Hypogea 2023

100 The ancient aqueducts of Asolo (Italy): new investigations and acquisitions fossores (ancient miners) to build a tunnel, about 180 m long, at an average depth of 20 m (fig. 1). The north entrance of La Bot gives access to an 80 m long tunnel called Cava delle Monache. This gallery identifies the upper gallery of La Bot, while the hydraulically active conduit parallelly runs under it. For the first 60 m, the tunnel runs almost straight along a north-south direction, with transverse sections reaching 180 cm in width and 150 cm in height. The first section of the gallery is dug into conglomerates, it has no covering and crosses three rooms created by the collapse of clay lenses. From the progressive distance of 60 m, the upper gallery enters a layer of softer marls requiring, from the pillars to the barrel vault, a facing of bricks, which dates back to the 17th century AD, at the time of the renovation commissioned by the Serenissima Republic of Venice (Mondin, 2022). The upper gallery ends 80 m from the north entrance, with a brick wall, built to contain a major landslide coming from the mountain. The course of the tunnel also coincides, in this last section, with that of the lower hydraulic conduit, which begins to deviate in a southwesterly direction, probably to bypass a clay lens. Is the presence of these clay lenses that led to the excavation of the upper gallery, which we interpret as coeval with the lower hydraulic conduit, previously considered of medieval age (Riera, 1995), perhaps due to the larger size of the upper gallery, compared to that of the lower hydraulic tunnel. The cleaning of the base of the upper gallery, and the study of the bricks, which were an integral part of the vault of the underlying hydraulic tunnel (fig. 2), corroborate the thesis that the upper gallery of La Bot was a service gallery, used for extracting the material needed for the covering of the lower hydraulic work. (Mondin, 2022). From the north to the south, the access to the lower hydraulic conduit of the aqueduct is through the second of the five inspection wells, located along the floor of the upper gallery, 25 m from the entrance. The conduit crosses a poorly cemented conglomerate, which required the use (from the progressive 25 m to the 32 m one) of large slabs of soft stones for covering both the walls and the roof of the gallery. It is worth of note that the roof coverage is fulfilled with horizontally mounted slabs, differently of a short previous stretch of the gallery, vaulted with stone ashlars laid in barrels. The difference between the two vaulting techniques suggests that the horizontal slab mounting was adopted during later renovation works. Fig. 1 – Layout of the ancient aqueducts in the historic center of Asolo city: track 1, La Bot Roman aqueduct; track 2, Il Gattolo medieval aqueduct (graphic elaboration by M. Zago).

101 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa From the progressive 32 m to the 57 m, we find the conduit with bare walls without cladding, vaulted with stone ashlars, as in the previous section but resting directly on the rocky step (fig. 3), with a constructive solution that has a character of unicity (Riera, 1991). Southwards, the wall cladding changes again following lithological changes. In correspondence of a large clay bank crossed by the superior gallery (from progressive 57 m to 79 m), a third cladding technique is found. It consists of abutments covered with Roman Northern Italic sesquipedal bricks, with the vault laid with staggered bricks (stepped vault), covered by a further protective layer of sesquipedal bricks, visible at the bottom of the upper service gallery. At the proFig. 2 – La Bot Roman aqueduct: view of the external part of the key ashlars of the stone barrel vault, from the upper gallery (photo M. Zago).

102 The ancient aqueducts of Asolo (Italy): new investigations and acquisitions gressive 79 m the covering of the hydraulic changes again. Here, a sudden lowering of the vault marks the conjunction of the two different stretches of gallery, which were almost certainly built starting from the two opposite directions. From this point on, and up to the progressive 161 m, the hydraulic tunnel is still covered with sesquipedal bricks, but with a classic barrel vault (fig. 4). The dimensions of the tunnel are smaller in this section, down to minima of 65 cm width by 140 cm height. It is also possible to observe some deformations, caused by strong lateral thrusts, while some 19th century alterations (basins and walls for contain water) have modified the original appearance. The ancient hypogeum ends at the base of a modern internal stairway (from progressive 161 m to 175 m), which exit in Brugnoli’s square. This construction, dated to 1937 AD, completes the overall length of the hypogeum, which is about 230 m long. Based on both the morphological analysis of the two different sections of the hydraulic conduit, and the planimetric trend of the excavation, the authors believe that La Bot aqueduct was built in two different moments. In the first phase, excavations of the lower gallery advanced from south to north for a hundred meters, with the expectation of lining the specus with bricks, laid from the inside with the use of wooden arches, which were then removed. As they progressed, the fossores encountered increasingly less compact rocks, and they had to vary the direction of work two times at least, definitely stopping in correspondence of an important clay bank. In the second phase, excavations resumed from the opposite (north) side of the hill, first excavating the upper gallery, and then the lower hydraulic conduit until connecting with the tunnel previously dug. A confirmation of this thesis comes from the measurements carried out in the north section of the hydraulic conduit. It is evident that the transport and installation of the heavy ashlars and stone slabs could not have taken place from inside the conduit, but from above, therefore from the upper service gallery of the Cava delle Monache. The medieval aqueduct Il Gattolo The study of the medieval aqueduct of Il Gattolo is included in this work as a term of comparison with the aqueduct of La Bot. From documentary sources, its construction dates to the 11th century AD, and its function was a 170 m westerly extension of the water path from La Bot aqueduct, crossing the orographic obstruction of Mt. Ricco at its thinnest point. Its construction was probably dictated by the unavailFig. 3 – La Bot Roman aqueduct: stone vault ashlars, in the northern section of the hydraulic duct (photo D. Davolio). Fig. 4 – La Bot Roman aqueduct: brick barrel vault in the southern section of the hydraulic duct. At this point the gallery floods after heavy rains (photo D. Davolio).

103 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa ability of La Bot, as reported by some technical annotations from the late Middle Age (Mondin, 2022).The specus is 75 m long and runs with an almost straight direction from NW to SE, with unfinished sections of average width and height of 65 cm and 110 cm, respectively (fig. 5). At the end of the aqueduct there is a large cistern 13 m long by 5 m wide, originally covered with a brick vault and 2 m deep, from which the supply pipes of the Renaissance city fountain still start today. The specus has no lining, it is excavated in conglomerates, and a foot ledge (called risega) is engraved in the piers for almost its entire length, testifying that water flowed freely. Other elements characterizing this aqueduct are the presence of numerous lamp niches (lucerne), and an evident connecting step in the vault (at the progressive 56 m from the north entrance), which confirms that, also in this case, excavations were made from two opposite directions. The analysis of this medieval artifact suggests, albeit indirectly, the different excavation periods of the Cava delle Monache gallery of La Bot. The medieval excavation is in fact made in a coarser way, with smaller transverse sections allowing the passage of only one man at a time, and the 1 m step on the vault at the junction of the two trunks highlights some difficulties in managing the altimetric trend during excavations (Riera, 1991). These macro misalignment errors are not found in the Roman aqueduct, where the upper gallery seems specially sized to serve as a construction site gallery, with an original average section that is almost constant for the entire length, Recent investigations and acquisitions In the period 2020 - 2022, the La Bot and Gattolo aqueducts were the subject of a new in-depth study and documentation, carried out by some speleological associations of the provinces of Treviso and Venice. The work began with the production of new digital videophotographic documentation, useful for assessing the state of conservation of the artefacts by comparing them with the old photographic material from the 1990s. The use of new high-performance photographic equipments, with modern LED illuminators, has allowed the creation of a new photo-plan of the external cladding bricks of the vaults of the hydraulic duct, as well as the creation of three-dimensional video footages for assembling “immersive 3D projections. The need to perform a new topographic survey of the La Bot aqueduct has directed the authors towards the use of the new three-dimensional scanning technology LiDAR, a non-strictly professional tool first introduced in the entertainment arena by Apple on the iPad Pro 2020 device, and then available on the mobile product line starting with the iPhone 12 Pro smartphone. The LiDAR sensor scans the surrounding environment with a laser and shares the information with the device’s photographic sensors, thus allowing to obtain both very high-resolution stills and videos, and to focus on any object even in low light conditions generating, after post processing, a scan of the environment in augmented reality. The scanning function useful for topographic surveying is done through the App ‘Scaniverse’, an open-source software. In order to be able to geo-reference the cavity the scan took place, starting from the outside, on the same place where four markers are previously placed. For long scan runs, the use of intermediate markers is necessary to allow, during post-processing, to easily combine the various point clouds, saved by the device in separated files. The scan of the aqueduct took place with a series of scans about 20 m long each, with four markers placed at the point of interruption of the single scan. At the end of each scan, the following one is restarted, overlapping the previous one to resume the same markers in order to have unambiguous references. The optimal scanning method for this type of duct involves a slow passage through the center of the gallery, making an initial scan of the base plane and the vault, and then alternately scan first the right wall and then the left wall, without ever returning backwards. This scanning technique was used constantly for the entire length of the hypogeum, proceeding in sections of about three meters at a time in repeating the same movements, since the maximum scanning range allowed by the LiDAR sensor is five meters. In the upper gallery, about 80 m long, seven point Fig. 5 – Il Gattolo medieval aqueduct: central section of the specus (photo D. Davolio).

104 The ancient aqueducts of Asolo (Italy): new investigations and acquisitions clouds were generated, due to the high precision desired in the scanning of the collapsed voids and in the relief of the stone coverings present on the floor. The lower hydraulic tunnel, 136 m long, but smaller in size than the Cava delle Monache gallery, required the construction of only six point clouds, also dictated by the homogeneity of the lining for long stretches of the tunnel. Finally, the restitution of the entire threedimensional survey (fig. 6) was obtained by recomposing the various ‘point clouds’ with a post-processing process, managed by the ‘CloudCompare’ software. The working hours required for the survey phase is just one, while processing took about two hours. Three-dimensional print of a scale model of the La Bot hypogeum The three-dimensional survey using the LiDAR technology opens up to various possibilities about its renFig. 6 – Extract of the three-dimensional scan of the Roman aqueduct La Bot, obtained with LiDAR technology. Frame 1: external view of the scan; frame 2: internal view of the scan (processing by M. Pellegrini and M. Zago).

105 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa dering. One of the non-virtual options consists in the three-dimensional printing of what has been scanned, allowing an overall visualization of the shape and development in the space of the hypogeum. The 3D printing requires the creation of a solid model, namely the starting point for defining the paths necessary to physically deposit the printing material. Once the area of interest has been selected, the first processing phase consists in approximating the surfaces of the relative point cloud and assigning them normal vectors, which indicate the direction and orientation of the surface itself. In order to obtain a processable solid, it is very important that all adjacent surfaces have a normal vector with the same direction. It is hard to obtain this result if the scan do not completely cover the target areas, or details has much smaller dimensions than the entire scan. A cleaning phase of the cloud was therefore necessary for the printing of the southern terminal section of La Bot, where both the external vegetation and some internal details, such as the lighting lamps and a section of handrail, had to be removed from the scan. The creation of a continuous surface starts from the point cloud, on a second phase, using a plugin that allows to perform the ‘Poisson Surface Reconstruction’. This process evaluates the point cloud and reconstructs a single surface that can be exported as a mesh. This action produces an output in a scalar field, indicating the density of the points for each area of the surface, allowing that the regions with an insufficient number of points are excluded from the image. Once the surface has been created, it is possible to export the mesh in the ‘.stl’ format, for the second processing stage. The ‘Autodesk Meshgrid’ software was used to create the solid, starting from the mesh, by performing an offset of the surface, in order to create a solid wall. Creating an offset surface requires a continuous starting entity that has no intersections. This condition is difficult to achieve directly by the process described above, so the breaks are manually corrected, and all the areas that are disconnected or have intersections, are modified or eliminated. Before performing the offset, the object is brought back to the desired final scale, which for this model has been set at 1:100. The object thus obtained is ready for the creation of the offset surface. It can be created on the outside, if you want to keep the cavity true to scale in size, or inside if you want to create a “negative” of the cavity. The offset thickness can vary in the range 0.40 ÷ 3.00 mm. A smaller thickness allows the creation of a lighter object that better approximates the original surface, however creates a more fragile model that could present difficulties in the printing phase. Conversely, a greater thickness is less difficult to print and requires less support, but requires more material usage and longer printing time. In this specific case, the best compromise was obtained with a thickness of 1,5 mm. Finally, the surfaces are automatically interconnected by the software, creating a solid ready for export. A part of La Bot hypogeum was printed with the ‘Fused Deposition Modelling’ technology, using a Prusa i3 MK3 printer. This printing technology starts from a filament of polymeric material, which is deposited through a nozzle that superimposes numerous layers. The process starts with the creation of the code necessary to the machine to deposit the material, using the ‘PrusaSlicer’ software (fig. 7). Fig. 7 – View of the printing plate on the ‘PrusaSlicer’ software for creating the paths of the 3D printer (processing by M. Pellegrini).

106 The ancient aqueducts of Asolo (Italy): new investigations and acquisitions The software allows the setting of the object on the printing plane, dividing the model in horizontal sections, creating different layers with a thickness set by the user. A good compromise is a layer height of 0.20 mm, that was the value chosen for this print. The positioning and orientation of the object on the printing surface is of fundamental importance for the success of the whole process, using as few supports as possible. Due to the nature of the process, each layer of material must rest on the previous layer, and cantilevered geometries or walls with an angle of inclination greater than approximately 45° are not manageable. Exceptions are allowed for “bridge geometries” resting on supports no more than about 20 mm apart. This feature was exploited for the construction of the vault of the tunnel, without relying on internal supports. Once the model and the correct printing parameters have been set, the software generates the ‘.gcode’ file, containing the paths for the printer. The model that reproduces the southern part of the aqueduct is 26 cm long at the end of the print, and required about 7 hours to print with 55 grams of material used, supports included (fig. 8). Acknowledgments The authors thank the Municipal Administration of Asolo city and the Director of the Civic Museum, Dr. Cristina Mondin, for the cooperation during the last study campaign on the underground aqueducts of the city. Based on the research carried out, we can argue that the work of speleologists is still confirmed as a multidisciplinary practice useful, if not indispensable, for archaeological investigation in underground environments, primarily aimed at researching and conserving important underground architectural works otherwise destined, by their very nature, to oblivion. Bibliography Mondin C., 2022, Asolo invisibile, 232 pages, Nodo Edizioni. Riera I., 1991, L’acquedotto romano di Asolo (Treviso), in Quaderni di Archeologia del Veneto VII, pp. 181-197, Ed. Canova. Riera I., 1995, Asolo (Treviso): nuovi dati sull’acquedotto romano “La Bot”, in Quaderni di Archeologia del Veneto XI, pp. 183-187, Ed. Canova. Riera I., 2006, La Bot, in Sedran S., Speleo per tutti, pp. 138-149, Studio Grafico Miotto. Fig. 8 – Three-dimensional print of the entrance of the southern part of the Roman aqueduct La Bot (authors R. Sordi and M. Sordi).

107 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa 1   Associazione Cocceivs 2 Ministero della Cultura *  Reference author: Graziano Ferrari, Associazione Cocceivs, via della Grotta Vecchia 3, 80125 Napoli - associazione. [email protected] The Civil Forum cisterns in Pompeii (Italy) Graziano Ferrari1,*, Daniele De Simone1 , Alberta Martellone2 , Bruno De Nigris2 , Massimo Osanna2 Abstract Thanks to an agreement between the Pompeii Archaeological Park Applied Research Laboratory and the Cocceius Association, we investigated a rain-water drainage system located under the Pompeii Civil Forum. The system was recovered and cleared-up in 1900 and it is still in operation. The pivotal element of such system is represented by twin water tanks, designed to hold about 350 m3 of water. Their structure resembles the Jupiter temple favissae, ascribed to the 2nd century BC. The tank development is strictly related to the Forum evolution in pre-Roman and Roman age. We were able to enter and inspect the southernmost tank, well-preserved and still holding rain-water. This essay illustrates the safety measures applied to the exploration operations, the system description, evolutionary hypotheses, conservation and maintenance suggestions. The tank is 27.80 m long and 1.91 m wide. The overall height is about 4 m, but the bottom is filled with 2.90 m of thick debris and mud layer, collected after the 1900 clean-up. The northernmost tank has been identified through the outtake hole, although no vertical access is evident on the Civil Forum surface. Rainwater is still collected by Augustan age channels under the Civil Forum southern area and stored into the southern tank, where debris settles. Water then fills the northern tank. The excess water spills over the outtake hole and flows westwards into a 160 m long drainage channel under Via Marina, then reaching a square hole in the Samnite walls near Porta Marina. Keywords: Pompeii, Ancient drainage systems, Roman hydraulics, Samnite hydraulics. Introduction The hydraulic management of the ancient city of Pompeii is a subject of considerable interest for understanding ancient water technologies and the historical and social evolution of the city. Several researchers investigated the topic of fresh water supply in Pompeii, with particular attention to the castellum aquae at Porta Vesuvio and to the distribution network to the public fountains (e.g. Eschebach, 1998; Ohlig, 2000; Olsson, 2015). Some papers have been dedicated to private cisterns and latrines (e.g. Jansen, 2000), but little attention has been paid to the drainage systems of public spaces. Understanding these systems is of considerable interest not only from a scientific point of view, but also to allow accurate and sustainable management of the Pompeii site, through the correct re-functionalization of the ancient conduits. As early as 1861, the Director of the Excavations Giuseppe Fiorelli decreed that: “Contemporaneamente alle strade dovranno essere scavati e puliti gli antichi condotti di acque, e le fosse del loro assorbimento1 ” (Fiorelli, 1861). 1 “At the same time as the roads, the ancient water conduits and their drainage ditches will have to be excavated and cleaned” (translation by the authors). Thanks to an agreement signed in 2018 between the Applied Research Laboratory of the Pompeii Archaeological Park and the COCCEIVS Association, it was possible to examine a system intended for the drainage of rainwater in the Civil Forum, as well as its storage and delivery outside the ancient city walls. The key element of this system consists of two underground cisterns located under the very level of the Forum. They were intended to guarantee a reserve of water for collective use at a time when Pompeii did not have an aqueduct and the related distribution network yet. The paper presents the preliminary results obtained from an inspection inside the southern cistern, the only one whose accesses are currently visible on the Forum floor. Materials and methods The Civil Forum cisterns are connected to a network of underground ducts, that run at a little depth under the surface of the southern part of the Civil Forum and under Via Marina, with an east-west direction, until they emerge from the ancient walls just south of Porta Marina. In the modern age, the system was largely cleared of ancient deposits, in order to restore its rainwater drainage functions. Summary information on this cleaning operation, carried out in 1900,

108 The Civil Forum cisterns in Pompeii (Italy) emerges from only one publication (Cozzi & Sogliano, 1900), which ascertains the existence of a single cistern and supposes the presence of a second similar structure, located immediately to the north of the first one. An architectural survey of the southern cistern is also reported (Fig. 1). The restoration operation was followed by a structural investigation on the upper surface of the cistern system (Sogliano, 1925), which also demonstrated the existence of the second cistern, with interesting considerations about the possible genesis and evolution of the Forum (Fig. 2). Recently, the drainage system has been intercepted by the works for the construction of easy access routes on the surface, and this has allowed a new speleological exploration of the hypogeal conduits. During this research, it was possible to identify the intake and outtake holes of the cisterns and to obtain an initial documentation of their characteristics, including that they are still occupied by a considerable amount of water. In the southern portion of the Civil Forum floor, two circular stone manholes covers are visible, as parts of the Sarno limestone paving, whose construction is attributed to the early imperial age (Pesando & Guidobaldi, 2006). They allow access to the southern cistern, while similar manhole covers for the northern cistern are not evident on the surface. In accordance with the direction of the Applied Research Laboratory of the Pompeii Archaeological Park, an inspection was therefore planned and carried out inside the southern cistern, to visually establish its state of preservation. Given its peculiarity, in addition to the normal use of speleological techniques and equipment, the exploration required the draft of a specific safety protocol and the application of operating methodologies suitable to allow the presence of an operator in a confined, flooded space, filled with thick mud and with a narrow vertical access. On January 17th, 2019, with the support of ALES company technicians, a worksite was installed that included both the southern cistern western manhole and the access manhole to the drainage tunnels to Via Marina. A drainage pump was then activated, its discharge pipe positioned inside the underground conduit of Via Marina, thus avoiding water spills on the surface by using the conduit under via Marina in its original function. Before any drainage procedure was carried out, speleologists conducted an initial exploration inside the cistern to check its condition. The operation strictly followed the safety procedures for accessing a confined environment with suspected hazards, with the use of a five-sensors personal gas analyser (O2 , CO, CO2 , NH3 , H2 S). Dry suit and a metal frame for lowering the operator into the manhole were also employed. At all times, the operator was secured to the frame by means of a caving rope in case of an emergency rescue. In addition, a two-way voice communication protocol was activated to assess the operator’s safety conditions in the tank for the whole duration of the process. Only after the operator had completed the inspection and left the tank, the pumping started. Unfortunately, the intake pipe did not allow to reach the furthest and deepest part of the cistern, it was only possible to drain the water under the access manhole, about 0.50 m high. After removing the drainage pipe, a further inspection into the cistern was carried out, in order to document its overall condition. Some samples of hydraulic plaster were also collected for laboratory analyses. The operations were performed without any risk, despite the fact that the confined environment was particularly demanding. It was only observed that the operator’s footsteps in the mud caused the release of non-dangerous quantities of methane. System description The access pit is made up of an opus incertum facing, topped by a stone ring about 0.15 m high and about 0.50 m in diameter. The pit is circular and leads to the top of the vault of the cistern with a 0.8 m high vertical, near the western wall of the structure. In the area immediately below, there is an airspace of about 0.6 m Fig. 1 – Pompeii: the south cistern survey (from Cozzi & Sogliano, 1900). Fig. 2 – Pompeii, Civil Forum: plan showing the twin cistern position (from Sogliano, 1925).

109 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa in height and water 0.5 m. deep. The southern cistern has the following characteristics: • Measurements: length 27.80 m; width 1.91m; the total height, from the Cozzi survey (Cozzi & Sogliano, 1900), appears to be about 4 m; during the inspection, a free span of 0.60 m from the vault to the water surface and 0.50 m from the water surface to the bottom solid deposit was verified, at the western end of the tank. These data would imply a solid deposit thickness of about 2.90 m, decreasing to about 2 m at the east end. • Cross section: rectangular vaulted. • Roof: lowered arch vault, covered with hydraulic plaster. • Side walls: vertical, covered with hydraulic plaster. • Floor: not visible, as it is covered by water and a thick layer of fine debris and mud. In addition to the access manhole, located at the western end of the cistern, there is a similar manhole at the eastern end, and two inlets located in the centre of the shorter walls (fig. 3). The eastern inlet is clogged with soil and is no longer functional, while the western inlet still allows for rainwater to enter the cistern. Interestingly, the inlet holes open as breaks in the cistern walls. On the vault of the cistern, two sub-circular breaks covered with stone slabs are visible. These are probably holes opened to facilitate the cleaning operations of the 1900s. According to Cozzi’s survey (Cozzi & Sogliano, 1900), on the north wall of the southern cistern there are four windows in direct communication with the northern cistern. The limited amount of water that has been removed has not made it possible to identify these windows yet. The northern cistern has so far only been examined from its western outlet; its characteristics appear quite similar to those of the southern cistern. The northern cistern is equipped with at least one inspection pit at the western end, located above the discharge hole. It is assumed it should open on the level of the Civil Forum, currently not visible on the surface. It is also conceivable that there is a similar well at the eastern end. The walls and vaults of both cisterns are made of opus incertum covered with a layer of hydraulic plaster. It is currently possible to define the following water path (fig. 4): rainwater is drained by circular inlets made in the pavement of the southern sector of the Forum; water flows into a system of channels that conveys it towards the inlet of the southern cistern, working as a decanter for suspended solids while water flows towards the northern cistern through the connecting windows. The overflow of the northern cistern runs through the outlet towards the network of tunnels under Via Marina. Discussion Sogliano (1925) pointed out that the internal measurements of the south cistern are multiples of the Italic foot (27.5 cm), respectively 100 feet long, 7 feet wide and 14 feet high. Analogies were also found with the structure of the Temple of Jupiter favissae, located at the opposite end of the Civil Forum with respect to the cisterns and whose construction is attributed to the late 2nd century BC. (Pesando & Guidobaldi, 2006). Fig. 3 – Pompeii, Civil Forum, south cistern: the western side and the intake hole (photo G. Ferrari). Fig. 4 – Pompeii, Civil Forum: plan of the hydraulic system with present water flow directions (drawing by D. de Simone). Fig. 5 – Pompeii, Civil Forum: the south cistern present status (from Cozzi & Sogliano, 1900, modified). The mud filling is in brown, while the water left after the pumping operation is in cyan.

110 The Civil Forum cisterns in Pompeii (Italy) Although the Forum and the surrounding area have yielded much older traces, the analysis of the structures in the cavities allows to chronologically date the underground installations at least to the 3rd-2nd century BC. A whole series of changes in the urban structure occurred in this period. They followed the Hannibalic War when the city became an important centre in the district, thanks to the presence of the port and the river Sarno as well as to the position along the routes that crossed the Gulf of Naples. The oldest phases of the development of the drainage system of the Civil Forum could be placed in the 2nd century BC. At that time, the Via Marina - Via dell’Abbondanza axis perhaps represented the southern and lowest limit of the Forum. The twin cisterns system is located just south the above mentioned axis. Sogliano (1925) already indicated a limit along this axis, perhaps also in relation that, up to at least the 2nd century BC, the Forum could be divided into two distinct squares, placed at different heights and connected by a flight of stairs. This way, the square would fully fall within a widespread model in the Hellenistic period, similarly to the nearby Neapolis (Giampaola, 2010). In this phase, the cisterns had to be equipped with a rainwater collection system that is no longer recognizable today. In the Augustan age, the Forum underwent a series of interventions aimed at reorganizing the spaces, to ensure their correct use and functioning, including the repaving of the square using travertine slabs. As a result, the water disposal system was adapted, still functionally linked to the needs of the previous structure of the square, from the Sillan era (80 BC), which, as far as it has been possible to reconstruct through the excavation data, had been enlarged to the south. The new paving led to an increase in the walking surface, which can be observed thanks to the analysis of the facing of the western manhole cover of the southern cistern. The archaeological analysis of the cisterns and the relative drainage system can therefore provide new information to contribute to the definition of the evolutionary phases of the Pompeii Civil Forum. Several private cisterns are known in Pompeii, in the typical setting of the domus italica, equipped with an impluvium to collect rainwater. Similarly, thermal establishments were equipped with a dedicated cistern, typically placed on an elevated level. However, in the case of these specific cisterns, it is a system of rather considerable capacity, estimated at about 350 m3 , located under the main public space of the city. The strategic and social value of this system is easily understood. A similar water reservoir is known under the other main Pompeian public space, the Triangular Forum. In this case, however, it consists of a series of long and narrow tunnels, with barrel vaults (Osanna, 2019). In regard to the protection and maintenance of the system, the exploration allowed to obtain the following information: the south cistern appears to have a plan area of approximately 53 m2 ; at the time of the inspection, the height of the filling, estimated on the basis of the Cozzi survey (Cozzi & Sogliano, 1900), was approximately 2.9 m on the west side and approximately 2 m on the east side. This filling therefore has an estimated volume of about 130 m3 . If the cistern had only been emptied in 1900, the filling rate would therefore amount to just over 1 m3 /year and would be higher if the cistern had also been emptied again in later periods; the western inlet is placed 0.5 m above the current fill; on the basis of the estimated filling rate, the vent would therefore be blocked at a date that can be estimated between 5 and 30 years from present. Acknowledgements The inspection in the south cistern was possible thanks to the collaboration of cavers Berardino Bocchino and Elena Rognoni; the technicians of the ALES company took care of the workspace and of the drainage pump operations. The Hans Brand company in Milan provided the portable gas analyser used to check the safety of the air in the hypogeum. Alessandra Ressa proofread the text. Bibliography Cozzi S., Sogliano A., 1900, La fognatura di Pompei, Notizie degli Scavi di Antichità, pp. 588–599. Eschebach L., 1998, Wasserwirtschaft in Pompeji, in Cura aquarum in Campania, proceedings of the Ninth International Congress on the History of Water Management and Hydraulic Engineering in the Mediterranean Region, Pompeii, 1-8 October 1994, pp. 1-12. Fiorelli G., 1861, Giornale degli scavi di Pompei. Napoli, 160 pages. Giampaola D., 2010, Il teatro e la città: storia delle trasformazioni di un comparto urbano, in Baldassarre I., Giampaola D. (eds.). Il teatro di Neapolis, scavo e recupero urbano, Napoli, pp. 21-33. Jansen G. C. M., 2000, Systems for the disposal of waste and excreta in roman cities. The situation in Pompeii, Herculaneum and Ostia, in Dupré & Remolà (ed.) Sordes Urbis. La eliminaciòn de residuos en la ciudad romana, Roma, pp. 42-44. Ohlig D. P. J., 2000, De aquis Pompeiorum. Dissertation, Katholieke Universiteit Nijmegen. Olsson R., 2015, The water-supply system in Roman Pompeii, Dissertation, Lund University. Osanna M., 2019, Pompei, il tempo ritrovato, Rizzoli. 444 pages. Pesando F., Guidobaldi M. P., 2006, Pompei, Oplontis, Ercolano Stabiae, Roma, 502 pages. Sogliano A., 1925, Il Foro di Pompei, Atti della R. Accademia dei Lincei, Memorie, Classe di scienze morali, storiche e filologiche, s. 6, 1 (3), pp. 220-272.

111 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa 1   Associazione Cocceivs *  Reference author: Graziano Ferrari, Associazione Cocceivs, via della Grotta Vecchia 3, 80125 Napoli - associazione. [email protected]. Aqua Augusta in Campania New section under the Posillipo ridge (Naples, Italy) Graziano Ferrari1,*, Daniele De Simone1 Abstract The Campanian Aqua Augusta is one of the major civil works in Roman antiquity, built in the last decades of the 1st century BC to supply fresh water not only to the naval harbour in Misenum, through the famous Piscina Mirabile, but above all to the city of Puteoli and the luxury thermal installations at Baia. An estimated length of 105 km of the main axis alone, and at least 135 km with side branches, made it the longest Roman aqueduct of the time. It was also the only one to supply several cities. Researches carried out in the 16th and 19th centuries asserted the existence of an important side branch that ran along the Posillipo ridge towards Fuorigrotta-Coroglio; it is assumed to have reached the island of Nisida via a channel-bridge. This branch would be about 5 km long, but only a 250 m long section was known in the Discesa Coroglio area, near the Nisida end. Only in 2019 another section about 270 m long was explored, still near Discesa Coroglio. At the beginning of 2023, thanks to the authorization by the Extraordinary Commissioner for the Reclamation of Bagnoli, and with the collaboration of the facility management Invitalia company, the Cocceivs Association found a previously lost section in an excellent state of conservation. The development reaches 800 meters; this qualifies it as the longest known segment of the Aqua Augusta, also showing fourteen horizontal accesses (adits). The paper reports preliminary information on the new aqueduct section, with details on the adits and the channel and some information on the expected water flow amount. Keywords: Augustan aqueducts, Roman aqueducts, Ancient Campania, Ancient drainage systems, Roman hydraulics, Bagnoli reclamation action. Introduction The Campanian Aqua Augusta aqueduct is an ancient public work whose springs are located in the Apennine mountains, presently at Serino (Avellino). They are still collected for the water supply of the city of Naples. The main purpose of this ancient aqueduct was to supply the Phlegrean area, including the merchant port of Puteoli, the harbour of the Tyrrhenian Navy and the numerous public spas and noble villas. The development of the main axis was 105 km; several side branches added further development, estimated at 135 km at least (Keenan-Jones, 2010a; 2010b). This made it the longest Roman aqueduct of the time and the only one to supply several cities. As most of the Roman aqueducts, its course runs mostly underground (Abate, 1864). In the transition period between the republican and the imperial age, imposing noble villas were built in the Phlegraean Fields, requiring large quantities of fresh water. The two best-attested properties were located at the south-western end of the Posillipo ridge, overlooking the sea, in an impressive panoramic position, and on the nearby island of Nisida. The first (Gunther, 1913) belonged to the knight Publius Vedius Pollio, who bequeathed it to Octavian Augustus upon his death. The literary tradition located the villa of M. Licinius Lucullus on Nisida (Jolivet, 1987). These villas were provided with thermal establishments, fisheries and harbours, so needed a large fresh water supply. However, up to now, little is known about their hydraulic management systems. Research carried out in the 16th and 19th centuries asserted the existence of an important side branch of the Aqua Augusta, that ran along the Posillipo ridge towards Fuorigrotta-Coroglio and presumably reached the island of Nisida via a channel-bridge. This branch would be nearly 5 km long. In the Coroglio and Nisida areas, no large civil settlements are known, but the Coroglio side branch would probably have supplied the two above mentioned villas and possibly several other unknown ancient settlements. Recently, the Cocceius Association explored a long section of the Aqua Augusta in the Coroglio branch. Preliminary information about the channel characteristics are reported here. Literature Shortly before the middle of the 16th century, the Spanish Viceroy, Pedro de Toledo, charged Pietro Antonio Lettieri with the task to investigate the ancient Roman aqueduct, to verify the possibility of restoring its ancient functionality and bringing fresh water to the city of Naples, experiencing at that time a phase of high demographic growth. This feasibility study lasted four years, but the results have been lost.

112 Aqua Augusta in Campania. New section under the Posillipo ridge (Naples, Italy) Lettieri’s research is known only thanks to a manuscript containing a later summary report, drawn up in 1560. The manuscript, which is still available at the Naples National Library, was then printed by Lorenzo Giustiniani in 1803. From it, we can get the main literature information on the branch to Nisida: “… se sparteva in dui rami; et l’uno andava ad mano mancha per la falda dela pred. montagna de Posilipo dala banda de ponente per fi ala sua punta et de più passava più oltre per sopra archi fatti sopra mare per insino all’isola de Nisida; secondo appare evidentamente in molti lochi;...” (Lettieri, 1560, in Giustiniani, 1803, p. 403). Lettieri states that the branch left of the main axis, near the entrance of the Neapolitan Crypta at Fuorigrotta, ran along the slope of the Posillipo ridge up to the Coroglio headland, and crossed the sea towards the island of Nisida through a canal-bridge. He also states that the presence of the aqueduct was evident at several sites. Later scholars and commentators generally limit themselves to reporting Lettieri’s information. Only Italo Sgobbo adds that he explored a long section: “… il ramo, che ho in gran parte esplorato, seguiva a mezza costa la falda occidentale del monte [di Posillipo]” (Sgobbo, 1938)1 . Unfortunately, it does not appear that Sgobbo has ever published further information regarding these explorations. The Superintendence for Archaeological Heritage archive (catalogue: Bagnoli B 4/6) holds some information: on 15 June 1927, the officer Salvatore Albanese mentions the discovery of three aqueduct entrances about 20 m apart from each other in the Bagnoli area. They were 1.5 m high and 0.8 m wide, dug into the tuff along the hillside, with side lining in opus signinum up to a height of 0.6 m from the floor. Morphology, position and the fact of being located on the hillside seem consistent with the attribution of the sections to the Coroglio branch of the Aqua Augusta. However, a more precise position is not available. A 250 m long aqueduct section was known at the far end of the Coroglio branch, aside Discesa Coroglio (Mariniello, 1981). Within the Cocceius Association research on the Aqua Augusta, Ferrari (2019) provided a first synthesis of the knowledge on the Coroglio-Nisida branch and on a previously unpublished 80 m long section within the Pausylipon villa; finally, the paper defined preliminary hypotheses on further research on the Coroglio branch. In the meantime, again in 2019, a team of cavers reached an entrance in the Posillipo cliff, using a rope rappel, and succeeded in exploring an aqueduct section about 270 m long, again near the far Discesa Coroglio area (Palumbo et al., 2020). No channel section was known in the more than 3 km long (as the crow flies) Posillipo cliff between Fuorigrotta and Discesa Coroglio. 1 “The branch, which I explored in large part, followed the western slope of the mountain [of Posillipo], at middle height”. Translation by the author. The discovery In 2022, Mr. Giuseppe Scodes contacted the Cocceius Association via Facebook and explained that, as a boy, he and his companions frequented the area about 40 years ago and enjoyed exploring the channels. At that time the land was not guarded and the children ran around happily, being careful not to trespass on military areas. Furthermore, they had to avoid meeting smugglers, who happened to use some spaces in the channels. Now, access must take place through the reclamation area of Bagnoli. Until 1992, the Bagnoli area hosted the huge Italsider steel plant, but the cliff area connecting Bagnoli to Posillipo was left to grow wild. After Italsider’s shutdown and dismissal, the whole area was left abandoned; later on, it was the subject of several reclamation projects. Between 2007 and 2010, a Sport Park was established in the area next to the Posillipo ridge, with several structures designed for sport and leisure. Later on, the Italian Government appointed a Special Commissioner, in charge of accomplishing the Bagnoli area reclamation. Work is ongoing in areas next to the Sport Park. We asked for authorization to access, which was kindly granted by the Special Commissioner; thanks to the cordial collaboration of the managing entity, the Invitalia Company. In 2022 we made an initial inspection with Mr. Scodes, who showed us the area where their entrances were located, right on the tuff ridge wall of Posillipo. However, presently the ridge is heavily wooded and full of brambles, so we tried to go up to the slopes underlying the cliff, cutting brambles, reeds and branches and looking for accesses. Finally, on the 2nd January 2023 we succeeded in reaching an entrance on the tuff wall. Aqueduct description The recently discovered aqueduct section runs underground, within the Posillipo ridge, in the south-west Neapolitan area, between the Fuorigrotta and Coroglio districts (fig. 1). Locally, the Posillipo ridge raises as a vertical cliff in Neapolitan Yellow Tuff, with several volcanic seams and compacted debris layers. Above the cliff, the land slopes upwards to the Via Manzoni (Posillipo) urbanized area. At the foot of the cliff, fallen debris is present, covered by wild vegetation (blackberry bushes, ferns, bamboo). The flatland at the base is a presently unexploited strip alongside the Sport Park area; it belongs to the Bagnoli reclamation project. The aqueduct channel runs at approximately 38 m a.s.l. elevation, within the vertical cliff. Up to 800 m of tunnel were investigated, with fourteen horizontal entrances (adits). We provisionally numbered adits assigning the letter E (Entrance) to the first entered one, and a progressive Left (upstream – towards Fuorigrotta) number, or a Right (downstream – towards Coroglio) one (fig. 2).

113 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa The adits Adit E (fig. 3) is 0.69 m wide and 1.97 m high, with an unlined rectangular section and a flat ceiling. It tunnels 8.40 m into the cliff, before joining the aqueduct channel. The channel itself is 0.55 m wide. The following table shows the adits’ main characteristics. Fig. 1 – Digital terrain model of the Fuorigrotta-PosillipoBagnoli area, west of Naples. The white strips mark the Augusta Aqueduct course, the northern one the main branch, the lower one the Posillipo-Nisida branch. The green section is the newly explored one (from the Campania Region public GIS, modified). Fig. 2 – The adits of the explored section (from Google Earth, modified). Adit number Access Length (m) Distance from previous (m) Notes L3 Open 7.70 1 m higher than the aqueduct. L2 Open 6.76 24.33 At the aqueduct level. Two small rooms just inside, on the channel course. L1 Blocked > 5.06 60.22 Higher than the aqueduct. E Open 8.40 31.31 First entrance. At the aqueduct level. R1 Open 6.00 36.69 At the aqueduct level. There is a band of loose pyroclastic material on the ceiling. R2 Open 8.55 25.98 At the aqueduct level. Larger than other adits. Continues beyond the aqueduct channel. Pyroclastic band on the ceiling and on the sides. R3 Blocked > 3.68 11.98 Pyroclastic band on the ceiling. R4 Open 6.50 44.61 At the aqueduct level. R5 Open 9.03 57.48 At the aqueduct level. R6 Open 7.96 47.05 At the aqueduct level. Large room instead of adit. R7 Open 6.75 46.63 At the aqueduct level. R8 Blocked > 8.68 44.86 Aqueduct floor is 1.61 m lower than adit floor. R9 Open 7.40 36.83 Aqueduct filled with debris to 0.40 m lower than adit floor. R10 Blocked > 4.95 52.02

114 Aqua Augusta in Campania. New section under the Posillipo ridge (Naples, Italy) The adit width ranges from 0.55 m to 0.69 m, with the exception of R2, that is 0.68 wide at the entrance and up to 1.10 m wide inside, and of R6, that was purposely enlarged to obtain a 4 m wide room. Distances between adjacent adits are quite different. Probably, adit R2 was carved at a later time with respect to the aqueduct construction. In this case, the distance between R1 and R3 would be about 38 m, in range with the other distances, with the exception of R3-R2 and R1-E, which are shorter. The encasing rock is a Neapolitan Yellow Tuff with few enclosed volcanic bombs. Just in the R2-R4 area, a loose pyroclastic seam was met (fig. 4), causing ceiling collapses. This seam angles in from the cliff face, that is north-west (upper) toward south-east (lower). In some places, for example near R1 and R3, vertical fractures are visible, sometimes causing rock collapses. Fractures are parallel to the external cliff face. The channel Out of the overall 800 m surveyed passages, adits and side branches amount to 135 m, while the aqueduct proper channel totals 665 m. However, the channel is winding. The distance between the two current extremes is 542 m as the crow flies. Most of the channel consists of a specus between 0.52 m and 0.70 m wide, with a hydraulic plaster lining (opus signinum), 0.03 m thick, on the floor and at the lower 0.64 m of the side walls. The corners between the floor and the walls are reinforced with a 0.10 m radius quarter-round mould. The side walls linings are Fig. 3 – The first entrance (adit E) (photo by Giovanni Grasso – Associazione Cocceius). Fig. 4 – Pyroclastic seam encased into Neapolitan Yellow Tuff rock, at adit R2 (photo by Graziano Ferrari – Associazione Cocceius).

115 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa in turn covered by a rounded limestone deposit, up to 0.10 m thick at its mid-point (fig. 5). Above their lower hydraulic lining, the walls and ceiling are lined with a thin layer of mortar, greyish in some sections, reddish in other ones. In several sections, the floor is covered with a very thin layer of dust and some garbage, so the hydraulic section is evident: E-L1, L1-L2-L3, E-R1, R4-R5-R6- R7-R8. In these sections, one can easily walk upright inside the channel (fig. 6). Elsewhere, surface storms and floods spread tuff dust and soil inside the adits. This means that some channel sections had to be accessed by crawling on one’s stomach or on all four: L1, R3, just before R9 and just before R10. The two extremes too, beyond L3 and R10 are filled with tuff debris so as to prevent any further progress. Often, electric cable sheaths in various diameters are found in the floor debris. The presence of other garbage (wine bottles, rusty cans, etc.) highlights a relatively recent presence in the channel. For instance, at adit R8, a bright red writing dated 1960, and attributable to a US Army lieutenant, was inscribed in the wall. In 1960 the external area was occupied by a U.S. military settlement. At several points in the channel there are traces of bricking up with tuff stone masonry, now no longer in place; furthermore, nails were driven into the walls, to hold up electric wire for lighting, this too no longer in place. On the other hand, where the hydraulic channel crosses the adits, a low wall should have supported the opus signinum lining and kept the water inside the channel. However, any such walling has long been removed in order to allow an easy access to the channel. At adit R1, towards R2, the channel digging started straight, but soon met the pyroclastic loose seam together with a vertical fissure in the tuff ceiling. Ceiling collapses started, so the tunnellers were forced to stop digging and to support the ceiling and the walls with tuff masonry. The channel was dug to the left, inside the cliff, presumably to avoid fissures and to ensure the pyroclastic layer was on the walls and not on the ceiling. Similarly, along the channel at adit R3, a longitudinal crack fissure in the rock threatened a rock collapse from the ceiling. The channel was protected by two short masonry sections before and after the adit junction. Again, at adit R9 the tunnelling toward R8 started at first at a higher level than the required one. A side rock collapse happened, so the underlying channel was protected by a 4 meters long masonry vault and walls. Other similar examples are visible at adits R8 and R10. Just after R9, a 1,10 m high wall partly closed the channel. The cross-over between the channel and the R2 adit is the most intriguing part of the whole aqueduct secFig. 5 – Channel cross-section, with the hydraulic lining and the sinter deposit (photo by Graziano Ferrari – Associazione Cocceius).

116 Aqua Augusta in Campania. New section under the Posillipo ridge (Naples, Italy) tion. The adit is wider than the other ones and continues 17 m beyond the aqueduct channel into the ridge. Its construction looks to be later than the channel one. The function of this branch is currently unknown. Unfortunately, this branch also is almost completely blocked by recent powdery deposits. The aqueduct course shows several twists and turns, usually in the middle point between adjacent adits, due in part to errors in the excavation directions between two adjacent tunnelling teams, sometimes to the need to avoid areas where the encasing rock is affected by seams of loose volcanic material. The latter case has already been mentioned at adit R1. Some very sharp turns and bends are present between R4-R5, R5-R6, R7-R8 (fig. 7), R9- R10. Often, in sharp turns, one can spot the abandoned straight channel which was then walled up and plastered. The junction between R8 and R9 is a special case, since the course was forced to divert further into the cliff in order to avoid a small valley (fig. 8). The analysis of channel longitudinal sections shows where tunnelling crews were forced to adjust incorrect junctions. At midpoint between L1 and L2, a little step in the ceiling marks the point where the two crews joined; an abandoned and walled up straight continuation is evident on the wall. Similarly, two steps in the ceiling between R9 and R10 mark the point where the crews joined with a sharp turn and the need to correct an elevation error. Discussion The surveyed 800 m long development results in the longest section presently known in the whole Campanian Aqua Augusta course. For the first time, a long continuous stretch of an ancient aqueduct in excellent conservation conditions, is available. It will be possible to obtain an accurate measurement of elevation differences between the flow levels at two distant points. Fig. 8 – Deviation from the straight line, in order to avoid a valley (from Google Earth, modified). Fig. 7 – Narrow bend between adits R7 and R8 (photo by Graziano Ferrari – Associazione Cocceius). Fig. 6 – The typical aqueduct cross-section (photo by Graziano Ferrari – Associazione Cocceius).

117 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa The archaeological analysis of the whole underground structure is promitting, in order to determine the methods of construction, the presence of flow control structures, branches, inflows and outflows. Up to now, the flow direction appears from north-east (Fuorigrotta) toward south-west (Coroglio), with a small elevation reduction across the known course. So, the inflow comes from beyond adit L3, while the outflow is beyond adit R10. In the middle, no side outflow is evident, but we still have to ascertain the function of adit R2. Ferrari & Lamagna (2015) reported on a preliminary analysis of several sections in the Aqua Augusta main branch, in order to estimate the aqueduct flow rate at several points. This analysis provided some hints at the fresh water distribution in the Roman time Phlegraean Fields. New data about the hydraulic section in the Coroglio branch can be added to previous data, in order to compare water flow in the main Augusta branch with the Coroglio branch one. Upstream from both branches, alongside the Crypta Neapolitana, the hydraulic section area is about 0.80 m2 . Downstream from Pozzuoli, the area is about 0.225 m2 . It looks like most of the Augusta fresh water flow was to Puteoli. However, the Coroglio branch hydraulic section is about 0.35 m2 . Thanks to the presence of a well-preserved sinter deposit, we can even estimate an average hydraulic section, that is, the one where the deposit is thicker: about 0.15 m2 . This means the Coroglio branch was a very important one, because it drained a fair percentage of the overall Aqua Augusta flow. Since no large urban Roman settlements are known in the Coroglio area, these data support the hypothesis that the Coroglio branch would have supplied the two main villas in the Posillipo and Nisida areas: the Pausylipon and the Lucullus villa on Nisida. A rough flow computation, with the assumption of a low 0.5 m/s flow speed, provides the figure of an average 75 l/s, that is 6480 m3 /d. The intermediate sections between two adjacent openings show evident errors in the excavation directions, both in plan and in elevation. This caused a winding morphology of the channel, in contrast with the precepts of hydraulic management, thus suggesting a poor quality of the workers involved in the construction of the aqueduct, probably combined with short construction times. As a final consideration, interviews with elder residents could provide information about the purpose of the apparently modern additions to the cavity, such as the large room at adit R6 (fig. 9). Fig. 9 – The R6 adit converted into a room (photo by Graziano Ferrari – Associazione Cocceius).

118 Aqua Augusta in Campania. New section under the Posillipo ridge (Naples, Italy) In conclusion, we believe that there are good chances for the definition of a research and exploitation plan about this important discovery, which adds a significant element to the knowledge of the ancient population of the Phlegraean Fields and to the modern development of the area. Acknowledgements We are grateful to the Special Commissioner for the Environmental Reclamation and Urban Regeneration of the Bagnoli-Coroglio Relevant National Interest Area, the Major of Naples, Eng. Prof. Gaetano Manfredi, to the Commissioner Structure, especially to Sub-Commissaries Eng. Prof. Filippo De Rossi and Notary Diomede Falconio, and to the Invitalia Company, especially to Eng. Michele Pizza, for the authorization to enter the Bagnoli reclamation area and for the lasting moral and logistic support. The Campi Flegrei Regional Park (President Arch. Francesco Maisto, officer Mr. Giulio Monda) granted permission to clean-up weeds and brambles in the area. Mr. Giuseppe Scodes’ friendship was decisive in re-discovering the neglected aqueduct section. The Hans Brand company lent a professional gas analyser, needed as a safety measure in exploring underground confined spaces, especially in a volcanic area such as the Phlegraean Fields. The Cocceius Association members provided invaluable support in the research, exploration, survey and documentation phases, especially Berardino Bocchino, who succeeded in finding the first entrance, and to Raffaella Lamagna, who helped in the survey sessions. Finally, supporting member Eng. (ret.) David Millar proofread the text. Bibliography Abate F., 1864, Studii sull’acquidotto Claudio e progetto per fornire d’acqua potabile la città di Napoli, Stamperia del Giornale di Napoli, Napoli, 78 pages. Ferrari G., 2019, Acquedotto Augusteo della Campania: la diramazione per Nisida ed il Pausilypon, Opera Ipogea, Journal of Speleology in Artificial Cavities, no. 2-2019: pp. 47-66. Società Speleologica Italiana, Bologna. Ferrari G., Lamagna R., 2015, Aqua Augusta Campaniæ: considerazioni sulle morfologie degli spechi in area flegrea, Atti del 22° Congresso nazionale di speleologia “Condividere i dati”, Pertosa-Auletta (SA), 30/05-02/06/2015: pp. 435-440. Günther R.T., 1913, Pausilypon, the Imperial Villa near Naples, Hart, Oxford, XII, 294 pages. Jolivet V., 1987, Xerxes togatus: Lucullus en Campanie. Mélanges de l’École française de Rome. Antiquité, no. 99: pp. 875-904. Keenan Jones D., 2010a, The Aqua Augusta. Regional water supply in Roman and late antique Campania, unpublished PhD thesis, Department of Ancient History, Faculty of Arts, Macquarie University, Sidney. Keenan Jones D., 2010b, The Aqua Augusta and the control of water resources in the Bay of Naples, Proceedings of the 31st Australasian Society for Classical Studies Conference, Perth, Australia, 2-5 February 2010, 18 pages. http://ascs.org.au/ news/ascs31/Keenan-Jones.pdf Lettieri P. A., 1560, Discorso dottissimo … circa l’anticha pianta, et ampliatione dela Città di Napoli et del’itinerario del acqua che anticamente flueva, et dentro, et fora la pred. Città per aquedocti mjrabili qvale secondo per più raggioni ne dimostra, era il Sebbetho celebrato dagli antichi avttori, Manoscritto presso la Biblioteca Nazionale di Napoli. In: Giustiniani L., 1803, Dizionario geografico ragionato del Regno di Napoli, Napoli: [s.n.], v. 6: pp. 382-411. Mariniello A., 1981, L’acquedotto augusteo nel tratto Napoli-Miseno, Mondo Archeologico, vol. 61: pp. 18-23. Palumbo M., Cristiano M., De Santo L., Ruocco M., 2020, Nuovi ritrovamenti e studio del tracciato dell’Acquedotto Augusteo che costeggia il versante occidentale della Collina di Posillipo (Napoli, Campania). Atti del IX Convegno nazionale di speleologia in cavità artificiali, Palermo, 20 marzo 2020, Opera Ipogea, Journal of Speleology in Artificial Cavities, no. 1-2-2020: pp. 129- 136. Società Speleologica Italiana, Bologna. Sgobbo I., 1938, L’acquedotto romano della Campania: Fontis Augustei Aquaeductus, Notizie degli Scavi di Antichità, 1938: pp. 75-97.

119 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa 1   Associazione Roma Sotterranea, Viale dell’Esperanto 71, 00144 Roma, Italy *  Reference author: [email protected] The water storage system of Marcigliana (Rome, Italy): an unusual representative of a Roman cistern Andreas Schatzmann1,*, Mara Abbate1 , Andrea Peresso1 Abstract The aim of this contribution is to present a multi-part hydraulic system located in the Parco della Marcigliana, a still completely undeveloped agricultural zone east of Rome. Among the elements of this hydraulic system is a still active water conduit lying under a plateau, the course of which is evident in the landscape from well shafts lined up at regular intervals. Where the plateau drops off steeply towards one of the side valleys, this conduit was intercepted from a lower level, thus diverting the water into a system of galleries lined with hydraulic cement (opus signinum) up to a considerable height. Structural features indicate these plastered galleries as a typical example of a cistern consisting of a network of impermeable galleries, a typical pattern of the 1st century BC in the Roman countryside. As a rule, such systems were fed by rainwater. In contrast, this is a rare case in which the cistern takes its water supply by secondary docking to an already existing water conduit, which in turn means that the latter remains profoundly impaired in its original function. In our contribution we will discuss different functional hypotheses and contextualization of the hydraulic system using a diachronic perspective. Keywords: water conduits, irrigation, water storage systems, cisterns, villae rusticae, defunctionalization of preceding structures. Introduction Southern Etruria, as well as the hilly countryside around Rome, is an ‘eldorado’ for cuniculus researchers. It is not surprising that, already during the 60’s, a systematic study was completed under the aegis of the British School at Rome, focused on the typologies and presumed functions of these water conduits. A distinction was made between channels leading a stream underground, parallel to the valley axis, and those diverting a stream from one valley to another (Judson & Kahane, 1963: 84). In a nutshell, one can say as a rule that a certain area was drained while another was irrigated, which means that the most frequent purpose was to create new agricultural land. A series of wells has been recognized, already at the beginning of the 1900s, on the plateau of the Casaletti farm house, at the outskirts northeast of Rome and within today’s Parco della Marcigliana (Ashby, 1906: 50) as a typical indicator of an underground water course (Quilici & Quilici Gigli, 1993: 109-110; site 22). These wells were visible also on aerial photographs (fig. 1; see on these photos Quilici & Quilici Gigli, 1993: 110, nota 32). Furthermore, at a lower level, at the edge of a side valley (41°59’33.9’’N/12°34’39.6’’E), the outlet of a tunnel can be seen, with an active water flow feeding a cattle trough by means of a metal pipe (fig. 2). The terminal part of this system was the object of a quick survey, in the context of a thesis in 2000, when an approximate plan was drawn (Dell’Era, 2000: 257-258). The dry areas, filled by soil, had not been accessed at that time, nor had the entire inner part of the system been analysed, with the consequence that the most revealing details of the structure had remained unobserved, and the structure as a whole had been erroneously regarded as a water draining system dating at Roman times. This hypothesis seemed confirmed, at a first sight, by the significant water flow encountered in 2011, when we began to inspect the corridors. Fig. 1 – Aerial photograph of the area (1943): the chain of wells is clearly recognizable (from Quilici & Quilici Gigli, 1980: tav. LXXXIX).

120 The water storage system of Marcigliana (Rome, Italy): an unusual representative of a Roman cistern A new functional explanation As it is shown below, this explanation is too simple. The case, like many others, underlines the importance of a complete inspection of the entire structure, using speleological techniques, for a correct functional determination. Corridor B (see plan in fig. 3) starts more or less at a right angle from the refurbished entrance (corridor A). It is completely dry and filled by soil, especially in its central part; it is necessary to crawl directly under the vault to reach its end. Despite being detached in various parts, the omnipresence of cocciopesto (hydraulic cement) strikes the eye in this zone. It covers not only the side walls, up to the start of the vault, but also part of the vault itself, which means that the water must have reached a considerable height here. Corridor B ends with a reasonably well-preserved plastered wall, on the sides of which two vertical quarter rounds reaching up the arch are clearly visible (fig. 4). About halfway along corridor B, a short passage C opens up on the left, probably connecting to a further corridor, parallel to B, but impossible to explore due to the fill which almost reaches the vault. However, at the end of the branch, a widening is visible, with a wall at the end, which could indicate the presence of an intersection with a new branch. The cocciopesto, together with the quarter rounds and with the apparently net-like structure of the branches, showing a dead end in corridor B, leads to the inevitable conclusion that the interpretation of a water catchment system cannot be correct 1 . Instead, these elements suggest the interpretation of the structure as a particular type of water reservoir, with several intersecting passages. Indeed, this type of cistern is very frequent in the countryside around Rome (cf. below). It is designed for distributing the wall pressure, caused by the accumulated water over several arms. The construction of vaulted cisterns with large chambers became possible only later, when the development of technology led to an improved stability of huge vaults in the I century AD. Turning back to the intersection of branches A and B, and going further ahead in the direction where the water comes from, we have to pass under a small road (via di Tor San Giovanni). In this section, on the original plastered corridor, a vaulted brick cladding was applied, which is very reminiscent of some hydraulic conduits built in Latium, in the realm of reclaim works 1 Dell’Era, 2000: 258 had assumed that the waters coming from two different sources would meet shortly before the outlet. Fig. 2 – Site entrance and location on the Istituto Geografico Militare (IGM) map.

121 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa of the first decades of the 1900s. The reinforcement of the corridor is directly related to the road running above it, which means that, whenever it was decided to enhance the stability of the water conduit, it was considered important not to interrupt it. After crossing the reinforced section, at the beginning of branch E, signs of a major collapse are visible almost in the entire stretch (fig. 5). In addition to the presence of large stones, the course is rather irregular, part of the ceiling is replaced with large brick boards, and another consolidation in form of an arched vault can be seen. However the most evident sign of a collapse is an irregular wall, erected with stone blocks that, at the arrival point of the reinforced corridor D, closes corridor E towards east. In other words, corridor E almost certainly once continued in this direction (marked as E’ and with dashed lines in the plan). If this assumption proved to be correct, E’ would come to join corridor C, therefore the open space mentioned above, visible from corridor B, could actually be nothing else than the point of intersection with E’. Only a few meters to the north-west, another intersection is encountered, after which the ceiling of corridor E has collapsed, leading to daylight over a cone of rubble. Continuing outside in the same direction, after a few meters there is another section of the corridor (about 30 m long), with exactly the same direction (E’’), which without any doubt represents the continuation of E. Fig. 3 – General plan of the structure (M. Abbate, A. Schatzmann; graphic restitution A. Schatzmann).

122 The water storage system of Marcigliana (Rome, Italy): an unusual representative of a Roman cistern This “detached branch” is plastered with the same type of cocciopesto, covered in some places by considerable calcite deposits (fig. 6). It is heavily filled by soil, whose thickness increases towards the end of the branch. Here, through a tiny opening, a small water conduit (M) is visible, running at a lower level and leading out of the tuffaceous ridge. There is a further shaft (L), too, with which it seems to be indirectly connected. The deeper parts of the system At the last intersection of corridor E, in the main part of the system, another corridor branches off at a right angle, characterized by a first straight section of about 3 meters (F), followed by a curvilinear course (marked with the letter G). While the original height can no longer be determined, due to the filling, the transition between the two sections is marked by a striking step in the ceiling height. Because of the low ceiling and the presence of water, section G is difficult to cross. The presence of well-preserved hydraulic cocciopesto suggests that the entire corridor G was still a part of the cistern up to its end. It is surprising to see that corridor G, at its end, seems to be walled up, intersecting a water conduit, indicated with I on the plan (fig. 3), whose floor is presently about 1.20 m higher than the current level of G. Originally, the height difference was probably more than 2 m, but this is impossible to establish due to the filling in corridor G. At a closer sight, it appears that the rock on which conduit I was running was undermined in an obviously purposeful act, creating a small and irregularly shaped basin in the middle of its former course. Its back wall was certainly cut at an oblique angle, to prevent the water from hitting the baffle with full force, and thereby releasing erosive energy. The effect of this, however, is that all the water arriving from the conduit falls down into the basin, and is diverted towards the cistern. The bottom of the basin has been deeply dug over time by the erosion of the falling water. Another effect of this drastic measure was that the original continuation of conduit I became a dry branch. As to block even more definitely the original water course towards west, a brick wall (referred to as H in the plan), reinforced with opus caementicium and plastered with cocciopesto (fig. 7), was raised on the part not cut out, and on the original floor level of the water conduit, further increasing the capacity of the cistern. The wall H, together with the back wall of the basin under the original course of the conduit, is therefore to be considered the limiting element of the cistern. As a confirmation of this, there are further vertical quarter rounds in two of the corners created through this water diversion system: one on the southwest side between wall H and the west wall of corridor G (the opening, visible today between this corner and the plastered wall of the cistern, is probably caused by subsequent erosion), and the second one on the inner edge of the basin (fig. 8). It seems that, in a second moment, the effectiveness of the block was affected when Fig. 5 – The collapsed part of corridor E with the refurbished ceiling (photo A. Schatzmann). Fig. 4 – The vertical quarter rounds at the end of corridor B (photo A. Schatzmann).

123 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa a small channel was dug on its surface. It may have served as an overflow, probably in order to prevent water from stagnating in the conduit I when the cistern was completely full. In that case the continuation of I towards west would have at least partly entered into function as a water conduit. While the dry branch is heavily filled by soil, it is possible to enter in the adductor for a few meters in the direction of the flow origin. From the basin one has to climb onto its floor, on which there are numerous blocks of stone, but also modern rubbish and glass remains, fallen in the well K and then carried by the water towards the cistern (some material is still encountered in the curved corridor G). After little more than two meters, a rectangular well K (60 x 85cm) opens up. The well, washed out in the lower part, is 10.4 m high and closed on the surface with modern hollow blocks. Under its opening there is the usual accumulation cone, which creates a sort of dam for the water coming from the east. This well is in perfect line with the chain of inspection shafts mentioned above, around which large craters have formed over time (fig. 1), even though no structures in exact correspondence with well K are visible on the surface. Continuing eastward from beneath well K, the progression in the conduit I becomes soon impossible, due to a layer of very dense mud about 80 cm thick. In any case, the tunnel is completely sealed off after 10 m, probably in correspondence with the filling of the next shaft. Fig. 6 – The detached branch E’’, and detail of its plastered wall, with lime deposit above it and digging traces below the plaster (photo A. Schatzmann). Fig. 7 – Intersection between conduit I and corridor G (from where the photograph is taken). The vertical quarter round is well visible in the left angle of wall H (photo A. Schatzmann).

124 The water storage system of Marcigliana (Rome, Italy): an unusual representative of a Roman cistern At the north-west edge of the studied area, in the forest along the foot of the tuffaceous ridge, other remnants belonging to the hydraulic system here described have come to light. A small channel opens from the rocky slope, revealing itself as the same structure (M) we have noticed below the terminal wall of branch E’’. Although it is impossible to establish by direct survey, M is unlikely to be anything else than a branch of conduit I. Its function could have been that of a nutrition/ overflow of branch E’’, but again this cannot be definitively determined because of the overall soil deposit. About 5 m higher there are the borders of well L (fig. 3), whose present depth does not exceed 3.80 m, although it was probably deeper. On its south-east wall the vault of another channel is visible, but it cannot be crawled into because of the 20 cm free space remaining between the filling and the ceiling. As it is visible on the plan, this vault is located at a point that seems to correspond with a hypothetical extension of channel I, with which it has a comparable appearance. In this case, well L could be another example of the shafts accompanying at regular intervals the water conduit I. The interpretation of all these structures, however, leaves open a number of questions. It is not clear, for example, if L was the last of the series of wells, or if conduit I continued in the same direction beyond well L. If such a continuation existed, it would be today deeply buried in the filling of the well. As already mentioned, also the existence of a connection with conduit M still needs to be checked. The most probable hypothesis is that channel M started as a fork from channel I, in the vicinity of well L. Discussion The system we describe is a typical example of a cunicular cistern, a widespread typology in central Italy and in agricultural villas of the I century BC around Rome. These cisterns are normally fed by drainage conduits or, when they are directly integrated into the platform of a villa, by various impluvia, and equipped with wells for drawing water (Riera, 1994: 313-330; Schingo, 2004: 53-56; De Franceschini, 2005: 306-307). From this point of view, it is immediately clear that we are dealing with a particular case. The quantity of water arriving from conduit I suggests that the latter was not dug as a classical water extraction tunnel. Rather, it is highly probable that it took, and still takes, water from an underground flow. It was certainly pre-existing to the water cistern, but at a certain moment, which for the reasons discussed above we believe to be the I century BC, it was intercepted in a very particular way, with the purpose of feeding a (newly dug?) cistern. This reduced its previous function of a water transport conduit to the exclusive function of an adductor tunnel for the storage structure, with its westbound continuation serving as a mere overflow. As regards the cunicular cistern, we were unable to find a directly comparable structure, but we believe that this is due to the state of research and to the general conservation situation of such systems. As a comparison with other areas in the immediate vicinity of Rome shows, such hydraulic systems are rarely preserved to a major extent, namely if they had not been destroyed by expansive construction activity or from intensive agriculture. For our cistern, a primary question is whether it was, like most of the others, originally designed as a rainwater cistern. One of the rare examples of a direct contact between a waterproofed cistern structure and a conduit for continuous water transport has recently been documented below the gardens of Villa Medici at Rome, where, in antiquity, the Horti Luculliani were located. But in this case things seem to have gone differently. In a first phase, a cunicular cistern filled with rainwater and drainage water met all the requirements of the horti. Later, this cistern fell out of use when there was the possibility to increase significantly the amount of water, with the help of an off-branch arriving from a nearby aqueduct (Fratini & Moriconi, 2015: 119-121). Fig. 8 – The construction that blocked the water flow, seen from east (photo A. Schatzmann).

125 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa Since in the explorable parts of the Marcigliana cistern no remains of impluvia are detectable,2 it might not be entirely unreasonable to assume a very unique and bold project. Its builders must have been inspired by the presence of the ancient water conduit under the plateau, whose shafts had always remained visible. So they decided to dig and divert the watercourse inside the hill, with the help of a cistern of the classical cunicular shape (this might also explain the curved design of corridor G). Once the storage capacity of the cistern was reached, water could continue to flow in the inner conduit. The conduit which determined the position of the cistern runs under a plateau where, on the base of sparse pottery findings3 , the presence of a Roman villa has been hypothesized. On the other hand, the cistern by itself does not extend under the plateau, but rather under the hillside that, like today, was probably covered by dense vegetation even in ancient times. It can therefore be deduced that, if such a villa existed, the cistern was almost certainly not in direct relation to it. Putting all the clues together, namely the absence of impluvia or drawing wells, which would instead attest a direct relation between the two (Mari, 1991: 39; De Franceschini, 2005), and the marginal position of the cistern under the hillside, but not under the plateau, the general impression is that our cistern was rather used to irrigate a structure different from a residential unit, probably an agricultural area that required not only a continuous water flow, but also a large amount of stored water. The small valley where the cattle trough and the entrance to the structure are located, is and was undoubtedly suited to agricultural cultivation. It is easily accessible from both the main valley of the Fosso di Tor San Giovanni and from a villa on the other side of the flat valley.4 It is evident that the cistern at a certain moment went out of function, and the system was reduced to a mere sequence of conduits (see the arrows on the plan, fig. 3) drawing water from the adductor out to a historical fountain, attested on the Catasto Alessandrino. This fountain was located at the same place as today’s cattle trough (Dell’Era, 2000: 258). The availability of a perennial source of water seems to have been highly regarded, deserving accurate restorations after any significant threat or disaster, such as the construction of the modern street above it. In extreme synthesis we can thus distinguish three phases in the history of use of our system: 1) A probably pre-Roman cuniculus, which ensured irrigation for agricultural purposes; 2) An apparently purposeful interruption of this cuniculus in Roman times and the drainage of its water in order to fill a cunicular cistern that most probably served for agricultural purposes; 3)  The flow of running water through the disused remains of this cistern, feeding the fountain in early modern times and, up to today, the cattle trough. Acknowledgements This study was commissioned by the former official of the Soprintendenza Archeologica di Roma Francesco Di Gennaro, whom we thank for his constant interest in the project. We are also extremely grateful to the current official, Marta Baumgartner, for her interest and permission for this preliminary presentation. Also we would like to thank Gianluca Schingo, a specialist of Roman cisterns, for the extremely useful discussion on functional issues of this typology. Many members of Roma Sotterranea were involved in the exploration and documentation process: Antonella Boccone, Maurizio Massimo Cappa, Gabriele Catoni, Vittorio Colombo, Donatella Ertola, Manuel Gentili, Luciano Meloni, Lorena Nubile, Gianluca Spuntarelli, Ivano Stranieri, Alberto Tancredi and Beatrix Tejero. Bibliography Ashby Th., 1906, The Classical Topography of the Roman Campagna, part II, Papers of the British School at Rome III: pp. 3-212. De Franceschini M., 2005, Ville dell’Agro Romano, L’Erma di Bretschneider, Roma, 564 pages. Dell’Era F., 2000, Villa e paesaggio: gli impianti idraulici, Bullettino della Commissione Archeologica Comunale di Roma LX: pp. 249-262. Fratini G., Moriconi F., 2015, Lo scavo dietro la falegnameria e le ricerche a Villa Medici, le indagini nel convento, nel giardino e nella Casa d’accoglienza della Trinità dei Monti: Una nuova lettura dei resti archeologici, in: Bollettino di Archeologia online VI: pp. 81-138. 2 We have to remind however that probably only a small part of its original extension has been preserved and remains explorable to these days. 3 Fragments of concrete and fragments of tiles, referable to the imperial era, and transported by agricultural work, are mentioned in Quilici & Quilici Gigli (1993: 110) in order to argue for the presence of such a villa on the plateau (see also Dell’Era, 2000: 258). However the observed remains could also be referred to a small secondary building (stable, granary or similar). But as mentioned above, the remains we found during our exploration were all of recent times. 4 The closest example of a known Roman villa is the ‘villa di Marcigliana’ close to today’s Casale di Belladonna (Di Franceschini, 2005: 48-51).

126 The water storage system of Marcigliana (Rome, Italy): an unusual representative of a Roman cistern Judson S., Kahane A., 1963, Underground Drainageways in Southern Etruria and Northern Latium, Papers of the British School at Rome XXXI: pp. 74-99. Mari Z., 1991, Tibur, pars quarta, Forma Italiae 35, Olschki, Firenze, 324 pages. Quilici L., Quilici Gigli St., 1980, Crustumerium, Latium Vetus III, CNR, Roma, 325 pages. Quilici L., Quilici Gigli St., 1993, Ficulea, Latium Vetus VI, CNR, Roma, 552 pages. Riera I., 1994, Le cisterne: Le testimonianze archeologiche: in AAVV. (eds), Utilitas necessaria. Sistemi idraulici nell’Italia romana, Progetto Quarta Dimensione, Milano: pp. 297ss. Schingo G., 2004, La cisterna a cunicoli, in Campitelli A., Cremona A. (eds), La Casina Valadier. L’edificio e il suo sito, Electa, Milano 2004: pp. 53-56.

127 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa 1   Martin (Szusz) Department of Land of Israel Studies and Archaeology, Bar-Ilan University, Ramat Gan, Israel *  Reference author: [email protected] An Archaeological Survey in the Jerusalem Hills and Water Facilities for Pilgrims during the Early Roman Period Boaz Zissu1,*, Danny Bickson1 , Dvir Raviv1 Abstract The Shalmon spur is located in the Jerusalem hills, about 6 km south-west of Jerusalem. The site dominates Nahal Refaim, a natural route leading to the city. Another ancient road to Jerusalem diverges from the main Roman road from Beit Guvrin to Jerusalem and passes nearby. The site consists of two separate concentrations of ancient remains: the eastern spur, where two towers and two ritual baths were found, and the lower slopes of triangulation point 691, where many ancient remains are covered by terraces. The pottery collected from the surface was dated to the Iron Age II and the Persian, Hellenistic, Hasmonean, Roman, Byzantine, and Early Islamic periods. The article presents the results of comprehensive documentation of this previously unexplored site. In this project, we surveyed remains of a Roman bath, a church or chapel, and various artificial cavities: two ritual baths, an impressive rock-cut cistern, open and underground quarries, rock-cut tombs, and more. One of the ritual baths (118) and one of the cisterns (107) are well preserved. These are among the largest water installations known outside Jerusalem. The article describes these two installations and discusses the possibility that they were part of the services provided by Jewish authorities to pilgrims heading toward the Second Temple during festivals. Keywords: Jerusalem Hills, ancient roads, ritual baths, cisterns, pilgrims, pilgrimage, Second Temple. The Shalmon spur is located in the Jerusalem hills, on the northern bank of Nahal Refaim, about 6 km west of Jerusalem (UTM 69962/35140). The spur is surrounded by a bend in Nahal Refaim and is essentially bipartite: the main branch turns to the southeast while the secondary branch heads west (fig. 1). This paper presents a previously unknown site on the slopes of the secondary branch of the spur and the south-western slope of triangulation point 691; these are separated by a small ravine (figs. 2, 3).1 The total settled portion of the site is estimated at 4 ha; the ancient ruins, including burial caves and other rock-cut features, are scattered over a total area of 17 ha. The site has been damaged by antiquities looters. The slopes of the secondary branch of the Shalmon spur (UTM 69974/35139) are steep and craggy, with exposed rock along most of their length. Aside from towers 111 and 112, this section was apparently not built on in antiquity. The built-up area of the site extended over the south-western slope of triangulation 1 “The site was surveyed by the authors on behalf of the Department of Land of Israel Studies and Archaeology at Bar-Ilan University (IAA lic. S929/2019), with support from the Jeselsohn Epigraphic Center for Jewish History. They were assisted by Yotham Zissu (measurements), Avner Ecker, Noam Bar-David, Yuval Shtober, Gilad Zissu, and Nurit Shtober-Zisu (geology and geomorphology). point 691 (around UTM 69970/351411). This area is currently covered by terraces that conceal the ancient buildings, and characterized by anthropogenic soil, remains of walls and foundations, rock-cut cisterns and other artificial cavities. Presumably, the poor preservation of the site is mainly due to later preparation of fields for cultivation, terrace building, rock removal, and so on. The top part of the site rests on the Kesalon formation, whereas the bottom part is on the Soreq formation. The geological sequence of the Soreq formation is characterized by beds of argillaceous dolomite with thickness ranging from 0.5 m to 1.0 m. These beds are interlayered with softer dolomites and marls. Additionally, in parallel orientation to the bedding planes, intermittent layers of quartz spherulites and crystals, along with chert concretions, can be observed. The Kesalon Formation overlies the Soreq Formation. The dolomite within this formation exhibits a more substantial lithology. Within the lower beds, there are occurrences of narrow bands of quartzolite, quartz crystals, and chert concretions (Shtober-Zisu and Zissu, 2018). The stratification of the Kesalon formation facilitated the quarrying of large rectangular stone blocks. This was done in open quarries (visible on the surface) and in underground quarries. The open quarries left cliffs, steps, and even a flat open area (no. 196). The under-

128 An Archaeological Survey in the Jerusalem Hills and Water Facilities for Pilgrims during the Early Roman Period ground quarrying created cavities of various sizes, some with straight, stable ceilings; after the quarrying was finished, these cavities were converted for use as cisterns, a ritual bath, underground storerooms, burial caves, and so on. The secondary Shalmon spur extends over the Nahal Refaim riverbed; its slopes are steep and craggy. Slightly to the east of it, in the opposite bank of the steep wadi, is a spur of Mt. Refaim. Between the two spurs is a narrow, winding passage dominated by the secondary Shalmon spur. An ancient road leading from the foothills to Jerusalem ran along the gentle route of Nahal Refaim, giving this spur added importance. Evidently, the road followed the channel of the wadi and was damaged when railroad tracks were laid from Jaffa to Jerusalem in the late 19th century. Another road, marked “Ancient Road” on the Survey of Western Palestine (SWP) map (sheet XVII, squares Fig. 1 – Map showing the location of the site against a backdrop of roads on the SWP map, sheet XVII (B. Zissu). Fig. 2 – Oblique aerial photograph to east showing the area with main locations discussed in the paper (B. Zissu).

129 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa uK–uL), split off from the main Roman road linking Jerusalem to the Coastal Plain nearby, slightly to the west of the Shalmon spur (See map of Roman roads in Tsafrir, Di Segni, and Green, 1994). This road ascended toward the southern slope of Mt. Shalmon via the Shalmon spur. According to the SWP map, this road followed a gentle route along the northern bank of Nahal Refaim, passed the village of Malha, and ascended to historical Jerusalem (see fig. 1). The importance of our site apparently stems from this location near the roads leading to Jerusalem from the south-west. In the center of the secondary Shalmon spur we found two towers (111 and 112), two ritual baths (118 and 119), an open quarry (196), two underground quarries (114 and 115), and additional remains (fig. 3). Two ritual baths were found on the north-western slope of the spur. The smaller one (119) is located 10 m north of the upper tower and slightly lower than it. The entrance to the ritual bath is approximately 0.9 m wide and is oval. Its walls were covered with light-colored plaster. Its ceiling collapsed or was deliberately removed when the ritual bath was converted into a storage and rainwater extraction facility. Ritual bath 118 is preserved in full and is one of the largest and most impressive ones known outside of Jerusalem (see discussion below). This ritual bath was carved out of the Kesalon formation rock and its maximum volume is estimated at 150 m3 (figs. 3,4,5,6). Presumably, there was a rock-cut, stepped vestibule at the entrance to the ritual bath in an area now covered with dirt, which took in surface runoff. The ritual bath has two archways separated by a square stone pillar. The northern one is exposed and currently serves as an entrance, whereas the southern one is blocked by later construction. The dimensions of the archways allow for easy entry (northern archway: 1.3 m wide, apFig. 3 – Vertical aerial photograph showing the location of the remains discussed in the paper (D. Bickson and B. Zissu). Fig. 4 – Ritual bath 118: plan and section (measurements: Y. Zissu; drawings: B. Zissu).

130 An Archaeological Survey in the Jerusalem Hills and Water Facilities for Pilgrims during the Early Roman Period prox. 2 m high; southern archway: 1.4 m wide, 2.1 m high). The cavity is rectangular (7 x 9 m). Its rock ceiling is slightly slanted due to the natural stratification. Seven rock-cut steps along the length of the installation facilitate descent to the immersion pool. The floor of the installation is exposed, with the exception of the bottom two steps and the pool itself, which are currently covered by water and sediment. The immersion pool is rectangular (6.7 x 4.5 m); its depth cannot be measured without cleaning and excavation. The floor and walls of the installation were covered with three layers of hydraulic plaster: (a) a foundation layer of gray plaster containing smooth pebbles; (b) light-colored compressed, smoothed plaster of a type familiar from late Second Temple–period water installations; (c) light-colored plaster containing crushed pottery and ribbed potsherds and used for repairs, especially in the lower portion. These repairs may have been carried out in the Byzantine period when the ritual bath was converted into a water storage facility. Significantly, no opening for the extraction of water, of the sort typical of later use, was made in the roof of the installation. The purpose of having two separate doorways in a ritual bath was to keep the ritually impure people going down to immerse apart from the ritually pure people coming up from the immersion pool. This arrangement is discussed in the Mishnah: “All vessels that are found in Jerusalem on the way down into the immersion area are ritually impure, whereas those found on the way up are ritually pure, as one does not descend and ascend the same way” (M Shekalim 8:2). Lieberman (1967) mentions two non-canonical sources that shed light on this halachic reality. The first is an external gospel discovered in Oxyrhynchus (Upper Egypt), which describes a dialogue between the Pharisee high priest and Jesus. Jesus asks the high priest if he is ritually pure and the high priest replies: “I am pure, for I bathed in the ritual bath of David. I descended by one staircase and ascended by another.” Fig. 6 – Ritual bath 118, view to the south-east (B. Zissu). Fig. 5 – Ritual bath 118, view to the north-west (B. Zissu).

131 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa The second source is the Letter of Aristeas to Philocrates, where he describes the Temple Mount: “There are steps too which lead up to the cross roads, and some people are always going up, and others down and they keep as far apart from each other as possible on the road because of those who are bound by the rules of purity, lest they should touch anything which is unlawful” (Letter of Aristeas, 106, trans. R. H. Charles; but see Regev, 1996). Ritual baths with a double doorway or a partition are discussed by Reich (2013); such installations were discovered first in Jerusalem and the vicinity (Reich, 2013) and later in the northern Hebron hills (Zissu and Amit, 2008), in Jericho (Netzer, 1978), Qumran (Reich, 2013), in the Jerusalem hills (Klein and Zissu 2012). in the Benjamin region and the northern Judean Hills (Raviv, 2018), and in the Lod Shephelah (Reich, 2013). The installation discussed here can be added to the seven ritual baths with double doorways in the A ritual bath of this size would have exceeded the needs of the residents of an ancient agricultural settlement. In our opinion, this ritual bath served pilgrims (see below). The slopes of triangulation point 691 are gentle and covered with broad terraces. After the settlement was abandoned, the area was prepared for agriculture by means of the construction of terraces combining segments of ancient walls and ancient building stones. Without excavating, it is impossible to determine the precise extent of the structures buried by the builders of the terraces. Two loci are particularly interesting: at no. 185 is a terrace that involved the use of massive amounts of Roman concrete; the walls of a building made of large ashlars were found nearby, as were vestiges indicating the existence of a bathhouse: Tubuli of the sort found in caldaria of Roman baths, fragments of tiles, ceramic pipes and tesserae. Rock-cut cavities survived among the terraces; e.g. small, square, rock-cut chambers (e.g., no. 109), at least seven cisterns (e.g., 101, 104, 105), and additional rock-cut installations. The most impressive of the remains at the site is cistern 107 (figs. 3, 7,8). The meticulously hewn cistern is located at the center of the site at the end of a broad terrace but is not disFig. 7 – Cistern 107: plan and section (measurements: Y. Zissu; drawings: B. Zissu). Fig. 8 – Cistern 107: view to the east (B. Zissu).

132 An Archaeological Survey in the Jerusalem Hills and Water Facilities for Pilgrims during the Early Roman Period cernible on the surface. It is rectangular overall (average dimensions: 11 x 10 m; maximum depth at present: 4.5 m from the ceiling to the sediment and stones covering the floor). Two monumental pillars left in place during the quarrying support the ceiling. The bases of the pillars are roughly square (southern pillar: 1.5 x 1.5 m; northern pillar: 1.8 x 1.6 m), whereas their bodies are rounded. The cistern was carved out of Kesalon formation rock, taking advantage of the natural stratification to create a straight, stable ceiling and erect walls. The walls were coated with at least three layers of hydraulic plaster: (a) a bottom layer of lightcolored plaster; (b) a layer of pink plaster containing fragments of ribbed jugs; (c) a top layer of gray plaster that was repaired (especially near the north-eastern corner) with ribbed potsherds. Near the south-eastern corner of the cistern, at the top of the wall, is a small square opening through which water flowed in, as evidenced by several layers of chalky sediment. We have no way of knowing whether this was surface runoff that flowed through a crack or pipe or a small spring that dried up. The cistern has two mouths: the southern (original) mouth from which the cavity was hewn (107 in the survey) has a square shaft measuring 1 x 1 m with a maximum depth of 1.9 m. The upper part of the shaft is lined with stones, whereas the lower part is carved out of the rock. The northern mouth (106 in the survey) was made in the ceiling of the cistern and seems to have been added in one of the stages of use, perhaps to facilitate extraction of water or filling of the cistern from an additional spot. The northern mouth measures 1.1 x 0.9 m and its shaft extends 2.5 m down from the surface. Here, too, the lower part of the shaft was carved out the rock, while the upper part is lined with stones. The top of the shaft is currently covered with an intact pierced stone. A large piece of another pierced stone, possibly originating in the southern shaft, sits on top of the hole in the first one. The cistern has features familiar from cisterns in southern Israel (e.g., pillars supporting the ceiling) (Kloner, 2002), but it differs from them in various ways, including the absence of an opening on the side. It is not clear what the surface looked and functioned in antiquity. The material that collapsed into the cistern includes large building stones originating in ancient structures that stood on the surface. Approximately 140 potsherds from the Iron Age II (6%), Persian period (3%), Hellenistic period (9%), Hasmonean period (16%), Herodian period (35%—a rather large quantity, including fragments of grinding tools and knife-pared or lathed stone tools), and Late Roman, Byzantine, and Early Arab periods (31%) were gathered from the surface of the site, and especially from the debris left by the antiquities looters. One finds elicited particular interest: a stone artifact, perhaps a Roman military bread stamp, inscribed with the name Terenti(us). Terentius Rufus was a Tenth Legion officer mentioned once by Josephus in connection with the destruction of the Second Temple (Josephus, Jewish War 7.31). It is, of course, impossible to connect his unit to our site with certainty. A total of 36 rock-cut tombs were hewn around the site. The majority were dated to the Early Roman period. Evidently, most of these family tombs (Kloner and Zissu, 2007) were created and used in the late Second Temple period. Their existence shows conclusively that this was a settlement site. Discussion and Conclusion The Nahal Refaim basin was densely populated in antiquity. There is a substantial amount of arable land in the channel of the wadi, and terraces were made on its slopes for growing unirrigated crops (Edelstein, Milevski, and Aurant, 1998; Amit, 2007). A series of springs made settlements and irrigated agriculture possible (Yechezkel, Frumkin, and Tzionit, 2021). Nevertheless, the site unique location – near arable land and at crossroads by a spur that is the only means of passage, and which dominates the road leading to Jerusalem through the wadi – created special conditions for the growth of a settlement and a way station that supplied services to pilgrims. The discovery of a group of rockcut family tombs indicates that it functioned as settlement site from the Second Temple period. The discovery of a monumental ritual bath with a double doorway and a cistern of impressive size enables us to associate the site with pilgrim traffic. These finds mandate a discussion of installations associated with the needs of pilgrims en route to Jerusalem in the Second Temple period. The pilgrimage to the Temple was one of the central events in the lives of Jews in the Second Temple period (Safrai, 1965). The presence of large numbers of pilgrims along the roads at festival times necessitated finding solutions to the problem of supplying water for drinking, bathing, and ritual purification. The Mishnah describes the situation: “And on the fifteenth of [Adar] … they repair the roads, streets, and ritual baths, and they do everything necessary for the public welfare, and they mark the gravesites” (M Shekalim 1:1; see also Moed Katan 1:2). The Tosefta (Shekalim 1:1) gives further detail: “And on the fifteenth of [Adar], agents of the court go out and repair the roads and streets that have developed holes during the rainy season, before the festival, just before the pilgrimage, so that they are in good repair on these three festivals. On the fifteenth, agents of the court go out and dig cisterns, trenches, and caves and repair the ritual baths and aqueducts. Every ritual bath that contains 40 seah is suitable for immersion; for any that does not have 40 seah, they extend the aqueduct to it and fill it up to 40 seah so that it is suitable for immersion.” Although the late date of redaction of these sources (early 3rd century AD and later) puts their reliability in question, as they might possibly be tendentious and might reflect an imagined situation (see, e.g., Mali, 2021 and bibliography there), the discovery of dozens

133 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa of water installations from the Early Roman period that were clearly associated with the Temple and with the main roads leading to Jerusalem (below) supports the assumption that these descriptions reflect actual procedures in the Second Temple period. There are more than 40 known ritual baths, pools, and other water installations near the Temple Mount. Some of the ritual baths are huge and have two doorways (Gibson, 2005; Reich, 2013; Gurevich, 2017). Ritual baths with two doorways have been discussed extensively by scholars (see above). As we explained, they are customarily regarded as a way of preventing physical contact between those going in and out of the bath as part of the purification process (Zissu and Amit, 2008; Reich, 2013; Raviv, 2018). Adler (2011) lists approximately 60 ritual baths with a partition on the staircase and/or two doorways. We now know of circa 70 such installations, some of them alongside roads leading to Jerusalem, together with cisterns and camping areas that clearly served the pilgrims, and others in settlements unrelated to the road system. Grossberg (1997) believes that the ritual baths with partitions were intended to serve pilgrims to Jerusalem, as they had to purify themselves in close quarters in order to enter the Temple. He discusses only ritual baths with partitions discovered in Jerusalem; he does not address similar ones in Second Temple–period Jewish settlements scattered around rural Judea. Adler (2011) discusses the nature of the partitioned installations. Based on the Mishnah (Hagigah 2:6–7), which describes an increasing scale of purity, he demonstrates that someone who was careful to maintain the required level of ritual purity for a certain level of sanctity would have to regard those who did so for a lesser degree of sanctity as being as impure as an av tum’a (a certain type of severe impurity). Therefore he believes that the need to prevent contact between people on different levels of purity led to the installation of partitions in ritual baths. This subject was discussed recently by Raviv (2018), who examined the geographical location of the installations in view of the pilgrimage routes, the topography of the public camping areas, and the function of the ritual baths used by the pilgrims. The discovery of ritual baths and cisterns hewn near the roads to Jerusalem has contributed to a new understanding of rabbinic sources and pilgrimage routes. Raviv notes the concentration of public ritual baths within one or two days’ walk of Jerusalem. Because ritual purity only takes effect at sunset after immersion, early immersion in these ritual baths enabled many of the pilgrims to enter the Temple as soon as they arrived in the city, or the next day (following another immersion at the entrance to the Temple Mount as required by early halachic sources) (Adler, 2011). This means that they would be able to enter the Temple, bring their meal offerings, and watch the priestly service as soon as they arrived in Jerusalem. Moreover, by making sure to purify themselves en route, they were able to come in contact with teruma offerings and other objects intended for the Temple, for which they would have to wait until sunset according to the Sadducean laws practiced in the Temple in the late Second Temple period. As for the remains of the bathhouse discovered at the site, because these have not been excavated, we cannot date them precisely within the Roman period, and consequently we cannot know whether the bathhouse was used in the Second Temple period or in the Middle/Late Roman or Byzantine period. In both of these periods, the site was bustling, as attested by the potsherds, the bread stamp that may be related to a Roman military presence, and some ruins of a Byzantine-period church (?). In the Middle Ages and the Ottoman period, the site changed beyond recognition. The area was cleared of stones and prepared for planting, and the ruins of the ancient buildings disappeared and were forgotten. The cistern and ritual baths continued to be used for water storage until recent generations. Bibliography Adler Y., 2011, Archaeology of purity: archaeological evidence for the observance of ritual purity in Eretz-Israel from the Hasmonean period until the end of the Talmudic era (164 BCE–400 CE), Ph.D. diss., Bar-Ilan University (Hebrew), 457 pages. Amit D., 2007, Remains of Jewish settlements from the Second Temple Period near Teddy Stadium, Jerusalem, Eretz-Israel vol. 28: pp. 152–158. Israel Exploration Society, Jerusalem (Hebrew). Edelstein G., Milevski Y., Aurant S., 1998, Villages, terraces and stone mounds: excavations at Manhat, Jerusalem, 1987–1989 (IAA Reports 3). Israel Antiquities Authority, Jerusalem, 149 pages. Gibson S., 2005, The pool of Bethesda in Jerusalem and Jewish purification practices of the Second Temple period, Proche-Orient Chrétien vol. 55: pp. 270–293. Grossberg A., 1997, Ritual baths in Second Temple period Jerusalem and how they were ritually prepared, Cathedra vol. 83: pp. 151–168. Yad Izhak Ben-Zvi, Jerusalem (Hebrew). Gurevich D., 2017, The water pools and the pilgrimage to Jerusalem in the late Second Temple period, Palestine Exploration Quarterly vol. 149: pp. 103–134. Palestine Exploration Fund, London. Josephus, 1961, The Jewish war, trans. H. St. J. Thackeray, vol. 3. William Heinemann, Harvard University Press, Cambridge, MA. Klein E., and Zissu B., 2011, Ritual immersion baths (Miqwa’ot) with double entrances in the Jerusalem Hills. in: Baruch E., Levin Y., and Levy-Reifer A. (eds.). New Studies on Jerusalem 18. Bar-Ilan, Ramat-Gan. pp. 225-245 (Hebrew).

134 An Archaeological Survey in the Jerusalem Hills and Water Facilities for Pilgrims during the Early Roman Period Kloner A., 2002, Water cisterns in Idumea, Judaea and Nabatea in the Hellenistic and Early Roman periods, Aram vol. 13–14: pp. 461–485. Aram Society, Oxford. Kloner A., Zissu B., 2007, The necropolis of Jerusalem in the Second Temple period. Peeters, Leuven, 820 pages. Lieberman S., 1967, One does not descend and ascend the same way, in Rosenthal E. S. (ed), P’raqim: yearbook of the Schocken Institute for Jewish Research of the Jewish Theological Seminary of America, vol. 1: pp. 97–98. Schocken Institute, Jerusalem (Hebrew). Mali H., 2021, A walled city: Tannaitic descriptions of the first fruits procession, in Noam V., Boyarin D., Rosen-Zvi I. (eds), To be of the disciples of Aharon: studies in Tannaitic literature and its sources: in memory of Aharon Shemesh: pp. 411–449. Tel Aviv University, Tel Aviv (Hebrew). Netzer E., 1978, Miqvaot (ritual baths) of the Second Temple period at Jericho. Qadmoniot vol. 11, no. 42–43: pp. 54–59. Israel Exploration Society, Jerusalem (Hebrew). Raviv D., 2018, The pilgrims’ ritual baths: a reconsideration based on discoveries in northern Judean Hills, in Zelinger Y., Frankel N. (eds), Studies on the land of Judea: proceedings of the 2nd annual conference in memory of Dr. David Amit: pp. 11–25. Etzion, Kefar Etzion (Hebrew). Regev E., 1996, Ritual baths of Jewish groups and sects in the Second Temple period, Cathedra vol. 79: pp. 3–21. Yad Izhak Ben-Zvi, Jerusalem (Hebrew). Reich R., 2013, Miqwa’ot (Jewish ritual baths) in the Second Temple, Mishnaic and Talmudic periods. Yad Izhak Ben-Zvi and Israel Exploration Society, Jerusalem, 352 pages (Hebrew). Safrai, S. 1965, Pilgrimage at the time of the Second Temple. Am Hasefer, Jerusalem, 252 pages (Hebrew). Shtober-Zisu N., and Zissu B., 2018, Lithology and the distribution of Early Roman-era tombs in Jerusalem’s necropolis. Progress in Physical Geography: Earth and Environment 42.5: 628-649. https://doi.org/10.1177/0309133318776484 Tsafrir Y., Di Segni L., Green Y., 1994, Tabula Imperii Romani: Iudaea-Palaestina: Eretz Israel in the Hellenistic, Roman and Byzantine periods: maps and gazetteer. Israel Academy of Sciences and Humanities, Jerusalem, 263 pages. Yechezkel A., Frumkin A., Tzionit S., 2021, Ancient spring tunnels of Jerusalem, Israel: physical, spatial, and human aspects, Environmental Archaeology. https://doi.org/10.1080/14614103.2021.1888613. Zissu B., Amit D., 2008, Common Judaism, common purity and the Second Temple period Judean miqwa’ot (ritual immersion baths), in McCready W. O., Reinhartz A. (eds), Common Judaism: explorations in Second-Temple Judaism: pp. 47–62. Fortress, Minneapolis.

Rock-cut settlement works

Roberto Bixio, 1998 Cappadocia Sunset on the rock-cut site of Uçhisar (Cappadocia, Turkey). Icon of the full original painting. (Watercolour and golden foil, 45×30 cm)

137 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa 1   Zefat Academic College, I.C.R.C. (Israel Cave Research Center) - [email protected] Quarried underground hiding complexes in the Galilee, Israel: new evidence for their use in the Second-Century CE Revolt against the Romans Yinon Shivtiel1 Abstract Since the discovery of hiding complexes in the Judean Foothills during the seventies, there has been almost complete consensus among scholars of the Bar Kokhba Revolt that these complexes closely parallel the description of the Roman historian Cassius in the third century CE. He describes how during the Bar Kokhba Revolt, Jewish rebels hid in underground spaces carved out by them in the Land of Judea, both as preparation for, and during, the revolt itself. The question as to whether the Bar Kokhba Revolt spread as far as the Galilee has been a subject of controversy for many years. Theoretically, Cassius has nothing to say about this region, and there is a paucity of other historical sources. When they do occur, they are unclear about the question of the Galilee’s participation in the rebellion, leaving the question unresolved. Until recently, the prevailing view has been that the Galilee took no part in the revolt, in light of the absence of clear historical sources or archaeological evidence. The exposure, exploration and documentation in the Galilee of 75 hiding complexes whose characteristics resemble those of complexes in the Judean Foothills sparked renewed interest in the search for evidence of the spread of the revolt into the Galilee, despite the virtual absence of historical sources. Due to typological similarities between some of the Galilean hiding complexes and those of the Judean Foothills, the above question has recently been asked with increasing frequency. Unlike the finds from the Judean Foothills, until recently the Galilean material finds provided insufficient evidence to establish a definitive new position on this matter. However, the discovery and excavation of 14 hiding complexes in the Lower Galilee revealed significant finds that increase the likelihood of the Galilee’s involvement in the Bar Kokhba Revolt. In my lecture, I will first present the theory, based on historical research, that in using the expression ‘Judea’, Cassius actually refers to the entire Province of Judaea, including the Galilee. Secondly, I will make a typological comparison between the hiding complexes in the Galilee and those in the Judean Foothills that have been firmly identified as having been used during the Bar Kokhba Revolt. Finally, I will present a number of recently excavated Galilean hiding complexes that have yielded finds dating from the period of the Bar Kokhba Revolt. Keywords: Bar Kokhba Revolt, Galilee, hiding complexes, Romans, underground. Since the 1960s, researchers have been exploring underground cavities connected by narrow, low passages quarried beneath the remains of ancient Jewish settlements in Galilee. The passages were deliberately cut with sharp, right-angled bends. They can only be accessed by belly-crawling, doubling up around the bends to continue, and eventually reaching man-made cavities large enough for a few families to sit or lie down in (figs. 1-5). Niches for oil lamps are hewn along the walls of the passages and cavities, which have yielded various finds including storage and cooking pots and tableware, coins, jewelry, arrowheads and oil lamps (fig. 6). Some of the passages are continuations from other underground cavities used as storerooms, ritual baths, cisterns, oil presses and sometimes even burial caves (fig. 7). From the beginning of their exploration, it was clear that these locations served as underground hideouts from an enemy who, even if those hiding there were discovered, would find it extremely difficult to reach them, weighed down with heavy equipment. These rock-hewn systems have been classified as ‘hiding complexes’. Since the beginning of their research, 80 such hiding complexes have been discovered in Galilee. 1 Similar systems, found in much larger numbers (c. 500) have been discovered in Judea, Samaria and the Land of Benjamin. Research in these regions has yielded abundant finds and enabled them to be classified according to quarrying methods and typology and, above all, dated to specific periods of distress during which the Jewish population retreated to underground shelters, mainly during the Roman period (Shivtiel and Osband 2019). We know of two fateful rebellions that had severe consequences for the Jews in their resistance to the Romans: the Great Revolt of 66–70 CE and the Bar Kokhba Re1 74 hiding complexes were recently published by the author (Shivtiel 2019) and the remainder are published in various articles (see references), other still under research

138 Quarried underground hiding complexes in the Galilee, Israel Fig. 2 – Hewn cavities reached by crawling (photo Y. Shivtiel). Fig. 1 – Hewn tunnel in hiding complex (photo Y. Shivtiel).

139 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa volt of 132–136 CE. Since the hiding complexes are not proper dwellings, they were evidently prepared by the Jewish population to hide from the might of the Roman army during the Great Revolt—when a force of 60,000 soldiers was sent to suppress the rebellion—as well as during the Bar Kokhba Revolt when the Romans sent an army of 70,000 soldiers to crush the second rebellion (Rappaport 1984:283; Gichon 2016:158). The hiding complexes were quarried out wherever there was Jewish settlement, in both Judea and Galilee. When it became evident that the hiding complexes discovered in Judea were comparable with those in Galilee, the same classification was adopted in the two regions (Shivtiel 2019:98–102; Shivtiel and Osband 2019). In investigating the use of the Galilean hiding complexes during the Great Revolt, researchers can also rely on the writings of Josephus, who describes all the events of the Great Revolt and the use of caves for hiding. Furthermore, Josephus distinguishes between natural caves and man-made caves (Shivtiel 2011:24–25; Shivtiel 2016a:180–181). In contrast, the paucity of historical sources regarding the Bar Kokhba Revolt poses a major problem for scholars. The only historical description of this rebellion was written by Dio Cassius, a third-century CE Roman historian. Apart from never having visited the country, Dio Cassius devotes barely two pages of his treatise to a description of the rebellion. The reliability Fig. 3 – Crawling along a hewn tunnel (photo Shahar Banin). Fig. 4 – Hewn passage between chambers (photo Y. Shivtiel). Fig. 5 – Long, meticulously hewn passage with right-angled bend (photo Y. Shivtiel).

140 Quarried underground hiding complexes in the Galilee, Israel of this source is further compounded by the fact that the original manuscript was lost and is only known from a later epitome by the eleventh-century Greek monk Xiphilinus (Gichon 2016:11). Nevertheless, Dio Cassius’s source specifically describes the fact that the Jews quarried out subterranean passages and cavities in order to hide from the Roman legions (Dio Cassius, Hist. Rom. 69.13 [Cary and Foster]). Elsewhere in the same source, Dio Cassius states that the Jews rebelled against the Romans in Judea. Based on this account, combined with the fact that no second-century CE destruction layers have been found in archaeological excavations throughout Galilee (with the exception of three sites, see below), many scholars concluded that the Bar Kokhba Revolt failed to reach the Galilee. Over the years, several archaeologists, the current author in particular, have mapped and documented some 20 of the 80 hiding complexes so far discovered in Galilee, recovering many second-century CE pottery vessels in the process. Archaeological excavations at 16 of these complexes have all yielded finds dating from between the period of the Great Revolt and the second century CE (Table 1). The table reflects a new archaeological-historical perspective on the Galilee and the defensive preparations for the Bar Kokhba Revolt made by its Jewish population. The hiding systems are evidence of a concerted, collective effort to build underground shelters by the Fig. 7 – Hiding complex leading from oil press (photo Y. Shivtiel). Fig. 6 – Niches for oil lamps (photo Y. Shivtiel).

141 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa Tab. 1 – Archaeologically excavated hiding complexes with second-century CE finds. Location Excavation type Sources Finds 1 Gush Halav Gisela (Jish) Salvage excavation Damati and Abu-‘Uqsa (1992) Greek inscriptions, one base and four lids of 1st-century CE jars. Oil-lamp niches. Assemblage dated exclusively to 1st–2nd centuries CE. 2 Horbat Meroth, Meroth synagogue Cleaning/partial excavation Ilan and Damati (1987) The hiding complex was identified as having been used in two periods: Early and Middle Roman. 3 Mount Hazon Kh. Hazzur Partial excavation Bahat, D. (1974; 1983) 1st-century CE potsherds, fragment of Sixth Legion Ferrata roof tile. Dated by Dan Bahat to the 2nd century CE. 4 Horbat Huqoq synagogue complex Cleaning/partial excavation Shivtiel Y. (2016b) 1st–5th-century CE potsherds. 5 Horbat Khukha Excavation (Barshad Dror) Aviam 2004:125. Potsherds dated to the second half of the 2nd century CE and ostracon with Hebrew inscription. 6 I‘billin Excavation Mukari (1999). Fragments of jars, stone vessels and pottery dated to the 2nd–3rd centuries C.E. Early Roman parednozzle lamps. 7 Horbat Ruma Kh. Ruma Salvage excavation Rochman (1986) Plaster dated to the first half of the 1st century at earliest. 1st–2ndcentury CE pottery. 8 Karm er-Ras, Kfar Cana. Area T and Area W Karm er-Ras Salvage excavation Alexandre (2008a; 2008b; 2008c). 11 intact jars of a type used during the Great Revolt. Two bronze Jerusalem-mint coins from the second year of the Great Revolt. Further excavations yielded 2nd–3rd-century CE finds and a habitation layer. 9 ‘Enot Sho‘im –‘En Mahel ‘Enot Sho‘im Excavation Leibner, Shivtiel and Distelfeld (2015). 1st-, 2nd- and 3rd-century CE pottery and coins; 2nd-century CE type gemstone ring. 10 Zippori citadel complex Saffuriya Excavation Strange and Longstaff (1985). Intact Early Roman jars, bone hairpin, pottery lamp and potsherds dated to 1st–3rd centuries CE. 11 Zippori, southwest of citadel Saffuriya Excavation Strange and Longstaff (1987). Middle Roman cooking-pots, jars and sherds. 12 Migdal Ha-‘Emeq Excavation Shalem (1996). 1st century CE coin from Tyre, coin of the Governor Felix, 3rd-century CE Severan coin, 17 4th-century CE coins, Early Roman pottery and glass. 13 Bet She‘arim Excavation Erlich, et al, 2022 2nd-century CE finds. Meager finds from one room may be of a later date (3rd–4th centuries CE), suggesting renewed use of the passageway. The passage was interpreted by its excavators as having been hewn in the context of the Bar Kokhba Revolt, with possible later use. 14 Geva‘ Parashim Excavation Safrai and Linn 1988: 208–9; Shivtiel and Safrai 2021. Mixed Hellenistic–Mamluk finds (from surface intrusion). Intact Byzantine Samarian lamp with a possible menorah motif. Various 1st–6th-century potsherds retrieved from building where the hiding passage was located. 15 Horbat Huqoq 2 Excavation Shivtiel et al. (2022) Predominantly 2nd-century CE pottery and other finds. 16 Horbat Beit Netofa Excavation Still under research Predominantly 2nd-century CE pottery and other finds.

142 Quarried underground hiding complexes in the Galilee, Israel inhabitants of Jewish settlements (fig. 8). New hiding complexes were hewn underground and previous ones dating from the time of the Great Revolt were expanded and reused according to archaeological reports. (some hiding complexes have also yielded finds from even later periods). The complexes therefore provide hitherto unrecognized archaeological evidence of the Galilee’s preparations for the Bar Kokhba Revolt (figs. 9-11). The discoveries call into question the interpretation of Dio Cassius’s historical source. Early researchers of the Bar Kokhba Revolt, although unaware of the existFig. 8 – Map of hiding complex distribution (drawing by S. Add).

143 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa ence of the hiding complexes, drew an important conclusion regarding the rebellion’s geographical extent from the work of Dio Cassius, which was overlooked by subsequent scholars. In their early work (Yeivin 1946; Alon 1955; Avi Yonah 1984) understood Dio Cassius’s use of the word ‘Judea’ as referring to the Roman Province of Judaea, namely the entire region ruled by the Romans after the Great Revolt. At that time, the Province of Judaea included all the coastal cities from Caesarea to Rafiah, Idumea, Judea and Samaria, Peraea, Galilee, and almost all the cities of the Decapolis. After the death of Agrippa II (92 CE), Agrippa’s provinces of Peraea, Tiberias, Migdal and the Golan were also added to the Province of Judaea (Avi Yonah 1984:68). In the third century CE, Dio Cassius was referring to the Roman province he knew of, and thus described the Bar Kokhba Revolt as having occurred throughout the entire country, not confined to the limited region of Judea. Discussion and Summary It is now apparent that we have two significant kinds of evidence attesting to the spread of the Bar Kokhba Revolt to the Galilee. The absence of archaeological evidence for the destruction of settlements in Galilee in the second century CE as a result of the suppression of the Bar Kokhba Revolt does not necessarily detract from this evidence, since (a) in view of the limited extent of archaeological excavations in Galilee, it is difficult to claim that the archaeological contains a sufficient number of Jewish settlements from which to infer their situation Fig. 9 – Plan of hiding complex in Galilee (drawing by A. Lev-Ari and Y. Shivtiel).

144 Quarried underground hiding complexes in the Galilee, Israel Fig. 10 – Map of hiding complex in Galilee (drawing by A. Cohen).

145 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa during the second century CE, and (b) three settlements that have been archaeologically excavated in this century (Migdal, Wadi Hamam and Tel Rahash) exhibit clear signs of destruction and/or abandonment during this period (Zissu and Eshel 2013:64, note 1; Leibner and Bijovski 2013; Aviam 2018:1–9). In any case, the evidence emerging from new research into the hiding complexes in Galilee, coupled with insight into the implications of Dio Cassius’s testimony – by which ‘Judea’ refers to the entire Province of Judaea – prompts renewed discussion into the likelihood of the Bar Kokhba Revolt in Galilee. It is now abundantly clear that the Galilee did not remain peaceful during the Bar Kokhba Revolt and that it prepared underground shelters for the eventuality of the revolt reaching the Galilee, an eventuality that can no longer be ruled out. Bibliography: Alexandre Y., 2008a, The Archaeological Evidence of the Great Revolt at Karm er-Ras (Kfar Kanna) in the Lower Galilee. Pp. 73–79 in The Great Revolt in the Galilee Catalogue 28, ed. O. Guri Rimon. Alon G., 1955, The Jews in Their Land in the Mishnaic and Talmudic Periods, II. Tel Aviv: Hakibbutz Hameuchad (Hebrew). Aviam M., 2004, Jews, Pagans and Christians in the Galilee (Land of Galilee 1). Rochester: University of Rochester Press. Aviam M., 2018, A Jewish Settlement (Farmstead?) on the Summit of Tel Rekhesh: a Contribution to Galilean History between the Two Revolts. Eretz-Israel 33:1–9 (Hebrew; English summary, 185*). Avi Yonah M., 1984, A Historical Geography of Eretz Israel. Bialik Institute, Jerusalem. Bahat D., 1974, A Roof Tile of the Legio VI Ferrata and Pottery Vessels from Horbat Hazon. Israel Exploration Journal 24:160–169. Bahat D., 1983, Mount Hazon – Hideout Complex. Niqrot Zurim 7:29–52 (Hebrew). Damati E., Abu-‘Uqsa H., 1992, Gush Halav. Excavations and Surveys in Israel 10/2:11–13. Erlich A., Binshtok D., Kaftory R., 2022, New Hiding Complexes in Beth She‘arim and Their Dating. JJAR, (Jerusalem Journal of Archaeology), 3/2 134-163. Fig. 11 – Rappelling into a cistern at the bottom of which is a hiding complex from the second century CE (photo V. Boslov).

146 Quarried underground hiding complexes in the Galilee, Israel Gichon M., 2016, A Star Came Out of Jacob, Bar Kokhba and His Time, Modan, Ben Shemen (Hebrew). Ilan Z., Damati E., 1987, Meroth: The Ancient Jewish Village. Tel Aviv (Hebrew). Leibner U., Bijovski G., 2013, Two Hoards from Khirbat Wadi H. amam and the Scope of the Bar Kokhba Revolt. Israel Numismatic Research 8:109–134. Leibner U., Shivtiel Y., Distelfeld N., 2015, An Early Roman Hiding Complex at ‘Enot Sho‘im, Central Lower Galilee. Tel Aviv 42:127–143. Mukari A., 1999, I‘billin. Hadashot Arkheologiyot 109:18*–20*. Rappaport U., 1984, A History of Israel in the Period of the Second Temple. Amikai, Tel Aviv (Hebrew). Rochman A., 1986, Excavations in the Hiding Complex at Khirbat Rumah. Niqrot Zurim 11–12:32–36 (Hebrew). Safrai Z., Linn M., 1988, Excavations and Surveys in the Mishmar Ha-‘Emeq Area. Pp. 167–214 in Geva: Archaeological Discoveries at Tell Abu-Shusha, Mishmar Ha-‘Emeq, ed. B. Mazar. Jerusalem: Israel Exploration Society (Hebrew). Shalem D., 1996, Migdal Ha-‘Emeq. Excavations and Surveys in Israel 15:31. Shivtiel Y., 2011, Hiding Complexes in the Galilee: Update and Renewed Discussion. Cathedra 142:7–26 (Hebrew; English summary, p. 189). Shivtiel Y., 2016a, Because of Midian the people of Israel made for themselves the dens which are in the mountains, and the caves, Historical Sources for the use of Rock-Cut Caves in Times of Distress. In the Highland’s Depth 6:175–200 (Hebrew). Shivtiel Y., 2016b, The Hiding Complexes Beneath the Huqoq Synagogue: a Comparative Perspective and a Test Case for Hiding Complexes under Synagogues in the Galilee and the Southern Hebron Hills. Pp. 175–204 in New Galilee Studies, Vol. 2, ed. T. Grossmark. H. Goren, M. Abbasi, Y. Seltenreich and Z. Greenberg. Tel Hai (Hebrew). Shivtiel Y., 2019, Cliff Shelters and Hiding Complexes in Galilee during the Early Roman Period: The Speleological and Archaeological Evidence. (Novum Testamentum et Orbis Antiquus), Gotingun Vandenhoeck and Ruprecht. Shivtiel Y., Zingboym O., Badihi Z., Berger U., 2022, A Hiding Complex from the Period of the Bar Kokhba Revolt at the Ancient Settlement of Huqoq. JJAR (Jerusalem Journal of Archaeology) 3/2:110–133. Shivtiel Y., Osband M., 2019, A Methodological Perspective on the Chronology and Typology of the Hiding Complexes in the Galilee, Pp. 237–259 in Cornucopia, Studies in Honor of Arthur Segal, eds. M. Eisenberg and A. Ovadiah. Rome. Shivtiel Y., Safrai Z., 2021, The Hiding Complex at Tel Mishmar Ha-‘Emeq (Abu-Shusha), Geva Parashim: a Jewish Defense System in a Gentile Polis. In the Highland’s Depth 11:97–120 (Hebrew). Strange J.F., Longstaff T.R.W., 1985, Sepphoris (Sippori), (II). Israel Exploration Journal 35:297–299. Strange J.F., Longstaff T.R.W., 1987, Sepphoris (Sippori). Israel Exploration Journal 37: 278–281. Yeivin S., 1946, The Bar Kokhba War. Jerusalem: Bialik Institute (Hebrew). Zissu B., Eshel H., 2013, Coins and Hoards from the Time of the Bar Kokhba Revolt. Hoards and Genizot as Chapters in History. [Hecht Museum Catalogue no. 33]. Haifa, Hebrew Version, pp. 59–65. Classical sources Dio Cassius, Roman History, Volume VIII: Books 61-70. Translated by Earnest Cary, Herbert B. Foster. Loeb Classical Library 176. Cambridge, MA: Harvard University Press, 1925. Online sources Alexandre Y., 2008b, Karm er-Ras (Areas A, B). Hadashot Arkheologiyot 120. https://www.hadashot-esi.org.il/report_detail_eng. aspx?id=600&mag_id=114 Alexandre Y., 2008c, Karm er-Ras (Areas C, D). Hadashot Arkheologiyot 120. https://www.hadashot-esi.org.il/report_detail_eng. aspx?id=602&mag_id=114

147 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa 1   OBRUK Cave Research Group, Açıkhava Apt. 16/7 Nişantaşı, Istanbul, Turkey - [email protected] Castle of Gaziantep (Turkey) Tunnels, Dwellings, Excavations and Earthquakes Ali Yamaç1 Abstract Gaziantep Castle is located in southeastern Anatolia, in the center of Gaziantep city, on a 35 m high hill. Most of the hill is natural, but there is a mound dating to the Early Bronze Age in the southwest corner of the castle. Artifacts dating to around 3750 BC were found during the excavations carried out between 2003 and 2005 in four different trenches. Although the first construction date of the castle is not known, it has been understood by archaeological excavations that it was built as a watchtower between the 2nd and 4th centuries AD and took its current form in the 6th century AD. On the other hand, the most interesting structures of Gaziantep Castle are not on top of this hill, but inside it. The tunnels, which start with two different entrances on the northwest outskirts of the hill, continue both downwards and upwards towards the castle, and there are two cisterns at the deepest point. These structures were explored and surveyed within the scope of the “Gaziantep Underground Structures Inventory Project”, which we started as the OBRUK Cave Research Group in 2012. Apart from this, there are defensive tunnels on the southwest outskirts of the hill and many rock-cut dwellings on the eastern side of the hill. Gaziantep Castle, which has been destroyed many times over the past centuries, was seriously damaged by two consecutive earthquakes of magnitude 7.8 and 7.4 on February 6, 2023. In this presentation, the underground structures of Gaziantep Castle will be explained. Also, it will be discussed how the recent earthquake’s devastation was effected by archaeological excavations, modifications made to underground structures, and inadequate castle restorations. Keywords: Gaziantep, Gaziantep Castle, underground structure, cistern, earthquake. Introduction Castle of Gaziantep is located in the center of this city in southeast Turkey and it is the symbol of the city (fig. 1, fig. 2 and fig. 3). Although there is no definite information about the first construction date of Castle, it is assumed that it may have been built after the second century AD. Two altars were unearthed during the 2003 excavation of the mound, located to the southwest of the hill. On one of the limestone altars dating to the Roman Period, the Greek inscription “To the Great Zeus” can be read. The altar was found in waste material over the Early Bronze Age layer, which was largely destroyed. The presence of an altar in the castle also evokes the existence of a temple. However, it is not possible to talk about such a structure today. Also, we still don’t have enough material to answer the question of whether there was only a castle in this period or if there was a city at the foot of the castle. Fig. 1 – Location map showing Gaziantep (after Google Mapselaboration A. Yamaç). Fig. 2 – Gaziantep Castle before the February 6, 2023 earthquakes. View from the northeast (photo S. Savcılı).

148 Castle of Gaziantep (Turkey) - Tunnels, Dwellings, Excavations and Earthquakes In a travel book, it was stated that “Antep was built on two hills, a stream flows between the hills, the city is three miles around, the Sacır river passes through the east of the city and there are aqueducts here, the castle was built on a round hill, and there are rock tombs around it…” (Pococke, 1743). There are no aqueducts in the city today, and Pococke may have mistaken the dwellings on the east side of the castle for rock-cut tombs. Although the finds we have are limited, the altars found in the excavations in the mound in the castle show the existence of a Roman period settlement in the area where the present city is located. When the existence of the castle and the existence of the city are considered as two separate issues, it can be said that the castle was a Late Hellenistic or Roman period outpost, a Byzantine period castle, the region gained its present shape after the Mamluk administration, and the city has been inhabited since the Bronze Age. However, when the information in the ancient written sources is evaluated, the city remained in the shadow of the ancient city of Dülük from the ancient period to the 10th century. According to the Chronicle of Mateos from Urfa, Dülük was a province, and its center was Antep Castle in the 10th century (Andreasyan et al., 2019). The situation of the region where Gaziantep is located in the Middle Ages is quite complicated. In 636, Antep and its surroundings came under Arab rule. In the following years, Antep region, which was the scene of Byzantine-Arab conflicts for a long time, was ruled by the Abbasids (782), Byzantines, Hamdanis (951), Seljuks (1084), Ayyubids (1183), Mamluks (1277), Mongols (1281), Dulkadirs (1390), Timur (1400), Karakoyunlular, and again passed to the administration of the Dulkadir Principality. After the city was a battleground between the Mamluks and Dulkadiroğulları for a long time, after the Mercidabık War in 1516, it passed from the Mamluks to Ottoman rule (Beyazlar, 2003). Gaziantep Castle is described as follows in a manuscript dating from the period of Nur al-Din Mahmud Zengi (1118–1174): “Perimeter of the wall of the citadel, 540 fathoms at qastmi, six towers; enclosure perimeters (haouch), 66 (?) 1/2 fathoms, three towers; bâchoûra under the markaz, 307 fathoms at the qâsimi, and five towers; median fortress, 343 fathoms to the qâsimi; small bâchoûra, 234 fathoms at the qâsimi; large inhabited enclosure, 382 fathoms 1/2 to the qâsîmî; enclosure of the gate of the citadel, 105 fathoms at the qasimi; and 3 towers.” (Cahen, 1940). A fathom is 1.83m, therefore, according to this source, the perimeter of the wall of the citadel is 988 m with six towers, and the enclosure perimeter is 122 m with three towers. On the other hand, it is evident that the castle underwent major changes after the 12th century, when this work was written. For example, even if the outer wall of the castle is close to this size, today there are 12 towers instead of six. The inner city wall with three towers mentioned in the article is more than 300 m long. Although the word ‘bâchoûra’ in the article is used in the sense of gallery, even ‘small bâchoûra’ is 234 fathoms, i.e. 428 m. However, the battlemented defense gallery, which is located on the outskirts of the castle and is described in detail below, is only 150 m long. The ‘bâchoura under the markaz’ (‘markaz’ is Arabic, meaning ‘center’) no longer exists. Tunnels and Cisterns The northwest part of Gaziantep Castle sits on rocky ground. There are two different tunnels in this part of the hill; the first of these tunnels, the entrances of which are almost close to the ground, reaches two different cisterns 11 m below the entrance level (fig. 4). Another branch of the same tunnel bends to the south 22 m after the entrance and leads up to the castle plain above with wide stairs carved into the rock (fig. 4 and fig. 5). The entrance of the second tunnel is 4 m further south than the first. There is a connection between the two tunnels just behind the entrances, but it is currently blocked. In a part of this second tunnel after the entrance, the ceiling has collapsed, and the tunnel does not have a ceiling in this part. After this section, the tunnel divides into two, and the section that turns to the north ends after 12 m. On the other hand, the other tunnel, which continues towards the east, reaches the castle by rising on neatly cut steps. West Side Defence Galleries The 150 m long, battlemented defence gallery, located almost on the ground on the western slope of the castle, is not original. This gallery, which was almost completely destroyed during the Independence War, was rebuilt during the restoration after 1950. Based Fig. 3 – Aerial view of Gaziantep Castle (Google Earth).

149 Fourth IC of Speleology in Artificial Cavities Hypogea 2023 - Genoa on the battlement hole in the western part of the castle, it is thought that this gallery was built on the bedrock where the moat of the castle started and formed the west, south, and east parts of the castle. Due to the excessive destruction of this gallery in the depressions in the south of the castle, it collapsed during the repairs made after 1950, and a new gallery with a symbolic 1.20 cm width and a loophole was built in its place (fig. 6). On the other hand, many old tunnels that lead up to the castle from this newly built defense gallery remain intact. East Side Dwellings Similar to the ‘West Side Defense Galleries’, there are more than thirty interlocking rock-cut dwellings dug into the bedrock on the eastern slope of the castle. When we, as OBRUK, started the ‘Gaziantep Underground Structures Inventory Project’ in 2012, these dwellings surrounding the entire east side of the hill were completely filled with rubble and clogged. A few years later, when a project was made to clear these rock-cut dwellings and open them for tourism, we objected that such an arrangement would reduce the static resistance of the structure. Even if this part of the hill is solid rock, the top cover is all soil, and even if these dwellings did not collapse, it could lead to erosion of the top cover. Despite all our objections, these dwellings were cleared and opened to tourism, and two years later, we are very sorry to see how right we were in our objections (fig. 7). Archaeological Excavations Artifacts dating to around 3750 BC were found during the archaeological excavations carried out in four difFig. 4 – Plan of tunnels and cisterns under the northwest part of the hill (drawing M. Egrikavuk and S. Çoltu). Fig. 5 – Large rock-cut stairway in the main tunnel climbing to the castle (photo A.E. Keskin).