Research article

The water supply of Constantinople:

archaeology and hydrogeology
of an Early Medieval city
P. Bono á J. Crow á R. Bayliss

tures expected of an eastern classical metropolis: the fora,

Abstract This paper describes the longest ancient baths, colonnaded streets and a hippodrome (Mango 1990;
aqueduct system in the Mediterranean World, built Ricci 1999). At a time when many western cities stagnated
in the 4th century A.D. to supply the city of or shrunk in size, Constantinople expanded and ¯ourished
Constantinople, then capital of the Byzantine in both cultural and physical wealth.
Empire. Results are presented from ongoing inves- Only fragmentary remains survive from the late antique
tigations in the city and hinterland of modern and Byzantine city because much has been lost or lies
Istanbul by an international team of archaeologists buried beneath later Ottoman and Turkish Istanbul. Yet
and hydrogeologists. The hydrogeological setting of the fundamental necessities for urban existence in the
the region resulted in the development of a unique early medieval world, religion, security and sustenance, are
water supply system, which survives both as represented in three of the city's greatest surviving mon-
archaeological remains and in the writings of uments. The ®nest achievement of the 6th-century em-
contemporary historians. peror Justinian, the church of Hagia Sophia, still
dominates the Istanbul skyline and the western extent of
Keywords Aqueduct á Constantinople á Karst á the old city is marked by the Land Walls of Theodosius,
Water resources commenced in A.D. 415. The third structure is the Bo-
zdogÆan Kemeri (Fig. 1), a great water bridge of around 86
arches spanning one of the city's busiest highways and
normally identi®ed as the Aqueduct of Valens (Dalman
Introduction and others 1933; Mango 1995). It remains the most con-
spicuous surviving element and engineering feat in a vast
Constantinople, renowned and admired as a centre of and complex system of water channels and aqueducts,
Christian art and culture until the 15th century, was which supplied the city with water from sources in Thrace,
founded in A.D. 330 by the emperor Constantine on the the region commonly de®ned as `European Turkey'
natural boundary between Europe and Asia. The emperor (Fig. 2). According to a contemporary source the 4th-
extended the walls of the old city of Byzantium and within century emperor Valens welcomed the Thracian nymphs
three decades the urban population had begun to expand and waters to the thirsty city, transported there by the new
signi®cantly. Taking on the mantle of old Rome, the city overground and underground river (Themistius, Orationes
became a focus for imperial patronage and display, ac- 13. 168a-b; Mango 1995, p. 13).
quiring within a century many of the grand urban struc- At over 250 km, it is the longest water supply line known
from the ancient world and it remains one of the greatest
achievements of hydraulic engineering. It has however
Received: 30 January 2001 / Accepted: 8 May 2001
been largely ignored in all standard accounts of Roman
Published online: 1 August 2001 hydraulic engineering (see Hodge 1992). The principal
ã Springer-Verlag 2001 reason for this is that for much of their length, the water
channels run through dense and inhospitable forest,
thereby effectively deterring archaeological investigations
P. Bono (&) until recently.
Dipartimento di Scienze della Terra, More than 30 stone water bridges and many kilometres of
CNR ± Centro di Studio per il Quaternario e l'Evoluzione
Ambientale, UniversitaÁ `La Sapienza',
underground tunnels carried the water over mountain and
P. le A.Moro, 5-00185 Roma, Italy plain from the plentiful springs of the Istranja mountain
E-mail: bono@uniroma1.it range near Vize directly to the heart of the city (Mango
Tel.: +39-6-4994925 1995; CË ecËen 1996a). Such was the magni®cence of the
Fax: +39-6-4454729 undertaking that it even appears to have received its own
J. Crow á R. Bayliss popular mythology so that medieval writers claimed that
Department of Archaeology, University of Newcastle, its source was the great Danube River (Dirimtekin 1959).
Newcastle upon Tyne NE1 7RU, UK They were not far wrong. The known system is at least two

DOI. 10.1007/s002540100365 Environmental Geology (2001) 40:1325±1333 1325

 Research article

Fig. 1
The Aqueduct of Valens, Istanbul

Fig. 2 In many respects the completion of this new water-supply

Schematic geological and hydrological map of Thrace system inaugurated and con®rmed the city as the new
capital of the Roman world, the New Rome. Not only was it
built to ful®l the daily needs of the growing population,
and half times the length of the longest recorded Roman but it also supplied the great bathing establishments
aqueducts at Carthage and Cologne, but more signi®cantly (thermae) and monumental fountains (nymphea) expected
it represents one of the most outstanding surveying in any classical metropolis (Crow and Ricci 1997). To the
achievements of any pre-industrial society. resident emperors, the water supply was not simply a

1326 Environmental Geology (2001) 40:1325±1333

 Research article

functional requirement for sustenance; it was a symbol of

wealth, power and continuity. The enormity and com-
plexity of its construction demonstrates the absolute
con®dence in the future prosperity of the city maintained
by its benefactors.

The Archaeology of the Thracian

hinterland
Outside the city the archaeological reminders of this water
supply system survive in the forests of Thrace as impres-
sive aqueducts and subterranean water channels. Very
little work had been carried out on the Thracian aqueducts
until a recent publication by the hydrological engineer
Professor KaÃzim C Ë ecËen, whose investigation of the long-
distance system demonstrated the enormity of the con-
struction (C Ë ecËen 1996a; see also important earlier studies
by Oreshkov 1915, pp 88±103; Dirimtekin 1959, 1968).
C
Ë ecËen also produced the ®rst map, showing in outline, the
course of the water supply line based on interpolation
from topographical maps and the fragmentary remains he
observed. Although the scale of the construction has been
recognised a more comprehensive understanding of the
entire system was wanting.
The Anastasian Wall Project, established in 1994, has
contributed signi®cantly to our knowledge of this long-
distance system in the vicinity of the city's 6th century
outer forti®cation, the Anastasian Wall or Long Walls of
Thrace (Fig. 3). The project was inaugurated to investigate
the archaeology of the city's hinterland, with the primary
objective of studying the remains of the 56-km-long linear
forti®cation that once secured the Thracian peninsula.
This was located some 60 km west of the 5th century walls Fig. 3
Map of central Thrace, showing the water supply system and the
of the city and protected the hinterland of the Constanti- Anastasian Wall
nople for over a century (Schuhhardt 1901; Crow 1995;
Crow and Ricci 1997). The associated investigation of the the channel masonry is usually made of metamorphic or
water supply system in the region has gathered pace with limestone blocks.
each successive season and much of its complexity has In the vicinity of the Anastasian Wall the landscape is
now been revealed (for archaeological reports see Crow more undulating and dense forestation has helped to
and Ricci 1995, 1996a, 1996b, 1999a; Crow 1997, 2000; preserve both the Wall and the aqueducts. Present ®eld-
Crow and others 1998, 1999b; Bayliss 1999; Bayliss and work has focused on the water system in this region where
Crow 2000; Crow and others 2000). the existence of a second channel of a narrower gauge has
The long-distance line of the aqueduct system originated been discovered, complete with its own massive aqueducts
at springs just west of the small town of Vize (Bizye) and and located at higher level than the main line. The authors
despite the attribution of the BozdogÆan Kemeri to the have been able to date this high-level system to the 5th
emperor Valens (A.D. 364±78) construction work probably century (Crow and Ricci 1997, pp. 257±258).
dates from A.D. 357, in the reign of Constantius II (Mango The main source for this high-level system has been located
1995). The fragmentary remains of water bridges survive around PapucË in the Mandara Dere (Fig. 3), where a large
in the ®rst 200 km of the aqueduct's length and the cave and spring have been located and two separate supply
channels have been identi®ed in numerous locations. channels survive along the side of the valley above the river.
These are found snaking around the contours of the hill- Further along the line in the KursËunlugerme valley it is
sides and were normally 1.60 m wide with a height of possible to observe the high-level channel running at about
1.70 m. They were constructed by `cut and cover' tech- 6.45 m above the lower main line and crossing the valley on
niques and were roofed with rough stone vaults. In places a remarkably well-preserved aqueduct (Figs. 4 and 5),
the hydraulic lime-mortar lining of the sides of the tunnels which was built across the top of the lower channel (Fig. 6).
survives with later accretions of distinctive travertine or The fragmentary remains of an earlier aqueduct for the
sinter deposits. Both the hydraulic mortar plastering and lower channel still survive to the east.

Environmental Geology (2001) 40:1325±1333 1327

 Research article
Fig. 4
KursËunlugerme aqueduct from the east
Further east towards Istanbul the authors discovered
parallel aqueducts in the dense woodland below Kemi-
kharman Tepe, but until the 1998 season it remained un-
clear how far both high- and low-level systems continued
to the east before they reached a similar elevation. At
BuÈyuÈkgerme near CË iftlikkoÈy however, it was observed that
only a single aqueduct crosses the steep valley (Fig. 3). To
the north-west and upstream from this aqueduct, parallel
water channels running at approximately the same height
were recorded for the ®rst time. These were preserved
below and alongside a forest track, which had utilised the
level platform created for the construction of the two
tunnels. The outer tunnel was broad, with a width of 1.6 m
and a surviving height of 1.7 m, making it comparable
with the wide tunnels on the low-level system. Beside it ran
a second narrower tunnel, which was apparently slightly
higher although its base was in no place visible. This was
only 0.9±1.0 m wide and can, therefore, be compared with
the narrow, high-level channel. It was clear that the broad
lower tunnel was the ®rst to be constructed, before the
second narrower tunnel was cut in to the hillside beside it.
This evidence demonstrates that the high-level system
gradually converged with the earlier channel over a dis-
tance of around 15 km, between KursËunlugerme where the
1328 Environmental Geology (2001) 40:1325±1333
two systems are visibly separate and BuÈyuÈkgerme where
the channels cross the same aqueduct. It is not yet clear
however whether the later system actually ever fed directly
into the main line, or whether it existed as a separate
channel all the way to the city. The latter method appears
to be a more widespread solution, as seen particularly
clearly at the Porta Maggiore in Rome, where the Anio
Novus crosses this converted aqueduct above the channel
of the Aqua Claudia (Steinby 1995, pp. 304±305).
Whether or not the high-level channel was therefore
simply a tributary for the main line, it is clear that other
smaller tributaries from local spring sources fed into both
the main channel and also the high-level channel itself.
These smaller feeders were observed in the KursËunlugerme
valley (Fig. 5) and also in the vicinity of BuÈyuÈkgerme. The
latter was discovered in a valley to the west (upstream) of
the main line and was a narrow-gauge channel with its
own aqueduct.
One of the key issues that has been targeted in the inves-
tigation of the water supply is the question of the temporal
relationship between the two systems, in particular to
understand whether the 5th century high-level system was
built as a replacement for the earlier long-distance line. In
1997, an inscription was discovered on a remote aqueduct
at Elkafdere (Fig. 3), which named Longinus, the Prefect of
Constantinople under the emperor Justinian (A.D. 527±
565). The authors deduced that this aqueduct was part of
the 4th century low-level system and were therefore able to
conclude that the low-level system continued in use for
some time after the construction of the high-level system
in the 5th century and that both presumably functioned
simultaneously, in the 6th century at least.
Closer to the city the situation becomes much more
complicated as supply lines converge from the various
sources to the west and north. One of the major compli-
cations in addressing issues of chronology, sources and
destinations is caused by the fact that several different
spring sources have been exploited throughout the history
of the city and the early lines were often rebuilt and ren-
ovated. The Valens Aqueduct itself carried a number of
channels dating, it seems, from various different periods
Fig. 5
KursËunlugerme aqueduct; schematic plan
showing the position of high- and low-level
water channels

Fig. 6
KursËunlugerme aqueduct; north±south section showing low-level
channel passing beneath arch of later high-level aqueduct
(C
Ë ecËen 1996a). The closer sources in the Forest of Belgrade
and at Halkalõ were the most enduring and were individ-
ually exploited by the Roman, Byzantine and the Ottoman
city. The substantial redevelopment of these lines after the
15th century however, has meant that it is hard to unravel
the earlier material from the Ottoman (C Ë ecËen 1996b).
Details of the complete system become increasingly com-
plex as this survey progresses and it is not yet clear why it
was necessary to create such a double system in the central
part of the long-distance aqueduct. However, as awareness
of the system as a whole is expanded, more correlation can
be made between individually observed anomalies.
Geological and hydrogeological
context
The geological framework of Thrace, from the Bosphorus
in the east to the Ergene River Basin in the west, is char-
Research article
acterised by a Palaeozoic basement of granite igneous with
an extensive range of metamorphic rocks (gneiss, schist,
metaquarzite, marble) up to the highest elevation
(1,031 m) of the Istranja massif in the west (Fig. 2). It is in
this western zone in the vicinity of Vize where the best
spring sources can be found today as in antiquity, when
they were exploited extensively by the Byzantines to sup-
ply water to the city of Constantinople (Forchheimer and
Strzygowski 1893).
The Istranja mountain range borders the northern ¯ank of
the Ergene Plain, with a NW±SE basement axis that deep-
ens towards the Sea of Marmara (around BuÈyuÈk C Ë ekmece),
where igneous and metamorphic rocks are progressively
covered by the Palaeozoic sedimentary formations of
sandstone, silts and clays through which the Bosphorus is
cut. The Palaeozoic basement in Thrace is overlain by
Cenozoic sedimentary formations, whose sequence begins
with Palaeogene (sandstone, sandy-clay, limestone), that
overlie unconformably the metamorphic complex.
Eocene is widespread in Thrace and generally consists of
limestones extending along the southern slopes of the
Istranja massif and towards the south-east along the axis
¯exure of the metamorphic basement. South of the Istranja
mountains in the Ergene Plain, the Eocene limestone for-
Environmental Geology (2001) 40:1325±1333 1329
 Research article

mation is buried by Oligocene, Miocene and Pliocene (Aicher 1995; Bono and Boni 1996). However the Con-
terrigenous deposits of marine and continental facies. The stantinople system, at more than 250 km in length, with its
thickness of sediments in the Thrace basin reaches up to complex articulation of two principal levels of water
6,000 m (YuÈzer 1997), including Quaternary continental channels and with the convergence of several minor
deposits mostly seen as alluvial sand and lacustrine silts branches into the main line, is in many ways very different
and clay. and altogether more complex. General evidence from the
Because of erosion and tectonic displacements, the Eocene aqueducts of Rome shows the use of much simpler hy-
limestones are scattered along the southern slopes of the draulic schemes of construction. This was not a result of
Istranja metamorphic range forming discontinuous and any apparent technological advances achieved by the
limited outcrops at high elevations between 400±500 m. At Byzantines. It was mainly because of the greater avail-
lower altitudes (200±300 m) on the ¯anks of the Istranja ability of water resources within a shorter distance from
massif and towards the topographic depression of the Rome and also the greater hydraulic potentials and the
basement, the Eocene limestone formation is better pro- more stable discharge of these sources, tapped from either
tected from this erosion and forms outcrops with larger karstic or volcanic hydrostructures of regional extension.
extension and thickness. The Thracian water scheme was instead conceived to
Concerning the groundwater resources of Thrace, the most overcome problems related to a general shortage of
extensive and abundant aquifers are in the alluvium de- ground-water resources, particularly during droughts.
posits of the Ergene basin and are represented by uncon- On the basis of the geological and hydrogeological setting
solidated formations of Quaternary±Pliocene age. Most of of Thrace, it is possible to hypothesise that the scarcity of
the Ergene tributaries are perennial and their discharge springs with signi®cant discharge during the dry season
during the dry season is supplied either by peripheral could explain the necessity to extend the catchment to the
karstic springs from the Istranja limestone aquifers or by perennial sources of higher discharge in the distant area of
linear springs located at lower levels in the plain, where Vize. For the ®rst 100 km of the supply line, from the
there is ground-water seepage to the river bed. principal sources in the region of Vize (east) to modern
Although much of the ground-water resource is related to BinkilicË (west), the main water channel is mostly supplied
the unconsolidated aquifers of the Ergene basin, the major by karstic springs related to aquifers located on the
spring discharge (associated with highest hydraulic po- southern slopes of the Istranja massif and belonging to the
tential of the region) derives from the Eocene karst res- Ergene River basin. This was also the case for the spring
ervoirs. These are located mostly along the southern ¯ank sources close to the Anastasian Wall such as PapucË and
of the Istranja massif bordering the Ergene Plain, where Pinarca (Fig. 3). The implication of this feature for the
the metamorphic basement outcrops. Similar conditions, water supply is that there is a very signi®cant difference in
at lower altitudes, can be observed towards the south-east the discharge of the springs between the winter and
in the Palaeozoic basement in the C Ë atalca area betweensummer months, with a particularly rapid response fol-
Terkos Lake (Black Sea) and BuÈyuÈk C Ë ekmece (Sea of lowing heavy rain.
Marmara). The igneous±metamorphic rocks of the Palae- Further to the north of Vize, where the underlying geology
ozoic basement generally have a low permeability and, as a is metamorphic in character it is estimated that the dis-
result, make a limited contribution to the potential of thecharge would be less than those from the karst springs. As
ground-water resources of the region. a measure of the extent of the demand for water it is ev-
In the basement rocks, the marble sequence displays a ident that even the sources from these metamorphic
different hydrology. This formation is located in the upperaquifers of scarce potentiality (but with constant dis-
part of the metamorphic series and outcrops over a large charge), were branched to the main supply lines. In ad-
area of the Istranja massif at 600±800 m altitude. At pre- dition, several tributary channels were provided to divert
sent, the extent and thickness of these carbonate rocks is the base ¯ow of perennial streams crossing the meta-
overstated in the existing survey of this region. For much morphic basement. A number of these sources can be
of the metamorphic sequence, surface run-off is the associated with supplementary channels to both the high-
level and low-level systems located east and west of the
principal process. It is signi®cant that these resources have
recently been used to provide a major new water supply Anastasian Wall, at KursËulugerme and BuÈyuÈkgerme (see
for Istanbul and the cities of western Thrace. This scheme above). Considering the morphology of the valleys and the
low permeability of the basement rocks, the use of dams to
is based on a series of small reservoirs feeding into a large
isolate water reservoirs even of limited capacity cannot be
pipeline ring on the north side and parallel to the Istranja
ridge (YuÈzer 1997). excluded, although structural evidence has not yet been
found within the investigated area.
Maintaining a stable and consistent water supply to the
city would clearly have been a priority for much of the
Hydrogeology and water supply Byzantine period while the city enjoyed relative prosperity.
It can be suggested that the maintenance and protection of
In terms of scale and complexity, the closest parallel for the aqueduct system was managed through a complex
the water supply of Constantinople is the system that program of control and defence, for which the Anastasian
evolved in Rome between the 1st and the 3rd centuries A.D. Wall might have played an important role. The role of the

1330 Environmental Geology (2001) 40:1325±1333


military presence on the Wall in the protection of the
water supply line remains a mute point, largely because of
the dif®culty of providing protection for the entire length
of the aqueduct system.
However, it is likely that survival strategies were in place to
cope with the cyclical drought periods that must have
occurred because of the natural decline of the source
discharge as a result of low precipitation in the summer
season. These ¯uctuations would have necessitated an
appropriate management strategy for the unhindered
continuation of normal daily life.
The dramatic seasonal variability of the Thracian supply
was noted by the historian Procopius in his account of the
construction of the Basilica cistern (Yerebatan Saray) in
6th-century Constantinople, for which ``the emperor Jus-
tinian made a suitable storage reservoir for the summer
season, to contain the water which had been wasted be-
cause of its very abundance during the other seasons''
(Buildings I, xi, 13±15). The construction of the Basilica
cistern is attributed by Malalas, another 6th century
chronicler, to Longinus, the Prefect of Constantinople
under Justinian also mentioned in the inscription found at
Elkafdere, who was presumably therefore involved in
restorations and improvements to the system both in the
Research article
city and the hinterland (Malalas, 482B). Although it may
be doubted that Justinian was the ®rst to recognise the
problem of seasonal variability, a consideration of these
hydrogeological issues may explain why Istanbul possesses
so many open and covered cisterns of Byzantine date. At
least 70 cisterns are known from archaeological and lit-
erary sources in the city, ranging from small private
establishments, to enormous open-air reservoirs (Fig. 7).
According to C Ë ecËen (1996a), the discharge of the Istranja
aqueduct varied between 0.3 and 1.0 m3 s±1 in accordance
with the Thracian springs. His results were based on hy-
draulic calculations considering a mean gradient of 0.6&
and a water channel section of 0.85 m (width) by 1.7 m (to
the top of the vaulted roof). The maximum discharge of
about 1 m3 s±1 refers to a high water level (Hw) of 1.2 m.
Although these ®gures are merely approximations, C Ë ecËen's
¯uctuation value estimates allow signi®cant comparisons
to be drawn with Rome. Sextus Julius Frontinus, the water
resource administrator (curator aquarum) during the
Fig. 7
Map of Constantinople and its immediate vicinity, showing the
principal line of the Byzantine water supply system and the known
cisterns and reservoirs within the city (after C
Ë ecËen 1996a, p. 40)
Environmental Geology (2001) 40:1325±1333 1331
 Research article
reigns of Nerva and Trajan (97±100 A.D.), records a mean
water supply of the aqueducts in operation during his
mandate, equivalent to about 12 m3 s±1. More than 8 m3 s±1
were supplied by karstic springs while the additional
discharge came from catchments in the volcanic aquifers
near the capital (Bono and Boni 1996). This is considerably
more than the maximum values of 1 m3 s±1 estimated for
the Istranja aqueduct.
Investigations of the springs that provided the majority of
Rome's water supply have deduced that the channel net-
work bene®ted from a substantial and steady discharge
(Bono and Percopo 1996). This conclusion comes from the
assessment of large water resources, most of them related
to a karstic hydrostructure of regional extension, with a
rock reservoir that is more than 3,000 m thick (Boni and
others 1986).
Rome's aqueduct discharge was, therefore, perhaps 10
times more proli®c than that of Constantinople during the
dry season (summer season±early autumn). Although at
different times both these systems were devised to supply
populations in excess of half a million. In the context of
such a quanti®cation, the distribution, quantity and scale of
the cisterns recorded in Istanbul perhaps ®nds explanation.
Three of the reservoirs (Aetius: 197,000 m3; Aspar:
220,000 m3; Mocius: 250,000 m3) were established in the
early 5th century on highpoints (50±60 m a.s.l.) in the ci-
ty's suburbs (Fig. 7). In part they seem to demonstrate the
greater concern for security witnessed by the near-con-
temporary Theodosian Walls because they are located in-
side the defensive circuit. Another reservoir however, the
Fildamõ or `Elephant's Stables' is known to the south-west
of the city, close to the assembly point of the Byzantine
campaign armies at the Hebdomon. Its role appears to have
been fundamentally to provide water for the substantial
palace that developed in the Hebdomon in late antiquity.
No aqueduct has been identi®ed leading to this extra-mural
reservoir although it seems more likely to have been sup-
plied by the nearby sources of Halkalõ than by the long-
distance line. In practice it remains unclear how any of the
reservoirs worked within the system as a whole. However,
the piezometric towers (aqua castellae/control and regu-
lation towers) evidenced in the Mocius and Fildamõ res-
ervoirs demonstrate their function not only as large
capacity storage tanks, but also as control points in the
distribution network of the city (Crow and others 2000).
It could be argued that the three intra-mural reservoirs
were associated with the construction of the 5th century
high-level channel found near the Anastasian Wall. This
substantial new channel (discussed above) would have
provided a greatly increased discharge to the city, perhaps
creating the opportunity to store water on such a large
scale. C
Ë ecËen (1996a) estimates that the total water storage
capacity of the city's cisterns and reservoirs was between
800,000±900,000 m3.
The solution to both the military problems and the hy-
drogeological constraints seems, therefore, to have lain in
the mass storage of water in and around the city itself,
primarily within the bounds of the city's massive defensive
circuit. Sustained use of such a system required a careful
1332 Environmental Geology (2001) 40:1325±1333
day-to-day balance to be maintained through hydraulic
control, between the acquisition and public distribution of
the water resource. The daylight activities of the urban
community most probably depleted the available water
considerably. At night when demand would be less, but
average supply through the aqueduct system the same, the
reservoirs would be progressively re®lled.
It might expected that during the dry season, the input of
water in the storage reservoirs was inadequate to balance
the daily consumption of the city. Therefore, the necessity
to store and save water as much as possible to overcome
the critical dry periods of the year or military crises would
represent a crucial strategy for the survival of the Empire's
capital.
Water storage structures on the scale of Justinian's Basilica
Cistern and the open-air reservoirs of Constantinople are
not found in Rome. In this paper the authors have argued
that explanation for this dichotomy can be found in the
substantial difference in the availability of water resources
between the two capitals as a consequence of the pecu-
liarities between the geological and hydrogeological
framework of Thrace and Latium. For the ®rst time, po-
tential links have been made between the aqueducts of the
hinterland and the cisterns within the city, and this ar-
chaeologically derived account has been situated within an
environmental context. It is clear, however, that the pic-
ture is much more complicated and that the city with its
baths, fountains and numerous underground cisterns, was
supplied from a variety of sources whose points of origin
are not yet exactly located.
Epilogue: medieval Constantinople
The historical sources record the continuing maintenance
of the water supply system until the early 7th century.
However in A.D. 626 the Avars besieged the city and the
Byzantine chronicler Theophanes reported that the Aq-
ueduct of Valens was severed. Restoration of the long-
distance system was not recorded before A.D. 767 in the
reign of Constantine V (Theophanes, Chron. pp. 439±440).
Some historians have interpreted this as an 150-year hiatus
as an indication of the end of the `classical water supply
system' with disastrous implications for the maintenance
of a large urban population (Mango 1995). Magdalino
(1996) takes a less pessimistic position and has recently
questioned this interpretation of the city's demographic
decline.
Neither account however considers the possibility that the
sources closer to the city were continuously (or even in-
creasingly) exploited. A recent study of early medieval
Rome (Coates-Stephens 1998) has demonstrated how the
imperial and papal authorities maintained at least part of
the old imperial Roman system into the 10th century. It
would seem correspondingly unlikely that the imperial
capital at Constantinople was not able to maintain at least
a part of the late antique system up until the 8th century
restoration reported by Theophanes. The clearest evidence

for distinct later construction at this time survives from
the aqueducts at Talas and Leylek Kale near to CiftlikkoÈy
where the rebuilding incorporated much of the earlier 4±
5th century works (Dirimtekin 1959, 1968; Fig. 3). Later
restoration to strengthen earlier masonry vaults using
brick arches can be seen at Ballõgerme and BuÈyuÈkgerme
(Crow and Ricci 1997).
Throughout the middle ages the water bridges and chan-
nels were disrupted by earthquakes and possibly also as the
result of military con¯ict and invasion. The long-distance
system is said to have been abandoned in the early 12th
century as a result of cumulative damage. Like the Anast-
asian Walls, the ®nal reference to the aqueducts was in a
nostalgic late 15th century list of the Wonders of Con-
stantinople. The Ottoman water supply, which formed an
essential element of the regeneration of the metropolis in
the 16th century, was based entirely on the closer sources
at Halkalõ and in the Forest of Belgrade (CË ecËen 1996b).
Acknowledgements The `Water Supply of Constantinople' is a
3-year Leverhulme Trust sponsored project directed by James
Crow and Paolo Bono. Richard Bayliss and Glyn Goodrick are the
project researchers (Newcastle University), with major academic
support also provided by Alessandra Ricci (Matera University)
and Carlo Percopo (La Sapienza, Rome). This new research forms
part of the Anastasian Wall Project, an ongoing programme of
®eldwork that focuses on the 6th century outer defences of
Constantinople. Principal sponsors are the British Institute of
Archaeology at Ankara, The Arts and Humanities Research
Board, the British Academy, the Society of Antiquaries and the
Seven Pillars of Wisdom Trust. Our continuing ®eldwork expe-
ditions rely on the generous support of the Turkish authorities
and, in particular, the Ministry of Culture.
References
Aicher PJ (1995) Guide to the aqueducts of ancient Rome. Bolc-
hazy-Carducci Publishers, Waucondo, Illinois
Bayliss R (1999) Anastasian wall project. http://muse-
ums.ncl.ac.uk/long_walls/index.html
Bayliss R, Crow J (2000) The forti®cations and water supply
systems of Constantinople. Antiquity 74:25±26
Boni C, Bono P, Capelli G (1986) Schema idrogeologic dell'Italia
Centrale. Mem Soc Geol Ital 35(2):991±1012
Bono P, Boni C (1996) Water supply of Rome in antiquity and
today. Environ Geol 27:126±134
Bono P, Percopo C (1996) Flow dynamics and erosion rate of a
representative karst basin (upper Aniene River, central Italy).
Environ Geol 27:210±218
Ë ecËen K (1996a) The longest Roman water supply line. TuÈrkiye
C sar, Roma
Sõnai Kalkõnma Bankasõ, Istanbul
C
Ë ecËen K (1996b) Sinan's water supply system in Istanbul. Istanbul
BuÈyuÈk Sehir Belediyesi ISKI, Istanbul Su ve Kanalizasyon
Idaresi, Istanbul
Research article
Coates-Stephens R (1998) The walls and aqueducts of Rome in the
early middle ages, A.D. 500±1000. J Roman Stud 88:166±178
Crow J (1995) The Long Walls of Thrace. In: Mango C, Dagron G,
Greatrex G (eds) Constantinople and its hinterland: papers from
the Twenty-seventh Spring Symposium of Byzantine Studies,
Oxford 1993. Variorum, Aldershot, pp 109±124
Crow J (1997) The Anastasian Wall. Anatolian Archaeol 3:15±16
Crow J (2000) The Anastasian Wall Project 1999. Bull Br Byzan-
tine Stud 26:33±34
Crow J, Ricci A (1995) Anastasian Wall Project. Anatolian
Archaeol 1:12±3
Crow J, Ricci A (1996a) Anastasian Wall. Anatolian Archaeol
2:16±17
Crow J, Ricci A (1996b) The Anastasian Wall Project. Bull Brit
Byzantine Stud 22:30±34
Crow J, Ricci A (1997) Investigating the hinterland of Constan-
tinople: interim report on the Anastasian Wall Project. J Roman
Archaeol 10:235±262
Crow J, Ricci A (1999a) The Anastasian Wall Project 1996±1997.
ArasËtõrma SonucËlarõ Toplantasõ 16:239±249
Crow J, Ricci A, Bayliss R, Bono P (1998) Anastasian Wall Project.
Anatolian Archaeol 4:19±20
Crow J, Ricci A, Bayliss R, Bono P (1999b) Anastasian Wall
Project 1999. Anatolian Archaeol 5:15
Crow J, Bayliss R, Bono P (2000) The Anastasian Wall and the
water supply of Constantinople survey 2000. Anatolian Archaeol
6:16±18
Dalman KO, Wittek P, Schede M (1933) Der Valens-AquaÈdukt in
Konstatinopel. J.M. Reindl, Bamberg
Dirimtekin F (1959) Adduction de l'eau aÁ Byzance dans la region
dite Bulgarie. Cahiers ArcheÂol 10:217±243
Dirimtekin F (1968) Leylekkale, un aqueduc byzantin aÁ Byzance
dans la reÂgion dite `Bulgarie'. Byzantinische Forschungen 3:117±
119
Forchheimer P, Strzygowski J (1893) Die Byzantinischen
WasserbehaÈlter von Konstantinopel, Wien
Hodge AT (1992) Roman aqueducts and water supply. Duck-
worth, London
Magdalino P (1996) Constantinople MeÂdieÂvale. De Boccard, Paris
Mango C (1990) Le deÂveloppement urbain de Constantinople :
IVe±VIIe sieÁcles. De Boccard, Paris
Mango C (1995) The water supply of Constantinople. In: Mango
C, Dagron G, Greatrex G (eds) Constantinople and its hinter-
land. Variorum, Aldershot, pp 9±18
Oreshkov P (1915) Vizantinski Starini okolo Tsarigrad. Spisanie
Bulg Akad Nauk 10:71±118
Ricci A (1999) Constantinople. In: Bowersock GW, Brown P,
Grabar O (eds) Late antiquity : a guide to the postclassical
world. Belknap Press of Harvard University Press, Cambridge,
pp 391±392
Schuhhardt C (1901) Die Anastasius-Mauer bei Constantinopel
und die Dobrudshca-Walle. Jahrb Dtsch Archaol Inst 16:107±
115
Steinby EM (1995) Lexicon Topographicum Urbis Romae. Qua-
YuÈzer E (1997) Hydrogeology of northwestern Turkey. In:
Schindler C, P®ster M (eds) Active tectonics of northwestern
Anatolia. VdF Hochschulverlag AG an der ETH ZuÈrich, pp 125±
137
Environmental Geology (2001) 40:1325±1333 1333