Lakeside Cemeteries in the Sahara: 5000 Years of Holocene Population and Environmental Change

Background Approximately two hundred human burials were discovered on the edge of a paleolake in Niger that provide a uniquely preserved record of human occupation in the Sahara during the Holocene (∼8000 B.C.E. to the present). Called Gobero, this suite of closely spaced sites chronicles the rapid pace of biosocial change in the southern Sahara in response to severe climatic fluctuation. Methodology/Principal Findings Two main occupational phases are identified that correspond with humid intervals in the early and mid-Holocene, based on 78 direct AMS radiocarbon dates on human remains, fauna and artifacts, as well as 9 OSL dates on paleodune sand. The older occupants have craniofacial dimensions that demonstrate similarities with mid-Holocene occupants of the southern Sahara and Late Pleistocene to early Holocene inhabitants of the Maghreb. Their hyperflexed burials compose the earliest cemetery in the Sahara dating to ∼7500 B.C.E. These early occupants abandon the area under arid conditions and, when humid conditions return ∼4600 B.C.E., are replaced by a more gracile people with elaborated grave goods including animal bone and ivory ornaments. Conclusions/Significance The principal significance of Gobero lies in its extraordinary human, faunal, and archaeological record, from which we conclude the following: The early Holocene occupants at Gobero (7700–6200 B.C.E.) were largely sedentary hunter-fisher-gatherers with lakeside funerary sites that include the earliest recorded cemetery in the Sahara. Principal components analysis of craniometric variables closely allies the early Holocene occupants at Gobero with a skeletally robust, trans-Saharan assemblage of Late Pleistocene to mid-Holocene human populations from the Maghreb and southern Sahara. Gobero was abandoned during a period of severe aridification possibly as long as one millennium (6200–5200 B.C.E). More gracile humans arrived in the mid-Holocene (5200–2500 B.C.E.) employing a diversified subsistence economy based on clams, fish, and savanna vertebrates as well as some cattle husbandry. Population replacement after a harsh arid hiatus is the most likely explanation for the occupational sequence at Gobero. We are just beginning to understand the anatomical and cultural diversity that existed within the Sahara during the Holocene.

(1) The early Holocene occupants at Gobero (7700-6200 B.C.E.) were largely sedentary hunter-fisher-gatherers with lakeside funerary sites that include the earliest recorded cemetery in the Sahara. (2) Principal components analysis of craniometric variables closely allies the early Holocene occupants at Gobero with a skeletally robust, trans-Saharan assemblage of Late Pleistocene to mid-Holocene human populations from the Maghreb and southern Sahara. (3) Gobero was abandoned during a period of severe aridification possibly as long as one millennium (6200-5200 B.C.E). (4) More gracile humans arrived in the mid-Holocene (5200-2500 B.C.E.) employing a diversified subsistence economy based on clams, fish, and savanna vertebrates as well as some cattle husbandry. (5) Population replacement after a harsh arid hiatus is the most likely explanation for the occupational sequence at Gobero. (6) We are just beginning to understand the anatomical and cultural diversity that existed within the Sahara during the Holocene.

Introduction
The ''greening'' and ultimate desiccation of the Sahara rank among the most severe climatic fluctuations during the Holocene [1]. Driven by variation in orbital insolation and magnified by feedback between monsoonal rainfall and vegetation [2], ecosystem succession in the Sahara is well known from many lines of evidence such as pollen spectra [3], paleolake levels [4][5][6], and, most recently, high-resolution paleolake sediment cores [7].
Human adaptation during this period of climate fluctuation is best known in the Eastern Sahara to the west of the Nile valley. This region witnessed continuous occupation from 8500 B.C.E., when hunter-gatherers using a distinctive Epipaleolithic tool kit expanded across open grass savanna habitats, to about 3500 B.C.E, when aridification drove pastoralists from most areas of the desert [8]. Occupational patterns in low-lying regions elsewhere in the Sahara most closely resemble the Eastern Sahara during the early Holocene (,8000-7000 B.C.E.), when pottery-producing hunter-fisher-gatherers resided beside paleolakes, utilizing a tool kit including microliths and harpoon points and fish hooks of bone [9][10][11]. By the mid-Holocene, occupational histories diverge in the Central and Western Sahara [12] as a result of distinctive local humid-arid cycles [13][14][15], diversified economies and lifestyles tied to ephemeral paleolakes [10], upland refugia [16] and rivers [17], and marked variation among the human populations themselves [10,18,19]. Despite increasing knowledge regarding occupational succession in the Sahara from early to late Holocene [8,10,11,16,17], that record is based on individual sites that typically preserve short intervals of occupation, include few if any intact burials, and rely largely on indirect dating of human remains and artifacts [20].
We report here on a new site complex called Gobero located at the western tip of the hyperarid Ténéré Desert in the southern Sahara in Niger ( Figures 1A, 2). Approximately 200 burials ranging in age over five millennia are present in the upper level of several paleodunes that are situated adjacent to a paleolake deposit. Gobero preserves the earliest and largest Holocene cemetery in the Sahara, opening a new window on the funerary practices, distinctive skeletal anatomy, health and diet of early Holocene hunter-fisher-gatherers, who expanded into the Sahara when climatic conditions were favorable. The site complex also preserves numerous mid-Holocene burials, some indicating funerary rituals with grave inclusions. Associated middens and an exceptional faunal and pollen record provide a chronicle of episodic human occupation in the Sahara under conditions of severe climatic change.

Geologic Setting
Gobero is located on the northwestern rim of the Chad Basin, approximately 150 km southeast of the Aïr massif ( Figure 1A). Isolated on a vast peneplain of mid-Cretaceous sandstone between fields of migrating barcan dunes (Figure 2), the most important cluster of sites at Gobero is located in low, calcrete-fringed paleodunes that are partially surrounded by paleolake deposits [ Figure 1B, C). The paleodunes accumulated at Gobero over a period of at least seven millennia from the Late Pleistocene to the early Holocene (,14,000-7000 B.C.E.), as determined by optically stimulated luminescence (OSL) dating of paleodune sand at various depths ( Figure 1F, Table 1). During the best preserved intervals of occupation, the core paleodune sites (G1-3) formed islands that may have been originally partially conjoined as a narrow peninsula into Paleolake Gobero, a shallow, freshwater lake no more than 3 m in depth and 3 km in diameter ( Figure 1C, D). The lake occupied a small endorheic basin located on a low rise between drainages southwest to the Niger River and southeast to Paleolake Chad. Fed periodically by surface water from the Aïr massif that pooled against a low east-west fault scarp (Mazelet) immediately to the south, Paleolake Gobero appears to have been more closely associated with the Chad Basin, as there is no trace of the giant catfish (Arius gigas) common to drainages of the Niger [21]. The archaeological sites were submerged when the lake filled to a depth greater than 5 m ( Figure 1D). Over time the human bone in submerged burials darkened and hardened to resemble the pyrolusite (MnO 2 )-darkened vertebrate bone in the adjacent paleolake deposit.

Model for the Gobero Sequence
Based on geochronological data (Tables 1, 2) with input from archaeological, craniometric, zooarchaeological and archaeobotanical analysis, we divide the record preserved at Gobero into four occupation phases (Figure 3).

Phase
1-Paleodune Accumulation (14,000-7700 B.C.E.). Phase 1 is predominantly an arid interval at the close of the Pleistocene and beginning of the Holocene, when dune sands accumulated in the Gobero area over Cretaceous bedrock under climatic conditions characterized by weakened monsoons and the spread of aridification across northern Africa [4][5][6][12][13][14].
The base of the paleodune deposit at site G1 dates to ,14,000 B.C.E. (Figure 1B, F, section 12; Table 1, dates 3,4). Although humid intervals are recorded elsewhere in the central Sahara [5,13,14] and may have occurred during phase 1 at Gobero, Paleolake Chad never grew to megalake dimensions until later in the early and mid-Holocene [5,6], and the paleodune sequence at Gobero accumulated without a detectible hiatus. Deflated Ounanian artifacts [22][23][24] were recovered suggesting that during this phase there were transient hunter-gatherers in the Gobero area who seem not to have left any burial record in the lower portion of the paleodune sequence.
Phase 2-Early Holocene Occupation (7700-6200 B.C.E.). During phase 2, wet climatic conditions attracted a population of hunter-fisher-gatherers to Gobero sites, which served both funerary and habitation functions. In one area of site G3 no larger than 50 m 2 , 17 closely interspaced (#4 m), undisturbed burials contain dark-stained skeletons composing a cemetery ( Figure 4A, B). Direct dates for five of these burials indicate an age of ,7500 B.C.E. and range over only ,250 years ( Figure 3; Table 2, dates 1-7). This cemetery is considerably larger than a small cluster of burials at a somewhat younger site along the upper Nile (El Damer) [11] and predates by three millennia the oldest cemetery in Egypt's Western Desert (Gebel Ramlah) [25]. Phase 2 is delimited in time by burials within site G3 with direct dates (Table 2, dates 1, 7), the oldest a subadult individual in the cemetery (7730-7580, midpoint 7655 B.C.E.) and the youngest an adult located nearby (6380-6210, midpoint 6295 B.C.E.). Phase 2 peoples are tall in stature, approaching two meters for both males and females. Hyperflexed, supine burial postures predominate, their compact configuration and anatomical articulation suggesting that their bodies were tightly bound with animal skin, ligament or basketry binding, although no trace of these perishable materials are preserved ( Figure 4C). Their crania are long and low and are characterized by a distinct occipital bun, flattened sagittal profile, pentagonal posterior outline, broad proportions across the zygoma and interorbital region, broad nasal aperture, and negligible alveolar prognathism ( Figure 4D), features that are apparent in juveniles as young as four years of age   Figure 4E) and absent in skulls from mid-Holocene burials ( Figure 5C).
Craniometric data from seven human groups (Tables 3, 4) were subjected to principal components analysis, which allies the early Holocene population at Gobero (Gob-e) with mid-Holocene ''Mechtoids'' from Mali and Mauritania [18,26,27] and with Late Pleistocene Iberomaurusians and early Holocene Capsians from across the Maghreb (see cluster in Figure 6). The striking similarity between these seven human populations confirms previous suggestions regarding their affinity [18] and is particularly significant given their temporal range (Late Pleistocene to mid-Holocene) and trans-Saharan geographic distribution (across the Maghreb to the southern Sahara).
Microliths, bone harpoon points and hooks, and ceramics with dotted wavy-line and zigzag impressed motifs were found in the burial fill, in an associated refuse area, and in nearby paleolake deposits ( Figure 7C-F). These artifacts exhibit attributes of the Kiffian technocomplex, named after the type site Adrar-n-Kiffi at Adrar Bous some 500 km to the north [28,29]. Direct dates on bone harpoon points and plant temper in Kiffian sherds confirm the association of several of these artifacts with this occupational phase ( Figure 3; Table 2, dates 26-28, 64, 65). Nile perch (Lates niloticus) and large catfish dominate the midden fauna, which also includes bones and teeth from hippos, several bovids, small carnivores, softshell turtles and crocodiles (Table 5, refuse area 5). The burial density, tool kit, ceramics, and midden fauna suggest a largely sedentary population with a subsistence economy based on fishing and on hunting of a range of savanna vertebrates.
Pollen from phase 2 burials at site G3 indicates an open, lowdiversity savanna with grasses and sedges, and an arboreal component including fig (Ficus) and tamarisk (Tamarix). Hydrophytes and rushes (Juncus) suggest the presence of permanent water and marshy habitats [15,30]. Nevertheless, xerophytes including saltbushes (Chenopodiaceaea) are significant, indicating that sandy habitats were also present, although perhaps distant, in the region.
Toward the end of phase 2 (,6500-6300 B.C.E.), the level of Paleolake Gobero rose, at least episodically, submerging the paleodunes and forcing the relocation of occupants ( Figure 1D). Well-aerated permanent water at depths of 5 m or more are suggested by large vertebrae (up to 5 cm diameter) of the Nile perch (Lates niloticus), which correspond to a body length of up to 2 m [31]. These vertebrae were found in situ in paleolake sediments and directly dated to the end of this phase ( Figure 3; Table 2, dates 57, 58; midpoint average 6525 B.C.E.). The darkened bone color of all human skeletons in phase 2 burials is indicative of sustained inundation. Inundation, nevertheless, may have been episodic, as a dark-stained human skeleton (G3B28), Kiffian potsherd, refuse area, and hartebeest skeleton (Alcelaphus buselaphus) are also directly dated to this interval ( Figure 3; Table 2 Table 2. cont. Because spillways external to the restricted catchment for Paleolake Gobero may have kept inundation levels under 10 m ( Figure 1D), it is possible that floodwaters may not have displaced lakeside occupants by a long distance or over a long temporal span. The inundation history at Gobero is doubtless complex, as the earliest mid-Holocene burials are also variegated or darkly stained (Table 2, dates 9-13; midpoint range 4825-3860 B.C.E.). Resolving additional inundation events of short duration will require further direct dates on paleolake fauna and sediments.  Table 2 using current atmospheric standards [55]. All of the burials that have been dated at Gobero fall within phases 2 and 3, which are shown as green to indicate favorable humid climate conditions; more arid intervals are shown as tan including occupation phases 1 and 4. Multiple dates on individual specimens or features are boxed. A dotted line separates early and mid-Holocene human burials. Abbreviations: B.C.E., before current era (registered to calendar year zero); B.P., before present (1950); G1B8, burial 8 on G1; G1B11, burial 11 on G1; G3B8, burial 8 on G3; K, Kiffian; LT, Late Tenerean; T, Tenerean. doi:10.1371/journal.pone.0002995.g003 Occupational Interruption (6200-5200 B.C.E). A harsh arid interval separates early and mid-Holocene populations at Gobero, when the paleolake appears to have dried out and the area abandoned. Although we have no means to directly assess aridity, no terrestrial or aquatic vertebrates or lakebed sediment have been dated to this interval, which lasted approximately one millennium ( Figure 3). The only specimens dated within this interval were found in the paleolake deposit and consist of a cluster of the small gastropod Melanoides tuberculata, a species that prefers periodically flooded habitats to permanent water bodies.
This occupational hiatus correlates well with the ''arid interruption'' in the central Sahara [12], a somewhat shorter interval (,6400-6000 B.C.E.; ,400 yr) of severe climatic deterioration across the Chad Basin [5,12] linked to cooling events in the North Atlantic [13,14]. This overlaps the early portion of the occupational hiatus at Gobero, the beginning of which is set after the youngest burial dated so far in occupation phase 2 ( Figure 3; Table 2 Phase 3 is a long interval over two millennia in length that witnessed the return of humid conditions and the re-occupation of Gobero by humans that account for approximately one-half of excavated burials ( Figure 3; Table 2, dates 9-25). The lower boundary of this phase marks the end of the occupational interruption and, as discussed above, is based on dates on lakebed sediment, a potsherd from burial fill, and the oldest burial in the occupation phase ( Grave goods occur in approximately 20% of the burials excavated thus far (7 of 35 burials) and include bones or tusks from wild fauna, ceramics, lithic projectile points, and bone, ivory and shell ornaments ( Figures 5B, 7A, B, H, K-M). Although the disc knife that characterizes the mid-Holocene Tenerean industry [28,29] has never been recovered in situ, small projectile points that also characterize the Tenerean tool kit are present in some burials that are directly dated to the mid-Holocene ( Figure 7H;  A conspicuous fine-grained green rock (with tan and brown variants) was often used for points, scrapers and adzes ( Figure 7G, I) and always used for the delicate Tenerean disc knives [28,29] at sites east and south of the Aïr massif. Previously this distinctive rock was identified as microcrystalline quartz, initially as jasper [32] and later as silicified vitric tuff [28]. For some 50 years, its source has remained a matter of speculation. Along with amazonite, the green rock was cited as evidence of trade over distances of one thousand kilometers or more, in order to link the Aïr with Tibesti (northern Chad-southern Libya) or regions farther afield [28,33,34].
Thin sections of artifacts made from the green rock and its color variants show the near absence of quartz. It is correctly identified as a felsite, a fine-grained volcanic rock composed of microcrystalline feldspar [35]. During reconnaissance a short distance (160 km) north of Gobero on the edge of the Aïr massif, we located a narrow outcrop of green felsite rock near the Alallaka wadi ( Figures 1A, 8A). The site is littered with debitage from a longstanding knapping operation ( Figure 8B). The frequent use of this rock at Gobero and Adrar Bous and its availability in nearby wadis along the eastern edge of the massif (Takolokouzet highland) suggests there were multiple felsite sources local to the Aïr. Given the absence of thorough geologic reconnaissance in the Aïr or any comparative trace element analysis, neither felsite nor amazonite provides evidence to infer trans-Saharan trade in the early or mid-Holocene.
At least semi-sedentary occupation is inferred from the number and density of undisturbed mid-Holocene burials, lack of evidence from strontium isotope analysis for enhanced mobility, abundance of lithic debitage on site, presence of groundstones, and numerous rodent marks on human bone and ornaments indicative of commensal species ( Figure 7L) [36,37]. The prevalence of juveniles of very young age among the interred also favors longer term occupation by family groups over transient visitation of the area as a watering hole or attractive hunting ground.
Clams (Mutela), small catfish (Clarias) and tilapia dominate the midden fauna (Figure 9), which also includes bones and teeth from hippos, a small antelope, small carnivores, softshell turtles and crocodiles (Table 5; middens 1-4). Domesticated cattle (Bos taurus) are present but unlike Adrar Bous [28,29] comprise only a minor component of the midden and area fauna (Figure 3, Bos partial mandible on chart sidebar; GF10). The scarcity of bones or teeth of domesticated cattle suggests a subsistence economy emphasizing fishing in shallow waters and hunting of a range of savanna vertebrates. The gathering of grain and cattle pastoralism may also have played important nutritional or economic roles; further data relevant to occupancy and subsistence patterns are needed, such as seasonality data from piscine otoliths recovered from middens. Elsewhere in the southern Sahara, diversification of dietary resources that combine gathering, hunting, fishing and pastoralism seems to occur under less certain climatic conditions [34,38,39], a pattern that may well accommodate the emerging archaeological record during the mid-Holocene at Gobero.
Pollen spectra from phase 3 burials indicate a mosaic of habitats. Open savannas with shrubland and grassland vegetation dominated, with sporadic presence of a fairly diversified Sudanian and tropical tree flora ( Figure 10). Plants linked to wet environments include hydrophytes, which indicate the presence of shallow freshwater lakes. Xeric and psammophilous plants indicate the presence of sandy soils.  Conclusions. In a recent (2003) review of Niger's archaeological record [40], the near absence of human remains across the broad expanse of the hyperarid Ténéré Desert was noted. Like many other regions of the Sahara, humid intervals in the early and mid-Holocene created attractive areas for human settlement in Niger. The Gobero site complex, which includes as many as 200 burials, underscores the scale and complexity of human occupation in a ''greener'' Sahara as well as the fragility of that record under present conditions. We draw the following major inferences from the rich and relatively continuous Holocene record at Gobero:

1)
Early Holocene sedentism. The early Holocene occupants at Gobero (7700-6300 B.C.E.) were largely sedentary hunterfisher-gatherers with lakeside funerary sites that include the earliest recorded cemetery in the Sahara dating to ,7500 B.C.E.

2)
Trans-Saharan craniometry. Principal components analysis of craniometric variables closely allies the early Holocene occupants at Gobero, who were buried with Kiffian material culture, with Late Pleistocene to mid-Holocene humans from the Maghreb and southern Sahara referred to as Iberomaurusians, Capsians and ''Mechtoids.'' Outliers to this cluster of populations include an older Aterian sample and the mid-Holocene occupants at Gobero associated with Tenerean material culture.

3)
Arid interruption. Early and mid-Holocene occupation phases 2 and 3 at Gobero are separated in time by a barren interval (6200-5200 B.C.E), which is associated with a period of severe aridification recorded across the Sahara.

5)
Population replacement. Population replacement rather than gradual phenotypic evolution best explains the distinctive craniofacial morphology and funerary practices of the human occupants during phases 2 and 3 in the early and mid-Holocene, respectively, particularly considering the relatively short intervening occupational hiatus. 6) Regional differentiation. The timing of population change observed at Gobero may only characterize a restricted area. Other areas in the southern Sahara, even those with comparable environmental conditions such as Hassi-el-Abiod in Mali, appear to show a later transition between human populations. The data from Gobero, when combined with existing sites in North Africa, indicate we are just beginning to understand the complex history of biosocial evolution in the face of severe climate fluctuation in the Sahara, a vast region that was occupied for much of the Holocene by an anatomically diverse series of human populations.

Discovery and Excavation
The Gobero site complex was discovered in 2000 during paleontological reconnaissance in central Niger, briefly revisited in 2003, and noted in the press the following year [41]. The Gobero Archaeological Project was established to coordinate field work and research by an international, multidisciplinary team. Two field seasons (2005,2006) have been undertaken, during which 67 burials were excavated and the main paleodune burial sites and adjacent paleolake deposits were mapped and sampled. A minimum of 182 human burials are preserved in the central area of the Gobero site complex (G1-5) buried within paleodunes on the edge of a paleolake. With few exceptions, only those burials that have reached the surface have been mapped and/or excavated thus far, and the total number of burials will very likely exceed 200 when the site complex is fully explored and excavated. The paleodunes and lakebed sites also preserve several middens, fossils of high quality documenting the invertebrate and vertebrate faunal record, and a pollen record. After setting a corner datum stake, a total station was used to map each site using a square meter grid. The total station was also used to integrate geologic sections and to log several topographic profiles across the paleolake catchment. Given the absence of any obvious bedding planes in the paleodune sediment or grave pit boundaries around any of the burials, excavation proceeded by dividing the sediment into a surface layer and burial levels of set depth. All sediment removed was screened (3 and 4 mm sieves), and lithics, ceramics and skeletal fragments were collected and labeled according to burial, level and grid square. The total station was also used to log positional data for several skeletal landmarks as well as significant grave goods. Sediment samples were taken for biogeochemical, pollen and parasite analysis.
Burials of significance for interpreting funerary practices were removed in plaster jackets. Individuals were excavated according to the general site protocol (outlined above) until the integrity of the specimen was at risk. All pollen and parasite sampling was then completed and the skeleton was collected intact using conservation grade consolidant (B72) and an encasing jacket of plaster. Laboratory preparation of these specimens was conducted under a stereomicroscope in the Fossil Lab at the University of Chicago, resulting in detailed preservation of skeletal position and the relationship between the skeleton and particular grave inclusions. Burials collected using this protocol include an adult male in hyperflexed supine pose ( Figure 4C), an adult male buried in recumbent pose and associated with the carapace of a mud turtle ( Figure 5A-C), a triple burial of an adult female and two juveniles ( Figure 5E, F), a juvenile in vertical burial pose, and a juvenile with an upper arm bracelet ( Figure 5D).

Optically Stimulated Luminescence Dating of Paleodune Sand
When grains of quartz are hidden from exposure to light, they accumulate trapped charges due to ionizing radiation from radionucleotides in the sediment. This energy accumulates predictably over time and is released as luminescence after controlled light stimulation in the lab. The age of the sediment (last exposure to daylight) is a measure of the paleodose (radiation dose to which the crystalline material has been exposed) divided by the total radiation dose for each year [42,43]. For the paleodune samples from Gobero, the cosmic radiation dose was estimated assuming a sediment water content of 562% at an altitude of 560 m. Samples were collected on a moonless night from a newly excavated section of paleodune sediment, with a depth ranging from 1-3 m below the paleodune surface (Table 1).

Direct 14 C AMS Dating of Enamel and Bone
Bone collagen is usually thoroughly degraded in arid environments, which is true for all skeletal samples from Gobero (human, faunal). Darkly-stained specimens that were likely subject to prolonged periods of inundation did not preserve collagen; in bone samples lighter in color, collagen preservation was poor as well. We therefore turned to 14 C AMS dating of the carbonate component in bioapatite in bone and enamel. Few publications have reported 14 C dates on tooth enamel and bone bioapatite, as the method is destructive and less well established than that utilizing bone collagen as a source of radiocarbon. In pre-treatment of enamel, we used a vacuum as an alternative to acetic acid [44].
Our sampling strategy for human enamel involved replacing all sampled crowns with exact replicas in their original position. We first took photographs in multiple views of the teeth to be sampled. A silicone mold was made for each crown, from which we made a urethane cast. That cast was fixed in the position of the original crown, using the photographs of the original tooth row prior to sampling. Using this sampling protocol, we retained all information for subsequent dental anthropological analysis.
General concordance between dates on human burials with similar taphonomic signatures and buried in close proximity suggest that the apatite dating is yielding accurate results and that the burials represent a contemporaneous early Holocene cemetery. Five individuals dated to a narrow range of about 250 years (8640-8330 BP; Table 2, dates 1-7).
To test the reliability of using the carbonate component in bioapatite as a source of radiocarbon, we took multiple samples using different source materials from the same individual or   Table 5. Faunal list at Gobero showing taxa recorded at archaeological sites (G1-3), in early and mid-Holocene middens (middens 1-4, refuse area 5), and in the Gobero area (in paleolake deposits or on deflation surfaces).

Gazella rufifrons +
Gazella dama + feature ( Table 6). The general concordance of dates validates the method, as seen in three dates from an early Holocene skeleton ( Table 6, dates 1-3). The enamel sample, the crystalline structure of which makes isotopic exchanges more difficult, and the core of the femur, which is especially porous after hydrolysis of collagen, still provide concordant dates. This demonstrates the near absence of isotopic exchange under arid conditions. The third estimate of 8220640 BP suggests slightly greater sensitivity to exchange in bone compared to enamel as expected.
A second multiply-sampled individual (G1B11) is mid-Holocene in age (Table 6, dates 4-6). The discrepancy between dates from this individual is slightly greater and may be due to insufficient purification during pre-treatment. The enamel for one of these samples may have been imperfectly separated from the dentine. Under arid conditions, carbon isotope exchange almost always involves rejuvenation, during which total inorganic dissolved carbon (TIDC) from surface waters tends toward equilibrium with atmospheric CO 2 ,. This process results in a younger date. Excluding other complicating factors, the older date (5940640 BP) is more likely to be the most accurate.
A third human skeleton was sampled twice, using enamel and bone, and also yielded concordant results (Table 6, dates 7, 8). A concordant pair of dates was also obtained using enamel and bone from a cattle mandible (Bos taurus, GF10), which was found deflated on the surface of the paleolake deposit (Table 6, dates 9, 10).
The differences that are apparent between enamel and bone samples from the same individual suggests that bone shows some isotopic exchange with the TIDC surface waters. This rejuvenation results in an average decrease in age of 250 years, a minor difference for dates in the early and mid-Holocene. The isotopic exchange of carbon, in addition, does not seem to be linked to time (e.g. 8420 and 8470 BP), but rather to local variation in inundation or sediment. The d 13 C from bones that experienced early diagenesis remains slightly more positive (e.g. G1B8), but this criterion is too tenuous to be reliable.
Convergence of dates based on multiple sources was also used to evaluate the age of features such as the middens or lakebed fauna. Clam shell from middens or the lakebed that was sampled in multiple ways (leached; chalky exterior shell; dense opalescent, unaltered inner shell) yielded generally concordant results.
In conclusion, dating of the bone and enamel bioapatite has provided the backbone for our chronology (Figure 2). If significant diagenesis were a confounding factor, the dates would show greater dispersion.

Comparative Craniometric Analysis
Craniometric measurements were selected to provide maximum correspondence with previous work on North African Late Pleistocene and Holocene human remains [18,26,27]. Seven chronospatial sample units were defined in accordance with published radiometric dates and allocation of specific sites to particular North African cultures (Table 3, numbers 1-7). The principal components analysis includes all individuals for which data were available and the maximum number of craniometric variables. After individuals were allocated to one of the sampling units, variable means were calculated from the raw data for the following variables: LGO, BPX, LBN, HBB, AFR, APA, AOC, CFR, COC, HNP, HNZ, BNZ, BZY, BFW, BFX, HORC ( Table 4). The resulting matrix of means was subjected to principal components analysis to capture the correlations among variable means in a more manageable number of dimensions ( Figure 6). The strength of this approach is that it allows maximum inclusion of individuals and variables and thus captures the greatest possible genotypic structure. This approach, however, does not consider within-sample variability, and population affinity is based solely on mean differences.
Three of the loadings from the principal components analysis returned eigenvalues greater than 1.0 (Table 7). Factor loadings transparently reflect anatomical or functional units. The first principal component appears to most closely approximate a size vector, but the loadings are not uniformly positive. The pattern of positive and negative loadings for principal components 2 and 3 are not interpretable.

Strontium Isotope Analysis
In order to investigate residential mobility and sedentism at the Gobero site complex, strontium isotope analysis was performed on archaeological human tooth enamel as well as modern baseline soil and faunal samples. At the Archaeological Chemistry Laboratory at Arizona State University (ASU), the teeth and    [45].
Uranium and neodymium concentrations were obtained on a quadrupole inductively-coupled plasma mass spectrometer (Q-ICP-MS) to monitor diagenetic contamination in the samples.
Since biogenic enamel and bone should have very low concentrations of these elements, elevated concentrations of uranium and neodymium can be used as proxies for the presence of diagenetic contamination of trace elements such as strontium. The mean U/Ca and Nd/Ca ratios are low and within the range of published uncontaminated U/Ca ratios from enamel samples (U/Ca = 1.08610 26 [46,47]. Our preliminary archaeological human enamel data from Gobero (Table 8, 87 Sr/ 86 Sr = 0.7118960.00039, n = 28, 1s) closely matches the baseline data outlined above, which suggests that the sampled individuals were not moving between, or migrating from, areas with much lower strontium isotope signatures, such as the Cenozoic volcanic terrain of the Aïr highlands in northern Niger and Hoggar in southern Algeria ( 87 Sr/ 86 Sr = 0.703) [48][49][50][51].  [35], is exposed as an oval outcrop approximately 0.8 km in width and 2.5 km in length near the Alallaka wadi on the southeastern edge of the Air massif (an area known as Takolokouzet), situated 160 km north of the Gobero site complex Figure 1A). The exposed rock shows the characteristic green hue, variable lamination, and vesicular texture common to many of the felsite lithics from Gobero. (B)-Abundant debitage as well as large groundstones attest to a longstanding knapping operation at Alallaka. Scale bar equals 3 cm. doi:10.1371/journal.pone.0002995.g008   Long distance transport is low, as indicated by the paucity of Mediterranean taxa [15]. Crushed or damaged pollen grains that could be identified were summed in the relevant taxa. Only pollen that could not be identified were included in the deteriorated sum. Abbreviations: CONISS, constrained incremental sum of squares; p/g, pollen per gram. doi:10.1371/journal.pone.0002995.g010

Pollen Analysis
A total of 59 pollen samples were collected from 23 burials (Table 9). Pollen samples were prepared using 10% Napyrophosphate, sieving using 200-300 mm and 7 mm meshes, 10% HCl, acetolysis, heavy liquid separation (Na metatungstate hydrate with specific gravity of 2.0 and centrifugation at 2000 rpm for 20 minutes), 40% HF, and mounting with glycerol jelly on permanent slides. Lycopodium spores were added to calculate pollen concentration (pollen grains per gram = p/g), and pollen analyses were performed at 4006and 10006with immersion oil for critical determinations. Graphs are drawn with Tilia and TGView [52], and results are presented using pollen sums of 300-400 grains ( Figure 10).
Preliminary data on early Holocene pollen were obtained from seven samples from three burials on site G3 (G3B17, G3B8, G3B9). Concentrations are less than 500 p/g, and two of the samples are sterile. Most pollen is crumpled. Poaceae comprise half of the spectra with a significant amount of Cyperaceae and Cornulaca. Arboreal pollen includes Ficus, Tamarix and Myrtus. Aquatics include Juncus, Nymphaea and Potamogeton, and many hyphas and fungal spores were observed in G3B17 and G3B8. Microscopic charcoal is present (.125 mm) in burial in G3B8, suggesting the presence of a hearth in the vicinity of the burial.
Mid-Holocene pollen was analyzed on the basis of 13 samples from burials at sites G1 and G3 (Table 9). Pollen is present in all samples, although concentrations are very low, frequently less than 300 p/g and always less than 600 p/g. The diversity of pollen taxa per sample are also low (minimum 14, maximum 32). When samples from a single burial are combined, diversity remains low in burial G1B18 (53 taxa) but increases in the other two burials (68-78 taxa). Pollen clusters of several taxa are present in sample 3a between the arms of the triple burial (Table 9). Pollen spectra are herb-dominated with prevalence of Poaceae (ca. 60% on average). Cyperaceae and Cornulaca are common with significant percentages in many samples. Cornulaca, for example, is greater than 30% in samples 3a (G1B8-10) and 4b and 5b (G1B18). The ratio D-desert / Sh-Sahelian elements was less than 1.0, the only exception being sample 4b from burial G1B18. Arboreal pollen, which is always less than 5% of total pollen, includes the ubiquitous Sudanian Ficus plus Moraceae undiff. and the Sahelian Salvadora and Ziziphus ( Figure 10B). The Sahelian Capparis is present in burials G1B18 and G3B36. Desert trees such as Acacia are rare, occurring only in sample 1a of the triple burial G1B8-10. Sudanian alluvial or riverine vegetation, such as Celtis, Ficus, Salvadora persica and Hyphaene thebaica, is more common in G1B8-10 (1.5%) than in the other two burials (0.5-0.8%). Saharo-Sindian shrubs and herbs, such as Cornulaca, Cichoriae undiff., Ephedra, Plantago, Heliotropium, Zygophyllum, are widespread in the spectra. Pioneer elements, such as Compositae, Polycarpaea and Plantago, are present especially in burial G1B8-10. Boerhavia-type pollen that characterizes a semidesert environment is significant (3.1%) in sample 1b from burial G1B18. Celosia-type trygina (wool flower) was unusually common (32.4%) in sample 3a from burial G1B8-10, where it occurred in clusters.
Herb plants of wet environments, which measured 0.6-0.7% excluding Cyperaceae, were found in the majority of samples and were slightly more common in burial G3B36. The hydrophytes Lemna and Potamogeton, both of which have submerged and floating leaves, were present in G1B18 and G3B36 ( Figure 10C). Typha (cat tail) was found only in trace amounts in sample 2c (G3B36). Algal elements were observed in G1B8-10 and G1B18 (here also including Concentricystes which are algal spores of Zignemataceae) [53]. On the other hand, graminoid-type phytolits, which are silicate or ossalate inclusions of plant epidermids, were abundantly observed in sample 1b from burial G1B18.
Pollen data from the burials are fairly concordant for interpreting the Gobero plant landscape, which was largely dominated by an open grass savanna [15]. Sahelian elements (Poaceae+Cyperaceae) predominate over desert vegetation (Che-nopodiaceae+Asteraceae). Under conditions of increasing aridity and near salty environments, the Sahelian savanna shifted toward   Sudanian and tropical elements, such as Celtis, Ficus, Salvadora persica, Hyphaene thebaica, could have been part of riverine forests that also included Combretum, Cadaba, Ziziphus, Pterocarpus and Capparis. None of these taxa, however, occur in very high frequencies among non-arboreal pollen ( Figure 10C). Pollen from burial samples, in sum, suggests that the landscape was a mosaic of xeric and wet environments.