Materials and Design in the Roman Harbour of Soli-Pompeiopolis:
The ROMACONS Field Campaign of August 2009
To the west of Mersin, on the southeast Mediterranean coast of Turkey lie the ruins of Soli/Pompeiopolis, now surrounded by the modern town of Mezitli (Figure 1). The city has a long maritime history. Initially named Soloi, the settlement was supposedly founded by Argives following the Trojan wars and populated by Rhodians from Lindos some four centuries later. Subsequently it was ravaged during the Mithradatic wars (89 – 81BC) and eventually abandoned (Strabo 14.5.8). In 67 BC, Pompey the Great restored the city and colonised it with survivors from his successful campaign against the Cilician pirates who re-named it Pompeiopolis.
Since the harbour installations visible today date to the Roman period, we will subsequently refer to the site simply as Pompeiopolis. The portion of the harbour still well preserved presents an atypical example of Roman maritime engineering in which well clamped ashlar masonry encases a hydraulic concrete core. Although founded in part on a natural reef, it was largely an artificial harbour laid out to a symmetrical geometric design. The harbour was sheltered by two opposed, curving moles 320m long and approximately 23m wide, set 180m apart. They joined on the landward end in a semicircle. The seaward ends curved inwards to frame the harbour entrance (Figure 2).
Most of the eastern mole has now disappeared, and the landward half of the moles is now surrounded or covered by silt and sand. The shape of the harbour was first illustrated in the modern era by Sir Francis Beaufort in his book, Karamania, published to accompany his survey of the southern coast of Turkey between 1811 and 1812 (Figure 3; Beaufort, 1817: 248-56).
An ancient, stylised plan of the harbour also appears on a coin struck by Antoninus Pius between 143 and 145 AD (Figure 4)(Boyce, 1958: 67-78).
Although the harbour has continued to deteriorate and has in part been built over since it was recorded by Beaufort, it is still possible to make out its plan from aerial photographs, and a substantial section of the western jetty survives in surprisingly good condition (Figures 5, 6).
Beaufort’s plan shows what is called a “sluice” on the eastern mole near the shoreline of that period. There is now no clear evidence for it, although Vann suggests that a 3 m gap in the concrete near the present beach restaurant could be part of it (Vann, 1995: 531). Ancient harbour engineers knew about the need for flushing devices that would, if properly designed, allow water to flow through the harbour basin and prevent it from silting up. How effective they were is a matter of debate. A complicating factor is the possibility that the Mezitli River (modern name) might once have run through the centre of the ancient city alongside the colonnaded cardo and flowed into the harbour with all its accompanying burden of silt. At some point well before Beaufort’s visit the harbour was completely clogged with sand and fell out of use. Beaufort’s plan shows it much as it is today, with over three quarters of the basin landlocked and sand dunes covering most of the western side of the harbour.
Although the western mole is the best preserved of the two, only 160 m of the seaward leg is now visible. The curved seaward head lies in ruins, scattered on the seabed, while the landward length is buried under the sand and under the road skirting the ancient basin. The surviving western section, standing in the sea, did so because it was founded on a natural reef whilst the eastern arm was built on sand. The absence of a firm foundation allowed its seaward length to collapse and essentially disappear. The seabed in this area is now strewn with ashlar blocks and rubble and there is no visible coherent structure.
Both moles were composed of large boxes built of ashlar blocks, a type of permanent ashlar formwork filled with hydraulic concrete. The outer faces of the two outside wall were approximately 23 m apart The lower portions of these walls were approximately 2.8 m thick, constructed with approximately uniform stone blocks, 1.6 m long by 0.6 m wide and 0.6 m deep. A well preserved section of the outer wall of the western mole clearly shows the layout of a course of stone blocks (Figure 7).
The design consists of two outer and inner stretcher blocks laid on either side of five headers followed by a double row of headers. Each block was secured to the adjacent blocks with enormous butterfly clamps set into the upper surface of the stone (Figure 8). No clamps have survived, but deep cuttings 35 cm long, 5 cm deep, and varying in width from 6 cm at the ends to 3 cm at their midpoints. There were up to 6 clamps per block. The size of the clamp sockets suggests that they were originally wood rather than metal (Vann, 1995:533).
The upper surface of the western mole is at 1.8 m above sea level, and where stretches of the original paving stones remain, they are 1.3 m long and 0.63 m wide, laid out in alternating rows of header and stretcher. Four cross walls are clearly visible on the exposed surviving length of the western breakwater, are set at 34 m, 30 m, and 14 m apart to form the cells into which the concrete was placed. Most of the cross-walls are 1.6 m thick, built with alternating courses of headers and a line of double stretchers alternating with a header. One cross-wall on the landward end is only 60 cm thick on the upper course, consisting of a single line of stretchers, whilst it widens to a double row of stretchers at a lower level. The cells were built out into the sea one by one, and as each was completed it would have been in-filled with concrete (Figure 9). This form of enclosure was not watertight and the compartments would have flooded to sea level, requiring that the lowest stratum of the concrete be laid underwater. The upper layer of the cells was then filled with a poorer quality concrete, ultimately paved with stone slabs.
Beaufort’s plan shows the colonnaded street, part of which is still visible today, running inland from the harbour along the central axis of the basin. Aerial and land surveys now show that the street was not on axis with the harbour, off centre by some 20 m to the east. This was obviously a deliberate design decision, but one that seems perverse when set in the context of geometrically symmetry harbour basin. This deviation from symmetry would make more sense if at some stage the Mezitli river or a canal leading from it ran through the city and flowed into the harbour.
In the short time available, the ROMACONS team took two cores on the west breakwater, using our standard diamond core-drilling rig. We only worked above water, since the top surface of the mole is currently 1.80 m above sea level. The structure is so well preserved that we were able to drill down through the complete height of the structure and well into the bedrock foundation (core POM.2009.01). The coring was carried out over two days, on the 13th and 14th August 2009. The first core, POM.2009.01, was drilled 5.7 m in from the west face of the structure, 13.2 m from cross wall 01, and approximately 5 m north of a point where bedrock protrudes through the constructed part of the mole (Figure 10). The upper layers were very difficult to drill through due to the friable nature of the binding mortar and the very hard, large aggregate composed of closely packed, smooth riverbed cobbles and pebbles, circa 15 to 20 cm in diameter.
The mortar in the top 0.75 m is a poorly mixed, very pale brown material that is fairly soft, containing much micro-aggregate and many pebbles circa 4-18 mm in diameter. The micro-aggregate consists of rounded sea or river sand, including many grains that are red, green, and blue. There are also many white nodules 4 – 8 mm in diameter that are either un-mixed lime or a product of the long term reaction of the lime with the pozzolana, and also some small fibrous nodules that could be pumice, brownish yellow in colour. There is a laminar deposit across the core at a depth of 0.15 m below the top surface, either a product of the evolution of the mortar, or laitance created during the pour.
From -0.75-0.95 m the mortar was mostly ground away by the coring, although several hard river stones remain, and some pumice lapilli, 20-30 mm in diameter.
From -0.95-1.40 m, the mortar is a very light grey to white, with limestone and other smooth cobbles as aggregate. There are small fibrous pumice inclusions and much rounded sand micro-aggregate, some brightly coloured as noted above. (Figure 11)
At a depth of -1.35 m there is a rounded lump of volcanic tuff, light greenish brown in colour with yellow brown inclusions.
From-1.40-2.20 m, the mortar was the same type as above. Although poorly compacted, it nevertheless is quite hard and varies in colour from white to light green. The change in colour and composition may have to do with the proximity of the water level. There are many voids in the material and many large white nodules, along with lumps of fibrous pumice and particles of green sand, possibly olivine.
Below -2.20 m the core consists of a yellowish red to pink limestone bedrock with a layer of very fine mud just above it. It was impossible to determine precisely where the mud came from, because it infiltrated the core hole each time the tubes were removed to take out the cores. There are no apparent fissures in the rock that could have contained the mud, so it probably had been deposited on top of the bedrock by the river that flowed into the basin or harbour that preceded the Roman mole.
The core POM.2009.02 was taken on top of a flat concrete surface 0.49 m above sea level, and inside the line of the blocks framing the upper part of the mole, 3 m from cross wall 02 and 3.1 m from the western marginal wall outer face (Figure 12). The level surface seems to be the top surface of the lower level of hydraulic concrete, exposed by erosion of the upper level of concrete after the outside ashlar wall was breached at this point. The core hole depth was 0.90 m, although only 0.80 m of material was recovered. The mortar is clearly pozzolanic in character, even containing some tuff aggregate. This same type of tuff is seen in the Italian and Caesarea cores, and it probably arrived as an accidental component of the pozzolana sand shipped from the Puteoli area. The piece of limestone aggregate that forms the bottom of the core appears weathered and does not show any traces of adhering mortar, so it probably represents the bottom surface of this layer of concrete. The core tube went approximately a metre beyond this point, seemingly going through layers of hard and soft material, and jamming frequently. Nothing was recovered from this layer, which may have consisted of a rubble footing.
The mortar is very homogeneous throughout the core: very hard, well mixed, and clearly containing much pozzolanic material. The mortar of the upper portion is a yellowish brown colour, drying to a very pale brown. The mortar contains many nodules of pumice, 11-18 mm in diameter, and many angular fragments of a white material, 2-10 mm across, that are either unburned limestone, un-mixed lime, or a product of the pozzolana-lime reaction that has occurred over time. The mortar is well compacted, but there are numerous very small spherical voids 1-3 mm across, perhaps resulting from some circumstance during placement, or from the chemical evolution of the mortar. The aggregate is the same round river bed limestone cobbles as noted in the first core, with occasional lumps of tuff 23-53 mm in diameter. The tuff also contains nodules of pumice. Wet, the tuff is greenish blue in colour but dries to a light yellow brown. There is no sign of the coloured sand particles seen in the upper portion of Core 01. (Figure 13)
From -0.33 to -0.70 m there is a rapid change in colour of the mortar, to a bluish green colour or possibly greenish grey, drying to a light greenish grey. At -0.70 m the mortar returns to its brownish colour. In our other cores of pozzolanic mortar, the bluish green colour is typical of mortar exposed to sea water, or kept moist by sea water infiltrating the block.
At -0.65 a small fragment of a fibrous material, possibly a reed or twig, D 6 mm, was embedded in the mortar. At -0.75 m a small fragment of wood was embedded in the surface of the core. Oxford University has carried out a C14 dating analysis on the organic material recovered from the cores. Calibrated results suggest that the concrete was placed around 147 AD Ī 48.
After the cores were extracted, the holes were filled with inert sand and sealed with a weak hydraulic lime mortar and parts of the top section of the core and river cobble aggregate, were reinserted as a cap to the filled core holes.
It is apparent from the initial visual inspection that there are two distinctly different concretes, the lower layers of a clearly hydraulic material and the upper layers of a lime or weaker pozzolanic lime mortar. The chemical and thin section analysis will confirm the specific make up of each. One very marked difference between the Pompeiopolis concrete and that we have studied at other sites is in the proportion of large aggregate to mortar. We have seen the percentages range around 40% aggregate to 60% mortar in the concrete sampled at sites along the Italian coast, Alexandria and Caesarea, whereas at Pompeiopolis the percentage varies from 64 to 54% aggregate and 36 to 46% mortar. This ratio is more akin to that found in Roman terrestrial structures (DeLaine, 1997:123).
In line with the analysis and mechanical tests that have been carried out on previous ROMACONS concrete samples; the cores have been taken to Italcementi’s research laboratories in Bergamo, Italy. The C14 date corresponds well enough with the date of the coin of Antoninus, suggesting that the coin commemorated a recently completed project. All the data from the 2009 season will be added to that collected from earlier work and will be published in a final report that is currently being prepared.
We wish to thank Dr Remzi Yağci of the Department of Archaeology, Dokuz Eylul University in Izmir for his assistance and for generously agreeing to support us with the permit request. This project would have been impossible without the practical and scholarly assistance of Dr Nicholas Rauh, of Purdue University. Thanks also go to Dr Enrico Borgarello, of Italcementi who has supported our research and to his colleagues Mr Dario Belotti, Mrs Isabella Mazza, Dr E Gotti, and Dr G Vola for their logistical and scientific input. We especially thank Akin Kaymaz for his assistance in the field.
Beaufort, F., 1817. Karamania, or a brief description of the South Coast of Asia-Minor and of the Remains of Antiquity. London, esp. pp. 248-256.
Boyce, A.A. 1958. “The Harbour of Pompeiopolis. A Study in Roman Imperial Ports and Dated Coins,” American Journal of Archaeology 62: 67-78.
DeLaine, J., 1997, The Baths of Caracalla, a study in the design, construction, and economics of large-scale buildings in imperial Rome, Journal of Roman Archaeology Supplementary Series Number 25. Portsmouth RI: JRA.
Oleson, J. P., Brandon, C., Cramer, S. M., Cucitore, R., Gotti, E., and Hohlfelder, R. L., 2004, “The ROMACONS Project: A Contribution to the Historical and Engineering Analysis of the Hydraulic Concrete in Roman Maritime Structures,” International Journal of Nautical Archaeology 33.2: 199-229.
Vann. R.L. 1995. “Survey of ancient Harbours in Turkey: the 1993 Season at Pompeiopolis,” in XII Arastirma Sonućlari Toplantisi, Ankara, pp. 529-534.
Figure 1 Location plan (C. Brandon).
Figure 2 Sketch impression of the 2nd Century AD harbour (C. Brandon).
Figure 3 Sir Francis Beaufort’s plan (F. Beaufort 1817).
Figure 4 Antoninus Coin, Obverse (American Numismatic Society).
Figure 5 Aerial photograph (Prof. Remzi Yağci).
Figure 6 Plan of harbour (C. Brandon, L. Vann).
Figure 7 Photograph of ashlar marginal wall (C. Brandon).
Figure 8 Clamp cuttings in blocks (J.P. Oleson)
Figure 9 Reconstruction sketch of concrete laid in ashlar permanent formwork (C. Brandon).
Figure 10 Core 01 in progress (J.P. Oleson).
Figure 11 View of core 01 (J.P. Oleson)/
Figure 12 Core 02 in progress (A. Kaymaz).
Figure 13 View of Core 02 (J.P. Oleson).