Introduction

This article presents the results of a palaeoenvironmental study that aims at the reconstruction of the vegetation cover and use-history of non-built-up, empty and relatively large areas that occur within the Late Iron Age urbanised enclosed sites—oppida and hillforts. It is based on the combination and correlation of five different archaeobotanical proxies—pollen, phytoliths, diatoms, charcoal, and seeds—with analyses of parasites and archaeological stratigraphy obtained from one of the large open spaces at the oppidum Bibracte in central France.

Archaeological traces of populations living during the Late Iron Age (5th–1st century bc) in temperate Europe from the Atlantic to the Carpathian Basin are assigned to the La Tène culture that bears many common features but can differ considerably regionally. These Late Iron Age societies, traditionally connected with the Celts or Gauls, known from ancient written sources, witnessed profound social, political and cultural transformations. One of them was connected to urbanisation and in the 2nd–1st century bc resulted in the construction of oppida. These large, fortified ‘Celtic towns’, that occur in the vast territory of transalpine Europe stretching from the British Isles to the Danube Bend, fulfilled complex functions including roles of political, religious, and economic centres. Large-scale excavations conducted in many oppida open insights into the general urbanistic organization of the site and use of space within it. They almost exclusively focus on the originally built-up areas. Yet, unbuilt areas within these sites, which appeared as empty spaces for prolonged periods may have played an important role in the oppida urbanism and were possibly created just as deliberately as the architecture itself (Smith 2008). The relatively small areas, up to hundreds of square metres, surrounded by buildings and sometimes paved, are generally identified as public spaces (‘squares’). Less clear is the function of large empty spaces that cover up to several hundred or thousand square metres and usually separate (or connect) different residential zones and/or line the fortifications. The debate on their function has taken place for over a decade (cf. Fichtl 2005). Hypotheses on their usage as fields, pastures, spaces for social gathering, markets, refuge, and spare land for urban development, or various combinations thereof, have been put forward (Metzler et al. 2016; Winger 2016; Moore 2017; von Nicolai 2017).

Most of these roles and functions are drawn from theoretical frameworks, ethnography, history, or sociology but are seldom tested and supported through bio-, geo- or other archaeological evidence. The exception is the use of open spaces for farming. Based on the distribution of rare finds of farming implements and scattered reworked pottery fragments, recovered from the top-soil horizons during prospections or excavations it was argued that during the life of Iron Age oppida (or hillforts) the open spaces were used as arable land or as grazing grounds (Křivánek et al. 2013; Winger 2016; Knopf et al. 2000). Further, a system of channels delimiting rectangular plots and plant macro-remains (seed) assemblages from their fills, which differed from assemblages from habitation areas, were interpreted as direct evidence that the open space at the lowland oppidum at Manching was used for arable farming (Küster 1992). Pollen data originating from waterbodies such as lakes or cisterns situated directly in some of the hilltop oppida/hillforts were also used for reconstruction of land management and vegetation cover. However, the sources of pollen might originate from further afield (cf. oppidum Corent, Ledger et al. 2015) or might date from the period after the abandonment of the hilltop settlement (cf. hillfort Vladař, Kozáková et al. 2015) it was thus concluded that, if on their own, they do not represent an ideal resource of information about on-site vegetation or land-use history.

Long-term research of ancient field systems (so-called Celtic fields) demonstrated that a combination of analyses of plant macro (seeds) and micro (pollen) remains with soil chemistry, micromorphology, radiocarbon and OSL dating of sediments at field plots and their banks/lynchets is a useful tool for verification of their formation through farming activities and for unravelling the developmental histories of plots used as arable land, fallow or grazed pasture (Nielsen and Dalsgaard 2017; Arnoldussen 2018). Yet, for understanding the use-histories of open spaces situated amid “the urbanised space” of the oppida, and without field parcellation, palaeoenvironmental multiproxy analyses, which combined not only those mentioned but also other methods, like analyses of diatom and parasites, were (to our knowledge) carried out only at Bibracte oppidum (cf. Goláňová et al. 2020a). The sediments studied did not originate from the open/empty space itself or a bank, but from the ditch that in the 1st century bc ran between the two summit plateaus of Mont Beuvray (La Terrasse and Le Porrey) and delimited an open space. The results yielded no evidence for the presence of cultivation of crops or animal grazing in the area; nevertheless, it proved that the area has been maintained as grassland, possibly with some planted Tilia (lime) trees (Goláňová et al. 2020a).

The main aim of this paper is to discuss how to reconstruct the vegetation cover and (land) use-history of the open, non-built-up areas situated within the urbanised landscape of the Iron Age oppida/hillforts by use of multiproxy analyses of various types of ecofacts of mostly plant origin. In this, it will aid the unravelling of the developmental history of preserved archaeological soil deposits and try to detect/eliminate human activities that formed them.

The case study concentrates on the Bibracte oppidum, one of the key sites in oppida research first excavated in the second half of the 19th century (Buillot 1899). Excavations were resumed in 1984 and have continued since then almost without interruption (Guichard et al. 2018). The oppidum of Bibracte, situated on Mont Beuvray at 821 m a.s.l. (near Autun, Burgundy, France), is a well-known site that was described in Caesar’s Gallic Wars as the capital of the Celtic tribe of Aedui. It was founded at the end of the 2nd century bc and reached its peak in the second half of the 1st century bc, shortly before its rapid abandonment in favour of a new city, Augustodunum (Autun) (Guichard et al. 2018). In the Middle Ages, human activities on Mont Beuvray were limited to the Franciscan monastery built in the 14th century ad, and the area near the chapel of St. Martin which was constructed at the site of a Gallo-Roman sanctuary from the 1st–3rd century ad. The chapel was a focal point of annual livestock fairs that took place here from at least the Medieval period (Beck and Saint-Jean Vitus 2018).

The studied material comes from an enigmatic and archaeologically seemingly empty space known as La Terrasse situated at one of the two summit plateaus (810 m a.s.l.) in the proximity of the chapel of St. Martin and for which various functions, including the location of a citadel (Garenne 1867), Camp of Mark Antony (Buillot 1899) or a location of a Gallic cult (Gruel and Beck 1996) have been proposed. It is a ca. 1 ha (120 × 85 m) large and relatively flat space enclosed on three sides by a bank/rampart (Fig. 1). Except for the rampart, no built structures were detected either by archaeological excavations (Gruel and Beck 1996; Goláňová et al. 2020b), or geophysical prospection (Milo 2020). The 1986–1995 excavations resulted in the recovery of a very small number of artefacts (pottery and amphorae fragments, tegulae, an iron ring and a nail) mostly from the 1st century bc, though prehistoric, probably Neolithic/Bronze Age, pottery fragments and chipped stone finds were also reported (Beck and Gruel 1989). The results of soil geochemistry and micromorphology from La Terrasse (Lisá et al. 2022) indicate that (i) prior to the rampart construction in the 1st century bc (at the latest), soils were repeatedly eroded, (ii) the former prehistoric surfaces were indistinct and occurred deep, just above the frost-weathered bedrock, and (iii) based on an OSL date, the material from the upper 50 cm represents a levelling deposit dating to the 6th century ad and points to a later construction episode.

Fig. 1
figure 1

Bibracte, Mont Beuvray. Bottom left—plan of fortifications and position of La Terrasse (A) and Les Grandes Portes (B). Top—plan of the trenches re-excavated in 2019 and position of test pits A–H at La Terrasse; Bottom right—position of the test-pit I next to the rampart at Les Grandes Portes. Graphic: J. Geršl

Full size image

This on-site study was conducted to test existing archaeological hypotheses and to answer archaeological questions. It focuses on and evaluates materials preserved in archaeological sedimentary archives that are difficult to work with due to a bad state of preservation and disturbed stratigraphy, unlike for example continuous chronological palaeoecological records from a natural lake or peat bog deposits.

Methods

Sampling in the field

To cause a minimal impact to the site, four trenches excavated in 1986, 1987 and 1993 (Fig. 2 in Gruel and Beck 1996) were reopened. Three were placed on the internal part of the La Terrasse plateau, and the fourth cut on the western rampart (Fig. 1). The profiles of the sections were cleaned and documented. The eight new test pits (1 × 1 m) extending from the main section and labelled A to H were laid down for systematic sampling. One additional test pit—labelled I—was excavated and sampled near the main fortification at Les Grandes Portes (Fig. 1), one of the lowest points of the oppidum, and was hoped to represent a sedimentary trap.

Fig. 2
figure 2

Bibracte, Mont Beuvray. Sections of test pits from La Terrasse (A–H) and Les Grandes Portes (I). Test pits are visualised according to their elevation. Red triangles mark the surface of the test pit in m a.s.l.; light grey colour in the profiles indicates soil/deposit, dark grey colour indicates stones. Graphic: J. Geršl, M. Hajnalová

Full size image

In each test pit two types of samples were collected—composite samples (5 l) for seeds and charcoal and small volume (ca. 200 ml) point samples for pollen, phytolith, diatom and parasite analysis. Bulk samples were collected during the excavation in a sequence from top to bottom, each sample representing a 10 cm thick layer (Fig. 2). Point samples were taken from freshly acquired profiles and taken in a reversed sequence from the bottom to the top, with maximum care to prevent contamination.

Sample processing, sample selection and laboratory analysis

Seeds and charcoal

Plant macrofossils—seeds and charcoal—were extracted from the deposits by a combination of flotation in the tank and subsequent manual wash-over, and wet-sieving of the remaining mineral matrix (Goláňová et al. 2017). A sieve with a mesh of 0.25 mm was used for flotation and wash-over methods, and 1 mm for wet sieving. The dried fractions were sorted, and seeds were identified under a stereo microscope (Leica M80 at max 50×). For charcoal identification, the refractive surfaces of fragments larger than 2 mm were studied under a microscope with reflected light (Olympus BX 51 at max 200×). For seeds and fruits, all items were analysed. For charcoal, the analyses stopped after reaching 30 or 50 items. Taxa identification for seeds and charcoal was based on the available literature and modern and archaeological reference collections. In addition to quantification by counting (NISP), the weight of individual charcoal specimens (WISP) was measured. The uppermost samples representing the layers from 0 to 10 cm (the turf) and 10–30 cm (decomposing plant litter and/or with visible traces of recent human activity) were omitted from the analyses.

Pollen

Complete sequences were prepared and analysed from test pits A (A10 to A70), E (E10 to E100), and H (H10 to H100), organised along the East–West transect through the terrasse and test pit I (I10 to I70) situated to the south and further down the slope. Sampled deposits were dry and acidic. Pollen and spores were extracted by standard methods using both HCl and HF (Moore et al. 1991). Lycopodium spores were added to determine pollen concentration. If possible, at least 350 terrestrial pollen grains were counted per sample. The determination of pollen grains was based on Moore et al. (1991) and Beug (2004). The programme POLPAL (Nalepka and Walanus 2003) was used for plotting pollen, and ascertainment of local pollen zones according to ConSLink Rarefaction and PCA analysis. Total pollen concentration (TPC) was calculated as the number of counted pollen grains in the sample divided by the number of counted Lycopodium spores and multiplied by the total Lycopodium spores (cf. Bonny 1972).

Phytoliths

Altogether 10 samples from profiles A, E and H were subjected to phytolith analysis (A30, 40, 60, 70; E60, 80, 100; H50, 70, 90). The phytoliths (biogenic silica or plant opal particles) were recovered from the deposit samples in a multi-step process conducted following the methodological guidelines set up by Pearsall (2000) and Piperno (1988). The amount of raw soil used for the recovery was uniformly 2 g. The process included the destruction of organic matter by H2O2 (30%), the separation of the coarse sand fraction by wet sieving, the separation of clay fraction by gravity sedimentation and the recovery of plant opal particles by heavy liquid centrifugation using sodium polytungstate. Phytolith extracts (1 ml/sample) were mounted on a microscope slide and counted under a light microscope at a magnification ranging between 100–400×. The aim was to count at least 100 identifiable phytoliths. However, this was not achieved for samples with extremely low phytolith concentrations. The nomenclature follows the guidelines of the International Code for Phytolith Nomenclature 2.0 (ICPT—International Committee for Phytolith Taxonomy 2019). Morphotype names are also given based on the ICPN 1.0 (Madella et al. 2005), due to the fact the reference data from soil profiles also utilised ICPN1.0. C2 software (Juggins 2007) was used for descriptive statistics of the data and the output diagrams.

Diatoms

In total, 34 diatom samples were inspected, representing entire sequences from test-pits A (5 samples), E (10 samples), H (10 samples) and I (9 samples). Pre-screening of samples for the presence/absence of diatoms was conducted with an Olympus BX 53 light microscope at 400× magnification. Identification was carried out using the compendium by Lange-Bertalot et al. (2017). Since samples contained very few or no diatom frustules, permanent diatom slides were not made.

Parasites

In total, 46 samples from test-pits A, B, D, E, H and I were studied for parasites (as described in Flammer et al. 2018). In the laboratory, a sub-sample (5–10 g) was re-hydrated and disaggregated overnight in 20 ml of ultra-pure water (Sigma-Aldrich). For identification a Nikon Eclipse E400 microscope with Nikon 20×/0.25 Ph1 DL and 40×/0.65 Ph2 DL objectives was used. Prior to microscopic analysis, only agitation before pipetting was carried out to stop denser material from settling. A QImaging MP5.0 RTV camera was used with Qimaging Qcapture Pro to record any suspected eggs and these images were assessed against reference images before the final count number was confirmed.

Radiocarbon dating

The charred hazelnut shells, wood charcoal (fragments of twigs smaller than 5 mm in diameter and with an uncounted number of annual rings) and mineralised Sambucus seeds were submitted for radiocarbon (AMS) dating (ESM 1). We assumed that the charred seeds and wood charcoal (in all layers) are the result of human activity in the area and that the mineralized seeds come from vegetation that grew on the spot at the time of the formation of the layers. By dating these finds, we aimed:

  1. 1.

    To establish when the flat area inside the rampart was created;

  2. 2.

    To determine when it was the most intensively used (the assumption is that most data indicate the time when it was most used);

  3. 3.

    To confirm the OSL dating indicating that the construction/formation of the terrasse took place ca. ad 561;

  4. 4.

    To verify the results of the previous geological, pedological and micromorphological study (Lisá et al. 2022) that the terrace was created by layering the sterile earth (i.e. the part of the soil that has never been in contact with vegetation) unmixed with settlement debris and that the deposits are strongly bioturbated.

Samples were dated at the Institute for Nuclear Research in Debrecen. Charred remains were treated with two-step combustion (Molnár et al. 2013). Calibration of data was done in OxCal 4.4 (Bronk Ramsey 2021) with calibration curve Intcal20 (Reimer et al. 2020).

Results

Chronology

Relative chronology

From the test pit I at Les Grandes Portes, no artefacts were recovered. At La Terrasse artefacts were rare and even if present in each archaeological stratum (stratigraphic unit), they did not occur in every excavated (10 cm thick) layer. They usually represent few or single pieces of very small size (units of grams). The majority of datable artefacts are attributed to the La Tène (oppida) period. From other periods there are Neolithic or possibly Mesolithic chipped stone artefacts (silex) which occur in the upper (A20, C30) and the lower layers (E60, F80, F140, H50), and are always in association with La Tène period finds. Pottery that can be dated only broadly to the period of prehistory originates especially from the lower parts of the profiles (from D50, E50). However, in the case of profile E, these were situated above a (bank forming) layer with oppida pottery. Due to the residual occurrence of prehistoric finds (pottery and silex) in the stratigraphically younger layers, their archaeological dating is impossible.

Absolute chronology

Radiocarbon AMS dates on plant macrofossils retrieved from layers at La Terrasse are diverse, spanning nine millennia (Table 1; Fig. 3). There are single dates for Mesolithic, Late Neolithic (Eneolithic) and Hallstatt/Early La Tène periods; two dates for late La Tène/Gallo Roman and Migration/early Medieval periods and three dates for a later phase of the early Medieval period. This contrasts with artefactual evidence from built-up occupational zones of Mont Beuvray that dates almost exclusively to the late La Tène and Gallo-Roman periods.

Table 1 Bibracte, Mont Beuvray. La Terrasse and Les Grandes Portes. AMS dates on plant macroremains
Full size table
Fig. 3
figure 3

Bibracte, Mont Beuvray. Calibration of single radiocarbon measurements on plant macroremains; in grey samples of charred wood, in green charred Corylus nutshells and in purple mineralised Sambucus seeds. Graphic: P. Barta

Full size image

Nevertheless, AMS dates seem to agree with the OSL date obtained previously from the vicinity of test pits A and E (Lisá et al. 2022), which signals a later, 6th century ad construction episode for the middle and possibly upper strata of the enclosed area.

The calibrated radiocarbon dates within individual test-pits are in line with the stratigraphic position of samples collected from layers of La Terrasse (A, B, E, H) and at Les Grandes Portes (I). The only exception is layer E100 where remains from three prehistoric periods are mixed (Table 1; Fig. 3). We believe that dated plant macro-remains became part of the layers during the time of their formation partly through natural and partly through anthropogenic processes and as such, can be used to date the deposits/layers themselves.

Palynology

Despite the dry nature of sediments, the acidity of the bedrock and overlying deposits facilitated relatively good preservation of pollen grains, which occurred in relatively high concentrations. Some broken and corroded pollen grains could not be identified. Sample A70 wasn’t counted due to extreme pollen degradation and an unreliable palaeoecological record. Microcharcoal from wood species predominated.

Pollen diagrams for each test pit are presented in Fig. 4 and pollen concentrations (number of grains per ml) and absolute pollen frequencies (sensu Dimbleby 1985) in Fig. 5 (for detailed results see ESM 2).

Fig. 4
figure 4

Bibracte, Mont Beuvray. Palynological diagram composed for test-pits from La Terrasse (A, E, H) and Les Grandes Portes (I). Graphic: L. Petr

Full size image
Fig. 5
figure 5

Bibracte, Mont Beuvray. Variation of pollen concentrations calculated for each pollen sample from La Terrasse (A, E, H) and Les Grandes Portes (I). Possible buried surfaces marked with arrows. Graphic: M. Hajnalová

Full size image

For a better understanding of similarities and differences between the layers within and across the test-pits, the pollen data percentages were subjected to detrended correspondence analyses (DCA). Results of DCA visualised in a plot (Fig. 6, ESM 2 Fig. 1) show, that within each test pit, pollen spectra change with depth (Fig. 6a) but correspond with samples from other test pits (groups I to VI). Along the horizontal axes, samples separate in dependence on the AP/NAP ratio. Samples dominated by arboreal pollen have higher positive loadings, while non-arboreal pollen dominated samples with lower positive loadings (Fig. 6b, c). Along the vertical axes samples separate mostly depending on the percentages of tree pollen taxa. Samples dominated by Quercus and Fagus show higher positive loadings while Tilia and Corylus have lower positive loadings (Fig. 6d). Based on pollen taxa percentages the samples can be grouped into three horizontal strata, roughly corresponding to the bottom, middle and upper part of each section. The width of each stratum varies depending on the depth and stratigraphy of each test-pit (Fig. 7).

Fig. 6
figure 6

Bibracte, Mont Beuvray, La Terrasse and Les Grandes Portes. Detrended correspondence analyses (DCA) plot of pollen taxa percentages showing a test pits (A–I), depth (in cm) of individual samples and pollen groups I.–VI. based on pollen data similarities (top left); b proportion of trees, shrubs, and other pollen types (top right); c proportion of grasses and weed taxa (bottom left); d proportion of main trees species to other pollen types (bottom right). Graphic: M. Hajnalová

Full size image
Fig. 7
figure 7

Bibracte, Mont Beuvray, La Terrasse and Les Grandes Portes. Schematic grouping of pollen assemblages based on pollen taxa percentages within and across the test pits as detected by the DCA. Graphic: M. Hajnalová

Full size image

The first group is represented by the samples from the deepest layers in the test pits E and H (E100, E90 and H100) (Figs. 6a and 7), according to AMS the oldest from the period between the Mesolithic and La Tène periods. They are characterised by a very low ratio of NAP and a mixture of various tree taxa, among which stand out the highest abundances of Tilia and/or Corylus (Fig. 6b, d). The other abundant taxa are Quercus, Fagus and Pinus. Amongst herbaceous taxa pollen of Poaceae, Asteraceae and Artemisia are common, though Cerealia pollen and other pollen indicating farming and/or settlement activities (Fig. 6c) is present in low abundances in E100 and E90, but not in H100, suggesting localised human impact to the landscape.

The second group is formed by samples situated in the stratigraphy just above the first group (E80, E70, H90) and also includes the lowermost sample in test-pit A (A60) (Figs. 6a and 7). It differs from the previous group (Fig. 6b) by higher abundancies of Poaceae and synanthropic herbaceous vegetation—arable (Cerealia, in all samples), pasture plants (Plantago lanceolata-t., P. major, Rumex acetosella t., in E80 and 70) and ruderals (Polygonum aviculare t., in H90) (Fig. 6c) indicating stronger human impact and a more open landscape. In the AP the same main taxa as in the previous group are present, though Fagus and Pinus decrease more than other species.

The third group is formed exclusively by samples from the middle section of the test-pit H (H80 to H50) (Figs. 6a and 7) and is characterised by the highest, and gradually increasing, percentage of NAP (Fig. 6b), including species of grasslands (Poaceae, Artemisia), grazed land (Plantago lanceolata t., P. media, Carduus) and/or abandoned pastures (Pteridium) (Fig. 6c, ESM 2).

The fourth group consists of samples from the middle strata of the test pits E and I (E60-40; I60-50) (Figs. 6a and 7) and is characterised by ca. 30–40% of NAP (Fig. 6b) and an equal proportion of Quercus, Corylus, Fagus and other arboreal taxa (Fig. 6d). It indicates semi-opened oak and hazel woodland with the presence of arable (Cerealia) and grazed (Plantago lanceolata t.) land in the vicinity.

The fifth group, formed by samples from the upper part of the test-pit I (I40 to I10 and I70) (Figs. 6a and 7) differ from the previous only by slightly lower NAP (although Cerealia increase) and higher abundances of Quercus pollen, indicating spread of denser, possibly thermophilous, oak and hazel woodland.

The last, sixth group is formed by the samples from the upper part of all the test-pits at La Terrasse (A30-A10, E30-E10, and H40-H10) (Figs. 6a and 7). It is characterised by high, (but lower than the 3rd group) abundances of NAP including Cerealia and Rumex acetosella t. (ESM 2) and the highest abundances of Fagus. It reflects the mosaic of beech growths, degraded forest openings and pastures.

To summarise, pollen sequences from all test pits and all three/six strata reflect a relatively opened landscape, with a mixture of forest stands with oak and/or beech and woodland openings with pastures, meadows, and fields. The latter are possibly situated at a greater distance, as the pollen of arable taxa never reaches over 5%. In addition, based on the results of pollen spectra changes (Fig. 4, ESM 2), pollen concentration and absolute pollen frequency (Fig. 5), there is a possible presence of buried surfaces (Dimbleby 1985) in all test pits—A (A30), E (E50, E70), H (H70) and I (I50).

Notable are pollen grains of Castanea sativa (chestnut). One pollen grain was present in the sample (E60) connected with the bank of the prehistoric (most probably late Iron Age) fortification. Its frequency has risen since the early medieval period, the 9th century ad. These findings are in accordance with the data from stratified peat-bog deposits in the region. Jouffroy-Bapicot (2010) and Jouffroy-Bapicot et al. (2013) also record the first, though still rare finds of Castanea pollen in the late Iron Age (La Tène period) and their sharp increase since the early Middle Ages. Local cultivation of chestnut trees in the region of Bibracte during its occupation in the late Iron Age is also supported by wood charcoal data (see below).

Anthracology

The abundance of charcoal was extremely low (905 fragments, with a total weight of 29.01 g) and the size of fragments rarely reached over 3 mm. Fourteen taxa of trees and three taxa of shrubs were recorded (Table 2). Wood taxa composition and abundance varied between and within test pits (Fig. 8, ESM 3 Tables 1 and 2). Four taxa—Fagus, Quercus, Corylus and Cytisus—form 75% of the assemblage (both in NISP and the WISP; Table 2; Fig. 8) and the remaining 13 “other/minority” taxa form 25% (Fig. 9).

Table 2 Bibracte, Mont Beuvray, La Terrasse and Les Grandes Portes. Wood charcoal taxa weight (mg) from individual test-pits, and their ubiquity/frequency. The number in brackets indicates number of studied samples
Full size table
Fig. 8
figure 8

Bibracte, Mont Beuvray. Visualisation of DCA analyses on charcoal abundances data, showing proportions of the most common and “other” taxa in individual samples. Graphic: M. Hajnalová

Full size image
Fig. 9
figure 9

Bibracte, Mont Beuvray. Percentages of ‘minority’ wood taxa in individual charcoal assemblages based on WISP. The numbers in brackets indicate percentage of all minority taxa from all identified items in a given sample. Graphic: M. Hajnalová

Full size image

Results of DCA (Fig. 8) indicate that Quercus with a mixture of other taxa is characteristic for the stratigraphically earliest layers (the deepest samples of E and D). Accompanying Corylus and Cytisus indicate relatively open oak woodland. The plant macrofossils dated from this stratum (E100) are from the Mesolithic, Eneolithic, and La Tène periods. The youngest date (late Iron Age) is on mineralised Sambucus seeds. Surprisingly, sample H80, very different to all other samples from H, with relatively numerous finds of Castanea sativa charcoal also associates with these samples in statistical analyses.

Fagus dominates the samples from the middle part of A, B and D. These are also the only samples with Tilia (found in individual fragments) and with very few species typical of open forests. Small quantities of C. sativa are also present. Radiocarbon-dated plant macrofossils from this stratum (B50, E80) suggest its formation during La Tène and the beginning of the Gallo-Roman period. As this is the phase with the most intensive occupation of the hilltop, the picture given by the charcoal data—that of dense canopy forest—might be biased, probably by the high quality fuelwood (Fagus and Quercus) imported to the site. Objects made from the wood of Tilia and Castanea, species not naturally present at the site, similarly might also be imported. However, the pollen present from associated deposits indicates the local plantation or cultivation of these trees.

All samples from H, apart from one (H80, see above), have equal proportions of one of the main species and the “other/minority” taxa (Fig. 8). The composition of “other/minority” taxa strongly varies between samples (Fig. 9), and the spectrum contains species with poor quality fuelwood from wet stands like Alnus, Betula/Alnus, Populus, Populus/Salix, and species from secondary woodlands (Carpinus—high quality fuelwood) or shrubland (Cytisus). This indicates an anthropogenically changed, managed or maintained, substantially opened landscape. The AMS dates obtained in the lower strata (H90, H70) indicate their formation during ca. 5th century ad.

The last group represent samples from “I”, where Corylus and Cytisus prevail suggesting the presence, and maintenance, maybe by burning, of opened landscape (Figs. 8 and 9). The date obtained from the lower part of the test-pit I (I60) places the start of the formation of the deposit to the 9th century ad.

Seeds and fruits

The seeds and fruits were extremely rare. In 32 evaluated samples (over 160 l of deposit) from the test-pits A, B, D, E, H, and I there were 45 charred and 173 uncharred seeds (Table 3). The charred specimens were badly preserved and often fragmented. Seven taxa were taxonomically determined. The most numerous were fragments of shells of Corylus avellana. From crops, only one Panicum miliaceum grain was determined. In addition to a Hordeum/Triticum grain fragment, there was a fragment of cereal straw node. The other finds from wild and or ruderal species included exocarp fragments of Fagus sylvatica, nutlets of Rubus fruticosus, R. cf. idaeus and Galeopsis sp. The majority of the charred seeds occurred in deeper layers (Fig. 10).

Table 3 Bibracte, Mont, Beuvray, La Terrasse and Les Grandes Portes. Counts (NISP) of charred and uncharred seed and fruit remains. Asterisk* indicates occurrence of remains fossilised through mineralisation
Full size table
Fig. 10
figure 10

Bibracte, Mont, Beuvray, La Terrasse and Les Grandes Portes. The Number (NISP counts) of charred and uncharred seeds recovered from individual samples; shown only samples containing seeds or fruits. Graphic: M. Hajnalová

Full size image

Uncharred remains, some possibly mineralised, of 7 species were recorded and included exocarps and fruitlet fragments of Fagus (beech), Quercus sp. (oak), Sambucus spp. (elderberry), Rubus fruticosus (blackberry), Cytisus scoparius, Chenopodium album agg. and Polygonum aviculare. The mineralised Sambucus seeds from sample E100 are dated by radiocarbon (AMS) to the La Tène period, indicating that at least part of the uncharred material is of prehistoric origin (Table 1).

Phytoliths

The detailed results of the phytolith analysis are presented in ESM 4, where ICPN transformation codes are also listed.

Four samples from test-pit A (A30, A40, A60 and A70) yielded 470 disarticulated phytoliths of 12 different morphotypes. These consist of epidermal short (SC) and long cells (LC), as well as of trichomes (T) (Fig. 11). All epidermal long cells (Elongate entire, Tabular elongate sinuate, Tabular elongate clavate) are considered indicators of Poaceae stem and leaf epidermis, while Rondels (grass silica short cell—GSSC) can also be associated with the leaves of grasses. Three samples out of four yielded indications for the presence of strongly fragmented and/or corroded Elongate dendritic morphotypes, which can be considered as the direct indicator of cereals.

Fig. 11
figure 11

Bibracte, Mont, Beuvray, La Terrasse. Diagram of the results of phytolith analysis from samples from test-pits A, E and H. The horizontal axis shows the percentages. Graphic: Á. Pető

Full size image

Three samples from test-pit E revealed altogether only 160 specimens of 9 different morphotypes. Two samples (E80 and E100) are considered as sterile as the total surface count (TSC) did not provide the minimum number of phytolith specimens, set here at 100 pieces (ESM 4). Sample (E60) yielded well-preserved Elongate dendritic morphotypes that could be attributed to inflorescence bracts (lemma, palea and glumes) of old-world cereals. As such, they directly indicate the presence of cereal chaff. Based on Rosen’s (1992) cell wall pattern identification key, it is presumed that the specimen represents the wheat genus (cf. Triticum sp.). However, the signal on its own is very weak.

In three samples from test pit H altogether 267 disarticulated phytoliths were counted. The minimum number of phytolith specimens, set at 100 pcs, was not reached in the lowermost (H90) sample. The remaining two samples (H70 and H50) are composed of 12 different morphotypes, dominated by similar ones as in the assemblages of test-pits A and E, although in H Rondels and Lanceolate entire morphotypes show higher frequencies (Fig. 11).

It can be concluded that samples are relatively poor in phytoliths, with a rather limited, monotonous spectrum and the same dominant morphotypes in all three records.

Qualitative evaluation of diagnostic features of phytoliths can be used as a key for the reconstruction of the original environment, especially vegetation and plant habitats (e.g. Bertoldi de Pomar 1971; Powers 1992; Bowdery 1998, 1999; Golyeva 2001a, b). Using a model developed for the temperate climate zone, the majority of dominant morphotypes from La Terrasse fall within the group of the general indicators (Fig. 8 in Pető 2013). One of the most common morphotypes is the Elongate entire, which was found under all studied vegetation covers and land use types, thus having a low indication potential. Their presence itself indicates that the sampled sediments came in contact or were associated with the vegetation. This seems to be a straightforward conclusion, but not in the case of the studied Bibracte assemblages. Since many samples here are extremely phytolith-poor, their absence might indicate non-surface samples of lower soil horizons. It should be emphasised that the poorest phytolith samples were collected from lower layers in test-pits E (E80, E100) and H (H90). It can be assumed that these have never been in contact with dense vegetation and the sporadic phytolith appearance might come from downward filtration (Fishkis et al. 2009, 2010a, b).

The other present phytolith morphotypes originate mostly from the vegetation of open steppe/grassland areas (e.g. Infundibulate rondel), but also abundant are types from stands which developed under (possibly periodically) changing forest and/or open grassland vegetation cover (e.g. Tabular, Tabular elongate clavate). The Infundibulate rondel morphotype has so far only been detected from chernozem soils in the Carpathian Basin, but it is possible that at La Terrasse they originate from the treeless grassland vegetation. One of the main and characteristic types is the Tabular elongate clavate, previously recovered from soils developed under forest (brown forest soils), as well as from soils developed under grasslands (meadow soils) (Pető and Barczi 2010, 2011; Pető 2013).

The longate dendritic morphotype is characteristic of palea and lemma of the cultivated cereals (Ball et al. 1996, 1999, 2017). Although the ‘cereal signal’ is weak, yet it can be seen as evidence that either cereal processing activities were performed in the close vicinity, or that the sediment was transported here from a place where such activity took place. Many of these morphotypes show a corroded and broken condition, which might refer to post-deposition processes (ex-situ).

Diatoms

The samples were very poor in diatom species and diatom numbers (recovered from 11 out of 34 samples). Except for one sample with two finds, diatoms always occurred singly. Two exclusively aerotolerant species were present, Pinnularia borealis and Hantzchia amphioxis, both euryvalent taxa known mainly from wet and moist or temporarily dry places (van Dam et al. 1994). According to van Dam (1994), both taxa are circumneutral, tolerating fairly high oxygen concentrations, elevated levels of salinity and concentrations of bound nitrogen. In addition, both taxa have robust frustules and tolerate freezing, experimentally down to − 20 °C (Hejduková and Nedbalová 2021).

Pinnularia borealis occurred in samples A10, A20, A60, E40, H70, H100, I20 and I50. Hantzchia amphioxis occurred only in samples E80 and I70. One diatom identified only to the genus level (Pinnularia sp.) was detected in sample H80.

Diatoms are unicellular micro-organisms inhabiting various types of environments. They are most species-rich and most abundant in aquatic (Round et al. 1990), but also inhabit, for them, much harsher, terrestrial ecosystems (Ress 2012). Diatoms also live in soils. Their productivity here is mainly controlled by soil moisture, while light availability for photosynthesis is also an important factor (Foets et al. 2020). For this reason, in soils diatoms inhabit the upper layer and are most numerous at depths down to 10 cm (Konrad Wolowski pers. comm.). Rarely they can pass down through macropores of the soil column (leached experimentally down to 50 cm; Coles et al. 2016), but stop being viable shortly after.

The absence of diatoms in the majority of samples, and a very small number of only aerotolerant species point to originally desiccating conditions. All finds from deeper in the soil profiles (from 50 to 100 cm) were not viable and represented empty diatom frustules without protoplast. As diatom frustules preserve in deeper soils (Barragán et al. 2018) for millennia (Spaulding et al. 2021), their low, but systematic occurrence in deeper layers possibly indicates buried surfaces or admixture of soils which were once on a surface.

Parasites

The general level of parasites is very low, with the majority of samples not containing any diagnosable parasites (present in 11 out of 51 samples; Table 4). Except for one sample, where Taenia sp. was identified, all diagnosed parasites were Ascaris sp. Parasites were absent in test-pit H. Elsewhere, they are present, but occur only in the upper layers—between 0 and 20 cm, at maximum down to 40 cm from the surface. The eggs detected from Taenia sp. are of human origin as this stage is only produced in the human definite host, and not in the animal (i.e. pigs or cattle) intermediate host. Ascaris sp. eggs are produced in infected humans and pigs. The eggs cannot be conclusively distinguished.

Table 4 Bibracte, Mont, Beuvray, La Terrasse and Les Grandes Portes. Parasites detected in studied samples. Only samples with positive results are shown
Full size table

While Ascaris sp. infects humans via the faecal-oral route, Taenia sp. infections occur through the consumption of raw or undercooked red meat. These parasites are no longer endemic in Europe, but they have frequently been reported in a variety of archaeological contexts (Reinhard et al. 2013; Araújo et al. 2015; Flammer et al. 2018; Flammer and Smith 2020). Although it is unclear when intestinal helminths declined in Europe, it has been reported that the prevalence levels in the medieval period were similar to modern affected countries (Flammer et al. 2020). Current literature indicates that the sanitary reforms of the 19th and early 20th century, in combination with modernised water and waste management, lead to the disappearance of these infections.

The detected parasite densities were consistent with findings reported earlier from the site (Goláňová et al. 2020a), where their variation indicates that the use of the features/areas changed over time and occasionally included the deposition of faecal material. While the densities were comparable to the one described in Goláňová et al. (2020a), several differences are notable. Firstly, each sample only contained one single species of parasite. Secondly, no Trichuris sp. were detected, and thirdly, instead of the fish-tapeworm Diphyllobothrium sp. the tapeworm obtained from red meat, Taenia sp. has been detected.

Although Trichuris and Ascaris infect humans through the same (oral-faecal) pathway, it has been noted that the population-wide prevalence is often lower for Trichuris (Flammer et al. 2020). Interestingly, this was not the case for the samples in Goláňová et al. (2020a). Similarly, Brinkkemper and van Haaster (2012) found a significant prevalence of Ascaris over Trichuris in post-medieval Dutch cesspits.

The occurrence of Taenia sp. and the absence of Diphyllobothrium sp. is of interest as these parasites infect humans through undercooked or raw freshwater fish (Diphyllobothrium) or red meat (Taenia). While the samples in the previous publication (Goláňová et al. 2020a) showed Diphyllobothrium in two samples, indicating the consumption of raw freshwater fish, the occurrence of Taenia here indicates the consumption of raw pork or beef. As only a few samples in each set contain tapeworm eggs these changes have to be considered carefully. However, they are clear indications of dietary culture.

Discussion

Detection of the old ground surfaces

The original aim of the bioarchaeological disciplines involved in the project was to reconstruct the appearance of the non-built-up enclosed area at La Terrasse at the Bibracte oppidum, specifically to characterise its vegetation cover and its management during the lifetime of the oppidum. The information as to whether it was covered by trampled soil devoid of vegetation, grassland, heathland, or woodland was to contribute to the archaeological discussion on the function of this and similar places at the oppida or any other type of sites.

Based on existing archaeological evidence of uniform stratigraphy (Beck and Gruel 1989; Gruel and Beck 1996), it was anticipated that the flat surface of this non-built-up plateau was created together with the rampart at the latest during the 1st century bc. Based on stratigraphic rules, the “archive” of botanical and other materials recording the use-history of the oppida period should be situated towards the present surface incorporated into the modern soil’s A and/or B horizons.

However, the results of soil chemistry, soil micromorphology, OSL dating and to some extent plant macro remains analyses have indicated that the upper part of the soil profile developed on the deposit brought here in the early Middle Ages and that old ground surfaces are probably eroded or occur much deeper in the profile (Goláňová et al. 2020b; Lisá et al. 2022). Soil chemical data—most importantly soil cation exchange capacity (CEC) (Lisá et al. 2022) from the studied profiles—indicated that if present, old surfaces are buried at different relative depths in each profile.

To meet the original project aim, it was important to verify the presence of old surfaces and determine if there are layers and archives contemporary with the oppida period. This was inspired by the work of Dimbleby (1985), who recognised that even if “the soil pollen profile is not truly stratified” (Dimbleby 1985, p. 6) the change in absolute pollen frequencies (APF) (or currently used TPC—total pollen concentrations) is useful for detection of old land surfaces buried beneath the man-made structures. Increased values of TPC indeed indicated the presence of old surfaces in all studied profiles/sections (sometimes possibly more than one; A30, A60?, E50, E70, H60, I40, I50?; Fig. 5).

Dividing profiles into strata, determining their age

Variability in the depth of old surfaces is in line with the topography of the studied area. The position of APF/TPC defined surfaces doesn’t fully correspond with borders of strata defined on the basis of taxa (presence, change, or finds densities) of pollen, phytoliths, diatoms, charcoal, and seeds (Fig. 12). The shifts are usually not more than one sampled level. Parasites only occur in younger upper layers and the division could not be made based on them.

Fig. 12
figure 12

Bibracte, Mont Beuvray. Schematic stratigraphy of the test-pits at La Terrasse (A, B, D, E and H) and Les Grandes Portes (I). The “xxx” indicate the borders of the strata as defined by changes in particular taxa numbers, densities and/or composition, brackets indicate uncertainty; • indicates samples with present remains; small dot indicates samples with no finds; blank cells—samples not available. Colour-coding indicates time period of strata formation. The colours refer to dating of the strata and/or material in it as indicated in the last column. Graphic: M. Hajnalová

Full size image

The difference/shift of the position of borders of different strata, indicated by individual proxies (materials) within the same test-pit, are most probably the result of their different origin and pathways of their incorporation into the soil matrix. In natural soils, all phytoliths and most pollen derive from the local plant community. Only a smaller part of the pollen is air-borne from further afield. The archaeological deposit that contains relocated soil will contain pollen grains of the original place of the deposit, however, it also gets enriched on its “new surface” by local taxa. In addition, part of the pollen drifts downwards. Thus at any depth of a soil profile, there is pollen of various ages (Dimbleby 1985, p. 3). The origin and pathways of the introduction of plant macro-remains to soils are quite different. The local vegetation is rarely represented because the uncharred seeds usually do not survive. Charred seeds and wood charcoal are in most cases results of human activities, and most often are of non-local origin. They are brought to sites from the economic hinterland. Their remains cluster mostly in residence, storage, and waste disposal areas, or occur in places where deposits that contain them were transported. While firewood for household fires is usually expected to originate from the settlement’s vicinity, the fuel for metalworking (wood or already prepared wood charcoal) might originate from a considerable distance. Similarly, herb and vegetable gardens might be placed near the houses, the arable fields further away, on more suitable soils/terrains. Last, but not least, the movement of botanical micro- and macro-remains, and their distribution within the soil/deposit profile, is influenced by bioturbation, caused by soil fauna and or roots of plants. The dated macro-remains from La Terrasse are in the correct time sequences (with one exception), thus the non-human-induced movement of materials is for a limited distance—up to a dozen centimetres, probably less. Therefore, we consider all proxies in a layer/stratum as contemporary.

Despite the complexity of formation of the studied deposits and different pathways of introduction of individual materials, each material divides the section into three or four main strata.

Their time of formation (Fig. 12) indicated by the radiocarbon dates obtained on plant seeds and charcoal is also supported by regional palynology. For example, Castanea is recorded in Morvan in the La Téne period, or Tilia is present elsewhere at higher altitudes in the Eneolithic (cf. Jouffroy-Bapicot 2010; Jouffroy-Bapicot et al. 2013; Chevassu et al. 2019), and on Mont Beuvray itself also in the La Téne period (Goláňová et al. 2020a).

Due to similarities in stratigraphy recognised through botanical materials and their radiocarbon dating (Fig. 12), we describe separately section (I) from Les Grandes Portes and combine all sections (A–H) from La Terrasse.

Vegetation cover reconstruction

La Terrasse

The oldest stratum, with one or two possible later prehistoric surfaces in E80/90 and A60, seems to represent remnants of a prehistoric cultural layer composed of a mixture of weathered bedrock (indicated by low counts of phytoliths) and plant materials dated to the Mesolithic, Eneolithic, Hallstatt and pre-oppida Late Iron Age. The samples from lower layers contain the lowest proportion of NAP (max 30%) and indicate a forested landscape, earlier on made up of Quercus (oak) and Tilia (linden), later on of Fagus (beech). However, the most abundant Corylus (hazel) and Alnus (alder), which probably grew further away (the closest springs are 250 m away and down the slope), suggest opened forest canopy. This is supported by the evidence from wood charcoal, most of which is from Quercus (Fagus is less common), although Corylus and Cytisus scoparius (broom) comprise a substantial part of the sample. As the layer is mixed, it is impossible to specify exactly how the vegetation has changed. Nevertheless, the pollen samples from higher positions within this stratum indicate a more open landscape with spreading grasslands. Also, the dated macro-remains—charred hazelnut shells from the Mesolithic, oak charcoal from the Eneolithic and Hallstatt period and mineralised seeds of Sambucus from the pre-oppida phase, and occurring for the first time, Castanea (E80 and 90, as yet undated)—agree in general with the regional landscape and vegetation changes indicated by the palynology of several stratified peat-bog deposits in the Morvan (cf. Jouffroy-Bapicot 2010; Jouffroy-Bapicot et al. 2013; Chevassu et al. 2019). In contrast to peat bogs the data from La Terrasse soil indicate a more open character of the Middle and Late Holocene woodland and repeated and more intensive human presence in this part of the Morvan (or at least on Mont Beuvray). The occurrence of cereal chaff phytoliths suggests handling of crops in the area and cereal pollen (3%) presence of arable plots nearby, but not at La Terrasse or neighbouring La Chaume (Goláňová et al. 2020a). The situation of this stratum directly on the frost-weathered bedrock might be the reason why low phytolith numbers (E100 and E80) indicate “soil without contact with vegetation”. Aerotolerant diatoms that live on the moist surface of soils occur at the same depth where an old, buried surface was indicated by pollen and charcoal (A60, E80).

The protohistoric stratum formed from the middle La Tène to the Gallo-Roman period represents the period of the most intensive occupation of Mont Beuvray. It is, therefore, surprising that it is very thin and only detected in the central and northern parts of La Terrasse. The surface was recognised in A40 and E70. The pollen samples from the central (A) and northern (E) parts are almost identical. AP/NAP ratios indicate the further opening of Quercus and Tilia woodland (30–50% AP). The most prevalent is Corylus pollen. This might originate from managed woodland openings and/or hedges, possibly signalling the parcelling of the urbanised landscape of the oppidum. In NAP grasses prevail over arable and ruderal taxa percentages, suggesting meadows and/or pastures on the site and arable plots further away. Phytolith numbers are higher than in the previous stratum, and those from grassland vegetation prevail. The cereal signal, even though weak, indicates manipulation of the crops. It cannot be excluded that the sample in E originates from the material used for berm/rampart construction (not recognised when sampled) and was brought here from another part of Mont Beuvray. In terms of charcoal, Fagus is for the first time more common than Quercus and Corylus indicating vegetation change and/or the importing of fuelwood for metalworking. Charcoal from Tilia (at the time naturally rare in the area) and Castanea (probably at the time introduced to the region) suggests the cultivation of these or the importation of their wood. Local occurrence of these trees on Mont Beuvray is supported by on-site pollen (Tilia, this article and Goláňová et al. 2020a) and regional data (Castanea, Tilia) (Jouffroy-Bapicot et al. 2013). Only mineralised Sambucus seeds and uncharred Fagus seeds were present. The absence of charred seeds of crops, weeds, and ruderals in this layer supports a function other than habitation for this area.

The Early Medieval stratum was only recognised in the deepest sections near the bank in E and in H. Even if dated to the same period by plant macro-remains, the two sections give very different pictures. In the north, a surface in E40 was detected by pollen, diatoms, and seeds. Palynology shows a similar taxa composition, but more forested landscape than in the previous period. The lower charcoal sample resembles protohistoric conditions, but the upper sample is very different. It shows open landscape/scrubland with Cytisus, Corylus and Betula. Fagus is missing, and Quercus is accompanied by Prunus and Carpinus. In the south, the buried surface lies deeper (H60 and H70). Pollen samples are very similar to each other and show a fully open, grassland, landscape (AP max 30%, mostly Corylus). The most common charcoal is Fagus. As during this period soil/deposit was brought to La Terrasse (Lisá et al. 2022), it is tempting to interpret these striking differences in contemporary pollen and charcoal assemblages from the north and the south as being due to the different origin of the deposit. Yet, the AMS dates are in sequence, suggesting the gradual development of the soil profile within this stratum.

The most recent stratum, containing materials from the early/late Medieval period to the Modern era, is only characterised by plant micro-remains and parasite data. Palynological samples from all test pits are very similar. They characterise open landscape (but less open than in the Early Middle Ages) with similar proportions of the main tree taxa, basically reflecting the present situation on the site. Aerotolerant diatoms are found in five of 14 samples, and parasites in eight.

Les Grandes Portes

In section I, four strata can be recognised, though the differentiation and dating of the two lowermost are uncertain. The whole profile is very different from La Terrasse. Throughout the sequence the highest proportions in pollen are Quercus (ca. 25–40%), in charcoal Corlyus (50–70%) and in the middle section, there are more charred (38) than uncharred (13) seeds.

The lowermost, prehistoric/protohistoric, stratum, was originally interpreted as frost-weathered bedrock. It is characterised mostly by charcoal (and the absence of seeds). Charcoal mostly of shrubs or low-quality firewood suggests opened woodland. The pollen sample shows semi-opened Quercus and Corylus woodland with arable and grazed land in the vicinity. The presence of diatoms suggests a buried surface in the upper part of the stratum.

During the early Middle Ages, the landscape opened up even more and fields and pastures expanded. Castanea charcoal is present. Another surface is revealed by APF/TPC and diatoms.

The uppermost late Medieval to Modern period stratum witnesses gradual forestation and increase of Fagus shown by pollen. Interestingly, the pollen of conifers planted nearby after WWII, was not incorporated in the pollen record.

Landcover and its maintenance at La Terrasse during the late La Tène and Gallo Roman period

Here we discuss what could and what could not be deduced from botanical proxies about the character of on-the-spot vegetation cover, its change, and human activities in the area at the time of the occupation of the oppidum.

The thin stratum connected to the oppidum occupation phase of Bibracte was detected only in the central and the northern part of La Terrasse. In its lower part plant macroremains were admixed from three other periods (Mesolithic, Eneolithic and Hallstatt). This points to some form of soil disturbance at the beginning of its use in the La Tène period. Very rare cereal pollen and phytoliths indicate that disturbance was not caused by farming.

The extremely low numbers and densities of charred macro-remains are in strong contrast to the picture obtained from the habitation areas (Wiethold 1996, 2011). This, together with the scarcity of artefacts, demonstrates the absence of habitation activities. Furthermore, no settlement debris like waste containing fragments of objects, kitchen wastes, or refuse from workshops was detected, indicating that the area has been kept clean.

Phytoliths and pollen document that the area was covered by grassland, even if the grass phytolith indicators are always overrepresented in the phytolith record. Low counts, or absence of pollen or spores indicative of grazing, and rare occurrence of grasses/herbs in otherwise rich microcharcoal assemblages (detected in palynology and micromorphology) suggest that cutting (mowing?) was used to maintain the area as treeless. The absence of, or only occasional grazing is also supported by missing parasites and low values of nitrates and phosphates (Lisá et al. 2022, ESM 2–4). Mineralised Sambucus seeds and shrubs of ruderal stands show that in some periods the area was not maintained or that Sambucus was locally grown.

The repeated presence of Tilia pollen (and its quasi-absence from the charcoal) in both dry deposits at La Terrasse and waterlogged layers dated to the Late Iron Age (Goláňová et al. 2020a) at La Chaume—both areas in close proximity to a Gallo-Roman sanctuary—raises the discussed questions of the role of trees in the urbanism of the oppida (or Iron Age in general). Tilia trees, often underrepresented in pollen spectra, had been repeatedly found near Gallo-Roman sanctuaries from the 1st–4th centuries ad elsewhere in France, where they are thought to have been deliberately planted, perhaps as part of the construction of the sacred landscape (Zech-Matterne et al. 2018). Similarly in the local soil, the ubiquity of Corylus pollen, fruits, and wood shows that these shrubs were present intra muros and might have been used as hedges delimiting various boundaries.

Conclusions

The open, non-built-up areas that occur in most of the late Iron Age oppida and hillforts have long histories of use. They have often been reused, sometimes continuously for centuries. Even if never settled or built over, there might have been pastures, meadows, or fields.

If the aim is to reconstruct their vegetation cover and its management during relatively short periods of time, it is important to select, detect, collect, and interpret the suitable type of botanical evidence.

Combining the results of a wide range of different types of plant macro- and micro-remains and parasite data from dry soils and deposits proved to be a suitable approach. Even though the study of waterlogged deposits in archaeological features (such as ditches) may give quite a precise idea of the management of the surrounding area during a particular short period, the analysis of soil sediments/layers brings information about diachronic changes in the local on-the-spot vegetation. Of key importance is the correlation of the results from archaeobotany with those from geoarchaeology, OSL and radiocarbon dating. To allow for the best possible correlation, it is essential that different, and not only archaeobotanical, types of samples are taken from identical and/or clearly corresponding positions.

Radiocarbon-dated plant remains from an enclosed, non-built-up large open space at La Terrasse at Bibracte oppidum detected periods in which human activities left very little or no archaeologically visible traces, not only there but on the whole site. Archaeobotany and radiocarbon dates verified the results of geoarchaeology (chemical analysis, micromorphology and OSL dating) that the substantial mass of the deposits that fill the space enclosed by the protohistoric bank was brought in during the early Middle Ages. The combined results of all the studied botanical materials have demonstrated that La Terrasse has a much longer and more complex use-history than recognised through traditional archaeological analyses.