Regional geomorphology and palaeoenvironments
To evaluate the terminal Pleistocene and early Holocene environments of the region, trenches were excavated at four playas in the vicinity of the archaeological sites, two of which contained sufficient sediment for palaeoenvironmental analysis and luminescence dating: ARN (Site 1) and JMI (Site 4), (Fig. 1; Supplementary Note 5 and Supplementary Fig. 42). As local centres of deposition, and areas where water accumulated in the past, these playas provide archives of hydroclimate. Gravels found at the base of both 2 m deep trenches were poorly sorted, containing pebbles of up to 5 cm, and were interpreted as alluvial fan deposits. A luminescence age of 68.8 ± 5.0 ka (MIS-B-1) from JMI indicates that these sediments were deposited at the Marine Isotope Stage 5a to 4 transition (Table 1). Palaeohydrological activation at this time is consistent with both broader regional records28 and a surface find of a Middle Palaeolithic Levallois core from JMI (Supplementary Fig. 32).
Table 1 Luminescence ages from archaeological sites and playa deposits Full size table
At both excavated paleoenvironmental sites, the overlying sediments consist of a thin interval of alternating much finer-grained and well-sorted aeolian sands and playa deposits. These are followed by a prolonged sequence of playa deposits consisting of quartz sand, clays (illite and kaolinite), and calcite, indicating more humid conditions with increased water accumulation in these local depressions. We interpret the onset of playa sediment accretion as representing the change in the balance between aeolian erosion and fluvial sedimentation. During the LGM, the hyper-arid environment meant that any fine-grained sediments deposited in ephemeral floods were subsequently eroded by the wind. As the climate became less arid, fluvial sedimentation increased, becoming greater than the aeolian erosion rate and sediments started to accumulate. However, the subordinate carbonate concentration and the absence of root traces and organic-rich layers suggest that conditions remained too dry for the establishment of more permanent water bodies, indicating that a dryland environment persisted around these ephemeral, and probably seasonal, lakes. Luminescence ages constrain the onset of surface water to between 17.1 ± 1.7 ka (MIS-B-3) and 15.5 ± 1.5 ka (MIS-B-2) at JMI (Site 4) and to 12.7 ± 2.1 ka (SAH-L3-1) at ARN (Site 1) (Table 1). The onset of sedimentation after a long hiatus, coupled with an increase in calcite noted in both sedimentary records, indicates that local conditions gradually became more humid after ~16 ka at JMI to ~13 ka at ARN, with these site-specific differences in timing likely due to differences in catchment hydrology (Table 1). These sites thus represent the earliest evidence from northern Arabia of increased humidity following the hyper-arid LGM. However, the lakes were ephemeral, indicating an arid or semi-arid climate prevailed in the region.
Monumental rock art
Surveys at ARN and JMI identified previously unknown rock art landscapes with life-sized depictions of wild mammals and human figures, and an individual panel at MLH. Across the three areas 62 rock art panels were recorded, containing 176 engravings. Of these, 130 were life-sized and naturalistic engravings depicting camels (90), ibex (17), equids (15), gazelles (7), and aurochs (1), with individual representations frequently measuring up to 2.5–3.0 m in length and 1.8–2.2 m in height. In addition, we identified 2 camel footprints, 15 smaller scale naturalistic depictions of camels, 19 human figures, 4 human faces or masks, and 6 unidentified, partial engravings (Supplementary Data 1). Most of the recorded engravings show camels in a detailed and naturalistic style that echoes the reliefs of the Camel Site to the north of the Nefud desert. This includes the frequent depiction of a bulging neckline, indicating they represent male camels in rut29,30 (Fig. 2A, B).
Fig. 2: Rock art panels at Jebel Arnaan. Tracings highlight the layering of engravings, showing phase 1 in green, phase 2 in yellow, phase 3 in white and phase 4 in shades of blue. Rock art scale is 10 cm wide. A Panel ARN21A: Several small, stylised depictions of women (phase 1, traced in green), superimposed with a large human figure (phase 2, traced in yellow). Large, kneeling camel engraved over the top (phase 3, traced in white). B Panel ARN22A : naturalistic camel (phase 3, traced in white), superimposed by a camel with stylised, rounded eye and standardised outline (phase 4, traced in blue); original and unfinished camel head (traced in dark blue); three gazelle engravings were added during a later part of phase 4 (traced in light blue). Additional examples are provided in Supplementary Fig. 4. C Panel ARN3B : ibex with cartoon-like eye and horn (phase 4, traced in blue), superimposed with life-sized aurochs (phase 4, traced in lighter blue). D Panel ARN 39 : equid with cartoon-like eye, and with a young (phase 4, traced in blue). Untraced photos are provided in Supplementary Figs. 2 and 3. Full size image
The depictions span multiple engraving phases, with images often overlapping on rock surfaces. Sometimes this was done to update an existing representation (Fig. 2B) or to depict a different animal species (Fig. 2C). We distinguish four phases here. Two early rock art phases: small, stylised depictions of women (phase 1, traced in green), followed by large human figures in frontal view (phase 2, traced in yellow). These human figures were always noted to be older than, i.e. underneath, the recorded life-sized animal representations (Fig. 2A), and they make up a much smaller proportion of motifs. The third phase shows detailed, extremely naturalistic representations of animals, where each depiction has individual characteristics (traced in white) (Fig. 2A, B). A later, fourth phase (traced in blue) shows more stylised depictions of animals with cartoonish features, including rounded eyes and horn ridges, and more standardised, near-identical depictions of animals (Figs. 2C, D and 3C).
Fig. 3: Monumental rock art panel at Jebel Misma (JMI18). A Location of the panels at 34 m and 39 m height, and trenches JMI7 and JMI8 (project vehicle for scale). An OSL date of 12.0 ± 1.8 was obtained from JMI8 (Table 1: JMI8-T1-4). Excavated playa deposits are located directly behind this spur (see also Figs. 1 and 8). B narrow, downward sloping ledges in front of the panel viewed from above. C orthophoto generated via a high-resolution 3D model of the panel, with tracings showing 19 life-sized camels and 3 equids (one further camel was documented on a collapsed fragment left of the image). Naturalistic animals belonging to phase 3 traced in white. More stylised and standardised depictions of phase 4 traced in blue, including two engravings of equids traced in dark blue, and superimposed stylised camels traced in light blue. Unidentified lines traced in black. White traced camels: 1.7–1.9 m length, blue camels: 2.15–2.6 m length. A human figure was added on the far left for scale (1.7 m). Full size image
Unlike the SAU site, where many of the engravings were found inside narrow gaps between boulders27, engravings at ARN and JMI were found on prominent locations on boulders or cliff surfaces, facing into the landscape (Supplementary Figs. 1 and 7). Some of these panels were etched onto cliff surfaces in inaccessible but highly visible locations (Fig. 3). The difficulty in getting to and engraving these rock surfaces, and their enhanced visibility by height were clearly attractive for the engravers. The precarious nature of the engraving process is particularly evident in the largest recorded panel. On the ground, panel JMI18 is today only visible in optimal light conditions for about 1.5 h in the morning, due to its elevated location and the varnish build-up on the engravings. This panel would have been accessed by climbing up a cliff and then engraved while standing on a downward sloping ledge, only ~30–50 cm in width (Fig. 3B). Today the sandstone is too degraded to reach the ledge safely, and the panel was documented using a drone. The friable nature of the substrate and the slope of the narrow ledges suggest the engravers likely risked their lives to create this art. Engraving at close range would have required them to use direct percussion, while also preventing them from being able to see the complete image. Twenty-three life-sized camels and equids, each with an individual length of ~1.7–2.6 m were engraved on this surface (Fig. 3C), with the engravings stretching ~23 m across two cliff surfaces at a height of 34 m and 39 m (Fig. 3A), giving this rock art a monumental scale.
All recorded rock art panels show a thick coating of dark rock varnish on the natural sandstone surface and inside the engraved lines (Fig. 2), although on some panels this varnish has partially eroded and is now only visible in areas that are less exposed (Fig. 3C and Supplementary Fig. 3C, D). The re-formation of rock varnish following the exposure of the fresh sandstone during the engraving process has been shown to take over 8 ka31,32, and provides a first indication of the antiquity of these images.
The appearance of multiple life-sized animal engravings is impressive today. Freshly engraved against the varnish, the images would have had considerable visual impact. The durability of the images may have facilitated the remembrance of meaning and symbolism across generations of people using these sites. The large naturalistic engravings, therefore, align with the definition of monumentality, which references great size and effort, but also longevity and remembrance within a community33,34.
Archaeological Excavations
Four trenches revealed stratified archaeological deposits. Two were excavated at ARN (ARN3, T1 and T2), and two at JMI (JMI7 and JMI8, Fig. 1), yielding artefacts including over 1200 lithics (Supplementary Note 2), and 16 bone fragments (Supplementary Note 4). Optically stimulated luminescence (OSL) and radiocarbon dating (Supplementary Note 6) samples from the archaeological deposits attest to human occupation during the terminal Pleistocene and earliest Holocene.
In front of panel ARN3, a 2 × 1 m trench (T1) was excavated directly below two life-sized engravings of camels, one superimposed on the other (Fig. 4). Below recently disturbed sediments, the excavation revealed intact archaeological deposits with layer 5 in particular containing dozens of flaked stone artefacts. This layer was sealed by several large sandstone fragments in layer 4 (Fig. 4), likely the result of a colluvial event. In ARN3 T1, the legs of the engraved camels were buried at a depth indicating that they were engraved no later than when layer 4 was deposited (Fig. 4). Layer 5 included a stone tool that may have been used for engraving the rock art (discussed below) (Fig. 4, orange). Two luminescence samples obtained from layer 5, dating to 12.2 ± 1.4 ka, and 12.8 ± 1.1 ka (Table 1), directly date the burial of the artefacts and pecking tool, and provide an indirect date for the engraving above. Analysis of the engraved lines shows that all images were made using a pecking stone. In some engravings, pecked lines were smoothed in a second step (Fig. 4 inset).
Fig. 4: Excavations at ARN3 Trench 1. South-western and north-western sections, highlighting the extent of the disturbance: original height of deposit (pink dotted line), and extent of disturbed sand (pink line). Colour grading of layers reflects artefact density. Location of OSL samples shown in black, location of engraving tool projected onto section from a plotted position 72 cm south-east. Earlier camel engraving (phase 3) traced in grey, later camel engraving (phase 4) traced in blue, dashed lines indicated parts of the engraving that were covered with sediment. Note that the legs of the earlier (grey) camel have partially eroded and are no longer visible. Inset shows a closeup of the engraved lines, the crudely pecked line of the earlier (grey) camel on the upper right, and the pecked and smoothed lines of the nose of the later (blue) camel on the lower left. Remains of dark rock varnish29,31,47 can be seen in several patches on both lines. Full size image
A second trench (T2) was excavated at ARN3 1.2 m to the east, beneath two more camel engravings at the other end of the same boulder surface (Supplementary Fig. 11). T2 also revealed layers containing hundreds of lithics (layers 6–8; Supplementary Fig. 13) with three small hearths in layer 7, one of which was radiocarbon dated to 11.44 ± 0.18 ka (cal. BPUGAMS65278, Table 2). Refits from layers 8 and 9 in T2 testify to the high integrity of the archaeological contexts (Supplementary Note 3).
Table 2 Radiocarbon date from excavations at ARN Full size table
At JMI, two trenches were excavated beneath boulders featuring rock art panels. At JMI7 a 2 × 1 m trench was excavated beneath life-sized engravings of two camels and an equid, which revealed a single layer with high artefact concentrations (layer 4; Supplementary Fig. 16). At JMI8 a 2 × 1 m trench was excavated below a life-sized engraved camel. Here moderate artefact concentrations were interspersed throughout the stratigraphic sequence, with the highest artefact concentration in layer 2, near the surface (Supplementary Fig. 19). A luminescence sample was obtained from layer 6, which provided an age of 12.0 ± 1.8 ka (JMI8-T1-4, Table 1), consistent with the luminescence and 14C ages at ARN.
Artefact Assemblages
The ARN3 excavations revealed a rich assemblage of occupation debris. In T1, 101 lithics were recovered, with over half in layer 5 (Fig. 4), while in T2, 532 lithics were recovered, with over 95% from layers 5–9 (Supplementary Fig. 13). Chert was the main material used, constituting 36% of the pieces, with obsidian and crystal quartz making up another 23% and 15% respectively (Supplementary Data 2). Lithics showed a diversity of reduction strategies, including small bipolar cores occurring alongside chert bladelets/blades with lamellar scar patterns and often ground platforms (Supplementary Note 3.1).
A total of 30 retouched pieces were recovered. Among these, notches were common with eight notched pieces and two opposed notch blades (Supplementary Table 1). Two chert drills were recovered in T2, as well as a single endscraper from the disturbed upper part of T1. The largest retouched artefacts were two tongue-shaped scrapers from T2 layer 8 (Supplementary Fig. 28). T2 produced seven marginally retouched convergent points made on chert bladelets, while the tanged butt of a point on an obsidian blade was recovered from T1 (Supplementary Figs. 25 and 26). The most distinctive artefacts were a single crystal quartz Helwan bladelet from T1 layer 4 (Fig. 5B), and three chert El Khiam points, a broken butt from T1 layer 5, and both a broken butt and a complete El Khiam point from T2 layer 5 (Fig. 5A and Supplementary Fig. 26). The luminescence age of 12.2 ± 1.4 ka (ARN-T1-2, Table 1) in T1 and the radiocarbon date of 11.44 ± 0.18 ka (UGAMS65278, Table 2) from T2 correspond with the age of El Khiam points from the Levant and therefore indicate that the main occupation was contemporary with the PPNA.
Fig. 5: Stone tools and beads recovered from excavations. A El Khiam point from ARN3; B Helwan bladelet from ARN3; C–E Ground stone disc beads from the ARN3 T2; F Dentalium bead from ARN3 T2. Photos by Antonio Reiss. Full size image
Excavations at JMI7 produced 100 lithics while JMI8 produced 519 (Supplementary Figs. 16 and 19). Quartz was the most commonly used material comprising 65% of pieces, with an additional 20% made from silcrete (Supplementary Data 3). The single core was a silcrete naviform piece from layer 4 in JMI7 (Supplementary Fig. 29). The JMI7 and JMI8 excavations produced nine retouched artefacts (Supplementary Table 2). An opposed-notch chert blade was recovered in layer 5 (Supplementary Fig. 31), with two other chert notched blades recovered from JMI8 layer 4, but these did not have the distinctive opposing notches. This same layer also produced a chert drill with a distal break. A broken Helwan point was found on the surface between JMI7 and JMI8 (Supplementary Fig. 30).
Several of the recovered artefacts have associations with the PPN and even the Natufian in the Levant. Helwan bladelets (Fig. 5B) and opposed notch blades are known from Natufian sites; El Khiam points (Fig. 5A) are an artefact type characteristic of the PPNA, whereas Helwan points and naviform cores are typical for the PPNB (Supplementary Note 3). These artefact types accord with the chronometric dating at ARN and JMI (Tables 1 and 2).
Four grinding stones were recovered from the ARN3 excavations and six from the JMI excavations, along with a facetted stone ball (Supplementary Fig. 34) and two stone platters in JMI7 layer 4 (Supplementary Fig. 35). Two mortars were documented on an exposed rock surface between JMI7 and JMI8 (Supplementary Fig. 33). Mortars are typical of the PPNA35 and have not previously been documented among grindstone assemblages of Neolithic northern Arabia36.
A total of five ground stone beads were recovered, four green pieces from ARN T2 layers 7 and 9, and one from JMI8 layer 2. The recovery of two broken beads and one unfinished bead, together with drills at both sites, suggests on-site manufacture. Two marine dentalium (tooth-shell) beads were recovered from ARN3 T2, layers 7 and 8 (Fig. 5F and Supplementary Fig. 41). Among mostly red pigment, noteworthy was a crayon of green copper ore pigment from ARN3 T2 layer 7. Green pigment and ground stone beads are characteristic of the earlier PPN37,38. Dentalium shells are known from the PPNA39 and had to be collected from the Mediterranean Sea or the Red Sea, making the latter the closest potential source at over 320 km away.
Three tools with clear battering marks were recovered from directly below engraved panels. One is from a stratified and dated archaeological context underneath a life-sized camel engraving at ARN3 T1 (Figs. 4 and 6), immediately below OSL sample ARN-T1-2, dated to 12.2 ± 1.4 ka, and above sample ARN-T1-3 dated 12.8 ± 1.1 ka (Table 1). Two additional tools were recovered on the surface beneath panels ARN30 and JMI8 (Supplementary Fig. 37). The ARN tools are made on clasts of ferruginous sandstone, a tough material not immediately available at the rock art localities, while the JMI8 tool is made of silcrete, a material available at SAU. The tools are all wedge shaped, and the two made on ferruginous sandstone were flaked to sharpen the pecking edge. The tools comfortably fit into a medium-sized adult hand, with the battered edges protruding (Fig. 6B, C). Battering is evident on multiple edges of all three tools, indicating they were used extensively. Due to their angular form, these tools are not viable as lithic hammerstones in the arced free-hand percussion motion; however, the tapered ends would make good surfaces for the direct end-on percussion required to peck petroglyphs (Supplementary Note 3.3). The tool recovered from ARN3 T1 was used to peck and could thus have been involved in the production of either of the two camel engravings above (Fig. 4). In Arabia, this type of tool was first noted at the Camel Site, where traceological analysis on silcrete pieces showed damage consistent with sandstone engraving40. Comparable tools and patterns of wear have also been observed at petroglyph sites in Europe41 and South America42,43.