Obsidian from the Bronze Age Village of San Vincenzo, Stromboli, Aeolian Islands: A Preliminary Investigation

: We present the preliminary results of the ongoing investigation of the obsidian from the Bronze Age village of San Vincenzo, Stromboli, Aeolian Islands, for the purpose of providing technological and typological characterization, and also provenance of the raw material, the latter with non-destructive p-XRF. Regarding provenance, the source of the raw material is likely to be neighbouring Lipari. It was transported to Stromboli and used mostly in a highly opportunistic manner and for the production of blade(let), non-bladelike tools (mainly scrapers) and micro bladelets. The obsidian distribution around the site shows concentration in both domestic and production areas.


Introduction
Stromboli is an active volcano and the north-easternmost island of the Aeolian Islands, an archipelago composed of seven islands in the southern Tyrrhenian Sea (Figure 1).The Archipelago has been inhabited since the 6 th millennium BCE, and Stromboli has an attested human occupation from the 4 th millennium BCE.
The Stromboli terrain is very steep (12.2 km 2 wide and with a volcano peak height at 926 m).There are only a few flat areas on the island, mostly on its northern coast, and San Vincenzo is the largest (a steepsided plateau about 6 ha wide resting between 40 and 100 m a.s.l.), located in the NE: a strategic position for the control of Southern Tyrrhenian sea but with no visual contact with the other Islands (Di Renzoni, Lopes, Martinelli, & Photos-Jones, 2016a).
The archaeological site of San Vincenzo was discovered in 1980 (Cavalier, 1981) and has been explored systematically since 2009 by the Stromboli Project: ArchEOLogIE1, an international interdisciplinary team with scholars and students from numerous institutions and countries2.After geophysical investigation (Zhao, Forte, Levi, Pipan, & Tian, 2015) a series of trenches have been opened along the southeast margin of the plateau and more than 700 m 2 have been so far investigated, divided in three areas: East, West and North.

Methods
The study of the San Vincenzo obsidian samples presented here involved several methodological approaches aimed at addressing questions of production/function, use/distribution and also to provide a first indication of provenance of the raw material (Calliari, 2011(Calliari, -2012)).
Provenance: non-destructive p-XRF analysis3 was carried out while on site in June 2012 of thirty nine archaeological and five surface obsidian samples, collected by one of us (DG) from the 'Spiaggia delle Pomici' (SdP) on NE Lipari.The only reference collection available at the time was from the Spiaggia delle Pomici: this obsidian is thought to have originated from the Mt.Pilato eruption dated to the 8 th c CE and thus postdating the Bronze Age village at San Vincenzo.In addition, six samples of obsidian from Melos collected by one of us (EPJ) from the two well-known prehistoric obsidian working sites on Melos, SW Cyclades (Nychia and Demenagaki) were also analysed (in Scotland) and presented here for purposes of reference.The instrument used is a NitonX3Lt-GOLDD with a 50kVAgX-raytube, 80 MHz real-time digital signal processing and two processors for computation and data storage respectively.Of the instrument's three calibration modes (TestAllGeo, Soils, Mining), the TestAllGeo (TAG) mode was selected since this gives the largest array of measurable elements.Analysis time was set at 60 seconds and three measurements were taken on each sample.Table 1 represents the average of these measurements.
Replicate analyses of NIST 2709a soil standard revealed satisfactory instrument precision (<10% Zr and <5% Rb, Fe, Ni and Sr), and good accuracy for Sr (<10%), but poor accuracy for Zr and Rb (larger than 10%).Y was not measurable in the TAG mode.We refer to the above elements in particular since obsidian provenance discourse is largely based on the plotting of the relative concentrations between these elements.Reporting ratios between elements further facilitates the comparison of data sets undertaken by different laboratories.
Although not directly relevant to the Stromboli finds, the Melos data set is introduced here, to confirm accordance with published work, outwith the Lipari non-prehistoric material.Further to our own three data sets (Stromboli archaeological finds, SdP surface finds and Melos (Glasgow data set)), we also discuss those of others pertaining a) to the two known Liparian sites (Gabellotto and Canneto Dentro by Tykot ( 2017)) and b) obsidian from both Lipari and Melos analysed by Acquafredda, Muntoni, & Pallara (2018).
Production and function: techno-economical characterization of 2508 pieces was carried out (1314 of which found in the Bronze Age Capo Graziano layers).The methodology involved the technological classification of the finds and the definition of the châine opératoire (Leroi-Gourhan, 1943;Laplace, 1968;Inizan, Reduron-Ballinger, Roche, & Tixier, 1999;Bar-Yosef & Van Peer, 2009;Arzarello, Fontana, & Peresani, 2012).Use and distribution: every find from the archaeological site was collected according to a 1x1m grid system, the topographical general framework of the excavation.The weight in grams of the different kinds of artifacts collected from a single Stratigraphic Unit in a single square was entered in a Geo-DataBase.Specific queries were applied to identify the quantities per square from the Bronze Age layers.The result was consequently displayed in a GIS environment assigning to every square of the grid a different color according to the quantity of all the obsidian items from the western area of the excavation (about 18x20, 350 m 2 ), expressed by weight.The distribution of the tools was tested for Trench 3 of the Western Area, measuring 14x8 m.The distribution of the obsidian is compared with other finds.

Provenance
p-XRF analyses are shown in Table 1 and Figs.2a-g and encompass the data deriving from: the Stromboli obsidian, the SdP obsidian, the Glasgow Melos obsidian, the Acquafredda et al. (2018) group and with regards to their own Lipari and Melos collections and the Tykot (2017) data set.We address particular issues as outlined below (the outlier #1654 is not represented in all the figures).

The Stromboli Collection and the SdP Obsidian
Figure 2a shows a plot of Fe/Sr vs Rb/Sr for two data sets: the Stromboli finds and the five SdP samples.We demonstrate that for these two ratios SdP samples fall within the Stromboli group.Similarly, figure 2b shows a plot of the ratios Rb/Sr vs Nb/Sr and again the SdP surface finds fall within the Stromboli group.We suggest that although the origin of these surface finds is not known, ie they may or may not be from the 8 th century CE eruption, there is nevertheless a good agreement with the Stromboli group and some obsidian from the latter must have originated from 'that' source.

The Gabellotto and Canneto Dentro Obsidian Sources and its Relevance to the Stromboli Collection
Figure 2c is taken from Tykot (2017, Figure 3) and shows the same element ratios as those shown in figure 2a.The Stromboli obsidian spans the totality of the Liparian group which is made up of the two sources (and perhaps more) since it has an Rb/Sr ratio between 5 and 15 as opposed to 5 and 20 shown by Tykot.The Stromboli Fe/Sr ratio falls between 400 and 600 which is still within the 'greater Lipari' group but not necessarily associated conclusively with either the Gabellotto or the Canneto Dentro sources.Instead the Stromboli collection ratios skim the edges of those for both localities.

Setting the Scene for the Lipari Obsidian: Lipari vs Melos
Figure 2d shows a plot of the Rb/Nb and Sr/Nb ratios for one additional group, i.e. the Glasgow Melos group.The Stromboli group clusters very tightly on the left hand corner of the plot and with a Rb/Nb ratio between 4 and 6 with one outlier at 9. The Melos group clusters on the upper right hand corner of the same plot and an Sr/Nb ratio which is between 12 and 16 and is very distinct from the Lipari/Stromboli groups.
Figure 2e shows the Tykot data for both Lipari and Melos but also other sites as well not relevant to this discussion.The ratio for Rb/Nb is between 6 and 10 (as opposed to our 4 and 6 excluding the outlier).For the Sr/Nb ratio for their own Melos data they shows a range between 7 and 15 which incorporates out data set (between 12 and 15).It should be mentioned here that our Melos data set for Rb/Nb is in better agreement with the Acquafredda et al. (2018) Melos data set (between 4 and 6) than with the Tykot (2017) data set.
Finally figure 2f sets our data set in the context of the results of the Acquafredda et al. ( 2018) group which combines their own Melos and Lipari samples and with respect to ratios Rb vs Nb/Sr.With regards to the Melos group theirs and our analyses are in agreement but the Stromboli group is set well apart from their Lipari group.
p-XRF results of the analysis of the Stromboli obsidian show that Lipari was a likely source for the raw material.There is a good agreement between the Spiaggia delle Pomici (SdP), surface samples and the Stromboli collection, for ratios of Fe/Sr, Rb/Sr and Nb/Sr.There is also good agreement with the data sets of other investigators (Acquafredda et al., 2018) and with regards to the Rb/Nb ratios (Figure 2e) but less also with the data sets of the same group and with regards to Nb/Sr (Figure 2g).We cannot ascribe the Stromboli finds to either the published Gabellotto or Canneto Dentro sources but perhaps another source within Lipari.

Production
The protohistoric assemblage, summarized in Table 2, is made up for 8% of blades l.s., for 18% of flakes, for 10% of unidentifiable débitage products, for 10% of cores, for 4% of tools that lack a distinct ventral surfaces and for 50% of debris (Figure 3).No refit has been found.
The lithic assemblage is characterized for: • the flakes are small-medium sized (only one is > cm 5x5) and were obtained through hard-hammer direct percussion.There are, though, also evidences of bipolar percussion on anvil; • the blades l.s.don't exceed cm 5 length (although few broken blades would have).There are not only very regular blades l.s. with linear or punctiform butts, that suggest standardized pressure débitage method, but also blade-like products, with irregular morphology and wide butts, that seem to suggest hardhammer opportunistic débitage; • the cores are mainly small sized and show the total depletion of the lithic volumes, mostly through opportunistic methods, mainly with the removal of bladelets and microbladelets, at least in the last moments of débitage; • natural cortex is attested and is found mostly on blades l.s., showing not only the importation of raw material blocks, but also how the reduction sequences were started; • the retouched products are the 21,8% of the lithic assemblage (excluding cores and débris).Retouch operations are mostly concentrate on small flakes, in order to create scrapers through simple and/or scaled removals.Amongst blade products, only few bladelets (~8,7%) show evidence of retouch, while the ~46,3% of blades s.s. is retouched.

Use and Distribution
The distribution of the obsidian in the Western area of the excavation (Figure 4A) clearly shows a significant concentration related to the Hut 3, especially near the hearth (Ferranti et al., 2015, Fig. 6;Cannavò et al., 2017a, Fig. 7a), and to the area upslope of hut 2, where a 'productive area' has been investigated (area W: Ferranti et al., 2015, Fig. 2, Cannavò et al., 2017a, Fig. 5) (Figure 5).Similar concentrations are also observed with the pottery (Levi, Bettelli, Cannavò, Di Renzoni, Ferranti, & Galliano, 2018, Fig. 4   The same pattern can be read when we consider the distribution of the tools only: the test performed in Trench 3 (located in the north-east portion of the Western area): the analysis shows a high numbers of items around the hearth of hut 2 (Levi, Bettelli, Cannavò, Di Renzoni, Ferranti, & Martinelli, 2012, Figs. 5, 6) and a significant concentration south of the same hut (Figure 4B).Summarizing, the obsidian finds are quite homogeneously widespread across the whole site, though there are 'qualitative' differences between dwelling and non-dwelling contexts.In the dwelling contexts retouched flakes are highly represented and they are associated with the few«big» flake cores found, while in non-dwelling contexts unretouched flakes are prevalent and the cores are mainly small-sized and exhausted; instead blades l.s. are widespread in every context, even if limited in number.This distribution pattern seems to point to the waste and/or residual nature of the lithic assemblages of the site, beside the ones found in the dwelling contexts.If we compare the obsidian assemblage in the San Vincenzo huts with the contemporary ones at Milazzo Viale dei Cipressi (Martinelli, 2006(Martinelli, , 2009)), despite the different preservations of some contexts (huts 2 and C at Stromboli are probably more disturbed), it is evident that there is a common prevalence of flake scrapers at  3) in the most preserved contexts: Hut 1 in San Vincenzo (Levi et al., 2011, Figs. 7c, d) and Hut 1 in Viale dei Cipressi (Levi, Tigano, Vanzetti, Alessandri, Barbaro, Cassetta, Castagna, Gatti, Sabatini, & Schiappelli, 2003;Levi, Prosdocimi, Tigano, & Vanzetti, 2009).

Discussion
The preliminary results of this study seem to show that the lithic industry of San Vincenzo site was focused on the maximum exploitation of the raw material arrived from Lipari, as suggested not only by the profusion of small-size exhausted cores that show highly opportunistic débitage practices, but also by the prominence of débris and by the lack of cores and débris of big dimensions (Figure 6).The analysis of the distribution, considering all the finds and the tools only, show significant concentration of the obsidian in domestic contexts such as huts 1, 2 and 3 but also in the productive area W. It seems that the depletion of the raw materials was mainly aimed, at least in the last moment of the reduction sequences, to the obtainment of blade-like products of small dimension and was carried out not only by hard hammer opportunistic débitage, but, as some finds seem to attest, also by more standardized bipolar reduction sequences.
Although both the products and the cores of blade débitages aren't numerically consistent, it is also possible to hypothesize that one of the principal aim of the obsidian exploitation at San Vincenzo site was the production of blades l.s.: this, not only considering the residual/waste nature of most of the lithic assemblage, associated with the presence of regular blade products that show the existence of standardized and (usually) highly-productive blade débitage methods, but also assuming the «exportation» in other contexts of the blades l.s.obtained through these methods.This hypothesis matches also with the fact that natural cortex is highly represented on the few blades s.s.found, therefore suggesting the existence of blade reduction sequences that started from unprepared obsidian blocks.Simultaneously to this blade productions, there were also reduction sequences which aim was the production of flakes, as attested in the dwelling contexts; however, since retouched flakes are found in the whole site, it is probable the utilization of the technical waste of the blade reduction sequences as blank for tools.

Figure 1 .
Figure 1.1.The Aeolian islands in the lower Tyrrhenian Sea. 2 The Aeolian Islands and the location of the Capo Graziano villages.3. The island of Stromboli with the San Vincenzo plateau in the Northeast (A.Di Renzoni).

Figure 2a .
Figure 2a.Fe/Sr vs Rb/Sr.Stromboli: red diamonds; SdP, Lipari reference obsidian: blue squares.The latter fall within the Stromboli collection group.The SdP obsidian does not constitute an obsidian source, merely presence thereof.

Figure 2b .
Figure 2b.Rb/Sr vs Nb/Sr.Stromboli: red diamonds; SdP, Lipari reference obsidian: blue squares.The latter fall within the Stromboli collection group.

Figure 2d .
Figure 2d.Rb/Nb vs Sr/Nb.Stromboli: red diamonds; Melos (Glasgow collection): black triangles.Stromboli archaeological material has an Rb/Nb which is tightely set between 4 and 6 (but with an outlier at 9) and low Sr/Nb; Melos, geological material, on the other hand has an Sr/Nb ratio which is between 12 and 16 and is very distinct from the Lipari/Stromboli groups.See also Fig 2e.

Figure 2e .
Figure 2e.Rb/Nb vs Sr/Nb from Stromboli and two published data set: Tykot (2017, compositional ranges from fig. 4a) and Acquafredda et al. 2018 (yellow circles).There is a good agreement between the Stromboli group (red diamonds) and the Acquafredda Lipari's data.Brought to you by | Universita degli Studi di Ferrara Authenticated Download Date | 4/29/19 11:01 AM ) A less significant Brought to you by | Universita degli Studi di Ferrara Authenticated Download Date | 4/29/19 11:01 AM

Figure 4 .
Figure 4. Obsidian distribution in the 1x1 m grid of the excavation (see Fig. 5): A. all the finds in the Western area; B. tools in the Trench 3 (north-east portion of the Western area) (V.Cannavò, A. Di Renzoni).

Table 1 .
Chemical compositions of the obsidian artefacts from San Vincenzo (SV) and reference material from Lipari and Melos, determined by p-XRF.Element contents are expressed in ppm.

Table 2 .
Obsidian tools at San Vincenzo according Laplace typology.

Table 3 .
Obsidian tools in Bronze Age huts at San Vincenzo and at Milazzo Viale dei Cipressi -MVC.