The paper describes the development of the Urban Water Supply System (UWSS) for Split (Croatia) in the second half of the 19th century. The selected concept of the water supply system was entirely based on the system of Emperor Diocletian's Palace from 305 AD, which has not been in operation since the 7th century. The research is based on the analysis of historical data and the actual state of the water supply system of Split. The study provides a clearer insight into the process of choosing the optimum concept of water supply system, the operational characteristics of Diocletian's system and the restored UWSS. The sustainability of both the Roman UWSS and the 1880 system, which has a significant part of the aqueduct still in use today, have been confirmed. It is an example of a zero-carbon water supply system running entirely on renewable energy: gravity.
The Roman Emperor Diocletian built his palace in 305 AD on the eastern coast of the Adriatic in the vicinity of Salona, the capital of the Roman province of Dalmatia and the largest Roman city on the eastern Adriatic coast (ca. 60,000 inhabitants) (Suić 1976; Cambi 1991). According to the usual Roman practice, both Salona and the palace had well-designed urban water systems. The nearby Jadro River karst spring ensured water supply for both systems. The barbarian invasion in the 7th century destroyed Salona as well as its Urban Water Supply System (UWSS). Part of its population settled within the walls of Diocletian's Palace, which became the core of Split with a current population of about 180,000. After both aqueducts ceased to function, the wells, cisterns, local water springs, and the Jadro River were used for water supply (Marasović & Margeta 2017).
During the summer, most of the local wells and springs would dry up, and therefore, water was transported from the Jadro River and handed over in exchange for money. In the middle of the 19th century, Dalmatia was under Austrian domination. This period was characterised by a rapid growth of the population and industrial development, and it became necessary to build a modern water supply system in the Split.
The UWSS planning represents a complex and responsible task (Angelakis & Zheng 2015). The aim is to achieve a sustainable UWSS that will permanently provide sufficient quantities of potable water to the community. To solve the problem, all four domains of sustainability must be taken into account. With the Multicriteria Decision Analysis, the variables could have a different weighting; in this case depending on the sustainability purposes and approach. The selection of sustainable options includes the use of multi-objective optimisation. This methodology provides a decision support framework that identifies which alternatives of the UWSS are optimal. Today, the planning of the UWSS should be addressed by applying a life cycle assessment (LCA) (ISO 2006).
The problem of the water supply to Split in the 19th century was solved systematically, i.e. several different solutions were taken into account in accordance with good engineering practice. The solution focuses primarily on technical, economic and social sustainability with less emphasis on environmental. The system approach was applied in selecting the optimal concept. Natural, technical and social constraints of the area were taken into consideration. The most acceptable solution was to restore the Diocletian's Aqueduct, which had not been in operation for 12 centuries. It may be concluded that the usual planning approach was applied in solving the problem of water supply. The application of a historical–cultural criterion was crucial for the selection of a solution.
The aim of the paper is to present the whole process of selecting and building the water supply system of Split from 1880, which was based on a Roman waterworks built in 305 AD. Based on original historic sources, this paper presents the planning framework and procedure for the selection of the UWSS to 19th century Split. It also presents the procedure and method of implementation of the selected solution, including integration elements with the remains of the ancient water supply system. Finally, the basic elements of the system and technological development over time are presented.
PLANNING THE UWSS
Planning the UWSS requires a detailed analysis of water demand and the data on available water resources (volume of available water, purity of raw water, permanency of source and elevation of water level with respect to the area to be supplied). The financial resources have always been an important factor in decision-making. Today, environmental criteria are equally important.
A present-day analysis of the UWSS demand includes estimates of household consumption, the amount of water required for public use and for industrial use. Yearly, monthly, weekly, daily and hourly variations in water consumption also have to be analysed. Planners of the late 19th century assessed the problem in the same manner; however, they also took into account a different way of water abstraction from the water supply network.
Today, users connected to the distribution system decide how much water they will consume and when, causing the water flow to vary significantly in the pipe system. At the end of the 19th century, only a small number of house connections were present in the water distribution system of Split. The majority of the population obtained water directly from public fountains. This meant that the water flow in the pipe network was not affected by water consumption, and hence, the water flow remained almost constant. This is a significantly different to today's practice. The flow regime in the water distribution system changed gradually with the introduction of numerous house connections.
Estimating water demand in the 19th century was different than it is today. A document from 1860 (related to the planning of the UWSS of Split) gives an insight into the estimate of water consumption, i.e. litres per capita per day (Radman 1860). A quantity of 20 L per inhabitant per day had to be ensured to satisfy domestic demand. The consumption increased to 50 L per inhabitant per day when the consumption of crafts and gardening were added. In other words, 550,000 L per day had to be available for Split to supply its 11,000 inhabitants. However, industry, which was then developing, was not included in the estimation and the amount of water required for industrial use was calculated separately, the same way as it is done today.
All available water sources (springs, surface and groundwater) were considered in the analysis of possible water abstraction. The main source of water was the Jadro River, as in Roman times (Figure 1). The Jadro River spring is situated about 7 km from Split at an elevation of 33 m above sea level (asl). The river, which is only 4.5 km long, has a relatively small river basin with no significant influence on the river flow or capacity. However, the Jadro River spring and its catchment area of about 450 km2 have a significant influence on the water source (Kapelj et al. 2002).
The discharge of the spring varies from a minimum summer flow of 3.72 m3/s to a maximum winter flow of 85 m3/s (Bonacci 2015). The specific hydrogeological settings of karst areas are responsible for a very dynamic recharge system manifested in the spring, which is highly responsive to precipitation. The spring is characterised by very sudden and substantial changes in water discharge as well as turbidity. The data for the period 1995–2001 indicate that the mean annual precipitation in the watershed region is 1,149 mm with a minimum of 904 mm and a maximum of 1,511 mm. Precipitation is, on average, lowest in July and highest in November. In the dry season, the saturated aquifer zone is drained by the karst conduit network and the water table is thus lowered. This process constitutes the slow flow or base flow, which is important for water supply because it defines the possible capacity of the water supply system during summer, i.e. the critical demand period. The spring and the Jadro River have sufficient capacity to provide water supply during the dry season. Planners of the UWSS of Split knew about the variability of water discharge at the spring and the occurrence of minimum flows, although they did not have access to the discharge data. However, they knew that the spring had been used for centuries as the water supply to the Roman city of Salona and the Diocletian's Palace. It was therefore considered extremely reliable.
Groundwater and small springs are also present in the Split area (Split peninsula). The groundwater is contained in a typical aquifer situated in the coastal region near the sea and is characterised by freshwater that floats over saltwater and by the low available volume. In the past, this water has been used as the water supply to Split (prior to the construction of Diocletian's Palace and in the Middle ages). Groundwater was tapped by shallow wells and water was also collected from several small springs. During summer, the wells and springs would dry up; hence, the Jadro River represented the only permanent source of water for Split.
In 1845 an analysis was made regarding the construction of the water supply system of Split in which four different UWSS configurations were presented (Figure 2).
The proposed variants were as follows:
Reconstruction of the Diocletian's Palace water supply system, from the Jadro River spring to the city, using a gravity-fed water supply system, a grade aqueduct and a pressure distribution system. The estimated cost of construction was 169,260 Austrian florins.
Use of the existing wells and excavation of the new ones in the centre of the Split peninsula, which would be connected to a single and unique water supply system. Water from the well would be pumped into a reservoir from which it would flow towards the city. The estimated cost of construction was 15,000 Austrian florins.
Surface water abstraction on the right bank of the Jadro River, approximately 3 km downstream from the spring at an elevation of 7 m asl. A mill and a small dam were built at the same site in 1711 providing a water supply via gravity-flow pressure pipes that were laid out in the terrain at the mill, at two small bays on the embankments built over the sea, and in fields and lowlands to the west edge of the city. If necessary, at the beginning of the aqueduct, water could be raised to a higher level for 12 feet (3.8 m) using the river flow to run the pump. The estimated cost of construction was 40,000 Austrian florins.
Surface water abstraction on the left bank of the Jadro River, approximately 1 km upstream from the river mouth at an elevation of 3 m asl. It included water transport to a higher level for 18 feet (5.6 m) for the gravity-flow transport towards Split using the river flow to run the pump. A tunnel approximately 200 fathoms long (380 m) would be made below the high terrain on the left bank towards Split, allowing supply by a gravity-fed water pressure system following low terrain levels all the way to the west part of the city. The estimated cost of construction was 45,000 Austrian florins.
The planning approach was comprehensive, reasonable and technically of good quality. The period of construction, costs, impact on the local economy and the wider impact on society were analysed for each of the variants. An extensive correspondence on planning and implementing the UWSS has been preserved. A contentious discussion between the investor, the proponents, the competent institutions and the academic community was held, and a final decision was made.
Variant (2) was discarded due to the insufficient long-term volume of available water, the permanency of a source, the purity of raw water and the elevation of the water level in relation to surface elevations of the area served. Variant (3) with the water abstraction of the Jadro River on the site of the existing dam was affordable. However, the risk involved the construction of a pipeline made of cast iron on the embankment over the sea where the pipes were in danger of corrosion. The problem was the closure of Jadro estuary and one smaller bay by building an embankment and bridges for the pipe installation. Variant (4) had water abstraction on the left bank of the Jadro River, which was quite low (approximately 3 m asl) and the water had to be raised by pumping. In addition, a 380-m long tunnel was planned to reduce the water level rise on the left bank.
Shortcomings of variants (3) and (4) included the purity of surface water. Water quality downstream from the spring is worse than at the spring due to biological activities in the water and the influence of surface waters from the topographic catchment area. Surface water contains suspended and dissolved impurities and is generally contaminated. It must therefore be properly treated before being supplied to a community. It represented a technical problem and presented an additional cost. In addition, the water supply pipeline ended in the city on low terrain levels; therefore, a pumping station would have to be built for the water supply to most of the city. The construction of technically very demanding pumping stations, especially the main pumping station and a tunnel (variant 4), and the sea crossing (variant 3), as well as the problematic surface water quality, are the reasons why these variants were discarded.
In the end, variant (1) based on the renewal of the Diocletian's Aqueduct was adopted (Figure 3). This option was strongly advocated by architect Vicko Andrić, the first conservator of Split, as well as other prominent experts. They emphasised the sustainability of the Roman solution and the importance of the aqueduct, not only for the water supply of the population but also for the preservation of the Roman monument, which was of great cultural and historical importance for the Austrian Empire. A cultural and social criterion was therefore brought into the analysis.
The main technical advantage of the Roman water supply system was the position of the water intake at the spring of the Jadro River, where the quality of the water was the best. The elevation of the spring is 33 m asl, which provided the gravity-fed water system to supply consumers in the city. The solution was energetically sustainable and the system's reliability was relatively high. In addition, the pipeline route is such that the water reaches the centre of the city at a favourable height, allowing easy gravity-fed water distribution to all parts of the city. The aqueduct capacity is very large, far larger than needed; therefore, water is intended for irrigation along the aqueduct's route. At the time, a significant part of the Roman aqueduct was preserved including complex and expensive tunnel sections (35% of the channel was completely preserved and 25% lacked a vault) (Bulić & Karaman 1927). The route and the concept of the water transport solution were known and reliable because the system had already functioned in Roman times. The risk was therefore insignificant (Figure 4).
The applied approach to the planning of the UWSS in 1880 took into account all the basic steps of the classic system approach, which is today the fundamental methodology for solving complex engineering problems (Jewell 1986). When selecting the most favourable variant of the water supply system, significant attention is paid to cultural and wider social values. In fact, the criterion of the preservation of cultural heritage seems to be quite significant when making the final decision.
In 1855 the mayor of Split, Šimun Michieli Vitturi, entrusted architect Vicko Andrić with the project design for the reconstruction of the Roman water supply system. Andrić carried out comprehensive research and documentation of Roman remains and tried to make the reconstruction as good as possible. In 1857 he went on a study trip to Rome; however, his work took too long and the project design was taken over by the municipal engineer Locati in 1859, who proposed three different channel design solutions at sites where the original was not preserved: (a) completely rebuild based on the original, (b) replace with stone pipes and (c) reduce the channel profile in those sections. Variant (a) was chosen. The reconstruction works were initiated by Mayor Antun Bajamonti in 1878. The water supply system was put into service on 14 March 1880 (Belamarić 1999). The city had approximately 14,000 inhabitants at the time.
Restoration of the aqueduct respected all technical elements of the original Roman aqueduct. The aqueduct was restored on the original route, with the original channel bottom slope and the original channel dimensions (Figure 4). Water, as in Roman water supply systems, was constantly flowing from the spring to fountains and other water users, while the surplus overflowed into the sewerage and the environment. Only a very few wealthy individuals had water piped directly into their homes (De Feo et al. 2011). The new part of the system included a distribution reservoir at the end of the aqueduct and a network of pipes for water distribution to fountains.
Restored parts of the water supply system of Diocletian's Palace
During the restoration of the water supply system of Diocletian's Palace, the water intake structure situated at the Jadro River spring and various parts of the channel which were not preserved, were reconstructed. A detailed analysis discovered that the water intake structure was reconstructed in the same manner as the Roman one (Marasović et al. 2014, 2017). The reconstruction project has not been preserved; however, the water intake structure from 1880 can be seen on a drawing of the spring from 1908 (Figure 5).
The channel built in the spring riverbed ensured the abstraction of water. This type of channel was set perpendicular to the direction of the water discharge (Tyrolean type of water intake). Downstream of the evacuation channel, a concrete submerged dam was built to increase the water level (33.85 m asl) (Figure 5). The evacuation channel is split into two channels after 5 m. Water flows from the channels to the collecting chamber (4.70 × 4.70 m) with the bottom placed at the same level as the Roman channel. The collecting chamber ended with a delivery channel, two sluice gates in a row, manhole, overflow and a vent pipe. A retaining wall was built just above the Jadro River spring in order to prevent rockfalls.
The dimension of the Roman free-water surface channel was 60/120 cm (2/4 Roman foot). It was entirely vaulted, except for the parts where bridges were present. The longitudinal slope varied from 0.065% to 0.266%, while the average slope was 0.135%. The channel could be entered through inspection pits (60/60 cm), which were preserved at several sites. The distance between inspection pits varied significantly (from 71 m to 15 m). The channel capacity was 300–400 L/s for the flow depth of 60–80 cm, corresponding to a daily water quantity of 26,000,000–34,000,000 L. According to today's standards (150 L per inhabitant per day) this amount ensures sufficient supply for 173,300–230,400 inhabitants.
Parts of the aqueduct channel that were not preserved were rebuilt, respecting the original route, elements, dimensions and material. However, the sections where the Roman vault was not preserved were covered with stone slabs.
Three large bridges – Karabaš (156 m), Bilice (69 m) and Smokovik (114.5 m) – were fully rebuilt as new bridges, whilst the Mostine bridge, which was only missing an upper part, was reconstructed as in Roman times. All the smaller bridges, which were not preserved, were also rebuilt (Katanić & Gojković 1972).
Three Roman tunnels existed on the aqueduct – 287 m, 120 m and 1,268 m long. All the tunnels were largely preserved in their original form together with the original channel (cross-section 60/120 cm). The longest tunnel had 32 vertical shafts (90/90 cm). The height/depth of the vertical shafts ranged from 5 to 18 m. The central and highest shaft was surrounded by a largely preserved quadruple flight of stairs (a tunnel entrance). The collapsed sections of the tunnels, shafts and stairs were reconstructed during the 19th century in the original Roman form, geometry and with the similar materials.
Newly constructed sections of the UWSS
The newly constructed sections of the UWSS of Split include a distribution or service reservoir, a distribution pipe network of cast iron pipes and 25 fountains within the city (Figure 6). The reservoir is perfectly preserved and consists of an entrance building, two underground compartments of 530 m3 and 250 m3, control building, operation equipment and access arrangements. It is a partially buried reservoir, with walls about 1 m thick, and is covered by a shallow vault. The compartments are 6 m high, the entrance building is 5 m high and the bottom elevation of the reservoir is 17.10 m asl (Figure 7).
The reservoir had the same elements as present-day reservoirs: an inlet channel, two outlet pipelines (the first is situated at an elevation of 19.87 m asl, while the second is at an elevation of 18.07 m asl), washout and overflow pipes. The overflow was at an elevation of 20.85 m asl, which represented maximum operational level. There were sluice gates in the inlet channel and valves on all pipelines except the overflow. Common rules and practices for the prevention of contamination or other chemical, physical and biological changes that are detrimental to the water quality were applied: water circulation, ventilation, prevention of contaminants and daylight, and thermal insulation. It has been established that the larger compartment was operational, while the smaller would serve as a substitute to ensure the continuity of water supply (security of supply) if the larger compartment was out of action. The volume of the reservoir with a unit water consumption of 50 L per inhabitant per day was sufficient for a 1-day supply for 15,000 inhabitants if the aqueduct was not operational.
The purpose of today's reservoir is to store the necessary amount of water required for the water supply in the area. To achieve this, the following should be provided: (1) equalise the difference between water intake and output; (2) maintain the required pressure in the water distribution system; (3) keep stocks in reserve in case of interruptions in the system; and (4) provide water for fire protection. Since the intake capacity was much larger than the peak water demand in 1880 and also after, it was concluded that the major function of the reservoir should be to maintain the required pressure in the water distribution system, as was the case of the Roman UWSS with the implementation of castellums (Monteleone et al. 2007). It is expected that the tank also had the function of providing water for fire protection purposes. However, we did not find anything in the available documentation on this issue.
A new branch-pattern network of pipes was used to distribute water to the users (Figure 6). Fountains were situated at the end of branches, and therefore, the distribution system did not have terminals or dead-ends because the water was constantly flowing out of the fountains (no need for flashing dead-ends). The distribution system was built with cast iron pipes of different diameters 200 mm, 153 mm, 111 mm, and 73 mm. The pipes were laid in the ground at a depth of about 1.5 m. The maximum head of water in the distribution system was about 20 m, while the minimum depended on the head losses in the system. The hydraulic design was not preserved; therefore, it is unknown how the distribution system was designed and calculated. It was established that the planned water supply network (Figure 6) could provide a water flow at the fountains of 0.2–1.0 L/s.
The maximum theoretical capacity of the pipeline is about 100 L/s for the 300-mm pipe (the second phase of the distribution system development) and about 40 L/s for the 200-mm pipe (the first phase of the distribution system development), i.e. a total of about 140 L/s. Given the capacity of the aqueduct, which is between 300 L/s and 400 L/s, it is obvious that the capacity of the aqueduct was much higher than the demand for the UWSS in 1880 when the system was first put into service. As the city expanded, the water consumption grew steadily. More and more direct connections have been installed with taps where the users can use the water. It is the so-called ‘closed’ system. Such intermittent use of water has resulted in an increasing unevenness of water flow into the distribution network (Figure 8). Pumping stations and associated reservoirs were introduced in the 20th century due to the development of the city in the area above the level of the old reservoir (20 m asl). The reservoirs store water during periods when the population demand is low in order to provide a flow when the demand increases, and hence, the water supply system from the end of the 19th century was gradually transformed to the water supply concept applied today (pumping system) (Figure 8).
The original Roman ‘open’ system gradually became a ‘closed’ system. The aqueduct had enough capacity until the 1970s when a new gravity-fed pressure pipeline was built parallel to the Diocletian's Aqueduct.
DISCUSSION AND CONCLUSION
The construction of the UWSS to Split in 1880 is a good example of the restoration of a Roman water supply system that had not been in operation for 12 centuries (Angelakis & Zheng 2015; Juuti et al. 2015). The configuration and technological characteristics of the Roman aqueduct were restored in a consistent manner. The solution for the water supply system to Split was carefully and meticulously selected by analysing several water supply options, taking into account the relevant sustainability criteria. The selected solution resolved the water supply to Split and also preserved a valuable historical monument – the Roman water supply system. The engineers had a broader visionary approach to solving the problem that considered not only the technical and economic sustainability but also the cultural and historical value of the urban environment. It can be concluded that planners of the UWSS of Split used a similar procedure as we did today. Half of the UWSS route, which was restored in 1880, is still operational. The planning process and solution for the water supply system to Split is a good example of the historical development of sustainable urban water supply systems. The Roman gravity-fed concept of aqueduct was used for the newest (2018) solution of delivering water under pressure via pipelines to Split. It is an example of a zero-carbon water supply system – running entirely on renewable energy. By studying the conditions and processes that influenced the development of ancient water systems, as well as their operation and management, it is possible to gain knowledge for future technological advances.