Following the Euphrates in Antiquity: North-South Routes around Zeugma
Author(s): Anthony Comfort and Rifat Erge
Source: Anatolian Studies, Vol. 51 (2001), pp. 19-49 Published by: British Institute at Ankara Stable URL: http://www.jstor.org/stable/3643026 Accessed: 25/08/2010 11:19 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=biaa. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. British Institute at Ankara is collaborating with JSTOR to digitize, preserve and extend access to Anatolian Studies. http://www.jstor.org Following the Euphrates in antiquity: north-south routes around Zeugma Anthony Comfort and Rifat Erge Franco-Turkish Mission to Zeugma c/o Institut Francais d'Etudes Anatoliennes, Istanbul Introduction This second article on ancient roads around Zeugma deals mainly with navigation along the river itself and north-south routes on both banks of the Euphrates. Like its predecessor (Comfort et al 2000), this one is based on extensive survey work in the years 1996 to 1999 during the excavations carried out at Zeugma by a Franco- Turkish mission. It discusses sites located in the region surrounding the ancient city on both east and west banks of the river Euphrates and seeks to identify the principal roads of the region used in antiquity. Maps to complement those prepared for the preceding article are included. In Comfort et al 2000, ancient crossing points of the river Euphrates in southern Turkey were discussed in the light of survey work conducted with the help of satellite photos in and around the reservoir created by the Birecik dam. This dam was completed in summer 2000. The lake created behind the Birecik dam stretches north about 100km as far as the Atatiirk dam, for the most part constrained in a narrow gorge; a second lake now also reaches for about 20km south from Birecik to the Carchemish dam, itself approximately 6km north of the Syrian border. The survey was conducted over six visits between 1996 and 1999. Given the shortage of resources and the vast area covered (about 2,500km2) it can only be considered a preliminary investigation and was indeed originally undertaken with a view only to proving the usefulness of satellite photographs as an aid to archaeo- logical survey work. Since completion of this project two technical developments have rendered the potential value to archaeology of satellite technology much greater. On 1 May 2000 the US government removed the artificial distortions, introduced for civilian users, of signals from its series of Global Positioning System satellites. This change means that individual readings can be taken with an accuracy of about 20m instead of 100m anywhere on the globe. Since the end of 1999 the first photographs with Im resolution from the Ikonos series of imaging satellites have become publicly available and it is now possible to purchase more detailed satellite photographs than ever before, as is shown by the photograph of the Coliseum included with their publicity material (http://www.spaceimaging.com/gallery/ioweek/ archive/iowl21399/iowl213.htm). The previous best commercial satellite imagery was the Russian KVR- 1000 series, used for this project, which reveals objects more than 2m across. The photos available dated from 1990 and 1992. Less detailed 'Corona' photographs, also from reconnaissance satellites but of US origin and from the 1960s have also been used. Archaeological survey work for this project was carried out in the context of excavations by the Franco- Turkish mission to Zeugma and- given the very limited resources available it concentrated on ancient roads, which were identified with the help of satellite imagery. Comfort et al 2000 addressed in particular east-west routes crossing the Euphrates in the survey area. This article is concerned with north-south routes, both those passing along the Euphrates valley itself and other routes in the vicinity linking the major centres of settlement in antiquity. It draws on information from the satellite photos, on roads and sites found in and around the valley, on discussions with local people and on occasional references in ancient sources or published accounts of early European travellers. Sites along the Euphrates itself which are mentioned below have for the most part previ- ously gone unrecorded. The earlier survey of the river valley conducted from north to south by Guillermo Algaze in 1989 (Algaze et al 1994) began only at Halfeti; but this town and areas further north were not investigated by him. Standing buildings of the region have recently been inves- tigated in the context of a project sponsored by the Giineydogu Anadolu Projesi (GAP project for the development of southeast Anatolia) (Durukan 1999). Ancient place-names of this region, as well as all Syria, are reviewed by Honigmann (1923/4), who has also discussed the larger centres of the region in a series of articles published in Pauly-Wissowa. His major article on Syria in this encyclopaedia (Honigman 1932) also contains in section 14 (Itinerarien und R6merstrassen) a discussion of many of the routes mentioned below. 19 Anatolian Studies 2001 Rifat Ergec was Director of the Gaziantep Museum and co-director of the Franco-Turkish Mission to Zeugma, but is currently lecturing at the University of Gaziantep. Once again the first author pays tribute to the assistance of Danis Baykan, now at the regional museum at Pergamon, who accompanied most of the survey expeditions and ably interpreted, but also to Irem Gokcaylh and to the many villagers who assisted us by providing information. Ntisret Ozdemir, former guardian of the site of Zeugma, was especially informative in regard to river traffic along the Euphrates. Information for sites in Syria was obtained on a trip to the Euphrates valley south of the border in the company of Justine Gaborit. Apart from those published here, a large number of other photographs of sites mentioned in this text may be seen on the Project Zeugma website (www.ist.lu/zeugma/). Maps for the routes along the Euphrates valley are included, but for areas further afield readers are referred to those included with the preceding article. Trade and communications along the river Euphrates For early periods, there is almost no direct evidence for use of this part of the river Euphrates as a means of transport. But the indirect evidence is substantial. The Ubaid and Uruk period settlements discovered along the river to the south of Birecik become less frequent to the north but right up to Samosata and beyond there are 'hoyuiiks' which indicate early settlements. These very probably used the river for transport, especially if they were involved in exporting local products and resources back to Mesopotamia, as is widely assumed (for example, Algaze 1993b). Two of the Chalcolithic and Bronze Age sites concerned have been excavated only recently- Horum and Tilbes -but the natural hill of Hacinebi excavated by Stein (1999), 8km east of Zeugma, and the h6yiik of Kurban near Samosata (Wilkinson 1990), excavated by Algaze, had already indicated much activity in periods going back to the Uruk. Algaze's survey of the area in 1989 (Algaze 1994) located even earlier evidence from the late Ubaid at nine sites between Halfeti and Carchemish, including Horum and Tilbes to the north of Zeugma, from the Middle/Late Halaf at three sites and from the Neolithic at two. Another recently discovered site in the basin of the Birecik dam reservoir further north at Irmakboyu (or Nehriseid see below and fig 3) now offers additional evidence of exchange of goods in the river valley as early as the Ubaid. Early Bronze Age cemeteries also existed at Zeugma itself (Sertok, Ergec 1999) and, probably, at Kalecik, 1.5km northwest of Halfeti. The goods exchanged between Mesopotamia and the region of Zeugma in the Uruk period since contact was established around 3700 BC are likely to have been primarily raw materials exported to the south in exchange for manufactured goods, as well as tin from Afghanistan. As Stein (1999: 157) remarks, for the Uruk colony at Hacinebi, upstream traffic would normally have been limited to high-value items of low bulk, which, apart from textiles, may have included oils and aromatics, while downstream traffic would have included copper, lumber, gold and semi-precious stones. Presumably this is also true for Halaf and Ubaid periods, i.e. from the sixth millennium BC. Long-distance exchanges involving obsidian, shells, semi-precious stones, copper and bitumen are known to have been conducted in the area even earlier, in the Aceramic Neolithic; since the sources of much of the obsidian found in the Near East lay in eastern Anatolia it would be unsurprising to learn that the Euphrates had provided a corridor for such exchanges even then, although evidence is currently limited to places much further south along the river such as Abu Hureyra and Bouqras (Roaf 1990: 34). Unfortunately, good evidence to confirm hypotheses as to the nature of trade along the valley is still lacking. Timber from Lebanon and Amanus is, of course, known to have been floated down to Babylon and Assyria at least since the ninth century BC and possibly since the third millennium (Sargon of Akkad mentions extension of his realm to the Cedar Forest around 2300 BC) (Kuniholm 1997). But this wood seems likely to have been transferred to rafts at a point below Birecik. In any case, timbers were frequently acquired by military expeditions sent for this purpose and then as tribute, rather than through normal commerce. Forests along the higher reaches of the Euphrates valley must certainly have existed but it is not known when these were cut down. Ore from copper mines opposite Malatya and east of the river was probably also transported down the Euphrates but no evidence is yet available to confirm this, although Stein did find evidence of copper working at Hacinebi. Below the Chalcolithic and Bronze Age mound of Horum (fig 1; 5km north of Zeugma) squared stones placed in the river and visible at times of low water for considerable distances along the banks must have been either towpaths or quays (fig 6). Those at Horum were originally tied by metal cramps for which the holes were still visible and seem to be of Roman origin. The area around the mound of Horum is thought to have been an important late Roman town called Urima or Antiocheia. Excavation reports for the Chalcolithic and Bronze Age mound have been published since 1997 in Anatolia Antiqua. 20 Comfort and Ergeq (via Uzunburc) 0 Legend * Village A Site * Ancient quarry t Confirmed ancient road A Bridge footing Q Crossing o 0 00 0o Assumed course of ancient road To Birecik Dam 5Km Fig 1. Map of Euphrates valley from Zeugma to Halfeti 21 )es 'III I I o 0 0 0 0 0 0 0 i Anatolian Studies 2001 To Germanicsia (Marash)N 0000 000 000 0 0000000000 ? Kayalar A Kosk Legend * Village A Site / Ancient quarry li?, Confirmed ancient road A Bridge footing Ql Crossing 0000o 00 o Assumed course of ancient road Fig 2. Map of Euphrates valley from Halfeti to Siipiirgui 22 0 0 0 0 0 0 a90g ^ov * Comfort and Ergeq On the right (west) bank these quays seemed to disappear under the riverbank to the north of the mound at Horum, possibly as a result of a change in the course of the river. It was not therefore possible to judge their former length. Nevertheless about 30m were visible in the river. On the left bank, near the similar mound at Tilbes and for a considerable distance to the south (at least 100m), a quay or towpath was seen just below the surface of the river. Here it apparently consisted of smoothed natural rock with occasional deep artificial holes, possibly for wooden mooring posts. The interpre- tation of these artificial 'banks' depends crucially on the river level in antiquity. Both at Horum and at Tilbes the level of the platforms below the river level prevailing from 1996 to 1999 implies that the Euphrates was about 2m lower in the Roman period if they were in fact towpaths or quays. Although visible evidence is also lacking for commercial activity along the valley in Greek and Roman antiquity, apart from these quays or towpaths, there are references to the valley in ancient authors. These sources are reviewed by Kennedy (1998b: 139-62), with particular reference to Zeugma. However, in regard to the use of the river for transport, references are very few. The best concerns a journey downstream of the bishop of Samosata. Theodoretus recounts that the bishop was ordered into exile by the emperor Valens around the mid-fifth century AD. [Eusebius] confided his intentions to one of his household servants who followed him carrying nothing but a cushion and a book. When he had reached the bank of the river (for the Euphrates runs along the very walls of the town) he embarked in a boat and told the oarsmen to row to Zeugma. When it was day the bishop had reached Zeugma, and Samosata was full of weeping and wailing. ... Then all the congregation bewailed the removal of their shepherd, and the stream was crowded with voyagers. When they came where he was and saw their beloved pastor, with lamentations and groanings they shed floods of tears, and tried to persuade him to remain, and not abandon the sheep to the wolves. But it was all of no avail, and he read them the apostolic law which clearly bids us be subjects to magistrates and authorities. When they had heard him, some brought him gold, some silver, some clothes, and others servants, as though he were starting for some strange and distant land. The bishop refused to take anything but some slight gifts from his more intimate friends, and then gave the whole company his instruction and his prayers, and exhorted them to stand up boldly for the apostolic decrees. Then he set out for the Danube, while his friends returned to their own town. (Hist. Eccl. 4.14.3, 364/378, trans Jackson 1892) There are, of course, many references to crossing the river at Zeugma and there is also an interesting reference in Ammianus to the bridge at Capersana. These are reviewed in Comfort et al 2000. The question of the valley as a border is addressed below, but the evidence for military transport along the Euphrates quoted by Dabrowa (1997) refers to more southerly stretches of the river. In living memory the river was used for both downstream and upstream traffic, although boats were dragged upstream unloaded. Goods carried included grain, pistachios and timber. Grain in particular was trans- ported downstream from Samsat (ancient Samosata, now drowned by the Atatiirk dam) to Belkis and Birecik. Pista- chios are reported as having been regularly exported early in the 20th century from the large village of Sanlar (opposite Ayni/G6zeli) to Djerablus presumably for onward carriage to Aleppo via Membij. Except for timber, which descended the river as rafts of interlocked logs ('keleks'), the boats used for transport were flat-bottomed and rectangular (known as 'sahtur'; see fig 17 in Comfort et al 2000) and approximately 6 x 2m. For the downstream journey they were often linked in pairs abreast, while for the upstream haulage they were linked in lines of four, towed by teams of up to eight men, with one man left on board the first boat to guide a steering oar, also used for punting. Arab boats in use on the rivers Tigris and Euphrates up to the early years of the last century are discussed by Ritter (1919); although large masted boats were then in use around Baghdad, it is unclear whether these have ever been in use around Zeugma, the common type for both crossing the river and downstream navigation having apparently been the 'sahtur'. Fishing was still conducted in 1999 from small boats or rafts supported on inner tubes for lorry tyres, the successors of the round coracles or 'kuffas' mentioned in the sources as in use into the 20th century. Some villages, such as Kizilin opposite Col, had a reputation in the area for living at least partly from fish caught on downstream journeys and sold to settlements along the river. Evidently, the downstream journey by boat while catching fish would have been followed by a long walk home. Upstream towing was not always possible from both banks. It was frequently necessary to cross the river to continue on the other bank because of cliffs, which often made the riverbank impassable. The towpaths had largely disappeared by 1996 when the survey began, presumably because of flooding and landslides, but in any case seasonal variations in the river level may have precluded the estab- lishment of permanent towpaths. The satellite photos reveal banks above the level of the river at some points that may have been used as towpaths (fig 5), but may also have been used for caravan traffic along the valley. Freya Stark described such a journey in a boat towed and punted up from Halfeti to Rumkale (Stark 1966: 110). 23 0 \ f Turus (Kuyulu) o o/ onument 0 A 0 0 0 0 0 Legend * Village A Site A Ancient quarry li!? Confirmed ancient road A Bridge footing D Crossing o00 o 0 o o Assumed course of ancient road Fig 3. Map of Euphrates valley from Vahne to Adilpazar I I II Comfort and Ergeq ShiukhTahtani !j Tell lddaye g A Tell Qumluk oo000 00000 ,0????0 o o 0? o oo o o ? , o ? To Aleppo ?o? Legend oo00o o? Ancient roac - ? *O Frontier betv * Village f* Ancient site A Tell/huyuk/n 0 o o 0 0 0 0 o o o oK 0 0 0 0 o ? K o o ?o 0 o ?00 o ?0 0 o o o O o0 Manbij/Hierapolis oo oo?00 o 0O ~ ~ ~~o oO 0 O 0 0 Qo o oo oo o o oo o o o o o o oo d 0 0 o o 0 o o ween~~~~~~ Tuke an Syi MabHeaoli o o o o 0 o o o i~~~~~~~~~~~ o o o \ Qo< \000 j 0~~~~~~~~~~~ nound ?o ?ig o~~~~~~~~~~~~~~~~~~ 0 ?o? il0 00 18 10 Km I I I I I I L 0 Abu Hanaya (?Eragiza) o0 p o 0 000000000000000000000, ^ o00 P .....--......... ..... A Mahdum o 0 0 Salt-lakeait Jabbul To aa To Barbalissus Fig 4. Map of Euphrates valley below Carchemish 25 o ! 0 Tell Amama X noo 0o 0 0 o? 0 0 0 00 0 o o 0 Caecili o o o o o o0 o o o o o ?o O0 t o 1.k. o , Dam abel Khaled 111111 i I I I I O o o oo oo o o oo 00 o o 0 0 0 o o o o o 20 Km I . . Anatolian Studies 2001 Fig 5. Euphrates at junction with Kara Su; showing possible former tow-path on east bank (extract from KVR-I 000 photograph of 22 May 1992) 26 Comfort and Erge? Until modem times this part of the Euphrates valley seems likely to have been used in many periods for both land routes and fluvial transport for downstream traffic. The use of the river as a means of transport must have fluctuated according to the political and economic environment. The complete abandonment of the valley as a thoroughfare, apparent in the 1990s, resulted princi- pally from the arrival of the internal combustion engine and construction of passable roads, which changed the economic balance in favour of road transport. The physical barriers represented by dams just confirmed this abandonment, although the creation of very long reser- voirs navigable in both directions is already producing a revival of river traffic, at least for tourism. New sites and finds along the valley The sites mentioned in this section are the most important new finds along the valley itself. Many will have been affected by the reservoir, either directly or because access has now been impeded. Further sites are mentioned in the succeeding sections as evidence of the ancient roads of the region. They are presented roughly in chronological order; some have already been briefly mentioned in the preceding sections and in Comfort et al 2000. The mound Nehriseid (official name: Irmakboyu; fig 3) was the most important pre-historic site discovered during the survey. It is located on the left bank of the Euphrates (here the south bank) 32km northeast of Halfeti. This area along the south bank was previously unexplored and several other finds are described below. Although not investigated in detail the mound showed evidence in the form of painted potsherds of occupation in the Ubaid period and seems similar in size to that at Kurban. There is a track recently built leading to a house on the summit and there are some other houses on the east flank, but most of the small modem village has moved from the riverbank to the plateau to the west. Excavation of the track and on this east flank had exposed ceramics at several points. The river is about 100m to the northwest by a track leading to the cave- church described below. There is a further mound, which is smaller and seems at least in part to be natural, a further 6km to the northeast, also on the riverbank, at Ozgeren (formerly Adilpazar). For the Iron Age, evidence of travel along the valley is particularly patchy: castles found since 1996 above the river at Kaleboyu (near Elif) and Bozyazi (fig 2) must indicate a desire to control traffic but the period of construction and occupation is unclear. The former, mentioned briefly in Comfort et al (2000: 117), lies on the west bank in a remarkable position on cliffs with views to the north and to Ayni, and close to the route from the river-crossing to the settlements around Elif (4km northnortheast, on the Roman road from Doliche to Samosata). Little was found to date the site; the existing walls are of small irregular blocks, with at least one square tower projecting from the north wall, which was visible for a stretch of 150m, terminating at the cliffs which fall almost vertically to the Euphrates, 250m below. Footings for large cut blocks may be seen as well as a grape-press and foundations of internal structures, but our investigations were superficial. There were two deep cisterns on the summit, partly plastered, and a third cistern at least 4m deep was said to lie below the castle at a level 40m above the river, but the means of access to it were unclear. A substantial ruined settlement (Kucuk Kirkap), with some large door-posts of roughly shaped stone and with cisterns and rock-cut tombs, apparently of the Roman period, lay on the edge of the limestone plateau nearby. An even larger settlement of the same type with about 40 large rectangular houses, with walls up to a metre thick, and with enormous cisterns, lies 5km northnorthwest at Hasanoglu, and a further one was reported nearby on the cliff-top to the northeast, about 2.5km southwest of Elif. Hasanoglu, Elif and another village nearby called Hisar all still have remarkable, tall funerary monuments, which are comparatively well- known (see, for example, Wagner 1985: 32). Fig 6. Towpath or quay? - below Horum Hyviik 27 Anatolian Studies 2001 ^Ton g| mill & cros Falen slabs Level approx. 20Or ve river Fig Plan f temple at Kelefs, near Kllk Fig 7. Plan of temple at Kelefis, near Killik At Bozyazi, on the east bank some 5km to the north of Kaleboyu, the cliffs fall away from the village in a series of remarkable parallel natural collapses. On the final ridge above the river a long and thin fortification exists (20 x 190m), partly cut into the natural rock but with some walling still evident. The blocks in place are larger than at Kaleboyu but equally difficult to date. Several large standing stones on the summit ridge were shaped, there was a substantial circular cistern and, on the river side (northwest), an underground room with perfectly smoothed, square walls, a small rectangular window and a gently curving vault. The visible remains, in particular a semi-circular bastion carved in the natural rock at the south end, seem to indicate that this was an outpost of some importance. A further two natural caves on the southeast, 'inland', side of the castle had been enlarged and fitted out as houses. A track leads down from the plateau to the river at this point and the east bank below the castle is impassable because of the very steep cliffs. There were reports of a spherical stone ball found nearby which had been placed on a field wall and then fallen into the river. An underground temple in a cliff-face above the river was found at Kelefis, near Killik (fig 3), which may on stylistic grounds be dated to the Achaemenid or early Hellenistic period; in particular, several bulbous, and one hexagonal, column capitals remain carved in the natural rock ceiling of the hall, while the columns themselves had disappeared (fig 7). What may be an underground garrison or fortress called Avdules was also found some 5km east, opposite the junction with the G6ksu river (fig 3). This astonishing building on five levels carved into a cliff above the river was extremely difficult to reach on foot; an inscription promised by local guides was not found- possibly the 'building' is Hellenistic in origin, but some pointed arches in niches may indicate a medieval period of occupation. At both these sites passage along the river downstream will now be impos- sible except, of course, by boat. At the hamlet of Col, between Killik and Nehriseid (fig 3), evidence was found of a substantial town, probably dating from Hellenistic and Roman times, which has hitherto gone unreported. On the plateau above the river, with remarkable views into the valley, lies the existing small settlement with buildings that include a large number of reused blocks, probably from a temple located nearby on the top of the cliff overlooking the town and the river. Below, ancient steps and pathways lead down past an ancient press still in use for pressing grapes on the day of our visit to a ravine close to the Euphrates. A hundred metres to the west, water emerged from a spring in a cave just above the river that had once been used to power a mill, now vanished; but ancient aqueducts were visible in the cliff on both sides of the cave. 28 Comfort and Ergec Fig 9. The mausoleum at Nehriseid (Irmakboyu) Fig 8. Col- structures visible on the hillside opposite the modern hamlet Fig 10. The church at Nehriseid (Irmakboyu) A Underground mll Tombs: . path to Ciftekoz + a ...- . steps and arch ? - ".. .. ..a ~ CcCaveas ' v .. Cave tombsaves Euphrates 4 t CARAVANSERAI ? << A VA-,. X S Al track to Sirataslar (Vahne) Fig 11. Plan of t-he S crossin g a t /ahne rata0 100 m Fig 11. Plan of the crossing at Vahne (Slrataslar) 29 Anatolian Studies 2001 Above and on the side of the ravine facing away from the river, a large number of tombs and foundations of buildings cut into the rock are visible (fig 8). The site was not discovered by a German expedition of 1918 sent to chart this portion of the river (Noldeke 1920), possibly because it was invisible from the river itself. This town must constitute a candidate for the Porsica mentioned by Ptolemy and previously identified with locations on the bank opposite Samosata or with the modem regional centre of Yaylak, a small town with a large mound in open country to the southeast of Col. Epiphaneia, a site mentioned by Pliny (NH 5.21), would be another possi- bility since no other comparable town on the east bank of the Euphrates was found over the four years of this survey. However, the latter town is placed by Jones opposite Horum (1937: 441, n 3; see Kennedy 1998: map on 5). A site located here by Algaze at 'Colfelek Tarlasl' and identified as of the Hellenistic and Roman periods may therefore have been Epiphaneia, although no tombs in the cliff-face nearby were found on the occasion of our visit despite the existence of natural caves (Algaze et al 1994: site 20). Continuing along the valley in the northern part of the survey area, further evidence of occupation in Roman times came from the next village from Col along the south bank towards the east again at Nehriseid (official name: Irmakboyu). In addition to the large mound described above, this small settlement boasts a cemetery on a neighbouring hilltop about lkm from the river which has a remarkable small Roman mausoleum still standing, apparently to its original height of 4.20m (fig 9). Originally circular, only the northern half is preserved with the south side having fallen to reveal an altar. The round exterior of the surviving part is decorated with pilasters. No inscription was found as such but the letters 'GIII' may be read with difficulty above the altar and 'III' is roughly carved at two points on the exterior. It seems possible that these 'graffiti' were inscribed by visitors from Samosata, the base of Legio III Gallica. This mausoleum has apparently been preserved because it was believed by local people to constitute a shrine for a Moslem saint, i.e. a 'tiirbe'. It was last described by N6oldeke (1920) and had apparently not been visited since. One hundred and twenty metres west of the mound and the modem houses around it at Nehriseid, there is a large artificial opening in the cliff-face immediately above the river. A tomb with seven loculi is cut into the rock some metres to its east but the principal opening seems very probably to have been a church. About 40 large reddish stone blocks have fallen into the river and may have constituted a platform and facade for the church. The excavated chamber is 10.40m deep and 7.20m wide; the opening itself is approximately 10m high, forming the shape of an upturned bucket. The cliff leans out over the river and a further three small openings are placed left, above and right of the principal room (fig 10). Since the dam downstream will have raised the water only by a few metres at this point (Killik is about 65km upriver from the Birecik dam), the sites described so far will probably be unaffected although access will now be possible only by boat in some cases. Others which are lower-lying will however have already disappeared: in the northern section of the survey area they include a probable ruined caravanserai below Varne (Sirataslar, opposite Ciftek6z, figs 3, 11); a mill at Kelefis near some remarkable springs which rise partly in the river itself; and cave dwellings at Ina in the gorge opposite Hisar. Traces of mills were found at three points along the river in 1999: Rumkale, Kelefis and Col. N6oldeke found a further such mill at Adilpazar (now known also as Ozgeren) about 10km to the east and the last point inves- tigated by this survey before the Atatiirk dam. His description reads as follows. On the left bank above Adil-Bazar a notable feature is a cave mill, which like all those observed in this area must date back to early Christian times, and it is still in use [1912]. Recently it still belonged to a Christian family of millers and had probably done so since time immemorial... In the mills there is usually a stream of water led through a deep, narrow channel hewn in the rock, generally directed straight onto the horizontal millwheels. To ensure the supply during droughts, water was stored in reservoirs in the open among the bushes a little way from the river bank. Mills hereabouts consist of two large adjacent rooms and a common holding area, all cut out of the cliff. A single door leads from outside into the holding area; there are no openings for light. On the right of the door inside the holding area there is a chicken house cut into the cliff with a hole to the outside. The left-hand room serves as a donkey stall; in the right-hand one the millstones lie one behind another in a rectangular hollow; below them the water rushes past in a covered conduit. In the open near the mill chambers there are two interconnected containers; one reaches far into the cliff and is dammed by a small retaining wall, the other has been dug out artificially and is walled on three sides and sealed with plaster in the Roman manner. The mill wheels are, as I said, laid horizontally. They consist of two concentric iron rings set on their edge, between which paddles have been riveted. The square, wooden axle is bedded with an iron 30 Comfort and Ergeq connecting piece into a hard rock at the bottom of the channel. At the top the axle extends into an iron bar which is stuck through the middle of the fixed, lower millstone and firmly connected with an upper millstone by means of a four-edged head and double- winged, flat splint. The axle hole of the lower stone is covered by a leather cover, to prevent the corn falling through the hole. The water runs out of the channel in a strong stream onto the slanting paddles of the wheel and drives the wheel and the upper stone. The problem of translating the vertical power of the falling water into a horizontal, turning motion has been solved with a certain amount of wastage of water, but without recourse to technically difficult installations. Probably a very old solution, handed down by the Christians... (Noldeke 1920: 18, trans Philippa Seymour). Metal parts of mills were found at Kelefis in the river and in the Merzumen valley above Rumkale. These seem likely to have been in use until around 1920, but the associated aqueducts and tunnels are probably much earlier in origin. Large flat grinding stones were also found outside the valley itself at (iftekoz, Haydarahmet and Haciobasi. In all three cases these were in large underground rooms apparently constructed in antiquity. Other similar rooms were found in the Bahqe Dere, (a valley containing a large necropolis just north of Zeugma) and at Bumus (see below). Cave dwellings were evident at several points in cliffs along the valley, some apparently of great age but impos- sible to date in the absence of potsherds. There was a settlement on three levels, with some caves for animals and others enlarged for dwellings, at an abandoned settlement just above the river at Ina (fig 2, northeast of Bozyazi and opposite Hisar) and another below (iftekoz for which external staircases in the cliff face were still visible. This was probably the crossing point used by Humann and Puchstein on their journey from Halfeti to Nemrud Dag (1890: 180) since they refer also to many troglodyte dwellings. The crossing at Ciftekoz is discussed in Comfort et al (2000: 121). Roads along the valley The remains of the Roman bridge found in October 1999 (and briefly discussed in Comfort et al 2000: 117) on the west bank at Habe? (fig 2), some metres north of the junction of the Kara Su and Euphrates, are proof that there was an ancient north-south route along the river -at least for the section from Ayni and Kenk northwards. Because of landslides and erosion no road was visible from the east bank, for example looking down from the castle at Bozyazi, but animals and shepherds still passed along a riverside footpath below the enormous cliffs of Elif and Hisar during the period of our survey. It must be assumed that the caravans, which local people remember as using this route, crossed the river at Ayni, since to the south of Ayni the west bank of the Euphrates is again impassable. This valley route was still called by local people on both banks 'Top Yol' or Cannonball Road. The bridge at Habe~ was only partially preserved, its northern pier having been swept away in a storm in the 1960s. There was a house built on the remaining pier on the occasion of our visit and we were informed that the road had formerly followed the steep cliff on the north side on raised piers until it turned the corner into the main Euphrates valley. The Kara Su river had to be waded or crossed by boat and the route along the river had fallen into disuse. The existence of the Roman bridge at Habe~, constructed by the Legio IIII Scythica (as proven by the nearby inscription on the cliffs from which stone had been quarried, Comfort et al 2000: fig 16), poses a conundrum. For if the river was indeed the frontier between Rome and Parthia from first to third centuries AD then it is difficult to see how Roman caravans and military convoys can have used the road along the east bank of the river valley. But both the bridge at Habe? and inscription at Kenk imply a crossing to the east bank, presumably for caravans continuing either east to Urfa/Edessa or south along the Euphrates, as early as the time of Vespasian. In his discussion of the Euphrates crossings, Syme (1995: 95) mentions the belief that Pompey the Great assigned a portion of Osrhoene to Commagene opposite Samosata. The crossing at Ayni (fig 2, 10km north of Halfeti) and the road along both sides of the Euphrates valley is perhaps an indication that control of the east bank of the Euphrates passed to Rome rather earlier than thought previously. It also indicates that both castles of Kaleboyu and Bozyazi may have Roman rather than Parthian origins even though the latter is on the east bank. The possibility of early Roman activity on the river's east bank finds support from Wheeler (1989: 507). To the north of Ayni, the east bank becomes precip- itous and impassable below the small castle near the village of Bozyazi. Travellers' reports of the 19th century (Ainsworth 1888; Guyer 1916) indicate that caravans coming from the south and following the valley climbed to the plateau at Ayni after crossing the river and could then rejoin the valley to the north just beyond this castle, which was therefore probably intended to control caravan traffic along the east bank. 31 Anatolian Studies 2001 Fig 12. The watchtower 2km south ofAyni (Gozeli) Fig 13. The fort above Tilobiir i., '^ sVy ' -^ Illli""'' *'' - . - |l r.^ 1^ ^ ! wii^ :X;:| - - * '' ^ ,- ... .:: ' . ..... .:.:.:.: . . .......:. ., .": : : :? ''' : - . ." 7 4 '.. . :~T - ' ?. v : - :: :.~ ::- . ~ :" - -. .... ...... * Fi 'A15. The.~ church. .wita.......:;cornice... . . , '. . ' *'....... . . '; ; , , . 5 ......................... . ...... w..i .....,. ".. ....... g......n.... Fig 15. The church with a cornice at Ehne~ (Giimiigiin) Fig 14. Piers for a bridge in a ravine 2km northwest of Zeugma Fig 16. A hairpin bend in the road to Rumkale 32 Comfort and Erge? Until creation of the reservoir in 2000, a relatively good, if unpaved, modem road followed the east bank of the Euphrates all the way from Tilmusa (Apamea) up the gorges to Ayni- a distance of about 40km- and this may of course have been built on top of an ancient route. There was no direct evidence of this, but the landscape along the valley seemed to have changed less than along the west bank and an ancient road may therefore have been visible for much of the valley's course until improved by the bulldozers in recent years. Two kilometres before Ayni a substantial watchtower of rough stones without mortar was found above the road and on the right (fig 12). This was- surprisingly - the only visible indication of military activity along the valley between Zeugma and Ayni on either bank, but its construction was impossible to date and it may well have been medieval or even more recent. Traces of 'Achaemenid/Hellenistic occupation' are mentioned by Stein (1999: 119) at Hacinebi, 9km east of Zeugma on a prominent outcrop above the river at the point where it bends south to Birecik (and previously the site of a large Uruk settlement). However, the only firm evidence of a fortification on the east bank around Zeugma - other than the Hellenistic walls of Apamea was found in 1996 at the site of the Birecik dam itself, on a bluff, again overlooking the river but just 500m east of the small village of Tilobuir, directly opposite Belkis. This site was partly eroded, but traces were visible of about 40 houses, in one case with a smooth column drum of apparently late Roman date, a date confirmed by scraps of surface potsherds. On the cliff itself there was a section of wall with both faces visible but only one course of masonry visible. This course was of a double row of large rough blocks, with the smoother sides facing the outside; about 12m remained parallel to the river and a comer on the southeast end and was followed by approximately 4m of similar walling leading towards the cliff edge (fig 13). The wall then disappeared, apparently because of natural erosion. This building was provi- sionally interpreted as a small fort, but access was soon cut because of the dam construction and the houses, if not the wall, have since been removed by bulldozer before they could be properly investigated. The Roman frontier road? By contrast, the road north of Zeugma along the west bank, attested in the Peutinger Table, seems likely to have been covered by several metres of colluvium for its course from Zeugma as far as Ehne, (modem name: Gumufiyguin; fig 1). This village, 16km north of Zeugma, is the location of the well-known Roman quarries, first discussed in detail by Cumont (1917: 151-66). The only names known from antiquity for places on this bank are Urima, discussed elsewhere in this article, and Arulis. According to the Peutinger Table, Arulis was 24 miles from Zeugma and a further 24 miles from the bridge over the Singas river (usually identified with the G6ksu - but there are many inaccuracies in the Table's indications of distance). The identification of Arulis has been contested. Ehne?, Rumkale and Elif have all been selected by different specialists. Dussaud and Wagner chose Ehne3 (1927: 450; 1976: Karte 1); French (map in French forthcoming) chose Elif; Honigmann (1923: 165(76)) and Miller (1916: 758) chose Rumkale. The Goksu is about 85km from Zeugma, following the routes seen here as the most likely course of the Roman road along the Euphrates; a point half-way would indicate Elif. Fortifications along the west bank of the Euphrates are notable for their absence: Dabrowa (1997: 109) comments on the recently discovered documents (Feissel, Gascou 1989: document 11), ...which indicate that, at least in the third century AD, specially trained and equipped soldiers of Leg XVI Flavia had regular military service on the Euphrates, at many posts along the river. In fact, the papyrus document concerned, which was found near Dura-Europos, refers to the sale of a boat in poor condition, lying in the river. The boat was sold in AD 232 for 75 denarii by Aurelius Corbulo, a pilot ('gubernator') of Legio XVI Flavia Firma, then based at Samosata. No mention is made of 'posts along the river', although possibly the castles at Kaleboyu and Bozyazi, mentioned above, were constructed in this period. There are no other such fortifications to be seen today on the right (west) bank, although Commandant Marmier (1890: 535) stated that he found, apparently in the 1880s, ... une plateforme rectangulaire couronnant un massif artificiel en pierre et terre, ayant 9 metres sur 20 metres, et etablie au bord meme de l'Euphrate, au pied de la falaise calcaire. From the indications given and a reference to Chesney's map, this platform seems to have been subse- quently the site of a modern villa on a natural outcrop south of the village of Eren (see below). The surroundings of this site, indicated by Wagner (1976: Karte 1) as a h6yuk called ?eyit, were investigated by us but there were only a few scraps of pottery to indicate early occupation. A couple of kilometres to the south, a watchtower indicated by Wagner (1983: map at fig 8.1) on a hill north of Horum near Yukan Cardak was not found, although there were traces, below and near the river, of founda- 33 Anatolian Studies 2001 tions of an ancient building, which may have been a roadside watchtower (Algaze 1994: 28, site 4, Pinar Tarlasi). These foundations of walls could be followed along the nearby modem track leading over the hill to Yukan (ardak, but were not mentioned by Algaze. Similarly, 4.5km east of Zeugma, on Tilhane Tepe (Algaze 1994: 42, site 37, opposite Tilves Hoyiik; see Comfort et al 2000: fig 8) traces were found of a substantial wall on the side of the hill facing the river, which might be evidence of a further fort, possibly in this case guarding the road ascending to the plateau. In the absence of proven forts throughout the valley, it is difficult to imagine a defensive system being organised by the Romans on the right (west) bank. The idea of a permanent Roman fleet on the Euphrates, as on the Rhine or Danube (Samowski, Trynkowski 1983), must also be discounted, if only because of the strong current which impedes upstream traffic. At two points north of Zeugma in ravines descending to the Euphrates from the west, bridge piers were found: in one case cut blocks still visible as the footings of a bridge were close to the existing road, but on a different alignment (see photograph in Gregory 1996: 178); in a second, deeper ravine, the piers themselves were still visible in both banks of the ravine 100m east of the modem road. They rose from a point about 2m above the floor of the ravine for a height of a further 3m and up to a level about 1m below the modem surface of the fields (fig 14). An ancient stone capital was seen in front of a small agricultural building 50m to the south. Although one of these piers had apparently been recently reinforced with concrete, they were not in use for a bridge and had been constructed long before of large rounded pebbles and not from cut stone; no date could be given for the period of construction. The ancient road to the north would have passed by the mound at Horum, thought to be the site of Antioch- on-the-Euphrates, a town mentioned by Pliny (NH 5.86) whose coins have been published (Wroth 1899: 49, 113). Jones assumes that this was the Greek name of Urima (1937: 452, n 30). Urima itself is mentioned by Ptolemy (5.14) and if indeed the identification with Horum Hoyuiik is correct there is a remarkable conti- nuity of names. Although occasional fragments of cut stone were seen by the excavators in the fields around the mound of Horum, with remains of the Chalcolithic era and subsequent periods, almost nothing was discovered about the Greek and Roman town. If the cut stones below water at the edge of the Euphrates were indeed a quay then presumably this town had a busy trade. But since arable land is scarce, exports are unlikely to have included grain, while timber seems also most unlikely to have survived the Bronze Age on the surrounding hills in any great quantity. No mines or quarries are known in the vicinity, other than the ones on Fakir Dag described below. The coinage may indicate a city of some wealth and it seems likely that, like 'Epiphaneia' believed to lie opposite, it received the name of Antioch from Antiochus IV Epiphanes, ruler of the Seleucid Empire from 175 to 164 BC, who, according to Rostovtzeff (1941: 703), transformed many 'oriental' towns into Greek poleis bearing this dynastic name. It is nevertheless surprising that an important city could have developed so close to Zeugma. Around Horum irrigation using the stream descending from the villages of (ardak had created in recent years a densely cultivated series of terraces with fruit trees, presumably thus reproducing a landscape similar to that of earlier cities here. To the north the road led firstly to the village of Eren (formerly Kahtin, fig 1) where Cumont reported rock tombs and a decorated fragment of marble (1917: 152). No such traces of antiquity could be found on the occasion of our visit. Continuing northwards, the next place of any size is the town of Ehne?, mentioned above. As already indicated this is one candidate for ancient Arulis (Aroudis in Ptolemy), but there is no agreement on this point. Apart from some ancient tombs and the inscrip- tions in the quarries, there are remains of three churches, some rock-cut tombs and several fine houses, apparently from the 19th century. The church by the road to Burnus (and on to Gaziantep) is mentioned in Algaze's survey of the valley without comment, although it is one of the most interesting monuments of the area. According to Cumont (1917: 152) this church was dedicated to Saint Sergius. It contains inscriptions in Syriac and also some interesting architectural fragments, which appear to have been re-used from an earlier church or temple on this spot. A detailed study of the church and of most, but not all, of the inscriptions has been carried out by Palmer (1993a). A further publication is in preparation by Alain Desreumaux, which will analyse an additional inscription discovered in 1998. A second church at the north end of the village, also mentioned by Cumont, still exists, partly in caves and with a large number of crosses and niches, some of which give an impression of great antiquity. Cumont mentions the possibility of a temple of Silvanus, but one niche seemed to have had carvings removed and therefore was possibly the site of a Mithraeum. A third church in the centre of the village, perhaps that which was still standing at the time of Cumont's visit, is now recog- nisable only through a large arched opening carved into the natural rock, with a decorated cornice (fig 15). 34 Comfort and Ergec After Ehne? it was possible to continue by jeep along the river as far as Kalemeydan, the village opposite Halfeti, and there was evidence of ancient occupation in the form of tombs and substantial rock-cut houses at Kami?li, an attractive village perched above the river (fig 1). But the difficulty and poor condition of the modem track, which had to mount very steeply the full height of the cliffs above Kami?li, indicate that this route was unlikely to have been followed by the Roman frontier road. This probably crossed the plateau further 'inland', a course which would also have cut across a substantial bend in the river. The shorter distance would presumably have been considered by caravans and military convoys as ample compensation for the fairly gentle climb of 300m from river to plateau and then back down into the depression of the river Merzumen towards Rumkale. Two alternative routes which led north from Ehne? to K6seler, the village on the top of the plateau, were inves- tigated with the help of the KVR-1000 satellite photo- graph. The first left Ehne? along the modem road to the west and passes the first of the churches mentioned above, a remarkable free-standing building about lkm from the village and high enough not to have been affected by the reservoir. Although ruined, it was used as a place of worship until approximately 1980. Soon after, the ancient road would have left the modem one and continued northwest to the small village of Bumus (fig 1). A large underground hall opposite the village and in a ruinous state may also have been a mill, although small pits in the floor of an adjacent room, together with a niche and 'trough' could have had some religious significance. A powerful spring emerges here that constituted the main water supply of Ehne?, to which it was carried by an apparently ancient underground pipe (as revealed by the masonry around the point of its arrival below the village). Up the small valley from Bumus a route along the limestone valley bottom would have been passable for carts and arrives after 600m at a point where there is evidence of substantial ancient quarrying. There are a few tombs and cave houses in the hillside above the stream and, on the top of the escarpment to the north, there is a large abandoned settlement known to local people only as 'Harabe' (ruins). Remains of about 40 houses of basalt blocks were found but there was no indication in the form of potsherds to date the settlement. Although the route up the escarpment was not investi- gated, there was a substantial modem track visible; over the plateau itself from 'Harabe' and heading north to the modem village of K6seler, 2km distant, villagers showed us an ancient roadway, which was still referred to here as 'Top Yol', indicating that the southern terminus for this route may have been Zeugma. The paving stones had been removed but some were still visible in field-walls. The second route to Rumkale would have started heading north from Ehne? before turning west up a long valley known as Serkiz Dere. Beyond the quarries to the north of the village and about 50m above the river several rock-cut tombs were found. Although much overgrown, Serkiz Dere also showed evidence of occupation, in the form of abandoned houses cut into the rock - probably in antiquity. After lkm a small side-valley to the southwest provided an ample flow of water even in summer; there was evidence of ancient occupation in the form of several presses, a tomb and, possibly, libation pits, all situated above a small flat greensward near the point where the valleys joined. A narrow, ancient track led steeply up from this point to the plateau above, but this was not investigated. From the junction of valleys, the main route continued up the Serkiz Dere to the northwest along a gently sloping path towards K6seler, still in places well-preserved and wide enough for carts. At a point about two thirds of the way to the top, underground cisterns had been carved into the natural stone wall next to the path, fed apparently by artificial runnels to attract rainwater. Local people stated that this road had been in use in living memory as a long-distance route from K6seler to Nizip (presumably passing near Belkis). On the plateau the route leading west towards Doliche and Gaziantep passes a large number of abandoned villages made of basalt boulders, which are described briefly in Comfort et al (2000: 121). Koseler itself seems likely to have ancient origins since, like other similar settlements now abandoned, it is built largely of basalt and has a pond excavated into the layer of basalt boulders which provides water well into the summer. Continuing north from K6seler and down into the Merzumen valley (at whose junction with the Euphrates lies Rumkale, figs 1-2), much more evidence was found of an important ancient route (mentioned in Comfort et al 2000: 115). Several cisterns - now filled in - were found on the outskirts of the village to the north and there were alignments of stones, similar to those mentioned above on the cliff opposite Belkis, that may indicate a former fortifi- cation. The route to the escarpment and continuing on towards Rumkale is clearly visible on the satellite photograph (fig 17). From the lip of the escarpment the road itself had evidently not been used for a very long time but was massively built up on the down-hill side and included two sharp bends. The width of the road at the first turn was 3.10m with the height of the embankment about 2m both below and above the turn. The built-up stretches were visible on the ground for approximately 350m (fig 16). 35 Anatolian Studies 2001 EUPHRATES Merzumen val Fig 17. The road to Rumkale (extractfrom KVR-1000 photo of 22 May 1992) 36 Comfort and Erge? Although the track is less clear as it approaches the river Merzumen, we were informed that there had previ- ously been two bridges. Crossing at the time of our visit was by stepping stones, apparently created from the rubble of one such bridge. The south and west bank of the Merzumen is precipitous and impassable as it approaches the castle of Rumkale. Two aqueducts have been carved into this bank, respectively 2 and 8m above the Merzumen, and a third follows the north bank around to the former settlement of Kasaba, opposite the castle and near the piers of a Roman bridge and the junction with the Euphrates. This village was abandoned earlier this century and two new villages constructed above. Below the castle and at the point where the Merzumen turns north before joining the Euphrates, a large underground house was investigated on the south side of the river. A room with a pointed arch and an inscription- apparently in Armenian, possibly a shrine -was found just above the upper of the two aqueducts. An opening nearby led up a staircase to several decorated rooms cut into the rock. One of these contained finely-carved inscriptions in Arabic script and a rectangular window with a fine view of the castle. In the same room local people showed us graffiti, which apparently represented the river Euphrates, with several animals and at least one masted ship (a 'muhele'? - see Ritter 1919: 122). A copy of one of these graffiti is shown in fig 18. Evidently, this drawing could be of significance for the history of navigation on the river, but it seems probable, given the fact that the Arabic inscription would almost certainly have preceded the graffiti, that the drawing was made fairly recently and possibly reproduced from an image seen elsewhere, perhaps in a schoolbook. On the wall opposite, graffiti of a minaret could be discerned with difficulty. The Roman bridge, whose remains were discovered by Jorg Wagner (1977a) across the Merzumen at a point 150m from its junction with the Euphrates, may have served only an earlier fortification on the site of the medieval castle, of which there is now no trace. It seems unlikely that a Roman road passed along the Euphrates to the south of Rumkale, because the path is narrow and subject to flooding. However, immediately below the castle, a stretch of about 30m had been excavated from the rock at a level used by the modem footpath and just below the aqueduct (fig 19). The latter emerges near this point through a tunnel dug through the neck of the mountain before turning south and continuing along the river (origi- nally about 15m below) to Kalemeydan, opposite Halfeti and 3km to the south (fig 1; Algaze et al 1994: site 1). The aqueduct is partly covered; underground sections have regular access points, which indicate Roman construction. It was in operation until 1997 (Comfort et al 2000: 115). Despite the aqueduct and excavated section under the castle itself, large boulders partly cut away (in antiquity?) for a mule-track (fig 20) at a point about 50m south of the castle and the narrowness of the passage between cliffs and river seemed to rule out a Roman road along the river itself. At one point about 1.5km north of Kalemeydan a tunnel had been begun but not completed at a level about 10m above the river. It seems possible therefore that Roman engineers began to construct a river road but abandoned it as being too difficult or too subject to flooding. North of Rumkale remains of ancient roads were found at three different points, as indicated on the map (fig 2). One, below the village of Begendik, was a short stretch of narrow, paved track, about 1.50m wide. In the other two cases ruts and some embankment on the downhill side were the only evidence. Although one route seems to have initially followed the Euphrates, this one too appears to have steadily climbed and eventually to have left the valley, crossing towards the ancient abandoned settlement of Giimriikkiiyu, 2km west of the modem village of Sanlar (fig 2). Tombs, quarries, houses and some decorated, carved stones, apparently from a temple pediment, indicate that this point was of substantial importance. It lay near the road from Doliche to Samosata, clearly visible on satellite photographs, and was also close to the temple above on the bare mountain ridge at K6sk (see Comfort et al 2000: 117). A road, possibly of ancient origin, leads from Sanlar to the river bank opposite Ayni, but north-south passage along this stretch of river seems impossible along the west bank because of the cliffs, in particular that associated with the 'cable-car' suggested by Comfort et al (2000: 116). It is for this reason that the crossing point of Ayni seems to have been especially important. The traffic heading north along the west bank would therefore have left Rumkale by one of several possible routes converging on Guimriikkuyu and then have followed the Doliche-Samosata road over the eastern extension of the bare limestone range today called Kara Dag. At the highest point of this road a large area of arable land was the site of at least five important villages in antiquity. Three of these still exist and have substantial Roman mausoleums to prove their former status: Hasanoglu, Elif and Hisar. For details of other finds in this area, see Comfort et al 2000. From Kara Dag, the Roman road to the north crossed the next river flowing into the Euphrates from the west - the Kara Su (or Araban Cay)- by the Roman bridge near Siipiirgiic (Akbudak) and then linked up with the west-east road from Germaniceia (Maran) and with the road rising from the other Roman bridge at the junction of the Kara Su with the Euphrates (at Habe~). 37 Anatolian Studies 2001 Fig 18. Graffiti in a cave house beneath Rumkale (right- hand end of longer picture apparently representing the river Euphrates; now under water; boat is 19cm long and 17.5cm tall) Fig 19. The track beneath Rumkale; the 'excavated' passage Fig 20. The track between Rumkale and Kalemeydan, Fig 21. Tomb at Turus (Kuyulu), showing the decorative showing narrow path rim around the doorway 38 Comfort and Erge9 '~ . ? . :? : : '!i.:; .. -.. ... __l: ' - - _ >~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~', 39 9- T %mi .
. . rIl i ..... r \l to EZI Anatolian Studies 2001 After 20km heading east towards Samosata this road in turn crosses another important tributary of the Euphrates, the G6ksu. Arches of a fifth impressive Roman bridge are still standing here at a point 3km north of the Euphrates, although it is no longer possible to cross the bridge. The road continues north from here to the Roman tower at Burq, discussed in Blaylock et al (1990: 122). Shortly after, the road reaches Turus or Tarsa, mentioned below and about 30km from Samosata. The latter town, winter capital of the kingdom of Commagene and later base of Legio III Gallica, is of course beneath the waters of the Atatiirk dam. Quarry roads Kennedy (1998b: 57) discusses a road to the quarries found by Christopher Lightfoot in 1992 on Kalazan Dag, a table mountain on the east bank 5km north of Tilmusa (Apamea), whose highest point is 696m (fig 1). These quarries are one of the few ancient features of the landscape around Zeugma, other than mounds and roads, which appear clearly on the satellite photographs (fig 22). The ruts of the approach road are evenly spaced at a distance of 120-5cm and there are frequent cross channels dug into the stone for braking the heavily laden vehicles in their descent from the quarry. At the sharpest turn, very close to the point from which the photograph of this road published by Kennedy (1998b: fig 3.33) was taken, there is a relief statue of Hercules holding a club (fig 23). This relief seems not to have been seen by either Lightfoot or Kennedy and is more visible at certain times of day when the light is coming from the west. The question of the direction taken by this road after it descends the southern slope of the mountain was investigated; it appears to cross the plateau towards Apamea and not to take the shortest, but steep, route straight down to the river, opposite Horum. However, once it leaves the limestone of the mountainside, the road's course is soon lost in the rich agricultural land of the Hobab Ovasi. Other features of the limestone plateau at the top of the table mountain of Kalazan Dag include the 'recti- linear outpost' mentioned by Kennedy (1998b: 57-9), whose size is 7.30 x 23.50m, with a maximum height of wall still standing at 1.40m. Its position on the top of the steep hill opposite Horum might indicate a hermitage or shrine from the late Roman period, but the shape of this building does indeed seem more reminiscent of a barracks or other military structure. There is a small round cistern nearby, a square one 100m east, a large round one at the quarry and at least two more on this plateau. On top of Belkis Tepe, about 8km away to the south and assumed to be the acropolis of Zeugma, three of the cisterns visible are also round and one is square; square cisterns are usually shallow and left open, while round ones in this area sometimes have steps cut into the rock from one side, leading down into the cistern which may be as much a 5m deep, for example, that at the roadside between Elif and Sanlar. It is possible that round cisterns were constructed in the Hellenistic period or before, while rectangular and square cisterns are Roman or Byzantine, but this is not yet possible to prove. In some cases ancient cisterns have been intentionally filled in, presumably to stop livestock falling in. Opposite Kalazan Dag, 5km to the west and slightly higher (747m), lies Fakir Dag, another table mountain whose neolithic remains are discussed by Bourguignon and Kuzucuoglu (1999). Just behind the highest point, visible from Zeugma, and along the top of a limestone escarpment, a series of eight quarry 'workshops' or separate small quarries, each approximately 30m square, were found. There was no road leading to these quarries and a wooden chute down the mountainside towards Zeugma is therefore hypothesised. These quarries seemed of earlier date than those on Kalazan Dag and may therefore represent the source of stone for Hellenistic structures at Zeugma and possibly of Apamea too. Fig 23. The Hercules relief on the quarry road descending from Kalazan Dag 40 Comfort and Ergeq Other routes above the east bank Caravans passing from south to north were not confined to the valley itself. Some routes were investigated which passed over the plateau and above the valley on either bank. A Roman road from Zeugma to Samosata, which cut across the northwest comer of Osrhoene, is also known to have existed (Wagner 1983) and its possible course was studied on the available photographs from the US Corona series. Wagner identified stretches of Roman road near Ank (now Ye?ilozen) and at Eski Hisar. This substantial fort, on a hilltop 22km east of Halfeti, was also identified on the satellite photographs and the nearby village of Giirkuyu (formerly Nuhrut) is said to contain substantial remains of an ancient church. Ten kilometres eastnortheast of Guirkuyu lies Uzunburc, a village with many ancient cut stones, particularly around three well-heads still in use, and with some fragments of floral decoration, apparently from a substantial Roman building. Unfortunately the tower after which the village is named was destroyed a few years ago to provide building material for a mosque. It seems probable from the satellite photographs that the Roman road passed from Ank to Nuhrut and Uzunburc and then either to the small town of Yaylak, where there is a large ancient mound now covered by modem buildings and then to a point opposite Samosata or else direct to the Euphrates crossings at Nehriseid or Adilpazar, before proceeding along the north bank of the Euphrates via Turns. The existence of other north-south routes across Osrhoene from Suiruc (ancient Batnae) to Samosata may perhaps be deduced from the large caravanserai at Charmelik (now Biiyiikhan, see Durukan 1999: 318-27), which is situated on one of the main east-west routes from Zeugma to Edessa, as well as from the enormous ancient quarries 2km northeast of Kural (near the regional centre of Kanliav?ar). The remains of another ancient church and underground dovecote at Uckilise and of a Roman temple and fortification at Uzunburc are further evidence of substantial occupation of this region of Osrhoene in antiquity, apparently part of the diocese of Suiiric in the Christian era. However, much work remains to be done on surveying what is still a comparatively unknown part of $anliurfa province. The large Roman or Hellenistic granary at Kantarma (Comfort et al 2000: fig 12) has also been studied in Durukan (1999: 268-9). This building may indicate that the northwestern area of Osrhoene, apparently known as 'the White Region' in late antiquity (Palmer 1993b: 78, 267), was of substantial agricultural importance and that its population and wealth were correspondingly large. Since the Euphrates as it descends the reaches above Ayni curves from a westerly to a southerly direction, caravans following south-north routes on the plateau above the east bank were obliged to cross the Euphrates. Crossing points for east-west routes were covered in Comfort et al 2000, but at least three other points along this northern stretch of the river were also investigated because of the impact of the Birecik dam (fig 3). The level of the water will of course have risen much less at these points than at Belkis but still sufficiently in some cases to have removed any trace of the crossing point and associated remains. That near Killik at Kelefis was in use until quite recently and was said by local people to have been the crossing used by the Ottoman army retreating from Yemen in 1918. We were shown a former cistern south of the crossing point on a caravan route; caravans presumably came from Halfeti and the south via Yukan Goklii or the important centre of Kayalar (formerly called Keferhan or Kefren). They would have continued on the northern side of the Euphrates through the gap in the mountain chain (Kizil Dag) opposite the village of Dikilita?. On the summit of the mountain nearby is the important tumulus of Sesonk ('Three stones' in Kurdish, fig 3), a royal burial place for the dynasty of Commagene, discussed briefly in Wagner (2000: 23). This tumulus is not accessible by road; a very rough jeep track was followed from a point to the northwest for about 2km and thereafter a difficult walk of 40 minutes is necessary to reach the tumulus. Some of the sculptures on pillars illustrated by Humann and Puchstein (1890: 212-7) had fallen, but a chamber underneath the mound was investigated on the north side, which had been blocked on the occasion of this earlier visit. A first chamber, 4m square with a simple carved linear decoration, gave access to a passage 15m long. At the end there was a substantial pit, on a slightly different alignment from the tunnel and about 1.50m deep, apparently intended for a coffin. A deep vertical and rectangular pit on the east side of the mound, also mentioned by Humann, could not be explored without ropes, but there were signs of handholds carved into the side of the pit. To the northwest there is a fairly easy route from the low pass in the Kizil Dag, west of Sesonk, to the former monastery and medieval castle of Keysun and to the town of Besni, an important centre in the Seljuk period and believed to have formerly been the Roman waystation of Octacuscum. French's map shows this route as a Roman road which leaves the Doliche- Samosata road at a junction near Dikilita?. To the northeast, across the G6ksu, it would have been possible to reach Perre and Malatya. Crossings were also evident at Nehriseid and at Adilpazar. The former would appear to have been used by caravans or military convoys taking a 'short-cut' from 41 Anatolian Studies 2001 Zeugma to Samosata by a route across Osrhoene via the fort at Eski Hisar (discovered by Guyer 1939) and the settlement of Uzunbur? (see above). Thereafter cliffs along the south bank of the Euphrates that fall vertically into the river make passage along the bank impossible for several kilometres. Caravans crossing at Nehriseid or Adilpazar would probably have continued on the north bank via Turns or Tarsa, a town mentioned in the Peutinger Table as lying 19 miles before Samosata. More than 40 rock-cut tombs were counted in a necropolis, probably that of Tarsa, 5km after the modem bridge over the Euphrates and then, not far from the village of Kuyulu (fig 3), to the left of the modem road leading from the Atatiirk dam northnorthwest towards Adiyaman. Many of these tombs have a curious decorative rim around the entrance (fig 21), which distinguishes them from similar tombs in the neigh- bourhood of Zeugma. The site is signposted on the main road as a point of interest for tourists ('Turns Kaya Mezalari'). Other routes above the west bank South of Zeugma If the road along the Euphrates leading north from Zeugma was of military significance, this is probably also true of the route south along the Euphrates, even if commercial caravans normally travelled east from Zeugma to Edessa (Urfa) or Harran before turning south. The Peutinger Table does indeed indicate such a road along the river as far as the Euphrates bend at Barbalissus, 120km south of Zeugma, although no such route is mentioned in the Antonine Itinerary. A road south from Zeugma would have reached the Hellenistic foundation of Europus, probably at Carchemish which is 28km southsoutheast of Belkis, via the large h6yiik of Guinalti or Tilmiyan (in turn, 6.5km southeast of Belkis). Such mounds are clearly visible on the KVR-1000 satellite photographs. In contrast to the region north of Zeugma, the Euphrates passes through relatively flat or undulating countryside, although the settlements along the west bank are often located on cliffs or hills 50m or more above the river. Arable crops are grown between Aleppo and the Euphrates today but rainfall is unreliable and the harvest is presumably uncertain. The lake created by the Carchemish dam, now also completed, will not affect this route that is some distance away from the river. There may, however, also have been ancient roads along both banks of the Euphrates, despite the marshes along the west bank, since there are many mounds on the river banks. On the west bank that at Seraga is particularly notable, but there are several which will have been partly or wholly inundated along the east bank. Most of the latter have been excavated to some extent since 1997 in the course of a project directed by Professor Numan Tuna of Middle East Technical University (Tuna, Ozturk 2000/2001). The site at Zeytin Bahseli, 6km south of Birecik, has been excavated by Dr Frangipane; it has a high conical mound, apparently an Iron Age monument, built on top of the earlier Bronze Age hoyuk and commands fine views over the flood- plain of the Euphrates. But these mounds, listed in Algaze's survey of 1989 (Algaze et al 1994), which were active centres of commerce from the Chalcolithic, may have been linked more by the river than by roads. Identifications of modem settlements with sites along the west bank mentioned in the Peutinger Table are made by its editor, Miller (1916); the Table does not mention Europus/Carchemish but shows a wayside stop with a villa symbol at a point supposedly 24 miles from Zeugma. Unfortunately the KVR- 1000 photographs -the most detailed of the satellite imagery available for the project did not cover the region to the south much beyond the mound of Giinalti, but the course of the Roman road is clear. It headed southeast towards the riverbank, which it joined near the h6yiik of Elifoglu (10km south of the Birecik bridge); thereafter it is assumed to have followed the modem track along low hills west of the river to Carchemish on the current border with Syria. (Prelim- inary investigations were made of the first part of this route also south of the Syrian border - without the help of detailed maps or satellite photos.) Ptolemy, writing at Alexandria in the second century AD, lists the sites south of Zeugma as Europos, Caeciliana, Bethamaria, Serre, Arimara and Eragiza (Geographia V.15, ed Nobbe 1966: 61). An early discussion of the route from here to the south, as well as the routes to Antioch and Edessa, is contained in Regling (1901: 469-76). This article concerns only the right (west) bank below Birecik. For a description of the route along the east bank from the Sajur/Euphrates junction see Bell (1910). The next point south of Europus mentioned in the Peutinger Table is Caeciliana, supposedly 16 miles distant. This is probably correctly assumed by Dussaud (1927: 450) to lie near the former Euphrates crossing at the mouth of the river Sajur. Sresat (fig 4), believed to be Caeciliana by Ainsworth (1888: 224), is a village on cliffs 3km north of the junction of the two rivers and would therefore be a more likely site for Caeciliana than a point further to the south as was thought, for example, by Miller (1916: 758) and others (see below). Traces of ancient buildings and some tombs were visible at Sresat, but the modern village lies only about 14km from Carchemish (Gaborit: forthcoming). Mounds between Carchemish and Sresat include Tell Alariyeh (or Tell Djerablous Tahtani) and Tell Amara; excavations by the University of Edinburgh led by Prof E Peltenburg at Tell 42 Comfort and Ergeq Djerablous Tahtani and by a Belgian team at Tell Amama the latter will shortly be re-opened - have revealed important Hellenistic and Roman levels. Trenches for irrigation pipes had also uncovered substantial ancient walls built of large limestone blocks at the time of our visit. The village 2km south of Sresat at Mokar Mazar had at least five column capitals lying in the yards of village houses, as well as a fine hexagonal Roman tomb cut into the rock. This village may therefore be considered another possible site for Caeciliana. From Caeciliana the Peutinger Table shows a route to Hierapolis, now Membij (fig 4) and also known in the ancient world as Bambyke; this route seems to run parallel to that between Zeugma and Hierapolis, although the latter lies beyond Sresat. Possibly, there was a confusion here. For both routes a distance of 24 miles is indicated, although the direct distances are 60km from Zeugma to Membij and 25km from Sresat to Membij. The good modem road from Jerablus (Carchemish) to Membij leaves the route heading south to Sresat and the mouth of the Sajur after mounting some low hills above the river, near the mound of Amama. Membij today boasts a number of sculptured stones in a small park: two reliefs with eagles (figs 24-5), a ceiling coffer relief with a mythological scene (fig 26), a statue of a lion (fig 27) and three seated statues (figs 28-30). Their source is unknown and there are no other traces of its early impor- tance. The city walls and the relief of two sirens with a naked woman, mentioned by Sachau (1883: 147-9) as almost the only traces of antiquity in his day, have also disappeared. (The site is also described by Hogarth, who visited it in 1908.) Continuing south, the Sajur valley joins the Euphrates from the northwest at a point nearly opposite Tell Ahmar on the east bank. The area was first described by Hogarth in 1909 and was surveyed by Moore in 1977 (Sanlaville 1985: 41-66). The reservoir created by the Teshrin dam, now completed and located some 25km southeast of Membij, reaches this point and was rapidly filling on the occasion of our visit in August 1999. Although Tell Ahmar or Til Barsip, the site of an Assyrian palace and river port (Thureau-Mangin 1936; Mallowan 1937), will certainly have been drowned, the west bank of the Euphrates is higher along this stretch and ancient settlements along the cliffs will be unaffected, although some rock tombs in these cliffs may now be under water. The mound of Tell al'Abr, an Ubaid site on the east bank about 2km north of the Sajur/Euphrates confluence, will now also have disap- peared (Hammadi, Koike 1992), although the very large mound at Tell Badaye, 6km north of Tel Ahmar, should have been preserved, together with those further north at Shiukh Tahtani and Shiukh Fogani (fig 4). The next point mentioned in the Peutinger Table on the west bank is Bettamali (presumably Ptolemy's 'Bethammaria') at a distance of 14 miles from Caeciliana. The site of Tell Shiukh Fogani (or Faouqani), recently excavated by a team from Syria, France and Italy (Bachelot 1996) is called Burmarina in Aramaic. But this cannot be connected with Bettamali because it lies on the east bank only 5km south of Carchemish. Miller (1916: 758) identifies Bettamali with ruins south of Qalat an-Najm (fig 4), but these were not found by us. Bachelot (1996: 35) includes a map of rescue excavations in connection with the Teshrin dam, which shows that a site has been excavated near Qalat an-Najm at Tell Djurun el-Kebir. The castle at Qalat an- Najm is said to be originally Umayyad (Ball 1994: 160) and built to guard a bridge, of which some remnants may be visible on the west bank. The present castle ruins are not earlier than the 11th century, but were nevertheless identified by Chapot with Caeciliana (1907: 281). Serre and Apammari, the following stops on the Peutinger Table at distances indicated as 13 and 8 miles, were not found by us either, although Miller (1916: 758) believed that Serre may have been Ptolemy's 'Gerre' and gives the modem equivalent as Kara Bamb6dsch (Membij/Bambyke) - an identification accepted as probably correct by Dussaud (1927: 451). This is the old name of a location currently occupied by the Teshrin dam itself. A site northeast of 'Scheich Arad bei Djerram' was identified by Miller (1916: 758) with Apammari. The latter is probably the same as the Arimara of Ptolemy (placed by him between Serre and Eragiza) and thought by Dussaud to be the modem village of 'Khirbet es-Soude' (marked on modem maps as Kheurbet al Sauda, 5km from the Euphrates and 8km northwest of Qalat-an-Najm). The riverbank is hard to follow and is now in parts inaccessible because of the new reservoir. Just south of the Teshrin dam and 17km southsouthwest of Qalat an-Najm lies the Hellenistic fortified site of Jebel Khalid (Connor, Clarke 1997). The identification of this site with one of the places mentioned in the Table is unlikely because it does not appear to have been occupied after 100 BC. The importance of this area during the Hellenistic and Roman periods remains difficult to assess, but the many ruins found by Sachau, described in his account of a journey from Aleppo to Urfa (1883: 110-210), and in Poidebard's aerial surveys in the 1930s (1934; especially the map of 1943 in Mouterde, Poidebard 1945) seem to indicate a moister climate and cultivation in areas which are now wholly abandoned. Early excavations in the area gave little attention to layers of Hellenistic and Roman occupation, but these were clearly important at many sites and the military importance of the area seems to have provided a boost to its economic development (Clarke 1999). 43 Anatolian Studies 2001 Fig 24. Stele with eagle from the park in Membij Fig 26. Ceiling coffer (?) with mytho- logical relieffrom the park in Membij Fig 25. Stele with two eagles from the park in Membij Fig 27. Statue of lion from the park in Membij Figs 28-30. Seated figures from the park in Membij 44 Comfort and Ergeq Eragiza, the next stop on the Peutinger Table, is a site identified by Sachau (1883: 133-7) and Chapot (1907: 282) with 'Abu Hanaya', apparently 8km from the Euphrates. Here Chapot found, '...des debris de grandes dimensions; de formidables tambours de colonnes rappelant celles de Baalbeck'. Dussaud (1927: 452), however, believed Eragiza to be a separate site called 'Aroude' nearer the river (now under the water of Lake Assad created by the Tabqa dam and probably Tell el Hajj). He preferred to identify Abu Hanaya with Neocaesarea, another place mentioned by Procopius (De Aedificiis 2, 9). This latter town is however placed by modem specialists about 50km to the southeast, beyond Barbalissus (e.g. Wagner 1985: 13, map 'Die Ostgrenze des r6mischen Reiches von Augustus bis lustinian'). Dussaud's identification of Eragiza with Aroude is supported in reports of excavations at Tell el Hajj (Stucky 1973), but Stucky relied largely on the Peutinger Table distances to support this thesis and these distances would be about the same for Abu Hanaya. Poidebard's map of the region (Mouterde, Poidebard 1945), prepared on the basis of his aerial survey, shows Abu Hanaya only as a Roman 'locality' at Eragiza, but indicates Roman ruins, with a cistern and castrum, some 7km southsouthwest near a mound at 'Tell Mahdofim'. Although Abu Hanaya is not identified on modem maps and was not visited by us, there is an Al Mahdum on the modem main road to Raqqa 63km east of Aleppo, while on Kiepert's map of Sachau's journey in Syria (published with the latter's account) Abu Hanaja is shown about 25km southwest from what is now the north end of Lake Assad (and thus from the Teshrin dam and the nearby Jebel Khalid). Poidebard shows a Roman road heading due north direct from Abu Hanaya to Hierapolis/Membij, as well as the road hugging the river bank. It is assumed here that the Peutinger Table route followed the river along a road in the hills above to a point close to Jebel Khalid, turned inland for a few kilometres to Eragiza at Abu Hanaya and then continued southeast to Barbalissos/Meskene, formerly Emar and a site which is now beneath Lake Assad. This large artificial lake was created in 1973 by the Tabqa dam 40km west of Raqqa. Southwest of Zeugma North-south routes on the west bank, further away from the river Euphrates, were also investigated. In particular, attempts were made to identify the routes from Zeugma to Aleppo and, even further west, to Cyrrhus and Antioch. David French's map of Roman roads (French: forthcoming) was invaluable and use was also made of a survey of mounds in the late 1960s (Archi et al 1971). For this section please consult map at fig 8 in Comfort et al (2000: 109). Aleppo lies 110km southsouthwest of Zeugma; although as Beroea in the Hellenistic and Roman period it seems to have been relatively unimportant, it was of major significance in the Bronze Age. A trade route must have passed this way from early times if Zeugma was a major crossing in the pre-Hellenistic period, as argued by Gawlikowski (1996) and provi- sionally accepted in Comfort et al (2000), although Carchemish must have also had an important crossing. The Roman archives found near Dura-Europos (Feissel, Gascou 1989: document 16) contain a letter concerning purchase of camels for a caravan travelling between Beroea and Zeugma at some time in the third century AD. After the Byzantine defeat at the Yarmouk in AD 636 Aleppo again became one of the most important trading centres of the region; the Umayyad Great Mosque was built between 715 and 717, soon after that of Damascus. Aleppo was later the adminis- trative centre of the Ottoman empire for the region up to the Taurus, as well as to the Euphrates crossing at Birecik. Caravans coming from the east since the Middle Ages therefore had Aleppo as their prime desti- nation. Possible routes were examined on the satellite photos and field boundaries indicated that a long- distance route left the col of the tombs above Zeugma and headed southsouthwest via a number of mounds, including one at Yolcati where the ancient road crosses the railway. This route was investigated at several points, although it passes through a small reservoir called Hancagiz and so could not be followed in its entirety. The village of Kale on the south bank of this reservoir has no castle to show- despite its name but there were rock-tombs and ruins of at least one ancient building. The mound, or h6yiik, of Sariko9 is still the site of a village, whose inhabitants claimed that the last caravans passed this way in 1960. Roman pottery on this and several other mounds of the area was still visible. It was believed by Miller to correspond to the stop on the Peutinger Table called 'ad Zociandem' (Miller 1916: 768). An ancient well had been recently cleared in part and a coin with Arabic inscription found then is dated AD 1837 (AH 1255). The remainder of the route to Aleppo was off the available satellite photos but it would appear to have crossed the modern frontier near Elmali. The old road to Antioch was still in use in the 19th century. The plates and boilers for the two small steamships used by the Chesney expedition down the Euphrates in 1835 were brought by oxen along this road, but the course today is difficult to ascertain, even though Chapot found clear traces in the first years of the 20th century (1907: 339). 45 Anatolian Studies 2001 La voie principale de l'extreme-nord syrien etait 'a peu pres rectiligne d'Antioche a Zeugma. Entre Killis et Nisib, j'en ai observe les vestiges sur un parcours de plusieurs kilometres : elle etait large de 4 a 5 metres ; aucun pavage, mais une accumulation de gros cailloux et sur chaque rebord des pierres plus volumineuses.... Leaving from Zeugma, this road may either initially have followed the route towards Aleppo indicated above or else have passed via large mounds at Nizip and Hamius (near a settlement called today Bahcede) and then via the modem village of Uluyatir (formerly Mizar). Fifteen kilometres southwest of this village there is the large Bronze Age site and later Crusader castle of Tilbesar, clearly visible on a Corona satellite image (for reports of recent excavations, Kepinski-Lecomte et al 1996 and 1998; for its role as a crusader castle, Hellenkemper 1976). French's map indicates the Roman road as passing this way and identifies the 'ad Zociandem' mentioned above with Uluyatir and another way-station on the Peutinger Table, 'ad Secta', with Tilbesar. However, on the occasion of our visit no direct route was found on the ground, although a route is visible on the Corona photograph of 1968. The Peutinger Table indicates a distance of 12 miles between the two stations and a further 12 miles between ad Zociandem and Zeugma, which would match either Uluyatir or Sankoc. On the railway line that is the current border between Turkey and Syria, near the point where it crosses the valley of the river Sajur, lies the Achaemenid site of Dere Hiiyiik (Moorey 1980). This site could not be visited but must also be close to this route from Zeugma to Aleppo. The Hellenistic province in which Zeugma lay was called 'Cyrrhestike' after the city of Cyrrhus, now 63km northnortheast of Aleppo and just on the Syrian side of the border. The origins of this city remain obscure although the site has been investigated and partially excavated by Frezouls (1954/5). It was visited by us in summer 1999. Both the direct roads from Antioch to Doliche and to Zeugma would have left Cyrrhus off to the left, although the Peutinger Table does include Cyrrhus between Gindarus (now Jindairis) and 'Channuma', placed by David French (forthcoming) 15km northeast of Kilis (formerly Ciliza). This would have involved a detour of some 20km but Cyrrhus was clearly an important place for much of antiquity and there must indeed have been heavy traffic between it and both Zeugma and Doliche/Samosata. Roman bridges on the routes heading east from Cyrrhus are still extant and in use at two places; they are also mentioned in Maundrell's account of his journey in 1699 (Maundrell 1817; he refers to Cyrrhus as 'Corus'). Northwest of Zeugma The map in French (forthcoming) shows the course of the Roman road from Doliche to Arulis (identified by him with Elif) as proven for the segment east of Yavuzeli. Apart from the Roman bridge and watchtower at Yanmca (discussed in Comfort et al 2000: 117), there are long stretches of an older road alongside the modem road from Gaziantep to Adiyaman; southwest of the village of Buiyiikkarakuyu, this road, partly tarred but now replaced by a wider road in parallel, follows the supposed alignment of the Roman road and crosses at one point an old culvert which is possibly Roman (fig 31). The existence of large Roman bridges on the Karasu and G6ksu makes it clear that there was indeed an important route to Samosata from Doliche. The upper part of this route was first discussed in an interesting article by the Commandant Marmier (1890). Comfort et al 2000 drew attention to the large number of remains from antiquity in the area of Elif and Hasanoglu, the plateau at the eastern end of the Kara Dag which is situated on steep cliffs above the Euphrates. Some of these remains are also discussed above. The Roman road to Samosata passes over this plateau before descending to the Roman bridge on the Karasu, near the Hittite(?) relief and settlement of Suiipiirgfic. Immediately to the north of the village of Hasanoglu there is another road which crosses the Karasu south of the village of Karababa at a point 6km westsouthwest of the Roman bridge. This may have also been an important old road at some time since there are faint but clear ruins of a former bridge evident also here (fig 32). There must certainly have been another ancient route heading due north from Yavuzeli (formerly Cingife and also site of a Crusader castle), which lies in one of several rich valleys aligned east-west between arid plateaux or mountain ranges. The modem road north to Araban (formerly Altintas) climbs steeply from Yavuzeli to a pass at about 1000m on the top of the Kara Dag range, near a group of large dolmens one of which is well-preserved (fig 22). On the far side two older routes descend to the right of the modem road and the section beyond, heading east to Araban along the valley bottom, has been straightened, showing an older road at several points. Araban has a large mound that was also used as a Crusader castle, the third in this area (Hellenkemper 1976); some fragments of defensive wall are still visible and at the village of Fakill, 5km to the southeast, a very large quarry was discovered with a short Arabic inscription. But Araban itself has little now to show in terms of ancient buildings which would have justified such a large quarry. The neighbouring village of Eskialtintas has no antiquities at all. 46 Comfort and Ergeq Fig 31. A culvert 2km south of Biiyiikkarakuyu on the old road from Doliche to Samosata There was presumably an ancient road along the course of the modem road heading north from Araban to Besni (Octacuscum), which again passes through mountains and gorges that limit the possible alternative routes. The Seljuk remains of Besni are sadly neglected and little is known of this region's early history, except that it was a stronghold of Armenian princes in the Middle Ages (Dedeyan 1996). Discussion The routes and sites discussed here further attest to the wealth and high level of development of the area on both banks of the Euphrates around Zeugma, especially in the Roman period. The GAP (Project for Southeast Anatolia) and the construction of the Birecik dam have provided the occasion for the first examinations of the riches of this area in terms of ancient buildings and history (see also Durukan 1999). Much more remains to be done for the region's archaeology. It should be noted that in the course of this survey it became apparent that the remains of some ancient buildings had been destroyed comparatively recently. In particular, churches Fig 32. Remains of a bridge in the bed of the Kara Su, north of Hasanoglu (photo: Danl; Baykan) at Kayalar and Bozyazi, as well as a Roman tower at Uzunburc, have disappeared within the last ten years. More intensive and mechanised agricultural techniques are also contributing to the disappearance of archaeo- logical sites. It is imperative that, in addition to the work already published, a full inventory of all sites, ruins, reliefs and other artefacts still remaining from antiquity be carried out soon and that the sites mentioned here be more fully documented, where they have not already been destroyed by the waters of the dam. There are good prospects that a new emphasis on cultural tourism will increase awareness of the impor- tance of some of these sites, but dam construction is only one aspect of the process of modernisation which is threatening the survival of the region's cultural heritage. It is hoped that, together with the remarkable mosaics discovered at Zeugma itself, this article and its predecessor (Comfort et al 2000) will contribute to drawing attention to the archaeological wealth of those parts of Gaziantep and Urfa provinces adjoining the Euphrates and that more will now be done to study and preserve them. 47 Anatolian Studies 2001 Bibliography Ainsworth, W F 1888: A Personal Narrative of the Euphrates Expedition. London Algaze, G 1993b: The Uruk World System. Chicago Algaze, G, Breuninger, R, Knudstad, J 1994: 'The Tigris- Euphrates archaeological reconnaissance project: final report of the Birecik and Carchemish Dam survey area' Anatolica 20: 1-96 Archi, A, Pecorella, P E, Salvini, M 1971: Gaziantep e sua regione. Rome Bachelot, L 1996: 'Burmarina, decouverte d'une cite arameenne' Archeologia September: 34-9 Ball, W 1994: Syria - a Historical and Architectural Guide. Buckhurst Hill Bell, G J 1910: 'The east bank of the Euphrates' Geographical Journal 5: 513 Blaylock, S R, French D H, Summers G D 1990: 'The Adlyaman Survey: an interim report' Anatolian Studies 40: 81-135 Bourguignon, L, Kuzucuoglu, C 1999: 'Les occupations prehistoriques de la moyenne vallee de l'Euphrate au nord de Belkis (Turquie). Etudes geomor- phologiques et archeologiques: premiers resultats' Anatolia Antiqua 7: 265-83 Chapot, V 1907: La Frontiere de I'Euphrate de Pompee a la conquete arabe. Paris, reprinted 1967 Rome Chesney, R A 1850: The Expedition for the Survey of the Rivers Euphrates and Tigris (2 vols). London Clarke, G W 1999: 'The upper Euphrates valley during the Hellenistic-Roman period' in del Olmo Lete, Montero Fenollos (eds), Archaeology of the Upper Syrian Euphrates. the Tishrin dam area. Barcelona: 637-42 Comfort, A M, Abadie-Reynal, C, Ergec, R 2000: 'Crossing the Euphrates in antiquity: Zeugma seen from space' Anatolian Studies 50: 99-126 Connor, P J, Clarke, G W 1997: 'Jebel Khalid in north Syria: the first campaigns' Mediterranean Archae- ology 9/10: 151-83 Cumont, F 1917: Etudes syriennes. Paris Dabrowa, E 1997: 'The rivers in the defensive system of Roman Syria (from Augustus to Septimius Severus)' in W Groenman-van Waateringe et al (eds), Roman Frontier Studies 1995. Oxford 109-11 Dedeyan, G 1996: 'Les princes armeniens de l'Euphratese et les Francs' in R H Kevorkian (ed), Armenie entre Orient et Occident. Paris: 164-73 Dillemann, L 1962: Haute Mesopotamie orientale et pays adjacents. Paris Durukan, A (ed) 1999: Birecik, Halfeti, Suru?, Bozova Ilceleri ile Rumkale 'deki Ta;sinmaz Kiiltiur Varliklari. Ankara Dussaud, R 1927: Topographie historique de la Syrie antique et medievale. Paris Feissel, D, Gascou, J 1989: 'Documents d'archives romains inedits du Moyen Euphrate (IIIeme siecle apres J.-C.)' Comptes rendus de l'Academie des Inscriptions et Belles-Lettres (janvier-mars): 535-61 French, D 1983: 'New research on the Euphrates frontier: supplementary notes 1 and 2' in S Mitchell (ed), Armies and Frontiers in Roman and Byzantine Anatolia (BAR International Series 156). Oxford: 71-101 -Forthcoming: Roman Roads and Milestones in Asia Minor 9: Commagene and the Euphrates Road Frezouls, E 1954/5: 'Recherches historiques et archeologiques sur la ville de Cyrrhus' Annales archeologiques de Syrie. Damascus: 89-128 Gaborit, J, Leriche, P 1998: 'Geographie historique de la vallee du Moyen-Euphrate' in P Amaud, P Counillon (eds), Geographica Historica. Bordeaux: 167-200 Gawlikowski, M 1996: 'Thapsacus and Zeugma: the crossing of the Euphrates in Antiquity' Iraq 58: 123-33 Guyer, S 1916: 'Reisen in Mesopotamien - von Diyarbakir fiber Urfah nach Aintab' Petermanns Geographische Mitteilungen, 62 Gotha: 299-301 1939: 'Eski Hissar, ein r6misches Lagerkastell im Gebiet von Edessa' in Melanges syriens offerts a Rene Dussaud 1. Paris: 183-90 Hammade, H, Koike Y 1992: 'Syrian Archaeologicial Expedition in the Tishreen Dam basin: excavations at Tell al-'Abr 1990 and 1991' Damaszener Mitteilungen 6: 109-75 Hellenkemper, H 1976: Burgen der Kreuzritterzeit in der Grafschaft Edessa und im Konigreich Kleinar- menien (Geographica Historica 1). Bonn 1977: 'Der Limes am nordsyrischen Euphrat: Bericht zu einer archaologischen Landesaufnahme' in J Fitz (ed), Studien zu den Militdrgrenzen Roms 2. Koln: 461-71 Hogarth, D G 1908: 'Hierapolis Syriae' Annual of the British School at Athens 14: 183-96 1909: 'Carchemish and its neighbourhood' Liverpool Annals of Archaeology and Anthropology 2: 165-84 1914: Carchemish 1. London Honigmann, E 1923/1924: 'Historische Topographie von Nordsyrien im Altertum' Zeitschrift des Deutschen Paldstina-Vereins 46: 148-93 (A to H, with map at table VII) and 47: 1-64 (H to Z, with indices) 1932: 'Syria' in W Kroll, K Mittelhaus (eds), Paulys Real-Encyclopddie der classischen Altertumswis- senschaft 4A. Stuttgart: 1549-727 Humann, K, Puchstein, 0 1890: Reisen in Kleinasien und Nordsyrien (Die zweite Reise nach dem Nemrud- Dagh 1883). Berlin Jones, A H M 1971: The Cities of the Eastern Roman Provinces (2nd ed). Oxford 48 Comfort and Ergec Kennedy, D 1994: 'Zeugma, une ville antique sur l'Euphrate' Archeologia 306: 26-35 - 1998a: 'Declassified satellite photographs and archae- ology in the Middle East: case studies from Turkey' Antiquity 72: 553-61 1998b: The Twin Towns of Zeugma on the Euphrates (Journal of Roman Archaeology, supplementary series 27). Portsmouth, RI Kepinski-Lecomte, C, Gerard, F, Jean, E, Vallet, R 1996: 'Tilbeshar 1994, 1995' Anatolia Antiqua 4: 291-301 Kepinski-Lecomte, C, Erge9, R 1999: 'Tilbeshar 1997' Anatolia Antiqua 6: 337-42 Kuniholm, P I 1997: 'Wood' in E Meyers (ed), Archae- ology in the Near East 5. Oxford: 347-9 Mallowan, M E L 1937: 'The Syrian city of Til-Barsib' Antiquity 11,43: 328-39 Marmier, G, Le Commandant 1890: 'La route de Samosate au Zeugma' Bulletin de la Societe de Geographie de l'Est. Nancy: 518-35 Maundrell, H 1817: A Journey from Aleppo to Jerusalem at Easter AD 1697 to which is Added an Account of the Author s Journey to the Banks of the Euphrates at Beer and to the country of Mesopotamia (April 1699) (12th ed). London Miller, K 1916: Itineraria Romana. Stuttgart Moorey, P R S 1980: Cemeteries of the First Millennium BC at Dere Hiiyiik, near Carchemish, salvaged by T E Lawrence and CL Woolley in 1913 (BAR Inter- national Series 87). Oxford Mouterde, R, Poidebard, A 1945: Le Limes de Chalcis - organisation de la steppe en haute Syrie romaine. Paris Nobbe, C F A (ed) 1966: Claudii Ptolemaei Geographia. Hildesheim Noldeke, A 1920: 'Der Euphrat von Gerger bis Djerebis' Petermanns Geographische Mitteilungen 1920: 15-20 Palmer, A 1993a: 'The Messiah and the Mahdi - history presented as the writing on the wall' in H Hokwerda, E R Smits, M M Woesthuis (eds), Polyphonia Byzantina, Studies in honour of Willem JAerts. Groningen: 45-84 1993b: The Seventh Century in the West Syrian Chronicles. Liverpool Poidebard, A 1934: La trace de Rome dans le desert de Syrie. Le Limes de Trajan a la conquete arabe. Recherches aeriennes 1925-32. Paris Regling, K 1901: 'Zur historischen Geographie des mesopotamischen Paralellograms' Klio, reprinted Aalen 1966: 443-76 Ritter, H 1919: 'Arabische Flussfahrzeuge auf Euphrat und Tigris' Mesopotamische Studien 1, Der Islam, 9: 121-43 Roaf, M 1990: Cultural Atlas of Mesopotamia and the Ancient Near East. New York and Oxford Sachau, E 1883: Reise in Syrien und Mesopotamien. Leipzig. Reprinted 1995 by the Institute for the History of Arabic-Islamic Science, Frankfurt Sanlaville, P (ed) 1985: Holocene Settlement in North Syria (BAR International Series 238). Oxford Samowski, T, Trynkowski, J 1986: 'Legio I Italia- Libuma-Danuvius' in Studien zu den Militdr- grenzen Roms 3 (Vortrage des 13. Internationales Limeskongressses, Aalen). Stuttgart: 536-41 Sertok, K, Ergeq, R 1999: 'A new early Bronze Age cemetery: excavations near the Birecik dam, south- eastern Turkey' Anatolica 25: 87-107 Sinclair, T A 1990: Eastern Turkey: an Architectural and Archaeological Survey. London Stark, F 1966: Rome on the Euphrates. London Stein, G, Misir, A 1994: 'Mesopotamian-Anatolian inter- action at Hacinebi, Turkey' Anatolica 20: 145-90 Stein, G J 1999: Rethinking World Systems: Diasporas, Colonies and Interaction in Uruk Mesopotamia. Tucson Stucky, R A 1973: 'Ausgrabungen auf Tell el Hajj 1971/2' Antike Kunst 16: 83-6 Syme, R 1995: Anatolica - Studies in Strabo (ed Birley A). Oxford Thureau-Mangin F, Dunand M 1936: Til-Barsib. Paris Tuna, N, Ozturk, J 2000/2001: Salvage Project of the Archaeological Heritage of the Ilisu and Carchemish Dam Reservoirs: Activities in 1998; Activities in 1999. Ankara Wagner, J 1976: Seleukeia am Euphrat/Zeugma. Wiesbaden 1977a: 'Vorarbeiten zur Karte 'Ostgrenze des romischen Reiches' im Tiibinger Atlas des vorderen Orients' in J Fitz (ed), Limes, Akten des XI. Interna- tionalen Limeskongresses. Budapest: 669-93 - 1977b: 'Legio IIII Scythica in Zeugma am Euphrat' in J Fitz (ed), Studien zu den Militdrgrenzen Roms 2, Vortrdge des 10. Internationalen Limeskongresses in der Germania Inferior. Koln: 517-39 1983: 'Provincia Osrhoenae' in S Mitchell (ed), Armies and Frontiers in Roman and Byzantine Anatolia. Oxford: 103-23. 1985: Die Romer an Euphrat und Tigris (Antike Welt, Sondemummer). Mainz 2000: Gottkdnige am Euphrat (Antike Welt, Sonder- nummer). Mainz Wilkinson, T J 1990: Town and Country in Southeastern Anatolia, 1. Chicago Wheeler, E L 1991: 'Rethinking the upper Euphrates frontier: where was the western border of Armenia?' in V A Maxwell, M J Dobson (eds), Roman Frontier Studies 1989. Exeter: 505-11 Wroth, W 1899: The Catalogue of the Greek Coins of Galatia, Cappadocia and Syria. London 49