Zeugma From Space

Embed Size (px)

DESCRIPTION

An analysis of this important ancient site using satalite analysis.

Citation preview

  • Crossing the Euphrates in Antiquity: Zeugma Seen from SpaceAuthor(s): Anthony Comfort, Catherine Abadie-Reynal, Rifat ErgeSource: Anatolian Studies, Vol. 50 (2000), pp. 99-126Published by: British Institute at AnkaraStable URL: http://www.jstor.org/stable/3643016Accessed: 27/08/2010 05:13

    Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available athttp://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unlessyou have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and youmay 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 athttp://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 printedpage of such transmission.

    JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

    British Institute at Ankara is collaborating with JSTOR to digitize, preserve and extend access to AnatolianStudies.

    http://www.jstor.org

  • Crossing the Euphrates in antiquity: Zeugma seen from space

    Anthony Comfort, Catherine Abadie-Reynal, Rifat Erge9 Franco-Turkish Mission to Zeugma

    c/o Institut Francais d'Etudes Anatoliennes, Istanbul

    Abstract This article reviews progress in extensive survey work using satellite photographs along the course of the reservoir of the Birecik dam, completed in June 2000. Particular attention has been paid during field work to routes both along and across the river; here it is the east- west routes crossing the river which are examined in detail and in the context of the region's historical geography. The city of Zeugma, which has been the object of recent excavations, was the principal such crossing point but by no means the only one on the middle course of the Euphrates. A subsequent article will discuss north-south routes in the vicinity of the middle Euphrates and the question of its navigability.

    Introduction In 1996 the Polish ancient historian Michal Gawlikowski published an article entitled 'Thapsacus and Zeugma: crossing the Euphrates in antiquity' which concluded that the two places were one and the same, despite the scholarly consensus which had prevailed up to then and which placed Thapsacus about 130km to the south of Zeugma, on the Euphrates bend. This article reviews the evidence for this and other crossings of the Euphrates resulting from ground survey work using satellite photo- graphs in the area of the reservoir created by the Birecik and Carchemish dams (see fig 1, map of Zeugma and the Middle East).

    Thapsacus is a Greek name derived from the Aramaic 'Tiphsa' which means simply 'crossing point'. Zeugma, similarly, means the 'link' or 'bridge' in Greek. After establishment of the twin cities of Seleuceia-on-the- Euphrates and Apamea on right (west) and left (east) banks respectively in about 300 BC by Seleucus I, the city on the west bank became known as Zeugma and was renowned as a crucial link in Seleucid communications between the western capital at Antioch-on-the-Orontes and the other major city of the kingdom at Seleuceia-on- the-Tigris, founded as the new capital of the satrapy of Babylon (and opposite the later Parthian capital of

    Ctesiphon). Seleuceia-on the-Tigris lies about 760km southeast of Zeugma at a point where the Tigris and Euphrates are only 40km apart and not far from modem Baghdad.

    For most of the historical period one of the two main routes between the Mediterranean and Mesopotamia proceeded from Antioch to Zeugma in Syria and then across the river Euphrates towards Edessa and Harran in the satrapy or province of Osrhoene. It did not pass directly down along the river bank towards Babylon, probably because of the desert conditions which made it difficult to obtain supplies. The route via Zeugma and Edessa to the Tigris lies for the most part in relatively well-watered countryside and rich agricultural land ('the Fertile Crescent'). The other route, which became important from the second century, passed through Palmyra in the desert 280km to the south of Zeugma. This route while affording more direct access to the Mediterranean via Damascus and Tyre- was practicable only when security for caravans could be assured against attacks by nomadic tribes and when a system of watering points could be established. Norman Lewis discusses the pattern of settlement in the region of Aleppo (1987: 56) and quotes Gertrude Bell (1910), who remarked that the plain of Jerablus was coming into cultivation again.

    She noted that ... the ancient crossing of the Euphrates at Tall Ahmar [once Til Barsip, 20km south of Carchemish] was now in use again; indeed it was 'invariably followed by caravans from Aleppo to Urfa, though but a few years ago, when the Mesopotamian desert was still more unsettled than it is at present [1909], caravans were obliged to go up to Birijik and skirt the northern limits of the desert to Urfa'.

    From Edessa (now ?anliurfa) the northern route continued in an easterly direction towards the Tigris, passing through Nisibis or Singara and then Hatra during the Roman period- before reaching Seleuceia- on-the-Tigris and Ctesiphon. The cities of Antioch and

    99

  • Fig 1. Zeugma and the Middle East

  • Fig 2. The middle Euphrates valley

  • Anatolian Studies 2000

    Seleuceia-on-the Tigris (Invemizzi 1993) were two of the largest centres of population and therefore of economic activity in the ancient world. After the Seleucid period, they were on opposite sides of an important frontier and Zeugma became the main point of contact between Rome and Parthia. When the frontier with Sasanian Persia moved east to Nisibis, Zeugma remained a substantial and wealthy city with some of the finest mosaics of the region and large public buildings, which have only recently been partly uncovered.

    Much of Zeugma (the former Seleuceia-on-the- Euphrates) and all of its twin city of Apamea on the left (east) bank were drowned in 2000 by the Birecik dam, built at a point on the Euphrates in Turkey about 25km north of the Syrian border. From 1995 responsibility for excavations at Zeugma and Apamea in the area of the reservoir lay with a Franco-Turkish archaeological mission, which in addition had responsibility for excavating a large mound nearby at Horum, that has also now been drowned by the waters of the dam; a Spanish team excavated a similar Chalcolithic and Bronze Age mound opposite at Tilbes, while a group of Swiss and German specialists in Roman military architecture sought to identify the site of the Roman legionary camp. Last minute excavations on higher terraces at Zeugma, but which were to be inundated during 2000, were led by David Kennedy who had previously excavated there in 1993. Widespread publicity has been given to mosaics discovered in third-century villas found on the lower terraces during the final months. For a period of about three centuries Zeugma was the base of a legion and for most of this period the legion concerned was Legio IV Scythica (Wagner 1977).

    The Museum of Gaziantep has conducted a number of other interesting excavations in the area of Zeugma (see especially Sertok, Ergeq 1999) and will doubtless continue to do so since a large area of the city on the right (west) bank will not be affected by the dam. Until recently there was a village named Belkis or Kavunlu between the distinctive hill possibly an acropolis and the river, but little remained above ground to show the former importance of the site. This village was the base for the Franco-Turkish mission and hence also for the survey work described here.

    The first writer was engaged to examine the possible usefulness of satellite photographs for the investigation of the remains from antiquity in the area of Zeugma. The question of the location of the crossing points of the river has been one of the main focuses of research. Although the area inundated is for the most part narrow, being usually confined between high banks and sometimes gorges, all of these crossings for about 100km of the Euphrates upstream from the Birecik dam have disap- peared beneath the waters of the dam. The reservoir is

    56km2 in area and is 45m deep at the dam. A further 20km of the valley below the town of Birecik, almost as far as the Syrian border, have also been inundated by a smaller dam, also recently completed, just to the north of Carchemish. 100km beyond the border with Syria to the south lies the Teshrin dam- now also being filled. Thus, for much of its course as it passes through Turkey and Syria, the river Euphrates is covered by dam reservoirs, the greatest of these dams being the Atatiirk dam near Bozova, north of ?anllurfa (Edessa), and the Assad or Tabqa dam near Raqqa in Syria. The Atatiirk dam was completed in 1991 and had already drowned Samosata, a major city of the Roman empire, legionary fortress and previously winter capital of Commagene (known as Kummuh to the Assyrians). This city had also been a major crossing point of the Euphrates since Assyrian times.

    Only limited research was undertaken before the water covered Samosata (see especially Ozdurgan 1977; Blaylock et al 1990). The Turkish dams concerned form a part of the Project for Development of Southeast Anatolia (GAP), which includes dams on both Euphrates and Tigris rivers; the cultural aspects of this project are discussed by Algaze (1993a). Earlier writers about the area examined here include Archi (1971), Hellenkemper (1977) and Wagner (1976). A detailed survey of much of the Euphrates valley affected by the Birecik and Carchemish dams was also carried out in 1989 (Algaze et al 1994). David Kennedy has recently published a review of the situation of research in regard to Zeugma up to the stage of the recent series of excavations (Kennedy 1998b). This book also contains a collection of extracts from ancient authors who refer to Zeugma and the surrounding area; some of these authors are referred to below. Excavations by the French teams working at Zeugma and Horum are the object of regular reports in the periodical Anatolia Antiqua, published by the Institut franqais d'Etudes anatoliennes of Istanbul.

    More broadly, the sources for the historical geography for the region south of Zeugma as far as Dura- Europus have recently been discussed by Gaborit and Leriche (1999). David French will shortly be publishing an article on the Roman roads of the region to the west of Zeugma. East-west routes across Mesopotamia and Syria, including those of the Classical period reviewed here, are also discussed by Algaze (1993b: 42-5).

    Mme Catherine Abadie-Reynal was leader of the French part of the Franco-Turkish Mission to Zeugma and through her invitation made this research using satellite photographs possible; this paper has benefited from her advice. Rifat Ergeq was, until recently, Director of Gaziantep Museum and leader of the Turkish team and provided valuable support. On occasion commissioners of the Gaziantep and ?anliurfa museums also accom- panied the survey team.

    102

  • Comfort, Abadie-Reynal, ErgeV

    Satellite photography and archaeology on the Euphrates An extensive survey making use of satellite photographs and the work of predecessors interested in the valley and its surroundings was carried out over four years from 1996. Although the images and photographs were initially obtained to investigate the immediate area surrounding Zeugma, it was soon apparent that their utility lay less in the identification of archaeological features in and around the site itself than in investigation of ancient roads and other large features over a wider area along and around the course of the reservoir which would be created by the dam.

    Use has been made especially of Russian high- resolution photographs of June 1990 and May 1992 taken by a camera known as the KVR- 1000. These are publicly available and are taken from a height of 220km. Objects 2 to 3m across may be seen in the photos which therefore constitute a good substitute for vertical aerial photog- raphy, not currently available in Turkey to archaeologists. When specific areas of the KVR photos are fully enlarged and scanned into a computer file individual trees are visible, but the expense both of purchasing and scanning such photos constitutes a serious constraint on their use. Some examples enlarged for the areas of Zeugma and Rumkale are annexed (figs 3, 4).

    In addition, much less expensive American photo- graphs of the 1960s from the Corona series have been used for surrounding areas outside those covered by the KVR- 1000 photos. The Corona photographs have a rather lower ground resolution of about 4 to 5m and are therefore a tool which is less precise (fig 5). SPOT and Landsat images were also obtained and investigated (resolution of 10m and 30m respectively- figs 6, 7). While useful for identifying modem road systems and patterns of vegetation, these images are less useful for archaeologists because the information contained is more generalised and less detailed. Corona photographs of the area of Zeugma were also published in David Kennedy's The Twin Towns of Zeugma on the Euphrates, mentioned above (1998b), and also used in his article on satellite archaeology (1998a).

    Some features such as the walls of Apamea and the citadel of Seleuceia/Zeugma may be noted in the Russian photos and indeed in the Corona and SPOT images. However, the substantial erosion from the conical hill of 'Belkis tepe' (whether this constitutes the acropolis of Zeugma remains a matter of disagreement) has precluded any possibility of identifying individual buildings from the sky because of the thick layers of colluvium resulting on the right bank. Some large artificial banks which may cover walls linked to the military role of Zeugma do however appear in the photos above the town on either

    side of the modem road from Nizip where it descends to the river. Erosion and colluvium, together with repeated destruction by Persian and other invaders, and more recent looting of artefacts, probably also account for the fact that the ruins visible today on the surface are so much less impressive than those of Palmyra or Bosra, other cities which became extremely wealthy from trade and whose surviving artefacts are no more impressive than the mosaics of Zeugma.

    On the opposite, left bank of the Euphrates, buildings which were revealed by geophysical work at Apamea did not appear on the satellite photos because of silt and alluvium brought down by the river. As today, most buildings seem always to have been constructed from mudbrick, easily washed away during the river's floods: even the upper courses of the city walls of Apamea had probably disappeared for this reason, although the general layout of the fortifications is clearly visible in the KVR photos and, somewhat less clearly, in the SPOT panchromatic image.

    Early on, therefore, it was decided to concentrate on using these satellite photos for extensive survey work on a broad scale. The first KVR photo acquired covered a parallelogram about 25km long and 10km wide. Because of its usefulness also for the work of the Swiss archaeologists interested in Roman military architecture, a second and third photo of this type was acquired by Dr Martin Hartmann and kindly made available. One of these covers an area of 40 by 40km to the north of Zeugma and therefore includes most of the valley of the Euphrates which was inundated at the end of 1999 by the Birecik dam.

    The team available for survey work was very limited, consisting usually only of one or two people, but with occasional reinforcement from the excavators at Belkis; from 1998, Dani? Baykan, a doctoral student of archaeology from Istanbul University, participated in the survey work and provided an invaluable contri- bution, especially through the questioning of local people. He has also commented on this article. The pace of change in the area had been particularly rapid over the preceding decades, with many ancient buildings having disappeared and the river itself having lost all its former importance as a thoroughfare. Nevertheless, it proved possible to speak to many local people with memories, for example, of caravan traffic and of ruined churches, now almost entirely vanished. The site guardian of Zeugma, Nusret Ozdemir, has been especially helpful both in showing remains of aqueducts and other features in the area surrounding Zeugma itself and in providing details of a way of life which has only recently disappeared.

    103

  • Anatolian Studies 2000

    Apamea I I

    I Public buildings- I I ---I

    | Military encampment? )

    I Belkis tepe (acropolis?)

    Fig 3. Zeugma: extractfrom KVR-I000 photograph of 11.6.1990; ground resolution of original = 2m

    First gate -

    Cave house 1

    Fig 4. Rumkale: extractfrom KVR-1000 photograph of 22.5

    Some progress has been made in mapping features of archaeological interest along the valley and its immediate surroundings. Most of the sites located in the valley by Algaze and his team (Algaze et al 1994) have been located again and given a UTM reference using a small Global Positioning System (GPS) instrument. Other sites in the area of Zeugma mentioned by Archi (1971) or Wagner (1976) have also been located and many new ones found in the course of field-work. Work has, however, been handicapped by the lack of large-scale maps, whose use is restricted by the military authorities.

    Pie r of Roman bridge

    K. 1992; ground resolution of original - 2m

    The KVR photos may be turned into maps at a scale of about 1:25,000 by geo-referencing ground control points visible from space, such as cross-roads, with the GPS instrument and then correcting and orientating the photos accordingly. Using the programme 'Idrisi for Windows' such procedures proved relatively simple, though time-consuming. A UTM grid, names and inter- pretation of features visible on these photos were added using the programme 'MapMaker Pro'. Printing and publication however is far from easy, although large- scale maps can be printed out even for an area of 20 by

    104

  • Comfort, Abadie-Reynal, Ergeq

    Fig 5. Zeugma. extract from scanned Corona photograph of 8.8.68; ground resolution of original = 4-5m

    Fig 6. Zeugma: extract from digital SPOT image of 7.9.91; ground resolution of original = 10m

    105

  • Anatolian Studies 2000

    Fig 7. Zeugma area: extractfrom digital Landsat image of 26.3.85, bands 3, 4 and 5; ground resolution of original 30m (Belkis tepe is left of centre to the south of the river)

    20km using a 'poster print' option, which allows individual sheets to be printed and then glued together. It is the subsequent reproduction and publication which have so far posed problems, while the inclusion of topographic information such as contours although theoretically possible from a pair of satellite images by a process known as 'interferometry' has proved to be beyond the budget available (see annex for current prices of the imagery and processing).

    The main advantage of the photos lies in their homogeneous coverage of a wide area which when scanned and viewed on a computer- permits creation of large-scale maps on which the topography is clearly visible; the interpretation of features of archaeological interest may be superimposed using colour symbols. The ancient features of most interest which are visible on the photos are usually roads and tracks, which may either be directly visible or may be deduced from field boundaries and other secondary evidence. Interpre- tation of the photos cannot often provide conclusive evidence of the antiquity of such roadways. However, even on the ground the evidence may be far from convincing and rarely datable. The few incontro-

    vertible pieces of evidence- such as Roman bridges - have confirmed that some routes previously identified as of ancient origin on the satellite photos correspond to the real course of such ancient routes. Ancient mounds 'hyiiks' in Turkish are also usually visible from space and are frequently located along roads used also in Classical antiquity.

    Results of the survey work were described in annual reports for the Mission archeologique de Zeugma of which a summary in French was submitted as a part of the formal document prepared for the project's sponsor, the Ministere des Affaires Etrangeres in Paris, and for the Turkish authorities. Brief summaries of the work using satellite photos have also appeared in published form (Comfort 1997, Comfort et al 1998). A website is also available containing more details of this project and some reduced versions of the photos and images being used (http://www.ist.lu/zeugma/). The site also contains a large number of colour photographs of sites mentioned here. Methodological aspects of the project have been the subject of a dissertation submitted to Leicester University in connection with an MA course by distance learning on Archaeology and Heritage.

    106

  • Comfort, Abadie-Reynal, Erge?

    Crossing the river One of the main tasks addressed in the survey work has been the precise identification of ancient roads leading to Zeugma. However, it soon became clear that, even if Zeugma was throughout antiquity by far the most important crossing point, there were others in the vicinity; the road network was not therefore limited to those routes leading to and from Zeugma. For place- names in this section see figs 1 and 2, maps of Zeugma and the Middle East, and the middle Euphrates valley.

    Since the Middle Ages it has been the town of Birecik, 10km east of Zeugma and which still has an important medieval castle, that has been the principal crossing point of the river for hundreds of kilometres up and downstream. The road bridge constructed in 1956 replaced a ferry which was notoriously busy and which involved crossing in small flat-bottomed boats. Until construction of the Atatiirk dam in 1991 it was the only road bridge over the Euphrates between the Syrian border and Malatya 150km to the north. The railway bridge on the line to Mosul and Baghdad, which now forms for most of its length the frontier between Turkey and Syria, was built shortly before the First World War. Today the railway is little used but, the Birecik road- bridge which is still the main link between the Mediter- ranean and the southeast provinces of Turkey continues to have enormous strategic and commercial importance. Birecik did not, however, become important as a crossing point until that at Zeugma had declined, probably in the early Middle Ages.

    For most of recorded history there has been no bridge at all on the Euphrates, although its width of between 180m and 200m at Zeugma is less than that of the Rhine and Meuse and hardly greater than the Thames at London. Crossing has usually been in small wooden boats or on rafts supported by inflated skins- or alter- natively, when the river is low, on foot or horseback by a small number of fords. The absence of a fixed link is easily explained by the frequent and destructive floods to which the river was prone until the construction of dams along its upper reaches. Nevertheless, such a fixed link did exist for several centuries in antiquity at Zeugma, although the nature of the bridge is still a matter of dispute and no trace has yet been found of stone piers which might incontestably confirm the precise location of the bridge at Zeugma. The lack of such stone piers today, despite an apparent sighting in the 19th century by a German visitor (Sachau 1883, as reported by Cumont 1917: 142), is the main reason for believing that the 'fixed link' may always have consisted of a roadway built on a series of boats tied together and able to rise with the level of the water, rather than a permanent and immovable structure built on wooden or stone piles.

    Given the rapid current of the river at Zeugma such a 'pontoon bridge' would require strong moorings and Pliny (NH 34.43) refers to an iron chain, apparently still visible in his day, which was shown to visitors as the chain used to make the bridge when it was crossed by Alexander.

    Pontoon bridges may of course be quickly dismantled to deny access to an approaching enemy (as was intended by Ammianus Marcellinus (18.8.1) in order to impede an advance of the Persians in AD 359), but the problem of assuring downstream passage past Zeugma for the rafts carrying timber and stone from quarries upstream is not easily resolved. It is possible to imagine a system whereby one part of the pontoon bridge was swung on a hinge downstream with the current in order to allow passage for boats and rafts, but there is no evidence of how this was done. The modem pontoon bridge at Jerablus on the Syrian side of the modem frontier has no provision for letting boats through, but there seems to have been no substantial traffic along the river in the post-World War II period. There are, however, reports of pontoon bridges being drawn in when required during the 19th century much lower down the Euphrates in Iraq (Kennedy 1998b: 41).

    The 'fixed link' of Zeugma may have been one of several pontoon bridges constructed in the Hellenistic period; there appears to have been one at Samosata (130km upstream and now covered by the waters of the Atatiirk dam) and the long confusion over the site of Thapsacus may have resulted from the existence of a third important crossing point in the Hellenistic and Roman period near Dipsa, a town by the Euphrates about 150km south of Zeugma and close to the former religious centre of Hierapolis (also known as Bambyce or Membij). It is in this area that Thapsacus was usually thought to have been located before the appearance of Gawlikowski's article (e.g. Bosworth 1980: 222). However, this latter crossing may never have involved a fixed link; although it lay on the direct route from Aleppo to Urfa (Edessa) and was important throughout the Ottoman period, crossing the river here even in 1907 involved a small ferryboat which had to be dragged upstream for a mile in order to reach the equivalent point on the opposite bank to the point of departure, since the current was still strong enough to make direct crossing by boat impossible (Soane 1926: 38).

    But there was probably a crossing point at or near our Zeugma much earlier than that established or reinforced by Seleucus. It seems possible that Darius on his attempted escape eastwards after his defeat by Alexander at Issus also crossed here -probably on a pontoon bridge- and that this may have been the main east-west route for the Achaemenid empire. The route of the

    107

  • Anatolian Studies 2000

    Persian Royal Road from Elam to Sardis has long been a source of contention. First Ramsay (1890) and then Calder (1925) argued strongly for the route's passing along this southerly trajectory and crossing at Zeugma, but many have believed its course to lie further north, probably crossing the Euphrates at Samosata which was the capital of the kingdom of Kummuh, later Commagene. Oates (1968: 7) mentions both crossings; for general discussions, see Dillemann (1962: 151), who is not convinced by Calder, and Forbes (1965: 138), who is. A fine summary is included in an unpublished paper by Starr (1962) referred to by French (1983: 91) and available in the library of the British Institute of Arch- aeology at Ankara. Starr explored possible routes for the Persian Royal Road in 1961 and concluded that the second pair of double 'gates' referred to in Herodotus (1,75) can only be the Cilician Gates, a natural rock formation on the pass between Tyana and Tarsus. This in turn means that the main crossing of the Euphrates was probably at Zeugma, although this route from Kayseri and central Anatolia to Mesopotamia was not the most direct. Starr, however, also identified an alternate route via Mara? and Samosata, citing in particular the Roman bridge on the G6ksu (see below), where there is evidence of an earlier bridge.

    However, before the Achaemenids, it is likely that Zeugma lay on a route frequented by Assyrian merchants on the journey between their homeland and the Anatolian trading colonies. A route via Urshu- usually identified with Gaziantep- is mentioned in a tablet from Kanesh as one of two possible routes from Assyria to the Anatolian plateau (Astour 1995). Although the northern route via 'Khakkum' is identified by Astour with the Mardin-Diyarbakir-Malatya road, he believes that the southern one passed from Urfa to Mara? via Zeugma and Urshu (Gaziantep). Old Assyrian texts, according to Astour (1995: 1410), place 'Abrum' on the left (east) bank (Apamea?) and Zaqaria on the right bank (Seleuceia/Zeugma?). Subsequent places on this route from Assyria mentioned are 'Ashikum' and 'Khaqa', which he identifies with the mounds of Turlu and Aril (both villages on the old main road from Nizip to Gaziantep see fig 2, map of the middle Euphrates valley). Astour goes on to consider the Persian Royal Road and, despite his insistence that this followed earlier Assyrian practices, he fails to identify a route across the Euphrates at this point, preferring an exclusive crossing near Khakkum and Malatya (1995: 1418).

    There seems no particular reason to prefer Birecik to Zeugma as the main early crossing point of the region unless it be that the river was broader there, hence less fast-flowing and easier for ferries to cross directly. In contrast to Astour, Beitzel (1992: 56) suggests that the

    crossing on the principal Assyrian route via Hahhum (or Khakkum) lay at Samsat but that an alternative via Ursum crossed at Birecik; this is apparently because he believes there to have been only 'four direct fords of significance' on the upper Euphrates - Kemah, Malatya, Samsat and Birecik. But the crossing at Birecik in historical times has been by ferry.

    For much of the Bronze Age and especially during the Hittite period the main settlement of the area was at Carchemish, now on the border with Syria 30km to the south and currently impossible to visit for this reason. Substantial remains of this city were excavated by Woolley before the First World War (Hogarth 1914; Woolley 1921) and it is likely that there was a ford here near the modem railway bridge (see below); no remains from the Hittite or earlier periods have been found at Zeugma itself but a large village with an Uruk settlement has recently been excavated at Hacinebi on the left bank between Zeugma and Birecik (Stein, Misir 1994; Stein 1999) and a very large cemetery of the Middle Bronze Age was discovered in 1997 during the process of dam construction one kilometre southeast of the village of Belkis (Kertep, Ergec 1999). The mounds at Horum and Tilbes, recently excavated, also indicate a high level of population and activity along the valley in periods preceding the Hittite dominance of Carchemish.

    Zeugma from 500 BC The use of Zeugma as a crossing point into the modem period is attested by the guardian of the site, recently retired, Nusret Ozdemir, who saw as a child caravans with camels crossing the Euphrates near Belkis village in wooden boats, four camels to a boat, with their knees bound to prevent them from rising during the crossing. The village on the opposite bank from Belkis was called Tilobur, or 'Gecittepe' in Turkish, 'the hill of the crossing'. The history of this crossing point remains unclear archaeological evidence has not yet been found for an origin preceding the Hellenistic era, but Gawlikoski has advanced convincing arguments for believing that it goes back at least to the Assyrian period and for identifying it with 'Thapsacus'. The latter is first mentioned in the Bible as the eastern limit of the kingdom of Solomon; it was probably the point of entry at that time from the east into the Persian satrapy of Syria and Palestine.

    Thapsacus was also known throughout the period up to the conquests of Alexander as the main crossing point of the Euphrates and therefore an essential stop for travellers passing from the Mediterranean to the East. For the geographer Ptolemy, distances in the East are measured from Thapsacus and this crossing point of the Euphrates is therefore of crucial importance for ancient

    108

  • Comfort, Abadie-Reynal, Ergeq

    To Samosata 0 0 0 oo 0o 0

    0 0 ?

    0 000 0 ?

    o or

    ooo o o o

    o

    o

    o

    0 0 o

    o

    o o o

    o o o o o

    0000

    0

    o

    Sarikoc )

    oo 0 0 0

    0

    Legend 0 Major hoyuk (mound) * Ancient town

    ,* Quarry o o o o o o Main ancient i

    *,* #O-*o Frontier

    Fig 8. The Euphrates valley from Carchemish to Halfeti

    109

    0 10 Km hvlII km ?

    o _ 0

    o

    o

    o o

    o 'o0

    0

  • Anatolian Studies 2000

    To Samosata , 00oATATURK DAM

    TuFus V o'

    r

    0- * ^ -- >

    Uckilise/Edessa

    A Crossing point

    A Roman bridge

    * Major hoyuk (mound)

    * Ancient town (Hellenistic/Roman)

    * Town/village

    ~:~ Castle

    jr Mausoleum

    aD Church or temple

    Contour (450 and 600m)

    o ? ? o 0 Main ancient routes

    Fig 9. The Euphrates valley from Halfeti to the Atatiirk Dam

    110

    To

    o

    Kosk

    To DC

  • Comfort, Abadie-Reynal, Ergeg

    geography. Alexander himself is said by Arrian to have crossed at Thapsacus during summer 331 BC by two bridges which were specially constructed (3.7.5); from here 'he continued inland through the country called Mesopotamia, keeping on the left the Euphrates and the mountains of Armenia.' This description corresponds more closely to a crossing point at Zeugma, or even further to the north, than to one at Meskene, the tradi- tional identification of Thapsacus, 150km to the south of Zeugma and far from the mountains of Armenia. The reference to two bridges also opens the possibility of two columns crossing in different places.

    (For place-names in the following section, please see figs 2 and 8, maps of the middle Euphrates valley and the the area surrounding Zeugma.) It has been thought that Zeugma was a new foundation- although it was never clear why Seleucus would have chosen to move the crossing point from 'Thapsacus'. No pre-Hellenistic levels have yet been identified at Zeugma, but if Gawlikowski's conclusions are accepted then the foundation of the twin cities on each bank which became known as Zeugma was simply a matter of fortifying the existing crossing point and perhaps constructing a new and semi-permanent pontoon bridge. The large Bronze Age cemetery excavated in the course of construction of the dam 1km southeast of the Hellenistic and Roman city is perhaps an indication that Zeugma too was occupied in much earlier periods, although archaeology has not proven this. No Achaemenid or earlier levels have yet been identified at Zeugma but they may lie hidden below many metres of colluvium and occupation levels of Byzantine, Roman and Hellenistic Zeugma. Hellenistic and Achaemenid pottery is in any case difficult to distinguish.

    Although the thesis of Gawlikoski is supported by Starr and others in regard to the crossing point of the Euphrates for the Achaemenid Royal Road (and by Astour for earlier periods), given the absence of arch- aeological evidence the question of the identification of Thapsacus with Zeugma has to remain unresolved. Although the arguments advanced by Gawlikowski seem strong, other sites such as Carchemish/Europus in the northern reaches of the Middle Euphrates also have claims. Carchemish is the 'Thapsacus' favoured by Farrell (1961). But in any case the question arises of the reason for construction of a crossing point at Zeugma and for its probable dominance over a long period, at least from Seleucid times. There is no obvious ford (although, because of dams upstream, the former seasonal highs and lows which would sometimes have permitted easy fording were no longer apparent in the region of this survey. There are - or were - however, islands in the middle of the river just upstream from the assumed

    crossing point between Seleuceia (Zeugma) and Apamea. These islands changed shape regularly and seem to have been created by the sharp slowing in current forced by the almost 90 degree turn which the river was obliged to make at 'Belkis tepe' (the large conical hill used by Zeugma as a symbol of the city on its coinage in the Roman period). Downstream from this point the current may on occasion have been rapid but the river was just about navigable to its mouth (Chesney 1850; Guest 1992). The islands would have permitted some sort of fixed structure to be built in mid-stream (of which no remains were found) from which the centre of a pontoon bridge could be moored.

    A small, deep valley called the Bah9e Dere and lined by tombs descends to join the Euphrates just north of the town on the west bank; it provides relatively easy access from the plateau to the riverbank at a gradient manageable for carts. The ancient roadway descending from the west is assumed to have followed roughly the course of the modem road from Nizip which drops 1 00m to the river from the plateau. On the east bank the ground slopes up more gently; a 'river-gate' was sought un- successfully in the walls of Apamea but this may have been eroded by the Euphrates. More generally, the geographic position of Zeugma favoured its development as a crossing point. It is situated near the gorges from which the Euphrates issues above Ehnes (16km upstream) and east-west travel at any point to the north of Zeugma is made more difficult by the mountainous terrain west of the river. The satellite photos show clearly the rich agricultural plain of Altminova (Hobab) on the east bank which would have facilitated the provision of supplies for travellers. Large arable areas also lie to the south of Zeugma on the west bank.

    Zeugma is also at a point closer to the Mediterranean (whether the ports on the Gulf of Issos or Seleuceia-in- Pieria) than anywhere else on the Euphrates. In fact it would be possible to carry goods with hardly any extra distance involved to a point much further south on the Euphrates bend near Hierapolis/Bambyce, but this meant crossing an area where water supplies were not so easy to come by and where for long periods physical security may have been at risk from nomad raiders. The steppe desert (in the sense of that area south of the 200mm isohyet line the minimum annual rainfall for rain-fed agriculture) begins about 60km below Zeugma, although such conditions did not prevent the establishment of trade routes in later periods.

    There is another less rational explanation which has not yet been subjected to critical analysis. The conical hill of Zeugma is still very prominent and may have been even more so in antiquity, given the large volume of colluvium eroded from its flanks since then which has

    111

  • Anatolian Studies 2000

    covered entirely the city and given the possibility of an artificial mound on top of the hill, similar to the burial mounds of the dynasty of Commagene at Dikilita? or Karaku?, which may have been subsequently washed away. Belkis Tepe is even today a landmark for miles around and may also have been endowed with some religious significance (the temple of Tyche, which seems to have been the principal temple in the Hellenistic and Roman periods, was within the walled precinct on its summit, together with four large cisterns). It may therefore be that the religious awe associated with a prominent landmark, easily fortified, contributed to the reputation of Zeugma as a place of pilgrimage and encouraged travellers to pass this way rather than elsewhere. In a similar fashion, Roman roads in this region often pass close by prominent artificial mounds which had long lost their significance as towns, presumably because of their usefulness as landmarks although Roman road-building activity also often seems to have involved only the improvement of previous long- distance road networks.

    Once the Achaemenids, the Hellenistic kings or the Romans had imposed a system of collecting taxes in the form of tolls on goods crossing the river, it was obviously in their interest to ensure that other points without customs offices were no longer permitted for caravans. The only firm evidence for such tolls is a story from Philostratus (1,20) concerning a declaration to a customs officer by Apollonius of Tyana dating to the third century AD (although the sage concerned lived in the first century), but it seems certainly possible that such tolls were exacted as soon as the states concerned had the necessary bureaucratic apparatus to do so. The customs officer understood the list of Greek virtues declared by Apollonius to be slave girls.

    The volume of traffic would be the primary consider- ation in whether or not a fixed link was required, since the extra cost of its construction would be balanced against the disadvantages in waiting time for caravans and armies, if only a few small ferry boats were available. The pontoon bridge would avoid a much longer wait for commerce and therefore justify payment of heavy tolls. The volume of traffic and the question of the importance of commerce as an element of Roman frontier policy has been discussed by Isaac (1993; 1998). The evidence is wholly insufficient for a comparison of the value of trade passing through Zeugma with that of the sea routes or the desert route through Palmyra, but in any case it would presumably have fluctuated through time.

    When tradition established a particular point as the main crossing of the Euphrates for both caravans and armies, it would seem highly probable that it should keep this role for long periods. Seleucus may well therefore

    have reconstructed an Achaemenid pontoon bridge, his innovation being rather the founding of cities on each bank to act as fortified bridgeheads which might protect this essential link in his communications from adver- saries and rebellious vassals. But in the absence of any Achaemenid remains, this is unproven.

    Investigation of the possible routes to Aintab and Doliche, west of Zeugma, has revealed that the most probable course of the ancient roads was that followed by the main road between Gaziantep and Nizip prior to the construction of the post-World War II highway to ?anliurfa. (This in turn will shortly be replaced by a motorway.) The old road also used by 19th-century travellers such as Chesney passed by a major mound at Aril and then followed the Nizip river to another such mound at Turlu before turning south to the town of Nizip. However, no evidence in the form of cisterns has yet been found to confirm this supposition.

    To the east, the road from Apamea to ?anliurfa, ancient Edessa, followed two different routes: the first about 10km north of the existing highway- left the river at the current important junction of Hacinebi (the site of American excavations of a settlement dating from the Uruk period reported in Stein 1994) and turned north- northeast along a 'hollow way'. It avoided the mountain of Arat Dag to the south, following its northern edge to the villages of Yuvacik, Haydarahmet and (armelik (now Biiyiikhan). The evidence for this route consists of cisterns, some roadside alignments of stones, scattered blocks at Yuvacik and Haydarahmet, local tales of caravans and reports of early European travellers, such as de Thevenot (part II: 41) and Buckingham (1827: 42), who stopped at the large 'han' or caravanserai of Carmelik. Impressive ruins of this building may still be visited today.

    The second route to the east is that described in the Parthian Stations of Isidore of Charax. After Apamea, this route mentions as its first stop Daiara, possibly the 'Thiar' of the Peutinger Table; investigations in summer 1999 have shown that the probable route passed through Birecik and then southeast to Siiriic, usually identified with ancient Batnae/Anthemusia or the 'Charax Sidou' of Isidore, via the villages of Kurucahiiyuik, Hanhiuyiik, Uctepe and Misgevli. The evidence -apart from the presence of the ancient mounds and associated ceramics

    is once again tales of caravans, which passed this way until about 1960, and the ruins of a troglodytic caravanserai near Hanhiiyiik, which may have been 'Thiar'. However, an alternative route to Batnae would have been east from Apamea to Haydarahmet and then following a track still marked by ancient mounds (Karahiiyiik and Kizilhuiiyk) towards the southeast. This route is very evident on the Corona satellite photographs although not currently in use.

    112

  • Comfort, Abadie-Reynal, Ergeq

    There are no traces of the ancient settlement of Batnae/Anthemusia to be seen at Siiriic. Jones states that the two were not identical (1937: 442, n 4), Batnae being located merely in the territory of Anthemusia. Some tombs and quarries were located on hills about 8km west of Siiriic which might be the remains of Batnae, but according to Ammianus (23.3.1), discussed by Millar (1993: 483), there was a fair every September in this town which brought traders in goods from as far as India and China. Such a fair seems much more likely to have taken place in the well-watered and fertile plain around Stiriic than on the bare hills above.

    A problem frequently found in tracing ancient roads in this area is the absence of a paved surface. In some cases local people speak of paving stones having been removed only recently (and we were shown such stones near Koseler on the west bank) but it is also true that camels prefer a soft surface. The Roman roads may therefore have been abandoned very early on, if only for routes alongside which could still exploit the cisterns, some of which are very large and have remained in use to modem times.

    Zeugma was far from the only crossing point in the region. Survey work conducted with satellite photo- graphs since 1996 has revealed at least five other routes across the Euphrates between Zeugma and Samosata which may have been important at various periods. Newly-obtained copies of Turkish 1:100,000 maps show many fords which- while not recently in use- never- theless constitute proof of crossing points, which were indeed used until the modem era. Survey work has attempted to investigate all such crossings in the 100km of the river Euphrates which will be affected by the reservoir of the Birecik dam. For place-names in the following sections see fig 9, map of the Euphrates valley from Halfeti to the Atatiirk Dam.

    Rumkale Two other crossing points are shown on the 1:100,000 maps near Horum and Cekem (respectively, 6 and 21km upstream from the Birecik dam); the first of these may have been important in the Chalcolithic and Bronze Ages especially, because of the two important settlements at Horum itself and at Tilbes on the opposite bank. Quays or tow-paths have been found in the water below both mounds. But the first point at which a major ford seems possible above Zeugma lies at the junction of the Euphrates with its important tributary called the Merzumen ('Marsyas' in antiquity), 26km upstream from Zeugma. Curves in the river combined with the waters of the tributary river to form a bank of gravel which stretches across the Euphrates diagonally in a northwest-southeast fashion. Although some parts of

    this bank lie deeper below the surface than others, it seems likely that the main river was crossable at this point at times of low water.

    The Merzumen, which approaches from the west, takes a sudden turn to the north just before joining the Euphrates, being deflected by an enormous promontory of rock which juts out from the mountain to the south. On this promontory, cut off by two deep artificial trenches dug into the neck of the ridge linking it to the mountain, lies the large and famous castle of Rumkale (fig 10). Wheeled traffic along the right bank of the Euphrates beneath the castle is currently impossible, because of the narrow space of flat land available, and may always have been so, although there are some signs of a passage being excavated in the rock immediately underneath the fortress. The Roman military road which is shown on the Peutinger Table (but not in the Antonine Itinerary) as passing along the right bank of the Euphrates between Zeugma and Samosata to the northeast, probably climbed to the plateau north of Ehnes (which may be the ancient Arulis and is the site of some well-known Roman quarries) via a route with ancient cisterns carved into the adjacent rock-face and- after passing through the modem village of Koseler on the summit- continued north down into the Merzumen valley and Rumkale. Remains of an ancient roadway about 3m broad were discovered on a zig-zag just below the lip of the escarpment north of K6seler. This route will be discussed in detail in a subsequent article.

    Similarly, on the northern side of the Merzumen, leaving Rumkale for the broad and fertile plain which leads westwards to the town of Yavuzeli (formerly Cingife), traces of a well-built roadway, though less than 2m wide, were found beneath the modem village of Begendik. This road presumably connected with the main route from Doliche (near Gaziantep) to Samosata, which passed along this valley, over the Kara Dag range via the villages of Hasanolu, Elif and Hisar (all of which have large and fine Roman funerary monuments) and then down to the Kara Su Roman bridge, of which one arch is still extant (see French, in press; fig 11 and below).

    The imposing presence of the medieval castle of Rumkale (Hellenkemper 1976: 51-6; Sinclair 1990: 166- 72) may indicate that this was always an important crossing point (figs 2, 6). However, despite rumours of earlier occupation, no remains of periods earlier than AD 1000 have yet been positively identified here, apart from the piles of another Roman bridge below the castle, which lie about 200m inland from the junction of the two rivers. Efforts were made to investigate the area which would be inundated (not including the castle itself, which will be relatively unaffected though isolated by the

    113

  • Anatolian Studies 2000

    Fig 10. Rumkale: fortifications from north side

    Fig 11. Bridge on the Karasu south ofAkbudak (Siipiirgiiu)

    Fig 12. Kantarma: remains of storehouse or granary (?)

    114

  • Comfort, Abadie-Reynal, Erge?

    reservoir). The lower part of the tributary valley was covered with lush and beautiful vegetation in stark contrast to the surrounding hills, before rising into a series of grandiose cirques to the west. Some carved stones in the river seemed to be associated with a nearby mill and a former bridge, now disappeared; a cave dwelling by the Merzumen had Arabic and Armenian inscriptions, as well as graffiti of boats on the river, but nothing was found to show that this area was intensely occupied in earlier periods nor the scene of a busy caravan traffic. The ancient roadway descending from the mountain to the south must have crossed the Merzumen before reaching Rumkale and local people confirmed that there had in fact formerly been two bridges, one near the stepping stones in recent use, for which a natural rock, partly carved, formed the central pier. The next two bridges (one medieval, one Roman) across the Merzumen nearer its junction with the Euphrates and whose remains were still visible, must have served the castle and- possibly -a minor road along the Euphrates only. North-south passage beneath the castle on the Merzumen (western) side was impos- sible even on foot and only a narrow track was feasible on the Euphrates side. However, an impressive double aqueduct had been dug along and through the cliffs here; following a tunnel of 200m through the promontory on which the castle is built, one of these aqueducts then descended to the south above the Euphrates for 3km as far as the village of Kalemeydan, opposite Halfeti, where it was said to have powered a mill. The aqueduct itself functioned until 1997.

    On the east bank opposite Rumkale, it is known that a route continued towards Edessa since Pococke passed this way in 1738 (1743: ch 18). Traces of a roadway rising from the river at a point some 500m south of Rumkale have been identified, near the recently ravaged remains of what appears to have been an early Bronze Age cemetery (Kalecik Dag, about 2km north of Halfeti; there are similarities with the necropolis southeast of Zeugma recently published by Sertok). Pococke continued northeast to the village of Cibin (now Saylakkaya, see below) and then east to $anlhurfa via a series of villages as yet unidentified. The recently discovered remains of an imposing building of the Hellenistic or early Roman period at Kantarma, a village east of the nearby regional centre of Yukan G6klu, may indicate that there was an early and important route leading eastwards from the Euphrates near Rumkale and Ayni (see below). The building concerned appears to be the ground floor of a large storehouse and consists of three rows of large arches, very strongly built, 22.5m long by 10.5m broad (fig 12). It was found by the first author in May 1999 but has also been the object of a recent visit by a team from Ankara University. There

    were also reports of a road leading east towards Urfa at the village of Argil, further south, so there may have been several possible routes from the Euphrates crossings leading eastwards.

    One of these seems likely to have passed through the village of Sava?an, opposite and slightly to the north of Rumkale. In addition to vestiges of two churches and of troglodyte settlement in and around the modem village, a further abandoned settlement, with tombs and houses apparently of the Roman period, was found 2km further east, up a narrow valley and at the foot of a mule-track leading to the plateau, which is still in remarkably good condition.

    Gozeli/Ayni A further 9km to the north the modem road along the left bank used to come to an abrupt end at the village of G6zeli, formerly Ayni (35km upstream from Zeugma). The village itself showed no obvious remnants of antiquity apart -possibly- from another aqueduct, still bringing a substantial flow of water from a side valley. There were no pre-modem buildings, although some cave dwellings were evident, possibly of great antiquity. However, there were a number of signs that this also constituted an important crossing point for a long period, despite the steep and rocky descent to the river on both banks. A route mounting from the village towards the east passes through a section, widened perhaps in modem times, but hewn straight from the rock and lined by ancient tombs on several levels sometimes linked by internal stairs- at the side of the road. Further up, this road divides, with one section about 2m in width (visible on the KVR-1000 photos) passing due east towards the abandoned village of Hamurkesen, which possesses ruins of a large building as well as a stone sarcophagus, underground rooms and many 'pressoirs', ancient presses probably intended for grapes. The second branch- now partly covered by the modem jeep-track but visible on the plateau in the form of a long line of upright stones bordering the ancient roadway - heads southeast to the village of Saylakkaya (Cibin) which has about 100 large abandoned cisterns cut into a broad limestone plateau lying northwest of the village, some at a considerable distance. It was deduced that this was a stopping point for caravans awaiting their turn to cross the river below. This village was formerly the site of a substantial church (as reported by both Pococke 1743: ch 17; and Ainsworth 1888: 193).

    On the right (west) bank of the Euphrates, opposite and a little upstream from the village of Ayni but at the end of another diagonal bank of gravel which would allow crossing on foot at low water, there is an Assyrian inscription of Shalmaneser III, just south of a dry gorge and near the top of the cliff (Tasyurek 1979, who refers

    115

  • Anatolian Studies 2000

    Fig 13. Kenk. relief of Shalmaneser III (photo: J-B Yon)

    to the gorge as 'Kenk'; fig 13); below lies a statue of a Roman river-god (fig 14) and an inscription from the time of Vespasian which speaks of construction by the legions of an 'opus cochli' (ILS 8903). Both inscriptions and the river-god will fortunately be above the level of the reservoir.

    The usual interpretation of the Greek word 'cochlias' (snail) is an Archimedean screw to raise water and this explanation has been given by Oleson for this inscription (1984: 55-6), but the height of the cliffs above (approxi- mately 300m) renders unlikely the explanation of a pump to provide irrigation water or for any other purpose. The material used would have been wood, and no traces remain of the mechanism. Although, as Oleson suggests, the current of the river might have been used to power the device, this also seems unlikely given the sometimes rapid fluctuations in the river level. An alternative expla- nation could be a mechanism to draw boats across the river: a crank handle attached to a screw would have turned a cogwheel and a drum with a cable. A similar mechanism would of course have been necessary on the far bank near Ayni village. However, no such mechanism is attested elsewhere and from the Middle Ages small ferry boats have operated on the Rhine (for

    example, at Basel) by a simpler means: an overhead cable is fixed at a high level to which the boat is attached by a sliding tether. It is sufficient to turn the rudder against the current for the boat to be drawn across. Such a system could well have operated here although no traces of a fixed cable were seen. At the time of Ammianus (see below) there seems in any case to have been another pontoon bridge at or near this point, but possibly 1km to the south.

    A more likely solution to the problem of the 'opus cochli' now seems to the first writer to be a cable-car. The volume of traffic in the area must have been substantial, given the bridges and other evidence cited below; there must therefore have been a serious problem in raising heavy loads from the river to the plateau, on which several large Roman settlements are known to have existed. There are remains of a substantial but steep zig-zag path supported on banks of boulders which leads up to the Assyrian inscription some 150m above the river and which might have been used by unloaded animals; there is also a rectangular edifice of large cut stones nearby whose purpose has not yet been identified. French (1994) thought that this building might be a reservoir; it seems to this writer that the building may either have contained mechanisms in the form of axles, cogwheels and crank handles to draw up the loads by cable from below or else that it did indeed act as a storage chamber for water which might then be used to counterbalance rising loads. A receptacle would of course have been necessary both for raising water from the river as a counterbalance for descending loads and then to facilitate raising them. The use of 'cochlia' with crank handles, cogs, screws and axles is attested in Hieron's Mechanica (11,18); the term is also regularly used in obscure circumstances in Biton's work on the construction of siege machines and catapults (probably for cables on drums). Unfortunately, the cable-car theory was not developed on the occasion of the visits made to the site; no axle-holes or other evidence of the use of the building can be cited, but many blocks had fallen down the slope in any case.

    Fig 14. Kenk: reclining river god

    116

  • Comfort, Abadie-Reynal, Erge?

    A general discussion of Hieron is included in Drachmann (1963: 19-141, but especially 60-1), who confirms that the endless screw in conjunction with a toothed wheel was invented by Archimedes about 250 BC. There appear to be no examples known from elsewhere of the type of practical machine proposed here, but it would seem strange that the simple technology of raising loads by cable, which is still widely used in mountainous regions today, was not used by the Romans, given that the theoretical basis clearly existed long before the time of this inscription.

    There are indications in the form of smoothed rock some 400m due west of the inscriptions that a route from the ford in the river and the Assyrian inscription led to an ancient settlement at Giimriikkiiyu (see below, near the modem village of Sarilar) and then on towards Yavuzeli (formerly Cingife). This track would have been used by caravans but not by wheeled traffic. However, the route to the top of the plateau, some 150m higher than the summit station of the cable-car if such it was would have passed northwest behind the Assyrian inscription and then up to Elif. This roadway, intended also for vehicles, passed over the hill and across the upper part of the ravine of Kenk, known locally as 'Midirin', via an apparent old caravanserai in caves at the locality of Altinmagara (see fig 9, map of the Euphrates valley from Halfeti to the Atatiirk Dam). It could then have continued either southwest to Giimriikkiiyu or north to the Roman bridge near Akbudak (Siipiirgiic) still extant on the Karasu. Remains of another Roman bridge were found in October 1999 on this tributary river at a point just 30m above its junction with the Euphrates, a fact indicating the existence of a regular north-south route also along the main river itself (see below, fig 15 and forthcoming article). A nearby inscription indicates that this latter bridge, at the hamlet of Habe~, was constructed by Legio IV Scythica (fig 16).

    The evidence for intense Hellenistic and Roman activity in this area was formerly limited to the large funerary monuments at the three villages of Elif, Hisar and Hasanoglu and to the road from Doliche to Samosata which passed through. This has now been supplemented by finds indicating religious centres at Kosk (fig 17) and at Giimriikkiiyu; for the former, a temple precinct remains well-preserved on a mountain ridge 8.2km west- southwest of Elif, while for the latter site some carved and decorated blocks on a mound adjacent to a well and an abandoned settlement with tombs, cisterns and quarries, near the course of the Roman road, seem also probably to have been the site of an ancient temple. In addition, abandoned towns were found near Hasanoglu and Elif, lacking fortifications but with large well-built houses possibly of the Roman period, and a hitherto

    undiscovered castle, probably of Hellenistic origin, was explored at Kaleboyu above Kenk. An exceptional concentration of four Roman bridges in a relatively small area also indicates heavy traffic at this period (Rumkale, Yanmca, Habe? and Siipiirgiic).

    The course of the main Hellenistic or Roman road which leads northeast from Doliche to the next Roman legionary fortress after Zeugma at Samosata was confirmed by the find in May 1999 of a watchtower (fig 18), 4.5m2 and still relatively well-preserved, 22km due west of Rumkale and also by that of the fourth major bridge, in this case over the Merzumen tributary a kilometre east of Yanmca (a village which is in turn 9km east of the important 'hoyiik' and later Crusader castle of Yavuzeli/Cingife). The bridge has a single large curved span of 60 blocks comprising a semi-circular arch with a diameter of about 15m still in place over the river (fig 19). The bridge is 7.1 m wide and fully comparable with the other large Roman bridges in the area to the northeast on the Karasu and G6ksu tributaries. (It appears on the plan of the Roman roads in this area kindly provided by David French in advance of publication of his article, but does not yet itself appear to have been the object of publication.) A quarry some 500m to the west is the obvious source of stone for construction of the bridge and has curious relief figures of an eagle and a god or goddess with upturned hands, one of which has enormously long fingernails. There are also remains of at least two abandoned villages near the bridge, one of which has a necropolis in which was seen a tomb with a bull's head with long horns.

    The modem track descends by a fairly gentle gradient from Sarilar to the river opposite Ayni and at a point 1 km to the south of Kenk gorge. It too might have Roman origins, since Ammianus speaks of destruction of pontoon bridges both at Zeugma and 'Capersana' (18.8.1), but this may have been on the site of the river crossing at Kenk discussed above, since the road could then have continued north along the right (west) bank. Passage along the river bank opposite Ayni is not possible. Capersana was thought by Dussaud (1927) to lie at Ayni and no other point between Samosata and Zeugma seems more suitable, although it also seems possible that Capersana was a name originally applied to the village of Saylakkaya (formerly Cibin), above Ayni on the plateau. A pontoon bridge at this point presumably implies wheeled traffic. It now seems likely that the Roman road to the west would have passed due north from Kenk/Midirin along the Euphrates, over the Karasu bridge near its mouth at Habe? and then up the easy climb to the village of Tarlabail on the plateau, whence it could join the nearby Germanikeia-Samosata road without difficulty.

    117

  • Anatolian Studies 2000

    Fig 16. Habe?: Roman inscription

    Fig 15. Habe?: bridge at the junction of the Karasu with the Euphrates

    Fig 17. K6sk: temple precinct wall

    Fig 18. Watchtower on the Doliche-Samosata road

    Fig 19. Yarimca: Roman bridge

    118

  • Comfort, Abadie-Reynal, Ergeq

    Dussaud also identified Capersana with 'Caphrena' (1927: 459-60), but this link is denied by others (e.g. Dillemann 1962: 169-70). Caphrena is mentioned by Pliny as a satrapal capital lying to the east of Zeugma but in his time reduced to the status of a citadel (NH 6.119). No such place has yet been found to the east of Zeugma. Ayni lies clearly to the north rather than east and Dillemann would seem to have been certainly right in denying this association were it not for some finds in autumn 1998 which appear to confirm the past impor- tance of this region in some periods.

    A small village called Kayalar, formerly known as Keferhan and about 10km north of Ayni, lies on the plateau near the left (east) bank and in the bend which the Euphrates makes coming from the east as it turns south- wards. This village was stated by a local village mayor to have been a centre of caravan traffic until recently. Local people also reported that the village used to have four churches (ruins of one are still visible), together with quarries, tombs, cisterns and many pieces of green glazed ware (which are difficult to date but probably late medieval). The village was visited by Guyer (1916), who subsequently crossed the river at G6zeli/Ayni. He published a photograph of another of these churches, then well-preserved and more similar in style to fifth century churches of the Syrian limestone massif than to that still standing in the nearby town of Nizip, which is declared by Sinclair to be 10th or 11th century (1990: 122). Guyer believed this town, not Ayni or Cibin, to be Capersana, a view already indicated by Kiepert on his map of the area (1913). He also mentions the ruins reported in this area to Chapot in 1901 (1907: 271); the latter however mentions them as lying to the east of the village of 'Boser', now Bozyazi (see below). It seems possible that by the end of the 19th century the village of Kayalar/Keferhan had been deserted. Much of the cut stone has been removed since then, but there are still substantial quarries and cisterns to be seen, as well as the remains of some subterranean rooms.

    The similarity of names (Keferhan and Caphrena) is beguiling but may be misleading because the Aramaic word 'Caph-' meaning town is commonly used in place- names. No evidence was found to indicate a satrapal palace at or near this village nor was any found on the satellite photograph, but closer investigation of the area is certainly warranted. A further village on the cliff-edge above the Euphrates and between Kayalar and Ayni on the east bank provided strong evidence of an important route along the Euphrates itself. Another small castle below this village- Bozyazi, formerly Bozyr- is built on a cliff immediately above the river and apparently guarding a path from the riverbank which ascends to the plateau at this point. Cisterns in the gully followed by

    the path from the north, as well as the castle itself, seem to show that caravans coming along the Euphrates have mounted the cliffs here before descending once again to cross the river, probably at Ayni, or else continuing east to Edessa. Neither of the two castles mentioned will be directly affected by the dam and both will be discussed in the forthcoming article on north-south routes in this region.

    A final reflection in regard to this crossing point concerns Arrian's reference (3.7.3, quoted above) to Alexander having crossed the Euphrates by two bridges and to his line of march having followed the Euphrates and the Armenian mountains on the left. Since the outlying mountains of the Taurus range are visible from the higher ground above Ayni on the east bank of the river and only a route crossing at this point could be said to follow the Euphrates 'on the left', then Ayni must also be a possible crossing point for Alexander's expedition, especially if two columns of advancing troops were involved.

    Other crossing points The crossing at Carchemish must have been important during many periods. Sadly, it lay outside the more detailed of the satellite images available for this project and it has not been possible to examine closely the terrain nor to study detailed maps of this area. The map entitled 'The Euphrates valley from Halfeti to the Atatiirk Dam' (fig 9) shows the position of Carchemish and adjacent mounds. Requests to visit Carchemish were refused by the military authorities and access to the east bank opposite the town was rendered difficult by the construction work on the dam 5km upstream and by the proximity of the Syrian frontier which, for the stretch along the Euphrates as far as the dam, lies only about 500m to the east of the river. The usual identification of Carchemish with the Seleucid foundation of Europus has been put in doubt recently (Gaborit, Leriche 1999: 195- 6). Nor has any evidence been found to establish the whereabouts of the city of Amphipolis-Nikatoris believed to lie opposite Europos (Jones 1937: map facing 217 and 219): careful scanning of the terrain from the opposite bank just north of Carchemish revealed no likely spots. Stephanus of Byzantium in his alphabetical list of place-names indicates the local name of Amphipolis as Turmeda; Jones links this to the name of a modem village called Zurme (1937: 441, n 3).

    Algaze's survey of both banks north of the frontier in 1989 found no remains opposite Carchemish, although Hellenistic sites were identified by him on the east bank at Harabebezlkan and Kiiciuk Kale Tepe to the north, while recent excavations of a site at Duyduk village, only 6km to the north and partly eroded by the river, have also

    119

  • Anatolian Studies 2000

    revealed Hellenistic levels, but more probably of a village than a town. It seems possible that the village of Beilune (Zurme?) on the Syrian side of the frontier is a more likely candidate for Amphipolis/Turmeda and that the route crossing the river and leading east towards Anthemusia/Batnae (Siuriic), Edessa and Harran passed where the existing rail bridge that constitutes the border between Turkey and Syria stands.

    An interesting ancient cemetery, overlain by a modem one, just to the north of the citadel of Carchemish was investigated and revealed a series of approximately 40 strange stelae with large square niches surmounted by three round cupules (fig 20). The stelae do not appear to have direct parallels elsewhere, although stones with three square cupules may be seen in the Museum of Damascus. They are not mentioned in the published volumes on the excavations at Carchemish (Hogarth 1914; Woolley 1921). No direct information was obtained about the ancient crossing or the roads in the area; currently there is a rope-drawn ferry to an island just upstream of the citadel and of the railway bridge which forms the frontier, but there is no sign of a ford. The general issue of crossing the Euphrates at this point is discussed by Winter (1983).

    In the opposite direction, 22km to the north of Zeugma, the small town of Halfeti on the left (east) bank constituted until the beginning of this century an important administrative centre for villages on both sides of the river including the castle of Rumkale. It appears to have grown in importance as Rumkale declined but shows evidence in the form of ancient stonework and quarrying beneath ruins of modem buildings to indicate earlier importance.

    Fig_~~~~ 20.w:~~~~ Carchem~: stl icetenrhotecta .e. : e Fig 20. Carchemish: stele in cemetery north of the citadel

    Fig 21. Crossing the river near Ayni

    The lowest part of the town, including some fine Ottoman baths, has been drowned by the reservoir but fortunately most will be unaffected. Although the ford at Rumkale lies upstream only 3 to 4km from Halfeti, the calm waters immediately in front of the town were themselves crossed regularly by a small ferry, which is even now indicated as such on the Turkish 1:100,000 maps. Even before the dam was completed, this ferry ran irregularly and might have carried a horse but no mecha- nised form of transport (fig 21). However, it must have been from Halfeti that Freya Stark was carried upstream on a flat-bottomed boat with her companions in order to visit Rumkale in 1908 (Stark 1966: 110). Hogarth (1925: 66-8) also crossed the Euphrates here by ferry some time before 1896 at a moment when the river was in spate and by an identical process to that described above by Soane for a crossing near Dipsa. Hogarth then continued northeast to recross the river at Samosata.

    There is still visible both from Halfeti and on the satellite photograph an ancient mule track leading up to the plateau from the village of Kalemeydan opposite the town (on the right, west, bank) and this track does not coincide with the modem jeep-track. It is possible that this track which would have crossed the north-south

    120

  • Comfort, Abadie-Revnal, Erge?

    ancient road on the plateau at the village of Koseler was part of an important route linking Halfeti to the west at Doliche and later Aintab (Gaziantep). The route follows a basalt ridge between deep gorges and passes several ancient villages built of basalt boulders, and even a town at a site known as Karaburq, all of which have been wholly or largely abandoned. They are distin- guished by the presence of one or more artificial basins which often remain full of water into summer even today. In some cases decorated tombs from the Roman period have been found below the villages (especially at Karaburc and Ucg6l) which are excavated in the under- lying limestone, but little pottery was available to date these settlements. At Karaburq some fragments of mosaic were found. Further west, the route passes a necropolis with tombs of Hellenistic, Roman and later periods at Haciobasi, near the basalt village of G6bek (Archi 1971: 66 refers to this site as 'Ciftlik').

    Lastly, there used to be a regular crossing point for boats and rafts 9km to the north of Ayni in the curve of the river and close to the mouth of the important Euphrates tributary, the Karasu. The latter river is in turn crossed by the newly-discovered Roman bridge at Habe, (about 50km above Zeugma). This tributary river passes through the Araban valley to the west of the Euphrates, a valley which has several important mounds, the substantial remains of the other Roman bridge mentioned above and an ancient settlement below the village of Suiipiirgiic (Akbudak)- only 3km from the junction with the Euphrates and still graced by a rock relief. Initially published by Burney and Lawson (1959), this was thought by Hellenkemper and Wagner (1977) to be Hittite, but Geoffrey Summers has expressed doubt on this attribution (pers com). Happily we are able to report that the relief, although damaged in the dynamiting mentioned at the end of the latter article, is still largely extant (fig 22).

    In the 19th century travellers reported that it was common practice to cross the river here on skins (Chesney 1850: see map 'The River Euphrates with the Cilician Taurus and northern Syria' which draws on an expedition along this part of the river by a Lieutenant Lynch in 1836), presumably when the water was too deep to be forded. As described above in relation to Ayni, this route mounted to the plateau at the village of Tarlaba?i. Ruts are reported to have been visible in the natural rock surface along this road until a recent re- surfacing.

    The nearby village of (Ciftekoz, also on cliffs above and to the northwest of the river Euphrates, has a further track descending to the river along which ancient tombs are visible in great numbers. There are also cisterns and a large underground square room apparently used as a

    mill (big millstones are still in place). The ground is carpeted with potsherds, as at the old settlement of Siipiirgiiuc, and a steep path leads down extremely worn and ancient steps to a ford of the Euphrates at a point at which it is enclosed on the north by cliffs. On the left (south) bank a cart-track descends to the ford from the village of Sirata?lar (formerly Vahne), on whose outskirts lies another remarkable concentration of cisterns. Just by the ford, ruins of a large building are visible on the south bank which may be those of a han or caravansaray.

    There are many caves in cliffs along the north bank, some enlarged to form dwellings on several levels linked by an external staircase. Most of these will be above the reservoir level, but there are also signs of buried tombs or inscriptions below this level in the form of 'runnels' carved into a cliff face at seven points. These runnels were apparently intended to deflect rainwater from entering the tombs or other structures beneath and hollowed from the cliff, but which are now covered by a bank of earth. Although progress along the bank towards the east seems impossible because of the cliffs, there is a pathway visible from the south bank with stairs at one point, but which is now disused.

    The crossings at Tarlaba?i and (iftekoz do not at first seem likely to have been associated with heavy caravan traffic, but it was reports among villagers of caravans following routes along this section of the river which led to the discovery of the Roman bridge at Habe?. This bridge, the associated road and the evidence found in 1999 for other north-south crossing points further east along the river will be addressed in the forthcoming article on north-south routes along the valley.

    Fig 22. Siipiirgiiu. relief of the God on the stag

    121

  • Anatolian Studies 2000

    Discussion Many of the ancient roads discussed here are difficult to date. The principal users of the routes mentioned in the preceding text were probably always commercial caravans whether using camels or donkeys as beasts of burden but the crossing points will also have had an important military significance at several periods, as well as a possible role in the state courier services which operated from the Assyrian period onwards. Wheeled vehicles were essential for Roman and Hellenistic armies, but disappeared entirely from the Middle East from about the time of the Arab conquests (Bullitt 1975). Unless there is convincing evidence in the form of Roman bridges or wheel ruts it has been assumed by us that transport involved pack-animals, not carts.

    Assyrian caravans passing between the Anatolian trading stations and their homeland could have used any of the crossings described here, but the existence of the late Assyrian relief at Kenk/Midirin, and possibly of the relief at Siipiirgiic, must surely imply the likelihood of an important crossing at Ayni. The assumptions of Astour and Beitzel that the main crossing point was at Zeugma or Birecik during earlier periods may therefore be insecure.

    The lack of sites with convincing evidence of occupation in the Achaemenid period makes identi- fication of Persian crossings difficult. But the arguments of Gawlikoski seem sufficiently convincing to the first writer to support the assumption that Zeugma was the principal point of crossing in this period. Achaemenid pottery is apparently difficult to distinguish from Hellenistic, and Stein, for one, declares that the Uruk site of Hacinebi 5km downstream from Zeugma had an 'Achaemenid/Hellenistic occupation (ca. 500-100 BC) over the entire area of the mound' (1999: 119). The castles at Bozyazi and Kaleboyu, mentioned above, may prove to be of Persian origin and this would imply a crossing also at Ayni.

    In the Hellenistic and Roman periods, Zeugma was clearly the central point, but if the identification of Ayni with Capersana is accepted, then the Roman monuments at Elif, Hisar and Hasanoglu and the other evidence cited above show that Ayni too was of great importance. Apart from the principal crossing at Zeugma (and later Birecik), it seems probable, for example, that the east- west route at Ayni formed part of a regular network of links during the Hellenistic, Roman and perhaps Byzantine periods between cities to the west such as Doliche and Mara? (Germanikeia) and those to the east, including Urfa (Edessa), Harran, Diyarbakir (Amida) and Nisibis.

    More recently, caravans passing between Aleppo and Mosul seem to have used several crossing points even to

    the north of Zeugma despite the long detours involved. Besni, a town with remains of Seljuk mosques and situated north of Araban, as well as Rumkale itself, are known to have been focuses of power in the Middle Ages. (For a discussion of the history of the area in this period see Sinclair 1990: 199.)

    In regard to periods before the Arab conquests, the significance of all these crossings should not be under- estimated; throughout the second and first millennia BC many of the eastern developments in religious thought, philosophy and science which had a major impact on the Mediterranean world must have reached Ionia and the west via routes such as those discussed here. The exchange of goods in the important cities of northern Syria was accompanied by an exchange of ideas which intensified when this area became the border between Rome and Parthia. It is sad that some sites where such exchange took place have now vanished and no further opportunity is available for finding material traces in the valley of the river Euphrates of this interchange between East and West.

    Technical annex Satellite imagery used in this project included:

    KVR-1000 Black and white photos are available in various formats up to 40 by 40km. Prices currently range from $2,500 for a 10 by 10km scene to about $5,000 for a 40 by 40km scene in digital format, but the basic product is negative or positive film in 40 by 40km squares. Ground resolution is 2m.

    KVR-1000 photos were initially obtained for this project from an intermediary in Potsdam, Germany. (Currently these photos are available exclusively from SPIN-2, a firm in the USA.) The first parallelogram of positive photographic film was purchased for $900 in 1996 and covered an area of approximately 10 by 25km around Zeugma. It dated from 11 June 1990 and was taken from a height of 230.1km with an f value of 1003.8. This was printed out by a professional photo- grapher on paper and various sections were enlarged for study. Subsequently the whole image was scanned using a drum scanner and it proved to be easier to process and study such images in detail when they had been turned into electronic files, in TIF or BMP format. The photos may now be obtained already in digital format. All of the detail - down to individual trees - was obtained in the scan of this first photo with a computer file of a size of 7MB. Larger photos processed subsequently proved much more problematic since the file size became too large for the PC to handle unless broken down into

    122

  • Comfort, Abadie-Reynal, Ergeq

    smaller areas. Two further photos from 22 May 1992 were obtained for ?4,000 in 1998 one of the area to the north of Zeugma including most of the sections of the Euphrates valley to be inundated (40 by 40km) and one of Zeugma itself and the south (15 by 15km), but sadly not quite stretching to Carchemish.

    The technical characteristics of the KVR-1000 film are as follows: black and white panchromatic film, 640- 70nm spectral response, available in 40 by 40km subscenes (but also for smaller areas). Extracts may now be ordered over the internet from the archive of Terraserver (see http://terraserver.microsoft.com), but this currently has very little in it from the Middle East; photos are nevertheless usually available, but have to be ordered from Sovinformsputnik via SPIN-2 (http:/www. spin-2.com) or Core Software Technology (fax 001(626)796-8574).

    The large-format camera is mounted in a recoverable module on the Cosmos series of remote sensing satellites and is discharged at intervals of 45 days, descending in the final stages by parachute to a landing site in Kazakstan. It provides high-resolution photography taken from an average height of 220km with average ground sampled distance of 2.5m. This may be enlarged to a scale of 1:10,000 if clear atmospheric conditions existed at the time of the photograph. It seems currently to be the most detailed satellite photography com- mercially available despite the plans of various US-led consortia to provide photographs with both higher resolution and greater spectral range.

    The Russian supplier is Sovinformsputnik and it seems to be able to provide photography for all of the Middle East (availability was confirmed for six sites in Turkey and Syria).

    Corona Corona photographs from US reconnaissance satellites have been released since 1995. However, the photo- graphs concerned come from programmes which ended in 1972 and the quality in terms of spatial resolution is poorer than for the KVR-1000. It is possible to study thumbnail versions on the internet before ordering, in order to ensure visibility; a sophisticated map system is available for identifying areas of interest and availability (as for the Microsoft Terraserver site and the KVR-1000 photographs, but with a much wider range of available photographs for the Middle East). The website concerned is that of the US Geological Survey http://edc.usgs.gov/webglis.

    Like the Russian photographs the film was ejected from the space platform and then recovered in capsules, but in this case by aircraft snaring the parachutes. The

    ground resolution varies according to the 'mission' concerned but is usually between 3 and 5m.

    Photographs take the form of long strips covering areas roughly 20 by 300km and may be ordered on paper or on film. For this project paper prints were used but the quality of image available from the photographic film is superior and the extra cost is surely worth paying. Current prices are $18 for each photograph on film and $14 for the paper prints. Although the prices are much lower than for other forms of space imagery discussed here, it should be remembered that the format in which the photos are available precludes analysis by computer of a homogeneous area stretching north or south of a single strip, since they are difficult to merge. On the other hand, for researchers interested in a specific site, or not intending to conduct computer analyses, they constitute the obvious choice being both cheap and easily available (Kennedy 1998a and, for Zeugma, 1998b: 30).

    SPOT Both multispectral and panchromatic images are available from SPOT Image, which also offers the possi- bility of special twin photos for assessment of topog- raphy by stereoscopy. The latter possibility would be of great interest to archaeologists but is very expensive and the offer to progra