Introduction

1Declared a World Heritage Site by UNESCO in 1990, the ancient city of Delos is located on a small island (360 ha) in the centre of Cyclades (Fig. 1). Today it is one of the major archaeological sites and tourist centres in the Mediterranean basin. Between the beginning of the 4^th c. BC and the end of the 1^st c. BC, the city of Delos thrived, particularly because of its sanctuaries and commercial harbour. At its height, between the end of the 2^nd and the beginning of the 1^st c. BC, archaeologists believe that more than 10000 people lived in Delos (Bruneau and Ducat, 2005). During this period, the water supply of Delos was only ensured by resources available on the spot, while nowadays tanker ships are necessary to supply the inhabitants of Delos and the neighbouring island of Mykonos.

2Since 1873, the site has been excavated and studied by archaeologists from the French School of Athens (EFA), who exhumed most of the remains of this period of occupation. At the beginning of the 20^th c., this institution was aware of the need to study the characteristics of the physical environment of Delos. This awareness resulted in a publication by the geologist Cayeux (1911). A second volume was expected to explain the links between the physical settings and the archaeological data, but it was never published. For most of the 20^th c., scientists were influenced by these initial observations concerning the ancient environment of Delos, in particular Cayeux’s thoughts on the old harbour and the island’s principal river. However, these elements were challenged by scientific work undertaken by geographers at the end of the 20^th c. (Dalongeville, 1996; Dalongeville, 1997).

3With this context, specialists in the environmental sciences and researchers in archaeological sciences developed a multidisciplinary research programme between 2000 and 2003. The aim of the programme, ‘Water in Delos’, supervised by M. Brunet and mainly financed by the EFA, was to understand whether the ancient population took into account the physical settings for the growth of the city, the sanctuaries, and the rural territory.

4Owing to its environmental characteristics and the evolution of the area after the desertion of the ancient city, Delos constitutes a remarkable case study for understanding whether the water supply was ensured, by the use of adapted and complementary hydraulic devices, in a physically restrictive environmental context.

Geographical and historical setting

5The island is located in the driest area of Greece. Its Mediterranean semi-arid climate is characterised by an average annual precipitation reaching approximately 350 mm (the average of rainfall data recorded in Naxos, Mykonos and Delos from 1955 to 2003; Desruelles, 2004). Rainfall occurs almost entirely between October and April, whereas the months of July and August are almost completely dry. During the six hottest months of the year, almost all the rainwater evaporates. Strong winds and high temperatures support evapotranspiration. The length of this hydrological summer is longer during the dry years. Moreover, annual precipitations vary greatly from one year to the next: between 1955 and 1997, the annual rainfall varied from 139 mm (in 1989-1990) to 683 mm (in 1981-1982) in the town of Naxos, whose climate is nearly the same as in Delos.

6Delos’highest elevation is only 114 m, but the island’s landscape is hilly. The topography follows a north-south crest, flanked by short, steep east-west slopes (less than 1 km), whose integrity is disrupted by small plain areas. The ancient city developed primarily on the openness of this region, the ‘Plaine principale’ (Figs. 1, 2, and 4), where the topography is more favourable. The landforms in Delos mainly result from Plio-Quaternary weathering and the erosion of fractured crystalline rocks (gneiss and mostly granite; Lucas, 1999; Fig. 1). The island consists of differential erosional landforms characteristic of this kind of rock, including granite domes, tors and pseudo-tors, rounded blocks, and alveoli (semi-circular basins of a few hundred square metres). Soils and superficial deposits (grus, mainly), which cover the bedrock, are generally thin (average thickness of 1 m) and discontinuous. At the top of the slope, they are located between numerous rock fragments. At the bottom of slope, they are thicker and consist of colluvium, even porous sandstone composed predominantly of algal grains and skeletal debris (calcarenite; Desruelles, 2004). The superficial deposits are thicker (1.5 m on average) in the alveoli and main valleys, which are located at various heights. In the island’s northwest region, the topography of the ‘Plaine principale’ was modified by the large volume of rubble resulting from the excavations and 60000 m³ was deposited in the principal bay of Delos, where the ‘Sacred Harbour’ had been developed during Classical Antiquity. The low and discontinuous plant cover is dominated by phrygana, a formation of the thermo-Mediterranean level composed of bushes that are often thorny and deciduous during the summer. It seems to be a scrubland facies, growing on very thin soil cover. These physical characteristics of the island would seem to not favour freshwater resources. Moreover, taking into account Delos’small area, the strong winds and the marine environment, the sprays produce very heavy saline deposits. In addition, saltwater intrusions occur in the coastal aquifers.

7Most ancient ruins currently visible in Delos were unearthed by the EFA archaeologists, starting in 1873 and in particular between 1904 and 1914 (‘grandes fouilles’). The current visible ruins offer a ‘fossil landscape’ of the city of Delos in the 1^st c. BC, resulting from several successive phases of occupation. The first traces of buildings date from the second half of the 3^rd millennium BC (Poupet, 2000), even if the human occupation of the island remained intermittent until the 7^th c. BC. The island became a sacred territory because of a poetic narrative: according to the Homeric anthem devoted to Apollo, Delos was the birthplace of Apollo, son of Zeus and Leto (Bruneau and Ducat, 2005). Thanks first to the sanctuaries and then because of the presence of a powerful economic centre, Delos acquired an important reputation. In the 3^rd c. BC, when the commercial rise of Delos began, only a few thousand people lived on the island (Vial, 1984). The decision, taken by the Roman Senate in 167 BC, to attach Delos to Athens and to exempt the harbour from taxes spurred economic progress and a sudden inflow of population. The population subsequently tripled and the city quickly developed to over one hundred hectares. Starting from the end of the 1^st c. BC, however, the city was gradually abandoned. The activity around the sanctuary of Apollo continued at least until the 2^nd and 3^rd c. AD, but the inhabited area contracted to the area around the sanctuary of Apollo. Delos was deserted, or only temporarily occupied by pirates, from the 6^th c. AD onward. In fact, it was used as pastureland until AD 1873. Currently, there are only a few inhabitants in Delos, whose ancient landscapes primarily seem to be modified by the excavations. The comparison between these ecological constraints and the importance and density of the city of Delos, even if only for a short period (from the 4^th to the 1^st c. BC), is the focus of the research completed in the research programme, ‘Water in Delos’.

A geoarchaeological multidisciplinary study

8This research was carried out by developing a geoarchaeological approach, combining geographical, historical, archaeological, and palaeo-environmental data. The study integrated information from maps and aerial photographs, as well as data obtained by field observations and laboratory analyses. This study is structured around three main axes or foci of investigation.

9The first axis aims to understand the distribution of contemporary water reserves and to estimate their temporal variability, in relation to climatic fluctuations. The replenishment of aquifers and the volume of the water reserves were estimated for the groundwater catchment in the area occupied by the ancient city of Delos. The climatic data, topographical characteristics, and the properties of the superficial deposits were taken fully into account

• In order to study the effects of the annual variations of rainfall on the water resources, we have chosen to focus on three hydrological years that represent three different types of conditions: one ‘humid’ year (1981-1982, 751 mm), one ‘dry’ year (1989-1990, 153 mm) and one ‘average’ year (1990-1991, 391 mm). Potential evaporation was calculated using the Folton and Lavabre formula (Desruelles, 2004). The replenishment of aquifers was estimated using a daily time-step.
• Geomorphological mapping and studies were undertaken in Delos and on the neighbouring islands of Mykonos and Rhenia, which present the same physical settings. Landforms were classified according to the topographical characteristics and the properties of the supericial deposits. Boreholes were drilled in order to better understand the terrestrial phase of the hydrological cycle. These results were taken into account in establishing the water balance (Cosandey and Robinson, 2000) of the selected catchments. The estimates of aquifer replenishment were crosscompared with piezometric levels and salinity measurements carried out in spring and autumn during the period 1998-2003.

10The second axis seeks to understand the historical evolution of the city’s physical settings and its efects on water reserves. The climatic, geological, and topographic settings were studied starting with bibliographical data and field observations. Sea-level changes in Mykonos, Delos, and Rhenia were reconstructed, mainly using radiocarbon dating of beach-rocks (fossil beaches, which indicate past relative sea levels; Desruelles et al., 2004; Fouache et al., 2005).

11The third axis links the hydraulic installations of the ancient city of Delos, the water resources, and the entire physical setting. It aims to understand the water management allowed by these devices. This approach was facilitated by the interdisciplinary collaboration within the framework of the ‘Water in Delos’ programme. The research is based on archaeological studies and surveys, which were compared with data collected on the past and present physical settings.

12The main limitation of this study concern the reconstruction of the climate and plant cover during antiquity. Indeed, in Delos there are no sedimentary archives allowing specific environmental analyses (palynology, in particular), which would normally be used to provide precise elements of landscape reconstruction, and no similar study has been undertaken on a neighbouring islands.

An important spatiotemporal variability of water resources

Rainfall irregularity exacerbated by geomorphological setting

13In Delos, available water resources are mainly constituted by underground water (groundwater). Concentrated flow and surface runoffare rare, even in the valley of the island’s main river (Inopos). Indeed, the differential erosion of landforms on Delos favour the infiltration of rainfalls:

• Rock surfaces (such as tors and pseudo-tors) favour the infiltration of water because of the dense fracture network and these surfaces cover 28% of the island and often correspond to areas located at the apexes of the relief.
• Hillslopes and depressions located below the apexes are characterised by the existence of superficial deposits of varying thickness (1.5 m on average in the alveoli), made up of grus (sand or silty sand produced by granitic bedrock weathering) whose mean porosity was estimated at 46.5% (Desruelles, 2004).

14The erosional processes and the surface runoff, only produced when abundant winter rainfalls saturate the ground, are thus limited. On the other hand, the water transfer within soil and superficial deposits causes evapotranspiration, even if the uptake by plants is limited. Amongst the three types of relief defined according to the geomorphic analysis, rock surfaces are more ‘effective’ for the replenishment of the aquifers. Groundwater infiltration is fast, almost without evapotranspiration, thanks to the dense fracture network. By contrast, inside the depressions filled by a thicker mantle of superficial deposits (whose ‘field capacity’ was estimated at approximately 50 mm; Desruelles, 2004), rainfall is partially or totally lost due to evapotranspiration: in the dry period and at the beginning of the hydrological winter when the water retention in the soil stops the replenishment of the aquifers.

15Rainwater replenishes fissured aquifers and superficial deposits. Freshwater streams down hillsides towards the sea and, more often, refill the alveoli aquifers. The geometry of crystalline landforms is favourable to the concentration of fresh water in these depressions: the alveoli, associated with their drainage basins, constitute hydrological units located at various heights. Due to the low permeability of the superficial deposits in contact with the crystalline bedrock, and the presence of less permeable outcrops infilling them, water is partially retained in these alveoli. Consequently, even if recharge of the aquifer is reduced by the water retention in the soil and superficial deposits, the alveoli are the landforms with the best hydrological potential in Delos.

16The water, which crosses the impermeable outcrops that fill the alveolus located, feeds flows, generally sub-surface flows, into the valleys located farther down the hillside. In some of them, temporary flows are created after abundant and intense winter precipitations. The low discharges of the topographically highest stream flows are illustrated by the case of Inopos, the island’s longest river (Figs. 1 and 4). The simulations produced an estimate that the river, whose catchment area is 0.21 km², had a peak stream flow of only about 0.15 m³/s at its mouth, before its management in the Hellenistic period (Desruelles, 2004).

17The morphology of Delos favours the formation of relatively high water reserves during humid periods. This feature can even cause temporary flooding of the alveoli. Part of this water then flows down the valley, or sometimes out towards the sea. On the other hand, the diversion of underground water into small and loosely connected water tables causes springs to dry up during dry periods. During the dry year, the three alveoli that supply Delos’s water were almost entirely dry and the volume of water stored during the hydrological winter was not sufficient.

18Measurements and modelling produced the conclusion that the rhythm of groundwater exhaustion and recharge is quasi-annual. The aquifers do not allow storage, one year to the other, of the water produced in humid years. The effects of a dry year on the water-table level are thus not reduced by underground water reserves. Piezometric level measurements in the ancient wells of Delos have confirmed these variations of underground water volume and the influence of the rainfall variations (Desruelles, 2004). However, during an average year most of the aquifers are not totally dry and the salinity is quite low (around 3 ‰ to 4 ‰ in the mean). In the low coastal areas, the upstream flow reduces the depletion of the water table due to evapotranspiration, causes flow towards the sea, and is available for human use (in the North-West of the ‘Plaine principale’).

A weak evolution of the physical settings since antiquity

19Even if no data directly related to the ancient climate of Cyclades is known to the authors, according to scientific studies of Holocene palaeoclimate in the southern Balkans (Denèfle et al., 2000; Fouache, 2001; Fouache et al., 2001) and Crete (Grove, 2001), climate variations associated with rainfall variations took place over centennial or decennial-scales. According to archaeological and archaeobotanical studies (e.g., Sampson, 2002; Megaloudi et al., 2003), leguminous plants and cereals, similar to the present cultures, were cultivated in the 5^th millennium BC. During the 6^th c. BC and the beginning of the 5^th c. BC, terraces were built on the major part of the island’s hillsides (Brunet, 1999; Poupet, 2000) in order to develop a traditional Mediterranean mixed-farming pattern, with barley, vines and fruit trees (Brunet, 1999; Charre and Couilloud-Le Dinahet, 1999). According to this geomorphic study, Delos did not experience a climatic phase more humid than the present one during the last millennia. The morphology of the calcarenite outcrops indicates little evolution of landforms since the end of the Pleistocene (Desruelles, 2004). Moreover, the valleys were shaped by ephemeral flows similar to the present ones. The channels cut down through unstratified superficial deposits. At the mouth of the river, the mantle of superficial deposits is very thin and primarily made up of colluvium and sometimes constituted by very fine argillaceous layers caused by water stagnation behind a beach ridge.

20The modifications of the island’s landscape are mainly a result of human action. The topography of Delos has changed due to ancient developments (the city in the North-West and cultivation terraces on the other part of the island) and then again during the archaeological surveys (Bruneau et al., 1996). However, the coastline has moved due to of the post-glacial sea-level rise, accentuated by a subsidence of the Cyclades Plateau. This relative sea-level rise was estimated at 2.5 m since the beginning of the Hellenistic period (Desruelles et al., 2004; Fouache et al., 2005). The Western shoreline of the ‘Plaine principale’ moved back from 5 m to 10 m in cliffed areas and several tens of metres in bays. However, the influence of this relative sea-level rise on the level and salinity of the groundwater was probably weak: in Delos, the water tables are small, often located at height and protected from saltwater intrusions by impermeable granitic outcrops (Desruelles, 2004).

Complementary and adapted hydraulic devices

21At the beginning of the 1^st c. BC, human water demand was greater than at present and the water reserves were probably as limited as today. Thus, during average and dry years, the risks of aquifers drying out would be strong at the end of spring. The lack of supply was partly, or completely, reduced by a water supply system connecting individual and collective hydraulic installations, which can be classified into three main categories: they are wells, cisterns and ‘reservoirs’. These devices make it possible to collect, store and use optimally almost all the available water resource (including rainfall and groundwater) in order to supply the city of Delos.

22Underground water was collected by wells, whose architecture and depth were adapted to the characteristics of aquifers (Desruelles et al., 2003). In the city, as in the sanctuary, wells are about the oldest and the most common means of access to water in Delos (Brunet, 2008). Wells were the last method of collecting the groundwater before its flow towards the sea and water from wells was probably particularly important during the winter. It would prevent the losses by underground flows to the sea and ensure the supply to the city. Due to the absence of recharge from springs, water-table depletion was to be too great in summer to provide the volume of water necessary to supply the residential districts. Water supply by wells was very random at the end of the summer, because of underground water pollution (by salt and sewage) and the fall in water-table levels.

23The ‘reservoirs’ received a mixed water supply from underground and superficial waters. In Delos, they can be considered ‘infiltration wells’ allowing the infiltration of rainwater from impermeable surfaces (roofs and terraces of houses as well as public building surfaces like the theatre). This artificial recharge was generally more ‘effective’ than natural aquifer replenishment, reduced by the retention of water in the soil. These installations created points of infiltration in the urbanised areas. Large collective reservoirs were built between the end of the 5^th and the first half of the 3^rd c. BC, within the framework of a policy of major public works (Brunet, 2008):

• The ‘Theatre Cistern’ (Fig. 3), located west of the theatre, allowed the infiltration of rainwater flowing from the theatre into the water table. This 1150 m³ reservoir (Bruneau and Ducat, 2005) protected the ‘Theatre Quarter’, located below, from runoff. It also constituted an ‘infiltration well’ for the water table located below, which supplied the numerous wells of the houses in the district (Desruelles et al., 2003). The reservoir allowed a direct recharge of aquifers at the beginning of hydrological winter, when rainfall does not reach the water table due to retention in the soil.
• The ‘Reservoir of Inopos’ (Figs. 1 and 4) is the other large reservoir on the island. Starting from the end of the 5^th c. BC (Fincker and Moretti, 2007), this reservoir collected almost all of the flow from the river Inopos, upstream of the ‘Plaine principale’. Watercourses connected to this reservoir supplied a sanctuary of Egyptian Gods with the water of this river considered as a diversion of the Nile (Bruneau, 1990). The water of the reservoir was also used to supply the urbanised area located in the small valley upstream.

24Cisterns were built in the majority of houses, primarily during the urban growth in the 2^nd c. BC. Their volume varied from 100 m³ to 600 m³ and they collected almost all rainwater falling on the roofs and terraces (Desruelles, 2004). The losses by evapotranspiration were low because water was quickly guided by the gutters towards the cistern, which was covered and waterproof. Thus, they could partially refill the devices in a dry year. According to the daily recharge simulations, the capacity of certain cisterns was higher than the volume of water collected during an average year (Desruelles et al., 2003). This oversize suggests that groundwater could be taken (using the wells) to refill the cisterns when a portion of the groundwater, by reaching the sea, could not supply the city or, while flowing to the low coastal areas, could flood them (as was observed during the winter of 2002-2003). In contrast, other cisterns were too small to store all of the rainwater contributions and they often had a draining device that siphoned offthe excess water towards the water table or into a nearby infiltration well.

25Any optimal management of the cistern water needs to be planned over several years: multiannual cycles of drought, similar to those, which can currently occur, were probably experienced during antiquity as well. The cisterns were intended to be full at the beginning of the drought to allow the continuity of the city’s water supply between the beginning of groundwater depletion and the first autumnal rains. This fact implies an initial use of the ephemeral resource of well water and reservoirs during humid periods to supply the city. In anticipation of the dry periods, the population probably used groundwater in wet periods for filling the cisterns ‘over-engineered’ when compared with the volume of water collected during average years.

26According to the daily simulations, if such management was carried out, the inhabitants of each house had a minimal ration of 10 litres per person per day (LPD) during one year of drought (Desruelles, 2004). For each person, 26 m² of roof used to collect the winter rains are needed for this supply even in very dry conditions. By comparison, average water consumption in the ancient cities is estimated to have been from 20 to 50 LPD, but it could have been reduced to 5 LPD if necessary (Bonnin, 1984). De Planhol and Rognon (1970) indicate that in the Mediterranean or Middle East, in traditional urban environments, 7 LPD was enough. Indeed, during the period of water shortage, it can be supposed that water was used for several purposes. For example, the same water could be used to cook, clean the house (the latrines, in particular) and then for draining offthe waste, which accumulated in the sewers during the dry season. In addition, part of the domestic water, which was used for the latrines in particular, was drained to the sea.

27However, such a scenario of multi-monthly, even multi-annual management of water is only possible based on the operating mode of hydraulic devices and on the optimal use allowed by their association in the residential districts in the 1^st c. BC. On the one hand, it does not take into account the relative contributions of rainwater and groundwater, depending on their usage. This scenario implies a preferential use of groundwater during most of the year, even a combination of this water and cistern water. The ancient texts, in particular those written by physicians, distinguished between groundwater (coming from spring or wells) and rainwater (coming from cisterns). Thus, the techniques of water supply developed in Delos seem, in this respect, very unusual (Brunet, 2008). On the other hand, there perhaps existed in Delos a regulation that specified the statute of the hydraulic devices. The cisterns, which collected water for only one house, probably constituted private water reserves and were only used by the people of this house. On the contrary, the groundwater was certainly considered a public wealth (Chevalier, 2001). Was groundwater exploitation regulated by altered rules governing water abstraction, especially in periods of drought? Such a regulation would have been supported, by an understanding of the risks of well water depletion, and the value of the construction of cisterns in each house. Lastly, devices were to be temporarily empty in order to be cleaned. The devices that allowed the cleaning of the cisterns seem rather widespread in the ‘Theatre Quarter’ (Chamonard, 1922-1924). The cleaning of the cisterns was probably carried out when they were empty, and also possibly when the wells and the reservoirs were well supplied. The population’s continued supply was then possible without the water roof of the cisterns, which could be refilled with groundwater after the cleaning. Moreover, the cisterns were not probably cleaned at the same time.

28The waste water and the water that were not collected for the supply of the cisterns and the reservoirs was evacuated by a dense network of drains, whose layout corresponded to that of the streets (as in the ‘Theatre Quarter’, for example; Chamonard, 1922-1924). The water drainage was necessary in the ‘Plaine principale’. In this low coastal area, where the sanctuary of Apollo is located, the waters from the hillslopes concentrated, but which were not collected to supply the residential districts. During the Hellenistic period, a network of drains took excess water to the sea (Desruelles, 2004). North of the ‘Plaine principale’, the groundwater of an alveolus supplied a temporary pool: the ‘Sacred Lake’ (Figs. 1 and 4). Until the beginning of the 1^st c. BC, a draining device took excess water from this temporary pool, especially when it was used as a fish basin, to the sea. Probably due to the fall in the water supply related to the development of the surrounding city, and the outfall of sewage, this basin was gradually abandoned during the 1^st c. BC.

Conclusion

29Due to its environmental characteristics and the remains of its hydraulic devices, Delos is a particularly interesting site for the study water resources and their management during the antiquity. Compared to many ancient cities of the Mediterranean basin (Bonnin, 1984), the hydraulic system built in Delos, composed of wells, cisterns, and a reservoir, is unusual. In most other cities, the water supply was often provided either only by wells (as in Thassos in Greece) or by cisterns (as in Alexandria in Egypt). In Delos, groundwater abstraction using wells allowed the population’s water supply during part of the year, primarily the hydrological winter. This groundwater, was not located at great depths and so was easily accessible. Furthermore, due to the high rainfall variability and negligible volumes at least during the summer, cisterns were necessary for the collection and storage of rainfall sfalling onto the buildings. These cisterns, whose size was proportional to that of the roof and thus to the number of people living in the house, produced a sustainable water supply when the wells dried up. This implies that the number of people in Delos depended partly on the size of built surfaces, where the runoffwas collected in the cisterns. Each district seems to have been independently accountable for its own water supply. This may explain why the excess water from the ‘Plaine principale’ was taken to the sea, instead of being used to supply the residential districts whose natural water supply was not as plentiful.

30The increase in roof surfaces and the building of cisterns between the 4^th and the 1^st c. BC caused an increase in the volume of the ‘artificial’ reserves, used to complement the small aquifers. The complementarity and the adaptation of hydraulic devices thus provided a sufficient water supply for several thousand people on a restrictive insular environment, characterised by strong rainfall irregularity and small and poorly interconnected aquifers, which contained poor quality water during periods of drought. The aridity of the present landscapes of Delos probably differs from the Hellenistic ‘wellkept’ landscape in spite of mostly similar climatic constraints. The city of Delos and the neighbouring rural territory of the island benefited from an optimal management of natural water resources and supply provided by the water stored in the cisterns. This study highlights the determination of the people of the island of Delos to ensure a sustainable water supply during the period of urban growth between the 4^th and the 1^st c. BC.