This article is available at the URI as part of the NYU Library's Ancient World Digital Library in partnership with the Institute for the Study of the Ancient World (ISAW). More information about ISAW Papers is available on the ISAW website.

©2019 Corinna Rossi; text and figures distributed under the terms of the Creative Commons Attribution 4.0 International (CC-BY) license.
Creative Commons License

This article can be downloaded as a single file

ISAW Papers 16 (2019)

Egyptian cubits and Late Roman architecture: the design of the forts of the Kharga Oasis (Egypt)

Corinna Rossi1

Abstract: The Kharga Oasis, in Egypt’s Western Desert, is endowed with a scatter of Late Roman forts and fortified settlements. The architectural remains of Umm al-Dabadib, the largest and best-preserved site, were surveyed in 3D in 2014 at a high level of accuracy and precision. The ensuing 3D model allowed a detailed metrological study of the fort, which revealed that this building was planned and built using Egyptian cubits. The other Late Roman forts of Kharga have not been surveyed as accurately as Umm al-Dabadib, but nevertheless a careful analysis of their architectural remains reveals that the same conclusions may be extended to these buildings also. This article presents the results of this investigation, which represent the latest attestation of the use of this unit of measurement in architecture, and suggest that these buildings were the outcome of a joint Roman and Egyptian effort to guard the empire’s frontier.

Library of Congress Subjects: Fortification, Roman; Architecture, Roman.


Kharga is one of the large oases of Egypt’s Western Desert, together with Dakhla, Farafra, Bahriya and Siwa (figure 1). These and other smaller oases that punctuated the desert to the north and to the south were all connected by a network of desert routes, that linked Egypt with Libya and Sudan. During the Late Roman Period, the Western Desert oases marked the frontier of the Roman Empire; Kharga, in particular, represented an important desert crossroad, and was endowed with a scatter of forts and fortified settlements guarding its outskirts.2

Between 2001 and 2007 the northern portion of the Kharga Oasis was explored and studied by the North Kharga Oasis Survey, co-directed by the author and S. Ikram, that produced a first survey of all the antiquities that are currently visible in the area. The picture drawn by this survey work indicates that the northern area was densely inhabited and exploited from the Late Ptolemaic through the Roman period, until a dramatic drop that took place in the early V century AD.

Our knowledge of the previous periods is rather patchy. The desert around the oasis’ core offers plenty of evidence of prehistoric occupation, especially Neolithic, corresponding to a period in which the environmental conditions were significantly different, and the desert resembled a savannah. When the hyper-arid climate took over, the population concentrated around the main water sources but kept travelling across the desert, leaving trails of rock art, animal remains and travel equipment along paths and tracks; those crossing northern Kharga offer evidence of Early Dynastic, pharaonic and Graeco-roman travellers, thus suggesting that these routes were used continuously throughout Antiquity and beyond.3 So far, however, differently from other areas of the oasis, northern Kharga has not yet yielded any specific and incontrovertible trace of stable occupation earlier than the Late Ptolemaic or early Roman Period; perhaps future excavation will access levels that are currently buried.4

In 2014 the Italian Archaeological Mission to Umm al-Dabadib performed a photogrammetric 3D survey of the buildings called Fort5 and Fortified Settlement;6 the 3D model generated from the survey data allowed an accurate investigation into the metrology of the building,7 that led to the discovery that the Fort was designed and built using Egyptian cubits.8

The current article explores the possibility of extending the same conclusion to three other Late Roman forts, Qasr al-Gib, Qasr al-Sumayra, and partly also Qasr al-Lebekha, which were built in the Kharga Oasis in the same period, as part of the same strategic programme.

Figure 1: Map showing the position of the main oases in Egypt’s Western Desert. Drawing by C. Rossi.

The metrological study of the Fort of Umm al-Dabadib

The 3D surveys of the Fort and the Fortified Settlement of Umm al-Dabadib currently represent the most precise and accurate surveys so far performed on any Late Roman building in the Kharga Oasis. The precision achieved is 1 cm, suitable to represent the entire complex to a scale of 1:50; peaks of 2 and 5 mm accuracy were achieved in the surveys of the external walls and of the internal rooms of the Fort respectively.9

Thanks to the accuracy of the 3D model, in 2017 it was possible to investigate which unit of measurement was used by the builders; both the Roman foot (the pes of ca 29 cm) and the Egyptian cubit (ca 52 cm) were tested first by superimposing an artificial grid to the Fort’s plan, aiming at highlighting possible patterns. Although the cubit fitted better than the pes, the evidence was not conclusive; the crucial clue came from the analysis of the masonry. Two elements recur in the Late Roman buildings of Kharga: the use of mudbricks ca 33-34 cm long, 16-17 cm wide and 8-9 cm tall (here summarized as 34×17×8 cm), and the walls made of two headers and one stretcher. This wall composition is slightly over 50 cm thick, which is just the length of the Egyptian cubit.

A careful analysis of the dimensions of the mudbricks revealed that their measures correspond to 4, 2 and 1 reformed palms respectively (figure 2), that is, to the subdivision of the Egyptian cubit of ca 52 cm into 6 parts, which was introduced during the Twenty-sixth Dynasty and superseded the older subdivision into 7 parts.10

Mudbricks inevitably function as basic modules, as walls and corners and therefore internal spaces are generated by their (generally regular) composition. Partly as a consequence of this and partly as the result of an intentional choice, the three internal dimensions of the rooms of the Fort of Umm al-Dabadib correspond to either whole numbers of cubits, or to modular fractions of them. In addition, the use of the Egyptian cubit can be also detected in the general layout of the building’s plan, where the presence of two modules is noted, one corresponding to 5 and one to 3 cubits.11 The ruined state of the central portion of the Fort prevents, for the moment, a precise reconstruction of the dimensions of the internal courtyard. Hopefully future fieldwork will be able to clarify this point and contribute further information to develop and refine the metrological analysis of the entire building

Figure 2: The dimensions of the mudbricks used to build the Kharga forts compared with the reformed cubit. Drawing by F. Fiorillo.

Looking for evidence in the other forts

A reliable survey is the conditio sine qua non to perform a metrological analysis of a building. Among the Late Roman buildings of the northern portion of the Kharga Oasis, only the Fort and the Fortified Settlement of Umm al-Dabadib have been surveyed to a high level of accuracy in all three dimensions. The architectural surveys performed by NKOS on the other major buildings of the area were carried out single-handedly by the author or by N. Warner, with the aid of measuring tape and/or laser distance metre. As clearly stated in the description of the methodology that was adopted during the fieldwork, the resulting drawings must be considered sketch plans.12

Therefore, a detailed, accurate and reliable analysis as the one that was performed on the Fort of Umm al-Dabadib cannot be applied to the other forts. However, taking as a starting point the dimensions that were actually recorded, an overall analysis of their proportions and general dimensions is possible. The element that encourages this attempt is the ubiquitous presence of both the mudbricks measuring 34×17×8 cm and of the one-header-two-stretchers bonding technique used in the construction of the walls, suggesting that the cubit lay at the basis of the design of all these buildings.

The numerical value of the Egyptian cubit is ca 52 cm13 and essentially this value did not change when the cubit was ‘reformed’, as the change only concerned the number into which this unit of measurement was subdivided. The new subdivision into 6 (instead of 7) parts meant that palms and fingers, the submultiples of the cubit, became larger. In this analysis presented below, the palms must be always intended as reformed (larger) palms.

As mentioned above, the analyses presented here cannot be compared to the one performed on the Fort of Umm al-Dabadib for two main reasons: they have been conducted on a limited set of measurements and only concern the plans of the buildings. The original measurements were taken at the best of the surveyors’ ability and can be considered reliable with an average approximation around 5 cm. Not all horizontal dimensions could be recorded, as often debris and sand prevented access to some points, and only a very limited set of vertical measurements could be taken. In contrast, the photogrammetric surveys of Umm al-Dabadib and subsequent computer-based alignment and georeferencing of the data enabled the generation of a highly detailed dense point cloud, effectively representing all visible physical dimensions and characteristics of the structure. From this, a 3D model of the Fort was produced with a considerably much higher level of accuracy and precision as to structural dimensions and aspects of construction, that cannot be matched by traditional survey and hand measurements. To sum up, the existing surveys of the other forts cannot match those carried out at Umm al-Dabadib; nevertheless, if one bears in mind these limitations, a discussion of the available evidence can be extremely productive.

For the Late Roman forts of Qasr al-Gib and Qasr al-Sumayra, two schemes are presented below (cf. figures 3, 4 and 6): the measurements actually taken and recorded in 2001 by the author (in blue, above), and a schematic reconstruction of the dimensional pattern in cubits (in red, below). The measurements are given in metres, with one decimal figure in most of the cases, to remark that the average error may reach 5 cm. The red schemes are, in both cases, directly based on the blue measurements, thus offering an acceptable level of reliability.

The situation of Qasr al-Lebekha is slightly different. This fort was originally surveyed by P. Deleuze and C. Braun, and plans and sections to scale were published before NKOS started its operations.14 N. Warner later integrated the original survey by adding further details and notes; however, measurements were not taken again.15 This means that, while for Qasr al-Gib and Qasr al-Sumayra we possess the original measurements, in order to derive those of Qasr al-Lebekha one further step must be added, that is, they must be retrieved by measuring the published drawing. Plans of the latter appeared to have been produced to a scale slightly inferior to 1:100, and therefore one must take into account a plotting error, potentially ranging between 5 and 10 cm, which adds to the inherent difficulty of precisely measuring such a small drawing. This difficulty, in turn, negatively reflects on a reliable conversion from metres to cubits, in particular when palms are involved. For this reason, the analysis of the dimensional pattern of Qasr al-Lebekha is limited to only a few elements. The analysis of this building, in other words, is less reliable than then other two, and by itself would not be able to support the argument; however, as some of its elements appear to fit with the pattern suggested by the other forts, it eventually offers a ‘second-hand’ confirmation to the general interpretation.

Figure 3: Sketch plan of Qasr al-Gib showing some measurements in metres taken in 2001 and a corresponding dimensional scheme in cubits. Drawing by C. Rossi.

Qasr al-Gib

Qasr al-Gib, perched on top of a hill, stands out in the flat landscape and, from the distance, looks like a large and solid building. In fact, it is smaller than expected, and also quite ruined inside; most of the material that once constituted the upper floors must have been removed over the centuries, as the central space is relatively empty. The two rows of lateral rooms at ground level were well-preserved, and offered the chance to take several measures and to reconstruct the overall layout. Little information could be retrieved on the first floor, and even less for the upper levels.16

The measurements taken in 2001 indicate that the plan of Qasr al-Gib corresponded to a square 30×30 cubits, into which a slightly rectangular space of 23×21 cubits was occupied by rooms and other spaces (figure 3). The plan was divided into three north-south bands, each 10 cubits wide; the central courtyard most probably measured just 10×10 cubits. The only entrance gate for the building17 pierced the southern wall on the right side of the central 10-cubit module and corresponded to 3 cubits.

Inside, the two rows of lateral rooms were 5+1/2 cubits deep and 5 cubits wide; the visible niches (not marked on the red scheme) are all 1 cubit (slightly over 50 cm) wide. The area of the staircase is very ruined and only a few measurements could be taken, but the overall impression is that it was included in a 5×5 cubit square and revolved around a central element made of 1-cubit thick walls.

The thickness of the external wall corresponded to 5 cubits along the northern and southern sides, and to 3+1/2 cubits along the eastern and western sides. The thickness of the internal walls at ground floor measured 1 cubit along the central courtyard and 2+1/2 cubits inside the rooms along the two sides: this value composed by one central wall of 4 headers (corresponding to 1+1/6 cubit, or 1 cubit + 1 palm), flanked by two layers of mudbricks set vertically to form the vaults, each ca. 30 cm thick, inclusive of mudbrick, mortar and plaster. The combination of 4 headers is also the thickness of the walls that flanked the fort’s gate along the southern side.

Qasr al-Sumayra

The plan of Qasr al-Sumayra was difficult to acquire, as the building is reduced to a heap of ruined mudbricks; only a few measurements could be taken in 2001 and 2002, and only a basic subdivision of the internal space could be reconstructed.18

The outer line of the building’s plan corresponded to a square of 22×22 cubits, consisting of a 3-cubit thick wall that outlined an internal square of 16×16 cubits (Figure 4). The round buttresses at the four corners do not form sharp angles with the outer wall, but are rather connected by a continuous curve, which gives them a slight oblong shape in plan, extending for 2 cubits in one direction and 2+1/2 cubits in the other.

The internal 16×16-cubit space was subdivided into three north-south bands: two lateral stripes each corresponding to 6 cubits, and one central passage of 4 cubits. The width of the lateral rooms is difficult to establish; the room located in the north-eastern corner might have been 4+1/2 cubits wide, and 5 cubits deep. The roughly-built room located immediately to the right of the entrance appears to be only 4 cubits deep; thick debris prevented any further exploration.

Figure 4: Sketch plan of Qasr al-Sumayra showing some measurements in metres taken in 2001 and 2002 and a corresponding dimensional scheme in cubits. Drawing by C. Rossi.

Qasr al-Lebekha

The topographic survey carried out by Deleuze and Braun highlighted the marked irregularities that characterized the construction of this building, as well as the widespread use of curving walls, absent in the other forts.19

The presence of sloping walls, the addition of layers of masonry, the overall irregularity of the entire building, and the scant information on the dimensions of various architectural elements all combine into a picture which is not easily interpreted. The lack of precise measurements of the interior prevents the published plans (both by Deleuze and Braun and NKOS) from being considered precise in terms of dimensions and proportions of the rooms that filled the building. For instance, from the published drawings some rooms along the southern and western sides appear to be about 3.5 cubits wide, but the plotting error prevents us from establishing whether this approximate value corresponded precisely to 3 cubits + 2 palms (or 3+1/3 cubits, that is 1.73 m), 3 cubits + 3 palms (precisely 3+1/2 cubits, that is 1.82 m), or 3 cubits + 4 palms (or 3+2/3 cubits, that is 1.90 m).