The Roman Shipwreck of Antirhodos Island in the Portus Magnus of Alexandria, Egypt



Between 1998 and 1999 three excavation campaigns were undertaken on a shipwreck at the now-submerged site of the ancient Portus Magnus, off the coast of Alexandria. The site, close to the island of Antirhodos, was identified through geophysical and archaeological surveys carried out by the Institut Européen d’Archéologie Sous-Marine (IEASM), directed by Franck Goddio. The remains of the ship lie c.5 m deep and are spread over c.350 sqm. No cargo has been found. Artefacts recovered, the details of the ship's architecture and radiocarbon dating all suggest it sunk between the end of the 1st century BCE and the 1st century CE. Its dimensions correspond to those of commercial ships of the Roman era. Identification of the wood used contributes significantly to our knowledge of materials used in naval architecture of this period.

The Roman-period shipwreck discovered adjacent to the Island of Antirhodos in the now submerged Eastern Port of Alexandria was excavated as part of a larger research project launched in 1992 by the Institut Européen d'Archéologie Sous-marine (IEASM) in cooperation with the Supreme Council for Antiquities of Egypt (SCA). The main objective for the larger project was to determine the precise ancient topography of the now-submerged elements of the Eastern Port of Alexandria (Goddio and Darwish, 1998; Goddio, 2008: 26–39; Fabre and Goddio, 2010) (Figs 1, 2 and 3). This research has allowed the contours of the former land surface to be traced with some accuracy and, in some cases, it has provided information on the infrastructure of the part of the port that led up to the Basileia; a seafront palace-complex containing a series of government buildings and cultural institutions. The IEASM's map of the submerged structures in the Eastern Port of Alexandria shows a vast bay containing a group of harbours. After the Roman conquest of Egypt in 31 BCE, the harbour was enhanced, while the facilities from the Ptolemaic period – bedrock-cut basins and the existing dikes, jetties, curved sea walls and so forth – continued in use (Fabre and Goddio, 2010; Fabre and Goddio, in press). The investigation of the Eastern Port has focused on determining depth and sedimentation type, tracing the plan of the docks, creating a timeline for the development, abandonment and destruction of the sites and, of course, identifying shipwrecks. It was in the course of a test sounding (I1) that the shipwreck of the island of Antirhodos was discovered.

Figure 1.

Map showing the location of Alexandria in the Mediterranean.

Figure 2.

Map showing the East Port of the Portus Magnus at Alexandria and the position of the Antirhodos wreck (Map by Franck Goddio, © Franck Goddio, Hilti Foundation.)

Figure 3.

Ship of Antirhodos Island in the Portus Magnus of Alexandria, Egypt. (Photo Christoph Gerigk © Franck Goddio, Hilti Foundation.)

Site and excavation

The island of Antirhodos is 350 m long by 70 m wide and located off the south-western tip of the Poseidium peninsula and can be divided into three distinct areas. The main area follows a south-west to north-east orientation and is aligned with the sea-wall and the pier that extends from the tip of the peninsula, parallel to the ancient coast. The second is formed by a 340 m long by 30 m wide sea-wall that protrudes from the north-western end of the island. The third area, oriented north-west to south-east, terminates in a jetty (J4) built from limestone blocks, which advances to the north-east. The latter forms, with the island, a small port (H1) fully sheltered from swell and waves, which may correspond to the ‘small port’ described by Strabo in the following terms: ‘… private property of the kings, as is Antirhodos, the island located before the man-made port, containing a royal palace and a small port’ (Strabo, Geography, XVII, 1, 9).

The ship was lying in the inner harbour (H1) of the island along the solidly built artificial jetty (J4). It was discovered at a depth of 5 m, under 500 mm of sand. The layer of dense grey clay beneath the sand perfectly preserved the lower part of the hull. The keel was aligned to 30° degrees, that is approximately north-east–south-west, thus parallel to the jetty. The remains of the hull were lying flat on the sea-bed without any trim to starboard or port. The ship was relatively well preserved. Slight settling of the sides was noticed on the bilge. The sternpost and, above all, the stern itself have been for the most part destroyed. All the upper pieces have been lost. A large hole approximately 1 m long and 700 mm wide was been discovered between frames 60 and 64 on the starboard side of the keel. This damage was probably inflicted by modern drilling as the wood appears to have been pierced by a powerful rotating mechanism.

The position of the shipwreck on the bathymetric chart is puzzling as it lies only 500 mm deeper than the ancient island's shore. This is the result not only of silting, but also a rise in sea-level. It may be assumed that the ancient coast had an elevation of at least 2 m above the water. At the same time, even a ship of average size would have required a depth of 3 m of water to approach the pier. Thus, we can estimate the difference between sea-level and the bottom of the port would have been at least 5 m.

The shipwreck has been the focus of two excavations in 1998 and 1999. After an initial evaluation, it was evident that the ship did not contain any cargo. The excavations thus aimed to identify the main structural elements of the hull and the means by which they were assembled. For these purposes, the surface of the ship was cleared from bow to stern. At the end of each season the shipwreck was covered again with a protective layer of fine sand and the weak parts of the construction were supported by sand bags. The poor visibility in the area (often falling to 0 m and never exceeding 1 m) was a constant hindrance to the excavations. Nevertheless, the ship was successfully recorded, making it possible to use the diverse elements of the structure that were still intact as evidence for understanding how the floor timbers were connected to the planking and how the pegs were driven in. The study of how the timbers were joined and the assembly of the garboards to the keel was conducted by observation of the ship's broken and detached fragments. A tunnel was dug under the central axis of the boat in order to carry out a precise analysis of the section of the keel and of the manner in which the garboards were joined to the bottom. Frame 19 was lifted and studied on board of the research vessel for a short period of time and then returned to its original position. No other major constructional elements were lifted or otherwise removed from the vessel.

Samples of wood were taken for radiocarbon dating, which, along with the analyses of the archaeological objects discovered on board and in the stratigraphy sealing the wreck, has made it possible to assign an approximate date to the ship. Identification of the types of wood sampled from the wreck provides information about the wood species used in its construction.

Date of the Ship

The objects discovered within the stratigraphic layer covering the wreck – a sand layer with a high concentration of limestone and potsherds − suggest that it was deposited over the wreck between the 1st century BCE and the 4th century CE.

The ceramic material consists essentially of table and cooking wares and amphoras. The earliest datable find was the base of a Cnidian amphora that dates from the second half of the 1st century BCE (Empereur and Hesnard, 1987: 21, 63 fig. 16; Senol 2003: 194, fig. 7). The table and cooking wares can be attributed essentially to the 2nd century CE (Abadie-Reynal, 2007: 213–214, pl. 56 360.1). For example, a pot with a horizontal rim and two small handles relates to a type of production that seems to appear in the Aegean world in the first half of the 2nd century CE and disappears sometime after the middle of the 3rd century CE. A mortarium can probably be interpreted as an Italic import produced between the middle of the 1st century and the middle of the 2nd century CE (Hayes 1991: 71, fig. 26.1). The excavation of the sector where the shipwreck was found also yielded an oil lamp (Grataloup, 2008: 363, no. 487), which from the fabric and slip would appear to be similar to a lamp type produced in North Africa in the 2nd century CE, although the shape itself is unusual for this region (Joly, 1974: no. 605, pl. 24; Bailey, 1988: 42). The amphoras (intact or fragmentary) found in the upper strata (of the wreck) stem from a period between the second and fourth centuries CE (in addition to the Cnidian fragment mentioned above). The majority of the amphoras are imports: wine amphora of the Agora F65/66 variety from Asia Minor (Tomber, 1996: 45, fig. 3; Lemaître, 1997; Senol, 2001: 383, fig. 10.22–25), Cretan Amphora 4 from the 2nd century CE (Empereur and Picon, 1989: 551.580, fig. 25 a–b; Hayes, 1991: pl. 25.20A, fig. 70.19; Goldstream et al., 2001: 158–159, Marangou-Lerat, 2002: fig. 1.4), carrot amphora produced in Syria in the course of the third and fourth centuries CE (Empereur and Picon, 1989: 232–233, 244–245, fig. 9), Gaza Amphora or Late Roman 4 from the 4th century CE (Peacock, 1986: 191–192, class 46; Majcherek, 1995: 167, pl. 5.3; Dixneuf, 2007: 541). Along with these imports, an Egyptian wine amphora found in this layer belongs to the ‘bitronconic’ (AE) variety, which corresponds to a type produced from the 1st century to the 3rd century CE (Empereur and Picon, 1989: 234; Marchand, 2007: 178).

A bronze coin was discovered in the course of the excavations of the upper layers to the north-west of the shipwreck. It could date to the reign of Antoninus Pius (138–161) (pers. com. Andrew Meadows, Deputy Director of the American Numismatic Society).

Two gold rings found in association with the shipwreck dated to between the first centuries BCE and CE. The first of these rings consists of a hoop made of three massive beaded wires. A gold granule was attached to each end of the hoop. The oval setting soldered to the ring contains a double-layered, dark blue and white intaglio. The engraved image shows a bird carrying a ribbon in its beak (Fig. 4a). The second ring is polygonal in shape and has a tiered bezel. It is undecorated except for some engraved lines on the edges of the bezel (Fig. 4b) (Bakr and Stolz, 2008, 361).

Figure 4.

a) and b) Gold rings found in association with the Antirhodos wreck. (Christoph Gerigk © Franck Goddio, Hilti Foundation.)

Other objects found in the stratigraphic layer over the ship, such as peach stones, walnuts, hazelnuts, small lead weights and ivory pins, do not provide any chronological evidence.

A total of six wood samples were collected from the wreck itself for radiocarbon dating. The analysis of the material was carried out at Archéolabs (St. Bonnet, France) by A. Cura and L. van der Plaetsen (ARC98/R2123C, ARC02/R2771C1, ARC02/R2771C2). As can be seen in Table 1, four standard dates are very close to each other, within the range of 10–60 ± 45 AD. The results were analysed using OxCal, which combines the radiocarbon data from the Antirhodos ship, and assumes that there is a relationship between all of the wood as they were part of the same ship built at the same time. The results suggest that there is a 95% probability that the ship was built between 75 and 211 CE and a 91% probability that it was built between 75 and 174 CE. The program highlighted two samples, the tenon and the plank, that are outliers, which stretched out the date range. Without these – if they were repairs for example – the combined date would be between 30 and 129 CE at a 95% probability.

Table 1. Main results of 14C dating six wood samples form the Antirhodos Island ship. Based on the analyses of Archéolabs (St. Bonnet, France) by A. Cura and L. van der Plaetsen. The highest probabilities and corresponding dating periods are in bold type
No.SampleConventional AgeCalibrated DateInterval 1Prob. 1%Interval 2Prob.2%Interval 3



1Keel10 CE±4090 BCE–130 CE90 BCE–1 BCE16.51 BCE–130 CE78.9  
2Peg60 CE±45160 BCE–80 CE160–120 BCE4120 BCE–80 CE92.4  
3Treenail40 CE±4595 BCE–125 CE95 BCE–65 BCE4.465 BCE–90 CE86.890 CE– 125 CE5
4Plank165 CE±45130–380 CE130–265 CE59265–345 CE33.1345–380 CE4.5
5Keelson15 CE±4590 BCE–130 CE90–55 BCE1.855 BCE–95 CE83.895–130 CE11.4
6Tenon165 CE±4590–390 CE90–125 CE2.8125–365 CE90.4365–390 CE4.2

Some of the details of how the ship was constructed also support a date in the 1st century CE to the beginning of the 2nd century CE. The ship is much broader in section than all known pre-Christian era shipwrecks, as reflected by the shape of the keel, which does not resemble a keystone but an anvil. An increased hold volume was only made possible by developing the ship's inner structure (Steffy, 1990) and such changes can be observed from the beginning of the 1st century BCE. It is clear that the construction of the Antirhodos Island ship included a well-built frame: the floors are bolted to the keel, which, according to available archaeological evidence, only occurs after the 1st century BCE. It can consequently be suggested that the Antirhodos ship was built at some point in the 1st century CE or the early years of the 2nd century.

Construction of the ship (Fig. 5)

Figure 5.

Plan and sections of the Antirhodos wreck. (Drawing P. Sandrin © Franck Goddio, Hilti Foundation.)

Longitudinal structure

The keel, made of Aleppo pine, was entirely preserved and measures 16.5 m in length. With the aft cut-up and the preserved fragment of the stem, the ship in situ has a total length of 24.6 m. The longitudinal profile of the keel is fairly even. At the level of frame 37, which corresponds to the region of the bilge pump, a joint with the aft cut-up was found. This detail was intended to soften the rise of the lower part of the hull towards the stern. The angle of elevation of the aft cut-up is 7° and its length is 5.9 m. It was not possible to ascertain the number of sections of the keel.

The keel has an anvil shape with the upper part overlapping the garboards. The lower part of the keel is visibly rounded. The form of the keel was studied at both its ends and in two of its sections (at the level of frames 37 and 21) (Fig. 6). The keel-sided dimensions vary from 462 mm at the keel-aft cut-up joint to 299 mm at the keel-stem joint. The moulded dimensions vary respectively from 460 to 390 mm. The angle of rabbet along the keel is fairly constant, varying between 61° and 65°. Overall, the keel of the Antirhodos ship resembles the anvil-shaped keel of the later shipwreck of Yassıada (Turkey, 7th century CE), which potentially allows for a more solid connection between the keel and the framework (Steffy, 2001). The keel of the Antirhodos ship clearly differs from the classic, keystone-shaped keel where the strakes protrude from the keel at a wide angle (rising rapidly) resulting in a hull shaped like a wine glass (Steffy, 1990). The body of the Antirhodos Island ship is much broader and does not belong to this type.

Figure 6.

Section showing anvil-shaped keel, garboards, bolt, tenons and nails (position shown on Fig. 4). (Drawing by P. Sandrin © Franck Goddio, Hilti Foundation.)

Both the keel/stem and a part of the aft cut-up/stern Z-scarfs have been preserved at either end of the ship. The stern scarf was attached by means of two copper bolts passing through the centre of the aft cut-up of the keel. The bolts secured the two parts of the scarf, but were broken when the stern fell off. The diameter of the bolt shafts at the upper surface of the scarf is 29 mm and the diameter of their heads is c.50 mm. The joint was also strengthened by five treenails driven in from above. A trapezoid-shaped mortise for housing the tenon of the stern part of the joint is preserved in the upper part of the aft cut-up.

The Z-scarf joining the keel to the stem of the ship measures 590 mm in length. At this joint, the keel is 290 mm sided by 390 mm moulded, while the stem is 280 mm sided by 440 mm moulded (Fig. 7). The joint is compressed by nine wooden waterstops with a diameter of 12 to 25 mm and secured by several nails. The remains of a horizontal key made of oak and measuring 307 mm long and 20 mm thick was discovered inside the scarf. The joint was reinforced by a bolt (with a shaft diameter of 20 mm and a head diameter of 45 mm) driven from below and penetrating through the centre of the scarf. While vestiges of bitumen covering Z-scarfs have been discovered on other shipwrecks (for example Pomey, 1978, 85–87) no traces of resin were preserved on this scarf's surface.

Figure 7.

Detail of Z-scarf showing waterstops and horizontal key. (Drawing by P. Sandrin © Franck Goddio, Hilti Foundation.)

The Z-scarf joining the keel to the aft cut-up was discovered at the level of frame 37. In contrast to the keel-stem joint, this scarf had no key and the upper stern section had no tenon on its upper surface.

Two side-keelsons made of Sylvester or Mountain pine are preserved over a length of 8 m from frame 34 to frame 59. The keelsons are rectangular in section and measure 400 × 350 mm (starboard keelson) and 450 × 350 mm (port keelson). Both keelsons were attached to the floor timbers and to the second planking strake by copper bolts, as can be seen in Figure 8. Trapezoidal cuttings in the keelsons correspond to the position of the bilge-pump well. The distance between the two keelsons just forward of the well is 430 mm. The detail of the mast-step that was normally placed between the side-keelsons is absent and it is likely that this was reclaimed not long after the ship had sunk, as is the case for many other shipwrecks (Rival, 1991).

Figure 8.

Section showing 1. anvil-shaped keel; 2. bolts; 3. planking; 4. pegs; 5. tenons; 6. shim ; 7. plank in the hold; 8. timber; 9. floor timber; 10. stringer; 11. limber hole; 12. nail; 13. caulking. (Drawing by P. Sandrin © Franck Goddio, Hilti Foundation.)

From frame 42 towards the stern, the longitudinal strength of the ship was reinforced by the central keelson made of Aleppo or Turkish pine. This triangular piece, with a side of 380 mm, is the continuation of two side-keelsons. In fact, all three keelsons overlap over some 2.5 m of the hull to strengthen an area of profile transition near the keel/aft cut-up joint. The central keelson was attached to the keel by copper bolts passing through the frames. It is preserved up to frame 28, measuring 4.5 m in length.

The stem, made of Stone pine, is incomplete and is preserved for a length of 2.2 m only. The stem is 280 mm sided by 440 mm moulded at the joint with the keel and the stem's moulded dimension decreases to 150 mm at its front (preserved) extremity. The angle of the stem rabbet is 42°. The aft part of the ship was also damaged and thus the stern of the ship is missing.

The carvel planking of the ship is made of Aleppo/Turkish pine and Sylvester or Mountain pine. The former was often preferred in ancient times for the planking of full-bodied trade vessels because of its availability and the ease with which it could be worked (Steffy, 2001). The thickness of the planking varies from 90–135 mm with an average value of 108 mm. The garboard is 115–120 mm thick. As it was not possible to make a complete section of the hull, only one wale has been identified with confidence: that is strake 5 (135 mm thick). The staggered mortises of the planking have average dimensions of 118 by 280 mm (Fig. 9) and are 100–110 mm deep. The tenons were cut from Golden, Kermes or Holm oak. They measured on average 90 mm in width and had rounded edges. They were secured inside the mortises by pegs with an average diameter of 17 mm. The average centre-to-centre distance between the pegs is 152 mm. Pegs securing the tenons inside the garboard were driven in from above the keel, these tenons were additionally supported by copper nails driven in from the sides of the keel. The mortises in the keel were cut at an average distance of 50 mm from its upper surface. This distance increases in the stem to between 70 and 130 mm.

Figure 9.

a) Planking in situ showing staggered mortises in starboard strake at frame 37 (Photo F. Pereira © Franck Goddio, Hilti Foundation); b) Drawing of plank with staggered mortises. (Drawing by P. Sandrin © Franck Goddio, Hilti Foundation.)

The planks of one strake were generally joined by a simple diagonal scarf, however, at least in one place – in port strake 15 at the level of frame 29 – a more complicated finger joint was used.

Now, let us examine the details reinforcing the longitudinal strength of the ship. Seven stringers preserved on the starboard side and two on the port side were cut of Sylvester, Mountain or Austrian pine. Only the aft parts of the stringers are still in place; the forward parts have not been preserved. A diagonal scarf was probably used for the longitudinal assembly of stringers. The length of the aft parts of the stringers that have been preserved ranges from 5.7 to 9.3 m. The stringers are rectangular in section: their width, and especially their thickness, decreases as they move from the keelsons to the outer board (Fig. 5). Thus, port stringer 7 (next to the side keelson) has dimensions of 380 by 280 mm while stringer 1 (the best preserved stringer on the port side) only measures 260 by 80 mm. Stringers were attached to the frames and planking by copper bolts riveted from both sides of the hull.

The floor timbers of the Antirhodos Island ship had no chokes. Chokes are indispensable in wine glass-shaped hulls in order for the floors to reach the keel. As the bottom of the Antirhodos ship was rather flat, however, a more simple solution was chosen by the shipbuilders. They used shims—thin longitudinal planks that were put on top of the keel and made to fill the space between the keel, garboard and floor timbers. Generally, there is one rectangular shim on the keel's centre line and two additional shims of slightly triangular section on each board. In the region of the joint between the keel and aft cut-up there are two layers of shims. A single, wider shim (500 × 40 mm) is situated atop the keel and is then covered by a pair of shims (measuring on average 300 × 70 mm) attached laterally by means of tenons (Fig. 6). The same feature was also discovered in construction of the Caesarea ship (Fitzgerald, 1995; 165).

The ceiling is best preserved on the port side of the shipwreck. Twenty-three planks about 300 mm wide and 85 mm thick were cut from Sylvester or Mountain pine. The ceiling planks were placed transversely on three stringers for each board. The outer board edges were carefully cut to closely fit the line of the hull, while the inboard sides were trimmed to leave the bilge pump open. At the aft end, the ceiling is preserved at the level of frame 24 and continues forward till the bulkhead at the level of frame 45. Ceiling planks limited by the longitudinal bulwark of the bilge-pump area were fixed to the hull by treenails of 25–30 mm in diameter and were, thus, stationary. At the same time no assembly to the hull was found among the planks aft of the pump area and clearly these limber boards were mobile.

The mast-step was probably removed for reuse in ancient times; however it must have been situated just forward of the bilge-pump well and supported by two side-keelsons. This section must have extended from frame 46 to frame 57. In the middle of this space two rectangular openings were found in the port side-keelson and stringer 5. It is probable that these openings housed stanchions that supported a beam at a strategic location on the mast. A piece of the mast-step must have been held in place by its own weight. From the other side, three short transversal pieces of wood were used to level the upper surface of the floor timbers and the half-frames in this area.

No fragments of the main deck or upper structures have been preserved.

Transversal structure

Regular alternation of floor timbers and half-frames is to be observed in the construction of the Antirhodos ship. Unlike many other shipwrecks (cf. Pomey, 1982), this is also seen in the bilge-pump area. The pump well is situated over half-frame 42 and there are two floor timbers next to it. Both the aft and bow of the ship were badly preserved and consequently it is not possible to see any change in frame alternation in these regions. Half-frames were placed symmetrically across the keel. In total 58 frames have been preserved including 29 floor timbers and 29 half-frames. Frames were numbered from 19 (aft most) to 76. The average space between the moulded faces of the adjacent frames is 97 mm. The centre-to-centre distance from one floor timber to the next is 603 mm, which is almost the same figure as that of the half-frames (604 mm). The average dimensions of all frames are 325 moulded by 222 mm sided. Generally, floor timbers are larger and measure 370 mm moulded by 235 mm sided, in comparison to the half-frames, which are 280 mm moulded by 210 mm sided.

It can be suggested that the vessel underwent repair during its life, as evidenced by frame 19, cut from sycamore fig (Ficus sycomorus) (Fig. 10), which appears to have been a replacement. This frame, which is the single naturally formed shape that was preserved in the construction, was found detached from the hull at the aft end. With a span of 2.3 m, this floor timber measures 350 × 350 mm in section. The lower surface of the frame was sawn and its outboard section was levelled to allow the frame to sit vertically on the keel. A triangular limber hole, two notches for stringers and traces of resin were observed in the lower part of the frame. A number of transversal treenails seemed to reinforce the frame along a natural crack in the wood. Its assembly seems to indicate that it replaced an earlier, deficient frame. Thus all treenails fixing the frame to the planking were driven at an oblique angle. There were no keel bolts but three small metal fastenings were used instead, all of them driven in from above. Frame replacements have been reported from many other Roman shipwrecks (for example Hesnard et al., 1978; Steffy, 1999).

Figure 10.

a) Frame 19 made of Sycamore fig being recorded. (Photo F. Pereira © Franck Goddio, Hilti Foundation); b) Drawing of frame 19. (Drawing by P. Sandrin © Franck Goddio, Hilti Foundation.)

Assembly methods for fixing the framing to the hull have been analysed mainly in the better-preserved aft part of the ship. It is possible to identify three different types of assembly. First, starting from frame 40, some aft floor timbers were fixed to the keel by means of copper bolts with a shaft diameter of 20 mm and head of 29 mm. Bolts passing through the floors from frame 40 to 30 were attached to the central keelson above the floors at regular intervals. Second, floor timbers and half-frames were fixed by bolts to the planking, side-keelsons and stringers. Third, each floor timber and half-frame was fixed to the planking with a great concentration of treenails. Treenails were staggered across the frame to prevent the wood from splitting. The distance between the treenails varies in most cases between 110 and 140 mm.

Futtocks were not attached to the corresponding frames, but were fixed independently to the planking with treenails. There is generally a space of 50–180 mm between the frame and its futtock.

The central limber holes in the floor timbers have a triangular, rectangular or (in the case of frame 51) circular form. It should be noted that limber holes in the floor timbers aft of the pump well are triangular, unlike the rectangular ones forward of the pump well. The average width of the limber holes is 109 mm and their height is 73 mm.

The bilge-pump area extends from frame 33 to frame 45 and measures 3.67 m. It is bounded by two longitudinal and two partially preserved transversal bulkheads. The front transversal bulkhead was bolted through the stringers and floor timber to the planking. The relatively thin bulkheads that border the area of the pump were supported by six stanchions approximately 150 mm wide and 150 mm thick. These can be seen in sections 3 and 4 (Fig. 5). Stanchions and side bulkheads were attached onto the outboard edges of the two side-keelsons. The bilge-pump well has a diameter of 570 mm. No fragments of the bilge pump itself were discovered. A wooden disk with a central opening was found but it seems too large to have been a part of the pump mechanism. It has a diameter of 129 mm, while most of the disks found on Roman shipwrecks vary in diameter between 60 and 96 mm (Joncheray and Joncheray, 2002).

The form of the hull has been studied in the sections shown in Figure 5. It can be seen that the ship's body is rather full and that the floor timbers lie flat along the largest preserved part of the hull. V-shaped floors were only found in the stern (frame 27) and they were either destroyed in place or detached from the hull. The floor timbers have no chokes; thin shims were used instead.

Remarks on measurements

Many scholars have already demonstrated that a detailed analysis of a shipwreck can reveal the sequence of its construction and even tell a lot about the phase of its design (Bonino, 1985: 37–53; Steffy, 1991: 1–9; Casson, 1995; Steffy, 1995: 417–428; Arnold, 1998: 73–90; Pomey, 1998: 49–72; Rieth, 1998: 91–108 ; Hocker, 2004; Harpster, 2009: 297–313). So, the standardization of some dimensions of the ship of Antirhodos Island can evoke its phases of construction. The most obvious trace of standardization can be found in the centre-to-centre distance between both the floor timbers and the half-frames as in this case we possess almost complete data along the hull. These distances are almost equal to two Roman feet of 296 mm (Hosch, 2010), or bipedalis, often used as a length guage for some building material (Ulrich, 2007). The frames’ moulded dimensions vary considerably and need deeper analysis, because the thickness of the frame depends on its longitudinal position in the hull. At the same time their side dimensions are more homogeneous with respective average values of 214 and 230 mm, that is very close to 3 palms (palmae) or ¾ of a foot (pes). While the average height of limber holes falls very close to one palmus of 74 mm, their width seems to be equal in most cases to 6 fingers (digiti) of 18.5 mm. Planking data is obviously insufficient for conclusions of this kind, nevertheless, it is possible to state that the diameter of wooden pegs in mortise-and-tenon joinery equals to 1 digitus while average centre-to-centre distance between the pegs is very close to a semipes (half-foot) of 148 mm (or 2 palms). Although the data on the mortises’ dimensions is also rather poor we can suppose that their length has been calculated as 6 digiti with the width of 1 digitus or 1.5 (sesquidigitalis) digiti, depending on the planking width.

Wood identification

A total of 49 wood samples were collected and analysed. Paleobotanic studies were performed by Archéolabs (St. Bonnet, France) under the direction of C. Orcel (ARC98/R2123B, ARC00/R2384B, ARC02/R2771B). Table 2 and Figure 11 show the wood species used for various constructional details of the Antirhodos Island ship. Although it was not always possible to determine the exact types of pine, at least eight different wood species were used in the construction, not counting the rigging. Coniferous species of the same genus of Pinus predominate. From the 4th century BCE through to the 5th century CE, pine is consistently spoken of in Greek and Latin literature as one of the best woods for shipbuilding (for example Theophrastus, V, 4, 4; V, 7, 1; summarized in Fitzgerald, 1995, 91–93). Apart from assembly details for which Holm oak and olive tree wood were used, other broad-leaf species (elm, fig and ash) only occur in the frames and rigging details (blocks). The keel of the ship is made of Pinus halepensis (Aleppo pine) or, more likely, Pinus brutia (Turkish pine) − its eastern, Anatolian variant. While Pinus halepensis is usually of medium size and often bent, its eastern counterpart reaches 20 m in height and remains straight (Quézel, 1976). Only two examples of the use of Pinus halepensis for the construction of a keel can be found in the available archaeological record and, unlike the ship of Antirhodos Island, both these ships (Kyrenia, 4th century BCE and Port Vendres I, 4th–5th century CE) were made almost entirely of Pinus halepensis. It is quite possible that the Aleppo/Turkish pine was the only tree available at the time of the construction of the Antirhodos Island ship that would have provided wood of sufficient length and solidity for the keel.

Figure 11.

Plan showing location of identified wood samples from the Antirhodos wreck. (Plan by P. Sandrin © Franck Goddio, Hilti Foundation.)

Table 2. Identification of wood used in the construction of the Antirhodos Island ship
Structural DetailsNumber of SamplesWood SpeciesEnglish Name
Keel1Pinus halepensis/P. brutiaAleppo or Turkish pine
Central keelson1Pinus sp.Pine
Side-keelson1Pinus sylvestris/P. mugoSylvester or Mountain pine
Stem1Pinus pineaStone pine
Stringer3Pinus sylvestris/mugo/nigraSylvester, Mountain or Austrian pine
Floor-timber3Pinus halepensis/brutia, Ficus sycomorusAleppo pine; Sycamore fig
Half-frame3Pinus pinea, Ulmus campestris/scabra/laevis, Ulmus minor/glabra/laevisStone pine; elm
Floor futtock2Pinus pinea, Ulmus minor/glabra/laevisStone pine; elm
Half-frame futtock1Pinus pineaStone pine
Stanchion1Pinus halepensis/brutiaAleppo or Turkish pine
Planking3Pinus sylvestris/mugo/nigra, Pinus halepensis/brutiaSylvester, Mountain or Austrian pine; Aleppo or Turkish pine
Ceiling1Pinus sylvestris/mugoSylvester or Mountain pine
Locks of scarf joint2Quercus alnifolia/coccifera/ilex, Pinus halepensis/brutiaGolden, Kermes or Holm oak; Aleppo or Turkish pine
Planking pegs11Olea europea, Quercus alnifolia/coccifera/ilexOlive; Golden, Kermes or Holm oak
Tenons4Quercus alnifolia/coccifera/ilexGolden, Kermes or Holm oak
Treenails5Olea europeaOlive
Block (26K)1Pinus sylvestris/mugoSylvester or Mountain pine;
Blocks (Fig. 12)2Fraxinus excelsior, Fraxinus ornus/augustifoliaAsh
Blocks (Fig. 12)2Quercus alnifolia/coccifera/ilex, Quercus robur/petraea/pubescensGolden, Kermes or Holm oak; English, Sessile or Pubescent oak
Cone1Pinus pineaStone pine
Total number of samples49  

The stem of the ship had to withstand far greater dynamic force than the keel, and thus a different species of pine was used for it. Stone pine (Pinus pinea) wood is characterized by a better mechanical resistance than Pinus halepensis/P.brutia (Rival, 1991). The use of Sylvester or Austrian pine for the keelsons, stringers and garboards of the Antirhodos ship also seem to reflect a rational choice. The two side-keelsons were the crucial element of the construction, as they provided the longitudinal strength of the ship and supported the frames and mast-step, which was subject to considerable stress. Sylvester pine wood has good durability and mechanical resistance. In fact, this species is the second best wood (after larch) for naval construction – only firs and spruces offer better resistance along the longitudinal axis (Rival, 1991). The fact that Pinus sylvestris was used so rarely in naval architecture is usually explained by its limited availability as the tree is found exclusively on the southern slopes of mountains, far from any coast.

It is possible to identify three groups of wood species in the transversal structure of the Antirhodos Island ship. The first is composed of two species of pine – Aleppo or Turkish pine and Stone pine. The second is a species of elm tree (Ulmus campestris/minor/scabra/laevis), which provides the wood for curved details as, due to the structure of its dense interwoven fibres, its mechanical resistance is barely affected by sawing. It is considered an ideal material for floor timbers, half-frames and futtocks (Rival, 1991).

The third group of wood species is epitomized by a single floor timber, number 19, which arguably provides the most interesting detail of the entire ship's construction. The piece exceeds 2 m in length and was cut from the sycamore fig tree (Ficus sycomorus). It could suggest a repair carried out in Egypt. Archaeological data outside of Egypt does not provide any evidence for the use of the sycamore fig tree in Greco-Roman shipbuilding. This is not surprising, as this wood did not grow on the northern coast of the Mediterranean and ancient shipbuilders usually had a wide selection of other wood species at their disposal, all better suited for naval construction. Nonetheless, occasionally, the wood from a sycamore fig's relative, the Ficus carica, was used. This tree was introduced from Africa in pre-Roman times for its flavoursome fruit (Ulrich, 2007). The excavations of a large Roman oared vessel from Pisa (Ship C, Italy, 1st–2nd century CE) demonstrated that the fig tree was sometimes used in naval architecture (Bruni, 2000). From early on, Egyptian shipwrights had to use the wood of the acacia, sycamore fig and other local species as trees with large trunks for long timber were very scarce. Documents dating from the 18th Dynasty (1550–1295 BCE) and the Ptolemaic period (305–30 BCE) attest that the sycamore fig was used for ship building (Gale, 2003). The hull of the boat found at Matariya (near Heliopolis) in 1987 was cut from sycamore fig (Vinson, 1994). Recently, another ship with sycamore fig planking dating from the 4th century BCE was discovered in the Grand Canal of the submerged city of Heracleion-Thonis (Goddio, 2007; Fabre, 2011, 15–16).

Two species were used exclusively for the assembly of the Antirhodos ship: the olive tree and the Holm oak. Treenails and a piece from the ship's tenon pegs were made of olive wood (Olea europaea). Olive tree wood is extremely strong and durable, which explains its use for such details. As a parallel, one may mention the Bourse at Marseille wreck (France, 3rd century CE), where out of seven identified treenails, six were made of olive wood (Gassend, 1982). Olive wood was also used on the ship that sunk at Cap Gros (France, 1st century BCE) (Joncheray, 1989: 65), on Port-Vendres I (France, 4–5th century CE) (Rival, 1991: 88, 269, table 11), and on Culip 4 (Spain, 1st century CE) (Parker, 1992: 158). Other tenon pegs (about 60%), as well as the tenons themselves were made of evergreen oak wood (Quercus alnifolia/coccifera/ilex). In antiquity, Holm oak (Q. ilex) was most customarily used for tenons and tenon pegs. The following wrecks feature tenons from evergreen oak (Fitzgerald, 1995: 107): Planier III (1st century BCE); Cavalière (1st century BCE); Madrague de Giens (1st century BCE); Anse des Laurons I (2nd century CE); Bourse à Marseille (2nd–3rd centuries CE) (summarised in Rival, 1991); Rabiou (1st century BCE–1st century CE) (Joncheray and Joncheray, 2009).

Two of the four blocks found aboard the ship were made of ash, which is remarkably flexible and elastic, while offering good mechanical resistance. One serious disadvantage of ash is that it does not last long in humid conditions. However, this was no impediment as long as it was used in the rigging. Two other blocks were made of oak.

It seems that the majority of the wood used in the construction of the Antirhodos Island ship did not occur in Egypt or elsewhere on the African coast. In antiquity, elm tree and Sylvester pine, which make up a considerable part of the construction, only grew on the European coast of the Mediterranean (Spain, France, Italy and Greece) (Quézel, 1985), although it should be noted that the natural habitat of the Holm oak and the Aleppo pine also includes some regions of North Africa. As reported by Theophrastus (De historia plantarum, Vol. II, Book II, 2, 8) and Pliny (Natural History, XIII, 63, 19), an evergreen species of oak was even present in Egypt in the region of Thebes.

The Aleppo or Turkish pine was used on our ship for both the longitudinal elements (keel, keelson, planking) and for the transversal details (frames). In fact, the former species was the wood of choice in Roman shipbuilding (Rival, 1991). This can be explained by the widespread availability of the Aleppo pine along the whole Mediterranean coast, and by the ease with which it can be worked. According to Rival, this supple wood was optimal for a ship such as the Antirhodos Island wreck which possessed symmetrically raised posts and a relatively broad bottom. Finally, available archaeological and historical data suggest that plantations of olive trees existed in Egypt from the Ramesside period (1295–1069 BCE) onwards. In Hellenistic times, surfaces planted with olive trees might have increased, and, by the Roman era, documentation bears witness to extensive olive plantations (Sandy, 1989; Meeks, 1993; Tallet, 2004).

Sheathing and sealing of the hull

The hull was not covered with lead sheathing to protect it from infestation by xylophagous shipworms and to prevent algae and shellfish from sticking to it. The absence of lead plates on the external planking is an interesting chronological indication, since it is known that it was very widely used from the 4th century BCE (ships of Porticello and Kyrenia) through to the middle of the 1st century CE (Gianfrotta and Pomey, 1981: 261; Steffy 1985; Fitzgerald, 1995: 182–195).

Thin bands of lead were found inside the hull in the upper part of the seam between the keel and the garboard, both in the forward and after parts of the ship. The width and the thickness of these lead bands vary between 10 and 15 mm. In all likelihood, the role of these lead bands was to compress the joint and make the seam watertight. It looks like the lead had been added during the construction of the ship and not in the course of the repairs as, at the level of frame 37, it was found beneath the floor timber and the shims had never been removed. These lead bands were intended to protect the joints of the assembly and/or to reinforce or repair certain defective parts of the planking. This practice is documented on the ships of Port Vendres I (Rival, 1991), Yassıada (Steffy, 2001) and Caesarea (Israel, end of the 1st century BCE) (Fitzgerald, 1995).

In several areas of the interior of the hull, the presence of a thick coating was observed. Traces of sealing are clearly visible between the first three strakes. Most likely the substance used consisted of the kind of pitch made from thick vegetal resin which was applied hot, before the deck of what would have been the hold was set in place, so that it could remain mobile (Gianfrotta and Pomey, 1981: 263). It sealed the cracks between the different pieces, thus ensuring the internal water-tightness of the hull and other elements.

Fittings and on board equipment

Unfortunately, the preserved sections of the ship do not allow for an evaluation of the vessel's nautical properties through hydrostatic or hydrodynamic calculations of the hull's structure. It was also impossible to determine the centre-point of the hull. As a result of the destruction of the rear section of the ship, the presence of a false sternpost outside the axial structure, which would have acted as a lee-board, as seen on the ships of Madrague de Giens (France, c.70–60 BCE), Cavalière (France, 1st century BCE), Dramont D (France, 2nd–1st century BCE) and Cap del Vol (France, 1st century BCE) (Pomey, 1982: 136–138), cannot be ascertained. The steering mechanism has been lost, as has the ship's system of sails. A few rigging elements, however, were discovered in the course of the excavations. In particular several ash (Fraxinus) and oak (Quercus) pieces of different sizes and shape were found. These were used to guide, tighten or set the running and standing rigging (Fig. 12). A few lead rings of slight dimensions (40–50 mm in diameter), some of which have rope holes, certainly served to guide the ropes (brail fairleads). These rings may very well have been attached to the sailcloth for the passage of the lines. Similar objects have been documented on other shipwrecks (for example Dramont D, Madrague de Giens, Chrétienne C, Caesarae) (Fitzgerald, 1995: 211–214).

Figure 12.

Running and standing rigging. a) sheaveless block; b) pulley wheel; c) block; d) pulley wheel; e) dead-eye. (Drawing by P. Sandrin © Franck Goddio, Hilti Foundation.)

Two pulleys (Quercus) have also been recovered. These are simple pulleys, made up of a cast shell with a turning sheave revolving around an independent axle. All attempts at determining the location and purposes of these two pulleys within the rigging have proved futile, as they could have been set in a number of places aloft. It is also extremely difficult to determine the weights that such pieces could have displaced for lack of comparison with modern pulleys. This type of pulley was recorded on the shipwreck of Madrague de Giens. It is different from the half shell and sheave (made from a single piece of wood) found on other shipwrecks from the 1st century BCE and the 1st century CE, such as those of the Grand Ribaud (Hesnard, et. al., 1988: 114, pl. 45, B15–16) and Cap del Vol (Nieto and Foerster, 1980: 163–177, fig. 10).


The Antirhodos Island ship belongs to a type of construction quite characteristic of Roman trade vessels at the turn of the millennium. The wood species found in its construction were used with great frequency in shipbuilding at this time, making it difficult to be certain where the vessel was constructed. Nevertheless, the use of some species of indigenous wood could indicate that the ship was a local Egyptian construction, or that it was repaired in Egypt during its active life. The ship features a well-developed internal structure that includes two side-keelsons on the longitudinal axis, one central keelson, and regularly spaced stringers attached to the keel and planking by copper bolts. The side-keelsons supported a now lost mast-step. The ship's simple planking was assembled by tightly laid, staggered mortises fastened with tenon pegs. The transversal structure features a regular alternation of floor timbers and half-frames. The majority of the floor timbers were bolted to the keel. The floor timbers had no chokes, and thus, special shims consisting of three longitudinal planks were used to fill the space between the lower part of the floor timber and the keel. The pairs of half frames found at several locations are positioned symmetrically across the keel. The dense network of mortise and tenon joints from the planking plays an important role in the transversal structure of the ship. The futtocks and half-frames were nailed to the planking. Copper bolts are quite characteristic of this ship's construction. Bolts were used to fix the floor timbers and central keelson to the keel as well as the first three strakes of the planking to the floor timbers and stringers. Bolts were riveted from two sides of the hull to give their heads a round shape. It is worth noting that the body plan of the ship was rather full and that the bottom strakes protruded from the keel at a gentle angle. According to Steffy, this peculiarity of the hull is characteristic of ships from the beginning of the Christian era (Steffy, 1990). The bilge pump and ceiling boards are also quite common for this type of construction.

The significance of the Antirhodos shipwreck goes further than just the vessel itself. The position of the wreck adjacent to the island of Antirhodos enables a better understanding of the development of this port within the larger complex of the eastern harbour of Alexandria. The shipwreck comes from a period − the High Roman Empire, just following the conquest of Egypt (end of the 1st century BCE−1st century CE) − during which major redevelopments took place. The underwater excavations have shown that the harbour's infrastructure underwent major modifications just after the conquest of Egypt (Fabre and Goddio, 2010; Fabre and Goddio, in press), most likely in order to ensure the continued expansion of maritime operations, and especially the maintenance of the wheat supply to Rome (Rickman 1980; Nicolet 1988: 203–204; Andreau 1994: 92–94; Legras 2004: 27, 142–161). Sources show that granaries and their adjoining administrative buildings were erected in the area of the ancient palace (Clauss, 2005), which was a part of the island of Antirhodos during the Ptolemaic period. The presence of the wreck of a Roman trading vessel in what was formerly part of the Ptolemaic palace complex adds additional archaeological evidence in support of the transition of this zone of the port to commercial use.

The ship must have been a trade vessel of relatively large size. Its overall length can be estimated at 30–31 m. We should be more cautious in gauging the breadth of the ship as only the bottom part of the hull was preserved. Yet, taking into account the angle of the rabbet, the shape of the floor timbers and the position of the half-frames, we can suggest that Antirhodos Island ship was c.10.5–11 m wide at the level of its middle beam. Accordingly, the approximate height at the level of the middle beam can be obtained by the classic formula of the ratio of one third of the breadth, which yields a measurement of 3.5–3.7 m. These estimated dimensions are close to those of other ships discovered in the Mediterranean, including the Grand-Congloué (23 × 7 m), Albenga (30 × 8 m), Mahdia (30 × 10 m) and Madrague de Giens (37.6 × 8.7 m). Its coefficient of elongation (length-width ratio of about 2.7) corresponds to the smaller size commercial ships of the Roman era, which range between 2.6–2.8 m (Cavalière, Bourse à Marseille) and 3.5–4 m (Planier III, Madrague de Giens). Finally, while aware of the methodological precautions that must be taken with such a calculation and keeping in mind the approximate nature of the latter (Pomey and Rieth, 2005: 41–44), the load capacity of the Island of Antirhodos ship can be estimated at around 250–260 tonnes. This figure falls within the middle range of the ships of this period. However, it should be noted that, according to legal texts, commercial ships of Imperial Rome had a load-bearing capacity of between 90 and 450 tonnes (Pomey and Tchernia, 1978: 233–251; Pomey, 1981: 96–101). Thus, on the basis of the available documentation, the Antirhodos ship would belong to the category of large commercial vessels. The Bingen Papyrus 77 (P. Mich. 5760a), reasonably attributed to the 2nd century, contains entries concerning 12 ships in an unspecified port of the Delta that would seem to be Alexandria. The majority of them are small coastal trading vessels with mixed forms of propulsions (akatoï). The tonnages of nine ships may have been given. Two-thirds are smaller boats (10–35 tons). Two ships have a load bearing capacity of 75–100 tons. A third, used for sailings to Ostia, is a ship of 238–318 tons (Heilporn 2000, 339–359).


We would like to express our respectful gratitude towards Franck Goddio, President of the Institut Européen d'Archéologie Sous-Marine and director of the excavations, for entrusting us with the publication of the findings of the shipwreck of the island of Antirhodos. We would also like to thank Dr. Michael Fitzgerald, who conducted the preliminary study, as well as Dr. Catherine Grataloup and Dr. Andrew Meadows who analysed the ceramic pieces and coins, respectively. We also address our thanks to Dr. Damian Robinson and Dr. Sabine Laemmel for help with proofreading. We would finally also like to thank all those who took part in the excavation of the shipwreck, especially Bernard Camier, Stéphane Brousse and Fernando Pereira.