Ancient watercourses and biogeography of the Sahara explain the peopling of the desert

African Humid Period Biogeography and Palaeohydrology

Reanalysis of the Saharan zoogeography (SI Appendix, Section 1 and Table S1) suggests that many animals, including water-dependant creatures such as fish and amphibians, dispersed across the Sahara recently. For example, 25 North African animal species have a spatial distribution with population centers both north and south of the Sahara and small relict populations in central regions. This distribution suggests a trans-Saharan dispersal in the past, with subsequent local isolation of central Saharan populations during the more recent arid phase. If a diverse range of species (including fish) can cross the Sahara, it is impossible to envisage the Sahara functioning as barrier to hominin dispersal. The zoogeography of the Nile suggests that it was a much less effective corridor (SI Appendix, Table S1). Only nine animal species that occupy the Nile corridor today are also found both north and south of the Sahara. This paucity of Nileotic fauna highlights the importance of a green Sahara corridor and indicates that the Nile route is, potentially, less significant. A further 13 species of animal are not found in the Nile corridor but are found both north and south of the Sahara yet not in the center (SI Appendix, Table S1). Although they must have dispersed across the Sahara, whether they used the Nile or Green Sahara route cannot be determined without additional fossil evidence.

Fish species and molluscs are found both in oases across the Sahara and along the Nile, suggesting they crossed the Sahara using both the Green Sahara and Nile routes. The majority of these species are thought to only disperse via water (15) (SI Appendix, Section 2), thus current drainage only explains the Nile populations and another mechanism needs to be established to explain the present trans-Saharan distribution.

Mapping the palaeohydrology of the Sahara, using satellite imagery and digital topographic data, yields a tenable model (Fig. 1 and SI Appendix, Section 3). The Sahara was formerly covered by a dense palaeoriver network with many channels containing very large alluvial fans where rivers divide into multiple branches in an inland area. Where these fans are located on the boundary between two river catchments, their distributary channels can temporarily link adjacent river systems, thus allowing water-dependant life to transfer from one basin to the next (SI Appendix, Section 4). Five palaeofans identified in Fig. 1 link river systems in this manner. The green Sahara also contained numerous closed basins that supported some very large lakes (10, 16–19) (Fig. 1). When these lakes were full, they would have overflowed, thereby linking adjacent catchments (20). Fig. 1 shows that many fans and lake spillways are located in strategic positions that link waterways across the Sahara (SI Appendix, Section 4). Furthermore, tectonic activity can cause river diversion that transfers aquatic biota between basins, which appears to have been an important process in many northern basins (SI Appendix, Section 4). These mechanisms not only link many Saharan basins together but also connect them to large rivers with their headwaters outside the desert that can feed aquatic biota into the Sahara during humid periods. This interlinked waterway explains the observation that the fish species of the Sahara, Niger, Chad, and Nile Basins form a single biogeographic province (20) (Fig. 2A). The low degree of endemism indicates that these palaeohydrological interconnections happened relatively recently (15), perhaps during the early Holocene “climatic optimum” when the Sahara was last humid (21).

Freshwater mollusc biogeographic provinces (22) have a similar spatial distribution to that of the fish (SI Appendix, Section 5 and Fig. S5) suggesting that they dispersed across the Sahara using the same routes (SI Appendix, Section 2). That they have no endemic genera, and only 11 endemic species (22), again indicates a recent dispersal into the Sahara. In contrast, the divide between amphibian biogeographic provinces (SI Appendix, Figs. S7 and S8) is located roughly equidistant from the headwaters of the Nile, Chari, and Niger rivers and the Atlas Mountains (23). The latter presumably acted as refuges for these animals during arid periods. The contrasting spatial distribution of fish to amphibians is explained by the restriction of fish to rivers and lakes but the ability of amphibians to disperse both along watercourses and overland. Consequently, during early Holocene climatic amelioration, Palearctic amphibian species moved south, tropical species moved north, each crossing catchment divides that formed insurmountable barriers to fish, allowing direct colonization routes and thus their meeting in the center.

Although there is evidence that animals crossed the Sahara relatively recently, the timing and routes taken are uncertain. To constrain these two issues, maps of the spatial distribution of the recent (∼1980) or historical ranges (∼1900) of selected North African aquatic (20, 22, 24) and savanna (25) animal species, found within and north and south of the Sahara were combined with maps of reported historical sightings (24, 26), Holocene fossils (15, 27–33), and their depictions in rock art (26, 28, 34) (SI Appendix, Section 5). The evidence suggests that many species inhabited the Sahara during the early Holocene. This appears to be the case for hardy fish such as Tilapia zillii (Fig. 2A) and Clarias gariepinus (SI Appendix, Fig. S8), for the Nile crocodile (Crocodylus niloticus) (SI Appendix, Fig. S9), and for savanna species such as giraffe (Giraffa camelopardalis) (SI Appendix, Fig. S10) and African elephant (Loxadona africana) (SI Appendix, Fig. S11). These distributions suggest that, during the early Holocene, the Sahara was a savannah with riparian corridors linking north and south. Patches of more arid terrain in mountain rainshadows may better explain the arid pollen found in North Atlantic cores (8).

Some sub-Saharan animals do not exhibit trans-Saharan spatial distributions (Fig. 2B and SI Appendix, Figs. S12 and S13). They are found only in the southern Sahara, with a northern limit near the Ennedi, Tibesti, Tasilli, and Ahoggar Mountains, representing a catchment divide in the central Sahara (SI Appendix, Fig. S12). These species all have specialized aquatic requirements, such as deep water in the case of Nile perch (Lates niloticus) and hippopotamus (Hippopotamus amphibius) or long-term water connections for mollusc species such as Bellamya unicolor, Cleopatra bulimonides Pila and Lanistes (22) (SI Appendix, Section 2). The fan to the west of the Ahaggar Mountains (Fig. 1, fan 5) provides the only hydrological connection across this divide to the northern Sahara, but appears to have been a barrier to specialized aquatic species, accounting for an impoverished aquatic fauna in the northern Sahara.

The Peopling of the Sahara During the Holocene

We hypothesize that the differences in animal resources between the northern and southern Sahara during the early Holocene influenced the way it was peopled by humans. The north–south contrast in Saharan species ranges are remarkably similar to some key lithic, bone tool, and linguistic spatial distributions, suggesting that the peopling of the region during the early Holocene humid phase was driven by cultural adaptations that allowed exploitation of specific fauna.

The early Holocene archaeology of the Sahara is characterized by a regional distribution of specific archaeological cultures, such as those defined by barbed bone points, fishhooks, Ounanian arrow-points, and, more controversially, pottery (32, 35–38). The Sahara today is largely populated by speakers of Afroasiatic languages, Berber and Arabic, with some Nilo-Saharan languages (Teda-Daza and Zaghawa) in the region of Northern Chad, and Songhay cluster languages scattered across Mali and Niger (Fig. 3). However, it is clear that this situation is recent; Berber-speaking Tuareg moved into the Central Sahara ∼1500 y ago and the spread of the Hassaniya Moors into Mauritania probably dates from the 15th Century (39). Before this time, the central and southern Sahara are thought to have been populated by Nilo-Saharan speakers. The Nilo-Saharan language phylum is both widespread and strongly internally divided, suggesting considerable antiquity (40) (Fig. 3). Its greatest diversity is in the east, where a large number of small branches are found (Fig. 3), suggesting the original locus of expansion. Although fragmented into enclave populations today, the presence and pattern of relic populations in the northern desert points strongly to a much wider distribution in the past, covering the region from the Ethio-Sudan borderland to Mauritania and southwest Morocco.

It has long been suggested that Nilo-Saharan languages might correlate with barbed bone points, the so-called “Aqualithic” (35). Fig. 3 superimposes the sites of known barbed bone points on a map of current Nilo-Saharan languages, showing a remarkable similarity in spatial distribution, and also a notable correspondence with Holocene distribution of large aquatic species (e.g., Fig. 2A and SI Appendix, Fig. S12). It appears that the expansion of aquatic resources in the Holocene made the Sahara attractive to populations with existing fishing and riverine hunting skills (SI Appendix, Section 6). Their ability to hunt hippopotamus and crocodiles and to catch a wide variety of deepwater fish species would have propelled a rapid dispersal from east to west and into the central Sahara, to judge by the numerous branches of Nilo-Saharan in the east (Fig. 3 and SI Appendix, Section 6). Their movement further north would have been restricted by the absence of many of these species. However, the presence of an isolated Nilo-Saharan population (the Koranje, a branch of Songhay) and a barbed bone point in Northwest Africa near the headwaters of the catchment of the Soura River, the river that links the Atlas mountains to the lakes in the Ahnet-Mouydir basin (Fig. 1), forming a corridor from the central Sahara to Northwest Africa, indicates that a few groups may have traversed the green Sahara using the most promising routes. There is direct linguistic evidence that Nilo-Saharan populations exploited these aquatic resources in the form of a widespread cognate for “hippo” from Gumuz in Ethiopia, to Songhay in Mali (SI Appendix, Table S2). SI Appendix, Table S2 shows similar forms for “crocodile,” although in this case the cognates are split between eastern and western languages.

We hypothesize that the other economic revolution that occurred in the Sahara at approximately the same time was the southward spread of the bow and arrow. North African hunters would have observed the new abundance of large and unfamiliar land mammals to the south, notably elephant and giraffe. In a dispersal inverse to that of the Nilo-Saharans, they would have been attracted southward to hunt these animals with the bow and arrow. The “Ounanian” of Northern Mali, Southern Algeria, Niger, and central Egypt at ca. 10 ka is partly defined by a distinctive type of arrow point (37). These arrowheads are found in much of the northern Sahara (Fig. 3) and are generally considered to have spread from Northwest Africa. This view is supported by the affinity of this industry with the Epipalaeolithic that also appears to have colonized the Sahara from the north (41). No Ounanian points occur in West Africa before 10 ka, suggesting the movement of a technology across the desert from north to south around this time.

Our model envisages the initial Holocene repopulation of the Sahara being carried out by two separate populations practicing two quite different resource exploitation strategies: (i) aquatic foraging using bone point and fish hook technology, and (ii) savanna hunting using the bow and arrow. By linking the distribution of the Nilo-Saharan language phyla to the archaeological distribution of aquatic- and terrestrial-adapted technologies, we explain the pattern of human repopulation of the desert in terms of the changing faunal distribution, which is in turn dictated by the nature of trans-Saharan hydrological linkages.

Older Saharan Occupation and Crossings

The movement of people across the Sahara during the Holocene may well have been the last of several important hominin dispersals. To demonstrate the possibility of earlier trans-Saharan migrations, it is necessary to determine the timing of pre-Holocene humid phases in the Sahara. This chronological evidence can be obtained by dating the deposition of pre-Holocene lake sediments and demonstrating synchronous palaeohydrologic changes across the Sahara. MIS5 contains the most abundantly dated pre-Holocene humid sediment sequences, yet only two speleothems, two regions containing spring tufas, and nine separate palaeolakes have been dated to this period (SI Appendix, Sections 4 and 7), thus there are many gaps and it is not possible to say if the entire Sahara was humid. For example, before this study, there was a complete lack of last MIS5 ages in the southernmost part of the Sahara. One way to overcome this problem is to date lacustrine episodes in the Saharan megalakes because they are located within abutting catchments and these catchments span the desert (Fig. 1). Because the high stands indicate significantly increased moisture availability, fluvial systems within the megalake catchments must be active at these times. Consequently, synchronous lake high stands would yield a humid corridor across the Sahara, allowing northward dispersal of hominin populations located in sub-Saharan Africa.

To evaluate the timing of pre-Holocene humid phases, it is necessary to locate and date extensive sequences of lacustrine sediments in megalake basins. Fortunately, Saharan megalakes appear to preserve the required long-term lacustrine sequences (16–19, 42) despite the general assumption that such records are rare in the Sahara because of deflation of lacustrine sediments during arid periods (43). We have applied optically stimulated luminescence (OSL) dating to selected sediments (SI Appendix, Section  7) and have found evidence for synchronous humidity in the Fezzan–Chad–Chotts and Chad–Chotts–Ahnet-Moyer megalake corridors during MIS5 (Fig. 1). Humidity in the Chad Basin was investigated by dating Lake Megachad beach ridges (SI Appendix, Fig. S16). The Bama Ridge in northeastern Nigeria forms the southwestern shoreline of the Holocene palaeolake Megachad with an area of 361,000 km² (10). To the southwest of the Bama Ridge, we identified a succession of older beach ridges that run roughly parallel to it (SI Appendix, Fig. S16). OSL ages for two of these ridges at Alhajari and Kawiya give ages of 114 ± 14 ka and 125 ± 12 ka, respectively, suggesting megalake conditions in MIS5. In the Fezzan Basin, OSL dating (SI Appendix, Section 7; ref. 16) of palaeolake sediments shows that two large lakes of 1,350 and 1,730 km² existed in this basin at a similar time (SI Appendix, Fig. S16). The Fezzan Basin ages are consistent with a U/Th isochron age for lake sediments from the basin of the Chotts (98 ± 5 ka; ref. 18) and the Ahnet-Mouydir basin [92(+20 - 18) ka; ref. 19). At this time, the Chotts basin contained a lake with an area of 30,000 km² while the Ahnet-Mouydir basin lake attained an area of at least 32,000 km², thus suggesting two green corridors across the Sahara during MIS5 that could have been used by hominins to cross the Sahara (Figs. 1 and 4). Once hominins had achieved a trans-Saharan distribution, it would have been possible to pass into the Levant via the Mediterranean coast. Because the Levant was humid at this time (44), onward dispersal would have been possible.

It is probable that increased humidity within the megalake catchments is indicative of regional scale humidity and hence dispersal routes may have existed throughout the desert, as appears to have been the case during the early Holocene, rather than being restricted to corridors. Other studies of Saharan fluvial and lacustrine sediments provide some evidence for humidity in other areas. A humid corridor has recently been proposed in the Sahabi palaeoriver system (13) (Fig. 4) and there are 41 other ages for 12 areas that contain humid sediments scattered throughout the Sahara that are indistinguishable from these “corridor” ages. Their integration with the palaeohydrological map provides evidence for a wider green Sahara during MIS5 as well as the present interglacial (Fig. 4 and SI Appendix, Section 3).

Using the Holocene biogeography and palaeohydrology of the Sahara as an analogue for the MIS5 humid period, it is likely that an interconnected waterway would have been available for faunal and human dispersal. This humid period corresponds very closely with the age of the first modern human occupation of the North African coast (45) and the Levant (46) by sub-Saharan populations, who may have been crossing the Sahara at this time (9). The occupation of the Mediterranean coast of Africa by these early modern human migrants appears to have lasted from ∼110 to ∼30 ka (45), though the Levantine occupation appears to have finished by ∼70 ka (47). Some view the out-of-Africa dispersal into the Levant as the start of the spread of modern humans onward into Arabia and India in MIS5 (48), whereas others believe it to be a “dead end” that was followed by a later more successful dispersal of modern humans out of Africa at a later date: 60 ka in MIS4 (49). There is little evidence for a green Sahara in MIS4, so the desert is unlikely to have directly played a role in this postulated second dispersal, however, the modern humans north of the Sahara could have provided the population required to drive it. Indeed, “spatially” this hypothesis is attractive because the Sinai corridor out of North Africa is close to the location of the earliest archaeological sites associated with this dispersal, which are found in the Levant at ∼47 ka (50). However, an alternative dispersal route for modern humans out of Africa via the Bab el Mandab has been proposed at about this time, largely on genetic grounds (51), thus if this dispersal occurred, a route out of Africa via North Africa is not necessary. To further complicate matters, the possibility of multiple dispersals of modern humans both into and out of Africa has recently been suggested (52). If these proposed migrations did indeed happen, it is feasible that these different dispersals employed different routes.

Notwithstanding the as yet unresolved issue of the timing and route taken by modern humans out of Africa, both the use of the Green Sahara route to the North African coast by hominins and its long-term viability are supported by archaeological evidence (53–56). Stone tools from Oldowan to Neolithic have been found in all the megalake basins that form corridors across the Sahara, suggesting that the region was not only periodically habitable for hominins, but that it was also regularly occupied by them.

Supplementary Material