Agia Marina and Peristereònas: Two New Epipalaeolithic Sites on the Island of Lemnos (Greece)

The surveys carried out along the coasts of the island of Lemnos (Greece) have led to the discovery of new Late Epipalaeolithic sites at Agia Marina and Peristereònas. Peristereònas yielded a knapped stone assemblage that is strictly comparable with that from Ouriakos, a site located along the south-eastern coast of the same island, while the artefacts from Agia Marina are more problematic to interpret because they are probably to be attributed to a slightly different period. However, the most characteristic artefacts recovered from the sites are represented by microlithic geometrics obtained by abrupt, bipolar, or direct retouch, end scrapers, and different types of exhausted cores and technical pieces, which help us reconstruct the operational sequence employed for the manufacture of the armatures. The aim of the paper is to contribute to the interpretation of the characteristics of the Late Epipalaeolithic assemblages discovered on the island and to frame them into the general picture of the end of the Pleistocene in this part of the Aegean. The artefacts from the sites show unique characteristics, without parallels to the knapped stone assemblages of the same period so far recovered along the coasts of the Aegean Sea, the eastern Mediterranean, the Levant, and the Black Sea.


Introduction
Excavations carried out during the last 11 years at Ouriakos, an open-air site located along the south-eastern coast of the island of Lemnos (Greece), and the surveys that followed its discovery, have improved our knowledge of the Epipalaeolithic in this part of the northeastern Aegean Sea (Efstratiou, 2014;Efstratiou et al., 2014).
of natural processes. The dune deposit, c. 70-cm thick, covers the fissured limestone and sandstone basements (Anifadi et al., 2016), which is typical of the geology of the area (Fig. 4, bottom).
Site 1 (27 m asl) is located on a terrace c. 100-m wide delimited to the north and south by two narrow, seasonal streams. The site looks like an oval spot of concreted sand (light olive-brown: 2.5Y5/3) on the top of which two scatters of artefacts were recovered. The western one (Spot 1) covers an area of c. 300 sq. m., while the eastern one (Spot 2) c. 150 sq. m. (Fig. 5). The artefacts were collected from the surface during separate surveys conducted between 2012 and 2021 following different  Biagi, 2018, drawing by E. Starnini) methods. Two systematic collections were made in 2013 and 2021. During the 2013 season, 204 artefacts were retrieved from the western Spot 1. Their position was recorded according to squares of 1 × 1 m covering an area of 18 × 18 m. The artefacts were collected either from the eroded surface or from the topmost part of the concreted sand deposit, which is most probably a Pleistocene fossil dune (Fig. 4).
During the 2021 season, a further assemblage of 174 artefacts was collected from the entire surface of the site. On this occasion, the position of each piece was recorded by a Garmin-GPS device (see supplementary databases 1 and 2). The distribution map of all the artefacts from the two spots is shown in Fig. 5. Besides Fig. 4 Agia Marina, Site 1: Erosion area of the surface collection with the location of Spot 2 (top) and fissured bedrock below the sandy soil (bottom) (photographs by P. Biagi, 2021) debitage, debris, and other pieces discarded during the different stages of core knapping and maintenance, only a few, irregular, unretouched bladelets and microbladelets were found. The retouched artefacts are even fewer, represented by a few end scrapers and backed tools (Fig. 7,nn. 3,4,and 6;Fig. 12,n. 1;Fig. 14,(5)(6)(7).
The surface collections made from Site 1 during the last 10 years yielded altogether 420 artefacts, most of which come from the two spots reported above (see Fig. 5). They are most probably to be interpreted as manufacturing areas, due to the presence of exhausted, discarded cores, technical artefacts, debitage, and debris pieces (Keeley, 1991, p. 258). Moreover, no difference has been observed in the typological composition of the artefacts from the two spots. Consequently, the techno-typological analysis has been performed on the entire assemblage.
The analysis of the lithic assemblage has taken into consideration the following main features of the artefacts: (1) morphology and size of the striking platforms, where preserved, (2) number and debitage axis of the negative scars, (3) presence of cortex, and (4) dimensions. The last-mentioned included the recording of the maximum length, width, and thickness of all the artefacts, and the attributes of the knapping technology (platform, bulb, etc.). The artefacts were also divided into cores, fragments, angular wastes, debris, flakes, and blades (see supplementary databases 1 and 2). These data, along with the presence of cortical pieces (divided into distinct percentage categories), allow for the reconstruction of the reduction sequences and the "chaîne opératoire".
In several cases, traces of crushing have been observed at the lower end (tips) of the cores (Fig. 7, n. 1b; Fig. 11, n. 10b), a marker of the use of direct percussion on an anvil using a hard hammer. The angle between the striking platform of the cores and the flaking surfaces is less than 90°. The main debitage goal was to detach small, narrow laminar blanks with a triangular cross-section (Fig. 7,nn. 7 and 11) to be abruptly retouched to obtain backed microbladelets (Fig. 10, nn. 5, 6, and 12) or microlithic geometric armatures. Among these latter are 1 lunate (Fig. 10, n. 7), 3 backed points (Fig. 7,n. 4;Fig. 9,n. 10;Fig. 10,n. 9), and 1 backed and truncated piece (Fig. 7, n. 3).
The numbers of the different types of artefacts and tools from the two spots is shown in Table 1. The dimensional diagrams developed measuring the length, width, and thickness of 145 complete, unretouched pieces show that the assemblage consists mainly of blanks of hypermicrolithic (<1.25 mm) and microlithic (1.25-2.50 mm) dimensions, which together represent 79% of the total (Fig. 6).
As reported above, no substantial difference has been noticed in artefact typology, technology, and raw material use between Spot 1 and Spot 2, except for the number of finds and the state of preservation of the deposit. Spot 2 is smaller and looks more eroded than Spot 1. Therefore, we can interpret the lithic clusters as two distinct, more or less contemporaneous knapping floors, showing the ephemeral settling of one or more groups of mobile hunters.
Site 2 (17 m asl): c. 50-m south-southeast of Site 3, a small assemblage of 69 knapped stone artefacts was collected in 2015 from the surface in a small area without vegetation, just south of the same road ( Site 4 (27 m asl): This site is very interesting, although its cultural and chronological attribution are uncertain because the only lithic tool discovered does not find parallels with any other Epipalaeolithic artefact ever recovered from the island. One long, straight perforator obtained by abrupt, deep, direct retouch ( Fig. 8, n. 11) was collected from the bottom of a profile opened by natural events in a deposit of concreted sand of dark yellowish-brown colour (10YR4/4) c. 80-cm thick. The site is located c. 70 m east-southeast of Site 1 (Fig. 3, Site 4).
Site 5 (13 m asl): Two different collections were made in 2017 and 2021 in a parking place opened by a bulldozer a few years earlier on a terrace facing the sea (Fig. 3, Site 5) (see supplementary databases 1 and 2). The site was probably buried by a Holocene dune. The artefacts come from small clusters, which were exposed during the construction of the car park. They are represented by cores (Fig. 16, nn. 6 and 9), 1 simple burin (Fig. 16, n. 2), 1 sidescraper, and a few unretouched artefacts.
Site 6 (11 m asl): Is located within the same parking area of Site 5, just a few metres north of it, at the beginning of the pathway that leads to the beach (Fig. 3, Site 6). Many artefacts were collected from the surface in the summer of 2017, among which are different types of cores, 2 end scrapers (Fig. 16, n. 8), crested blades, technical pieces, and unretouched artefacts (Fig. 16, n. 10).
Apart from the sites reported above, a few more isolated knapped stone artefacts of uncertain attribution were recovered from the area of Agia Marina. According to their typological characteristics, they are probably to be attributed to different periods of Palaeolithic frequentation (Fig. 16, nn. 1, 3-5, and 7). Some of the artefacts were collected in 2012 next to a natural profile that opens just to the west of Site 4 at c. 27 m of altitude (39°49′55.80′′N-25°20′41.60′′E). Two more were found during a brief survey made in 2013 (Fig. 16, nn. 5 and 7).
To sum up, in addition to Site 1, which yielded the most consistent number of finds, the other sites have also yielded important data, which improve our knowledge 5 Page 20 of 34 of the behaviour of the final Pleistocene hunter-gatherers who settled in the region and help interpret the complexity of their mobility patterns (Kent, 1991). The unique conditions of the island, whose coastal area is still in a reasonably good state of preservation, provide an occasion for recording traces of prehistoric occupations. Apart from cases of land exploitation for present-day and historical agro-pastoral activities, some territories are still little urbanised. However, the new policy of tourism development has favoured the exploitation of some coastal zones, increasing accessibility to beaches, and opening new earth roads and parking areas. Consequently, on the one hand, this policy may have caused the loss of archaeological sites, but on the other, it has favoured the discovery of new sites. In our case, the coastal surveys carried out over the last decade, have shown that groups of Epipalaeolithic hunters repeatedly frequented some localities which were particularly suitable for settlement due to their natural resources.

Peristereònas
The site of Peristereònas was discovered during the summer of 2017, located on a small, heavily weathered promontory that protrudes towards the sea along the northeastern coast of the island (Fig. 1, n. 3; Fig. 17). At present, the cape is covered with sand and spots of small Centaurea spinosa bushes. The area immediately inland is rich in springs and small, seasonal streams. The only systematic surface collection of knapped stone artefacts was made in the summer of 2018 within a grid of 10 × 10 m. The finds were recovered from an area of c. 95 sq m, with a collection unit of 2 × 2 m due to the low visibility of finds (Fig. 17, bottom). A small trial trench, measuring 1 × 2 m, was opened in 2019 to check for the presence of residual archaeological deposits in situ (Fig. 19, top). Unfortunately, the test pits showed that the site is eroded; the deposit yielded no evidence of archaeological features or charcoal remains, and the bedrock lies c. 20 cm below the land surface. Just four knapped stone artefacts were recovered in the first 10 cm by dry sieving (see supplementary database 3).
The industry from the surface collection made in 2018 consists of 588 artefacts (Table 2). A further twenty-two pieces were collected in 2017 and 2019. The spatial distribution of the artefacts is biased by the bush cover that affects the visibility in many squares (Fig. 17, bottom). The finds seemed to be randomly concentrated in a few squares. Most specimens consist of undiagnostic fragments, blanks, debitage, and debris pieces. Many artefacts are in a fragmentary state (536: 90.54%), a few are burnt (61: 11.38%), or heavily weathered (see supplementary database 3). Forty complete, unretouched artefacts were measured to develop the length/width/ thickness scattergram of Fig. 6, top. The results show that 75% of the unretouched pieces are of hypermicrolithic and microlithic dimensions. Despite the low number of artefacts available, the general picture is comparable with that obtained from Agia Marina Site 1 (Fig. 6, bottom: 79%).
The Peristereònas artefacts were produced locally from raw material retrieved from the same local sources exploited by the Agia Marina and Ouriakos inhabitants. The absence of any exotic or imported raw material is significant. It reinforces the suggested chronological and cultural attribution of the assemblage.

Raw Material Provenance and Exploitation
The surveys conducted for the identification of the lithic sources have shown that all the raw materials for making artefacts are available on the island within 15-20 km from the sites (see Fig. 1). So far, we have identified two primary sources. One, of sedimentary origin, is the Ifestia Unit (Innocenti et al., 2009), which outcrops in a few well-defined zones of the island. At present, it is exploited as a gravel for road construction. Over the last 20 years, a few gravel quarries have been opened around Moudros Bay from which it is possible to collect hundreds of chert and radiolarian pebbles, which are identical to those retrieved from the archaeological sites for making artefacts (Fig. 1b and Fig. 2b). Another abandoned gravel quarry, opened to exploit the same sedimentary formation rich in chert pebbles, is located on a river terrace along the right side of the Havouli Valley ( Fig. 1a and Fig. 2a1). The second type of source consists of hydrothermal siliceous rocks. They occur as seams in volcanic formations ( Fig. 1a and Fig. 2a2), outcropping along the eastern slopes of the Kalogiros Hill, in the uppermost part of the same valley (Efstratiou et al., 2014, Fig. 10). This raw material was exploited until a few decades ago by local farmers to make threshing sledge inserts. The large areas of debris flakes, and the percussion scars still visible on the multicoloured Kalogiros outcrop blocks, testify to this activity (Biagi et al., 2015). Another outcrop of siliceous hydrothermal rocks has been located near the village of Rossoupouli (Fig. 1c and Fig. 2c).
Finally, a secondary source of both chert and hydrothermal rock pebbles is the Agios Ioannis Prodromos Bay (Fig. 1d and Fig. 2d) where the beach is rich in all these raw materials. The chert pebbles collected from the shore of the bay show the same cortical scars caused by wave rolling that have been observed on some archaeological pieces (Fig. 7, n. 10). They demonstrate one of the ways used by prehistoric groups to collect raw materials. Other rocks were also employed, such as quartzite, which is also available from the Ifestia Unit conglomerates that outcrop at the northern edge of the Peristereònas promontory (Fig. 19, bottom).

Discussion
None of the sites discussed in this paper yielded organic material suitable for radiocarbon dating. However, the techno-typological analysis of the knapped stone assemblages can help us establish their relative chronology with some confidence. To give an example, a recent review of the MIS2-Late Upper Palaeolithic assemblages of eastern Central Europe (see Lengyel et al., 2021) has shown that the initial Late Glacial Maximum (LGM) assemblages are characterised by domestic tool dominance and the frequent use of flake tools, while the post-LGM assemblages are correlated with armature dominance and blade and bladelet tools.
A preliminary study of the Ouriakos lithic assemblages has shown the predominance of bipolar-retouched microlithic lunates. They were manufactured without the microburin technique, following a unique technology, which at present has no parallels elsewhere in the Aegean (see Efstratiou et al., 2014, p. 5 and Fig. 14). This observation is important because it is most probably a cultural factor, which challenges its suggested affiliation with the so-called Antalyan, since there are marked differences between the lunates from Ouriakos and those from the Antalya caves (Öküzini in particular), among which are the technology of production, methods of retouch, and dimensions.
The term Antalyan was first introduced by S. K. Kozłowski (1994, p. 145) to frame the Late Epipalaeolithic assemblages found in the Gulf of Antalya caves (Turkey), which according to Kozłowski are characterised by "small backed pieces, segments and isosceles triangles", into the final Pleistocene archaeology of the Levant. However, the term has been employed on very few occasions, mainly to emphasise the importance of lunate microliths in the Öküzini Cave cultural sequence (Kaczanowska & Kozłowski, 2013, p. 18), despite their variable typological characteristics (Kartal, 2003, p. 50).
Although assemblages with different types of microlithic lunates, chronologically attributed to the YD event, are known from a few sites along the entire coast of the southern and western Anatolian Peninsula, from Direkli Cave in the east (Arbuckle & Erek, 2012), to Lemnos and the Karaburun Peninsula in the west (Çilingiroğlu et al., 2020), all the lunate complexes show different techno-typological and dimensional characteristics. Moreover, they are different from those of the same period known in the Levant (Belfer-Cohen & Goring-Morris, 2021, Fig. 2).
Ouriakos is a unique site within the general, complex framework of the Aegean and north-eastern Mediterranean Epipalaeolithic. This is due to two main reasons: (1) our knowledge of the Aegean Epipalaeolithic, its origin, and chronological subdivision is very limited (Galanidou, 2011). Moreover, we do not know if, and how, it eventually took part in the formation and development of the Early Holocene Preboreal assemblages of the area, which likewise are almost unknown (Çilingiroğlu et al., 2020, p. 495); (2) Ouriakos is most probably to be interpreted as a complex, repeatedly settled, specialised site, where specific activities were carried out, more precisely weapon production and maintenance (O'Connell et al., 1991, p. 69). It consists of a series of partly overlapping knapping floors, distributed over a large area, resulting from the manufacture of lunate on microbladelet blanks, that are formed in a relatively short time , whose interpretation is made complex by the "rapid sequencing of events which characterizes the daily lives of living peoples" (Binford, 1981, p. 197). Owing to its uniqueness, extent, ideal sedimentary condition, sealed as it is between two dunes, and the presence of a buried soil (Efstratiou et al., 2013, Fig . 3), Ouriakos represents a unique site to interpret some aspects of the Epipalaeolithic archaeology of the region.
If we compare the Ouriakos assemblages with those collected from Agia Marina and Peristereònas, we notice a few small, but important differences. They concern mainly the typological characteristics of some artefacts, although all the sites are probably to be assigned to the YD event.
Regarding Agia Marina Site 1, the differences are mainly in the absence of bipolar-retouched lunates. In contrast, the site yielded 1 thin lunate obtained by abrupt, direct retouch (Fig. 10, n. 7), and 3 different microlithic abruptly retouched points, only 1 of which is bipolar (Fig. 7, n. 4), and 1 microlithic backed bladelet and truncation (Fig. 7, n. 3). The Peristereònas assemblage is more closely comparable with that of Ouriakos. The size of the lunates from Peristereònas falls into the same dimensional range as those from Ouriakos (Fig. 20). The same can be said of the dimension of the complete blanks (see Fig. 6). Regarding the other Agia Marina sites, their attribution to the YD is based mainly on the presence of bipolar-retouched lunates (Sites 2 and 3) and cores whose technology and shape can be compared with those from Agia Marina Site 1 and Ouriakos (Sites 2, 3, 5, and 6). The only artefact that does not find parallels elsewhere in the region is the straight perforator from Site 4 (Fig. 8, n. 11), whose chrono-cultural attribution remains uncertain.

Conclusions
At present, our knowledge of the Epipalaeolithic period in the Anatolian Peninsula is based mainly on the results obtained from the excavations carried out in the caves of Öküzini (Albrecht et al., 1992;López Bayón et al., 2002), Karain B (Albrecht, 1988, Kızılin (Erbil et al., 2021), and the Beldibi and Beldaşi rock shelters (Bostanci, 1968) in the Gulf of Antalya, Pınarbaşı, in the province of Konya (Baird et al., 2013), Girmeler Cave along the coast west of Antalya (Erdoğu et al., 2021), Esek Deresi Cave in the homonymous valley west of Mersin (Altınbilek-Algül et al., 2021), and Direkli Cave in the province of Kahramanmaraş (Erek, 2010).
All these caves and shelters yielded Epipalaeolithic assemblages characterised by microlithic tools obtained from bladelet blanks detached from subconical and prismatic microcores. They are marked by the presence of abruptly retouched lunates, other types of geometric microliths, and short end scrapers in the case of Girmeler (Erdoğu et al., 2021, Fig. 9). The Direkli Cave horizon with lunates obtained by abrupt, direct retouch, which the authors attribute to the Early Natufian (Baysal & Erek, 2018, p. 523), has been radiocarbon dated to 10,460 ± 179 BP; 12,471-11,712 cal BP (94.9%) (Beta-276742) from a sample of unidentified charcoals (Arbuckle & Erek, 2012, p. 695). A comparable date on bone has been obtained from the Epipalaeolithic layer of the Esek Deresi Cave (10,771-10,361 cal BC (95.4%): Altınbilek-Algül et al., 2021, p. 143). From this cave, the authors report three microlithic lunates obtained by abrupt, direct retouch, microlithic end scrapers and microcores. It is important to note that the two reported radiocarbon results can be compared with that from Ouriakos (GrA-53229). Moreover, they all fall into the time span of the YD cold event. The discoveries and the new radiocarbon dates suggest that the spread of a new type of microlithic assemblage characterised by the presence of lunates and other geometric microliths took place during this period (see Bar-Yosef, 2002, p. 377).
Geometric armatures have been recovered from the sequence of the Öküzini Cave, the most important Epipalaeolithic multi-stratified site excavated along the southern coast of Anatolia (Yalçinkaya et al., 2002). Their number increases in the upper sedimentary units (III-IV) where they are considered a kind of "fossil director" (Kartal, 2002, p. 239 Table 1). However, the knapped stone assemblages from the two YD horizons are very different from those known from the Lemnos Island sites, in general, and Ouriakos in particular. The differences consist in the larger size of the artefacts, the presence of many types of geometric microliths (lunates, scalene and isosceles triangles, and double truncations) and short end scrapers, and the use of the ordinary and abrupt microburin technique (Léotard & López Bay, 2002, pp. 176-182).
A similar trend has been noticed in the neighbouring Kızılin Cave, where the undated upper layers of the sequence show the predominance of microlithic lunates (Erbil et al., 2021, p. 162). However, the technological characteristics, and the retouch methods employed for making this type of microlith, have never been described in detail for the Anatolian assemblages, with the exception of those from the Öküzini Cave (Kartal, 2002). In effect, if we compare the data published from the Anatolian caves and rock shelters with those from the Lemnos assemblages, we can see that there are always important differences.
As reported above, the lunates from Ouriakos have been obtained by abrupt, bipolar retouch, without microburin technique. In particular, the bipolar technique can be considered a characteristic element of this assemblage. There is no doubt that this method of retouch has not been conditioned by the quality of the raw material employed for making artefacts because it was systematically employed for retouching microlithic lunates made from both hydrothermal siliceous rock and radiolarian chert. For this reason, we think that their retouch characteristics, and the unusual production technique, are to be considered a cultural trait, which so far does not find parallels in other Epipalaeolithic sites known along the Mediterranean coast of Anatolia and in the Aegean.
The Agia Marina Site 1 displays a few unique types and the use of the abrupt, deep, direct retouch. From these observations, considering the chronology of the Anatolian Epipalaeolithic sequences, and the characteristics of the assemblages, we suggest that the occupation of Agia Marina Site 1 took place slightly earlier than that of Ouriakos, perhaps during the same YD event, though this suggestion cannot be confirmed based on our present data.
To conclude, the discovery of new Epipalaeolithic sites with microlithic lunates in the island of Lemnos and Kocaman, in the Karaburun Peninsula (Fig. 1, blue dot), along the Aegean coast of Anatolia (Çilingiroğlu et al., 2018, Fig. 4), supports the impression of a westward spread of a cultural tradition whose origins are most probably to be sought either in the Levant or northern Mesopotamia (Golovanova et al., 2022;Nishiaki et al., 2011). This observation is reinforced by the discovery of a few new sites along the Mediterranean coast of Anatolia, which are characterised by the presence of microlithic lunates and other types of microlithic geometrics obtained by abrupt retouch. However, the assemblages analysed from the Epipalaeolithic sites discovered on the island of Lemnos partly contradict this suggestion.
Although their characteristics are most probably to be attributed to a Levantine tradition (Bar-Yosef, 2002, p. 377), their unique technological methods of production; the systematic use of the abrupt, bipolar retouch; and also their shape, which is due to the detachment of microbladelet blanks with a scalene, triangular crosssection, distinguish the Lemnos Epipalaeolithic assemblages, which so far find no comparison in the complexes of the same age recovered from the Levantine Mediterranean, Aegean, and Black Sea sites (Bibikov et al., 1994).

Supplementary Information
The online version contains supplementary material available at https:// doi. org/ 10. 1007/ s41982-022-00118-8. continued financial support. The authors want to thank Georgia Tampakopoulou for information regarding the location of some of the raw material outcrops, and Renato Nisbet for identification of the Centaurea spinosa bushes. Special thanks are due to Professor Clive Bonsall of Edinburgh University for the revision of the original English text.
Author Contribution All the authors contributed to the study's conception and design. Material preparation, data collection, and analysis were performed by Professors Paolo Biagi and Elisabetta Starnini. Fieldwork and artefact collections were made by all the authors. All the authors commented on the previous version of the manuscript. All the authors read and approved the final manuscript.
Funding Open access funding provided by Università di Pisa within the CRUI-CARE Agreement.

Data Availability
The dataset generated during and/or analysed during the current study is available from the corresponding author on reasonable request.
Code Availability Not applicable.

Competing Interests
The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.