This study was supported by the Partner University Fund, project « Great Oasis », with Pr Roger S. Bagnall (New York University, United States) and Gaëlle Tallet (University of Limoges, France) as principal investigators (2010-2013). Various members of the two teams involved in the project have been involved in this research at various stages and at different levels. To all of them we express our gratitude. The samples collected in the field were dated thanks to official permission from the Geological Department of Egypt. Prof. L. Calcagnile and his CEDAD team were involved in the C14 analysis through the collaboration of Paola Davoli in the Amheida Project. The English version was revised by Cary Bartsch. The authors thank Pr R. Bagnall for his useful comments and two anonymous reviewers who significantly improved the content of the manuscript.
1Environmental investigations in the Western Desert of Egypt have focused mainly on climate change since the Late Pleistocene, noting sparse human occupation during the humid Holocene (from 8000 to 6500 cal. BC), and identifying the modalities of desertification between 6500 and 4000 cal. BC (Haynes, 2001, and others, see below). During this period, human societies had to cope with changing environments and therefore selected refuges where water was permanently available (Kuper and Kröpelin, 2006). In the last dry period of the Holocene, permanent human occupation was restricted to sites presently named oases. Though the climate is generally thought to have been stable since 4000 cal. BC, limited climate pulsations may have occurred in a context of slowly decreasing groundwater resources. The scope of this environmental study is to document, using geoarchaeological methods, the use of water for irrigation in one habitable part of the Deir-el-Hagar area in the Dakhla depression since the onset of agricultural practices. The study builds upon previous research undertaken as early as the late 1970s, notably by Brooks (2006) and Brookes (1983; 1989a; 1989b; 1993) for the Dakhla Oasis Project (D.O.P.). The oldest recorded ages are from before 1400 cal. BC (Haynes, 1983, cited by Brookes, 1993). The Roman period and more recent settlements display sediments, several metres thick, accreted in association with irrigated agriculture on various substratum. These sediments are associated with “irrigation channels, defunct well-heads and enclosed potsherds” (Brookes, 1989a; 1993).
2The Amheida archaeological site (the Roman Trimithis) stands in the NW part of the Dakhla depression (fig. 1-2), in an oblong and north-south oriented barren area located SSW of El-Qasr (25°39 N, 28°52' E) (fig. 3). Archaeology has proven that the urban site was occupied between the Old Kingdom (2700 cal. BC) and the end of the 4th century AD. The present climate is characterised by hyperaridity with average rainfall of less than 5 mm per year (Egyptian Meteorological Authority, 1996). The oasis landscape is a slowly changing patchwork formed by belts of green cultivated land and north-south oriented dunes. Some of the present barren areas, such as Trimithis, have barchans and smaller dunes, shifting at an average velocity of about 7 meters per year (observed by Fabrizio Pavia through topographic surveys during the last 10 years in Amheida), thus encroaching upon irrigated fields, roads and villages. The landscape is therefore highly variable according to the movements of the dunes and water availability. It is necessary to have these dynamics in mind when studying past landscapes and any attempt to infer past conditions from present landscapes requires great caution.
Fig. 1 – Map of Egypt.
Fig. 1 – Carte de l’Egypte.
(by Bruno Bazzani, based on SRTM GDEM data by NASA/JPL/NIMA)
Fig. 2 – Map of the Dakhla Depression
Fig. 2 – Carte de la dépression de Dakhla.
(by Bruno Bazzani, satellite image: Landsat 7, 2001).
Fig. 3 – Map of Amheida
Fig. 3 – Carte d’Amheida.
(by Bruno Bazzani, satellite image: Ikonos by Digitalglobe, 2001).
3The present study deals mainly with the hypothesis that irrigation soils existed during the Antiquity around the Trimithis urban site. This study is geoarchaeologically oriented and was carried out in close collaboration with Amheida’s archaeological team. The objective of the present work originated with a geological survey undertaken by one of the authors and reoriented towards geoarchaeology in 2011. Field surveys revealed a series of north-south oriented and 2 to 3 metres high yardangs (elongated buttes sculpted by deflation in soft material), lying over low areas of Quseir (Mût) red shale extending south of Trimithis (fig. 4). These yardangs could be the eroded remnants of a former flat Holocene playa composed of interconnecting basins, over which some low hills protruded. However, the high artefact content within the yardangs around the archaeological site indicates that they could also have been composed of irrigation soils and aeolian sand, and that their almost complete destruction occurred after a phase of historical occupation, as proposed by Brookes (1993). The objective of this study is to document the complex history of the yardangs, as exemplified by the ongoing el-Deir study in the nearby Kharga Oasis (Tallet et al., 2011; Tallet et al., 2013; Bravard et al., 2016). This objective includes dating sediment layers through ceramics and radiocarbon samples, as well as understanding their construction and destruction. The datation with the ceramics is another way to precise chronology, specially when this artefact is in a sufficient quantity to confirm the time of the occupation in the irrigation structure. It is of course to date by the different criterias according to the ceramologist’s classification. At this stage of the survey, the purpose was not to study the geochemistry and the micromorphology of the soils. These analyses are planned in the future to the condition that their implementation becomes possible.
Fig. 4 – Map of the NYU excavations at Amheida and geomorphic features described in this paper around the city of Trimithis
Fig. 4 – Carte des fouilles archéologiques menées par la NYU à Amheida et éléments géomorphologiques décrits dans cet article autour de la cité de Trimithis.
(plan by Bruno Bazzani; satellite image: Ikonos by Digitalglobe, 2001).
4The geology of the Dakhla depression is composed of gently (~0.5°) northward-dipping Tertiary and Cretaceous sedimentary units with synclinal and anticlinal NE-SW trending folds and minor faulting (Hermina, 1990; Brookes, 1993). To the north of the Depression, the Eocene Garra and Kurkur Fm. limestones form a resistant cap for the Libyan Plateau, where variable chert content within the limestone beds led to the development of yardangs indicating northerly winds in this region (Brookes, 2001). These resistant Thebes Group limestones protect a series of less resistant units that form the scarp along the southern edge of the Libyan Plateau, including Dakhla Fm. shales and Duwi Fm. phosphorites (fig. 5A).
5The floor of the oasis consists of Mût (Quseir) Fm. mudstone to the north and a lowermost unit of Taref Fm. sandstone, which is exposed primarily along the southern margin of the depression or where anticlinal folds and faults have led to upthrust of the Taref and subsequent weathering of overlying Mût Fm. shale (Hermina, 1990; Churcher et al., 1999; Kleindienst et al., 1999) (fig. 5B).
Fig. 5 – Geological sketch of the Dakhla Oasis.
Fig. 5 – Schéma géologique de l’Oasis de Dakhla.
A. North to south geologic cross-section of the Dakhla Oasis (by Adelsberger according to Churcher, 1999), illustrating stratigraphy exposed at the surface along the escarpment to the floor of the oasis to the south. Taref Fm. is exposed to the south, often due to localized faulting as well as thinning of the overlying Mut Formation. Oasis units are not to scale. B. Surface geology of the Dakhla Oasis according to Adelsberger and Smith (2010). Taref Fm. to the south is exposed in part due to uplift and faulting associated with anticlinal structures, creating areas of significant spring mound formation (black dots). A - A' indicates a possible cross-sectional area illustrated in part A, though a similar section can be described in any N-S line between the Plateau to the north and the Taref Fm. outcrop to the south. The area near Amheida to the NW is not rich in spring mounds compared to other areas of the oasis, but it is has structural characteristics similar to other areas of significant groundwater discharge in the past.
A. Coupe géologique Nord-Sud à travers l’Oasis de Dakhla (Adelsberger d’après Churcher, 1999), montrant les couches exposées en surface depuis l’escarpement jusqu’au plancher de l’oasis au Sud. La formation de Taref, au Sud, est souvent liée à des failles locales et à l’amincissement de la Formation de Mut qui la surmonte. Les unités de l’oasis ne sont pas à l’échelle. B. Géologie superficielle d’après Adelsberger et Smith (2010). La Formation de Taref au Sud est exposée en partie à cause du soulèvement et de failles associés à des structures anticlinales, créant des secteurs de buttes à venues artésiennes (points noirs). A-A’ figure une coupe possible décrite en 5A, mais une coupe analogue peut être décrite sur tout transect N-S entre le Plateau au nord et les affleurements de la Formation de Taref au sud. Le secteur situé au NO d’Amheida n’est pas riche en sources artésiennes par rapport à d’autres secteurs de l’oasis, mais il est localisé dans des conditions structurales semblables à celles d’autres secteurs à débit souterrain significatif dans le passé.
6These bedrock structures served as topographic controls for later Pleistocene and Holocene basinal deposition during more humid periods and associated surface water accumulation (Brookes, 1993; Churcher et al., 1999; Kleindienst et al., 1999) (fig. 5A). Within the depression proper, lacustrine and paludal sediments can be found as erosional remnants within the lower portions of the basin (Frizano, 1996; Kleindienst et al., 1999; Kieniewicz and Smith, 2009; Adelsberger and Smith, 2010), representing both surface and groundwater deposition within areas of low topography during humid periods of the Pleistocene (McKenzie, 1993; Szabo et al., 1995; Kieniewicz and Smith, 2007; Maxwell et al., 2010; Smith, 2012). Unlike the nearby Kharga Depression (Smith et al., 2004a; Smith et al., 2004b; Smith et al., 2007), Dakhla contains no evidence of plateau-margin springs and associated tufa formation. However, the carbonate-rich nature of the lacustrine deposits found within the Dakhla depression indicates periods of surface-water runoff from the Libyan Plateau, as no other source of carbonate is found within the Dakhla region (Kieniewicz and Smith, 2009).
7In contrast to the carbonate-rich Pleistocene lake deposits found in Dakhla, iron-cemented paludal deposits are the result of groundwater discharge from the iron-rich Taref Fm. sandstone, a part of the Nubian Aquifer and therefore the local aquifer for the Dakhla region (Brookes, 1993; Sultan et al., 1997). Resistant iron-cemented units form spring mounds composed of near-vent and spring-vent sediments (Adelsberger and Smith, 2010) that have thus far been dated only via archaeological inclusions from the Middle Paleolithic (Wendorf and Schild, 1980), indicating Pleistocene spring activation that was likely coeval with lake formation in the Dakhla Depression (Adelsberger and Smith, 2010).
8Just as bedrock topography provided a control for surface water accumulation in Dakhla, the location and nature of bedrock folding determined the location of spring activity. The Mût Fm. shale serves as a confining layer for the underlying Taref Fm. aquifer units, leading to limited spring activity in the northern part of the Depression where the Mût Fm. is thick and the Taref Fm. is not found close to the surface (fig. 5A). Instead, spring mounds are found primarily in the southern portion of the depression, where the Mût Fm. is thin or comes into contact with the Taref Fm. Spring mounds are also found near the anticlinal fold lines of bedrock structures in the northern part of the depression, where faulting and thinning of the overlying Mût Fm. allowed for spring development (Adelsberger and Smith, 2010). Although spring mounds are not as widespread near the site of Amheida as they are along the Tawil Anticline to the east, Amheida lies along the western limb of the Budkhulu Anticline and close to its associated faulting. The site is therefore subject to the same bedrock controls over potential groundwater discharge in its vicinity (fig. 5B). The structural setting of this portion of the depression may therefore be responsible for water availability at Amheida during periods of occupation throughout the Holocene.
9The climate of the “Humid Holocene” combined summer monsoonal rains, some storms brought by the Mediterranean Westerlies and the active functioning of spring mounds. Recent studies demonstrated the influence of early Holocene hydro-climatic changes on the human occupation of the Western Desert of Egypt (Haynes, 2001; Hoelzmann et al., 2001; Bubenzer and Riemer, 2007; Kindermann and Bubenzer, 2007; Kröpelin and Kuper, 2007). Artesian waters and associated mineral precipitations explain the deposition of lacustrine and playa deposits fed by runoff during the wet early Holocene in the plain which extends to the foothills of the northern escarpment of the Dakhla depression. In the area of Northern Africa presently covered by deserts, current arid weather conditions progressively set in after 5300 BC, making the forms and densities of human occupation dependent upon the availability of non-meteoric water. During the mid-Holocene phase of “regionalization” (5300 to 3500 BC), human groups retreated from desiccating regions into ecological niches, before the late Holocene “marginalization” phase dated 3500 to 1500 BC (Kuper and Kröpelin, 2006). The change from the African Humid Period (de Menocal, 2001) to present arid conditions has been considered an abrupt event (Alley et al., 2003), a gradual linear process (Kuper and Kröpelin, 2006) or a progressive decline which lasted 2500 years due to complex feedbacks between the changing biogeophysical climate and the non-linear response of vegetation (Kröpelin et al., 2008). In their study of the Tanezruft, Cremaschi and Zerboni (2010) agreed with this model of landscape change, the most recent Fezzan sites being related to a higher degree of wetness. In Dakhla and other desert depressions such as Kharga, the availability of groundwater through active spring mounds enabled populations to adapt and even thrive under intensifying desert conditions (Caton-Thompson and Gardner, 1932; Embabi, 2004). However, recent studies have revealed that the picture may be more complex with both minor climatic fluctuations and a diversified regional pattern. For instance, climate fluctuations have been documented in different Negev cave environments (Bar-Matthews and Ayalon, 2011) with short wet spells during the 1st century AD and ca. 500 AD within a longer phase of dry periods and droughts during the Hellenistic-Roman and Byzantine periods (Ackermann et al., 2014). Izdebski et al. (2015) consider that the Levant and northern Sinai experienced a wetter phase between 200 AD at the latest and 350 AD before a drier phase between 350 and 470 AD. In the Hadramawt (Yemen) several phases were documented at 4.0-3.9 to 2.8-2.7 ka BP, at 2.65-2.15 ka BP (concentrated rainfalls) and possibly at 2 ka BP. The pluvial periods in this area of southern Arabia have been hypothetically related to southward switches of cyclonic winter rainfalls originating from Mediterranean systems (Berger et al., 2012). Mediterranean cyclonic rainfalls could explain the 2500 cal. BC (end of the 4th dynasty) floods described by Butzer et al. (2013) in the Lost City of the Pyramid at Giza (10 floods recorded in 45 years with 200 to 300 mm rainfall events every 4 years) as well as flash floods recorded in the el-Deir oasis and the Kharga depression, between the 2nd and early 4th century AD (Bravard et al., 2016. See complete ref. at the end of the paper). However, floods described in the Red Sea Hills, in eastern Sudan, 2500 years ago, have been attributed to short-lived northward pulses of the Indian Ocean Monsoon (Mawson and Williams, 1984).
10Holocene spring activation, or possibly reactivation, is indicated by residual spring sediment containing ceramics as well as by the preservation of irrigation canals leading away from extinct spring mounds in some areas of the depression (Adelsberger, submitted). Groundwater availability outside of periods of humid climate was therefore likely in Dakhla and may have been continuous throughout the Holocene. Satellite imagery enabled Zaghloul et al. (2013) to document irrigation canals dating back to Antiquity. The broad category of Holocene deposits includes “irrigation soils” identified by Haynes (1983) and Brookes (1989a). A subcategory of Holocene deposits was defined as “anthropogenic irrigation sediment” (Brookes, 1990). Most of the rolled and drifted wind-blown particles were trapped by vegetation and local wetness (Haynes, 1983, in Brookes, 1989a). Such soils, found in the vicinity of springs or wells, may cover vast expanses of land. The present resistance of irrigation soils to deflation is low because they include blown sand trapped by former vegetation. When exposed to drier conditions (linked to the dwindling of irrigation water), irrigation soils are easily deflated and sculpted by the wind into yardangs. Irrigation soils, preserved in yardangs, constitute a precious source of information on past agricultural practices and on the resources of water and food available to oasian populations.
11During the Roman period, the settlement was named Trimithis and became a polis at the end of the 3rd century AD, but an earlier date is also possible (Bagnall et al., 2006; Davoli, 2012; Bagnall et al., 2016). The visible buildings, mainly built with mud bricks, spread over an area of about 700 x 650 m (roughly 45 ha) forming an open settlement, without defensive or perimeter walls. The city certainly flourished during the Imperial Roman period (1st-3rd century AD) and was abandoned toward the end of the 4th century AD (fig. 4). The cemeteries spread for 1.5 km east and south of the settlement and hosted different kinds of tombs from different periods.
12The geomorphological study of the Amheida area started in 2006 with the documentation of the present landscape by the Amheida project team directed by Roger Bagnall, a project launched by Columbia University in 2001 and organised with New York University from 2008 onwards. The topographic survey of the visible buildings (already started in 2001) was flanked by a survey of the main natural features, such as sand dunes, spring mounds and yardangs, and by other artificial features such as channels and areas covered by potsherds. The observation of the denuded geological strata together with the archaeological remains suggested a deep change in the ancient landscape due to deflation. In fact, some graves were found completely exposed among residual mounds in the flat plain and cut by later channels running SE of a mud-brick pyramid located in the East cemetery, close to the Roman settlement. These graves probably date to the Old Kingdom (2613 to 2181 cal. BC) and the skeletons deposited at the bottom of the burial sites are now completely exposed and badly eroded. Harsh deflation is also evident on top of the temple hill, where it has been calculated that at least 1.5 m of stratigraphy has been lost, leaving on top the Second Intermediate Period deposits. The New Kingdom (1550 to 1069 cal. BC), Late Period (747 to 332 BC) and the Roman Period are well documented by pottery, ostraka and other artefacts, including blocks from temples, but the stratified deposits have disappeared completely. Human activities are also responsible for some of these destructions, such as sebbakh gathering and the recovery of stone and wooden architectonic elements from older architecture to be reused for new buildings. Similar activity was observed at Ayn Asil, an ancient settlement east of the oasis (fig. 2), and contributed to the destruction of the upper stratigraphy together with erosional phenomena (Soukiassian et al., 1990, p. 347).
13According to the pottery and the archaeological data collected (Davoli and Kaper, 2006), the buildings visible on the surface of the site belong mainly to the 4th century AD. Pottery and objects from earlier periods, found all over the area, point to the presence of a settlement from at least the Old Kingdom onwards. The older settlement was possibly located in the centre of the archaeological area where, after centuries of stratification, a real kom or tell (hill) progressively rose over an uneven topography shaped by wind erosion and sand accumulation. At present it is a central hill on top of which the temple of Thoth was built in the 1st century AD (labelled Area 4, fig. 4).
14The wide barren area south of the city is characterized by the presence of shale hills with tombs and spring mounds, one of which is of impressive dimensions (point AM 25, fig. 4). Potsherds included in the mud layers of a channel draining a spring mound suggest a long period of activity for the spring, possibly with some artificial reactivation. Given the deflation, it is extremely difficult to say when the flat low areas were covered by soil and crops. However, a pigeon house was built in the 4th century AD at the north border of the cemetery, presumably near a farmhouse.
15The present study is a geomorphic and chronological survey performed inside and around the urban tell of Trimithis, 20 to 30 metres high, in order to provide new information on the hypothetical agricultural area surrounding the tell during its occupation. Due to the dominant N-S wind, deflation has destroyed most of the playa-like deposits which formerly covered the area as demonstrated by their 2 to 3 metres high scattered relicts. Consequently, most of the surveyed area displays barren surfaces eroded into shale bedrock with some man-made features (irrigation canals) and artefacts (sherds) dating from the Antiquity to the Arabic period, some of them being recent. Another reason limiting research is related to recent dune shifting, which masks part of the yardangs and exhumes landforms (fig. 6).
Fig. 6 – Photo of the Western Roman necropleis and eroded area south of Trimithis.
Fig. 6 – Photo de la nécropole romaine ouest et du secteur érodé au sud de Trimithis.
(Photo: J.-P. Bravard)
16The area studied in this article is not as large as the surveyed area and is located mainly south of the Amheida tell (fig. 6), an area well protected from deflation. These conditions have allowed for the preservation of a few relict yardangs displaying a limited number of natural sections (located on Figure 4). In the future, trenching and frequent surveys will make it possible to study additional sections, notably across large spring mounds, 2 to 5 metres high, which are well preserved. Other key landforms and features relevant for a geoarchaeological study (former spring mounds, wells dug into the substratum) were positioned using hand-held GPS devices and incorporated into the overall Amheida site map using a geographic information system (GIS). Post-deflation irrigation canals are too narrow to be mapped using aerial pictures and segments of canals on top of yardangs cannot be traced using this source of information. Cross-sections were described for eight yardangs and three spring mounds and wells (tab. 1, fig. 4, 7). They were measured using a Jacob staff and Abney level when necessary, or with a measuring tape for smaller sections. The elevations of sections were measured with a total station.
Fig. 7 – Stratigraphies of the main cross-sections surveyed at Amheida
Fig. 7 – Stratigraphies des principales coupes observées à Amheida.
(drawn by S. Gaillot)
17The survey originally intended to set a preliminary chronology of the yardang sedimentary sequences with the hypothesis that they could derive from the Humid Holocene and/or from more recent cultivation practices including irrigation. Potsherds were used for determining the phases of occupation, notably for the Ptolemaic and Roman periods, given that ceramic typologies provide more precise results than radiocarbon dating. A criterion for the delineation of antique irrigated areas and irrigation soils is the present concentration of ceramics inside the deposits and on the ground, selective wind erosion having destroyed the irrigation soils and the top of the altered Mût Formation. The most ancient ceramics date from the oldest period of development. In contrast, land plots presently deprived of ceramics were assumed to have no irrigated fields. In the past (before their erosion by deflation), these nearby areas were probably higher than the irrigated areas and were thus deprived of water. Areas that were developed or that remained undeveloped during Antiquity may be mapped using this criterion. However, following the suggestion by Brookes (1989a), we acknowledge that the presence of potsherds on the ground may have been “collected from older cultural waste and spread on fields, either to improve tilth and to inhibit sand accumulation by enhancing impact-induced wind transport across a rougher surface”. That is the reason why we privileged potsherds sampled in the cross-sections themselves or on top of spring mounds or yardangs.
18Micro-charcoal were sieved and sorted in refreshed cross-sections of the dunes and yardangs, wherever it was possible, and were positioned in the cross sections. Eleven micro-charcoals were sampled in 6 sites (tab. 1, 2) and sent to be radiocarbon dated by AMS (Accelerator Mass Spectrometry) at the CEDAD (Centre for Dating and Diagnostics), University of Salento, Lecce, Italy. The charcoals were not determined. Some samples could not be dated because they did not yield enough purified material to allow accurate dating.
19As presented above, the floor of the Amheida area consists of reddish Mût (Quseir) Fm shale. This substratum was eroded by aeolian and water-related processes during the Pleistocene and parts of the Holocene. This evolution resulted in a smooth, hilly topography isolating small independent or poorly interconnected, shallow basins. Present mounds of red shale are in some cases protected from deflation by iron-rich top strata. Lying upon the basin floor, a few yardangs are the relicts of a former discontinuous cover of playa-like deposits. Yardangs are sparse and scattered due to their destruction by the wind and to large shifting dunes. Given the presence of potsherds from the base to the top of these formations, it was initially speculated that the deposits might belong to the second half of the Holocene and possibly originate from irrigation practices (even if in some places, scattered undefined Neolithic potsherds may be found in older playa deposits). This is the reason why exposed cross-sections of yardangs, herein called “sequences”, were systematically explored in order to understand and date possible historical features and processes.
20Basic sedimentary units visible in the yardangs sequences include i) sandy grey layers defined as soils, ii) consolidated or loose wind-blown sand with dune facies and iii) small blocks of red shale forming artificial mounds topping the deposits (tab. 1). The distinction between these units was based on stratigraphic features, cohesiveness, texture, color and ceramic richness. The description and delineation of these units in cross-sections made possible a stratified sampling of charcoal in order to date the processes responsible for their deposition. The yardang sequences will be presented first, followed by the spring mounds sequences and overlaying units of sand and reworked shale.
Tab. 1 – Surveyed and studied landforms.
Tab. 1 – Formes décrites et étudiées.
21From bottom to top, the sequence GS001 displays the following stratigraphic units (fig. 4, 8):
22The basal complex is composed of a 1.8 metres thick layer of consolidated sand alternating with thin layers of sand and shale deposited by the wind and possibly by water, as well as soils. Shale particles (from silt to aggregates) originate from the Mût Formation, the weathered shale mounds towering over the irrigation soils having been eroded by the wind and possibly by runoff. The top of the basal complex displays gypsum and nodules of secondary carbonates probably corresponding to evaporation of irrigation waters. Three cohesive soils interstratified in these deposits, rich in iron and manganese oxides, ceramics and ashes, shells and charcoals have been respectively dated to 980 to 790 ca. BC (sample Cha 1, LTL-8364), 840 to 590 ca. BC (sample Cha 6, LTL‑8369), and 800 to 520 ca. BC (sample Cha 3, LTL‑8366) (tab. 2). They are separated by sand sheets dipping to the South. The top of the basal complex displays a discontinuous surface due to wind erosion.
Fig. 8 – Yardang GS001.
Fig. 8 – Yardang GS001.
(Photo: J.-P. Bravard)
A. Photo of a N-S cross section of yardang GS001. B. N-S cross-section of yardang GS001.
A. Photo de la stratigraphie N-S du yardang GS001. B. Coupe stratigraphique N-S du yardang GS001.
23Above these deposits, a 1.50 metres thick unit is composed of horizontally bedded sand and silts without any artefacts, inter-layered with oblique layers of sand dipping to the south and probably deposited by the wind on the lee face of a dune pushed by the northern wind (tab. 3). The summit of this unit, capping and protecting the yardang, is a massive grey soil stained by manganese and micro-particles of charcoal. Potsherds, concentrated by deflation from a former upper unit, also contribute to protecting this unit. Cha 2 (tab. 2), sampled at a depth of 20 cm below the surface, was dated from 390 to 110 ca. BC (Cha 2, LTL-8365). The chronological gap between the Late Ptolemaic and the early Roman period is due to the deflation of the top of the yardang.
Tab. 2 – List of radiocarbon dates.
Tab. 2 – Liste des dates radiocarbone.
Tab. 3 – Description of soils and dune sediment.
Tab. 3 – Description des sols et des dépôts dunaires.
24Excavated into yardang GS001, a short canal reach has been preserved (fig. 9). It may have been fed by a nearby well providing water extracted by humans, i.e. spring mound AM25. This canal has a trapezoidal shape, rims made of cohesive sediment (compact dark silt) and a high density of artefacts contrasting with sediments almost deprived of artefacts outside the structure. The lower part of the canal was dug into soils dated from 390 to 110 BC, and its upper, currently truncated part may have been dug into a younger topstratum. The sediments collected in the fill were not analysed, however, based on the ceramic encountered, the canal seems to date mostly to the 3rd century AD. They are composed of jars painted with dark purple motifs on a white colour, and fragments of kegs from the same period. The material and these decorations are similar to pottery from Phase 2 in Douch (the Kharga depression). The rims of cooking pots found are similar in shape to the chytra type, however it is difficult to precisely date them. Small non-dated shells and charcoal were also visible in the section. The topography was continuous and levelled enough for the irrigation canal to distribute water to lower adjacent areas.
Fig. 9 – Photo of a 3rd century canal on top of yardang GS001.
Fig. 9 – Photo d’un canal du IIIe siècle ap. J.-C. au sommet du yardang GS001.
(Photo : C. Caputo)
25Several short sequences located Fig 4 were surveyed and described.
26▪ In yardang AM66-67, two stratigraphic units were identified before sampling (fig. 7). Their upper part had been eroded by the wind. The lower unit is made of consolidated dune sand. Sample Cha 10 (tab. 2) allowed dating its base from 1050 to 840 cal. BC (LTL-8373).
27▪ AM69 is a 3.70 metres dune laying directly on the Mût shale. It is made of homogeneous well sorted sand. Charcoal 12 could not be dated.
28▪ Sequence AM70 comprises one stratigraphic unit composed of loose sand (fig. 7). A charcoal (Cha 13), sampled 30 cm below the top of the yardang, was dated from 180 cal. BC to 20 cal. AD (LTL-8374) (tab. 2). The top of the unit has been eroded by the wind.
29▪ The visible section of yardand 72 displays well sorted and laminated silt (fig. 7). No charcoal was found in the basal unit, whose top was eroded. The upper unit was dated from 390 to 200 cal. BC (Cha 15, LTL-8375) (tab. 2).
30Spring mounds are conspicuous features throughout the larger site. A large feature, named AM25 (fig. 4, 10), 110 metres long north to south and 75 metres wide east to west, has a central depression surrounded by rims towering 8 metres over the shale bedrock. It is open to the east and the west with layers of sand dipping to the south and criss-cross structures. The large central “hollow” was very probably dug out by humans and is potentially an area of ancient spring activity whose natural characters may have been completely altered or destroyed. The eastern part of feature AM25 includes a cross-section named sequence GS022.
Fig. 10 – Photo of spring mound/well AM65 dated Old Kingdom and dug later.
Fig. 10 – Photo de la butte artésienne et du puits GS022 datés de l’Ancien Empire et creusé en puits par la suite.
(Photo: J.-P. Bravard)
View from the south with a small yardang in the foreground and the spring mound/well in the background. Red shale on the top of the rim corresponds to Roman digging.
Vue prise depuis le sud avec un petit yardang au premier plan et une butte artésienne à l’arrière-plan. Les schistes rouges correspondent au creusement d’un puits romain.
31A crust of consolidated sand protects the rim. The present surface is rich in iron oxides forming this crust, with coarse granules of quartz, ceramics, brick fragments and shells. Granules of quartz are included in the consolidated sand and on the soil at this specific location. They were a component of the Mût shale extracted during an early phase of well digging, then concentrated on the well rim by deflation. This spring mound was active during the Antiquity (ceramic fragments cover the entire surface of the mound rim, as well as part of the base to the south). Fragments dated from the Old Kingdom (2400 to 2200 BC), similar to the 6th dynasty specimens found in Balat, were observed. One fragment belonging to a so-called “Maidum” carinated bowl was also found and could have been imported from the Memphite area (4th dynasty or 2613 to 2494 BC). There were also fine wares, storage jars, beer jars, large-diameter pot stands and conical bread moulds. Many small shells were also found on top and around the rim surrounding the spring. To date, we have not found any evidence of irrigation soils dating to the Old Kingdom around this spring mound and in the Trimithis area. If they existed, they were probably destroyed by deflation before the formation of the oldest soils in sequence GS001, dated from the 9th century BC onward. We cannot yet explain the long chronological gap existing between 2200 BC and the 9th century BC.
32GS022 (fig. 7, 11) is located on the eastern rim of spring mound AM25 and is capped by a thick cover of blocks of red shale. The shale deposit, rich in ceramics (too altered to be datable), is visibly related to the digging of the former spring mound.
Fig. 11 – Cross-section of site GS022.
Fig. 11 – Coupe stratigraphique du site GS022.
(drawn by S. Gaillot)
33▪ The lower unit (III) is composed of sand with layers of silts and particles of shale. It was not dated.
34▪ In the middle of the sequence, a 50 cm thick unit (II) is composed of layers of cross-bedded sand and silt, sloping 20° to the south. It was dated from 510 to 200 cal. BC (see Cha 5, LTL-8368) (tab. 2), 15 cm above the bottom of the unit. Charcoal 5 was probably transported by the wind.
35▪ On top of the sequence, a 20 cm thick unit (I) is a light, poorly consolidated brown soil, with blocky peds, shells and potsherds, covered with coarse granules of quartz or concretions of silica. A charcoal (see Cha 4, LTL-8367A) (tab. 2), sampled very close to the top of this unit, made it possible to date this soil from 350 to 30 cal. BC. The top of the soil includes potsherds concentrated by deflation.
36▪ The anthropogenic layer of piled blocks of red shale, consolidated sand and eroded potsherds covers the sandy sequence. Charcoal 14 sampled in the upper part of the dune could not be dated.
37On top of the tell, dunes, sand sheets and clay mounds are the substratum on which most of the Roman-period settlement of Trimithis was built. These natural deposits have been reached in several trenches of Area 2.1 in the framework of the archaeological project (fig. 4) and have been explored in depth in Street 2 (also in Area 2.1), where a 3 metres deep and 1 x 1 metre large trench reached a geological layer rich in organic material. The fossilised dunes start at an elevation of 135.97 metres asl and stop at 133.5 m asl, the layer of sand and clay mixed with organics starts at 133.5 m asl. However, the trench could not reach the Mût formation that is clearly visible on the surface in lower areas nearby. A small reddish iron-rich mound belonging to the Mût formation is located SW of Area 1, on one side of a round depression and near a series of pottery workshops. It is likely that the reddish clay was dug out from deeper layers and used for manufacturing the pottery during the late 3rd century to early 4th century AD, as suggested by the composition of the local vessels. Mounds of light-brown clay were and are also present in the area and were used to manufacture mud bricks. This type of mound is still preserved near the modern village of Bir Itnayn el-Arab and it was used to make the mud bricks of the replica house of Serenos.
38GS011 (fig. 4, 7, 12A-B) is an ancient dune belonging to the substratum of the Roman settlement. Dune sand is covered with blocks of red shale. The top of the dune has been dated between 40 cal. BC and 140 cal. AD (Cha 7, LTL-8370). It has a dense cover of small badly eroded potsherds and may have served as a dump for the kilns present in the immediate vicinity. Above the sand unit is a layer of red shale buried by non-oxidized beige shale probably excavated from a lower area in the immediate vicinity. Blocks of shale were processed by workers working with small individual tools as testified by the small size of the blocks. Shale probably originated from re-dug wells (large “hollows” may be a better term) to gain access to water again. Sand sheets were deposited inside the blocky structure by runoff (fig. 11) and micro-mudflows. The action of water is visible in the texture of deposits and cannot be attributed to post-depositional processes. This means that the farmers had reached the water table and were excavating a mixture of sediment and water. A charcoal (Cha 9), sampled between blocks of red shale, was dated from 40 cal. BC to 140 cal. AD (see LTL-8372 on table 2). The comparison between the dates obtained in the dune and in the shale cover suggests that people reacted immediately or very rapidly to the aeolian crisis described above.
Fig. 12 – Site GS011.
Fig. 12 – Site GS011.
A (Photo: J.-P. Bravard) / B (drawn by S. Gaillot)
A. Photo of drift sand covered by block red shale following the digging of a well. B. Cross section of site GS011.
A. Photo de sable éolien recouvert par des blocs de schiste rouge après creusement d’un puits. B. Coupe stratigraphique du site GS011.
39Large tracts of land south of Trimithis, lying at a low altitudinal level, were managed during the Roman period. Two Roman necropolis were built in the southern outskirts of the city of Thrimitis, about 1 to 3 metres below the level of the irrigated area dating back to the Ptolemaic period and the 3rd century AD (top of yardang GS001). Dunes made of loose sand cover an uneven topography resulting from deflation and sand accumulation. Tombs were erected on red shales and sometimes on top of low dunes after the erosion of the irrigation soils and sand deposition in a zone that we may consider as a former cultivated area (AM65) (fig. 7). The necropolis have not been excavated or surveyed yet, and therefore are simply dated from the Roman Period. Our knowledge of this area is still based on limited surface observations. The low topographic position of the Roman necropolis may be explained by intense deflation, which destroyed the fields built up since the Old Kingdom period, down to an unprecedented level, i.e. the level of the shale, except for some relict yardangs.
40A second canal was discovered on the shale floor, 3 metres below the 3rd century canal dug on top of yardang GS001. The lower canal, cemented with red clay (probably to stiffen the canal section and reduce water losses) has been dated from the 3rd century AD thanks to cemented potsherds found in its fill. Potsherds were not rounded and altered and were considered as in primary position to the difference of potsherds lying on the ground. The difference in elevation between the two canals means that a major deflation phase occurred during the 3rd century AD.
41East of yardang GS001, the ancient irrigated soils (deduced from narrow canals dug into the shale, lay directly on the shale bedrock and could also be dated from the 3rd century AD and post-3rd century AD. Potsherds of this period and from the Late Pharaonic times originating from the destruction of yardangs may be observed in high densities on its surface. They probably originate from the destruction of former soils of different ages. It is quite difficult to date the polygenic topography of this low area, considering that deflation has been active since the Roman period at least (fig. 6). Moreover, site GS016 (south of the site and not represented in Figure 4) displays a set of interesting features. A well was dug into the red shale, probably into a former spring mound. Open to the north, this well delivered water to a 2 metres wide irrigation canal. The well and the still undated canal are located below the level of the pre-Roman irrigation soils, which confirms that deflation had eroded the fields before the digging of the well. The red shale displays features of erosion by the wind and many potsherds are scattered over the surface. Finally, the southernmost large spring mound was dug below a terrace with ancient Pleistocene spring mounds characterised by iron concretions. Undated irrigation canals delivered water from the spring mound. Consolidated silt deposited on the canal sides helped to preserve them.
42The type of soil observed in Thrimithis differs from classical soils as defined in pedology, such as the Neolithic dark soils studied in Mar’ib Oasis (Yemen) dating back to the wet early Holocene from 9000 to 6000 BC (Pietsch et al., 2010). In the Western Desert of Egypt, “irrigation soil” or “anthropogenic irrigation sediment” refers to ancient practices and to a geoarchaeological definition of layers rich in ceramics, charcoal and other artefacts (Haynes, 1983; Brookes, 1989a; Brookes, 1990). The presence of features related to organic activity is not relevant in this context due to dry conditions. Moreover this type of soil frequently includes layers of microlaminated sand of aeolian origin without any artefacts. They correspond to periods of deflation and sand deposition in the irrigated area, their origin being related to the deposition of wind-blown particles and to silt-rich sediment transferred by water over the irrigated areas. Recent soils in Trimithis may be equivalent to the aeolian sand and silts blown from dune fields in areas where aridity is responsible for the high degree of sediment instability (Coque-Delhuille and Gentelle, 1998; Pietsch et al., 2010).
43A limited number of exploitable sequences document the construction of historical sediments in Trimithis. Figure 7 presents a graph of the most important dated sites. The most relevant result is the relative importance of dune sand in the cross-sections compared to the thinness of “soils” (sandy matrix incorporating potsherds, ash, shells, charcoal) in horizontal or inclined positions over cross-bedded or sloping sand sheets (lee faces of dunes). Soils and dunes are locally truncated by episodes of deflation. On the basis of limited observations, “soils” probably correspond to areas frequented by humans and to irrigated fields located on an uneven topography shaped by the wind. The sandy matrix of the soils and of the underlying dunes point to their extreme porosity and probably to extremely high consumption of irrigation water. However, it is possible that dust deposition improved fertility of irrigated soils, enriching them in calcium and mineral-derived nutrients blown from the northern plateau (Nakase et al., 2014).
44The period of dune/soil construction was attributed to two main periods spanning from 1050 to 840 BC (dune, AM66-67) and from 350 to 30 BC (upper soil, GS022). The fact that an irrigation canal dated 3rd century AD was dug into the Ptolemaic soil on top of yardang GS001 suggests that irrigation was practiced on top of present yardangs during the Roman period, but we must acknowledge that no clues of earlier irrigation have been found at this stage. We also have no evidence of any soils and fields over yardangs dating back to the early Roman period, but they may have been destroyed by deflation.
45Three ancient soils were dated to the Third Intermediate period 1069 to 664 BC and to the beginning of the Late Period 664 to 332 BC (tab. 2):
46- 2710 ± 55 yrs BP, or 980 to 790 cal. BC (Cha 1, GS001)
47- 2599 ± 36 yrs BP, or 840 to 590 cal. BC (Cha 6, GS001)
48- 2527 ± 40 yrs BP, or 800 to 520 cal. BC (Cha 3, GS001).
49These soils were observed in yardang GS001 and belong to most ancient documented soils.
50A phase of dune formation was dated from 2315 ± 42 yrs BP, or 510 to 200 cal. BC (Cha 5, GS022), and from 2220 ± 33 yrs BP or 390 to 200 cal. BC (Cha 15, AM72). This phase roughly corresponds to the Persian and early Ptolemaic periods. No soil dating back to this period has been found so far.
51The most recent phase of soil formation was dated by three charcoal samples from:
52- 2186 ± 40 yrs BP or 390 to 110 cal. BC (Cha 2, GS001)
53- 2101 ± 35 yrs BP or 350 to 30 cal. BC (Cha 4, GS022)
54- 2062 ± 35 yrs BP or 180 cal. BC to 20 cal. AD (Cha 13, AM70).
55This last phase is Ptolemaic in age and may extend to the Early Roman period. The end of this phase may be close to the most recent dates obtained in sites where dune sand was deposited, such as 40-140 cal. AD (GS011). This unit is the richest in ceramics and has a loose sandy texture. Shells of molluscs may indicate the presence of water. Thin layers of deflated red shale particles may be observed in some places.
56This sequence of soil construction was interrupted by phases of aeolian erosion and sand drifting. It must be pointed out that the sequences described above represent a net budget of deposition and destruction phases, explaining a maximum relative elevation of ca. 2 to 3 metres above the uneven bedrock topography. The relative importance of units may vary according to the site topography, to the local importance of erosion by the wind and sediment deposition, and probably to water availability.
57The field study showed how difficult it was to maintain a rural and probably irrigated area from the Late Intermediate to the Ptolemaic period due to the harsh constraints of deflation and drifting sand which shaped an ever-changing landscape. These constraints paid a significant role in the following period. Four sequences named AM65-AM69-AM70-AM71 and GS011 (fig. 4, 7) served to describe non cohesive sand deposits with the structure of mobile dunes; however only AM70 and AM75 were dated. However, it is impossible to reconstruct the landscape during this phase of dune shifting. The sand units observed are made of loose yellow sheets of pure sand of variable thickness. Dunes do not have any artefacts inside the sand body, which suggests short aeolian episodes or the preceding abandonment of the areas. Potsherds present on top of the dune units may be related to a short phase of frequentation by local people or to agricultural practices. The topography was probably uneven because the dune landscape was associated with the process of deflation.
58Most of the ancient dunes have been destroyed by deflation, but some of them have been preserved thanks to a cover resistant to deflation. On site AM65 (a 2 metres high yardang), the top of the sand unit is an archaeological layer (stones, potsherds) protected from erosion by the fallen wall of a ruined mud-brick building (a tomb of the necropoleis?). Elsewhere, protection is ensured by cohesive red-shale blocks dug into the shale substrate. Site AM71 and dunes GS011 and GS022 are also topped by blocks of shale, which have resisted deflation since the Antiquity. We suggest that these covers of blocks are located in the immediate vicinity of the wells dug into less active spring mounds or close to dug-out spring mounds buried by drifting dunes. These features look similar to the refitted spring mounds described by Bousquet (1996), which were widened and surrounded by 2 metres high artificial mounds.
59Spring mounds have discharged water from the Nubian aquifer for thousands of years and Ball (1927) demonstrated that underground water comes from Sudan where infiltration of monsoon rains until 6000 BC created a strong head. The level of the water table was often close to the surface of the surrounding ground and the flow was active through spring mounds. Such low places may have been the source of water for irrigation, watering spots for humans and cattle. Later, it was shown that, at the local scale, free groundwater discharge created depressions in the piezometric surface of distinct, locally faulted, and folded aquifers (Lamoreaux et al., 1985). Relative depletion of water by artesian free discharge resulted in the human reactivation of springs by removing sliding sediment and sometimes digging wells flanked by log casing, thus further lowering the piezometric surface. In Amheida, the presence of several areas with shale caps proves that the digging was a rather common process, but we could date only one sequence of dune and shale cover (40 cal. BC-140 cal. AD). The level of water inside the springs subsequently decreased until their decay (fig. 11-13).
60Sudden and severe crises of wind erosion followed by phases of deposition destroyed part of the irrigated land to such an extent that farmers and city dwellers had probably to recover access to water by digging into the decaying spring mounds. Further work must be done to understand the nature of this crisis or of a series of crises throughout the area.
61Some elements made it possible to describe landscape changes south of Trimithis, in the rural outskirts of the city. Evidence is provided by the relative elevation of irrigation features and of buildings foundations.
62Roman Trimithis was built on an uneven barren area with consolidated dunes, mounds of hard shale and flat lowlands, where most of the wells supplying water were probably located. The head of one such well is visible at the elevation of 137.5 m asl, south-east of Serenos’ house (see Area 2.1 in Figure 4). The area was already inhabited at least from the Old Kingdom onwards, but at present it is not possible to know if it was only one settlement, how big it was and if it was continuously inhabited. Data collected by the D.O.P. survey and our observation at Trimithis would suggest the presence of different, probably small settlements and workshops located north and west of Trimithis during the Old Kingdom. The erosion and the accumulation of sand prevent us from a clear view of these older phases. The Roman-period settlement was thus built preferentially according to an orthogonal orientation of streets and alleys, but also following the natural layout of the land. Even though the maximum extension of the Imperial-period city is not clear yet, as far as could be ascertained in the excavated areas, some houses from the 3rd century AD continued to be used in the 4th century AD, while other buildings, e.g. a huge public bath, were demolished and replaced. Stratigraphies found in all the excavated areas testify to the absence of sand deposits between the different building phases in the 3rd and 4th centuries AD and this probably means that the city was surrounded by fields, similar to today or even more extensively.
63Probably at the end of the 4th century AD and after its abandonment, Trimithis was invaded by a field of sand dunes running from the north to the south, similar to present dune fields visible in satellite imagery. The city was covered to an elevation of 138 to 139 m asl. Most of the features standing above that elevation, both natural and man-made, suffered massive erosion. Mud-brick walls were eroded to the foundation and, in some places, for instance on the temple hill, even lower. Only the very strong clay hills east and west of the site maintained a considerable height. The accumulation of pottery on the surface of the hills is all that is left of the houses once standing on them. Most of what was lying below 138 to 139 m asl has been preserved due to the massive silting of aeolian sand that can still be observed today. Sand dunes west and south-west of the site, all above 139 m asl, still cover a substantial portion of the inhabited area. Taking them into account together with the hills covered with sherds and areas in between provides an idea of the overall extension of the urban area of the Roman period that is twice as large as what is visible today on the surface.
64We still do not know if the settlement was occupied continuously from the Old Kingdom to the end of the 4th century AD. It is a reasonable possibility, however we cannot determine the degree to which it may have flourished in the various periods. The reasons in favour of a permanent settlement in this location can be found in the geography of the oasis and in the presence of good supplies of water enabling agriculture to thrive. However, after being abandoned toward the end of the 4th century AD, the site was not rebuilt. The constraints that caused the abandonment of this area and of several other settlements in the oasis must still be clarified and they could be a combination of several natural and historical factors. Other investigations are needed before coming to a comprehensive conclusion.
651) No early Holocene deposits corresponding to a playa shaped during the wet Holocene were found in the vicinity of Trimithis. Such deposits probably existed but we assume that they were destroyed by wind erosion during the arid period, which followed the wet Holocene.
662) The present yardangs are the relicts of a former, unstable complex of barren land, dunes and irrigation soils built up during the phases of irrigated agriculture in an area prone to deflation. Active spring mounds, probably reactivated as wells by farmers and city dwellers during the late occupation of the site, delivered water through canals. In line with previous studies, it is thought that the moisture created by irrigation practices and the roughness of the vegetation cover may have triggered the construction of irrigated areas through the trapping of blown sand and the reworking of sediments by farmers (Brookes, 1989a-b; 1993). However drifting sand buried soils and changed the microtopography of such places.
673) In Trimithis, the relicts of irrigation soils and active canals linked to yardangs date to the Late Pharaonic, Persian and Ptolemaic periods (from at least 840-740 BC to the 3rd century AD). Due to the lack of significant exposure, their detailed history will be difficult to further document. Laboratory studies, such as micromorphology of soils, chemical anlyses, archaeobotany, must be implemented in the future for understanding the composition and genesis of soils, and the type of agriculture practiced at different periods.
684) In the surveyed cross-sections, deposits of wind-blown sand are greater than expected when compared to El-Deir, in the Kharga oasis (Bravard et al., 2016). Irrigation phases were separated by periods of deflation responsible for the shaping of dunes and for discontinuities truncating the depositional sequences. The cumulative 2 to 3 metres thick complex of dune/irrigation soils present at the end of the Ptolemaic period was the result of a net sedimentary budget including positive and negative events (aggradational and erosional phases). The sediment budget benefited from spring mounds able to deliver enough water to help trap blown sand and limit erosion.
695) Several distinct areas of irrigation soils may have existed simultaneously around the Trimithis kom, the irrigation soils being the product of distinct spring mounds flowing from different altitudes over an uneven topography. We suggest a series of semi-independent areas dependent on water availability.
70Another phase of strong aeolian activity of indeterminate duration occurred between 30 cal. BC and 140 cal. AD. The Ptolemaic cultivated topsoil was scalped and overlain by sand dunes. These processes resulted in net erosion because most of the area was lowered. Vertical erosion was followed by the development of the Roman necropolis standing below the Ptolemaic level of the soils studied in the SW section of the area, as testified by a 3rd century irrigation canal active on top of yardang GS001. Another severe phase of deflation was dated to the late 3rd century AD, as testified by the construction of a low canal on the Mût Fm shale. These phases of deflation may have been similar in function to previous phases which occurred between the Pharaonic and Ptolemaic phases, however increased activity of northern winds pushing sand dunes to the south may also have affected the site.
71The progressive depletion of artesian pressure probably reduced the amount of water available from spring mounds and wells and consequently the wetness of soils and the capacity of irrigation soils to aggrade further and even to stabilize. In other words, a reduction in available water could compromise the ability to form new soil as fast as old soil was lost to wind.
72In other words, the fragility of soils may have increased until crossing the threshold of minimum wetness, exposing them to severe deflation. As stated by Brookes (1989a), the change from spring-mound artesianism to wells and the use of saqyias lifting devices drawdown and lowered water tables surrounding the wells. These permanent cones reduced the available discharge, the wetness of irrigated plinths, and paved the way for uncompensated deflation. The increased sensitivity of drying irrigation soils to drought, wind and deflation processes may have been the cause of the onset of a negative sediment budget. During the 3rd century AD, wells still provided water to canals slightly higher than the present ground. After its abandonment at the end of the 4th century AD, the city of Trimithis itself was covered by sand and the agricultural area alternated burial under the sand and deflation down to the exhumed Mût Formation.
73Human settlements in association with irrigated land are places where interdisciplinary studies may be conducted and, combining the approaches, provide new perspectives. The preliminary hypothesis is that each oasis provides a specific set of natural conditions for irrigation (size, water table depth, richness, and techniques of water control and duration of water availability in aquifers). Each oasis may experience different external constraints (deflation, runoff) and demonstrate different abilities to recover depending on the resilience of the societies living in each environment. Types of aquifers and recharge models may differ between places due to complex structural controls and wind action, explaining distinct histories or functioning.
74On the eastern side of the Kharga depression, the ancient town of Kysis was active from the 3rd to the 5th century AD, developing an oasian-type agriculture based on wells and qanâts. In the so-called “Douch model”, the early depletion of artesian wells during the Roman period was compensated by digging qanâts to collect water from water tables perched in structural landforms (Bousquet and Reddé, 1994; Bousquet, 1996; Bousquet and Robin, 1999). Kysis was able to overcome a climatic degradation between the 3rd and 4th centuries AD, that is assumed to have occurred in light of the redevelopment of qanâts. Kysis was then abandoned during the 5th century AD due to its incapacity to resist the encroaching sand (Bousquet, 1996).
75North of Kysis, the el-Deir oasis experienced another set of natural constraints including deflation and episodic flash floods. The depletion of the spring mounds which once ensured the prosperity of this oasis was compensated by digging wells (qanâts not being possible due to specific geological characteristics). Destructions were severe during the Late Ptolemaic period (deflation) and during the 3rd and 4th centuries AD (flash floods). Human activity was maintained, despite the abandonment of agriculture during the 5th and 6th centuries AD, through the development of trade between Hibis and the Nile valley (Bravard et al., 2016).
76In the Dakhla oasis, Trimithis could be a case of progressive decay, with society facing the constraints of the northern wind (deflation, drifting sand) in a context of declining water availability (shift from active artesian springs to wells). The change may have been progressive and a threshold may have been crossed during the early Roman period due to increased water scarcity. Digging of wells may have lasted over several centuries and reinforced the resilience of the city to droughts before the final decline.
77A preliminary conclusion is the recognition of the relative importance of local conditions in terms of water availability, drifting sand and deflation, on the one hand, and of human action on the other hand during the Antiquity, whatever the political, economic and social conditions may have been. Natural external constraints related to drifting sand and deflation were of prime importance in conditions of hyperaridity. They were probably key factors in the decreasing economy despite their economic prosperity. For the historians of Antiquity, the question is whether natural constraints were dominant at the depression scale or just at the site scale, enabling the economic system to resist thanks to more favourable zones. In any case, they greatly affected the profitability of agriculture and thus the resilience of the system in the face of environmental variability.