General framework for the recent Holocene hydrological changes of the Nile River in Upper Egypt linked with the human occupation

1With a length of ~6500 km, the Nile is the world’s longest river, extending across northeast Africa through 35° of latitude to the Mediterranean coast of Egypt (Woodward et al., 2007). Charting the lateral migration of the Nile River in Upper Egypt (area located 700 km south of Cairo; Fig. 1) in relation to human occupation is a recent multidisciplinary topic although few studies adopting a geoarchaeological approach have been undertaken in the ancient Theban area (Traunecker, 1970; Hillier et al., 2007). Recently, palaeoenvironmental researches have been intensively developed through the strong partnership between CNRS (CEREGE and CFEETK laboratories) and the Supreme Council of Antiquities (SCA): pioneering and major works have been published and aimed at reconstructing the historical mobility of the Nile River (Ghilardi and Boraik, 2011; Ghilardi et al., 2012). Particular attention has been paid to the Theban valley where important archaeological sites including Coptos, Dendera, Karnak (registered in the UNESCO heritage since 1979), Luxor, Medamud and Tod temples and the Valley of the Kings (Fig. 1) are still visible. The geoarchaeological approach was adopted since there is a lack of studies combining landscape evolution and human occupation in Upper Egypt during the mid to recent Holocene (i.e., the last 6000 years). This time period corresponds to the pre-Dynastic (5000-3125 BC), Dynastic (3125-332 BC), Ptolemaic-Roman (AD 332-395), Byzantine (5^th and 6^th c. AD), Copt and Islamic (from AD 642) phases. At Karnak, the world’s largest temples complex built from ca. 2000 BC to Roman times, recent work combining a study of the archaeological material and the interpretation of digital data highlighted the important lateral mobility of the Nile River during the Dynastic period (ca. 3000 BC) until AD 300-400 (Hillier et al., 2007). However, these studies did not use chronostratigraphy to precisely reconstruct the palaeogeographic evolution of the Nile River around Ancient Karnak during the Dynastic period. The aims of the present study are to reconstruct the evolution of the Nile River and its ephemeral tributaries (wadi) and to closely examine the recent Holocene sediments deposited by this river in the vicinity of Karnak. The sedimentological study of a deep borehole (25 m deep) combined with a geoelectrical survey (Fig. 2) helped to reveal the presence of sandy levees (called gezirah, “island” or “peninsula” in Arabic) on which the first settlers built the first Karnak temples. Using chronostratigraphy, our work aims to better understand the period when the Nile River flowed in front of the Karnak temples, along the quay excavated by the SCA from 2007 to 2010 (Boraik, 2010).

Geological and geomorphological settings

2The Modern Nile catchment has a total area of ~2.9 M km² (Shahin, 2005), covering ten countries before ending in the Mediterranean, but almost all of this area is located in the upper reaches of the river. The source of the principal Nile tributary, the White Nile, is in the Great Lakes region of equatorial East Africa. It is joined by two other tributaries, the Blue Nile at Khartoum in central Sudan and the Atbara at Atbara in northern Sudan (Woodward et al., 2007), both of which drain the Ethiopian highlands (Zaki, 2007). Further north, in Egypt, where the Nile River lows through the Sahara Desert, it has no signiicant tributaries. Between Assuan and Cairo, tributary wadi typically reach no further than a few 10s of km from the main Nile channel (Zaki, 2007) and some of these intermittent streams join the Nile in the heban valley. For instance, wadis Hammamat and El Medamud have formed large detrital fans on the right bank of the Nile River near Coptos and South of Medamud (Fig. 1; Ghilardi et al., 2012). These Pleistocene fans are partly buried under the Holocene Nile deposits and most of the time only the apex is visible. From a geological point of view, Middle and Upper Egypt consists of an extended sedimentary plateau of Cretaceous to Eocene age (Fig. 2A), characterised by low relief topography that dips towards the west (Said, 1990; Badawy et al., 2006) and the predominance of carbonate-rich rocks such as Eocene chalky limestone and Pliocene sandstones/siltstones (Fig. 2). The plateau reaches a maximum elevation of 500 m asl with a sharp scarp facing the Nile valley on its eastern side (Badawy et al., 2006). Carved into the African plateau ca. 5-8 Ma ago and then mostly re-illed with sediment, the ~ 10 km wide Nile valley is clif-bounded (300 m high) and flat-bottomed in most of Upper Egypt (Butzer, 1980; Hillier et al., 2007). Said (1981) reveals that the Nile valley lies along a seismically active belt. The evolution of the modern drainage network and its luvial geomorphology relect both long-term tectonic and volcanic processes and associated changes in erosion and sedimentation, in addition to sea-level changes. One of the geological features of the drainage basin of the Nile River is the predominance of volcanic rocks containing an important proportion of magnetic minerals in its upper part (plateau of Ethiopia). In contrast, Upper Egypt is characterised by a local drainage (wadi streams) on carbonate-rich rocks (low magnetic minerals content). During the Pleistocene, the mineral composition changed drastically during the Upper Palaeolithic when luvial capture forced the rivers of Ethiopia to branch onto the Nile River (Hamrouch and Stanley, 1990; Williams, 2009).

3As a result, pyroxenes are lacking within the pre-Palaeolithic deposits and this could be a useful indicator for chronostratigraphy purposes (Ismaël and Delaune-Mayère, 1987). Geoelectric methods based on seismic techniques have helped to reveal the 3D geometry of the unconsolidated sediments deposited within the Nile Valley (Ismail, 2003; Ismail et al., 2005), from the modern Nile River course to the Eastern plateau (approximately 5 km long; Fig. 2B, profile BB’). Resistivity values are expressed in Ω. m and highlight a clear contrast between the bedrock (mainly composed of limestone; Fig. 2B) and unconsolidated gravel, sand and ine sediment. A major feature is that the recent Nile sediments deposited west of the Karnak temples reach a depth of + 53 m asl (almost 20 m thick). By contrast, towards the east of the Karnak temples, the thickness of clay/silt deposits (which represent a low resistivity) is less than 6 m. Therefore, an important asymmetry characterises Nile sediment deposition around the Karnak temples.

Archaeological context

4The hebes area was an important religious and economic centre during part of the Dynastic period and several sites such as Karnak, Luxor, Dendera, Medamud, Tod and Coptos played a key role in the emergence of a powerful state from ca. 2000 BC until the end of Roman times. Construction of the Karnak temples complex started ca. 2000 BC, during the 11^th Dynasty (Charloux et al., 2004; Charloux, 2005, 2007), and was completed during Roman times (Legrain, 1903; Barguet, 2006). It comprises a main temple dedicated to Amun Ra and a vast complex of divine temples dedicated to different gods from Egyptian mythology, chapels, pylons, and other buildings. During the Dynastic period, the spatial distribution follows an E-W direction. Another significant edifice is the quay built in front of the first pylon, at a distance of 90 m to the west of the first pylon (Boraik, 2010; Ghilardi et al., 2012). Recent excavations by the Egyptian SCA revealed the presence of this large 1 000 m long (the exact length is still unknown because excavations are ongoing around Luxor temple, situated 3 km to the south) and approximately 6-8 m thick structure (Boraik, 2010). Based on Egyptological and archaeological studies, the quay is hypothesised to date from the first quarter of the 10^th c. BC, during the 22^nd Dynasty (Lauffray, 1975). An important archaeological find is the ramp of the Pharaoh Taharqa (25^th Dynasty, 690-664 BC) superimposed on the quay and potentially dating from the 25^th Dynasty. The use and the environmental contexts of the construction of these large buildings is still unclear: Was the quay constructed to protect the Karnak temples complex from violent flood events or was the Taharqa ramp intended to facilitate direct access to the river? The presence of the Nile River in front of the first pylon has never been proven and our paper investigates its position in the western part of the Karnak temples complex.

Material and methods

5A 25 m depth well (DW; Figs. 2B and 3B) in front of the First Pylon, was studied and accurately levelled with a Total station. All sediments were analysed at the American Research Centre in Egypt laboratory (ARCE, Karnak) situated inside the Karnak temples complex.

Grain-size analyses and quartz microscopy

6Grain-size analyses of 150 bulk samples were performed using standard sieving and sedimentation techniques (Folk, 1974). The sediments collected were hand sieved with different mesh sizes from 20 µm to 2000 µm. Ultrasound was employed for the finest particles. The mean grain size (in µm) was calculated. These parameters were plotted against each other to define depositional environments. The medium-sand fractions were separated and examined using a polarising microscope (PM) at the ARCE laboratory in Karnak. Surface features of the quartz sand grains provide information about transportation processes and depositional history (Krinsley and Doornkamp, 1973; Whalley and Krinsley, 1974; Higgs, 1979; Williams and Morgan, 1993).

Magnetic susceptibility measurements

7Measurements were performed using a resolution of 10^-5 SI units at the American Research Centre in Egypt laboratory. The sediment cores were sampled at a ~ 5 cm interval, except at levels including reworked material, yielding 150 samples in total. These samples were placed in 10 cm³ plastic boxes, dried and weighed. In addition to the low-field magnetic susceptibility, usually measured at the 465 Hz frequency, measurements were also taken at the 4650 Hz frequency. The magnetic susceptibility values were divided by the density of the dried samples in order to derive specific susceptibilities (χ). Magnetic susceptibility is used as an indicator of the concentration of magnetic particles and can help to identify different sediment source areas in fluvial deposits (Ghilardi et al., 2008) characterised by magnetic minerals concentration. For our study, there is a possibility that Nile sediments deposited at Karnak and Coptos show a different signal from the local ephemeral streams draining the Theban plateau. The size of the ferromagnetic particles can also influence magnetic susceptibility values. Magnetic susceptibility measurements performed at two frequencies are used to detect the ultrafine (<0.03 µm) superparamagnetic particles, which are produced by bacteria or chemical processes during soil formation (Dearing et al., 1996). The contribution of finegrained viscous/superparamagnetic particles is given by the frequency dependant susceptibility (χfd).

Radiocarbon dating (conventional method)

8For the Karnak temples complex, radiocarbon dating (conventional method) was performed at the IFAO laboratory (with permission from the SCA) on material collected from stratigraphic profiles (Ghilardi et al., 2012) and the deep well (Figs. 2B and 4). Sampling was accurately levelled and elevations range from 56.09 to 72.14 m asl.¹⁴C ages were subsequently calibrated using OxCal version 4.1 (Bronk Ramsey, 1995, 2001; Reimer et al., 2009).

Rapid Nile River sediment accumulation from ca. 1600 BC to AD 350 to the west of the Karnak temples complex

9The deep well (DW) reveals the longest sedimentary sequence ever-observed in Upper Egypt (Figs. 2B and 4). It records different depositional environments and can be described as follows.

From about + 48 m to + 50 m asl

10Unit A (Fig. 4) is composed of angular and wellrounded pebbles, with some coarse sands and gravels. The stones show different sizes and some of them are 20 cm long with sharp edges that indicate ephemeral fluvial transport. In contrast, flat pebbles of 5-10 cm in diameter were present. An autochthonous origin must be envisaged and the local wadi streams are the only possible explanation for this deposit. No human artefacts or organic material were identified, suggesting that the deposit is older than Holocene, probably Pleistocene. OSL dating is a possible technique for dating this sedimentary unit.

From about + 50 m up to + 52 m asl

11Unit B is a mixture composed of coarse yellow to white sands and well-rounded pebbles (mainly limestone fragments). Magnetic susceptibility of the coarse sands show low values, between 21× 10^-8 m³ /kg and 38× 10^-8 m³ /kg (χfd), which is less than 1% and confirms the detrital origin of the signal. Grain-size distribution reveals that the mean grain size is between 400 µm to 500 µm and shows a multimodal distribution, consistent with a poorly sorted sample. A wadi stream transport together with an aeolian origin is suggested by surface features of quartz grains, observed under a polarising microscope (Fig. 4A).

From about + 52 m up to + 55 m asl

12Unit C comprises yellow to light grey medium to coarse sands. The magnetic susceptibility measurements reveal increasing values from the lower part (38× 10^-8 m³ /kg) to the upper part of the sequence (130× 10^-8 m³ /kg) with important values oscillating around 80× 10^-8 m³ /kg (χ fd), which is less than 1% and indicates a low contribution of the superparamagnetic minerals. The grain-size distribution indicates a unimodal distribution at 300-350 µ m. The depositional environment is consistent with a high-energy fluvial system where local wadi sediments (coarse material identified from about + 50 m up to + 52 m asl) were reworked and transported away by important flood events, probably from the Nile River. Indeed, the increasing signal of χ can be explained by a higher contribution of magnetic minerals from the Nile catchment (e.g., the volcanic plateau of Ethiopia). Recent research on the Nile delta (Tristant, 2004, in press) shows similar deposition where pre-Dynastic settlements have been built upon. Unit C can be interpreted as a sandy levee (i.e., “gezirah”) composed of aeolian sands episodically reworked during Nile floods.

From about + 55 m up to + 73 m asl (surface)

13Unit D is composed of dark grey medium sands. Human artefacts were recovered including intact pottery, pottery sherds, red bricks, etc. At about + 56 m asl, medium grey sands (mean grain size between 175 µ m and 275 µ m) showing high magnetic susceptibility values (χ), between 160× 10^-8 m³ /kg and 260× 10^-8 m³/kg, are mixed together with large pottery sherds and wood fragments (Acacia sp.; Fig. 4B). This unit is composed of well-sorted fluvial sediments (ranging from clay to medium sands) deposited by the Nile River. The high magnetic susceptibility signals (up to 470 × 10^-8 m³/kg) reflect detrital deposition since the χ fd is less than 2%. The geology of the Nile catchment is characterised by volcanic rocks in its most upper reaches (Woodward et al., 2007) and their erosion, under tropical to equatorial climate controls. Radiocarbon dating performed on one of these wood fragments reveals an age of 1494-1402 BC (3158± 47 BP). This wooden artefact can be attributed to the early New Kingdom/18^th Dynasty (1550-1292 BC), under the reign of Thutmose III (1479-1424 BC). Unfortunately, it was not possible to identify the function of this piece of wood. The fluvial sequence of the Nile River sediments (thickness is ~ 18 m) and its recent age of deposition (~3500 BP) shows a higher magnetic susceptibility signal and a finer grain-size distribution (from 10 µm to 180 µm) compared to the gezirah and the wadi deposits described above.

14The upper part of the sequence, close to the surface (+73 m asl) is affected by intense human activities and mud bricks are clearly identified by the recent archaeological excavations. The date of these remains, located between the tribune and Taharqa ramp (Ghilardi and Boraik, 2011; Ghilardi et al., 2012) can be reasonably placed at the end of the Roman period (around + 72 m asl), based on the study of material (presence of pottery found in situ and walls built with mud bricks; Fig. 4) and the radiocarbon dating results: 1736± 110 BP (AD 208-415).

15To summarise, we can assume that Nile River sediments (~18 m thick) overlie gezirah and wadi deposits. This sedimentary sequence can be related to the second geoelectric unit identified by Ismail et al. (2005; Fig. 2B). The period of accumulation for the fluvial material between the ramp of Taharqa and the tribune is ~ 3500 BP and can be placed from the early times of the 18^th Dynasty (mid-15^th c. BC) to the end of the Roman period (ca. 4^th-5^th c. AD) since early phases of construction of Karnak temples complex are dated to the 11^th Dynasty (ca. 2000 BC). Given this data the obvious question arose concerning the position of the Nile River from the 11^th to the 18^th Dynasty, however, the lack of palaeoenvironmental data for the most eastern part of the Amun Ra temple does not allow an accurate fluvial reconstruction. Some authors (e.g., Larché, 2007) assume that during the Middle Kingdom the main entrance of the temples was situated to the east (instead of the present position) and their theory, based on archaeological evidence, implies that the Nile River was probably flowing east of the temples at the beginning of the 2^nd millennium BC. However, this assumption must be checked in the near future.

General landscape evolution for the archaeological site of Karnak

16Based on the palaeoenvironmental results described in this paper, it is possible to reconstruct landscape evolution linked to human occupation in the vicinity of Karnak. Before the Dynastic period (i.e., before 3000 BC), the Nile River did not flow around the Pharaonic temple complex. Indeed, aeolian dynamics combined with intermittent hydrological activity in wadi streams was dominant: within the Nile valley, large detrital fans were formed comprising coarse material derived from local sources (associated with angular pebbles). Gradually, the influence of the Nile floods increased throughout Upper Egypt and alluvial material was deposited over a greater area: wadi fans and aeolian sediments were washed away by occasional high flood events and mixed with alluvial material. Sand levees (gezirah) formed at the interface between the desert area and the river channel. The development of small mounds, similar to the ones observed in the delta (Tristant, 2004, in press) probably facilitated the installation of the first settlers from the Sahara Desert, when arid and warm conditions affected the whole area between the mid-6^th and the 4^th millennium BC (Lario et al., 1997; Midant-Reynes, 2003; Kuper and Kröpelin, 2006; Kröpelin et al., 2008). Due to an intense phase of aridity since the mid-Holocene and an increase in deforestation in the Nile basin, landscapes changed drastically from the end of the pre-Dynastic period (Woodward et al., 2007). Our preliminary study of the Nile’s evolution during the recent Holocene revealed a novel feature: from the end of the Middle Kingdom/beginning of the New Kingdom up to the end of the Roman period, the main course of the Nile River flowed past the quay, built at the beginning of the 1^st millennium BC. The first chronostratigraphic sequence, based on a series of three ¹⁴C dates performed on material sampled around Karnak’s ancient quay, helped to estimate the period of alluviation of the Nile River. Approximately 18 m of fluvial material, composed of sediments ranging from clay to medium sands, accumulated in less than 2000 years (i.e., from ~ 1450 BC to ~ AD 350), in the vicinity of the quay (between the tribune and the Taharqa ramp). Archaeological evidence reveals that the Nile River was flowing close to the temples complex: the fluvial ramp built by Pharaoh Taharqa (25^th Dynasty) and dated from the 7^th c. BC was built over the main quay in order to facilitate the arrival of the royal barque. Sedimentological results show that a narrow channel connecting this part of the quay to the main Nile River course was still in use during late Roman times but definitely abandoned during Byzantine and Islamic periods.