
3	Brown clayey silt, compact and weakly structured, with mm-scale Fe-Mn concretion.
4	Grey-yellow silt with silty bedding.
5	Humic silt with well-defined mm- to cm-scale silty bedding at the top.
6	Brown-grey humic silt with more or less diffuse silty beds and burrows, especially at the base.
7	Brown-orange clayey silt with polyhedral structure and reddish illuviation features.
8	Beige-orange clayey silt with polyhedral structure.
9	Brown/white silts with doublets (silt) showing strong deformation.
10	Flint gravel (partly frost-fractured); (10.a) brown loessic matrix, ~30 cm, mudflow deposit; (10.b) grey–blue hydromorphic fine-sand matrix with oxidation, ~30 cm.
	Profile T4
Unit No.	Description
0	Topsoil.
1	Brown-grey silt disturbed by greyish bands.
2	Brown-orange clayey silt.
3	Silts with thick doublets (clayey, orange); lamellar silt structure.
4	Brown silt with diffuse white silt.
5	Silts with thick doublets, including 1–2 cm silty beds within an orange-beige matrix.
6	Grey-beige loess with less distinct doublets.
7	Grey-orange hydromorphic silt; convolute deformations.
8	Beige clayey loess.
9	Heterometric frost-fractured and weathered flint gravel in a brownish silty matrix, with Fe-Mn grains more abundant towards the base.
10	Chalk slope deposit composed of chalk, flint, and silt.
	Profile T5
Unit No.	Description
0	Topsoil.
1	Orange-brown clayey silt, structured.
2	Grey and light-brown laminated silts; small vertical white bioturbations. Stratification fades towards the base due to bioturbation.
3	3.a Dark-brown humic clayey silt with minor clay coatings. 3.b Humic clayey silt with abundant clay coatings; bioturbation (burrows) at the base. 3.c Grey-brown compact silt with burrows; Fe-Mn features; diffuse clay coatings. 3.d Chocolate-brown silt, polyhedral to subpolyhedral, with pockets, burrows, bioturbation, and networks of cracks and microcracks. A flint gravel horizon marks the contact between units 4 and 3.d.
4	Reddish brown-orange clayey silt, polyhedral, with numerous veins and fissures, visible illuviation features.
5	Lighter brown-orange silt, less clayey, less structured and less compact than unit 4.
6	Massive grey silt with oxides and clay coatings, 7–8 cm thick.
7	Orange silts with thick doublets, cut by silty veins in the upper part, with ferric illuviation features.
8	Gravel layer = loess-derived mudflow deposit.
S	Chalk with a very steep dip to the left.
	Profile T7
Unit No.	Description
0	Topsoil.
1	Brown-orange clayey silt, weakly structured: crumbly to polyhedral; bioturbation, small burrow casts, earthworm channels, and some illuviation features.
2	Brown-beige clayey silt with a slight greyish tint and faint oxidation traces at the top.
3	Light-grey highly hydromorphic silt with horizontal and slightly oblique orange oxides; abundant Fe-Mn and oxides at the top (possibly organic matter?).
4	Brown-grey clayey silt with granular to polyhedral structure and orange oxides.
5	Yellowish clayey silt; orange oxides.
6	Brown-yellow loess with fine black beds, micro- to mm-scale (Fe, organic matter?), and Fe-Mn features.
7	Two greyish hydromorphic horizons at the base of unit 6, about 5–7 cm thick.
8	Slightly orange-yellow loess with Fe-Mn features concentrated in the lower 20 cm.
9	Yellow silt, slightly more clayey and structured than the overlying unit, with Fe-Mn features.
10	Gravel in unit 9 matrix; transition to brown variably clayey silt with chalk concretions; chalk slope deposit.
	Profile T8
Unit No.	Description
0	Topsoil.
1	Brown-grey silt with pellet structure and a Nagelbeek tongue-shaped horizon at the base.
2	Beige-yellow calcareous silt, finely bedded with mm- to cm-scale grey/light-brown layers; very powdery, with undulations and occasional orange oxide spots.
3	Bedded calcareous silt with a strong predominance of grey silt and some deformations.
4	Bedded beige calcareous silt, predominantly pale grey, with abundant orange oxidation. Bedding deformations and freeze–thaw fractures preferentially oriented to the right. Deformation and fractures fade towards the base. Basal boundary diffuse.
5	Light beige bedded calcareous silt with beds slightly thicker than in unit 4. Orange oxidation and Fe-Mn features. Diffuse upper boundary.
6	Hydromorphic mottled grey–orange calcareous silt with abundant orange oxides.
7	Yellow-beige calcareous loess forming a mottled matrix with several gley horizons. Gleys become thinner towards the top and the base. A yellow loess band, not fully continuous, occurs between 3.08 and 3.14 m.
8	Grey hydromorphic calcareous silt.

. All profiles are composed of non-calcareous loess, except for the calcareous unit T8.

4.1.2. Morphostratigraphy

The eight stratigraphic trenches (Figure 2B) excavated between Hermies and Ruyaulcourt allow the identification of several pedosedimentary complexes, organised along a 350 m longitudinal transect (Figure 5) parallel to the Canal du Nord.

At the base of the profiles, the Upper Cretaceous chalk bedrock was observed in situ in sections T1 and T3 (unit 10, Figure 5). However, in mid-slope and lower slope positions, the chalk appears reworked, occurring as small blocks included within a heterometric flint gravel with a highly heterogeneous silty matrix (notably observed in T2). In addition, the base of section T4 consists of a chalky deposit containing a few flints overlain by the gravel. These observations reflect repeated episodes of solifluction and associated erosion on slopes (unit 9, Figure 5).

Overlying these deposits, calcareous loess form a dome centred between T2 and T3 (unit 8, Figure 5), showing both laminated and homogeneous facies, and reaching up to 3 m in thickness in zones less affected by erosion (T3). These deposits are capped by a Bt horizon of a leached brown soil (Luvisol), attributed to the Eemian interglacial (unit 7, Figure 5), as it underlies directly the succession of soils and loess corresponding to the Weichselian.

The humic soil complex, resting on the interglacial Bt horizon, is attributed to the Weichselian Early Glacial and represents a central stratigraphic marker (unit 5, Figure 5). It is particularly well preserved in lower slope positions (T1, T5, T6), where it exceeds 2 m in thickness, illustrating the importance of local sediment traps that favoured its preservation. In contrast, on higher slope positions (T3, T4), this complex is poorly preserved or even entirely eroded as in T4.

The overlying deposits, consisting of finely laminated colluvial silts of variable thickness (unit 4, Figure 5), mark a major morpho-sedimentary discontinuity above the humic horizons. Based on their facies and stratigraphic position, they can be correlated with the Hermies laminated silts, attributed to the Lower Pleniglacial. Derived from the erosion of the underlying deposits, they accumulated along the slopes (T1, T5, T6) and form a finely laminated unit, rich in periglacial features such as micro-faults, small frost cracks, and deformed laminae (T1, Figure 6, Table 4). These sediments were deposited by episodes of intense runoff (cross-bedding and erosion gullies) on bare soils under rigorous climatic conditions.

An unconformity separates the laminated colluvial silts from the overlying brown clayey loess horizons (unit 3, Figure 5), which are associated with weakly developed soils (Bw horizon of Cambisols) and preferentially preserved in lower topographic positions. In the eastern part of the transect, they reach almost 4 m in thickness in profile T7, where several greyish-green, non-calcareous tundra gley horizons are interbedded in the brown loam complex. In section T8, where the bedrock was not reached, only stratified typical calcareous loess exceeding 3 m in thickness is recorded (unit 2, Figure 5). The contact between units 2 and 3 is marked by a sharp sedimentary discontinuity.

The chronostratigraphic attribution of units 3 and 2 to the Middle and Upper Pleniglacial is based on their stratigraphic position and on comparisons with two nearby sequences where some luminescence ages are available. At the Palaeolithic site of Tio-Marché, located at about dozen metres to the south of the transect between T6 and T7, two TL dates on sediment provided ages between 54 and 56 ka at the base of the brown clayey loam complex [74]. At the Hermies-Cimetière site, located just opposite T8 on the other side of the Canal du Nord, an archaeological level preserved in a tundra gley horizon was dated by thermoluminescence on heated flint to 31.3 ± 2.1 ka [75]. The same gley horizon was more recently dated at about 30.5 ka cal. BP using ¹⁴C on earthworm calcite granules (Moine, written comm.), allowing its correlation with the base of the Weichselian Upper Pleniglacial loess sequence [3].

Finally, the profiles are capped by a Bt horizon of a leached brown soil (Luvisol), developed on the Pleniglacial loess during the Lateglacial and the Holocene (unit 1, Figure 5). This soil is particularly well developed in sections T1 and T6, although its upper part is locally truncated by ploughing or by more recent erosional processes linked to ongoing works. At higher slope positions (T2 and T3), the Bt horizon is eroded, leaving only the Ap surface horizon resting unconformably on Weichselian levels (unit 0, Figure 5).

4.2. Sedimentology

The results of grain-size, total organic carbon (TOC), and magnetic susceptibility analyses from profile T1 at Hermies are presented in Figure 7. These data provide a detailed characterisation of the different loess levels and soil horizons.

Unit 6, corresponding to a thick horizon of laminated silts, displays the lowest clay contents (14–20%) and the highest IGR and U-ratio values, ranging from 1 to 2 and 3 to 7, respectively. In contrast, the base of the sequence (units 13 to 8) is characterised by higher clay contents and low IGR and U-ratio indices, reflecting a predominance of fine particles, particularly clays, associated with more pronounced pedogenetic activity.

In T1 three main pedological horizons can be distinguished: units 13, 11, and 8. At the base of the sequence, unit 13 shows high clay contents (19–21%) and low TOC values (ca. 0.2%), typical of the Bt horizon of a leached brown soil. Unit 11 exhibits the highest clay values, with a progressive enrichment from base to top (19–22%) and two TOC peaks at 0.45%, indicating significant pedogenic development characteristic of the Bth horizon of a grey forest soil. Unit 8, darker-toned, also shows a slight increase in clay content (15–17%) and two distinct TOC peaks in sub-horizons 8a (0.7%) and 8c (0.6%), with significantly lower values in sub-horizon 8b (0.4%). These values are consistent with an Ah isohumic horizon (Chernozem).

Within this complex, unit 10 shows markedly different characteristics: indeed, clay contents, close to 20% at the base, decrease progressively to 15% towards the top. A slight increase in coarser particles is also observed (moderate rise in IGR and U-ratio indices), together with a significant decrease in TOC values (<0.2%). These trends contrast with the underlying and overlying horizons and suggest that unit 10 is redeposited material, more consistent with a colluvial horizon.

In parallel, a strong and progressive decrease in magnetic susceptibility (from 33 × 10⁻⁸ m³/kg to 5 × 10⁻⁸ m³/kg) is observed from unit 11 to unit 8. This trend reflects decreasing pedogenetic activity, as the formation of ferromagnetic minerals (maghemite, ultrafine magnetite) was most intense at the base of unit 11. Conversely, the upper horizons may have been depleted by leaching or diluted by aeolian inputs poor in magnetic minerals. An opposite trend is observed in unit 13, where magnetic susceptibility values increase from base to top (14 × 10⁻⁸ m³/kg to 33 × 10⁻⁸ m³/kg). The humic horizon 8a records a minimum of 5 × 10⁻⁸ m³/kg, the lowest value of the sequence. This feature is probably linked to marked hydromorphic conditions observed during fieldwork (greyish reduction patches and oxidation tracks) that are responsible for the dissolution of ferromagnetic minerals due to water saturation [58].

The brown-beige clayey silt of unit 5 shows, in its lowermost 20 cm, a progressive increase in clay content reaching a maximum of 17%, in contrast with the underlying loess deposit (unit 6). This evolution is accompanied by a slight enrichment in TOC (0.2%). These trends indicate the development of a moderately expressed pedogenic horizon of interstadial type [20,21].

4.3. Micromorphology

Observation of the thin sections (Figure 6) provides detailed insights into the pedogenetic and periglacial processes that affected the studied horizons. Although the presence of resin residues of varying abundance (manufacturing defect) hindered high-resolution observations, clay illuviation dynamics, evidence of biological activity, and freeze–thaw features could be distinguished. The thin sections are described from the top to the base of profile T1. These interpretations rely not only on the observations made on the Hermies sections, but also on comparison with published micromorphological studies of regional loess–palaeosol sequences, where similar pedosedimentary features recur consistently within specific horizon types [21,76].

U6–T1. Thin section T1-B1 (Figure 3F), extracted from the laminated silts of unit 6b, shows a laminated microstructure with clear grain-size sorting, reflecting strong hillwash processes and repeated freeze–thaw phases. The presence of small ovoid aggregates indicates reworking by surface runoff. Small organic aggregates are also observed. These features are consistent with slope processes acting under sparsely vegetated conditions.

U8–T1. Thin sections T1-B2 (unit 8a, Figure 3G), T1-B3 (unit 8b), and T1-B4 (unit 8c, Figure 3H) show a platy microstructure with traces of bioturbation. Sections B2 and B3 contain numerous elongated voids, reflecting high porosity that decreases with depth towards B4. In B4, the platy structure is clearer at the top, while the base is much more disturbed, with ferruginous hypo-coatings. Angular ferruginous papules with reddish-brown tones are also present. These features are comparable to those observed in thin sections T6-B1 (unit 4b) and T6-B2 (unit 4c) from profile T6 (Figure 6), which also display a platy microstructure with well-developed channels. In this profile, B1 shows traces of biological activity as well as fissures related to freeze–thaw alternations. The dominant tones of the clay coatings are orange (ferri-argillans) and mahogany brown, and reworked orange papules are also present. However, the absence of similar orange ferri-argillans in profile T6 suggests that they represent reworked relics of an earlier soil.

The high porosity, linked to bioturbation and the presence of abundant voids, indicates intense biological activity (roots, microfauna). The absence of clay illuviation suggests weak eluviation processes associated with dry conditions, whereas fissures related to freeze–thaw alternations point to cold-climate processes.

U10–T1. Thin section T1-B5, from unit 10a, shows a subangular blocky microstructure [77]. The sediment is composed mainly of angular to spherical quartz grains of about 30 μm. The voids are disorganised. The groundmass consists of very light quartz silt embedded in a clayey matrix with yellow birefringence. Traces of orange and mahogany-brown illuvial clay coatings (ferri-argillans) are present. The strongly disordered arrangement of the illuviations—some fractured, others partly integrated into the groundmass—suggests a reworked horizon. Traces of biological activity also cross-cut the thin section, indicating post-depositional dynamics.

U11–T1. Thin sections T1-B6 (top, Figure 3I) and T1-B7 (base) from unit 11 show a more pronounced platy microstructure at the top than at the base. Well-developed channels (biotubules) are present in both sections, often associated with clay coatings, reflecting multiple phases of illuviation. Some coatings display tones from mahogany brown to blackish, suggesting incorporation of organic matter. In situ coatings vary from mahogany brown to orange, typical of ferri-argillans, clearly visible in the channels and locally filling voids. In both sections, the distribution of clay is heterogeneous. Some illuviations show fragmentation and are cross-cut by ice-segregation cracks, reflecting freeze–thaw processes and post-pedogenetic reorganisation. At the base, grains of sand, flint fragments, and a fibrous ferruginous concretion probably formed around a plant remain are observed. The brownish red and blackish clay coatings indicate illuvial pedogenesis with incorporation of organic matter typical of greyzem (Bth). This soil type today is associated with forest-steppe environments and continental cool climates. Their fragmentation and the presence of ice-segregation cracks demonstrate subsequent reworking by periglacial processes.

U13–T1. Thin section T1-B8 (unit 13, Figure 3J) shows numerous large channels with well-developed clay illuviations (ferri-argillans) in brownish-orange to orange tones, visible in biological channels and inter-aggregate voids. Some of these orange illuviations are microlaminated, alternating with silty layers similar to the groundmass [77]. The thin section is strongly bioturbated, with orange aggregates and organic matter. These features indicate marked illuvial pedogenesis, in conditions more temperate and developed than those inferred for U11–T1. The horizon is an intergrade between a Bt and a Bth horizon, i.e., BSO soil.

4.4. Dating

4.4.1. Modelling of Moisture Content

The results (Figure 8) are consistent with expectations from broad-scale bioclimatic models based on soil-climate relationships [77,78,79,80]. For the moderately developed soils of the Early Glacial, Middle Pleniglacial, and Late Glacial, mean reconstructed precipitation values reach 327, 360, and 339 mm/yr, respectively, in agreement with the expected 300–450 mm/yr range. For the colder and more arid environments of the Lower and Upper Pleniglacial, the respective values of 296 and 278 mm/yr fall within the expected 250–400 mm/yr range for steppe-tundra vegetation. They are also consistent with the results of Prud’homme et al. [17], which range between 200 and 500 mm, mainly concentrated between 300 and 400 mm, for the Weichselian Pleniglacial at Remagen-Schwalbenberg (Lower Rhine Valley, Germany) and Nussloch (Middle Rhine Valley, Germany). Finally, the Holocene average of 772 mm/yr is consistent with expectations for temperate humid biomes (>700 mm/yr). The resulting long-term moisture estimates are presented in

Table 6. Results of Optically Stimulated Luminescence (OSL) dating from the Hermies site. Underlined OSL ages are from the Re.S.Artes laboratory and regular ages are from Archéosciences Bordeaux. nGGC = number of aliquots measured with the SAR protocol used to construct the GGC. nL_n/T_n = number of L_n/T_n signal measured and plotted on the GGC. n/N = number of accepted aliquots/total number of measured aliquots. Both CAM and minimum age estimates (corresponding D_e are shown in italics) are reported for T1-OSL1, 2 and 3.

A
Sample	Depth (m)	Units from Figure 5	Grain size (μm)		Dose rate (μGy/a)		Total
				alpha	beta	gamma + cosmic
T1-OSL1	5.41	5	80−125	0	1471 ± 94	1091 ± 55	2562 ± 108
T1-OSL2	4.23	5	80−125	0	1503 ± 93	1145 ± 58	2648 ± 109
T1-OSL3	3.9	4	80−125	0	1583 ± 91	1072 ± 54	2655 ± 106
T1-OSL4	5.89	5	20−40	222 ± 31	1805 ± 15	1321 ± 12	3348 ± 37
T1-OSL5	4.9	5	20−40	197 ± 28	1689 ± 15	1219 ± 12	3104 ± 34
T1-OSL6	3.46	4	20−40	180 ± 26	1582 ± 15	1159 ± 12	2921 ± 32
T1-OSL7	2.35	4	20−40	208 ± 30	1707 ± 15	1307 ± 12	3221 ± 35
T3-OSL1	3.2	8	20−40	202 ± 29	1944 ± 15	1375 ± 63	3522 ± 71
T3-OSL2	1.73	8	20−40	213 ± 30	1890 ± 15	1322 ± 59	3424 ± 68
T6-OSL1	3.51	5	20−40	210 ± 30	1632 ± 15	1264 ± 12	3106 ± 35
T6-OSL2	2.92	5	20−40	200 ± 28	1639 ± 15	1244 ± 12	3083 ± 34
B
Sample	U (ppm)		Th (ppm)		K (%)	Water content (%)
T1-OSL1	2.99 ± 0.49		9.83 ± 0.42		1.29 ± 0.05	10
T1-OSL2	2.73 ± 0.4		8.89 ± 0.42		1.37 ± 0.05	10
T1-OSL3	2.62 ± 0.33		9.48 ± 0.32		1.50 ± 0.04	10.5
T1-OSL4	3.56 ± 0.03		12.06 ± 0.09		1.55 ± 0.02	10
T1-OSL5	3.28 ± 0.04		10.35 ± 0.09		1.49 ± 0.02	10
T1-OSL6	3.02 ± 0.03		9.58 ± 0.08		1.42 ± 0.02	10.5
T1-OSL7	3.25 ± 0.03		11.69 ± 0.09		1.49 ± 0.02	10.5
T3-OSL1	3.1 ± 0.03		11.44 ± 0.09		1.85 ± 0.02	10
T3-OSL2	3.31 ± 0.03		11.93 ± 0.09		1.72 ± 0.02	10
T6-OSL1	3.36 ± 0.03		11.43 ± 0.1		1.36 ± 0.02	10
T6-OSL2	3.35 ± 0.03		10.54 ± 0.08		1.4 ± 0.02	10
C
Sample	nGGc	nLn/Tn	De (Gy)	Annual dose (µGy/a)	OSL age	(ka)	n/N
					CAM	Minimum age
T1-OSL1			231 ± 1	2714 ± 109	103 ± 5	85 ± 4	4/18
T1-OSL2			210 ± 1	2811 ± 109	86 ± 4	75 ± 3	3/18
T1-OSL3			212 ± 2	2821 ± 106	82 ± 3	75 ± 4	4/18
T1-OSL4	8	20	264 ± 10	2247.8 ± 35	79 ± 7
T1-OSL5	8	20	225 ± 6	3104.7 ± 32	72 ± 6
T1-OSL6	9	20	212 ± 6	2921 ± 30	73 ± 6
T1-OSL7	10	20	222 ± 7	3221.5 ± 34	69 ± 6
T3-OSL1	9	20	284 ± 14	3521.4 ± 70	81 ± 7
T3-OSL2	9	20	252 ± 12	3424.5 ± 67	74 ± 7
T6-OSL1	9	20	258 ± 6	3106 ± 34	83 ± 7
T6-OSL2	10	20	205 ± 4	3082.8 ± 33	66 ± 5

.

Although precipitation differs between climatic phases, its conversion into long-term moisture content yields relatively similar values between samples (10.0–10.5%; Table 6). This arises because precipitation was first calculated as a time-weighted mean over the entire burial period, and this long-term averaged precipitation was then converted into moisture content using a proportionality relationship. Given that these burial intervals span up to ~80 ka (from the Early Glacial to the Holocene), the contrasts in precipitation between individual phases are strongly smoothed when averaged over time.

To illustrate this process, a short example is provided: for sample T1-OSL4, the weighted mean precipitation reconstructed from all climatic phases represented during burial is 392 mm/yr. Using the proportionality relationship adopted in this study (modern 727 mm/yr corresponding to 19% moisture in northern France loess), this value converts to a long-term moisture content of 10%.

4.4.2. OSL Results

Figure 9 illustrates the dose–response curves and the resulting global growth curve obtained for sample T6-OSL1, taken as a representative example of the luminescence behaviour observed in the dataset (Figures S1–S4). The equivalent dose (D_e) values derived from these curves, together with those obtained for the other samples, were processed using the selected statistical model (CAM or minimum age). The results obtained for profiles T1, T3, and T6 (Figure 6) are presented in Table 6

The age results obtained at Archéosciences Bordeaux (profiles T1 and T6, Table 6) are consistent with the expected chronological range, notably when compared with thermoluminescence (TL) ages on heated flints from the site of Mauquenchy located about 120 km to the WSW of Hermies in Normandy: 84 ± 8 ka at the base and 78 ± 7 ka at the top of the humic horizon SS-1 [25]. These ages are considered particularly robust owing to the recognised stability of the TL signal in heated flints [81,82] and thus provide reliable chronological benchmarks. The results are also consistent with the proposed correlations with the NGRIP record, established through cyclostratigraphic analysis of the pedostratigraphic signal of the Last Glacial [3].

At the Re.S.Artes laboratory, applying the Central Age Model (CAM) to samples T1-OSL1 (103 ± 4 ka), T1-OSL2 (86 ± 4 ka), and T1-OSL3 (82 ± 3 ka) yields significantly older ages compared with both the results obtained at Archéosciences Bordeaux and the TL ages from Mauquenchy. For instance, at the same stratigraphic level, T6-OSL1 gave 83 ± 7 ka, whereas T1-OSL1 gave 103 ± 4 ka. Moreover, sample T1-OSL3, taken from the thick laminated loess deposit (unit 6) overlying the humic complex, produced an age older than the TL chronology of Mauquenchy.

These discrepancies may be explained in part by differences between the two laboratories protocols, notably the grain-size fractions analysed (80–125 µm at Re.S.Artes versus 20–40 µm at Archéosciences Bordeaux). Grain-size-dependent saturation behaviour has indeed been documented in some contexts [83], suggesting that different fractions may approach saturation at different rates. However, as shown by Fuchs et al. [84] at Dolní Věstonice, there is no systematic bias between ages obtained on fine, intermediate, and coarse quartz fractions when alpha and beta dose contributions are properly corrected. Quartz ages obtained on 4–11 µm, 38–63 µm, and 90–200 µm fractions—after HF etching (coarse grains) or H₂SiF₆ etching (taking into account the alpha dose rate for fine and intermediate grains)—remain consistent within 2σ uncertainties. In addition to grain-size effects, differences in dose-rate determination between the two laboratories may also contribute to the observed offsets, as dose-rate measurements were performed in situ in the field at Re.S.Artes, whereas they were conducted in the laboratory at Archéosciences Bordeaux.

In our case, however, the dominant factor is likely the depositional context rather than inter-laboratory differences. The samples analysed at Re.S.Artes come from humic and laminated colluvial deposits in which age inversions and age overestimations are frequently documented, owing to limited light exposure, short transport distances, and reworking prior to burial. Such settings are well known to favour incomplete bleaching [61]. Although the overdispersion values for the Re.S.Artes samples remain relatively low (3–12%), moderate OD does not exclude partial bleaching when using multi-grain aliquots, since aliquots may still incorporate mixtures of well- and poorly bleached grains. This depositional context therefore provides the most plausible explanation for the older CAM ages obtained at Re.S.Artes. In such contexts, proposing minimum ages may be better suited to represent the well-bleached grains. The new estimates obtained (T1-OSL1 = 85 ± 4 ka, T1-OSL2 = 75 ± 3 ka, T1-OSL3 = 75 ± 4 ka; Table 6) show improved agreement with the TL ages from Mauquenchy and with the CAM ages measured at Archéosciences Bordeaux.

For profile T3, samples OSL1 and OSL2 were collected from loess layers located beneath a Bt horizon of a brown leached soil attributed to the Last Interglacial (Eemian). This stratigraphic position provides a chronological control, as it sets a minimum age threshold for the underlying sediments, which should not be younger than ~130 ka. The OSL ages obtained (81 ± 7 ka and 66 ± 7 ka) are therefore strongly underestimated and in clear disagreement with the stratigraphic constraints. The cause of this underestimation remains unclear. The cosmic contribution and concentrations of radioactive elements are comparable to those of other samples, and the use of laboratory-measured water contents (20% for T3-OSL1 and 15% for T3-OSL2) would yield slightly older ages (101 ka and 86 ka) but remain far too young compared to the ≥130 ka expected.

To explain this discrepancy, processes linked to intense water circulation are considered since uranium is highly mobile in aqueous solution and susceptible to leaching [61].

Field observations of Saalian loess show the presence of doublet-type structures highlighted by micro-bands of illuviation of fine particles and clays. These structures are interpreted as the first signs of water circulation during the permafrost melting phases preceding the onset of Eemian interglacial soil formation. Their formation necessarily involves deep leaching of radioelements, with the possibility of local accumulation of K, U, and Th in these illuvial microhorizons. It is likely that other episodes of water circulation affected the Saalian loess, particularly during the grey forest soils phase of the Early Glacial (110–75 ka). This period is marked by deep leaching processes leading to intense dissolution of the substrate [3]. Furthermore, the position of samples T3-OSL1 and T3-OSL2 relatively close to the surface could also have exposed them to subsequent leaching phases, which left no obvious traces (e.g., permafrost melting phases, Holocene soil formation). All these processes may lead to an artificial enrichment in radioelements in the Saalian loess, which may lead to an artificial increase in the annual dose rate recorded by quartz and to underestimated OSL ages.

Finally, the mixing of younger grains through bioturbation cannot be excluded for T3-OSL2, located about 1 m below the Bt horizon of the brown leached soil. The mild climatic conditions of interglacial periods are favourable to intense biological activity, especially of earthworms, which can disturb sediments at noticeable depth in the absence of permafrost. This explanation is, however, less convincing for T3-OSL1, located more than 2 m below the same Bt horizon, which was probably less disturbed than T3-OSL2.

However, none of these processes alone can explain such a large discrepancy with the expected ages. Several studies have shown that quartz in loess contexts frequently yields ages that are younger than those expected for deposits underlying the interglacial soil. At Mircea Voda (Romania), for example, fine-grain quartz (4–11 µm) sampled immediately below the Eemian Bt horizon gave an age of 106 ± 16 ka, although the stratigraphy requires >130 ka, with the degree of underestimation increasing with depth [85]. Similar results have been reported in China, with ages of 81 ± 7 ka beneath the Eemian Bt at Luochuan [86] and 69.8 ± 3.8 ka in the lower part of the Eemian palaeosol at Zhongjiacai [87]. Likewise, Lai (2010) [88] showed that quartz OSL ages from Luochuan become systematically too young beyond 70 ka. In this context, the two OSL ages from T3 will not be considered further in this study.

Finally, the integration of the stratigraphically constrained OSL age results allowed the construction of an age–depth model for profile T1 at Hermies (Figure 10).

5. Discussion

The pedosedimentary records identified in the Hermies area provide an opportunity to reassess the climatic and geomorphological dynamics of the early Weichselian in northern France. The combination of sedimentological and micromorphological analyses with OSL dating from sections T1, T3, and T6 allows for a detailed characterisation of the different soil horizons observed in the field and their chrono-climatic attribution. Finally, the study of their spatial distribution, carried out along a transect linking eight trench sections, leads to the proposal of a model for the development of Weichselian Pleniglacial horizons in relation to permafrost dynamics. The Hermies sequence thus offers a robust reference framework for broader-scale correlations, particularly with sequences in Belgium and Germany. Furthermore, several Middle Palaeolithic occupation levels have been identified within the studied sections, associated with horizons attributed to the Early Glacial. These fit into a broader pattern of numerous and recurrent occupations documented in northern France during this period.

5.1. The Weichselian Early Glacial Humic Soil Complex

5.1.1. Grey Forest Soils SS-1 and BSO

The Early Glacial humic complex is well expressed in profiles T1 and T6 (Figure 6 and Figure 10), with ages of 85 ± 4 ka (T1-OSL1) and 83 ± 7 ka (T6-OSL1) obtained from the humic horizons U11 of T1 and U5 of T6, respectively. According to the sedimentological analyses conducted on profile T1, this horizon shows clay enrichment, distinct TOC peaks (up to 0.45%), and illuviation features in thin section, characteristic of grey forest soils (Bth horizon). The ages obtained are consistent with those of the Saint-Sauflieu 1 (SS-1) horizon, dated at Mauquenchy by TL on heated flints to 84 ± 8 ka at the base and 78 ± 7 ka at the top [25]. Until now, these TL ages constituted the only robust chronological references for this unit in northern France.

The age–depth model obtained for profile T1 places the pedogenesis of the SS-1 horizon (unit 11) between ca. 82 and 85 ka (Figure 10 and Figure 11), within the chronological range of Greenland Interstadial GI–21 (~78–85 ka) in the NGRIP record and Marine Isotope Stage (MIS) 5a, dated to around 82 ka (Figure 12). This period was characterised by strong seasonal contrasts indicated by intense freeze–thaw processes and substantial spring snowmelt, marked by the downward infiltration of leached (or bleached) silt, i.e., silt lightened by eluviation, throughout the profile and at its base. According to pollen data from Saint-Sauflieu [30], the vegetation cover was a forest dominated by Pinus and Betula (AP max. 80%). These conditions enabled intense pedogenesis, as evidenced by high TOC values and bioturbation indicated by the presence of well-developed biotubules in thin sections T1-B6 and T1-B7 and of earthworm diapause chambers.

Equivalent horizons are found in numerous western European sequences (Figure 13): in Belgium at the top of the Rocourt pedocomplex at Romont and Remicourt (Villers-Saint-Ghislain B, i.e., VSG-B) [1,5], in Germany at Nussloch and Mainz-Weisenau (“Mosbacher Humus Zone”, MHZ) [11,89], or at Garzweiler (“Holz Soil”) [90,91]. However, these horizons are either undated or associated with ages whose accuracy and/or precision are insufficient to assign their formation to a precise chronological interval. The datings obtained in this study are therefore of particular importance, as they constitute, together with those from Mauquenchy in Normandy, the only results to provide such a level of chronological accuracy for this horizon in northern France and more broadly in northwestern Europe.

In profile T1, below the SS-1 soil horizon and the bleached (or leached) layer of unit 12, unit 13 shows an intermediate facies between a grey forest soil (Bth horizon) and a leached brown soil (Bt horizon). It can therefore be correlated with the Bettencourt-Saint-Ouen soil (BSO), of similar nature, posterior to the Eemian Bt but anterior to the SS-1 Bth horizon. In northern France, the BSO horizon is rarely identified outside the Somme Basin [20,92] because it is often difficult to distinguish from the interglacial Bt horizon. This horizon has been dated to 96.3 ± 22.8 ka with the additive dose method in thermoluminescence (TL ADD) at Bettencourt-Saint-Ouen [22], while a mean TL age of 105 ± 12 ka BP (107 ± 13 ka and 103 ± 12 ka) was also obtained on heated flints from Villiers-Adam [31] in an archaeological level preserved within the BSO soil. These results are consistent with the chronology of the overlying SS-1 horizon and support a formation of the BSO during a very long period (≈20 ka) encompassing interstadials GI–24 to GI–22 (Figure 12). At the scale of western Europe, this horizon is also difficult to identify but has been described in Belgium (Figure 13), notably at Romont and Remicourt (Villers-Saint-Ghislain A, VSG-A) [1,6]. The TL age of 95 ± 11 ka obtained at Remicourt [93] agrees with a formation of this horizon during MIS 5c.

Figure 12. Stratigraphic synthesis of the Early Glacial horizons at Hermies (T1 + T3 + T6), correlated with the NorthGRIP palaeoclimatic record [94]. SS = Saint-Sauflieu soil, BHZ = Bleached horizon, BSO = Bettencourt-Saint-Ouen soil.

Figure 12. Stratigraphic synthesis of the Early Glacial horizons at Hermies (T1 + T3 + T6), correlated with the NorthGRIP palaeoclimatic record [94]. SS = Saint-Sauflieu soil, BHZ = Bleached horizon, BSO = Bettencourt-Saint-Ouen soil.

Unit 10, overlying the SS-1 soil in profile T1, corresponds to a compact orange-brown sandy clay loam. This unit appears to be highly localised (unit 7, Figure 5), as it is absent from the other profiles studied at Hermies (T2 to T8), from nearby borehole pits (notably SP-178, located less than 10 m from T1) excavated during the Paléo-3 diagnostic campaign, [42] and in all the other northern France profiles. Moreover, in this unit, clay content decreases from base to top, contrary to the trend generally observed in a soil developed in situ. Micromorphological analysis of thin section T1-B5 confirms this interpretation: voids are disorganised, and the illuvial features (orange to mahogany-brown ferri-argillans) show a highly disturbed arrangement, with some fractured or partly incorporated into the groundmass. These observations, together with traces of post-depositional biological activity, indicate a disturbed and reworked horizon. Unit 10 is therefore interpreted as a colluvial deposit, formed by the local and rapid trapping of silty material derived from the erosion and reworking of older horizons (Eemian Bt, loess, and earlier soils) from the adjacent slope.

5.1.2. The Steppe Soil SS-2

Overlying a second bleached horizon (unit 9), the top of the humic complex in profile T1 is marked by a 0.8 m-thick isohumic soil horizon (Unit 8). During fieldwork it has been subdivided into three sub-horizons (8a, 8b, and 8c) based on colour variations highlighted by its TOC content. However, the clay grain–size curve, which shows a gradual trend without sharp breaks and a maximum in the upper third, suggests that this is probably a single pedogenetic horizon. Thin sections (T1-B2 to T1-B4) display a marked platy microstructure (freeze–thaw alternations), high porosity, and traces of biological activity. TOC content, the absence of illuviation, and the presence of bioturbation support the interpretation of unit 8 as a steppe soil. These pedological features are consistent with the OSL ages of 75 ± 3 ka and 72 ± 6 ka (T1-OSL2 and 5, Figure 11), which allow correlation of this horizon with the steppe soil SS-2, generally the best preserved of the Early Glacial soils in northern France and Belgium, whereas horizons SS-3a and SS-3b are more frequently eroded. In profile T6, unit 4 also shows steppe soil features (silty texture, dark colours, well-developed channels, ferri-argillan papules), with an age of 66 ± 5 ka (T6-OSL2). Although this age is slightly younger than those obtained from unit 8 in profile T1 (75 ± 3 ka and 72 ± 6 ka), it remains consistent within uncertainties and supports the attribution of these horizons to the steppe soil SS-2.

The steppe soils SS-2, SS-3a, and SS-3b of the Early Glacial in northern France are only rarely dated, and ages are only available from Saint-Sauflieu [95]. Their short duration of formation results in overlapping error bars associated with luminescence ages, making their distinction difficult. In profile T1, unit 8 is attributed to horizon SS-2 on the basis of its sedimentological and micromorphological features, as well as the ages from this study. The age–depth model places this horizon between 79 and 74 ka (Figure 10), supporting a formation immediately after the development of soil SS-1, i.e., from Interstadial GI-20 onward (Figure 12), in agreement with the model proposed by Antoine et al. [3] and with ages from the site of Bettencourt-Saint-Ouen: 77.2 ± 8.2 ka (ADD, TL) and 73.2 ± 7.3 ka (ADD, IRSL) [96]. The OSL ages obtained at Hermies in the present study therefore represent a robust dataset providing, for the first time, a precise chronological attribution for this marker horizon of the Early Glacial sequences in northern France. As with the grey forest soil SS-1, this steppe soil is also recognised at the scale of northwestern Europe (Figure 13): in Belgium, at Romont and Remicourt, where it forms a key pedostratigraphic marker horizon (Humic Complex of Remicourt, HCR) [6,97]; in the upper part of the Mosbacher Humus Zone (MHZ) at Mainz-Weisenau [11,89]; and in the Holz Humus Zone at Garzweiler [91]. Some TL ages obtained at Remicourt, Belgium (69 ± 8 ka and 73 ± 8 ka) [93], also support the development of this soil at the end of MIS 5a.

5.1.3. Bleached Horizons of the Early Glacial

In profile T1, two bleached horizons (Figure 11) were observed: one at the interface between soils BSO and SS-1 (unit 12, Figure 6), and another at the interface between soils SS-1 and SS-2 (unit 9, Figure 6), both associated with a marked decrease in clay content. These horizons are particularly well preserved at Bettencourt-Saint-Ouen and Villiers-Adam and are recognised in numerous sequences across northwestern Europe (Figure 13). In Belgium, comparable horizons, referred to as the “Whitish Horizons of Momalle” [1,6,97], have been identified at Romont (between BSO/SS-1 and SS-1/SS-2), as well as at Rocourt, Harmignies, Remicourt (between SS-1 and SS-2), and near Maastricht [98]. They are also observed further east in the Lower Rhine Valley [90] and at Nussloch [11], although they are less well preserved at the latter site.

These bleached horizons are thought to have formed during stadial episodes GS-22 and GS-21 (Figure 12), between the pedogenetic phases of the interstadial soils that bracket them. In this context, stadial climatic conditions, characterised by strong seasonal contrasts and winters with heavy snow cover, may have favoured abundant water infiltration and the leaching of iron oxides as well as certain clay minerals, giving these horizons their eluviated character. Nevertheless, field observations and micromorphological analyses indicate that surface erosion by runoff on sparsely vegetated soils (or fully open) was the main factor at the origin of their formation. This process accounts for the presence of reworked pedological nodules within the bleached silts, whose organisation was later disturbed by bioturbation processes associated with the development of the overlying SS-2 soil. This dynamic is well illustrated at Saint-Sauflieu, in a slope position, where a small flint gravel layer is associated with the bleached level between soils SS-1 and SS-2, confirming their link to a phase of climatic degradation under stadial conditions.

Finally, in Central Europe at Dolní Věstonice, silt horizons (MS) are also associated with stadial episodes during the Early Glacial, but in this case, they are of local aeolian origin, resulting from local-scale dust storms [99].

5.1.4. Humic Soil Complex: Chronological Framework and Erosion

The Early Glacial humic complex represents a key stratigraphic marker in the Hermies sequences (Figure 5). It is particularly well preserved at the foot of slopes (T1, T5, T6), where it reaches more than two metres in thickness, illustrating the importance of localised sediment traps in its preservation, as also observed at Havrincourt [21], Bettencourt-Saint-Ouen [22], and Villiers-Adam [20]. This soil complex reflects a step-by-step climatic evolution to more and more continental cold and dry conditions, in agreement with the palynological data from Saint-Sauflieu [30]. The series of luminescence results presented in this study supports the subdivision of the Early Glacial (110–70 ka) into two periods, as proposed by Antoine et al. [3,32] on the basis of pedostratigraphic analysis of numerous humic soil complexes: a first phase marked by the development of grey forest soils (BSO and SS-1, 110–77 ka), followed by a second phase characterised by the development of steppe soils (SS-2, SS-3a, and SS-3b, 77–70 ka). This humic complex, comparable to those described in Belgian and German sequences, provides reliable chronostratigraphic correlations with these regions. In contrast, further east at Dolní Věstonice (Czech Republic), the much more continental setting led to a different evolution: from the earliest stages of the Early Glacial, colder and drier conditions prevailed, favouring the early development of steppe chernozem-type soils [84,99].

At Hermies, the upper boundary of the humic complex is marked by an erosional contact affecting the entire slope (Figure 5). In profile T3, the complex is much reduced (~40 cm, unit 2, Figure 4 and Figure 6) but displays infilled burrow structures containing beige-yellowish silts, probably derived from the erosion of the upper steppe horizons (SS-3a, SS-3b). Similar observations have been made at Romont, where the SS-2 soil (lower part of the Remicourt humic complex) is directly overlain, unconformably, by laminated colluvium incorporating reworked humic soil material at its base. Thus, in northern France as well as in other northwestern European sequences, the upper boundary of the steppe soil complex is systematically marked by a major erosional contact. The observations at Hermies suggest that this event reflects a regional-scale process driven by the onset of periglacial conditions and the rapid shift into the Weichselian Pleniglacial [100,101].

Figure 13. Correlation of the Early Glacial soil complex from reference sites of the Last Glacial. In France: Hermies (this study), Bettencourt-Saint-Ouen [22,96], Hermies Tio-Marché [74], and Saint-Sauflieu [32,95]. These sites are correlated with Remicourt in Belgium [5,6,93] and Mainz-Weisenau in Germany [22,102]. Legend: (1) Bedded silts, (2) local non-calcareous loess, (3) bleached horizon, (4) colluviated Bt, (5/6) Ah horizon/Steppe Soil, (7/8) Bth horizon of Greyzem on colluvium, (9) Truncated Bt horizon of Luvisol. BSO = Bettencourt Soil; SS-1, 2, 3a, 3b = Saint-Sauflieu Soils; HCR = Humic Complex of Remicourt; VSG-A = Villers-Saint-Ghislain A; VSG-B = Villers-Saint-Ghislain B; MHZ = Mosbacher Humus Zone.

Figure 13. Correlation of the Early Glacial soil complex from reference sites of the Last Glacial. In France: Hermies (this study), Bettencourt-Saint-Ouen [22,96], Hermies Tio-Marché [74], and Saint-Sauflieu [32,95]. These sites are correlated with Remicourt in Belgium [5,6,93] and Mainz-Weisenau in Germany [22,102]. Legend: (1) Bedded silts, (2) local non-calcareous loess, (3) bleached horizon, (4) colluviated Bt, (5/6) Ah horizon/Steppe Soil, (7/8) Bth horizon of Greyzem on colluvium, (9) Truncated Bt horizon of Luvisol. BSO = Bettencourt Soil; SS-1, 2, 3a, 3b = Saint-Sauflieu Soils; HCR = Humic Complex of Remicourt; VSG-A = Villers-Saint-Ghislain A; VSG-B = Villers-Saint-Ghislain B; MHZ = Mosbacher Humus Zone.

5.1.5. Human Occupations During the Early Glacial

The humic complex has yielded unpatinated lithic artefacts in situ, at the top of the BSO soil (MIS 5c, Figure 12) and at both the base and the top of the SS-1 soil (MIS 5a, Figure 12). The latter horizon corresponds to a level where a high concentration of Middle Palaeolithic sites is generally observed [26,27,30,103]. Indeed, the mode of formation of Early Glacial soils, characterised by slow and continuous colluvial sedimentation, favoured the preservation of trapped artefacts. In addition, an archaeological level had already been identified during previous studies at the site known as Le Champ Bruquette, at the surface of the basal gravel and attributed to MIS 5d [23].

In contrast, lithic artefacts are much rarer in the steppe soil part of the humic complex [41,104]. This difference in density must, however, be put into perspective: grey forest soils developed over a long duration (~30 ka), whereas the steppe soils that followed formed over only ~5–7 ka. Thus, the presence of the SS-2 steppe soil at Hermies, well developed and preserved with a thickness of more than 50 cm in several profiles (T1, T5, and T6), offers the opportunity to uncover artefacts during ongoing and future excavations linked to the Canal Seine-Nord Europe project.

5.2. The “Hermies Laminated Colluvial Deposits”: A Marker Facies of the Lower Pleniglacial

Overlying the humic complex, units 6 and 7 of T1 as well as unit 3 of T6 correspond to heterogeneous laminated silts with irregular brown and grey bands. By their facies and stratigraphic position, these units correspond to the Hermies laminated colluvial deposits, first identified in the 1990s at the Palaeolithic sites of Hermies Tio-Marché and Champ-Bruquette and attributed to the Lower Pleniglacial [3,21]. They form a thick (2–3 m) stratified colluvial deposit with lenses of reworked humic material and periglacial deformation features, well preserved at downslope positions or in large gullies, as at Havrincourt. At Hermies, thin section T1-B1 shows evidence of freeze–thaw processes and ferruginous oxidised laminae, consistent with the onset of cold and arid conditions of the Lower Pleniglacial.

This horizon is also described in Belgium as a thick, heterogeneous stratified deposit well preserved at downslope positions, as at Remicourt, or filling large gullies, as at Harmignies and Romont. There it is referred to as the “Harmignies Colluvium” (EB1) [5]. It is also recognised further east in profile P3 at Nussloch [11]. However, these colluvial deposits with a finely laminated facies result from a major erosional episode on slopes. This mode of formation explains the numerous inversions and the large scatter in luminescence ages. In Belgium, luminescence ages mainly range between 80 and 100 ka [6], meaning that only a few ages are available for this unit at the scale of western Europe [33].

The ages obtained in this study (T1-OSL2 at 75 ± 3 ka, T1-OSL6 at 73 ± 6 ka, and T1-OSL7 at 69 ± 6 ka) in profile T1 suggest that these laminated silts were deposited immediately after the humic complex (Figure 11). These results are consistent with the ages of 69 ± 8 ka and 73 ± 8 ka from Remicourt [93] and 65 ± 9 ka at the base of the overlying loess at Harmignies [33]. The new dating results thus complement the scarce available data at the European scale and constitute the first ages for this horizon at the regional level. They suggest that the Hermies laminated colluvial deposits represent a marker of the transition from the Early Glacial to the Lower Pleniglacial around 70 ka (Figure 10).

The climatic signal represented by the deposition of the Hermies laminated colluvial deposits around 70 ka can be compared to the palynological data from La Grande Pile [105,106]. Indeed, immediately after the Ognon II short interstadial around 70 ka, AP collapsed to ~15%, while NAP exceeded 80%. The assemblages are then dominated by Artemisia, Chenopodiaceae, and Poaceae, with occasional Juniperus and Hippophae. The pollen record from Saint-Sauflieu confirms these trends. Birch, although sporadically recorded [30], was progressively replaced by Poaceae and Asteraceae.

This gradual evolution reflects an opening of the landscape, with the marked shift from boreal taiga to cold, dry steppe formations, corresponding to the onset of Marine Isotope Stage 4. This pattern is consistent with the stratigraphic transition from the humic complex to the Hermies laminated colluvial deposits. The accumulation of thick laminated silts at downslope positions likely resulted from enhanced erosive runoff during episodes of snowmelt and spring thaw, favoured by the abrupt reduction in the vegetation cover at the beginning of the Pleniglacial. This layer can be compared to finely laminated colluvial deposits including bleached silt, soil nodules and clayey aggregates that can be observed at the footslope of bare soils after strong rain events in cultivated loess areas (Figure 14).

This well-defined horizon therefore represents a major stratigraphic marker. It is positioned immediately above a major erosional unconformity marking the onset of MIS 4 in Western European loess stratigraphy.

5.3. Middle and Upper Pleniglacial: Erosional Events and Slope Periglacial Dynamics

The chalk bedrock of the Hermies sequence shows marked evidence of dissolution (dolines and sinkholes) and is overlain by a gravelly deposit containing frost-shattered flints chalk blocks and occasionally reworked Palaeolithic artefacts, showing rounded edges and variable patinas (brown, orange, white, green). In addition, profile T2 reveals a highly heterogeneous silty matrix of the basal gravel (brown, blue-grey, ochre), indicating repeated episodes of solifluction and cryoturbation. All these observations demonstrate that the formation of the gravel was not related to a single event but to several major erosional crises, linked to permafrost dynamics [107,108]. Incision down to the chalk bedrock testifies to the high intensity of these events. At Hermies, a dry valley adjacent to the slope, oriented southwest–northeast (Figure 2C), acted as a sediment trap recording slope erosion and reorganisation phases.

The dating and correlation of these stratigraphic discontinuities allow two large-scale erosional episodes to be distinguished in northern France sequences (Figure 15): (i) a first event at the transition from the Lower to the Middle Pleniglacial (LPG/MPG), and (ii) a second event marking the transition from the Middle to the Upper Pleniglacial (MPG/UPG).

Both episodes were triggered by thawing and rapid melting of permafrost ice induced by abrupt stadial–interstadial warming as illustrated by Figure 15:

• Phase 1: A cold phase promoting the development of ice-rich permafrost and large ice-wedge networks in a very cold but sufficiently humid environment to allow major ice accumulation in the ground.
• Phase 2: A phase of rapid warming leading to ice-wedge melting, thickening of the active layer, concentrated slope runoff, and incision by thermokarst channels. These channels widened and deepened to form gullies strongly incising slopes. Such processes caused large-scale remobilisation of slope materials, enhancing colluviation and the reworking of older horizons.

The first event occurred at the boundary between the laminated silts of the LPG and the weakly developed soils attributed to the early MPG. Structures observed in the laminated silts (small graben-like subsidence faults, linear features, and orange oxidation spots) reflect the influence of an ice-rich permafrost prior to its degradation (Figure 15A). The abrupt warming (Figure 15B) triggered massive melting of ice wedges and slope destabilisation. This process is documented at the same stratigraphic level in other sequences of northern France such as Villiers-Adam [20], Bettencourt-Saint-Ouen [22]), and Morcourt [39], as well as in Germany at Nussloch (thermokarst structure TK-2) [2,11]. It generated thermokarst gullies and heterogeneous colluvial deposits, accompanied by extensive stripping of the chalk slopes. At Hermies, the Mousterian site of Tio-Marché yielded TL ages at the base of the MPG sequence ranging between 54 and 57 ka (N. Debenham). At Villiers-Adam, the final sandy infilling of these structures was TL-dated between 55 and 60 ka, placing this erosional episode at the onset of the first MPG interstadial (GI-17 or GI-16). At Nussloch, OSL and IRSL datings (infill of TK-2: 54.2 ± 4.1 ka [109]; lower Gräselberg soil: 57.0 ± 5.7 ka [110]) also support this attribution. Analogous structures exist in Belgium (Veldwezelt-Hezerwater, [98]) and at Garzweiler [111], although not yet dated.

The second intense erosional event is recorded at the boundary between the MPG and the UPG. Profile T7 displays an exceptionally thick record of MPG horizons (Figure 15C), with a well-developed lamellar structure and orange oxidation features reflecting freeze–thaw alternations. Moreover, in better-preserved regional sequences such as at Haynecourt [38] and Havrincourt [21], the MPG/UPG transition is marked by a tundra gley affected by a polygonal network of large ice-wedge pseudomorphs (level F-4 at Havrincourt), indicating continuous permafrost. This gley, capping the MPG soil complex, has been dated between 30.4 and 31.1 ka at Haynecourt, to 31.4 ± 2 ka and 31.2 ± 2.1 ka at Havrincourt, and to 30.7–32 ka at Morcourt [39]. Similar ages have been obtained at Nussloch for tundra gley G2b, with a calibrated mean of four radiocarbon ages on earthworm calcite granules ranging from 29,622 to 30,982 cal BP [14]. These results place the development of the gley during interstadial GI-5.1 and locate the erosional event around 30 ka. As with the LPG/MPG event, it corresponds to rapid permafrost degradation, causing ice-wedge melting, polygonal networks on plateaus, and massive slope remobilisations (Figure 15D). After this phase, aeolian dynamics resumed, with increased dust supply and stronger winds, leading to the accumulation of calcareous loess of the UPG unconformably overlying the MPG soil complex (Figure 15E).

These two episodes of rapid permafrost destabilisation, dated around 60–55 ka and 30 ka, therefore triggered intense slope remobilisation [112]. They are exceptionally well recorded at Hermies thanks to the position of the sections along a small valley and the presence of an adjacent dry valley acting as a sediment trap, whose axis progressively shifted during this period. The magnitude of these crises, cutting down to the chalk bedrock, as well as the signs of periglacial erosion (thermokarst) observed at the same levels in other regional [20,21,22,38] and European sequences [2,11,98,111], suggest that they represent two major events with an impact extending across western Europe. Furthermore, together with the few studied slope sites such as Renancourt [113], Saint-Sauflieu [19,32,114], and Morcourt [39], Hermies confirms the high potential and efficiency of sediment traps for recording climatic cycles and erosional crises in loess–palaeosol sequences.

6. Conclusions

The study carried out at the Hermies site, within the framework of the Canal Seine-Nord Europe project, has documented a pedosedimentary sequence covering mainly the Weichselian Early Glacial and Middle Pleniglacial chrono-climatic phases. By integrating stratigraphic, sedimentological, and micromorphological analyses with a series of eleven OSL datings, several soil horizons were characterised, thereby strengthening the chronoclimatic framework of the Last Glacial period.

The OSL ages obtained in the Early Glacial humic soil complex and in the overlying laminated colluvial deposits attributed to the Lower Pleniglacial represent a major advance. The consistency of the set of ages, and the comparaison with the TL ages from heated archaeological flints, reinforces their reliability. They constitute the first robust corpus of geochronological data for these units in northern France, complementing the scarce dates available for northwestern Europe. These results also highlight the importance of applying a humidity correction based on palaeoprecipitation modelling in OSL protocols used for loess sequences. Furthermore, the humic complex represents a central stratigraphic marker and a major asset for archaeological research, as demonstrated by discoveries, for instance, at Hermies (Le Champ Bruquette) and Bettencourt-Saint-Ouen. The ongoing and forthcoming excavations at Hermies and Ruyaulcourt, carried out within the Canal Seine-Nord Europe project, therefore hold strong potential for identifying new archaeological levels in this well-dated Weichselian Early Glacial soil complex, as well as in the Saalian sequence, for which the chronological framework is still under development.

In addition, the eight trenches opened on the southern slope of the present-day canal and correlated over 350 m have provided new insights into the spatial organisation of pedostratigraphic sequences and allowed the identification of stratigraphic discontinuities linked to erosional episodes. The stratigraphic analysis of these eight sections reveals a strong morpho-sedimentary response to climatic forcing, modulated by local topography. In particular, the presence of an adjacent dry valley, oriented southwest–northeast, played a decisive role as a sediment trap, favouring the preservation of successive phases of slope erosion and reorganisation. It is proposed that the most intense erosional crises, sometimes reaching the chalk bedrock, resulted from the rapid thawing and destabilisation of permafrost and ice wedges, triggered by abrupt climate warming around 60–55 ka and 30 ka. The analysis and dating of the Hermies sequences thus contribute to refining the regional pedostratigraphic framework and lead to the proposal of a spatio-temporal model of periglacial loess environments in northern France and, more broadly, across northwestern Europe.