Richard Bayliss and James Crow
Water Supply of Constantinople 2003

Our aims in the 2003 field season were to investigate in greater detail three distinct elements of the Byzantine water-supply system in Thrace. Our research over recent seasons has enabled us to review and reassess earlier interpretations of the complex pattern of aqueducts and water channels still extent in the countryside to the west of Istanbul. The fieldwork undertaken by the late Prof. Kazim Cecen and published in The Longest Roman Water Supply Line (Istanbul, 1996) provided the first coherent interpretation of the Thracian water supply for Byzantine Istanbul. Our own fieldwork and especially the research carried out over the last three field seasons has enabled us to develop a fuller understanding of the distinct phases of the creation of the system between the fourth and sixth centuries AD. In summary our current model envisages that the initial phase of the Byzantine water supply line was constructed in the later-4th century during the reign of the emperor Valens. The spring sources for this were a group of karst limestone springs located west of Catalca, close to the villages of Danimandara, Gumspinar and Pinarca (see Geology in Bayliss, Crow and Bono 2001.) a total length of about 180 km, equivalent to figure of 100 stades quited in the Themistius oration on the Aqueduct of Valens. Subsequently the system was extended to its full length of at least 280 km. as far as Vize and incorporated the major springs at Ergene and Pazarli. This phase not only included a major extension of the system, but at the same time as the line was lengthened most, if not all, of the earlier aqueduct bridges were replaced to support the two channels. This second major phase may be dated on the basis of art-historical evidence to the first-half of the fifth century (see Kursunlugerme below).

Stage 1 of the fieldwork was conducted in the Vize district of Kirklareli province. Our aim was to visit all the sites noted in Cecen’s survey and to study in greater detail the structural evidence from the sector. We were able re-visit a length of channel 0.68 m wide at Gazi Mahale in the west suburbs of Vize. This enabled us to establish a relative elevation for the water channel in the Vize sector. West of Vize the main source is located at the powerful springs at Pazarli. An old channel leading from the spring and series of modern pools were restored about fifty years ago using modern cement. This channel retains however the essential features of the Byzantine narrow channel although in places the vault has been replaced with concrete slabs. Villagers informed us that the ruined mortared walls located near the spring source were the remains of old demolished water mills. On the west side of the valley, on the opposite side to the restored channel, we were shown rock-cut containers excavated in an outcrop of limestone on the valley side. Although they are not part of the main water supply system they appear to be of antiquity and are similar to the rock-cut storage tanks for olive-oil production found in southern Turkey, however in Thrace an alternative use needs to be considered.

To the east of Vize the channel follows the edge of a limestone escarpment as far as the springs at Ergene. We identified the first aqueduct bridge east of Vize at Kemerdere (Cecen 1996, K1) surviving as two substantial abutments. Traces of the channel could be traced eastwards through thick scrub. North-east of Okcular the remains of the west and east abutments of a ruined aqueduct were located across the valley of the Indere. The line of the channels could be followed to the east and a raised embankment was noted where it crossed a gully. West of Evrenli the ruined remains of a large bridge were seen across the Akpinar Deresi. Having crossed the deep valley the monumental east abutment is aligned on a tight 900 turn. On the escarpment east of Cakilli we were able to locate a series of tunnel openings as the line swung north-east along the valley of the Ergene Dere. We revisited the Ergene spring, first seen in1998, a major tributary source for the line from Pazarli. The remains of a low embankment and water channel were located in subsidiary valley and the main aqueduct (Cecen K5) across the Ergene Dere was identified. We had formerly considered that aqueducts in the area of Ayvacik (Cecen K 7) showed signs of reduction from a broad to a narrow channel and we had interpreted this aqueduct in a similar way (Anatolian Archaeology, 4, 1998, 19-20). Further study indicated that there was evidence for two narrow channels across east embankment, rather than a radical change in width as first thought. We remain uncertain whether this shows either two contemporary channels from different sources at Pazrli and Ergene, or two subsequent channels not in use at the same time. Beyond the embankment to the east the line could be traced as a broad platform towards the Ayvacik dere.

The second area of our fieldwork was to complete the study of the mason’s marks and other symbols on the aqueduct at Kursunlugerme. We were able to complete the study within two days, however in course of this work we encountered very worrying evidence of an attempt by treasure hunters to demolish one of the aqueduct arches. Fortunately they had only recently started and by contacting the local Jandarma we were able to prevent any further damage. We did however recognise new damage on the west face on the south side of the aqueduct and it would appear that stones decorated with crosses and other Christian symbols were especially targeted. This is similar to the serious vandalism noted in 2001 at Karamanoglu Dere (Cecen K13) near Aydinlar village.

The third area of study was the spring at Pinarca and its association with the main water supply line to Istanbul. This is a major karst spring located to the west of Ihsaniye and within the line of the Anastasian Wall. We had studied the spring and the immediate channels as part of our survey of the wall but only in 2002 did we begin to understand how it was connected to the main water supply to the city (see map). We are especially grateful to Mr Andy Byfield and Dr Caroline Finkel for their advice on aqueduct at Kalederesi. This is a major surviving aqueduct bridge located north of Ihsaniye on the Kalederesi. Although the main arch has collapsed the two abutments still survive 15-20 m. above the stream below. Of particular significance was the presence of a second, presumably earlier, aqueduct constructed of smaller blocks and located about 200 m. to the west. This sequence matches the paired aqueducts found to the east of Baligerme and reflecting the development of the system noted in the introduction. East of Kalederesi we were able to identify the remains of small ruined bridges east of Akalan at the Buyuk and Kucuk Kameraderesi. Using GPS we were able to show that these points are part of the same line of channels noted last year close to Subasi. North-east from there it would be possible for the Pinarca line to join the main system between Kalfakoy and Dagyenice.

West of Kalederesi the remains of an aqueduct abutment were located in the Kurudere and sections through the channel were seen in exposures revealed by recent road construction between Ihsaniye and Pinarca. The evidence we observed this season demonstrated that the springs at Pinarca were another major source for first phase of the water-supply system in the 4th century. It further confirmed our earlier suggestion that the Byzantine water system in Thrace was not a single line but that it consisted of a dendritic network of channels developed over a century or more. An outstanding question is to understand how it was maintained and modified over the later five centuries of use.

We wish to thank the General Directorate of Antiquities and Heritage in Ankara for permission to continue our survey and we are especially grateful for the enthusiastic support from Fahriye Bayram our representative. We were warmly received by staff in the Kirklareli and Istanbul Archaeological museums and we would like to express our gratitude for their continuing interest in our project.