Ancient watercourses and biogeographyof the Sahara explain the peopling of the desertNick A. Drakea,1, Roger M. Blenchb, Simon J. Armitagec, Charlie S. Bristowd, and Kevin H. Whitee

aDepartment of Geography, King’s College London, Strand, London WC2R 2LS, United Kingdom; bKay Williamson Educational Foundation, 8 Guest Road,Cambridge CB1 2AL, United Kingdom; cDepartment of Geography, Royal Holloway, University of London, Egham, Surrey TW20 0EX, United Kingdom;dSchool of Earth Sciences, Birkbeck College, University of London, Malet Street, London WC1E 7HX, United Kingdom; and eDepartment of Geography,University of Reading, Whiteknights, Reading RG6 6AB, United Kingdom

Edited by Ofer Bar-Yosef, Harvard University, Cambridge, MA, and approved November 22, 2010 (received for review August 23, 2010)

Evidence increasingly suggests that sub-Saharan Africa is at thecenter of human evolution and understanding routes of dispersal“out of Africa” is thus becoming increasingly important. The SaharaDesert is considered by many to be an obstacle to these dispersalsand a Nile corridor route has been proposed to cross it. Here weprovide evidence that the Sahara was not an effective barrierand indicate how both animals and humans populated it duringpast humid phases. Analysis of the zoogeography of the Saharashows that more animals crossed via this route than used the Nilecorridor. Furthermore, many of these species are aquatic. This dis-persal was possible because during the Holocene humid period theregion contained a series of linked lakes, rivers, and inland deltascomprising a large interlinkedwaterway, channelingwater and an-imals into and across the Sahara, thus facilitating these dispersals.This system was last active in the early Holocene when many spe-cies appear to have occupied the entire Sahara. However, speciesthat require deep water did not reach northern regions because ofweak hydrological connections. Human dispersals were influencedby this distribution; Nilo-Saharan speakers hunting aquatic faunawith barbed bone points occupied the southern Sahara, while peo-ple hunting Savannah faunawith the bow and arrow spread south-ward. The dating of lacustrine sediments show that the “greenSahara” also existed during the last interglacial (∼125 ka) and pro-vided green corridors that could have formed dispersal routes at alikely time for the migration of modern humans out of Africa.

biogeography ∣ human migration ∣ animal migration ∣ palaeoclimate ∣Saharan palaeohydrology

Traditionally, it has been assumed that the Nile was the primarycorridor for the dispersal of hominins across the Sahara desert

and then out of Africa (1). However, this route has proven hard tosubstantiate for three reasons. First, considerable research effortshave failed to uncover evidence for its consistent use and themovement of its inhabitants into the Levant (2). Second, thechronology of the evolution of the modern Nile drainage, whichis essential to create a functional corridor, is controversial (3, 4).Finally, although many ancient fluvial deposits have been recog-nized in the Nile Valley, some with lithics that confirm the pre-sence of hominins, their stratigraphic relationships and ages arepoorly understood (5).

The need for a Nile migration corridor arises from the viewthat a belt of aridity has always separated the more humid regionsnorth and south of the Sahara, yet the evidence for this aridbarrier is contradictory. For example, genetics indicates thatthe Sahara has been a barrier to the dispersal of Macroscelidea(elephant shrews) since the late Miocene when the desert firstformed (6), whereas pollen analysis of terrestrial sediments fromthe central Sahara demonstrates that during the Holocene humidperiod an arid belt existed north of 22° (7). Furthermore, aridplant taxa are found continuously through Atlantic Ocean coresbetween 23° and 27° north for the last 250 ka suggesting a long-lasting Saharan arid belt at this latitude (8). In contrast, a greenSahara route has been proposed whereby this currently hyperarid

region could have been habitable during “green” periods of great-er humidity (9,10). This humid period was first suggested byDuveyrier (11) to explain the existence of the Nile crocodilesin isolated Saharan oases and has periodically been promotedsince, notably by Dumont (12) who noted the trans-Saharan dis-tribution of numerous aquatic animal species. Recent evidencefor this view is provided by isotopic data from a Mediterraneanmarine core and aquatic snails that suggest a river corridor acrossthe Sahara during marine isotope stage (MIS) 5e (13) and varia-tions in dust concentrations in a North Atlantic marine core thatindicates humid conditions during the last two interglacials (14).

The idea of the Sahara as a barrier to faunal dispersal isbecoming increasingly important because of its role in the out-of-Africa debate. Here we use numerous lines of evidence to eval-uate the barrier hypothesis, and use this evidence to proposemodels for the peopling of the Sahara during past humid phases.In the next section, we map the palaeohydrology of the Saharaand show that during humid periods it contains an extensive in-terconnected hydrological system that animals appear to haveused to migrate across the green Sahara. The following sectioncompares the spatial distribution of archaeology and languagesto that of the fauna and palaeohydrology to develop theoriesof the peopling of the green Sahara during its last humid phase.Finally, we evaluate dates of palaeolake sediments from theprevious humid phase and show that it corresponds with the ageof the first permanent modern human occupation of the NorthAfrican coast and the Levant, suggesting a crossing of the Saharaat this time.

African Humid Period Biogeography and PalaeohydrologyReanalysis of the Saharan zoogeography (SI Appendix, Section 1and Table S1) suggests that many animals, including water-depen-dant creatures such as fish and amphibians, dispersed across theSahara recently. For example, 25 North African animal specieshave a spatial distribution with population centers both north andsouth of the Sahara and small relict populations in central re-gions. This distribution suggests a trans-Saharan dispersal in thepast, with subsequent local isolation of central Saharan popula-tions during the more recent arid phase. If a diverse range of spe-cies (including fish) can cross the Sahara, it is impossible toenvisage the Sahara functioning as barrier to hominin dispersal.The zoogeography of the Nile suggests that it was a much lesseffective corridor (SI Appendix, Table S1). Only nine animal spe-cies that occupy the Nile corridor today are also found both northand south of the Sahara. This paucity of Nileotic fauna highlights

Author contributions: N.A.D. and R.M.B. designed research; N.A.D., R.M.B., S.J.A., C.S.B.,and K.H.W. performed research; N.A.D., R.M.B., and S.J.A. analyzed data; and N.A.D. andR.M.B. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. E-mail: nick.drake@kcl.ac.uk.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1012231108/-/DCSupplemental.

458–462 ∣ PNAS ∣ January 11, 2011 ∣ vol. 108 ∣ no. 2 www.pnas.org/cgi/doi/10.1073/pnas.1012231108

Dow

nloa

ded

by g

uest

on

June

3, 2

020

the importance of a green Sahara corridor and indicates thatthe Nile route is, potentially, less significant. A further 13 speciesof animal are not found in the Nile corridor but are found bothnorth 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 de-termined without additional fossil evidence.

Fish species and molluscs are found both in oases across theSahara and along the Nile, suggesting they crossed the Saharausing both the Green Sahara and Nile routes. The majority ofthese species are thought to only disperse via water (15) (SIAppendix, Section 2), thus current drainage only explains the Nilepopulations and another mechanism needs to be established toexplain the present trans-Saharan distribution.

Mapping the palaeohydrology of the Sahara, using satelliteimagery and digital topographic data, yields a tenable model(Fig. 1 and SI Appendix, Section 3). The Sahara was formerlycovered by a dense palaeoriver network with many channels con-taining very large alluvial fans where rivers divide into multiplebranches in an inland area. Where these fans are located on theboundary between two river catchments, their distributary chan-nels can temporarily link adjacent river systems, thus allowingwater-dependant life to transfer from one basin to the next(SI Appendix, Section 4). Five palaeofans identified in Fig. 1 linkriver systems in this manner. The green Sahara also containednumerous closed basins that supported some very large lakes(10, 16–19) (Fig. 1). When these lakes were full, they would haveoverflowed, thereby linking adjacent catchments (20). Fig. 1shows that many fans and lake spillways are located in strategicpositions that link waterways across the Sahara (SI Appendix,Section 4). Furthermore, tectonic activity can cause river diver-sion that transfers aquatic biota between basins, which appearsto have been an important process in many northern basins(SI Appendix, Section 4). These mechanisms not only link manySaharan basins together but also connect them to large rivers withtheir headwaters outside the desert that can feed aquatic biotainto the Sahara during humid periods. This interlinked waterwayexplains 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 thesepalaeohydrological 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 simi-

lar spatial distribution to that of the fish (SI Appendix, Section 5and Fig. S5) suggesting that they dispersed across the Saharausing the same routes (SI Appendix, Section 2). That they haveno endemic genera, and only 11 endemic species (22), againindicates a recent dispersal into the Sahara. In contrast, thedivide between amphibian biogeographic provinces (SI Appendix,Figs. S7 and S8) is located roughly equidistant from the head-waters of the Nile, Chari, and Niger rivers and the Atlas Moun-tains (23). The latter presumably acted as refuges for theseanimals during arid periods. The contrasting spatial distributionof fish to amphibians is explained by the restriction of fish to riv-

Fig. 1. Late Pleistocene and early Holocene palaeohydrology of the Sahara(∼11 to 8 ka). The catchments of the megalakes that form corridors across theSahara are indicated with letters A–D, while the river catchments that linkwith many of these megalake basins to form the vast Saharan inland water-way are indicated with letters E–I: (A) Lake Megachad, (B) Lake Megafezzan,(C) Ahnet-Mouydir Megalake, (D) the Basin of the Chotts, (E) Senegal River,(F) Nile River, (G) Sahabi River, (H) Kufra River, and (I) Niger River. Numbersindicate the location of fan, tectonic and lake outflow links between basins:(1) Nile Basin/Chad Basin, (2) Chad Basin/Niger Basin, (3) Niger Basin/SenegalBasin, (4) Niger/Basin West Ahaggar Mountains, (5) West Ahaggar Moun-tains/Ahnet-Mouydir Basin, (6) Ahnet-Mouydir Basin/Chotts Basin, (7) ChottsBasin/Fezzan Basin, (8) Fezzan Basin/Serir Tibesti, and (9) Serir Tibesti/KufraBasin. The 200-mm isohyet indicates the current limit of the Sahara Desert.

Fig. 2. Late Pleistocene and early Holocene palaeohydrology and biogeo-graphy of the Sahara (∼11 to 8 ka). (A) Biogeographic provinces for theAfrican fish species are indicated (1 Sudanian, 2 Upper Guinean, 3 Eburneo-Guinean, 4 Lower Guinean, 5 Congo, 6 Maghrebian) along with the distribu-tion of Tilapia zillii, both recently, as indicated by the hatched area and frommarking the location of Saharan refuges, and during the Holocene, as indi-cated by fossils and rock art. A trans-Sahara distribution is evident, bothacross the Sahara and down the Nile. (B) Distribution of Hippopotamusamphibious both historically (∼1 ka) as shown by the hatched area and dur-ing the Holocene by marking the location of older historical sightings, fossils,and rock art. A distribution restricted to the River Nile and the southern-central Sahara is evident.

Drake et al. PNAS ∣ January 11, 2011 ∣ vol. 108 ∣ no. 2 ∣ 459

GEO

LOGY

Dow

nloa

ded

by g

uest

on

June

3, 2

020

ers and lakes but the ability of amphibians to disperse both alongwatercourses and overland. Consequently, during early Holoceneclimatic amelioration, Palearctic amphibian species moved south,tropical species moved north, each crossing catchment dividesthat formed insurmountable barriers to fish, allowing directcolonization routes and thus their meeting in the center.

Although there is evidence that animals crossed the Sahararelatively recently, the timing and routes taken are uncertain.To constrain these two issues, maps of the spatial distributionof the recent (∼1980) or historical ranges (∼1900) of selectedNorth African aquatic (20, 22, 24) and savanna (25) animalspecies, found within and north and south of the Sahara werecombined 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 thatmany 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 theNile crocodile (Crocodylus niloticus) (SI Appendix, Fig. S9), andfor savanna species such as giraffe (Giraffa camelopardalis)(SI Appendix, Fig. S10) and African elephant (Loxadona africana)(SI Appendix, Fig. S11). These distributions suggest that, duringthe early Holocene, the Sahara was a savannah with ripariancorridors linking north and south. Patches of more arid terrainin mountain rainshadows may better explain the arid pollenfound in North Atlantic cores (8).

Some sub-Saharan animals do not exhibit trans-Saharan spa-tial distributions (Fig. 2B and SI Appendix, Figs. S12 and S13).They are found only in the southern Sahara, with a northern limitnear the Ennedi, Tibesti, Tasilli, and Ahoggar Mountains, repre-senting a catchment divide in the central Sahara (SI Appendix,Fig. S12). These species all have specialized aquatic require-ments, such as deep water in the case of Nile perch (Lates nilo-ticus) and hippopotamus (Hippopotamus amphibius) or long-termwater 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 thisdivide to the northern Sahara, but appears to have been a barrierto specialized aquatic species, accounting for an impoverishedaquatic fauna in the northern Sahara.

The Peopling of the Sahara During the HoloceneWe hypothesize that the differences in animal resources betweenthe northern and southern Sahara during the early Holoceneinfluenced the way it was peopled by humans. The north–southcontrast in Saharan species ranges are remarkably similar to somekey lithic, bone tool, and linguistic spatial distributions, suggest-ing that the peopling of the region during the early Holocenehumid phase was driven by cultural adaptations that allowedexploitation of specific fauna.

The early Holocene archaeology of the Sahara is characterizedby a regional distribution of specific archaeological cultures,such as those defined by barbed bone points, fishhooks, Ouna-nian arrow-points, and, more controversially, pottery (32, 35–38).The Sahara today is largely populated by speakers of Afroasiaticlanguages, Berber and Arabic, with some Nilo-Saharan languages(Teda-Daza and Zaghawa) in the region of Northern Chad, andSonghay 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 andthe spread of the Hassaniya Moors into Mauritania probablydates from the 15th Century (39). Before this time, the centraland southern Sahara are thought to have been populated byNilo-Saharan speakers. The Nilo-Saharan language phylum isboth widespread and strongly internally divided, suggestingconsiderable antiquity (40) (Fig. 3). Its greatest diversity is in theeast, where a large number of small branches are found (Fig. 3),

suggesting the original locus of expansion. Although fragmentedinto enclave populations today, the presence and pattern of relicpopulations in the northern desert points strongly to a muchwider distribution in the past, covering the region from theEthio-Sudan borderland to Mauritania and southwest Morocco.

It has long been suggested that Nilo-Saharan languages mightcorrelate with barbed bone points, the so-called “Aqualithic”(35). Fig. 3 superimposes the sites of known barbed bone pointson a map of current Nilo-Saharan languages, showing a remark-able similarity in spatial distribution, and also a notable corre-spondence with Holocene distribution of large aquatic species(e.g., Fig. 2A and SI Appendix, Fig. S12). It appears that the ex-pansion of aquatic resources in the Holocene made the Saharaattractive to populations with existing fishing and riverine huntingskills (SI Appendix, Section 6). Their ability to hunt hippopotamusand crocodiles and to catch a wide variety of deepwater fishspecies would have propelled a rapid dispersal from east to westand into the central Sahara, to judge by the numerous branches ofNilo-Saharan in the east (Fig. 3 and SI Appendix, Section 6).Their movement further north would have been restricted bythe absence of many of these species. However, the presence ofan isolated Nilo-Saharan population (the Koranje, a branch ofSonghay) and a barbed bone point in Northwest Africa nearthe headwaters of the catchment of the Soura River, the river thatlinks the Atlas mountains to the lakes in the Ahnet-Mouydir basin(Fig. 1), forming a corridor from the central Sahara to NorthwestAfrica, indicates that a few groups may have traversed the greenSahara using the most promising routes. There is direct linguisticevidence that Nilo-Saharan populations exploited these aquaticresources in the form of a widespread cognate for “hippo” fromGumuz 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 andwestern languages.

We hypothesize that the other economic revolution thatoccurred in the Sahara at approximately the same time was thesouthward spread of the bow and arrow. North African hunterswould have observed the new abundance of large and unfamiliar

Fig. 3. Sahara palaeohydrology overlaid with the spatial distribution of Ou-nanian and barbed bone points and Nilo-Saharan languages. The languagesmarked as “other” are too small to be depicted as separate colors; they areNyimang, Temein, Hill Nubian, Daju, Berta, and Gumuz. Note the similaritybetween the distribution of barbed bone points and the distribution of spe-cies that require deep water (Fig. 2B and SI Appendix, Figs. S12 and S13).

460 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1012231108 Drake et al.

Dow

nloa

ded

by g

uest

on

June

3, 2

020

land mammals to the south, notably elephant and giraffe. In adispersal inverse to that of the Nilo-Saharans, they would havebeen attracted southward to hunt these animals with the bowand arrow. The “Ounanian” of Northern Mali, Southern Algeria,Niger, and central Egypt at ca. 10 ka is partly defined by adistinctive type of arrow point (37). These arrowheads are foundin much of the northern Sahara (Fig. 3) and are generally consid-ered to have spread from Northwest Africa. This view is supportedby the affinity of this industry with the Epipalaeolithic that alsoappears to have colonized the Sahara from the north (41). NoOunanian points occur in West Africa before 10 ka, suggestingthe movement of a technology across the desert from north tosouth around this time.

Our model envisages the initial Holocene repopulation of theSahara being carried out by two separate populations practicingtwo quite different resource exploitation strategies: (i) aquatic for-aging using bone point and fish hook technology, and (ii) savannahunting using the bow and arrow. By linking the distribution of theNilo-Saharan language phyla to the archaeological distribution ofaquatic- and terrestrial-adapted technologies, we explain the pat-tern of human repopulation of the desert in terms of the changingfaunal distribution, which is in turn dictated by the nature of trans-Saharan hydrological linkages.

Older Saharan Occupation and CrossingsThe movement of people across the Sahara during the Holocenemay well have been the last of several important hominin disper-sals. To demonstrate the possibility of earlier trans-Saharanmigrations, it is necessary to determine the timing of pre-Holo-cene humid phases in the Sahara. This chronological evidencecan be obtained by dating the deposition of pre-Holocene lakesediments and demonstrating synchronous palaeohydrologicchanges across the Sahara. MIS5 contains the most abundantlydated pre-Holocene humid sediment sequences, yet only twospeleothems, two regions containing spring tufas, and nine sepa-rate palaeolakes have been dated to this period (SI Appendix,Sections 4 and 7), thus there are many gaps and it is not possibleto say if the entire Sahara was humid. For example, before thisstudy, there was a complete lack of last MIS5 ages in the south-ernmost part of the Sahara. One way to overcome this problem isto date lacustrine episodes in the Saharan megalakes because theyare located within abutting catchments and these catchments spanthe desert (Fig. 1). Because the high stands indicate significantlyincreased moisture availability, fluvial systems within the megalakecatchments must be active at these times. Consequently, synchro-nous lake high stands would yield a humid corridor across theSahara, allowing northward dispersal of hominin populationslocated in sub-Saharan Africa.

To evaluate the timing of pre-Holocene humid phases, it isnecessary to locate and date extensive sequences of lacustrinesediments in megalake basins. Fortunately, Saharan megalakesappear to preserve the required long-term lacustrine sequences(16–19, 42) despite the general assumption that such records arerare in the Sahara because of deflation of lacustrine sedimentsduring arid periods (43). We have applied optically stimulatedluminescence (OSL) dating to selected sediments (SI Appendix,Section 7) and have found evidence for synchronous humidity inthe Fezzan–Chad–Chotts and Chad–Chotts–Ahnet-Moyer mega-lake corridors during MIS5 (Fig. 1). Humidity in the ChadBasin was investigated by dating Lake Megachad beach ridges(SI Appendix, Fig. S16). The Bama Ridge in northeastern Nigeriaforms the southwestern shoreline of the Holocene palaeolakeMegachad with an area of 361;000 km2 (10). To the southwestof the Bama Ridge, we identified a succession of older beachridges that run roughly parallel to it (SI Appendix, Fig. S16).OSL ages for two of these ridges at Alhajari and Kawiya giveages of 114� 14 ka and 125� 12 ka, respectively, suggestingmegalake conditions in MIS5. In the Fezzan Basin, OSL dating

(SI Appendix, Section 7; ref. 16) of palaeolake sediments showsthat two large lakes of 1,350 and 1;730 km2 existed in this basin ata similar time (SI Appendix, Fig. S16). The Fezzan Basin ages areconsistent with a U/Th isochron age for lake sediments fromthe basin of the Chotts (98� 5 ka; ref. 18) and the Ahnet-Mouy-dir basin [92ðþ20 − 18Þ ka; ref. 19). At this time, the Chotts basincontained 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 duringMIS5 that could have been used by hominins to cross the Sahara(Figs. 1 and 4). Once hominins had achieved a trans-Saharan dis-tribution, it would have been possible to pass into the Levant viathe Mediterranean coast. Because the Levant was humid at thistime (44), onward dispersal would have been possible.

It is probable that increased humidity within the megalakecatchments is indicative of regional scale humidity and hence dis-persal routes may have existed throughout the desert, as appearsto have been the case during the early Holocene, rather thanbeing restricted to corridors. Other studies of Saharan fluvialand lacustrine sediments provide some evidence for humidityin other areas. A humid corridor has recently been proposedin the Sahabi palaeoriver system (13) (Fig. 4) and there are 41other ages for 12 areas that contain humid sediments scatteredthroughout the Sahara that are indistinguishable from these“corridor” ages. Their integration with the palaeohydrologicalmap provides evidence for a wider green Sahara during MIS5 aswell as the present interglacial (Fig. 4 and SI Appendix, Section 3).

Using the Holocene biogeography and palaeohydrology of theSahara as an analogue for the MIS5 humid period, it is likely thatan interconnected waterway would have been available for faunaland human dispersal. This humid period corresponds very closelywith the age of the first modern human occupation of the NorthAfrican coast (45) and the Levant (46) by sub-Saharan popula-tions, who may have been crossing the Sahara at this time (9).The occupation of the Mediterranean coast of Africa by theseearly modern human migrants appears to have lasted from∼110 to ∼30 ka (45), though the Levantine occupation appearsto have finished by ∼70 ka (47). Some view the out-of-Africa dis-persal into the Levant as the start of the spread of modern hu-mans onward into Arabia and India in MIS5 (48), whereas othersbelieve it to be a “dead end” that was followed by a later moresuccessful dispersal of modern humans out of Africa at a laterdate: 60 ka in MIS4 (49). There is little evidence for a green

Fig. 4. Palaeohydrology of North Africa during MIS5 with the location ofdated sites marked. The hatched area covers catchments containing largepalaeolakes and thus represents the minimum area which experiencedconsiderably enhanced humidity during MIS5.

Drake et al. PNAS ∣ January 11, 2011 ∣ vol. 108 ∣ no. 2 ∣ 461

GEO

LOGY

Dow

nloa

ded

by g

uest

on

June

3, 2

020

Sahara in MIS4, so the desert is unlikely to have directly played arole in this postulated second dispersal, however, the modernhumans north of the Sahara could have provided the populationrequired to drive it. Indeed, “spatially” this hypothesis is attrac-tive because the Sinai corridor out of North Africa is close to thelocation of the earliest archaeological sites associated with thisdispersal, which are found in the Levant at ∼47 ka (50). However,an alternative dispersal route for modern humans out of Africavia the Bab el Mandab has been proposed at about this time,largely on genetic grounds (51), thus if this dispersal occurred, aroute out of Africa via North Africa is not necessary. To furthercomplicatematters, thepossibility ofmultipledispersals ofmodernhumans both into and out of Africa has recently been suggested(52). If these proposedmigrations did indeed happen, it is feasiblethat these different dispersals employed different routes.

Notwithstanding the as yet unresolved issue of the timing androute taken by modern humans out of Africa, both the use of the

Green Sahara route to the North African coast by hominins andits long-term viability are supported by archaeological evidence(53–56). Stone tools from Oldowan to Neolithic have been foundin all the megalake basins that form corridors across the Sahara,suggesting that the region was not only periodically habitable forhominins, but that it was also regularly occupied by them.

ACKNOWLEDGMENTS.We thank the Society for Libyan Studies for funding theFezzan Project that inspired this work and the Desert Migrations Projects thatallowed research in the Fezzan to continue. Particular thanks go to DavidMattingly who led these research projects and encouraged this research.We are also grateful to Marta Lahr, Robert Foley, and Kevin MacDonaldfor helpful discussions on the peopling of the Sahara and to David Blackburnfor his help in understanding its biogeography. Grants from the RoyalGeographical Society—Institute of British Geographers Peter FlemingAward, the National Geographic Society, Repsol/YPF Murzuk S A, and theGreat Manmade River Authority also funded aspects of this research, forwhich they are thanked.

1. Vermeersch PM (2001) “Out of Africa” from an Egyptian point of view. Quat Int75:103–112.

2. Van Peer P (1998) The Nile corridor and the out-of-Africa model: An examination ofthe archaeological record. Curr Anthropol 39:S115–140.

3. Said R (1993) The River Nile. Geology, Hydrology and Utilisation (Elsevier, Oxford).4. WilliamsMAJ, Adamson D, Prescott JR, Williams FM (2003) New light on the age of the

White Nile. Geology 31:1001–1004.5. Tawadros E (2001) Geology of Egypt and Libya (Balkema, Rotterdam).6. Douady CJ, Catzeflis F, Raman J, Springer MS, Stanhope MJ (2003) The Sahara as a

vicariant agent, and the role of Miocene climatic events, in the diversification ofthe mammalian order Macroscelidea (elephant shrews). Proc Natl Acad Sci USA100:8325–8330.

7. Schulz E (1991) Holocene environments in the central Sahara. Hydrobiologica214:359–365.

8. Hooghiemstra H, Stalling H, Agwu COC, Dupont LM (1992) Vegetational and climaticchanges at the northern fringe of the Sahara 250,000-5,000 years BP: Evidence from 4marine pollen records located between Portugal and Canary Islands. Rev PalaeobotPalynol 74:1–53.

9. Lahr MM (1996) The Evolution of Modern Human Cranial Diversity: A Study in CranialVariation (Cambridge Univ Press, Cambridge, UK).

10. Drake NA, and Bristow C (2006) Shorelines in the Sahara: Geomorphological evidencefor an enhanced monsoon from palaeolake Megachad. Holocene 16:901–112.

11. Duveyrier H (1864) Exploration du Sahara: Les Touâreg du nord (Challamel Aine, Paris).12. Dumont HJ (1982) Relict distribution patterns of aquatic animals: Another tool in

evaluating late Pleistocene climate changes in the Sahara and Sahel. Palaeoeco Africa14:1–24.

13. Osborne AH, et al. (2008) A humid corridor across the Sahara for the migration ofearly modern humans out of Africa 120,000 years ago. Proc Natl Acad Sci USA105:16444–16447.

14. Tjallingii R, et al. (2008) Coherent high- and low-latitude control of the northwestAfrican hydrological balance. Nat Geosci 1:670–675.

15. Van NeerW (1989) Recent and fossil fish from the Sahara and their palaeohydrologicalmeaning. Palaeoeco Africa 20:1–18.

16. Armitage SJ, et al. (2007) Multiple phases of North African humidity recorded inlacustrine sediments from the Fezzan Basin. Quat Geochronol 2:181–186.

17. Drake NA, et al. (2008) Palaeohydrology of the Fazzan Basin and surrounding regions:The last 7 million years. Palaeogeogr Palaeoclimatol Palaeoecol 263:131–145.

18. Causse C, et al. (2003) Humidity changes in southern Tunisia during the LatePleistocene inferred from U-Th dating of mollusc shells. Appl Geochem 18:1691–1703.

19. Causse C, et al. (1988) Pleistocene du Sahara nord-occidental daterait de80–100,000 ans. C R Acad Sci Paris 306:1459–1464.

20. Lévêque C (1990) Relict tropical fish fauna in central Sahara. Ichthyol Explor Freshwat1:39–48.

21. Gasse F (2000) Hydrological changes in the African tropics since the Last GlacialMaximium. Quat Sci Rev 19:189–211.

22. Van Damme D (1984) The Freshwater Molluscs of Northern Africa. Distribution,Biogeography and Palaeoecology, , Developments in Hydrobiology (Dr. W. Junk Publ,Hague), 25.

23. Borkin LJ (1999) Patterns of Distribution of Amphibians: A Global Perspective, edWE Duellman (Johns Hopkins Univ Press, Baltimore), pp 329–420.

24. de Smet K (1998) Status of the Nile crocodile in the Sahara desert. Hydrobiologia391:81–86.

25. Le Berre M (1989) Faune du Sahara Volume 1: Poisons, Amphibians, Reptiles(LeChevalier–Chabaud, Paris).

26. Mauny R (1955) Segmentation of the great Ethiopian fauna of North-West Africa fromthe Paleolithic to the present day (Translated from French). Proc Third Pan-AfricanCongress on Prehistory (Livingstone, Edinburgh), pp 102–105.

27. Petit-Maire N (1993) Map of the Sahara in the Holocene. (Commission for theGeological map of the World/United Nations Educational, Scientific, and CulturalOrganization, Paris), 1 p 500000.

28. Vernet R (1995) Climats anciens du Nord de L’Afrique (L’Harmattan, Paris).

29. Pieters J, von den Driesch A (2003) Deciphering Ancient Bones: The Research Poten-tial of Bioarchaeological Collection, eds G Grupe and J Peters (Verlag Maria Leidorf,Rahden, Germany), pp 265–284.

30. Gautier A (1987) Prehistory of Arid North Africa, ed A Close (Southern Methodist UnivPress, Dallas), pp 163–87.

31. Jousse H (2006) What is the impact of Holocene climatic changes on human societies?Analysis of Neolithic populations dietary customs. Quat Int 151:63–73.

32. Sereno PC, et al. (2008) Lakeside Cemeteries in the Sahara: 5000 Years of HolocenePopulation and Environmental Change. PLoS One 3:e2995.

33. Churcher CS, Kleindienst MR, Schwarcz HP (1999) Faunal remains from a MiddlePleistocene lacustrine marl in Dakhleh Oasis, Egypt: Palaeoenvironmental reconstruc-tions. Palaeogeogr Palaeoclimatol Palaeoecol 154:301–312.

34. Le Quellec J-L (1993) Symbolisme et art rupestre au Sahara (L’Harmattan, Paris).35. Sutton JEG (1977) The African aqualithic. Antiquity 51:25–34.36. Yellen JE (1998) Barbed bone points: Tradition and continuity in Saharan and sub-

Saharan Africa. Afr Archaeol Rev 15:173–198.37. Vernet R, Ott M, Tarrou L, Gallin A, Géoris-Creuseveau J (2007) Excavation of the

mound of FA 10 (Banc d'Arguin) and its contribution to the knowledge of theculture paleolithical Foum Arguin, northwestern Sahara (Translated from French).J Afr Archaeol 5:17–46.

38. Wendorf F, Schild R (1998) Nabta playa and its role in northeastern African prehistory.J Anthropol Archaeol 17:97–123.

39. Abun-Nasr JM (1987) A History of the Maghrib in the Islamic Period (Cambridge UnivPress, Cambridge, UK).

40. Bender ML (1997) The Nilo-Saharan Languages: A Comparative Essay (Lincom Europa,Munich).

41. Clark JD (1980) The Sahara and the Nile, eds MAJ Williams and H Faure (Balkema,Rotterdam), 527–582.

42. Brunet M, et al. (2002) A new hominid from the Upper Miocene of Chad, CentralAfrica. Nature 418:145–151.

43. Swezey CS (2003) Climate Controls on Stratigraphy, eds CB Cecil and NT Edgar (Societyfor Sedimentary Geology Special Publication, Tulsa, OK), 77, pp 207–225.

44. Vaks A, et al. (2007) Desert speleothems reveal climatic window for African exodus ofearly modern humans. Geology 35:831–834.

45. Barton RNE, Bouzouggar A, Collcutt SN, Schwenninger J.-L, Clark-Balzan L (2009) OSLdating of the Aterian levels at Dar es-Soltan I (Rabat, Morocco) and implications forthe dispersal of modern Homo sapiens. Quat Sci Rev 28:1914–31.

46. Grün R, et al. (2005) U-series and ESR analyses of bones and teeth relating to thehuman burials from Skhul. J Hum Evol 49:316–34.

47. Shea JJ (2008) Transitions or turnovers? Climatically-forced extinctions of Homosapiens and Neanderthals in the east Mediterranean Levant. Quat Sci Rev 27:2253–2270.

48. Petraglia MD, Haslam M, Fuller DQ, Boivin N (2010) The southern dispersal routeand the spread of modern humans along the Indian Ocean rim: New hypothesesand evidence. Ann Hum Biol 37:288–311.

49. Mellars P (2006) Why did modern human populations disperse from Africa ca.60,000 years ago? A new model. Proc Natl Acad Sci USA 103:9381–9386.

50. Mellars P (2006) A new radiocarbon revolution and the dispersal of modern humans inEurasia. Nature 439:931–935.

51. Macaulay V, et al. (2005) Single, rapid coastal settlement of Asia revealed by analysis ofcomplete mitochondrial genomes. Science 308:1034–1036.

52. Gunz P, Bookstein FL, Mitteroecker P, Stadlmayr A, Seidler H, Weber GW (2009)Early modern human diversity suggests subdivided population structure and acomplex out-of-Africa scenario. Proc Natl Acad Sci USA 106:6094–6098.

53. Lahr MM, et al. (2008) DMP III: Pleistocene and Holocene palaeonvironments andprehistoric occupation of Fazzan, Libyan Sahara. Libyan Studies 39:263–294.

54. Mattingly D, et al. (2008) Desert migrations: People, environment and culture in theLibyan Sahara. Libyan Studies 38:115–156.

55. Tillet T (1985) The Palaeolithic and its environment in the northern part of the Chadbasin. Afr Archaeol Rev 3:163–177.

56. Aumassip G (2004) Prehistoire du Sahara et de ses abords (Maisonneuve andLarose, Paris).

462 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1012231108 Drake et al.

Dow

nloa

ded

by g

uest

on

June

3, 2

020