1The Bou Ferrer shipwreck is situated at a depth of approximately 25 m off the coast of La Vila Joiosa, Alicante, Spain (Cibecchini et al. 2006). The modern city corresponds to ancient Allon, situated in the southern maritime zone of the Cap de la Nao, a geographical feature that divided the sea routes heading to Ibiza in antiquity and the Tarraconensis in the Early Empire, with the influential areas of Tarraco in the north and Carthago Nova in the south. At the time of its discovery by two sports divers, after whom the underwater archaeological site is named (De Juan et al. 2008, p. 270), the wreck occupied an oval area of roughly 24 x 8 m on the seafloor. The relatively poor conditions for recreational diving in the area in which the Bou Ferrer is located have contributed to the excellent state of preservation of the shipwreck, which remained free from looting and largely intact (with the exception of some remains of more recent fishing equipment) until it was ultimately discovered.
2The wreck is unique among those so far discovered along the Spanish coast due to its size, accessibility, and state of preservation. From the beginning of the excavation (Cibecchini et al. 2006), we have suggested that the ship most likely did not sink while hugging a coastal route, which would have been unsafe and difficult for a large merchant vessel. Instead, we believe that the Bou Ferrer was driven from its route in the open sea by strong easterly winds while on a heading to the Strait of Bonifacio.
3Based on the results of amphorae studies to date, the Bou Ferrer was an Early Imperial merchant vessel. The ‘tumulus’ dimensions, verified by tests, have shown larger dimensions than other known wrecks of the period, as we shall present later. When the vessel sank, it was carrying a homogeneous cargo of Dressel 7‑11 amphorae, whose number is currently under discussion, and which F. Cibecchini has differentiated into four main subtypes with parallels from workshops like that of Villanueva de Puerto Real in Cádiz (García Vargas 2000; De Juan et al. 2008, p. 188–189). The majority of this main cargo seems to consist of Dressel 11-type amphorae, stowed in three layers with very likely a fourth (fig. 1), located amidships with each container packed in vine shoots used as dunnage. We have studied, thanks to the contact marks apparent on a large number of amphorae, how they were stowed in a quincunx pattern. According to this, we propose that four layers of amphorae would reach a height of about 2.10 m (De Juan 2018b, p. 137–141).
Fig. 1: View from the southeast of the slope of amphorae in the northern section of the central trench
Four levels of amphorae are clearly visible: the layers have been numbered.
(photograph C. De Juan)
- 1 A systematic analysis of the amphorae contents by Gaël Piques (CNRS-Lattes) is currently underway.
- 2 Study conducted by Christian Rico and Claude Domergue (University of Toulouse).
- 3 Roman pound of 327 g.
- 4 This matter is discussed in a dedicated paper presented by Ch. Rico and the authors to the Instrum (...)
4Regarding the content of the amphorae1, we found the remains of what was a highly refined, semi-liquid fish sauce, obtained through the filtering of a product that was made with a large variety of small fish and others of larger size (De Juan et al. 2008, p. 190). Fish sauces and their derivatives, such as garum, muria, liquamen and hallec, from the Roman province of Baetica were sought-after trade products, some of them of great value in Roman gastronomy. In addition, there is a complementary cargo of lead ingots with a triangular section2 stowed at the bottom of the ship along its axis (De Juan et al. 2014, De Juan 2018). The weight of each single ingot is close to 200 Roman pounds3 (65.5 kg). They have a rich collection of hammered epigraphic stamps (IMP AVG GER NER CA), which according to Christian Rico, means Imperator Augustus Germanicus Nero Caesar. This could be a mark of ownership4.
5Based on the excavations carried out in the 2017 season, during which we revised the measurable extent of the wreck site to be an oval of approximately 31 x 22 m, we wondered whether Bou Ferrer could be one of the big vessels of its period. Strabo refers to these when he mentions the excellence of exports from Turdetania (Baetica) to Rome and Italy, describing the vessels as the largest of all the cargo ships that come to Puteoli and Ostia (the port of Rome), and in numbers nearly equal to those arriving from Africa (Strabo, III, 2, 5–6).
Fig. 2: Orthophotograph of the central trench
Orthophotograph of the central trench (22 x 6 m), from where close to 700 amphorae were removed to the surface, showing the portion of the investigated hull (about 55 m2).
(C. De Juan modified from Moya 2019, p. 307)
6Over the course of the 2017 campaign, we completed a large transversal trench (fig. 2) that bisected the vessel where the tumulus of amphorae exhibited its maximum width at the level of the seabed. The slow work of underwater excavation had started in 2006 with an initial survey of 3 x 3 m to assess the state of conservation of the wreck after some unfortunate episodes of looting. We continued in 2007 with the excavation of an area of 8 x 8 m. As a result of those first two campaigns, we were able to develop a clearer picture of the magnitude of the archaeological site, and we realised that extended excavation by ‘horizontal layers’ was not a reasonable approach. After the scientific discussions that followed those campaigns and the subsequent economic crisis that made funding extremely difficult, we resumed excavations in 2012 with the goal of enlarging the 2006 trench to the east and west, producing the aforementioned transversal trench of 22 x 6 x 3 m, which was not a perfect size given the nature of the cargo, from which we recovered close to 700 amphorae. This work, often entailing excavating more than of 1.5–2 m below the level of the seafloor, required the investment of numerous man-hours. It has, however, produced significant information with a large amount of archaeological data for the interpretation of the site.
7The study of two coins that were discovered in the 2017 campaign, (one dupondius and one sestertius) conducted by Tomás Hurtado (University of València), has given us a terminus post quem for the wreck of 66 AD, a date that is consistent with numerous other aspects of the wreck, such as amphorae typology, ingot epigraphy and pottery related to the crew. Taking all of the data into consideration, we propose that the wreck event occurred prior to the end of Nero’s reign (54‑68 AD), giving us a terminus ante quem of 68 AD, and allowing us to assign a date range to the shipwreck between 66 and 68 AD.
- 5 The pottery, under study, shows only two products from port markets: Rome and Gades.
8The east/west cross trench excavation (fig. 2) situated in the amidships area has provided a lot of information about the architecture of the ship, since about 55 m2 of wooden structures were recorded. A new cross trench under excavation from 2017 to 2019, 14 m north of this amidships trench, revealed mariners’ pottery5 that could be associated to a store or kitchen area, where three bilge pump wood plungers appeared. These elements suggest that the stern area of the ship is towards the north.
9The hull of the Bou Ferrer ship is assembled by standard Roman mortise-and-tenon shipbuilding technique within a shell-first construction principle and process (Pomey 2004a). The plank widths are between 22‑29 cm and the average thickness is 6 cm, with some planks reaching 8 cm. Only a few measurements were taken on the assemblages: the mortises are 9.5 cm wide, 1 cm thick, and 10 cm deep. The mortises edge-to-edge distance is 10 cm. The centre-to-centre average distance of pegs is 18 cm. The inner average diameter of pegs is 1.4 cm, while along the keel their diameter is bigger (2 cm). In some parts of the hull, a resin or pitch is preserved (analysis pending), demonstrating that the hull was internally protected by this material.
- 6 The samples are under treatment: the keelson by ARQUA Tech-Murcia: the keel, the starboard side ga (...)
- 7 Alba Ferreira (Aix Marseille University, CNRS, CCJ, France) carried out the identification of the (...)
- 8 We thank Prof C. Pulak for his suggestion.
10In 2012 a sample of the keelson was retrieved from the site of frame C123, and in 2017, taking advantage of that previous intervention, it was possible to sample the keel, the starboard side garboard and the second strake6 The keel (Quercus ilex7) was in a good state of preservation, but we did not expect to find a double garboard on both keel sides, as has previously been found all along the keel on the 1st century AD Ses Llumetes wreck at Porto Cristo (Munar et al. 2018). The Bou Ferrer keel has no rabbets or chamfered works. There is a kind of step, 3‑4 mm deep, only on one side of the keel for the lower garboard. As the samples were only about 15 cm long, there is no way to verify whether the double garboard will appear along the entire keel or just in the amidships area, maybe as a kind of reinforcement to fix unknown issues in standard garboards or even as a repair8. Further research to check for the presence or absence of the double garboard along the whole keel is absolutely necessary.
11The keel section is 26.5 x 22.5 cm, while mortises are 11 cm deep and 1.1 cm thick. The upper garboard is 8.5 cm thick and rectangular in cross section. The second strakes joined to it. It is 6 cm thick and the difference from the upper garboard creates a profile previously seen in other wrecks of the period (fig. 3). The lower garboard is 7 cm thick where it touches the keel and the sample showed a bevelled surface towards the second strake, as seen in the Ses Llumetes wreck (Munar et al. 2018). As the samples were taken between frames, it was also possible to check treenails in frames C123 and C124 in an attempt to understand the shipbuilding sequence along the ship’s axis.
12The study of the keel and the garboard cross section gave us the opportunity not only to carefully analyse these structures, but also to advance comparisons with other Western 1stcentury AD wrecks with a flat bottom and round turn of the bilge.
13In the eastern part of the trench, the preserved wood ends at the vessel’s lower wale (fig. 2–5). It is a massive piece with rounded inner and outer surfaces, 31 cm wide and 20 cm thick. Rounded inner faces for wales are not very common. In the Bou Ferrer wreck, despite the damage caused by Teredo navalis, it was possible to record the beginning of the arc in the cross section (fig. 3): this is visible in the orthophoto (fig. 2–5). Unfortunately, frames were not preserved at this spot.
Fig. 3: Cross section of frame C123, from which samples were taken, and C127, the longest preserved frame
a) represents a proposed reconstruction of the Bou Ferrer’s axial carpentry.
(drawing R. Geli and C. De Juan)
14An imperfectly preserved longitudinal piece that looks like an upper wale or even a bulwark appeared at the end of the western sector (starboard). Another element, which could be interpreted as 90º knee, appeared to be associated with it, as well as five aligned wooden sail rings underneath the wale (fig. 2–6). The piece is 29 cm wide and 20 cm thick, and on most of it there appears to be no evidence of mortise-and-tenon joinery with another plank, which would support a bulwark hypothesis, but towards the stern it is joined to massive pieces, not just planking, which are too eroded to be fully understood. The timbers continue towards the northwest under 1.5 m of sediment. Since it is possible to retrieve more information from this sector, new excavations are necessary to understand the role of this piece.
15The frames seem particularly robust in the excavated area: floor timbers are placed consecutively without the presence of half-frames. Although alternating floor timbers and half-frames are common in the Graeco-Roman framing pattern, there are other examples which do not follow this arrangement, for example, the 1st century AD Sud-Lavezzi 2 shipwreck (Liou, Domergue 1991, p. 121). In the Bou Ferrer wreck, the frames are 12 cm sided and 25 cm moulded under the keelson, lowering in height to 15 cm towards the turn of the bilge. Room and space measurement is 12 cm on average. The joining system of frames to hull is by treenails, 1.8‑2 cm in upper diameter, driven from outside, combined with blind iron nails also driven from outside, set 10 cm apart on average. The nails are probably made of iron (not yet analysed) until the portside strake 15, before the turn of the bilge curve, thereafter they are copper (analysis completed). The floor timbers have a round central limber hole above the keel, about 7 cm wide and 5 cm high. Other triangular limber holes are situated close to the turn of the bilge at the level of strakes 9–11 (fig. 3).
16There is no precise evidence of beams, but we were able to propose the position of a beam, which could connect the sides between the lower wales (fig. 2–5). In 2013, excavations of the stowed amphorae documented an empty transversal row between the second and third tilted cargo layers. This empty amphora row matches the position of a stanchion mortise (fig. 2–7), recorded in 2014, in the thicker portside stringer above frame C128. Therefore, we propose that a transversal piece, such as a beam, might have run through the amphorae. Lower beams are apparent in the contemporaneous Naevoleia Tyche bas-relief (Basch 1987, p. 457) and in Caligula’s Nemi wrecks (Ucelli 1950, tav. 2). Maybe an upper beam could be related to the 90º knee that appears beneath the upper wale/bulwark (fig. 2–8).
- 9 Due to this and other characteristics, we propose (De Juan 2013) that 1st century AD shipbuilding (...)
17The keelson (fig. 2–1) probably played the role of a mast step and was set directly on the frames, as in other 1st century AD wrecks, for example, Sud-Perduto 2 (Bernard 2008)9 The keelson has a trapezoidal profile with its upper surface wider than the lower. Its lateral surfaces have the precise inclination to perfectly stow the rows of triangular lead ingots, which are in direct contact with the keelson sides. Two thicker stringers (fig. 2–2) also helped to hold them in place (De Juan 2018b).
18Despite the fact that the upper surface of the keelson is badly damaged by Teredo navalis, thanks to a sample, it was possible to measure the trapezoidal cross section, which has a maximum width of 60 cm, and a minimum of 50 cm on its lower surface. The keelson gently tapers longitudinally towards the stern (fig. 2). Given the position of the trunk centre and following the growth rings observed in the sample, we propose its height to be 30 cm. The system for fixing it to the floor timber upper surfaces, following the Hellenistic and Roman Republican tradition (Pomey 2004b; Pomey, Rieth 2005, p. 163–164), is through a series of notches in its lower face to allow coupling with the floor timbers, and resulting in a precise fit. The craftsmanship on the lower face recalls that observed on the Madrague de Giens shipwreck (Pomey 1982) and other similar examples of the same date.
19As was common in Roman times, the internal carpentry is composed of stringers and movable planks. The stringers are quincunx nailed to the frames (figs 1 and 2–3) by iron square section nails (now disappeared). The stringers are placed alternately with rows of shorter and thinner movable planks, which could be dismantled if necessary (fig. 2-4). The stringers have a trapezoidal profile, with a 30 cm upper surface and 35 cm lower, for a thickness of 6 cm. The movable planks have a complementary inverse trapezoidal profile, 17 cm for the upper face width and 15 cm for lower, and 4 cm thick, which made them fit perfectly between the stringers. From both sides of the keelson, the first row of movable planks held lead ingots, which were placed longitudinally. The two rows of ingots, on the starboard and port sides of the keelson, are flanked by two thicker stringers (fig. 2-2), 15 cm in height. No nails or treenails have been found in their upper surfaces, nor recesses in the lower surfaces. The thicker stringers also have a slight trapezoidal profile, with an upper face 22 cm wide and lower 25 cm wide. The keelson and the thicker stringers create two parallel longitudinal channels to hold the lead ingots and restrict their movement. A stanchion mortise for supporting a beam was observed in the thicker port stringer above frame C128. The Bou Ferrer timbers, creating a longitudinal channel to stow the ingots between the keelson and thicker stringers, are unusual in Roman shipbuilding, without any another architectural example to compare with.
20From the thicker stringers and towards the turn of the bilge, movable planks and regular stringers alternate (fig. 3). There is a kind of stringer in the west sector (starboard) with an angled orientation (fig 2–3’) related to transversal carpentry, very well set above the frames thanks to grooves in its lower surface. This piece is still under study and a sample is necessary to understand it.
21The Bou Ferrer wreck dimensions are still uncertain, given that the size of the tumulus at seafloor level (22 x 8 x 1.5 m) does not correspond to its dimensions, since the ship is hidden under almost 2 m of mud. The width of the tilted amphora mound exceeds 22 m, with a height of almost 2.5 m in the central part that protrudes from the substrate. The test trenches made on the axis of the keel have revealed the presence of amphorae 14 m to the north from frame C120 and 17 m in a southeast direction, representing a spindle or oval shape of 31 x 22 m for the tumulus.
22We must logically assess these measurements, which correspond to the dispersion of the cargo, also recently provoked by fishing activities, and do not indicate the original dimensions of the ship. On the longitudinal plane, those 31 m may be close to the real length of the ship, either more or less, but logically we must interpret the transversal width as the result of the opening of the sides, due to the fracture of the hull as documented in the portside turn of the bilge (fig. 3), and amphorae falling to both starboard and port where a large number of them appear upside down.
23With other 1st century AD wrecks in mind, we would like to address two specific and separate issues: the architectural system, in particular, the internal axial carpentry; and the ship shape.
24The observations of preserved amidships structures in the transversal trench excavated in 2017 (fig. 2) have allowed us to compare the Bou Ferrer with other ships which show similar features, such as a flat bottom with rectangular cross section garboards, a rounded turn of the bilge which ends in a lower wale, and a large beam with respect to length.
25The Bou Ferrer shapes, though not necessarily the longitudinal timbers and mast system, might be closer to a group of ships composed of the Saint Gervais 3 (length to beam ratio or LTBR = 2.24–2.28), Tiboulen de Maïre (LTBR = 2.55), La Bourse (LTBR = 2.55) and even Port Vendres 1 (LTBR = 2.05): but not to those like the Madrague de Giens (LTBR = 4.4). This specific vessel group belongs to Pomey’s Western Roman Imperial tradition dated from the 2nd to the 5th century AD. Their shapes provided nautical capacities allowing long open sea routes, and also fluvio-maritime navigation in the deltas of the Tiber, the Guadalquivir and others rivers of the Baetican region. All theses zones were dangerous due to shallow waters and sand barriers. The Bou Ferrer ship probably also plied this transport zone, connecting the ports of Rome and Gades.
26With all the cross section drawings done in the 2017 trench (fig. 2), we have been able to fix manually, according to our personal perspective and after going over the reference group with their amidships cross section archaeological drawings and reconstructions, all the deformations in the frames caused by the passage of time (fig. 4–2.1). The mortise-and-tenon shipbuilding technique does not allow abrupt hull line changes, but involves drawing smooth lines from the ship’s bottom to the sides, giving rise to the characteristic rounded turn of the bilge of known Roman sailing ships. We made a sketch using circles to help us to draw the smooth arcs of the turn of the bilges (fig. 4–3), matching them in our initial drawings with the archaeological cross section wood pieces, and orienting and placing the drawings accordingly.
Fig. 4: Proposed breadth and depth measurements of the Bou Ferrer shipwreck
Proposed breadth and depth measurements of the Bou Ferrer shipwreck, developed from cross section drawings, based on the author’s personal opinion.
(drawing De Juan)
27This exercise suggested that Bou Ferrer’s maximum breadth might be 12.08 m. Seeing a coherence in the forms with the arcs that we drew (fig. 4–4), mostly in the turn of the bilge area, we examined the reconstructions of the shipwrecks La Bourse (Gassend et al. 1974), Laurons 2 (Gassend et al. 1984) and Port Vendres 1 (Liou 1974; Roman 1997), as well the archaeological cross section drawings of Saint Gervais III (Liou et al. 1990) and Tiboulen de Maïre (Poveda et al. 2016). In La Bourse, Gassend worked with the radius of a circle multiplied by four to get the beam as a ratio, in accordance with the proportions he observed in the wreck. In the case of the Bou Ferrer wreck, we realised that 12.08 m was very close to 11.82 m, which equals 40 Roman feet of 29.57 cm. According to Gassend’s radius and circle reconstruction device, it is perhaps possible to advance the hypothesis that the unit (regula) of a radius of 10 Roman feet was used to establish the Bou Ferrer maximum beam breadth of 40 Roman feet (fig. 4). In our opinion, these matters deserve more research in the future before trying to reach any kind of conclusion.
28From the previously mentioned examples, we think that the LTBR of 2.5 of La Bourse and Tiboulen de Maïre ships (Poveda et al. 2016), due to their proximity in shape and chronology, can be referents to assess the possible ratio of LTBR proportions for the Bou Ferrer ship. Using the Roman foot as a unit of measurement, and our theoretical 40 Roman feet beam multiplied by 2.5 LTBR gives a length of 100 Roman feet (29.57 m) for the ship. This value could fit with the dimensions of the tumulus we have recorded on the site.
29As regards the depth, we have no archaeological data yet, only our theoretical proposal of three or four layers of amphora cargo and its 2.10 m maximum height, so our argument is incomplete. A depth/beam ratio for the Bou Ferrer could stretch between 0.5 (1/2) as proposed by Pomey and Tchernia (1978, p. 234) for the Madrague de Giens ship, which has a higher LTBR ratio than the Bou Ferrer, and between 0.33 (1/3) or 0.4 (2/5) as proposed by Gassend et al. 1974. Using the first ratio of 0.5, we would get a depth of 5.91 m, and between 3.94 m and 4.72 m if using a lower depth/beam ratio (fig. 4), which we thought would fit our proposal. We are not able to see any equivalence in Roman measurements in Gassend’s values.
30The 18th century French navy cargo weight formula used to estimate the tonnage of old ships (Pomey, Rieth 2005, p. 44) would calculate the Bou Ferrer tonnage as between 310 and 375 tons following Gassend’s depth/beam ratio, and 468 tons following Pomey and Tchernia’s. The latter, using a 0.5 ratio, is probably outside the plausible range, because such a depth ratio might not apply to navis rotunda, which had an absolutely different shape than a merchantman like the Madrague de Giens.
- 10 The lowest amphorae layer has the mechanical strength to support the weight of consecutive upper l (...)
31In any case, this does not mean that 100 % of the usable cargo space in the vessel had to be occupied with amphorae10, as we understood after superimposing the reconstituted cargo load (2.10 m in height) upon the proposed cross section of the Bou Ferrer (depth between 3.94 m and 4.72 m) (fig. 4).
32The amphorae cargo would be placed low in the hold, helping to lower the vessel’s centre of gravity, and thus improving safety in bad maritime weather conditions. On the other hand, the commercial agents, the shipowner or the sailors themselves would not be keen to leave a large volume of hold completely empty. A complementary cargo may well have occupied part of that space. Classical sources talk about other very common exports from Baetica in the 1st century AD, which we know will not be preserved underwater. Such goods could have been transported above the amphorae (Pliny, NH, 27.94 and 31.86; Strabo, III.4.6): some of them are large in volume relative to weight, like wool.
33The Bou Ferrer is a very interesting shipwreck and an aid in the study of Early Imperial shipbuilding traditions in the western Mediterranean. From the structural point of view, Bou Ferrer together with other vessels of the same date (around the 1st century AD) shares some hull characteristics with the Roman Republican shipbuilding tradition, and others with the Western Roman Imperial shipbuilding tradition. This evidence suggests that these ships could belong to a hypothetical transitional tradition between the two well-defined traditions mentioned. This Early Imperial shipbuilding tradition developed in the western Mediterranean between the end of 1st century BC and the beginning of the 2nd century AD.
34Thanks to the data obtained through the ongoing excavation of the Bou Ferrer shipwreck, we hope to be able to reconstruct a set of hull lines drawings, to estimate the total number of amphorae stowed in the ship, and to estimate the weight of the ship structure. Thereafter, we should be able to calculate different displacements with different loads. In this way, and taking the lower wale preserved in the wreck as an important mark of the most suitable waterline, it will be possible to figure out the different possible draughts of the Bou Ferrer ship as was recently done for the 1st century AD vessel, Napoli A (Boetto, Poveda 2018).