Multianalytical approach to the exceptional Late Roman shipwreck of Ses Fontanelles (Mallorca, Balearic Islands)

The paper presents the results of an interim analytical approach to the Late Roman shipwreck of Ses Fontanelles recently found in the island of Mallorca (Balearic Islands, Spain). The excellent state of preservation of the hull and the cargo, including amphorae with painted inscriptions (tituli picti), and its location in shallow waters offshore of one of the main touristic beaches of the island makes this a unique �nding in the Mediterranean. A �rst season of excavations and study of the cargo triggered an analytical approach to solve some of the problems pose by archaeological research, mainly related to the possible origin of the vessel. The analytical strategy combines petrographic analysis for the study of the provenance of the amphorae, archaeozoology and residue analysis to identify their content, and analysis of the wood and plant remains to understand the use of vegetal resources in shipbuilding and in the stow of the cargo. The results of the combination of the petrographic analysis, the study of the ichthyofauna and the organic residue analysis suggest that the boat probably departed from the area of Cartagena in the southeastern part of the Iberian Peninsula, carrying a cargo of �sh sauce (liquaminis �os), oil, and probably olives preserved in wine derivatives, transported in three main types of amphorae. The analysis of the wood shows, as known in Roman shipbuilding, a clear selection of forest resources. The shipbuilders used pine for longitudinal parts of the hull, while for the small pieces related to the assemblage system (pegged, mortise and tenons) and subjected to a great stress they selected harder woods mainly Cupressaceae, Olea europaea, and Laurus nobilis. In addition, the study reveals that mainly branches of Vitis vinifera, but also other herbaceous plants were used as dunnage protecting the cargo during the journey. The results help to shed some light into different aspects of this unique vessel sunk in Mallorcan waters and contributes to show the bene�t of applying archaeological sciences in maritime archaeology.


Introduction
The Balearic Islands lay not far from the eastern coast of the Iberian Peninsula placed in a strategic position within the navigation and trade routes of the western Mediterranean in Antiquity (e.g., Mascaró 1971; Ruiz de Arbulo 1990; Izquierdo 1996; Moreno 2005; Guerrero 2006).The islands as steppingstones in these navigation routes were points of scale.Often, the boats experienced problems and sunk.This was the case of the shipwreck known as Ses Fontanelles.It was accidentally found in 2019 in one of the main touristic beaches not far from the city of Palma, the current capital of Mallorca, the largest island of the archipelago (Fig. 1).This prompted the authorities to undertake a rescue excavation in summer 2019 (Munar et al. 2021).Soon they realized of the exceptional state of preservation of the cargo and of the hull of the vessel.After a rst evaluation and the excavation of a section of the shipwreck, that corresponding to the bow, the rest of the remains were left in situ and covered again.After the rescue excavation, an interinstitutional project (ARQUEOMALLORNAUTA) to study the materials already recovered, and to ensure the continuity of the excavation and the full study of the shipwreck.
The ship transported a relevant cargo of products packaged in amphorae around the 4th century of our era, a time when commercial activities in the Western Mediterranean are still little known.This, together with the preservation of an impressive quantity of inscriptions depicted on the amphorae (tituli picti), makes this site a unique example in the Mediterranean and an excellent case for the study of commercial dynamics in Late Antiquity.
A rst campaign of study of the materials recovered in the 2019 excavation helped to classify the ceramic ndings and opened some major questions.To solve some of the main problems raised by the wreck and its cargo, an interim multi-proxy analytical targeted approach was designed based in a combination of methods and analytical techniques.The main aim was to use analytical techniques to help de ning the cargo, and the vegetal resources that could help to better understand the origin of this ancient vessel sunk in Mallorcan waters in Late Antiquity.To understand the provenance of the different types of amphorae a selection was subjected to petrographic analysis.To identify the products contained in the amphorae, some of them (including some that were still sealed) were micro-excavated and the content studied with an archaeozoological approach and organic residue analysis.Finally, a rst set of wood remains from the hull, plants found between the cargo, and bark used as stoppers to seal the amphorae were also studied to understand the use of vegetal resources and to shed light into the species used in shipbuilding.
Overall, this contribution presents the rst multianalytical results obtained from this important shipwreck and aims to contribute to the bene cial relationship between inorganic and organic analysis, and between maritime archaeology and archaeological sciences.Certainly, over the last decades the application of archaeological sciences to the study of shipwrecks have largely contributed to maritime archaeology worldwide for a wide range of chronologies and geographical areas.Apart from the studies of conservation science, it is worth mentioning the advance in the use of geophysics and in general remote sensing techniques in the detection and monitoring of shipwrecks with spectacular results (e.g., Caratori et al. 2006 Ward and Ballard, 2004).Archaeobotany in general has been essential for the study of the wood used in shipbuilding and providing important information on provenance and chronology (e.g., Colombini et al. 2003Colombini et al. , 2007 Traoré et al. 2016;2018) with important contributions from radiocarbon dating (e.g., Lorentzen et al. 2014).These approaches have been combined, of course, with the analytical study of the myriad of materials that formed part of the cargoes or of the crew's belongings on board, notably including inorganic materials such ceramics, glass, metals, stones, but also different organic materials and other substances (Colombini et al. 2003;Muller 2004;Lempiäinen et al. 2022).Also, archaeozoological studies have been contributing enormously in the study of shipwrecks, including full studies of the faunal remains oriented to dietary reconstruction (e.g., Hamilton-Dyer 1995), the identi cation of the content of the cargoes (e.g., Auriemma 2000; Barkai et al. 2013), or to singular applications like, for instance, the study of a mandibula of a mice found in the Late Bronze Age Uluburun shipwreck that allow to determine that Ugarit was probably one of the steps of the boat in its last journey (Cucchi 2008).All these studies and many others are excellent examples of the bene ts of the application of archaeological sciences in maritime archaeology.
Our interim multianalytical approach to the shipwreck found in Mallorca aims to modestly contribute combining organic and inorganic analysis and generalizing the analytical studies, even in cases when the resources, and therefore the number of samples, are limited.This is just a rst targeted step to solve archaeological questions waiting for a full excavation and a more developed strategy of application of archaeological sciences to this exceptional shipwreck.

The shipwreck and its cargo
The shipwreck lays just 65 m off the coast and just two meters depth in a very busy area of the coast, covering an area of approximately 12 m long by 5 m wide.The excellent state of preservation helps to reconstruct the composition and disposition of the cargo (Fig. 2).The hull was divided into horizontal sections by vertical bulkheads, which were used to organize the cargo inside the hold.It is also important to emphasize that the hull is very well preserved from bow to stern (approximate height of 1.5 m).The interim examination of the shipbuilding con rmed that the vessel belongs to shell rst with mortise and tenons assemblage tradition, being a typical example of western Roman imperial naval architectural type (Pomey et al. 2012;Munar et al. 2021).
The typological study of the amphorae recovered in the rst season of excavations has allowed to classify 249 pieces with diagnostic attributes that t into four main types (or groups of types) (Fig. 3) (Bernal et al. 2020).From a typological point of view, the Almagro 51c type amphorae are especially abundant, many of them exhibiting painted inscriptions (tituli picti) formed by a line with the names Ausonius et Alumni, and a second line allusive to the content: Liq Flos, restorable as Liquaminis Flos (liquamen ower), a liquid sh sauce appreciated in the Roman period (Soler et al., 2021) (Fig. 2C, and Fig. 3).The shipwreck is a unique example since is one of the few cases around the Mediterranean where, thanks to an exceptional state of preservation, is possible to relate the information of the painted inscriptions with the paleocontent.The second group identi ed is represented by a new type of amphora that we have named Ses Fontanelles I, but that could be probably considered as an imitation of the oil amphorae type Dressel 23, commonly manufactured in the Baetica (southern Spain).A third group of containers are the at-bottomed amphorae related to containers manufactured in the 3rd and 4th centuries, especially in the eastern part of Baetica, where some workshops have been excavated (Bernal-Casasola 2019).These containers have been traditionally considered as for transporting wine based in their typology and in their prototype the "Gaulish" amphorae produced in the Narbonensis.However, in our case some of these amphorae contained solid residues including what it seems olives pits.The amphorae repertoire is completed by type Keay XIX, but so far only two examples have been uncovered.This might indicate that this type maybe was more related to the crew than to the commercial cargo.The cargo was organized at least in two tiers that covered the whole length and width of the ship.The amphorae were vertically arranged to maximize the limited space available (Munar et al. 2021).Ses Fontanelles I type amphorae, which were larger and heavier, occupied the bottom tier of the center and the stern area (sectors 2 and 3).Smaller containers were mostly found near the bow (Sector 1) and the sides (Fig. 2).The amphorae were stooped with numerous quantities of what in the excavation seemed branches of vine shots, but also other plants, as dunnage to protect the cargo during the journey.

Petrographic characterization of ceramics
All amphorae and sherds recovered in the rst archaeological campaign were macroscopically examined.Of these materials, 25 complete or nearly complete amphorae were further examined with the help of a Dino-Lite digital microscope.In addition, ten selected samples were examined in the laboratory using an Olympus SZH stereoscopic microscope equipped with a continuous zoom between 7.5X and 64X, with 1X and 2X oculars, working between 7.5 and 128 magni cations.Systematic photographs were taken at 10X and 30X in both polished and fresh fractured sections.
The ten selected samples were also analyzed by Optical Microscopy (MO) using thin sections, for petrographic and mineralogical characterization (Table 1).Preparations were made from a fragment of ceramic included in resin, downgraded with a Struers Discoplan TS cutter-recti er, and manually polished using a powder abrasive to a thickness of 30µm, in which the quartz shows a grey-white rst order interference color.The fabric observations were made with an Olympus BX41 microscope, equipped with 2x, 4x, 10x and 20x lenses and 10x eyepieces, working between 20 and 200 magni cations.The petrographic descriptions of the fabrics were made following the system proposed by Whitbread (1989Whitbread ( , 1995) ) and Quinn (2013).Grain sizes for inclusions in petrographic descriptions were de ned using the American Geophysical Union (Udden-Wentworth scale) system.The following abbreviations are used in the description of the inclusions of each fabric or petrographic group: eq = equidimensional; el = elongated; a = angular; sa = subangular; r = rounded; sr = subrounded.All photographs, both in binocular and under the microscope, were taken with a computer-controlled Olympus DP70 digital camera.3.2.Archaeozoological approach: the study of the ichthyofauna in amphorae Some of the sealed amphorae were selected to sample their contents.The materials recovered from the microexcavation of these amphorae were wet sieved through a series of sieves of progressively smaller mesh size (i.e., 4 mm, 2 mm, 500 µm, 250 µm, 125 µm, 63 µm) and then air-dried.The only exception was the broken amphora DSF-231 that retained several bones embedded in the inner walls and its toe (Fig. 8.3).Most of the identi able sh bones were sorted through 500 µm mesh sieve, using a Dino-Lite digital microscope for its triage.Four of the six samples from Almagro 51c amphorae provided ichthyofauna remains: DSF-220, DSF-231, DSF-235 and DSF-257.The two samples from at-bottomed amphorae (DSF-266 and DSF-272) were sterile in terms of sh remains (Table 2).To estimate the body size (total length -LT) of the shes we followed the osteometric procedure indicated by Assis and Amaro (2006) for sardines.For the anchovies we built our own model from an extant reference population (n = 9) based on the length of the dentary (LD, Fig. 8.2 A) as this was the predominant element.
The following linear regression formula was used: LT = 8.95*LD-0.18(r 2 0.882).Only left-sided dentaries were used to avoid the repetition of individuals.The measurements were done with a Dino-Lite Edge digital microscope and using DinoCapture 2.0 software.).For a rst approach to the content, ve samples were selected: one from an Almagro 51c amphora (DSF220) for which the tituli picti informed of the content being liquaminis os (a liquid sh sauce), one from Ses Fontanelles I type (DSF002), and three from at-bottomed amphorae (DSF267, DSF274, and DSF272) (Table 3).
Each ceramic sample was cleaned mechanically, and then it was sub-sampled in the laboratory.The specimens were ground, and the powders were extracted by using three different extraction methods: All the extracts were trimethylsilylated with 30µl of N,O-bis(trimethylsilyl)tri uoroacetamide (BSTFA, Sigma-Aldrich) heating at 70°C.75 µl of hexane were added and the extracts analysed with gas chromatography coupled with mass spectrometry (GC-MS) using a chromatographer Thermo Scienti c TS GC ultra, with a silica capillary column and a mass spectrometer Thermo Scienti c ITQ 900 operated in electronic ionization (70 eV).The mass range is m/z 40-650.A blank sample was analysed using the same solvents employed for the analysis of the archaeological samples.

Analysis of wood and plant remains
After the rst excavation season and awaiting the full excavation of the shipwreck for an extensive sampling, 20 interim samples of wooden elements from various components of the hull were selected.The rest of samples up to a total of 101 specimens (Table 4), were selected from branches and herbaceous plants used as packaging placed between the amphorae to prevent their movement and breakage during the journey, and from bark used to seal the amphorae.A pinecone found within the cargo was also included.
The study of the cell structure of the wood was based on the observation of its three anatomical planes (transverse, longitudinal radial and longitudinal transverse) using a re ected light microscope Olympus BX60, with 50x, 100x, 200x and 500x magni cations.In addition, a MEB HITACHI S-3400N scanning electron microscope (MEB) has been used to obtain detail images at further magni cations.The taxonomic identi cation of wood remains is based on comparative analysis of the cellular anatomy of the woods of the different species.Wood anatomy atlases of European woods (Schweingruber 1990) and the reference collection of wood and coal of the ArqueoUIB Research Group (Universitat de les Illes Balears) has been used as comparative materials.The degree of accuracy of the taxonomic identi cation has varied in each case depending on the conservation of the wood and/or the di culty of obtaining thin sections from the different anatomical planes.
Several bunches of herbaceous plants were also analyzed.In this case, as these are non-woody plant remains, their taxonomic identi cation is not based on the analysis of the type of cells and cell structures typical of woody tissue.A rst approach to these materials has been carried out to characterize them both macroscopically and microscopically to provide a rst insight, awaiting further analyses.A test was carried out with one of these samples ("T.A. 2").The contents of the sample were dissed in water to identify the various elements (Fig. 9).Two of the herbaceous plant elements present were selected to be observed in the optical microscope with transmitted light without any treatment or preparation (stains, etc.).Different elements of his anatomy could be observed in this way.

Petrographic characterization: approaching the provenance of the amphorae
From a macroscopic point of view, although ceramics certainly share similarities, especially in the abundant micaceous content, the fabrics are not entirely homogeneous, with a diversity of colors and textures that could perhaps indicate a multiplicity of products.Thin-section petrographic analysis on the selected amphorae enabled the identi cation of two fabric groups.A full petrographic description of these groups is available in the Supplementary materials, and the discussion here is orientated towards the provenance issue.
Petrographic Group 1 (Fig. 4) includes almost all the samples analyzed (nine out of 10), including the ve samples of Almagro 51c amphorae, the two at-bottomed amphorae, and the two amphorae of the type Ses Fontanelles I (Table 1).The strong similarities in fabric between these nine samples provides evidence of a provenance, most likely, in the same area for all of them.The percentage of coarse inclusions is variable, although never very high, and, in any case, there is a continuous variability between samples with very little coarse fraction and others where it is more common.Similarly, the relative frequencies of the different types of inclusions vary between individuals, although again gradually and always repeating the same range of main inclusions, especially micas (mainly muscovite), monocrystalline and polycrystalline quartz/quartzite, and fragments of metamorphic rocks likely derived from mica-schist, along with other components in minor amounts (Fig. 4).
The typology of the amphorae -especially in the case of Almagro 51c-points to a provenance either in southern Spain or in Portugal.Comparison with regional geology helps to suggest various possible provenance areas in which there is an On the other hand, in the case of Petrographic Group 2 (Fig. 5), which includes only one of the samples analyzed (DSF277), corresponding to a Keay XIX amphora, the fabric observed in thin section indicates a provenance in an area with a dominant In any case, it should also be noted that the comparison with reference materials and the geology of other production areas for Keay XIX amphorae -like the cases of western and southern Lusitania, as well as the Baetican coast between Huelva and the bay of Algeciras, and the coast of Granada-suggests that none of them could be proposed as provenance area for sample DSF277.Something similar happens in El Mojón wokshop in Mazarrón Bay, in the Cartagena area, where amphorae like Keay XIX have also been attested (Berrocal 2007(Berrocal , 2012)); in this case, the fabric of sample DSF277 could be compatible from a strictly geological point of view, considering the presence of outcrops of phyllites and quartzites a few kilometers from the site (Espinosa et al. 1974).However, the products of this and other well-known workshops in the area are very different in terms of fabric both macroscopically and petrographically from the amphora DSF277.In summary, the integration of the existing archaeological information on the manufacture of Keay XIX amphorae together with the analytical evidence obtained from this study (and in combination with previous studies), suggest as a most likely hypothesis a provenance around the Málaga coast for the amphora DSF277 recovered at Ses Fontanelles shipwreck.In any case, both the lack of known workshops producing the morphological variant of Keay XIX to which this amphora is related to, and the differences mentioned above with respect to the most typical petrographic fabrics in Málaga amphorae, do not allow us for the moment to provide further details concerning its provenance.
4.2.The study of the ichthyofauna: which sh sauce?
All faunal remains recovered from the amphorae containing sh sauce reveals that the substance was derived almost exclusively from a single species: the European anchovy (Engraulis encrasicolus).The only exception was a hiomandibular bone of sardine (Sardine pilchardus) identi ed in the sample DSF-231 (Fig. 6).A selection of anatomical elements was used to estimate the minimum number of individuals (NMI) of each sample (Table 5).The anatomical representation of the anchovy is quite similar in all samples.The dentaries and other elements of the cranial region are predominant, with the rest of the appendicular skeleton appearing underrepresented, to a greater or lesser extent.Considering the dentaries as a reference for the minimum number of individuals (NMI), the percentage of representation of the vertebrae has been calculated by dividing the number of vertebrae observed by the expected number of vertebrae (e.g., the anchovy has 46-47 vertebrae, if 65 are observed, the estimated NMI will be equal to 2).The average percentage is 19%, ranging from a maximum of 33.3% (DSF-235) to a minimum of 7.1% (DSF-220).On the other hand, the number of vertebrae appears to be correlated to the estimated number of scales, since in those samples where there are with fewer vertebrae there are also fewer scales.Although it is tempting to suggest that the content of the Almagro 51c amphorae was composed almost exclusively of anchovies' heads, the embedded content in the resin of the inner walls of the amphora DSF-231 showed various articulated skeletons of anchovy (Fig. 7D, and Table 5), with their ribs and vertebrae still in anatomical position.The cranial parts of these specimens were recovered at the lower wall and toe of the amphora probably because of the detachment of the heads during the micronization process.For this reason, we suggest that, despite the observed partial anatomical representation in most of the samples, the sh sauce was made of whole anchovies.
The sardine found in sample DSF-231 had a total length of 11.4 cm based on the dorsal length of the hiomandibular bone (CRD, Fig. 8) and using the following linear regression: LT = 32.89*CRD+ 8.43 (r 2 0.931); where LT was the total length and r 2 was the coe cient of determination (Assis and Amaro, 2006).In the case of the anchovies, the size range is between 7 and 12.5 cm (n = 36; x = 8.7; CV = 13.5),although the vast majority were individuals with sizes between 8 and 10 cm.Although there are no signi cant differences in the size of the individuals present in the four samples, the occurrence of some larger shes is observed in the DSF-235 sample (Fig. 8, small box).

Organic residue analysis: further investigating the content of the amphorae Almagro 51c
The amphorae DSF220 (sample 4), Almagro 51c type, displays a painted inscription (titulus pictus) Alunnii et Ausonii NN() liq(uaminis) f < l(o)s informing of its content liquaminis os ( ower of liquamen, a sh sauce).Extract i displays abundant dehydroabietic acid followed by abietic acid, and methylabietate.Extracts ii and iii display very abundant dehydroabietic acid, abietic acid, methyldehydroabietate, 7-oxodehydroabietic acid, and retene is also present.Ses Fontanelles I amphora type Amphorae DSF002S Fontanelles I type (sample 1), display abundant residues.The oleic acid (C 18:1 ) is the most abundant fatty acid in extracts i and ii, followed by palmitic acid (C 16:0 ).In extracts i and iii there are also short chain fatty acids, being the C 9:0 the most abundant and dicarboxylic acids, among which azelaic acid is the most abundant.β-sitosterol is present only in traces in extract i.In general, the pro le of the three extracts suggests a plant oil content for the amphora, compatible with olive oil.However, the analyses performed do not allow for the identi cation of TAGs, therefore we cannot fully con rm the presence of olive oil.Sample 1 is the only sample, among the ve analyzed, in which tartaric acid is not identi ed.This datum reinforces the idea that this type of amphora was carrying a plant oil.Differently from the other analyzed amphorae, where resin/pitch biomarkers are abundant, dehydroabietic acid is present only in traces and there are no markers of the production process of the resin/pitch (i.e., heating or extracting it directly from the wood).
The at-bottomed amphorae Three samples were recovered and analyzed from at-bottomed amphorae.Olives were preserved in amphora DSF267.
The samples from amphora DSF267 (sample 2), DSF274 (sample 3) and DSF272 (sample 5), display abundant dehydroabietic acid followed by abietic acid.Methyldehydroabietate and 7-oxodehydroabietic acid are also present.These are markers of Pinaceae products (pitch) obtained burning the wood of Pinaceae trees.The abundance of abietic acid suggests that the pitch could derive from Abies sp.In sample DSF272 pitch is very abundant, even to the necked eye.
Possibly due to the abundance of Pinaceae products, or because we mechanically cleaned the samples in the laboratory (removing the pitch coating layer that possibly had absorbed most of the residues), it is di cult to identify other compounds in the lipid extracts.
Stearic acid (C 18:0 ), palmitic acid (C 16:0 ), and oleic acid (C 18:1 ) are present in extract ii of DSF267 (sample 2) and DSF274 (sample 3), being C 18:1 (oleic acid) the is relatively most abundant among them.However, none of these acids is abundant.βsitosterol is present in traces in extract i.These data are compatible with a plant oil.However, lipids are not abundant in the two amphorae.This is surprising mainly for DSF267, because olives were recovered in the bottom of the amphora.These data indicate that likely, if olives were preserved in a wine/vinegar medium (see later), they did not release abundant fats or that fatty acids possibly degraded and that possibly the abundance of pitch did not allow the abortion of residues in the pottery matrix.In DSF274, no solid residue was preserved, therefore we do not know if it contained olives.
Extracts i of the three amphorae displayed tartaric and succinic acids.Although tartaric acid is present also in other fruits (i.e., tamarind) (Barnard et al. 2011), it is possible to interpret it as a marker of grape derivatives for the context and period studied (Garnier and Valamoti 2016;McGovern et al. 2017;Pecci et al. 2013;Pecci et al. 2020).The presence of tartaric and succinic acids VEDERE MALICO, CITRICO con rms that the analysed amphorae also contained wine, vinegar, or OTHER derivatives, although it is not possible to differentiate among them.Extract i of samples DSF267 (sample 2), DSF274 also displayed syringic acid.This acid is related to the presence of red wine, however further analyses should be performed to con rm this hypothesis, as syringic acid could also derive from contamination.In general, the data could be compatible with the presence of olives in vinegar/other wine derivative, or with a rst wine content of the amphorae and a re-use of at least DSF267 for olives.
Sulfur is present in extracts i and iii of the three amphorae.

Wood and macroplant remains
Elements of the hull of the vessel The results of the taxonomic identi cation of the 20 wood elements analyzed are expressed in Table 5.In short, four different taxa have been identi ed, three corresponding to conifers, Cupressaceae (cypress and juniper family), Pinus (indeterminate pine) and Pinus t. pinea (Stone pine type); and two to angiosperms, Olea europaea (olive tree) and Laurus nobilis (laurel) (Table 5).
Among the larger elements that make up the hull of the boat only a single genus, Pinus sp, has been documented (Table 5).In two cases, it has been possible to differentiate that these corresponds to stone pine type (Pinus t. pinea) (Fig. 10).In two other cases only the genus, Pinus sp, has been determined.On the other hand, for the small parts of the hull three different taxonomic categories have been identi ed: Cupressaceae, Olea europaea, and Laurus nobilis (Table 5).In the case of fragments belonging to the family Cupressaceae (Fig. 10) anatomical elements have not been clearly observed to differentiate the genus Cupressus (cypress) and Juniperus (juniper).Similarly, in other cases it has not been possible to clearly differentiate between the two species of angiosperms identi ed, Olea europaea and Laurus nobilis (Fig. 10).The analysis of the wooden elements of the hull of the boat shows in any case a clear distinction in the woods used for large parts and for those of smaller sizes that formed part of the hull assembly system (pegged, mortise, and tenons).

Other plant remains
Several bunches of herbaceous plants interwoven with twigs, together with branches, were found between the amphorae as dunnage to protect the cargo during the journey.All branches were analyzed, together with some samples of the bunches of herbaceous plants and twigs and three fragments of wood of indeterminate shape and origin also found among the cargo (Table 6).Most of the branches (n = 61) were fragments of common grape vine (Vitis vinifera, Fig. 11) placed between the amphorae transported by the boat to prevent its movement and collision.One of the small branches was of a different species (DSF 40 sample), a plant in the Lamiaceae family.Two other herbaceous plants reveal the presence of paracytic stomata, in parallel to the axis of the leaf (Fig. 12), long cell structures and short cells could be observed, some of the long ones, with anticlinal walls profusely undulated (Fig. 11).The presence of tomentum was also observed (Fig. 12).This consisted of trichomes of various sizes, apparently single-celled, and especially associated with cells of the nerviations, where they concentrate very densely.Some images show structures that are possibly silica (phytoliths).Thus, the observed material has a diversity of wellpreserved cellular elements that could in principle be related to plants in the grass family (Poaceae), but a con rmation using phytolith analysis is already planned.
In addition, tree bark stoppers used to seal the mouths of the amphorae were analyzed.In all cases, the different elements of the cell structure of tree bark could be observed (Fig. 12).However, taxonomic identi cation beyond the determination of bark (cork) was not possible because of the taxonomic unde nition of this part of woody plants and by the conservation of the samples analyzed.However, due to the size of some of them, both in diameter and thickness, the cork pieces probably come from a tree species with a great development of this organ protecting the trunk.The most plausible hypothesis is that the cork oak (Quercus suber), a tree especially appreciated for its large cork production, was used to prepare the stoppers to seal the amphorae.
Finally, among the materials found on the boat a pinecone was identi ed according to the size and morphology of its components, mainly the peduncle (element that joins the pineapple to the branch) and the sheaths (Fig. 11).In this case, the morphological elements indicate that it is an Aleppo pine pinecone (Pinus halepensis).It presents a patent peduncle, elongated, and curved, typical of this species.The seminiferous scales have slightly convex apphysis.Finally, the measurements of this pinecone (7.1 x 4 cm) t quite well within the range of the pinecones of this pine species (6-12 x 3.5-4.5 cm) (Amaral Franco 1989).

Discussion
The multi-proxy approach to Ses Fontanelles shipwreck helps to a better archaeological interpretation of the boat and its cargo.
For the provenance of the amphorae cargo, the petrographic analysis provides additional and complementary evidence to the typological and macroscopic information.All amphorae of the Almagro 51c, Ses Fontanelles I, and at-bottomed amphorae are very similar to each other in terms of the petrographic fabric (Petrographic Group 1), indicating broadly the same provenance for all of them.The petrographic characteristics, in combination with the available archaeological evidence, point to the Cartagena area, in southeastern Iberian Peninsula, as a likely area of manufacture.On the other hand, for the only specimen of a variant of Keay XIX amphora a provenance in the current coast of Málaga may be proposed as the most likely hypothesis.Overall, it seems that the boat transported a cargo of products packaged in amphorae produced around Cartagena in southeastern Iberian Peninsula.
The archeozoological study reveals that the sh sauce transported in the Almagro 51c amphorae labeled as liquaminis os was prepared using primarily small-medium-sized anchovies (Engraulis encrasicolus), despite the occasional presence of sardines.The documented taxonomic selection and homogeneity of sizes between 7 and 12.5 cm is in line with that observed in other areas of the western Mediterranean (García-Vargas et al. 2018), highlighting that these small pelagic species were the subject of an intense and widespread shing for the preparation of salting products.From a historical and archaeological point of view, this practice can be contextualized in the renaissance of the production of sh salting during the Constantinian period with the generalization of new products based on small sh, such as sardine, wing, blotched picarel, or the same anchovy (Bernal-Casasola 2001; Bernal-Casasola et al. 2016, 2020; García-Vargas et al. 2018).At the same time, the samples studied here represents one of the few cases in the Balearic Islands where the sh remains were preserved in primary position (paleocontent in amphorae).
The results of the organic residue analysis provide further information about the content of the amphorae.The Almagro 51c amphora analyzed (DSF220), which displays a painted inscription (titulus pictus) informing of its content liquaminis os, displays residues that indicate the presence of animal fats, but cannot be surely attributed to sh at this stage of analysis.
The results are however interesting because they indicate abundant residues of resin or pitch used as a coating of the vessel and the presence of tartaric and succinic acids, indicating a wine/vinegar/defrutum content, which could have been used as a condiment to the sh sauce or the result of a re-use of the vessel.A combination of products was likely also preserved in the at-bottomed amphorae.In fact, in one of them, olives were preserved, and residues are compatible with a plant oil.It has been suggested that sterols can be indicative of the content of the amphorae coming from waterlogged contexts (Garnier et al. 2011).However, we also know that sterols could derive from post-depositional contamination (Hamman et al. 2018).In any case, in this regard, it is interesting that we could identify cholesterol in the Almagro 51c type sample, and β-sitosterol in the other samples, which agrees with animal products in the rst case and plant oil/olives in the second.In general, it is interesting that from the data obtained it is possible to suggest that substances that are identi ed as solid residues were stored in wine/vinegar/defrutum.An alternatively explanation could be the reuse of the amphorae, a practice that is more common than previously thought.If this would be the case, then wine/vinegar/defrutum or other wine derivatives would have been absorbed during a rst use of amphorae and then these would have been re-coated with pitch and used for a different content (olives, sh products).All these amphorae were coated with pitch, from Pinaceae, possibly from Abies sp., which was mostly visible also to the necked eye and could have been used as a waterproo ng and/or preserving agent for the content of the amphorae.In some cases, it is possible to identify the production process of pitch extraction, which was heated and/or extracted by burning the wood.Sulfur, which is also present in all of them, could derive from contamination or could have been mixed to the pitch to waterproof the amphorae and/or preserve the content.Finally, Ses Fontanelles I amphora type DSF 002 (sample 1), only displayed residues compatible with a plant oil content.Differently from the other amphorae, only traces of resin (and not abundant pitch), and no sulfur were identi ed in tis amphora.
The study of the wood used in the shipbuilding has identi ed ve different taxonomic categories: Cupressaceae, Olea europaea, Laurus nobilis, Pinus t. pinea, and Pinus sp.(Table 4).A selection of timber resources is clearly observed.For the construction of the large elements of the hull pine was used (Table 4), taking advantage of the wood of this tree that develops large trunks of straight growth, which, despite offering a wood of medium hardness and density, is especially optimal for the manufacture of these large elements which absorb water and expand transversely optimally for pegged, mortise, and tenon joints tightness'.In contrast, for the smaller elements of the hull assembly system (P, M & T) Cupresaceae, Laurus nobilis and Olea europaea woods were used (Table 4).The latter two species offer woods with a high density and ne grain, especially appreciated in cabinetry to make small elements and complex shapes.In the case of elements made of Cupresaceae wood, the genus or species used has not been more accurately determined.Considering the forest formations in which the identi ed taxa are developed, they could be Phoenician juniper (Juniperus phoenicea), a tree developing in coastal sand dune formation.However, we have not been able to be con rm this properly from the observation of the anatomy of the wood.In any case, the various species of the genus Juniperus also offer dense and ne grain woods specially indicated for the manufacture of this type of small objects.Furthermore, an association between taxon and object type is also documented considering the small elements of the assembly system (Table 4).For the manufacture of the pegs, only Cupresaceae wood has been documented, while only woods of Laurus nobilis and Olea europaea have been used for the tenons and treenails.The used of different woods is normal and has been widely attested in other ships similar chronology (e.g., Giacchi et al. 2003Giacchi et al. , 2017;;Beltrame and Gaddi 2006;Allevato et al. 2009).The identi ed species correspond to thermo-or meso-Mediterranean vegetation, without documenting, for example, taxa of broad-leaved formations (e.g., oaks, deciduous Quercus, or beech, Fagus sylvatica) or alpine (such as r, Abies alba, mountain pine, Pinus mugo, or Scots pine, Pinus sylvestris).Among the angiosperms identi ed, Olea europaea develops very extensively in all the low and middle lands of the circum-Mediterranean region, being in many cases the main tree component of open maquia formations or forming part of the undergrowth of other forest formations.Laurus nobilis (laurel) is also a taxon of clear development in the Mediterranean basin, although in this case the populations are more restricted to shady places and with greater humidity.As for conifers, pine woods have been identi ed, which in when type was possible to differentiate, they correspond to Stone pine type (Pinus t. pinea), so also discarding taxa typical of high mountain areas.Stone pine (Pinus pinea) is a species of lowlands and, although it grows in continental territories, develops signi cantly in coastal areas throughout the Mediterranean basin, also in sandy dunes.Among the 62 branches of woody plants analyzed used to secure the cargo (Table 3) most of them correspond to fragments of grape vine (Vitis vinifera), except for one corresponding to the Lamiaceae family.They were interwoven with herbaceous plants, and the analysis of few of them suggests that could be of the Poaceae family.They could be leaves and/or stems of common reed (Phragmites sp.), also a common coastal plant in the Mediterranean basin that was used in this case also to protect some of the amphorae during the journey.The use of vine shoots and other plants to secure the cargo was a common practice in ancient Mediterranean navigation, as witnessed in several shipwrecks such as Port Vendres (Liou, 1974), Cap del Vol (Spain) (Nieto and Foerster 1980), Madrague de Giens, Dramont E (Santamaria 1995), Gran Ribaud F (Long et al. 2006), Ma'agan Mikhael B (Israel) (Civkel 2020), or several of the shipwrecks found in Saintes-Maries-de-la-Mer (SM2, SM24, SM9) (e.g., Marlier 2018), just to cite a few examples.These plants were interpreted as for protecting the hull from the cargo but also to protect the cargo, as clearly demonstrated in the Madrague de Giens (Tchernia et al. 1978) and in Ses Fontanelles.Amphorae, sensible to friction and mechanical breakage, needed at least sometimes a proper a packaging.Finally, the pinecone analyzed has been identi ed as Pinus halepensis, a taxon that develops in large areas of the Mediterranean basin and it is also especially present in coastal areas.In the Iberian Peninsula, it is one of the main coastal forest formations south of the Llobregat River, which does not contradict the main hypothesis of the origin.The presence of pinecones in shipwrecks is not rare.Sometimes it has been interpreted as dunnage, as in the case of Madrague de Giens or more recently in the Ma'agan Mikhael B shipwreck in Israel (Cohen and Civkel 2018), but also as related to the sealing system of the amphorae.

Concluding remarks and future perspectives
Considering all the information obtained through the application of a multi-proxy approach to the exceptional shipwreck of Ses Fontanelles a few conclusions concerning its origin can be proposed.
All the analytical data suggest that Alumnius et Ausonius prepared a trade enterprise eeting a merchant ship with a cargo composed mainly of sh sauce (Liquaminis os), in Almagro 51c amphora, oil transported in Ses Fontanelles I amphorae (probably an imitation of Dressel 23 type), and vine derivates or fruits preserved in those substances in at-bottomed amphorae.The analysis of the ichthyofauna has contributed to understand that this sh sauce was basically prepared with small engraulidae particularly anchovies but with presence of sardine.It is possible that also a, so far, invisible cargo occupied part of the space in the galley (Munar et al., 2022).These products were carefully stowed in the hold of the merchant ship using vine shots and herbaceous plants as dunnage for protection.Petrographic analysis of the amphorae suggests an origin for the cargo in southeastern Iberian Peninsula, and particularly in Cartagena or its surroundings.The study of the wood of the elements of the hull and branches and herbaceous plants recovered suggest species of a temperate region, mainly in coastal areas and in lowlands ruling out the use of species of mountainous vegetation, compatible with the area of southeastern Iberian Peninsula.
On the light of this rst analytical approach, it is very likely that the vessel departed from some port of the area, perhaps from the main city of Carthago Nova or Carthago Spartaria (current Cartagena) in Late Antiquity.The boat probably departed from the city itself or a nearby port and sunk for reasons that we still do not know in the large bay of Palma in Mallorcan waters.

1 .
The alkaline extraction proposed byPecci et al. (2013bPecci et al. ( , 2020) ) was carried out on approximately 1g of sample to identify wine and fermentation biomarkers.Before extraction, 50 µl of internal standard (n-hexatriacontane) were added.2. The lipid extract was carried out following Chartres et al. 1993, modi ed by Pecci et al. (2013a).Before extraction, we added 20 µl of internal standard (n-hexatriacontane). 3. A further extraction with KOH and chloroform (Pecci et al. 2013a) was carried out on the powder already extracted by extraction ii.
metamorphic contribution, but different from the production area of Petrographic Group 1.The nature of the metamorphic rocks present in this fabric could be compatible with different areas with low to very low grade metamorphism associated with phyllite outcrops, as is the case -especiallyof the bay of Málaga, in relation to the outcrops of the Maláguide Complex(Aldaya et al. 1980; Fontboté 1983; Junta de Andalucía 1998; Martin-Algarra 2004; Serrano and Guerra 2004), and also from other more eastern areas of the southern part of Iberia related to the Alpujárride Complex, where phyllite outcrops can also be found, although in many cases along with other rocks related to a higher degree of metamorphism (Fontboté 1983; Junta de Andalucía 1998; Martin-Algarra 2004).Comparison with the petrographic reference collection at ERAAUB, as well as with comparative materials found in different publications, shows partial similarities with amphorae fabrics between the Punic and Late Roman periods in the Málaga region, including amphorae of the same type Keay XIX (Fantuzzi and Cau 2017; Fantuzzi et al. 2020).It should be noted that the Málaga coast was, based on the archaeological evidence known to date, the main production area of Baetican Keay XIX amphorae -in addition to the important production centers known in Lusitanian-with various documented workshops (Bernal-Casasola 1997, 2001; Serrano 2004; García Vargas and Bernal-Casasola 2008).There are, however, some differences with the most typical fabrics of coastal Málaga, especially the lower frequency of sedimentary rock fragments (argillites, sandstones, and limestone) in sample DSF277, as well as the absence of some inclusions (serpentinite and pyroxene) that are sometimes -but not always-found in accessory quantities in the products of Málaga due to the proximity of the ultra-ma c complex of the Ronda mountain range (Fantuzzi and Cau 2017; Fantuzzi et al. 2020).However, the characteristics of the metamorphic rocks found in the DSF277 fabric, as well as the other types of inclusions observed -e.g., the fragments of igneous rocks present in an accessory quantity, identical to those documented in other Málaga products (Fantuzzi et al. 2020)-suggest that the fabric of sample DSF277 could correspond to a different variant within the products of the Málaga coast.It is important, in this sense, to note that the amphora DSF277 corresponds to an atypical morphological variant of the Keay XIX type, different from that analyzed in previous petrographic studies (Fantuzzi and Cau 2017).For this formal variant we do not know possible workshops and there is no information concerning the chronological and geographical relationship with respect to the best-known variants of this type.The presence of a particular fabric could respond to various factors: a variant of Keay XIX produced in another pottery workshop in the same region?, use of different clay sources at different chronological times of Keay XIX production?

Figures
Figures

Figure 1 Situation
Figure 1

Figure 3 Main
Figure 3

Figure 5 Photomicrographs of the petrofabric Group 2 Figure 6
Figure 5

Figure 7 Fish
Figure 7

Figure 9 A
Figure 9

Table 2
Description of the samples for the archaeozoological study

Table 3
Details of the ceramic sherds analysed with Gas-Chromatrography.

Table 4
Synthesis of the wood and plant remains studied (Bernal-Casasola 1998, 2001;ntribution of metamorphic rocks, it is mainly through comparison with reference materials and previously known fabrics that it is possible to obtain a rst insight into the possible provenance based on the compositional compatibility.Thus, the fabrics observed in Petrographic Group 1 are clearly different from those de ned in previous studies for the Guadalquivir valley (e.g., Grubessi 1999; Grubessi and Conti 1999; Fantuzzi and Cau 2017, 2019), the coast of presentday Málaga province (e.g., Fantuzzi and Cau 2017; Bernal-Casasola et al. 2020; Fantuzzi et al. 2020), or the coast of Granada (e.g., Vigil de la Villa et al. 1998).Conversely, in the easternmost part of the southern Iberian Peninsula, particularly in the western zone of Cartagena, comparison with previously published studies reveals similarities between our Petrographic Group Berrocal 2012: 257), especially in the case of calcite, which might suggest the use of raw materials from various outcrops to produce the amphorae.For all the above, the relationship of the Petrographic Group 1 of Ses Fontanelles with the productions of El Mojón or, at least, of the Cartagena area is a very likely hypothesis, especially when comparing with the information previously published by other authors (Berrocal 2012).It is worth mentioning that the workshops in Cartagena (particularly El Mojón) and Granada (sites of Los Matagallares and Los Barreros, in Salobreña) are the only areas known to date in the Mediterranean coast of Spain with well-documented production of both Almagro 51c amphorae and at-bottomed amphorae(Bernal-Casasola 1998, 2001; Berrocal 2007 Berrocal  , 2012)); in the case of Málaga, production of Almagro 51c amphorae is well documented, but not of at-bottomed amphorae (Bernal-Casasola 1997; Serrano 2004).
important contribution of metamorphic materials, such as the Mediterranean coast of southern Spain (particularly from Málaga to Cartagena), or the Guadalquivir Valley(Aldaya et al. 1972(Aldaya et al.  , 1980;; Fontboté 1983; Junta de Andalucía 1998; Martín-Algarra 2004).Other areas such as the Atlantic coast of Baetica (e.g., Cadiz Bay), the bay of Algeciras, or southern and western Lusitania could instead be ruled out.In the latter area, corresponding to the lower valleys of the Sado and Tagus, although some rather micaceous fabrics of Late Roman amphorae have been reported (e.g.,Mayet et al.1996; Fantuzzi et al.   2015), the comparison with the samples from the wreck of Ses Fontanelles indicates that these are clearly different products, both in terms of clay matrix and texture and composition of the inclusions. 1 and the fabrics described for El Mojón workshop (Berrocal 2012: 256-257), where a ne fabric containing predominant inclusions of quartz, muscovite, and fragments of metamorphic rocks rich in quartz and muscovite, as well as iron oxides, has been recorded.It should be noted that we have compared the samples of Petrographic Group 1 with thin sections of four amphorae from the workshops of El Mojón and Mazarrón in Cartagena, and while clear similarities are observed in both the main types of inclusions and textural parameters, some differences are also noticed, especially in the relative frequency of inclusions.In this regard, we should also mention that the samples used as comparison elements are mostly spatheia, with fabrics possibly more re ned than in other amphora types.In addition, the frequencies of the different components in the amphorae from El Mojón are very variable (

Table 5
Results of the ichthyofauna analysis, number of remains of anchovies identi ed.The sample marked with * only counts for the remains recovered at the bottom of the toe of the amphora ( g. X), and those bones attached to the walls of the amphora container have not been considered.
(Garnier and Valamoti 2016;ae products, speci cally of pitch obtainPecci et al. , 2020))f Pinaceae trees(Colombini et al. 2005).Retene is the marker of heating.The abundance of abietic acid, which are quite uncommon in archaeological samples, suggests that the pitch could derive from Abies sp or alternatively that the residue is very well-preserved and giving a strong signal, not normally observed in archaeological cases.The abundance of Pinaceae products makes more di cut to identify other compounds.Nevertheless, extract i displays traces of tartaric and succinic acids, that, considering the context and period studied, indicate that the amphora contained wine or its derivatives(Garnier and Valamoti 2016; McGovern et al. 2017; Pecci et al. 2013Pecci et al. , 2020)).In extracts i, ii, and iii, palmitic and stearic acid are present, being stearic acid the most abundant.Cholesterol is also present in the sample, while β-sitosterol is absent.Although sterols could derive from post-depositional contamination (Hamman et al. 2018), this could also suggest an animal origin of the content of the amphora.In extract ii also C 20:0 and C 22:0 are present.These compounds are present in sh but are not exclusive markers of this product.The data indicate an animal origin of the content probably mixed with wine, vinegar, or other wine derivative.

Table 5
Result of taxonomic identi cation of recovered wood objects (excluding branches)

Table 6
Results of the taxonomical identi cation of branches and unclassi ed wooden fragments