The oldest Homo erectus buried lithic horizon from the Eastern Saharan Africa. EDAR 7 - an Acheulean assemblage with Kombewa method from the Eastern Desert, Sudan

Although essential for reconstructing hominin behaviour during the Early Palaeolithic, only a handful of Acheulean sites have been dated in the Eastern Sahara region. This is due to the scarcity of sites for this time period and the lack of datable material. However, recent excavations in the Atbara region (Sudan) have provided unique opportunities to analyse and date Acheulean stone tools. We report here on EDAR 7, part of a cluster of Acheulean and Middle Stone Age (MSA) sites that were recently discovered in the Eastern Desert Atbara River (EDAR) region, located in the Eastern Desert (Sudan) far from the Nile valley. At EDAR 7, a 3.5 metre sedimentary sequence was excavated, allowing an Acheulean assemblage to be investigated using a combination of sedimentology, stone tool studies and optically stimulated luminescence dating (OSL). The site has delivered a complete Acheulean knapping chaine opératoire, providing new information about the Saharan Acheulean. The EDAR 7 site is interpreted as a remnant of a campsite based on the co-occurrence of two reduction modes: one geared towards the production of Large Cutting Tools (LCTs), and the other based on the flaking of small debitage and production of flake tools. Particularly notable in the EDAR 7 assemblage is the abundance of cleavers, most of which display evidence of flake production. Implementation of giant Kombewa flakes was also observed. A geometric morphometric analysis of hand-axes was conducted to verify a possible Late Acheulean assemblage standardisation in the Nubian Sahara. In addition, the analysis of micro-traces and wear on the artefacts has provided information on the use history of the Acheulean stone tools. Sediment analyses and OSL dating show that the EDAR 7 sequence contains the oldest Acheulean encampment remains in the Eastern Sahara, dated to the MIS 11 or earlier. This confirms that Homo erectus occupied the EDAR region during Middle Pleistocene humid periods, and demonstrates that habitable corridors existed between the Ethiopian Highlands, the Nile and the Red Sea coast, allowing population dispersals across the continent and out of it.

the sample size was limited to 18 specimens. Every individual number of the artefact was entered to MS Office Excel spreadsheet (separately for each site) and the sample was selected using random sample formula. The artefacts subject to the analysis were made from various types of raw material of local origin (Table 1). Geometrical and morphological analysis followed the method proposed by A. G. Costa [40], which consists in the 2D analysis of semi-landmarks superimposed on the outlines of artefacts, allowing comparisons between stone artefacts using comparative points and identifying differences between given populations [33,[40][41][42][43]. The method has certain limitations as it analyses only artefact's silhouette [39,44,45].
The photographs and scans of the objects were aligned along the symmetry axis [46] and outlined with Inkscape (version 0.92.1). The resulting outlines were converted from JPG to TPS format using the tpsUtil64 programme (version 1.78). In the next step, 75 semi-landmarks (including one fixed landmark and 74 semi-landmarks) were superimposed, starting from one fixed landmarked located at the tip or maximal curve of the distal round extremity and also in the case of cleavers and "cleaver-like" hand-axes at the point of crossing main technological axis and distal edge (S1 Fig), with the use of the tpsDig2 automatic curve-tracing function. The landmark coordinates for all the objects were generated with the PAST 3 programme (PAleontological STatistics). All the objects were adjusted in the Procrustes analysis, yielding averaged and standardised images of all the artefacts in a given group (S2 Fig). The transformed objects were compared using principal component analysis (PCA) and the acquired image was tested using Thin-Plate Spline Deformations; this way we were able to see the tendencies in the standardisation of the acquired shape at the extreme points of both axes. Twenty-five landmarks selected proportionally to each object were examined with the use of Multivariate Analysis of Variance and Permutational Multivariate Analysis of Variance (MANOVA, PERMANOVA), to allow identification of statistical differences between the assemblages. These two tests make different assumptions about the underlying datasets, making our analysis more robust: the MANOVA tests comparing equality mean values for univariate population having Normal Distribution, the PERMANOVA tests distance of observation from centroids and is resistant to the assumptions of variance homogeneity and normality of distribution [47,48]. For each of the tests we assumed an alpha level (0.05) which defines their statistical significance.

Use wear analysis
In total, 15 quartzite artefacts that fulfilled the following criteria were selected for use wear analysis. Only the forms classified as tools during the technological analysis were included. The tools had to have at least one intact, functional working edge. Additionally, edge-rounding and micro-chipping was taken into consideration. Artefacts were cleaned for 2 to 5 minutes in an ultrasonic bath. All artefacts were observed under a NIKON Eclipse LV 100 with magnifications between 200× and 500×. Analysis included scanning of the entire edge of each tool, as well as the areas immediately adjacent to it. Additionally, ridges and surfaces in the middle of Table 1. Sample of hand-axes selected for analysis.
Site n artefacts were evaluated for post-depositional wear. Where evidence for use wear was found, it was documented with a Hirox RH-2000 digital microscope. Images were taken with a motorised extended focus system. Enhanced digital processing including anti-halation and contrast adjustments helped to visualise the traces better and reduce the glare of the highly reflective surfaces [49]. Due to the heterogeneous nature of quartzite, surface traces develop irregularly and in patches, being more abundant on the crystals than on the matrix [50][51][52]. Moreover, various diagnostic alterations such as fractures, impact pits or striations, may not necessarily occur on the same edge. These properties, combined with the high reflectivity of its surface, make quartzite more challenging to analyse than homogenous materials such as chert [53]. Therefore, greater magnifications, usually between 200× and 500×, are required to observe specific surface features [54].
Although far from being the most frequently studied raw material, quartzite has gained popularity in use wear analysis in the past few years [50,55,56]. Numerous experiments were conducted, concerning both the formation of use wear [53,57,58] and post-depositional traces [59,60]. Their results provided guidelines for observation and interpretation of quartzite use wear.

Optically Stimulated Luminescence (OSL) dating
Samples were collected in opaque metal tubes hammered into the face of a cleaned section. The samples were then processed in the Korean Institute of Geoscience and Mineral Resources (KIGAM) and in Gliwice Absolute Dating Method Centre (GADAM) laboratories under subdued red light. Sunlight exposed material was removed and retained for dose rate measurements.
Carbonates and organic matter were removed from the unexposed sample using 1M HCl and 20 volumes H 2 O 2 respectively. Pure quartz was extracted from the 90-212 or the 90-125 μm fractions using density separations at 2.62 and 2.70 g/cm 3 and a subsequent HF acid etch (23M HF for 40 min followed by an 10M HCl rinse). Refined quartz was deposited as a monolayer on aluminium discs using Silkospray silicone oil.
Single-aliquot regenerative-dose procedures [61] were applied to all samples using preheats of 260, or 220˚C for 10 s prior to measurement of the natural/regenerated luminescence intensity (PH1), and 160, or 220˚C for 10 s prior to measurement of the test dose luminescence intensity (PH2; S1 Table). These temperatures were deemed the most appropriate to preheat the samples following dose recovery preheat tests (S3 Fig).
At KIGAM measurements were carried out using a Freiberg Instruments Lexsyg Smart system TL/OSL reader equipped with a Hamamatsu bi-alkaline photomultiplier tube (H7360-02), while at GADAM measurements were performed on a Daybreak Model 2200 reader. Irradiations were carried out using 90 Sr/ 90 Y beta sources. Stimulations were carried out at 125˚C for 60 s using blue light emitting diodes. Aliquots were heated at 5˚C/s during all heating steps and a 10 second pause at 125˚C prior to optical stimulation was used, to allow for the lag in temperature between the thermocouple and sample. The OSL intensity is that recorded during the first 1.5 seconds of stimulation with a background signal subtracted. All growth curves were fitted using a saturating exponential plus linear function, a typical growth response curve is displayed in S4 Fig. The OSL characteristics of the quartz from Sudan show a rapidly decaying signal and continuously growing dose response curve, which makes it well-suited for application of the SAR protocol used in this study.
Recycling ratios and recuperation were calculated to monitor the performance of the SAR procedure [61], while sample purity was assessed by measuring an IR depletion ratio [62].
Aliquots not yielding recycling or IR depletion ratios consistent with unity or displaying recuperation greater than 5% of the natural signal were rejected. The sample equivalent dose was calculated using the Central Age Model (CAM) [63] on the D e values obtained from accepted aliquots. Radial plots showing the spread of the D e s and the calculated CAM for each sample  are presented in S5 Fig and Table 2.
Radioisotope concentrations were measured with Canberra gamma spectrometers with HPGe detectors-the samples did not display any sign of radioactive desequilibria. Since the moisture content of the samples is expected to have changed during burial, a value of 8 ± 3% was assumed for all samples, to account for the plausible range of past conditions and humidity changes.
Beta and gamma dose rates were calculated from radioisotope concentrations and moisture contents using the conversion factors of Guérin et al. [64]. The alpha dosed rind of the quartz grains was assumed to have been removed by HF etching [65] and hence the alpha dose rate was taken to be zero. Beta dose rates were corrected for grain size using the attenuation factors of Guérin et al. [66] and an etch attenuation factor after Bell [67]. Cosmic ray dose rates were calculated based on the altitude, latitude and longitude, present-day burial depth and overburden density of the sample [68]. Overburden densities of 1.8 g/cm 3 were assumed. The CAM D e , the dose rates and ages for each sample were calculated using DRAC [69] and are presented in Table 2.

Composite sedimentary stratigraphy of the EDAR area
A thick sedimentary sequence containing a Pleistocene to Holocene succession was excavated in the EDAR area. Detailed analysis of the Pleistocene fluvial sediments has been carried out through five correlated profiles, each subdivided into units (Fig 4). The simplified cross section of EDAR sedimentary deposits shows a~5 m thick stratigraphy divided into three units (Units I-III) bounded by erosional surfaces [32]. Unit I consists of a stratified pebble gravel unit (Unit IA) and massive sand with abundant calcium carbonate nodules (Unit IB), suggesting that the sediment deposition occurred in braided streams followed by calcium carbonate precipitation under arid climatic conditions. Unit II comprises planar-to cross-stratified pebble gravel (Unit IIA) overlain by a metre-thick, massive and carbonate-cemented sand (Unit IIB), indicating another episode of fluvial incision and sediment deposition followed by carbonate precipitation under more arid climatic conditions. Unit II is overlain by a thin and continuous gravel layer, probably representing a desert pavement or relict paleo-topographic surface which may be produced by the removal of sand and dust by wind and intermittent rain. The overlying Unit III comprises yellow to dark brown silt sands, which can be divided into three sub-units based on subtle changes in grain size and sediment colour. The unit is structureless but contains rare pebbles and tubular voids interpreted as rhizoliths.

Sediment lithology of EDAR 7
The 3.5 metre-thick sedimentary sequence at EDAR 7 sits upon rhyolite bedrock (Fig 4). The lowermost layer (UNIT IA) contains an Acheulean horizon. This unit is composed of a crossstratified thin alluvial pebble bed containing rare cobbles with a diameter of up to 30 cm, topped by well-rounded gravels supported in a well-sorted sand matrix. The orientation of elongated pebbles indicates north to south paleo-flow direction. Overlying Unit IB is 1.5 m thick and composed of massive medium-and coarse-grained sands cemented with secondary calcium carbonate, also containing a band of carbonate concretions. The overall features of this unit suggest rapid deposition of sand by floods, followed by precipitation of calcium carbonate under dry climatic conditions. Unit IIA starts at a depth of about 120 cm, and is

PLOS ONE
characterised by a cross-stratified gravel layer indicating the fluvial origin of the sediment. The characteristics of this deposit suggest that the sediments were deposited from braided streams, which were probably shallow and ephemeral and had variable paleo-current directions and discharge ( Fig 5). The variations in fluvial style and discharge between UNIT IA and IIA are connected with climatic changes and resultant river network modifications. Above this unit carbonate-cemented sands with concretions (UNIT IIB) appear again. The upper layers of the site comprise massive structureless fine-grained sands, indicating periods of slow but continuous sedimentation and slow denudation. An intermediate silty sand layer marks a period of reduced clastic input and morphological stability. Sediment facies starting with gravel bed stream deposit and terminating with structureless silty sands indicate environmental changes from stream channel to arid savannah and grassland with the development of a soil profile. The unconformity surfaces between Units II and III are erosive surfaces occurring locally. These are desert pavements which have formed through winnowing sand and dust by the wind and intermittent rain, leaving the larger clasts behind. The desert pavements indicate the onset of intense aeolian processes.

Multi-proxy records of EDAR 7
The mean grain size is highly variable through the EDAR 7 sequence. Grain size analysis and magnetic susceptibility therefore provides important clues to the sediment provenance, transport and depositional condition. It generally decreases with depth and the primary trends agree well with granulometric texture (Fig 4). Relative mean grain size decreases at two intervals, from 157 to 201 cm (Unit IB) and from 254 to 331cm (Unit IA), respectively recognised as fluvial levels. Magnetic susceptibility is the degree to which a material can be magnetised by an external magnetic field. Therefore, magnetic susceptibility provides information about sediment composition.

Chronology
The Acheulean layer (unit IA) was too cemented to sample, but a sample in the lowermost part of unit IB in EDAR 7 was taken, providing a minimum age for the Acheulean horizon. A similarly positioned sample was also taken at the nearby site EDAR 135, to refine the dating of the Acheulean layer. At EDAR 7, samples were also taken in the middle of unit IB, in unit IIA, and in the overlying sand layers units IIB, and III. The equivalent dose distributions from these samples have a range of overdispersion values (OD; the relative spread of equivalent doses after measurement uncertainties are excluded). The samples EDAR7-4 and EDAR7-5, which were taken from aeolian sand devoid of archaeological material overlying the entire studied sections, have OD values of~50 and 70%, which is far higher than what would be expected if they were well-bleached and unbioturbated. A geological analysis of units IIB and III shows that these layers are affected by soil formation processes which may have led to the vertical movements of sand grains. The use of the minimum age model was therefore not considered for these samples despite their large OD, since it is highly sensitive to the inclusion of younger grains into an older sediment body [70]. The CAM was used for the age calculations for these two samples, though if younger grains are intruded into these levels by pedogenic processes, this approach will yield age underestimates. Unit IIB and III were dated to 22 ± 3.4 (EDAR7-4) and 10 ± 2.1 ka (EDAR7-5) respectively, and most likely correspond to MIS2 arid episodes favouring the deposition of aeolian sands which were fixated and bioturbated at the beginning of the Holocene.
Samples EDAR7-1, EDAR7-2, EDAR7-3 and EDAR135-S6 have relatively low OD, ranging from~10 to~30%. These values suggest that the sediments were well-bleached prior to deposition, and not subsequently subject to bioturbation. Consequently, the Central Age Model (CAM) was used to calculate ages for these samples [63]. Despite the measurements of De's above 100 Gy the samples did not show any sign of saturation in EDAR7. The samples taken at the contact between units IA and IB were dated in EDAR 7 to 280 ± 27 ka (EDAR7-1) and to 391 ± 30 ka in EDAR 135 (EDAR135-S6), suggesting that the Acheulean level in EDAR 7 is MIS9, but possibly MIS11-13 or older. The morphology and wear-conditions of artefacts in UNIT IA indicate that they may have been reworked by overland flows and re-deposited into younger sediment (280 ka) reinforcing the hypothesis that these artefacts are at least MIS11-13.
The other sample taken in the middle of unit IB at EDAR 7 was dated to 199 ± 12 ka (EDAR7-2). EDAR7-3 was sampled directly at the interface between unit IIA and IB, and yields an age of 158 ±15 ka. This dates the deposition of unit IIA to between 199 ± 12 ka and 158 ±15, suggesting an MIS7/6 age. The~100 ka hiatus between the units IIA and IIB probably relates to erosion during intense arid episodes of the uppermost unit IIIA (visible in other sites in EDAR). Moreover, erosion of sediments laid down during periods of low sedimentation rates could have been enhanced by tectonic activity, i.e. during the uplift of the Nubian massif [71].

EDAR 7 lithic inventory
Artefacts taphonomy. Preservation state (S2 Table), degree of erosion and breakage patterns can be used to determine the taphonomic history of the lithic inventory. Less than 0.5% of the assemblage is fresh. The remaining 99.5% display only slight traces of abrasion or is abraded. The degree of abrasion in all raw material groups is approximately the same. Over 99% of both quartzites and rhyolites are abraded, which reflects the fluvial environment within which the assemblage was found. Such deposition, especially within a high-energy fluvial system, coupled with the brittleness of the raw material and different aspects of knapping and usage processes explain the breakage patterns:13.5% of quartzite and 65.9% of rhyolite artefacts are incomplete. No evidence of contact with fire was observed on the artefacts.

General structure
The EDAR 7 assemblage includes 918 artefacts weighing 115.4 kg in total (Table 3). Raw materials used are dominated by quartzite (90.5%) and fine-grained greenish rhyolite (8.7%), both available in the immediate surroundings. The Hudi chert, identified from outcrops in the vicinity of EDAR 7 and archival sites [27] was used only occasionally (only one artefact in EDAR 7, but this raw material was also present at other EDAR sites). The assemblage consists of products of all stages of core use and tool production. Among 138 tools (15% of the assemblage), 44 large cutting tools (LCT) were identified, which is 32% of the tool assemblage. Apart from one hammer stone, the remaining tools are retouched flake forms, hardly ever exceeding 3 cm in length (67% in the category of tools). Over 7% of the assemblage are cores, almost without exception made from quartz. These cores are very variable in size, ranging from giant cores to small specimens. Blanks with waste (debris) constitute over 77% of the assemblage. The most numerous in this group are chips, making up over 40% of the assemblage.

Cores
Cores are abundant at EDAR 7, constituting 7.5% of the inventory (see S3 Table for the frequency of individual categories of cores by raw material group). With the exception of two rhyolite examples, all cores are made of quartzite. In terms of degree of exploitation, over 70% of cores are considerably exploited blocks (S4 Table). Even though the cores belong to ten separate categories, closer examination reveals that three categories dominate, quite well reflecting the implemented technological measures. Initial cores and amorphic forms, which usually display negatives of removal situated chaotically on the perimeter of the blocks, jointly constitute over 40% of all the cores. This group also includes two giant cores, which do not display evidence of advanced reduction (Fig 6). The second group in terms of its size is single-platform, unidirectional cores with crude striking platforms-30% (Fig 7). The third characteristic group is the cores with negatives of flakes knapped on more than two platforms, where evidence of exploitation is seen on a greater part of core's surface. These three groups constitute over 85% of all the cores. Cores from the remaining categories occur only individually, even though discoidal cores constitute a substantial category (S6 Fig).
In terms of size and weight of cores the assemblage is diversified (S5 Table). Giant cores are present; the largest is longer than 34.0 cm and its weight exceeds 9 kg (Fig 5). These blocks were used in biface manufacture and have many analogies at Acheulean sites [72][73][74]. On the

PLOS ONE
other hand, the assemblage also includes quartzite cores of microlithic proportions, whose length and width hardly ever exceeds 3-4 cm (S7 and S8 Figs). These blanks were used to produce tools, such as small endscrapers and perforators.

Debitage and waste
Debitage and waste (debris and chips, i.e. flakes with length � 15 mm), numbering 710 artefacts, constitute over 77% of the EDAR 7 assemblage (Table 3). Blanks and waste weighed over 27 kg, which is~24% of the weight of the EDAR 7 assemblage. No products which could be interpreted as blades were observed among the blanks, so this category is made up by unretouched flakes, including core management pieces and technical forms, such as hand-axe shaping flakes.

PLOS ONE
(67% of blanks). Mean length and width of complete flakes is near 50 mm, with a mean thickness of~20 mm.
The analysis of natural surfaces on the dorsal sides of flakes shows that over 20% are noncortical. The remaining products, over 75%, are partially or completely cortical (Fig 9). If the two categories with the greatest contribution of cortex (51%-100% of cortex) are treated jointly, they constitute~40% of flakes. This suggests that preliminary flaking took place at the site. Common presence of natural surfaces is also clearly seen in the case of the direction of blow on the dorsal side (S7 Table) and the types of flake butts (S8 Table). In the case of the direction of blow on the flakes dorsal side, in as many as 30% of products they are invisible (dorsal side covered by cortex). Presence of unidirectional negatives is in principle the same as multidirectional

PLOS ONE
ones (more than two directions) and each constitutes~30% of the assemblage of flakes. This reflects preliminary unidirectional exploitation of frequently big cores on the one hand, and on the other it is the evidence of advanced multidirectional exploitation of small quartzite blocks.
Although flakes with plain butts were dominant (>40%), forms with cortical butts were present and constituted a significant fraction (>20%). Overall, butt attributes prove that limited preparation measures were used in core reduction (fewer than 3% are prepared). Hard percussors were used in stone working, which is substantiated by, e.g. thickness of the flakes and absence of faceted butts and lip [75,76]. One hand-axe edge modification flake was identified in the assemblage, but it is possible that some multidirectional quartzite blanks are also connected with this activity.
Two products with two ventral faces were identified among the flakes (Fig 9D). Small forms of this kind are known as 'dorsally plain flakes' as defined by Dag and Goren-Inbar [77], while larger ones are known as Kombewa flakes [78]. While the former are not intentional products, the latter were predetermined with the intention of biface production.
The blanks group includes 3 quartzite flakes, larger than 10 cm and of substantial weight (the heaviest specimen exceeds 1.5 kg-see S9 Table). Such forms can be defined as giant flakes sensu M. Kleindienst [79] or Clark [80]. Potentially, they could have been used in the production of bifacial tools [81].  Table 4). Some of the hand-axes could possibly be also included into a pebble tool category, i.e. specimen in Figs 11A and 13C could be classified as lateral-distal bifacial-choppers known from much older inventories i.e. from the Melka Kunture site in Ethiopia [82]. Most of the hand-axes and cleavers and all the choppers were preserved intact or with only a small breakage of the tip (S10 Table). Only four tools display a small degree of abrasion (2 choppers and 2 hand-axes), while for 40 out of 44 were determined as abraded.

Tools
In terms of raw material, the frequency of rhyolite is conspicuously larger among LCTs (over 18%) than in the remaining categories of the assemblage. Preference for this raw material is especially marked among cleavers: 66% of them were made of rhyolite, as well as two forms defined as hand-axes with "cleaver-like" edge (the crosswise edge smaller than half of the product's width) [72]. One of the choppers was made from chert-outcrops of this raw material are absent from the immediate surroundings of the site.
The most frequently used material to produce nearly 65% of bifacial tools were big pebbles and cobbles of naturally flat shape. Specimens based on giant flakes constitute an important contribution in the assemblage. One of them has two ventral sides, which testifies to the use of Kombewa flakes (Fig 13B). Such behaviour might be underrepresented due to later modifications and surface deterioration obscuring the original blanks. Use of the Kombewa method was also noticed at EDAR 6, an Acheulean site located~2 km from EDAR 7.
The cross-sections of the bifaces indicate a strong tendency to maintain flat-convex, trapezoidal or lenticular forms. The final shape may have resulted from either the selection of raw material blocks of preferred shapes and the exploitation of natural surfaces, or the use of flake blanks.
The sizes of hand-axes and cleavers are presented in S11 Table. The width is the most variable dimension, ranging from 51 to 179 mm. Some variation is also visible in length. Cleavers are slightly smaller than hand-axes, which may result from the fact that some of the former were based on flakes, while the latter on quartzite blocks of considerable sizes. Considerable variation of sizes is evident among choppers and chopping tools: minimum and maximum values of all four features are spread over a wide range. Medians and mean values reach approximate results, which indicates the absence of outliers.

PLOS ONE
Differences are seen in the general picture of use of both surfaces in the reduction of the bifacial forms from EDAR 7. The mean negative count for face one was 1.71 greater than for face two (S9 Fig); this is furthermore reflected by the higher maximum value and median for

PLOS ONE
face one. As far as the frequency of occurrence of negatives on surfaces is concerned, there is a clear tendency for reduction of the distal part and medial lateral part on face one, while face two is characterised by a great number of negatives in the distal part. Overall, EDAR 7 bifaces display considerable surface modification during production. Nineteen have less than 50% of

PLOS ONE
Thinning retouch and bifacial edge retouch were rarely implemented during the production and repair of LCTs. The former was observed on face one of five artefacts and on face two in three artefacts. The latter was infrequent as well and occurred on four tools (S9 Fig).

PLOS ONE
Bifacial tool production methods. Three significant production methods of bifacial tools are present in the EDAR 7 assemblage (S10 Fig). The first one focused on making hand-axes and cleavers out of cobbles and chunks. In the case of cleavers, production started with a flat cobble and was focused on bifacial shaping with blows perpendicular to the main axis; these blows tended to be invasive and removed most of natural surface (Fig 17B and 17C). One example of treating the natural surface as a striking platform in early shaping was observed; this was somewhat similar to unidirectional core reduction. Preparation of a working edge by either detaching one or two perpendicular flakes or using a series of bifacial parallel blows was the next stage in that method.

PLOS ONE
In the case of hand-axes, the strategy was somewhat different and aimed at preparing and shaping the tip and both sides without (or with minimal) modification of the base (Figs 10B,  The second method of producing hand-axes from cobble can be described as a classical, circular bifacial approach with three consecutive stages of reduction (Figs 10A,10C, 11A,11D, 12B,12C, 13A, 14A and 14B and 15B). The first stage is associated with general block shaping; the second was aimed at roughing-out and shaping the tip and sides. Small scars of thinning and bifacial retouch resulting from the last stage of faconnage are located near tool edges. Such hand-axes have almost no natural surfaces on both faces, are symmetrical and have a regular, cordiform, limande or oval shape. The last two stages of faconnage of hand-axes with cleaverlike edge were slightly different. The working edge was created by either two notch negatives or tip reduction using small perpendicular flakes.

PLOS ONE
The third method was based on using giant flake blanks (Figs 11C, 13C, 16A-16C). It is confirmed not only by flake LCTs, but also by two giant cores (Fig 6) and a few large flakes (Fig 9A-9C). Blank roughing-out is similar for both cleavers and hand-axes and it was focused mostly on dorsal side reduction. Most ventral side scars can be linked with butt thinning or

PLOS ONE
edge of cleavers and cleaver-like hand-axes was created by detaching notch-like blows located on both sides of the product. This method is also known from early Acheulean assemblages of Eastern Africa [83].
One hand-axe was made on a Kombewa flake (Fig 13B). Faces and edges of this artefact were reshaped only slightly; natural convexities and sharp edges of the blank were used instead. Thus, the reduction was mostly focused on removing the bulbs. As mentioned before, the Kombewa method could have been used more extensively, but it was later obscured by heavy modification and abrasion of artefact surface.
Two main factors had a significant impact on using different strategies of production as a well differently shaped degree of LCT's in the production process. The first concerns the transmission of knowledge within a group and also a skill level of individuals. Mastering "knowledge" (connaissance) as well as obtaining an appropriate level of "know-how" (savoir-faire) distinguished experts from novice knappers [84]. Expert knappers were better at planning and controlling the reduction process, including the application of appropriate solutions when mistakes were made or they were faced with raw material limitations [85]. The hypothesis of different skill levels could be confirmed by the presence of bifacial forms made of fine-grained rhyolite (see Figs 11C, 12A and 12C, 13A and 13C and 16) and quartzite (see Figs 10A, 11A, 12B, 14B, 14C, 15A and 15B) with strong control of both faces and regular and deep removals, which could be a result of a high skill of knappers. On the other hand the forms displaying a small degree of processing the face and the presence of numerous hinges point to an inefficient skill level.
The second factor is connected with the adaptation of production methods to raw material. To verify the possibility of using direct percussion technique with a hard hammer in the production of LCT's and reduction of giant cores, we conducted a short experiment with local raw materials which we gathered from the area close to the location of EDAR sites (S11 Fig). Two knappers with advanced skill level were involved in the knapping activity (M.E and G.M) (S11A and S11B Fig). Experiments showed that three kinds of raw material (fine-grained rhyolite, coarse grained rhyolite and quartzite) have varied fracture properties, which is indicative of different degrees of control in the detaching process (S11D-S11F Fig). Fine-grained rhyolite has the best fracture properties, which allowed better control of flaking angle and surface shape at each stage of shaping (S11E Fig). The quartzite raw material is characterised by good fracture but it required grater control skills-use of a hammerstone which was too heavy caused severe chipping and microcracks in the structure of raw material, which prevented continuation of shaping (S11A-S11D Fig). The lowest fracture properties are characterised by coarse-grained rhyolite, which causes presence of numerous hinges and difficulty in decortications and detaching overshot flakes (S11F Fig). A large number of hinged scares on faces and

PLOS ONE
EDAR 7 -an Acheulean horizon from the Eastern Desert, Sudan asymmetrical shape of LCT's is visible in the assemblage from EDAR7-bifacial tools made from coarse-grained rhyolite (Figs 11B, 15C and 16A). Retouched tools. The retouched tools, made on flakes and sometimes chunks, constitute 67,4% of the tool category; the remainder are LCTs and one hammer stone (Table 5 and Figs 21-24 and S12-S20). Such ubiquity of flake tools, some of which are very small, suggests differentiation of activities that took place in EDAR 7 area, and a functional complexity of the assemblage.
The state of preservation in this group shows that fewer than 10% of artefacts are fragmented. Core reduction stage was not a crucial factor for blank selection: non-cortical (52%) and cortical ones (47%) were used. Butts of more than 40% of the blanks were natural; another 41% were plain with only marginal participation of dihedral, linear, and punctiform types. Some of the tools, especially perforators (Figs 22 and S15) or endscrapers (Figs 23 and S17) were rather diminutive, with lengths~20 mm and width not exceeding 10 mm (S12 Table). Such small size could be a premise on which the possibility of hafting can be considered.
Both raw materials were used for the manufacture of flake tools. As in the entire assemblage, quartzite dominates over rhyolite. With the exception of quartzite-only endscrapers, all tool categories were made of both raw materials. Retouched flakes were the most common tool type. Denticulate, notches, perforators and sidescrapers were found in equal proportions, each type making up more than 10% of the tool group.
The most frequent retouch type is notch/denticulate, usually on the dorsal side of the flake. Straight delineation predominates, while concave type occurs less frequently. The distal part and the right edge are the most commonly retouched flake zones. The mean length of retouched edge is~40 mm (S13 Table).
Stone slab. The Acheulean artefacts were accompanied by a large, flat, rectangular quartzite block (151×140×44 mm, weight 1,57 kg) trimmed by several blows to the edges (Fig 25A). A similar block, but made of rhyolite and much bigger (220×180×75 mm, weight 4,25 kg) was also found at the nearby Acheulean site EDAR 6 ( Fig 25B and 25C) [32]. Such blocks are known among others from the Sangoan levels at the Sai Island site and are interpreted as grinding stones [86]. Basalt blocks of this kind have elsewhere been used primarily as an anvil, a passive percussion tool e.g. at Gesher Benot Ya'aqov [72,87].

Geometric morphometric analyses of hand-axes
Results of the principal component analysis (PCA) (S14 Table) indicate that the first 10 components describe 96.1 of the variance of the whole population, with the first two components accounting for 46.8% and 29.3% respectively. PCA results in the form of the scatter plot were

PLOS ONE
presented on the basis of the first two components due to the fact that when combined they account for over 75% of variance.

PLOS ONE
The Thin-Plate Spline Deformations analysis showed diversification of artefacts' shapes situated in relation to the axes of two main components. To determine the shapes of products, we resorted to the classification proposed by F. Bordes [88]. Two irregular subcordiform (up) and cordiform (down) LCTs are visible on the axis of the first component, while the second component is connected with the elongated (left) limandes shape similar to the cleaver-like edge and the shape approximating an oval (right). Location of the objects from assemblage EDAR 7 displays three tendencies in acquiring particular shapes: approximating elongated limandes and irregular subcordiform and regular limandes. A tendency similar to EDAR 7 is seen in assemblage Bir Sahara 14, yet in this case it differs in the location of the observed

PLOS ONE
extreme objects and increased frequency of observed objects of the limandes shape. Samples from Dakhla Oasis and Kharga Oasis show a great similarity in the dispersion of data, with a strong tendency to assume limandes and elongated limandes shapes. The PCA analysis yielded totally different results for assemblage EDAR 133. There the assemblage is characterised by strong standardisation of product shapes, which is seen in very small spread of observations (Fig 26A). The LCT's found at EDAR 133 mainly display irregular subcordiform shape and the shape approximating an oval.

PLOS ONE
Mardia's kurtosis test showed absence of normality of distribution for all variables (p = 2.414E-125). Because the assumption of normality of distribution was not met, we used the MANOVA test with Pillai's trace. The results of the MANOVA and PERMANOVA tests (S15 Table) did not reach the required value, which unequivocally indicates the rejection of the zero hypothesis about absence of diversification between the assemblages subject to the analysis. The results of similarity in pairs implemented in the PERMANOVA test indicate a similarity between assemblage EDAR 7 and Bir Sahara and a considerable similarity between Dakhla Oasis and Kharga Oasis (S16 Table). Results exceeding the assumed significance (p>0.05) are also visible when comparing EDAR 133 and Dakhla Oasis; yet this result should be treated with extreme caution, due to a low value exceeding significance (p = 0.0699).

Spatial arrangement of the assemblage
The artefact positioning indicates that the excavated area represents remains of an occupational surface with stone knapping activity. Analysis of the spatial distribution of lithics does not allow the identification of separate functional zones within the trench, although several clusters of stone artefact categories could be observed (Fig 27). This contrasts with the other Acheulean encampment remains from Sudan-Arkin 8, where traces of some of the earliest known domestic structures in Nubia have been documented. Arkin 8 consists of a main concentration of finds, several subconcentrations, cairns and possibly wind-shelters [28]. It was interpreted as a "prehistoric camp, with its living floor". The area documented at EDAR 7 is

PLOS ONE
much smaller than at Arkin 8 (9 m 2 and 64 m 2 respectively) and it is clear that only a small part of the site has been excavated.
The majority of artefacts are located in the northern zone of the trench, especially within metre no. 2, which contained 20% of the whole assemblage. However, the density maps show that the individual categories of stone artefacts concentrate in slightly different areas of the excavation. The bifacial component was most frequently recorded in the northern part within metres no. 1 and 4. Smaller clusters were also identified near the western and eastern margins of the trench. Retouched tools are located most frequently in the north-eastern part. In the case of cores, several clusters can be seen. The highest density of this type was recorded within metres no. 2 and 5. Smaller core clusters formed within metres no. 1, 4 and 6. The pebble component (5 chopping tools and 3 choppers) constituted a set too small to form any conceivable clusters.

Use wear analysis
The study of tools from EDAR 7 yielded results that are interesting both from the point of view of preservation and interpretation of traces. Macro traces suggesting possible wear were detected on the edges of all 15 artefacts. They included heavy rounding and numerous scars with step and feather endings. In 11 cases, more than one edge had macro traces suggesting use ( Table 6). The remaining 4 had natural surfaces which could have been used as a back for handling the tool.
Direction of tool movement was identified based on linear features [49,57]. Three distinct types of movements could be recognised on the tools: cutting, scraping and perforating. In most cases, the type of movement was consistent with the formal type of the tool. Each of the analysed artefacts tells a slightly different story of use and thus deserves a short, separate description.
A fragment of a tool (art. 357) with retouched edges that was used for working wood is an interesting case. All edges of the artefact were used and display different stages of development

PLOS ONE
of the same type of traces. Analysis revealed a smoothed surface covered with rough, domed and matt polish. A series of small, overlapping scars are apparent at low magnification. Wear of the left edge was least developed-irregular rounding and fractures were first to appear together with small concentrations of impact pits (marked with ellipse) (Fig 28A). Stronger traces were noticed on the proximal part of the tool. Linear features are crisscrossing behind the edge. The direction of grooves (marked with arrows) suggests more than one mode of movement of the tool: cutting and scraping (Fig 28B and 28C). In the distal part of the left edge, a well-flattened surface with large impact pits, grooves and striations perpendicular to the edge of the tool was located. This part of the tool was used for scraping ( Fig 28D).
Another example of distinctive polish was discovered on the surface of a small endscraper (art. no S9). Surprisingly, the tool was not used for scraping, but cutting and perhaps engraving https://doi.org/10.1371/journal.pone.0248279.t006

PLOS ONE
bones. Heavy traces were discovered along the left edge of the tool (Fig 29). Very little rounding could be observed with the naked eye, although series of visible, uneven scars were apparent. Well developed, highly visible, well flattened and shiny bright polish concentrated on the very edge of the tool, but this pattern did not continue onto the surface of the endscraper. Regular sleeks were found, going parallel to the edge (Fig 29A, 29B and 29D), as well as several perpendicular striations (Fig 29A and 29C). Two artefacts bore traces of both perforating and cutting (art. 340, 142). Both had a small, rather blunt tip, shaped with series of strikes to the edge. Extreme rounding of the edges and all protruding parts, especially the tips was visible to the naked eye Unevenly distributed small scars with step and feather ending could be also noticed. At the microscale, some differences between the two artefacts could be easily detected. On one (Fig 30A), polish appears in irregular, matt spots. There are also numerous cracks and impact pits forming linear features running perpendicular, around the perforator, and parallel along one of the edges of the tool. The other revealed regular, deep sleeks perpendicular to the edge (Fig 30B). The material worked was impossible to determine precisely due to heavy post depositional abrasion (Fig 30C). It is possible that the first of the artefacts was used on softer material, whereas the second on something harder.
Directions of traces, but not the specific material worked, were identified for two more artefacts. On one, some clues could be deducted from macro traces (art. S8). The left edge of the tool was visibly rounded, with many overlapping chips of step and feather ending. The

PLOS ONE
remaining edges appeared somewhat fresher. Further traces were observable in greater magnification only: spots of rough polish on the protruding part of the tool; several scattered impact pits and some irregular striations perpendicular to the edge (Fig 30C). Such traces could imply use on some kind of soft material, maybe hide or meat [89,90], but post-depositional traces make this interpretation difficult. A small knife showed similar wear (art. no 69). One of the edges, heavily crushed, bore all the traces. This edge was opposed by a natural surface that could be used for handling the artefact. Most of the micro traces were obscured by post depositional wear. It was only established that some patches of more regular, bright polish sleeks going parallel to the edge could signify that this tool was used for cutting ( Fig 30D).
Heavy post-depositional modification made the interpretation of function of half of the observed artefacts impossible. It can be distinguished from other types of wear by several

PLOS ONE
features. It appeared in the form of scattered polish and chaotic abrasion of surface of tools. In extreme cases, almost the entire surface appeared dull (Fig 31A). Fresh scars visible on the heavily abraded surface are evidence of damage during or after excavation (Fig 31A). Numerous impact pits and irregular striations are distributed randomly and do not form any pattern. Impact fractures and cracks are found mostly on flat fracture or cleavage surfaces of the tools (Fig 31B).

Discussion
Results of the EDAR (Eastern Desert Atbara River) area study indicate the presence of Acheulean occupation remains situated in the paleo-river system. EDAR 7 is mainly typified by fluvial sedimentary environment, characterised by two sedimentary facies. During pluvial

PLOS ONE
periods channels bars (UNIT IA and IIA) and floodplains associated with grasslands and abandoned channels prevailed in the braided river (Unit IB and IIB). The advent of interpluvial periods interrupted the braided river system to activate aeolian and overland-flow process, typified by either relict stones as pavement, or silty sands possibly derived from wind-blown deposits under the arid-to semi-arid conditions. Fluvial environmental context in the Eastern Sahara is basically characteristic of most of the Acheulean locations besides those from Saharan oases or artesian wells. Sai Island [5], Nag Ahmed el-Khalifa [29], Arkin 8 [28], Khor Abu Anga [30], al-Jamrab [31] and others [38,91,92] are Acheulean sites located either on the Nilotic terraces, its tributary banks or in smaller, dry watercourse systems. The buried channel system of EDAR area is not connected with any present-day active river valley and represents an ancient river course. Especially the fluvial system of the Atbara River considerably contributed to the northward dispersion of Acheulean tradition in the Middle Pleistocene. Acheulean sites from the region of Khashm el-Girba in the upper reaches of the river were already mentioned by Arkell [27]. Their stratigraphic context of Acheulean artefacts within the fluvial context was presented by Chmielewski [93] and later by Abbate [71]. Abbate's research shows that a 50 m thick Pleistocene fluvial succession is extensively exposed in the area along the Atbara River from Khashm el-Girba to Halfa el-Jadida.
Outside Sahara, some of the main Eastern African Acheulean sites are similarly located in active fluvial environments, such as channel beds, with the occurrence of lithics accumulation [94]. One of such sites is Garba IV from Melka Kunture Formation along the upper Awash river in Ethiopia [95]. The sedimentary environment of EDAR 7 has certain similarity to that site. Morpho-sedimentary environments, however, differ from each other. The fluvial regime of EDAR 7 is predominantly controlled by paleoclimate change between semi-arid to arid and savanna climate, while that of Melka Kunture is mainly controlled by both paleoclimate and occasional volcanic activities in the highland area approximately above 2000 m (a.s.l.). In EDAR 7 sedimentary deposits are mostly composed of gravels, gravely sands and silty sands derived from the braided stream, while that of Melka Kunture is characterised by relatively fine sediments, i.e. silts and sandy silts in association with volcaniclastics or pyroclastics [95].
Unlike many Acheulean sites, which primarily contain bifaces with occasional debitage, the EDAR 7 assemblage consists of a full knapping sequence, from raw-material and initial cores, through debitage, to tools bearing evidence of usage. Consequently, EDAR 7 is interpreted as the remains of a buried Acheulean inventory, Homo erectus stone knapping leftovers. Furthermore, the assemblage contains not only hand-axes but also other large cutting tools. Our results highlight the co-occurrence of two reduction sequences in the same assemblage, one geared towards production of LCTs and the other based on the flaking of small debitage and production of flake tools, sometimes of microlithic proportions (endscrapers, perforators etc.). Amongst the LCTs, a considerable number of hand-axes and cleavers were produced from flake debitage. EDAR 7 is also one of the very few sites from Eastern Sahara where the Kombewa method was used in the production of LCTs. Assemblages produced with the use of giant flakes and the Kombewa method were previously only known in this region from a small Egyptian assemblage (site KAS-1) located near Bir Kiseiba in the Darb el Arba'in Desert, the area west of the Nile Valley between Dakhla Oasis in Egypt and Wadi Howar in central Sudan [96]. Outside the Eastern Saharan Africa, the Kombewa method has been documented from Eastern Africa [83,97,98], the Middle East [72] and lately also from the Arabian Peninsula [73,99]. A striking feature of the EDAR 7 assemblage is the abundance of cleavers and "cleaver-like" hand-axes, most of which display features testifying to flake production. Asymmetrical and not well-shaped cleavers along with proto-hand-axes and hand-axes have also been lately reported from Gebel Karaiweb in the Red Sea Mountains (Sudan). According to Kobusiewicz and colleagues, this surface inventory could be attributed, although judging only by the stylistic and morphological features of the artefacts, to the turn of the Late Early and Middle Pleistocene [91].
The morphometric analysis showed similarities between the EDAR 7 assemblage and Site 14 in the Bir Sahara depression; the latter has a large number of hand-axes in cordiform and irregular amygdaloid forms. Morphological variation between Acheulean assemblages in NE Africa appears to be the result of adoption of production methods to the locally available raw material.
Use wear analysis revealed numerous post-depositional alterations on the surfaces of observed artefacts (Fig 30). Analogous micro-traces and edge deformations were discovered on much younger (9500-5700 cal. BP) tools from Sai Island in Northern Sudan [59]. Experiments show that such patterns can be generated by water transport. Abrasion extent seems to depend on artefact stabilisation. Without it artefact roll downstream freely and abrasion, in extreme cases heavy and readily visible, could cover its entire surface. Stabilisation, e.g. by sand or pebbles, limits the changes only to the exposed parts of the artefacts [100]. Aeolian abrasion, on the other hand, creates numerous impact pits, flat fractures and cracks [60]. Both cases were apparent on the EDAR 7 sample, which made the analysis more challenging.
Use wear traces were observed on 7 out of 15 of the analysed tools (Table 6). Most artefacts combined more than one type of movement but were probably only used for one material, either wood or bone. Similar use wear was replicated during numerous experiments e.g. by I. Clemente Conte and J.F. Gibaja Bao [52], or by Knutsson's team [58,101] and has analogies in archaeological record. Several artefacts with wood use wear were found at the Oldowan sites of Konjera South (cutting and scraping) [53] and Koobi Fora in Kenya [102]. Early use of wooden tools is further supported by residue analyses, which revealed remains of phytoliths, for example on Acheulean stone tools from Paninj in Tanzania [103].
On three artefacts, the traces were less developed but could still be interpreted as resulting from cutting of undetermined soft material (Fig 28A, 28C and 28D). Similar forms are sometimes connected with butchering activities [90,104].
In the Eastern Saharan Africa the Acheulean appears to be stylistically diversified, judging by individual excavated sites all belonging to the late phases of the industry (~0,6-0,3 Ma), and despite a paucity of data [17,105]. Among those late assemblages, there are: sets of asymmetrical hand-axes from Egyptian oases at site E-72-1 [11]; a wide range of ovates (20% of the assemblage) and chopping tools from the Nubian Arkin 8 site in the area of Wadi Halfa [28], and the EDAR 7 assemblage described above, consisting of numerous cleavers and hand-axes prepared on flakes. The archaeological sequence at site 8-B-11 on Sai Island, which contains a Late Acheulean assemblage with fresh, large lanceolate hand-axes interstratified with horizons containing Sangoan hand-axes from the Middle Stone Age [5], makes this picture even more complex.
Wendorf, Close and Schild [106] pointed out that enormous geomorphic changes have taken place since the Acheulean. Writing in 1987 they stated that that no Acheulean living floors had ever been found and that the only geologically in situ artefacts recovered had come from the lower portions of almost completely deflated spring vents; none had ever been discovered "associated with ancient river deposits". By in situ, they surely meant the BS-14 site (trench 3), where a small assemblage of 237 artefacts embedded in spring sands below a calcium carbonate cap was excavated. These artefacts, which were made of brownish quarzitic sandstone, were not eroded but fresh and semi-fresh [12,107]. From 1987, with the exception of the Sai Island site commented below, not much had changed in the Eastern Saharan African Acheulean, until the discovery of the EDAR sites, where Acheulean assemblages have been found buried within a paleo-fluvial context. Few Acheulean assemblages from NE Africa have been reliably dated. Besides EDAR 7 and EDAR 135 in the Eastern Desert, there is only one Sudanese site, 8-B-11 on Sai Island in the Nile, where a layer of aeolian sand separates the Acheulean bed from the lowest Middle Stone Age deposits of Sangoan affinity [86,108,109]. OSL dating of this aeolian sand yielded an age of 223 ± 19 ka [5], providing a minimum age for the Acheulean on Sai Island. The Egyptian oases have yielded much more information regarding the chronology of the Acheulean [107,110]. In the Bir Sahara-Bir Tarfawi depression, episodes of lake expansion and regression suggest that episodically habitable landscapes existed in the region from approximately 400 ka (possibly as early as MIS 11) until the end of the Acheulean and into the MSA (up to MIS 3) [111][112][113]. An assemblage from Bir Sahara, BS-14 (E99C14), partly located in a primary context (trench 3), yielded an OSL age of 210 ± 18 ka which is considered to be too young (Romuald Schild, pers. com.). The Acheulean from Bir Sahara, although currently undated, is presumed to be older than 300 ka [112,114]. Several small Acheulean assemblages or singular bifaces were found in a stratigraphic contexts near Bir Safsaf, Bir Kiseiba and Wadi Arid [96,106]. Luminescence dating of the sediment just below the Acheulean horizon from site E-85-2 at Dag Dag Safsaf, Bir Safsaf, yielded ages of 308 ± 28 ka (GdTL-644) and 268 ± 26 ka (GdTL-645) [114].
The published ages for the oldest Acheulean from Eastern Sahara indicate human occupations during MIS 9, or possibly during MIS 11 as suggested by the depositional contexts of the Acheulean artefacts. The OSL dating of EDAR confirms both these assumptions. The OSL age from Unit IB (280 ± 27 ka), possibly indicates deposition of the underlying Acheulean levels during a warmer episode in MIS9 [115]. However, a sample from the same stratigraphic context, just above the top of the UNIT IA, from the site EDAR 135 (western wall) located~100 m from EDAR 7, gave an OSL age of 391 ± 30 ka, indicating that the Acheulean artefacts from UNIT IA could be older, i.e. MIS 11, MIS13 (500-400 ka) or earlier. This possibility is supported by the morphology of the artefacts, suggesting that they have been reworked into younger deposits, and the abundance of cleavers and the use of the Kombewa method in EDAR 7. Such developed Acheulean technologies of large flake production and high values of standardised made-on-largeflake cleavers and hand-axes first appear in Africa around 1.0 Ma continuing up to the Middle Pleistocene and are labelled as 'Large Flake Acheulean' stage [116][117][118][119]. This is not a characteristic feature of the other Acheulean Nubian sites dated to the~300-200 ka period, including some other Acheulean sites from EDAR-e.g. EDAR 133 [38] and EDAR 135 (lower cultural horizon) [32], therefore a wide possible chronology for EDAR 7 inventory could fall into the 'Large Flake Acheulean' phase. This phase could possibly be also represented by the Acheulean inventory from another EDAR neighbouring site-EDAR 6, where made-on-large-flake cleavers are common (S21 Fig) [120]. Corresponding assumptions were proposed by Haynes and colleagues [96], based on the technological features of a few Acheulean artefacts from KAS-1, whose subsurface inventory-however small-appears to be similar to EDAR 7 (presence of cleavers and Kombewa technology). This chronology would suggest that the stone tools found at EDAR 7 are older than those found at BS-14 [12], BT-A [121], KAS-1 [96], Site 047 [71], and KO10 [122]), making it the oldest Acheulean assemblage from the Eastern Saharan Africa dated to the early Middle Pleistocene or even to the turn of the Late Early and Middle Pleistocene. Chronological analogies to the EDAR area among Acheulean sites from the Arabian Peninsula are worth mentioning here as they demonstrate Acheulean presence there during MIS 7 [123,124], which is chronologically similar to the Acheulean horizon from EDAR 135 [32] but younger than site EDAR 7 discussed here.

Conclusions
This paper analyses Acheulean site EDAR 7 registered lately in the Sudanese Eastern Desert, where geoarchaeological methods identified this archaeological assemblage. Summing up, EDAR 7 has yielded a rare Eastern Saharan Acheulean assemblage, which has been dated and compared to other Acheulean materials from the region. The site has yielded new information on several aspects of Saharan Acheulean, including its chronology, environmental setting as well as behavioural and technological specificity of the lithic inventory. The chronology established for EDAR 7 points out that this is so far the oldest recognised Acheulean assemblage from the Eastern Sahara indicating its age as MIS 11-13 or earlier.
The EDAR 7 archaeological horizon is associated with the environment of the ephemeral braided stream and floodplain (UNIT IA). The buried channel system of the EDAR area is not connected with any present-day active river valley and represents an ancient river course not studied before. The advents of aridity during the interpluvial periods frequently caused drying up of the stream channels, bars or floodplains. It makes sense to presume that Homo erectus hominins occasionally occupied and exploited the gravel-beds of the dried stream or overbank areas of the stream, possibly for the purpose of procuring gravel stones. It is common to find Acheulean stone artefacts, made mainly from cobble-stones derived from the fluvial process contiguous to EDAR 7 in particular.
The archaeological record in the EDAR area, which is under constant threat from gold mining, is of considerable importance for understanding the earliest prehistory of the Eastern Sahara. Its scientific potential, highlighting the role of ancient Saharan watercourses has only been outlined. Further research in this region has the potential to shed more light on early migration routes out of Africa.