Avec un réseau d’aqueducs de 222 km de long, Lugdunum se distingue des villes de Gaule romaine par l’ampleur de son système d’adduction d’eau. Les quatre conduits ayant été construits à mesure de l’accroissement de la ville, le nivellement des aqueducs a été contraint par l’altitude des points d’approvisionnement en eau suffisamment abondants et par celle des quartiers à desservir. L’objectif est ici de mettre en perspective, à l’échelle de la ville tout d’abord, les moyens mis en œuvre pour garantir une pente constante, contrôler les différents écoulements (structures de régulation et de maintenance) et franchir les obstacles – plus particulièrement les pentes fortes au moyen de chutes en marches d’escalier et de puits de rupture de pente –, et de les comparer avec d’autres aqueducs du monde romain, principalement de la partie occidentale.
With a network of aqueducts 222 km long, Lugdunum stands out from the cities of Roman Gaul for the scale of its water supply system. The four conduits were built as the city grew. Consequently, the levelling of the aqueducts was carried out under several constraints. Firstly, the water supply point had to be both sufficiently abundant to supply a city and at a high enough altitude to ensure a gravity flow of water to feed the high points of the city. However, a distinction needs to be made between the underground and aerial parts of the conduits, and the parts with special features such as chutes, dropshafts or siphons that affect the overall slope of the aqueduct. Moreover, the route had to be kept as direct as possible by limiting the number of engineering structures (bridge, arches, tunnels, etc.) in order to reduce the cost and difficulty of construction. In addition, a decision had to be made about the type of canal to be built, as this would affect the speed of construction and hence the overall cost of the project, but would also affect ease of maintenance: we can therefore distinguish between the simplest cases –slab-covered concrete canals– and conduits with masonry walls and a vaulted roof through which a maintenance man could walk.The objective here is to put into perspective, on the one hand, the means employed to guarantee a constant slope and to control the flows (regulation and maintenance structures), and on the other hand to overcome the obstacles, particularly the steep slopes (with chutes or dropshafts) and to compare them with other aqueducts in the Roman world, mainly in the western part.The examination of the four conduits demonstrated the small number of regulation and/or decantation structures found across the whole of the Lyon hydraulic network, with three exceptions. Firstly, on the Gier aqueduct, low-bottomed manholes are traditionally interpreted as a decanting system. The first is located upstream of the Couttange bridge and measures 0.90 m on each side and 0.32 m deep relative to the bottom of the canal. Further downstream, at Saint-Joseph, a second manhole of equivalent dimensions was discovered. Its location at the exit of the tunnel is not insignificant, since it was there to regulate the flow of water at these particular points. Furthermore, these two examples can be compared to the “Sottizon tank”, a structure 0.78 m long and 0.45 m deep on the Brévenne aqueduct located downstream of a chute. According to H. Chanson (2000, p. 52-53), this lowering of the canal was used as an energy dissipation tank rather than a settling tank. It is therefore surprising that there are almost no regulation and/or decantation structures on all of Lyon’s aqueducts. Taking into account their length and the large number of structures, they needed constant maintenance, and for this it was necessary for the water supply to be temporarily cut off. There is therefore a kind of contradiction between the quality of the Lyon conduits and the limited number of cleaning and control devices on the network as a whole.Secondly, in order to cross very steep slopes, Roman architects resorted to smooth chutes, stepped channels and dropshafts. The second type is well illustrated by the Brévenne aqueduct, on which inclined channels are interspersed with regularly spaced perpendicular structures (low walls, slabs, wooden beams) designed to slow down the water flow. At least six waterfalls would have been constructed on this pipe. The second chute is the only one to be archaeologically documented (Fage, 2000) and is located at Chevinay, where the aqueduct drops 87 m in altitude in 275 m, i.e. an overall slope of 32%. The canal –the bottom of which is covered with slabs– is delimited on each side by walls supporting covering slabs. But the originality lies in the presence of wooden beams at roughly regular intervals (between 1.95 and 2.40 m): these are square-section beams with sides of 0.30 m, the anchoring holes of which remain in the side walls. This set thus forms “flat, downward sloping steps” (Chanson 2000, p. 9). Chutes of this type have been found at Zama (Tunisia) and at Andriake and Perge (Turkey).In addition, another technical solution has been observed on the Yzeron aqueduct: this is the dropshaft. The underground canal connects to a vertical, quadrangular or circular shaft, at the bottom of which is another canal. When the slope is too steep, several successive shafts can be constructed in series. On the Vaugneray branch, a shaft measuring 1.90 × 1.14 m was identified which must have belonged to a group of seven shafts, approximately 40 m apart and each dropping 2.55 m in altitude on a 7% slope. On the Grézieu branch, two square shafts, each measuring 0.17 m, were discovered 490 m apart, with a loss of altitude of 38 m, i.e. a slope of almost 8%. If one considers the altitude loss in each shaft of 1.80 m or 2.55 m, 21 or 13 shafts would have been necessary. However, the sector studied would only correspond to a part of a much more developed cascade because of the steep slopes upstream (20%) and downstream (6%) of the sector studied, which would result in a much higher number of shafts. J. Burdy’s assumption that there are about 50 shafts at 30 to 100 m intervals over a total length of 2 km (Burdy 2002, p. 130) seems to be quite acceptable. These hydraulic staircases can be compared to those of Autun, Cordoba (Spain), Cherchell and Rusicade (Algeria).It is still difficult to assess the reasons that led the builders to favour one or another slowing system, chutes or dropshafts. The angle of the slope does not seem to play a determining role, since the slopes in the three types of slowdown system in the Roman world vary greatly from one conduit to another. The choices were perhaps governed more by economic factors and speed of execution. In any case, the identification of these slow-down structures in the Western Roman Empire and even beyond testifies to great technical mastery, and given the number of dropshafts suggested on the Yzeron aqueduct, it would seem that Lyon would have largely outdone Cordoba.
