Procurement and mobility during the Late Pleistocene : Characterising the stone-tool assemblage of the Picamoixons site ( Tarragona , NE Iberian Peninsula )

The Picamoixons site is a rockshelter located in the province of Tarragona (NE Iberian Peninsula). It was object of two rescue campaigns during 1988 and 1993, which led to the recovery of a complete archaeological assemblage, including stone tools as well as faunal and portable art remains that date the occupation to the 14 to 11 millennium BP (calibrated). This study involves a petrographic characterisation of the stone-tool assemblage in order to establish: 1) the procurement areas, 2) the raw materials management strategies and 3) the mobility radius and territorial sizes of the hunter-gatherers groups that occupied the site. The method applied comprises in a multiscale analysis that includes systematic prospection, the petrographic characterisation of geological and archaeological samples, an analysis of the chert types represented in the knapping sequence, and the definition of the mobility axes and areas frequented according to lithic procurement. A petrographic analysis of the chert in the prospected area led to the definition of nine macroscopic varieties related to five types (Vilaplana, Morera, Maset, Vilella and Tossa cherts), related to Lower and Upper Muschelkalk (Triassic), Lutetian, Bartonian (Palaeocene) and Sannonian (Oligocene) deposits.The study of the knapping sequences indicates the main exploitation of Bartonian cherts (Tossa type), and the use of Lutetian cherts (Maset and Morera types) for configuring retouched tools. The exploitation of the remaining raw material types identified is considered sporadic and opportunistic.Defining the procurement areas enabled the mobility radius to be assessed as between 3 and 30 km, highlighting the importance of the fluvial basins as natural movement pathways. The results indicate that the main procurement territory was 16 km2 in area, associable with a forager radius. The most remote procurement distances suggest a maximum exploitation area of 260 km2, defining an intra-regional range. This range presents parallelisms with various contemporaneous hunter-gatherers groups in Western Europe, suggesting a progressive mobility reduction dynamic during the Late Pleistocene-Initial Holocene.


Introduction
An absence of regional petrographic studies has made it necessary to characterise the lithic raw materials used by the Late Glacial hunter-gatherers groups around the Prades range and Camp de Tarragona region (Tarragona, NE Spain).This paper presents a systematic petrographic analysis of the lithic resources available in the contact area between the Ebro distal basin and the central sector of the Prelittoral Catalan Mountains, as well as the lithic assemblage from the Picamoixons site.
For this purpose we petrographically characterised both the geological and archaeological records to achieve our three main objectives: 1) to define the procurement areas; 2) to analyse the distribution of the different chert types in the reduction sequence; and 3) to delimit the mobility radius, the frequented areas and the territorial sizes of the huntergatherer groups according to the lithic procurement data.

Picamoixons: site and stratigraphy
The Picamoixons site is located in the left bank of the Francolí river (UTM -ETRS 89 31N 348399.2E x; 4574166.7 N y) in a contact zone between the Central Catalan Depression and the Camp of Tarragona region, and close to the La Riba Strait, a connection between the Prades and the Miramar ranges (Vergés 1989).
The site was discovered in 1972 and between 1988 and 1993 it was excavated as part of a rescue program due to the deterioration of the archaeological deposits (Carbonell et al. 1989) (Figure 1).
The first intervention led to the definition of a sequence formed by two main units.The CI, at the top, relates to the collapse of the rockshelter cornice.CII, at the base is a 2.40m thick unit, formed by clays, silts, cobbles and boulders.Five different layers (CIIA to CIIE) were differentiated, three of them (A, B and D) presenting an archaeological record (Vallverdú 1994), although only the CIIA has been excavated.
A reanalysis in 2005 (Angelucci 2005) led to the partial restructuring of the stratigraphy: BR at the top, equivalent CI; four further units (CI1-CI4), equivalent to the CIIA-CIIC units; the TC unit, corresponding to the CIID; and finally, the Fl layer; at the base, formed by two units (Fl1 and Fl2), and relating to the fluvial deposits of the Francolí river.
The excavation was focused on the chronocultural characterization of the human occupations.The five 14 C results obtained date the Picamoixons sequence between 13-10 kyr cal BP (Table 1).The two field seasons allowed the recovery of rich archaeological record comprising faunal, technological and artistic elements.The faunal remains are dominated by lagomorphs, Capra sp. and Bos or Bison, with Equus and Cervus also being recorded.Furthermore, three human dental pieces and a fragment of distal phalanx from at least two individuals were also recovered.The lithic assemblage is composed of ca.900 remains, and displays the classical dominance of endscrapers and backed blades that regionally define the Late Glacial and Early Holocene periods (Morales 2012).Recently, the lithic assemblage from the CIIA layer has beem subdivided into two separate assemblages according to spatial distribution and the degree of patination (García Catalán et al. 2009).
Additionally, two fragments of portable art, a limestone slab with pigment strokes forming seven a crayon-lines were also documented (García Díez et al. 1997).

Materials
A regional lithoteca comprising more than 100 hand samples has been created as a reference collection.From these hand samples a total of 60 uncovered thin-sections have been prepared (Figure 2).This collection was used for comparative purposes during the analysis of the archaeological assemblage.The assemblage recovered during the 1993 field-season was selected for this study.These samples are characterised by the dominance of knapping products (67%) retouched elements (10%), cores (4%), fragments (13%) and natural bases (6%) (Figure 3).The relative variability of raw materials including limestone, schist, slate, quartz, sandstone and porphyr has been also documented.
In this work we only present the results of the chert analysis, not including the nonsiliceous rocks.During the work, the 16% of the sample was also excluded from the analysis due to the presence of heavy fire damage and several alterations.

Methods
The methods applied in this study consisted of a multiscale analysis, with the purpose of creating a catalogue of empirical descriptions for interpreting lithic procurement strategies and territorial management.
These analyses are grouped into four research aspects: the prospection, a petrographic characterization of the geological and archaeological record, statistical analysis to the employed raw materials, and the definition of the exploited areas and mobility routes.

Prospection
Cartographic and bibliographical research was used to delimit the geological chertbearing units, in order to establish the baseline information for the fieldwork.The field surveys allowed the outcrops to be located and catalogued, and both the siliceous nodules and their enclosing rocks were systematically sampled to reflect the variability and availability of raw materials in the landscape.

Petrographic characterisation
The macroscopic and microscopic characteristics of the reference collection and the archaeological assemblage were petrographycally analysed.
The morphoscopic characters observed a visu and as well as with a binocular microscope (at x32/40 magnification) to classify both the external traits and the internal features of the rocks (e.g., colour, transparency, texture, sedimentary structures, fissures, and inclusions…).
The thin-sections were analysed using a polarising light microscope (at x20/200 magnification) to define the texture and mineralogy of the siliceous and non-siliceous components.

Raw materials in the archaeological record
The petrographic analysis of the reference collection enabled a classification of the raw materials recovered at the Picamoixons site as well a statistical analysis of the distribution percentages in the structural categories of the knapping sequence.
The Chi-square distribution test was also applied in order to define the raw materials management dynamics at an intra-site scale.
This test (X 2 =Ʃ (O-T) 2 /T) is based on comparing the data observed in the archaeological assemblage (O) with respect to the theoretical values (T = (n remains associated to X structural category * no.remains of X raw material) / Total archaeological remains) that the CIIA level must present in order to consider its distribution random (X 2 < 3.84).For this reason, values above 3.84 inferred to represent differential management patterns in the exploitation of raw materials (Tarriño 2006).

Lithic procurement: mobility, frequented areas and territorial sizes
Combining the previous results allows a definition of the procurement dynamics and exploited territories, delimiting the principal areas frequented by the groups that occupied the Picamoixons site.
These most-frequented areas point to mobility routes, which we have mapped using a GIS procedure that defines the territories with least displacement cost according to the slope gradients (Binford 1983;Brannan 1992;Elston 1992;Hodder & Orton 1976).All these results allowed us to establish the mobility type and dimensions of the procurement territories (Dyson-Hudson 1978;Féblot-Augustins 2009;Grove 2009;2010;Wobst 1974).

Prospection and available chert resources
A total of 56 different primary chert outcrops were defined after the prospecting campaigns around the Montsant, Prades and Miramar ranges.These outcrops are related to four geochronological units: the Lower and Upper Muschelkalk (Triassic), Lutetian (Paleocene) and Sannonian (Oligocene) (Soto 2015) (Figure 4).
Five siliceous types were identified according to their petrographic characteristics and these were named according to the place name of the model outcrop, defined by the highest chert abundance ratios.These are Vilaplana chert (Vilaplana), Morera chert (Morera del Montsant), Maset chert (Ulldemolins), Vilella chert (Albarca) and Tossa chert (Montblanc).
These chert types presented several morphoscopic varieties that forced us to establish a catalogue containing the various recognizable macroscopic features.In the end, the reference collection comprises a total of 17 chert subtypes that can be compared to the archaeological assemblage from the CIIA layer.

Petrographic characterisation
The comparison between the lithoteca and the archaeological material from the archaeological CIIA layer of Picamoixons confirmed the presence of nine (Figure 5) of the chert sub-varieties described (Table 2).
The crossed-polar analysis shows that the dominant texture is microcrystalline quartz, as diagenetically replacing the carbonated matrix (Figure 7).Spherulithic chalcedonite and quartzine and lutecite complete the spectrum of siliceous components forming in the intraparticular pores and cementing together the allochemical elements.

Morera chert (Morera del Montsant)
Two macroscopic varieties of this Lutetatian chert (Morera 1 and 2) have been described in 32 remains from the CIIA assemblage (Figure 5 c and d) Morera1 is characterised by a whitish and greyish colours (Very Light Grey N8, Yellowish Grey 5Y 8/1, Greyish pink 5R 8/2) and opaque and matt surfaces, with fine or very fine textures and frequent fissures.There are occasional lenticular and tabular gypsum crystals, iron oxides and contorted dark patterns, related to the presence of chicken-wire anhydrite (Ortí 2012).
The microscopic analysis reveals a siliceous body mainly microcrystalline in texture with spherulitic length-slow fibrous quartz (40-50%) (Figure 8).Microsparitic calcite (20-100 μm), micritic mud (10-20%) and ooids with microcrystalline cement have been also described and are related to the primary depositional texture.Furthermore, isolated lenticular gypsum crystals (300-500 μm), clay minerals and globular aggregates and fine-grained haematite have been described in vugs and fissures.The features related to the primary depositional textures suggest a siliceous replacement process where the microcrystalline quartz substituted the matrix and the fibrous quartz cemented the dissolution hollows forms obliterated.

Maset chert (Ulldemolins)
Two macroscopic varieties (Maset 2 and 3) (Figure 5 e and f) have been identified in 75 remains form the archaeological assemblage.
Maset 2 is a white-coloured chert (Very Light Grey N8, White N9), with translucent surfaces and a fine or very fine texture.There are occasional carbonate inclusions, iron oxides and evaporitic relicts (gypsum pseudomorphs and lenticular crystals or chicken-wire anhydrite patterns).This variety presents endocortical laminations as sedimentary structures, but there are also post-diagenetic fissures present.
Maset 3 is a reddish-coloured chert (Pale Reddish Brown 10R5/4, Moderate Red 5R 5/4), with translucent and matt surfaces as well as a fine or very fine texture.Iron oxides and evaporitic relicts, gypsum pseudomorphs, lenticular crystals, and anhydrite structures have been also observed.They present occasional laminations and probably biogenic escape figures, related to subaerial exposure and karstification processes.There are also isolated cavities, less than 1cm, with gypsum and halyte crystals-covered walls.Some remains associated with this raw material have a high degree of white patina due to superficial silica remobilisation (Fernandes 2012;Schmalz 1960) causing a banded aspect in the colour distribution.
The isometric microcrystalline quartz, caused by cement replacement, is the main siliceous component, accounting for 50-60% of the samples.Macrocrystalline isometric quartz associations (100-120μm) are occasionally described filling the cavities left by the chemical dissolution of gypsum.
The petrographic analysis defines these materials as being a product of the diagenetic replacement of nodular primary gypsum.
It is characterized by reddish mottling (Pale reddish brown 10R 5/4, Moderate Red 5R 4/6, Dusky Red 5R 3/4), with matt and opaque surfaces and fine or very fine textures.Evaporitic crystals and relicts, forming enterolithic patterns, and iron oxides inclusions, have been observed.There are laminations and pressure figures, or stylolites, related to biogenic processes and phases prior to silicification.
Microcrystalline isometric and cryptocrystalline quartz (50-60%) are the main components seen under crossed polarized light.These represent the siliceous cement of a primary lutitic matrix.Fibrous quartz of both elongations is also present: quartzine and lutecite have replaced the gypsiferous pseudomorphs and chalcedonite cemented the occasional carbonate components prior to the silicification (Figure 10).This chert type presents frequent primary microstructures, tunnels or shafts, related to desiccation cracks or moderate biological activity, and is cemented by crypto-and microcrystalline quartz.
Secondary porosity or vugs, the product of chemical dissolution, and fractures, filled by secondary alabastrine gypsum, anhydrite and evaporitic breccias are also described.
This chert originated from the diagenetic replacement processes from gypsilutitic-oxiderich facies, with chemical dissolution, subaerial exposure and bioturbations.

Tossa chert (Montblanc)
Two varieties, Tossa 1 and 2, have been observed in 141 lithic remains (Figure 5 h and i).Tossa 1 presents grey colours (Light Grey N6, Medium Grey N4), matt and translucent surfaces and fine or very fine textures.Carbonate inclusions, iron oxides and lenticular gypsum can also be identified.Allochemical elements are poorly represented, suggesting silicification from mudstones or crystalline carbonates.Tossa 2 is characterised by blackish colours (Medium Dark Grey N4, Greyish Black N2), opaque surfaces and fine or very fine textures, as well as the presence of iron oxides, evaporite relicts, allochemical elements and a mudstone or crystalline carbonate chert texture.The siliceous body is composed of isometric microcrystalline quartz, lenticular gypsum pseudomorphs, prismatic microgranular anhydrite, ocassional ferruginous ooids, massive haematite, pyrite and fluorite.
The micro-and cryptocrystalline quartz cement the micritic matrix that is partially preserved in the samples.This matrix supports sparitic calcite, gypsiferous pseudomorphs and anhydrite microcrystals with fine haematite inclusions (Figure 11).
The moldic porosity of the gypsiferous relicts is cemented by quartzine and lutecite, while the carbonates are occasionally replaced by botroydal and spherulitic length-fast chalcedony.Thin-sections show frequent dissolution microstructures, defined as vugs.Clay minerals, brecciated gypsum crystals, partially cemented by tiled and puzzle macrocrystalline quartz (up to 300 μm), and equigranular sparitic carbonates related to microcodia (Figure 12) originated from the calcification of vegetal matter (Stoops 2003) are observed inside these vugs.
Planes and fractures are filled by alabastrine gypsum or by blocky-type calcite, with palisadic macrocrystalline quartz along their boundaries.
The petrographic analysis indicates that Tossa chert is product of early diagenesis from marly limestone in a phreatic or buried environment according to the cortical areas analysed and the similarities established with the enclosing rocks.
14% of the archaeological assemblage presents cortical surfaces, and therefore could be associated to the first knapping episodes.An association of the chert-types to the different structural categories points to the Tossa 1 and Maset 2 as being the most exploited and the only chert varieties seen in each category (Table 3).The analysis (Table 4) shows a homogeneous distribution for the Tossa 1 type refuting the existence of differential exploitation strategies for this chert.On the other hand, the results for the Maset 2 variety indicatean overrepresentation flakes and a lower number of retouched elements with regard to the expected values.
The absence of cores in most of the raw materials defined does not provide significant values in the distribution test, but justifies the overrepresentation of the retouched tools of Morera 1 and Maset 3 varieties.The absence of retouched tools of Morera 2 also causes the significant overrepresentation of cores.
The chi-square test indicates the existence of differential management strategies for the Maset and Morera cherts, proposing the selection of certain varieties for the retouched tools.In other cases, a significant absence in the assemblage, suggests their transport or disposal away from the site.
The low number of remains associated to the Vilaplana, Vilella and Tossa 2 varieties, and their classification as knapping products, suggests they were exploited sporadically and opportunistically.

Discussion
The petrographic analysis of the chert varieties defined in the assemblage of Picamoixons site determines the potential procurement of the Vilaplana chert from the outcrops located in the Francolí and Siurana basin.The procurement of Morera, Maset and Vilella siliceous rocks types is related to the outcrops sampled around the Siurana, Montsant and Francolí basins, and the procurement of Tossa cherts is established as being from the upper and middle Francolí river basin.
The definition of these provenances suggests a mobility radius between 3 and 25 km for the Vilaplana cherts; between 27 and 30 km for evaporitic varieties (Morera, Maset and Vilella types); and distances of 3-7 km for the procurement of Tossa chert.The areas of least displacement cost indicate the importance of the fluvial basins as natural pathways.The middle basin of the Francolí River is the access route to the outcrops located around, and the upper basins of the Francolí and Montsant are the most important areas for procurement in the western sector of the prospected area (Figure 14).The definition of this procurement area suggests an exploited territory of around 16 km 2 , including the mobility axes in an immediate-local range, associable with a forager radius.The most remote procurement distances suggest a maximum exploitation territory of 260 km 2 , which is therefore defined as an intra-regional range (Féblot-Augustins 2009).
This long-distance tendency seemed to be reversed during Magdalenian, due to a decrease in procurement distances.Some authors explain this tendency by the standardization of knapping strategies, a potential demographic increase or the combination of both factors, as well as other possibilities (Brown & Douglas Price 1985;Kelly 1992).
This reduction is still more visible in the Catalonian and Portuguese sites (e.g., (Bicho 2002;Bicho et al. 2009;Fullola et al. 2006;Terrades 1992) during Late Magdalenian and Mesolithic periods, where lithic provenance studies indicate the dominance of local materials, with maximum procurement distances being around 50 km.
The CIIA lithic assemblage at Picamoixons could be part of this progressive mobility reduction dynamic.In this particular case, the short-distances travelled could also be the consequence of the high resource availability and accessibility of resources, making the study area a territory with high raw material predictability.This scenario suggests a model where continuous provisioning is possible, allowing a high level of regional mobility that progressively led to reduced mobility in the last hunter-gatherer groups (Kelly 1992).

Conclusions
The main points that can be drawn from study of the lithic raw materials from the CIIA level of the Picamoixons site are: • Nine macroscopic chert varieties were exploited related to five siliceous rock types described from around the Montsant, Prades and Miramar ranges.
• There are complete in-situ knapping sequences of the most-commonly represented lithic supports.
• There were differential management strategies related to the Lutetian evaporitic cherts.
• The outcrops located in the middle Francolí basin, and in the head of the Montsant and Siurana basins are the potential procurement areas, suggesting a catchment model based on the frequentation of high availability areas.
• The circulation axes show a NNW direction, in line with the fluvial basins, and the areas of least displacement cost.
• The potential procurement areas suggest an exploitation area coinciding with a forager radius.

Figure 2 .
Figure 2. Thin-section and analysis equipment at the Geoarchaeology facilities at IPHES.

Figure 6 .
Figure 6.Thin section images of Vilaplana chert.Left: The laminated carbonate elements (Liesegang rings) create a packstone texture.Right: The samples in crossed polarised light reveal a differential siliceous content configuring the xyloidal aspect.

Figure 7 .
Figure 7. Xyloidal chert from the Picamoixons assemblage.Left: Images with plane polarizer where the primary depositional texture and Liesegang rings can be seen.Right: Crossed polarised light reveals the presence of micro-and cryptoquartz as well as detrital quartz.

Figure 8 .
Figure 8. Thin section of Morera1 chert.Images with normal polariser (left) and crossed polariser (right).Detail of a dissolution cavity filled with clay minerals, gypsiferous pseudomorphs, and fibrous and microcrystalline quartz.

Figure 9 .
Figure 9. Thin section images of Maset 2 chert.Gypsum pseudomorphs with haematite concentrations (Above); chemical dissolution cavities cemented by macrocrystaline quartz in mosaic (centre); replacement of the primary depositional texture by microcrystalline quartz and length-positive fibrous quartz (down).

Figure 13 .
Figure 13.Distribution of the chert varieties identified in the Picamoixons assemblage.

Figure 14 .
Figure 14.Map of the areas with lowest displacement cost marked in blackish colours.

Table 1 .
Radiocarbon dates from the Picamoixons site.

Table 3 .
Chert type distribution according to the structural categories of the knapping sequence.