sites on the landscape: paleoenvironmental context of late ... · sediments are shown in gray;...

at SciVerse ScienceDirect

Quaternary International xxx (2013) 1e11

Contents lists available

Quaternary International

journal homepage: www.elsevier .com/locate/quaint

Sites on the landscape: Paleoenvironmental context of late Pleistocenearchaeological sites from the Lake Victoria basin, equatorial East Africa

C.A. Tryon a,*, J.T. Faith b, D.J. Peppe c, W.F. Keegan d, K.N. Keegan e, K.H. Jenkins f,S. Nightingale g, D. Patterson h, A. Van Plantinga i, S. Driese c, C.R. Johnson j, E.J. Beverly c

aDepartment of Anthropology, Peabody Museum of Archaeology and Ethnology, Harvard University, 11 Divinity Avenue, Cambridge, MA 02138, USAb School of Social Science, The University of Queensland, Brisbane, QLD 4072, AustraliacDepartment of Geology, Baylor University, One Bear Place #97354, Waco, TX 76798, USAdHeritage Consultants, LLC, 877 Main Street, Newington, CT 06111, USAeDepartment of History, Fairfield University, 1073 North Benson Road, Fairfield, CT 06824, USAfDepartment of Anthropology, University of Minnesota, 395 Hubert H. Humphrey Center, 301 19th Avenue South, Minneapolis, MN 55455, USAgDepartment of Anthropology, Graduate School and University Center, City University of New York, 365 Fifth Ave., New York, NY 10016-4309, USAhHominid Paleobiology Doctoral Program, Center for the Advanced Study of Hominid Paleobiology, Department of Anthropology, The George WashingtonUniversity, 2110 G Street NW, Washington, DC 20052, USAiDepartment of Geology and Geophysics, MS 3115, Texas A&M University, College Station, TX 77843-3115, USAjDepartment of Anthropology, University of Connecticut, Box U-2176, Storrs, CT 06269, USA

a r t i c l e i n f o

Article history:Available online xxx

* Corresponding author.E-mail addresses: [email protected]

(J.T. Faith), [email protected] (D.J. Pconsultants.com (W.F. Keegan), [email protected] (K.H. Jenkins), [email protected] (D. Patterson), [email protected]@Baylor.edu (S. Driese), cara.johnson@ucon(C.R. Johnson), [email protected] (E.J. Beverly

1040-6182/$ e see front matter � 2013 Elsevier Ltd ahttp://dx.doi.org/10.1016/j.quaint.2013.05.038

Please cite this article in press as: Tryon, C.Afrom the Lake Victoria basin, equatorial Eas

a b s t r a c t

Open-air archaeological sites record only a small fraction of the behavioral traces of mobile foragerpopulations. Whereas caves and rockshelters were often occupied at least in part for protection from theelements, the reasons why human foragers occupied other places on the landscape (however briefly) arevaried and not always readily recoverable. We develop a framework for interpreting human use of thelandscape and modeling occupation of open-air sites using the archaeological and paleoenvironmentalrecord of Middle Stone Age (MSA) sites from Rusinga and Mfangano Islands, located near the easternmargin of Lake Victoria. Paleoenvironmental reconstructions using fossil faunas suggest an arid grasslandsetting unlike the present. Paleoecological modeling of the habitats of extant and extinct bovids, com-bined with GIS-based reconstructions of lake level change, indicate that human occupation of these sitescoincided with substantial declines in the level of Lake Victoria. During this time, both Rusinga andMfangano would have been connected to the mainland and represented local topographic highs withinan extensive grassland. Geological, ecological, and ethnobotanical observations suggest that thesetopographic high points would likely have been important sources of stone raw material, fresh water,and a variety of plant resources for food, fuel, and other purposes. In contrast, the grassy lowland plainswere probably exploited primarily as a source of large game, which included numerous species of largegregarious grazers, several of which may have followed now extinct migration routes.

� 2013 Elsevier Ltd and INQUA. All rights reserved.

1. Introduction

In part because they provide shelter from the elements, cavesand rockshelters often preserve long archaeological sequencesattesting to successive human visitations. In addition to recurrent

(C.A. Tryon), [email protected]), wkeegan@heritage-u (K.N. Keegan), holt0452@(S. Nightingale), dbpatter@(A. Van Plantinga), Steven_n.edu, [email protected]).

nd INQUA. All rights reserved.

., et al., Sites on the landscapet Africa, Quaternary Internati

usage, caves and rockshelters favor longer-term occupations, thematerial residues of which may sample a wide range of thebehavioral repertoire of any given population, and these behavioraltraces are often better preserved than in other settings. For theseand other reasons, caves and rockshelters have played a central rolein Middle Paleolithic/Middle Stone Age (MP/MSA) archaeology,particularly in regions such as Western Europe, the Levant, andsouthern Africa. However, among mobile groups of human foragersin tropical climates, the majority of daily activities are performedoutside of caves and rockshelters, creating a larger, potentiallymore diffuse and varied archaeological record. The material culturedocumented at open-air archaeological sites may reflect a range oftime spans, from minutes to decades, and result from a number of

: Paleoenvironmental context of late Pleistocene archaeological sitesonal (2013), http://dx.doi.org/10.1016/j.quaint.2013.05.038

mailto:[email protected]

















www.sciencedirect.com/science/journal/10406182

http://www.elsevier.com/locate/quaint

http://dx.doi.org/10.1016/j.quaint.2013.05.038



25˚ 30˚ 35˚ 40˚

-5˚

0˚

5˚

0 200 400

Lake Tanganyika

Lake Kivu

Lake Edward

Lake AlbertLake Turkana

Lake Naivasha

KENYA

TANZANIA

UGANDA

D.R.C

RWANDABURUNDI

MALAWI

ZAMBIA

ETHIOPIASUDAN

Inset

000 2000000000 4004004040444 000000LAAAAAAWIWWWWWWWWAA

ZAMZAMMAMAZAZAMAMZAMAA BIAIAIIIAIIABIAAAABIAA

LAAAA

3 km

N

A.

B.

Lake Victoria

Fig. 1. A. Schematic map showing Lake Victoria in relation to other key geographicfeatures of eastern Africa. B. Detail of Rusinga Island and Mfangano Island. Pleistocenesediments are shown in gray; hatched areas are land >100 m above present lake level.

C.A. Tryon et al. / Quaternary International xxx (2013) 1e112

possible activities, including habitation, social gatherings, andacquiring stone, plant, or animal resources, among other possibil-ities. The contribution of open-air MSA sites to our understandingof hominin behavior is particularly important in areas such aseastern Africa where caves and rockshelters are relatively rare.

Key elements in the study of open-air archaeological sites arethe natural and cultural factors that led humans to choose to useparticular places on past landscapes. For more recent time periods,research has included catchment analyses to examine the spatialrelation between sites and available resources (e.g., Jarman et al.,1972), and more formal predictive modeling of site location basedon resource distribution and ethnographic or ethnohistoricbehavior of foragers or farmers (e.g., Gaffney and Stan�ci�c, 1996;Westcott and Brandon, 2000). These approaches can be difficult toapply to the Paleolithic record, where the modern landscape maybear little resemblance to the ancient one.

Among Early Stone Age sites in eastern Africa, several long-termarchaeological projects have examined variation among archaeo-logical sites and reconstructed paleoenvironments within relativelynarrow temporal intervals to identify favored places on the land-scape (e.g., Potts et al., 1999; Braun et al., 2008; Blumenschine et al.,2012). Comparable landscape-scale reconstructions for MSA sites ineastern Africa have yet to be done, although there have been effortsin this direction (e.g., Barboni et al., 1999; Ambrose, 2001; Yellenet al., 2005; Assefa et al., 2008; Tryon et al., 2010, 2012; Faith et al.,2011; Brown et al., 2012). Here, we build on these efforts toexplore the reasons for open-air site occupation by MSA homininson the easternmargins of the LakeVictoria basin inEquatorial Africa,drawing on archaeological evidence as well as approaches derivedfrom ecology, geology, historical geography, and ethnobotany.

Lake Victoria is the largest body of water in Africa and the sur-rounding area today supports diverse flora, fauna and dense humanpopulations. However, the lake is very sensitive to variation inclimate, and multiple lines of evidence indicate Lake Victoria hascompletely dried up repeatedly during the Pleistocene. As such, thepresent is often a poor analog for the past. We emphasize here MSAartifacts and fossils from the Kenyan islands of Rusinga and Mfan-gano that date to one or more of these lake low-stands some timebetween w100 ka and w33 ka. During MSA hominin occupation,Rusinga andMfanganowere likely topographic high points within agrassy plain. In the absence of an adjacent lakemargin, hominin useof the area instead may have been driven by the availability of anumber of other resources, including fresh water from springs,good quality stone raw material, hunting overlook sites that likelyintersected with migratory game paths, and the presence of plantmaterials not found in the neighboring grasslands.

2. The paleoenvironmental history of the Lake Victoria region

2.1. Evidence from lacustrine deposits

Lake Victoria (Fig. 1) is the largest lake in Africa as measured bysurface area (66,400 km2). It started to form �400 ka when tectonicuplift along thewesternmargins of the Rift Valley caused progressiveback-ponding of westward flowing rivers that previously cut acrossthe basin (Bishop and Trendall, 1967; Johnson et al., 1996). Lake Vic-toria is also very shallow, with a maximum depth of 68 m, and thusmodest changes in lake level can result in large increases or decreasesin lake surface area. Seismic profiles across the lake show four surfacesthat mark former periods of lake desiccation. The uppermost of thesedesiccation surfaces dates tow15e18 ka (Stager et al., 2011), whereasthe lowermost is likely the pre-lake surface developed on basementrock andMiocene sediments. The remaining two surfaces are undated(Johnson et al., 1996), although Stager and Johnson (2008) suggest apossible Late Pleistocene age of 80 ka for one of these surfaces. Raised

Please cite this article in press as: Tryon, C.A., et al., Sites on the landscapefrom the Lake Victoria basin, equatorial East Africa, Quaternary Internati

beach deposits and wave-cut terraces at þ3, þ10e12, and þ18 mabove current water level attest to periods when Lake Victoriaexpanded, with only theþ3m terrace dated tow4 ka (Temple,1967).

Today, Lake Victoria’swater level is primarily controlled by rainfall(Piper et al., 1986; Sene and Plinston, 1994; Broeker et al., 1998), andthus the size and depth of the lake serves as a fairly sensitive raingauge. Sedimentological, isotopic, andmicrofossil (pollenanddiatom)data from multiple lake cores, as well as historical bathymetric re-cords, confirm that lake level has long been strongly tied to localprecipitation. For example, Lake Victoria’swater level declined duringthe Last Glacial Maximum or LGM (26.5e19/20 ka; Clark et al., 2009),fell rapidly with complete desiccation during Heinrich Event 1 (17e16 ka; Hemming, 2004), and expanded during the peak of the Holo-ceneAfricanHumidPeriod (15e5 ka; deMenocal et al., 2000), andhasvaried�3m indepth since the 1880s (Kendall,1969; Nicholson,1998;Stager et al., 1997, 2002, 2011). Pollen and isotopic data indicate theexpansion and contraction of adjacent grasslandswith fluctuations inrelative aridity (Kendall, 1969; Talbot and Livingstone, 1989; Talbotand Laerdal, 2000, Talbot et al., 2006). These data suggest similarlinkages among variability in rainfall, lake level, and the surroundingvegetation throughout the history of Lake Victoria.

2.2. Evidence from terrestrial deposits on Rusinga and MfanganoIslands

Despite good data for the history of Lake Victoria since the LGM,there is little detailed evidence for the pre-LGM history of the lakeor the surrounding region. This paucity of evidence exists becausesediment cores collected from Lake Victoria do not penetratebeyond the Heinrich Event 1/LGM lowstand surface and few pre-


C.A. Tryon et al. / Quaternary International xxx (2013) 1e11 3

LGM deposits have been identified or studied. However, artifact-bearing Late Pleistocene terrestrial sediments from Rusinga andMfangano Islands within Lake Victoria (Fig. 1) provide key com-plementary information on the pre-LGM natural and cultural his-tory of the region (Tryon et al., 2010, 2012).

Although connected to mainland Kenya by an artificial causewaysince the 1980s, Rusinga Island (w40 km2) was formerly separatedfrom the mainland by a shallow (w5e10 m depth) and narrow(w250 m wide) channel, whereas Mfangano Island (64 km2) liessome 10 km from the shore of Lake Victoria (Fig. 1). Maximum ele-vations above the lake are 297 m for Rusinga Island and 541 m forMfangano Island, with most of Mfangano dominated by steep lavauplands (Fig. 1). The area today receivesw800e1000 mm of rainfallper year, and prior to substantial recent human impacts, the islandswere likely covered by variably dense woodlands (Andrews, 1973).Both islands are the remnants of deposition related to the eruptivehistory of the Kisingiri volcano during the Miocene, and share asimilar bedrock lithology of Miocene lavas and volcaniclastic rocks(e.g., Van Couvering, 1972; Pickford, 1986; Drake et al., 1988; Peppeet al., 2009). On Rusinga and Mfangano Islands, poorly consoli-dated Pleistocene sediments accumulated at the base of each island’scentral uplands andunconformably overlieMiocenedeposits (Fig.1).These Pleistocene deposits are the Wasiriya Beds on Rusinga Islandthat crop out over an area of �10 km2, and the Waware Beds onMfangano Island that are exposed over an area of �4 km2

(Whitworth, 1961; Van Couvering, 1972; Pickford, 1986; Tryon et al.,2010, 2012). TheWasiriya andWaware Beds are primarily tuffaceousalluvial andfluvial sediments andpaleosols. Rare tufa deposits in theWasiriya Beds indicate the former presence of springs at theNyamitalocality onRusinga Island (Fig.1). The generally poorly sorted, gravel-and pebble-dominated fluvial sediments are consistent with flashy,high energy, episodic deposition and erosion and suggest seasonalprecipitation. The features identified in clay-rich paleosols onRusinga (Beverly et al., 2012) also indicate precipitation seasonality,as well as seasonal soil moisture deficit (Wilding and Tessier, 1988;Coulombe et al., 1996). The presence of pedogenic slickenside

Table 1Artifact counts from 2009 to 2010 field seasons on Rusinga and Mfangano Islands.

Locality Wasiriya Beds, Rusinga Island

Nyamita Wakondo Nya

Approximate site size (km2) 0.2 0.2 0.5Area excavated (m2) 4 4 0Artifact typeComplete flakesLevallois/Levallois related 5 1 (1) 3Non-Levallois flake 20 (3) 15 (2) 12Blade 0 1 0

Proximal flake fragmentsLevallois/Levallois related 8 (1) 2 1Non-Levallois flake 48 (6) 21 17Non-Levallois blade 0 (1) 0 0Flaking debris 55 (13) 25 (6) 14

Cores and core fragmentsLevallois 2 (1) 0 4Multiplatform 3 0 2Casual 2 (1) 1 1Discoidal 5 1 5Single platform 0 0 1Indet. 0 0 1

Shaped and retouched piecesBiface 0 0 1Point 3 0 1Retouched Levallois flake or fragment 1 1 (1a) 0Retouched non-Levallois flake or fragment 5 0 0

Total 193 78 63

Numbers in parentheses indicate artifacts recovered from excavation.a Artifact found in situ but not from the excavation.


surfaces, wedge-shaped peds, root traces and the abundance ofsmectitic clay identify many of the paleosols as Vertisols (Buol et al.,2003; Southard et al., 2011) that formedwhen the soil expanded andcontracted during repeated wet and dry cycles.

The Wasiriya and Waware beds are constrained to approxi-mately 100e33 ka. Geochemical analyses of tephra from the islandssuggest derivation from volcanic sources 250 km to the east, fromvolcanoes in the East African Rift Valley System that began eruptingw100 ka (Tryon et al., 2010). The timing of these eruptions providesa maximum age for the deposits. A suite of calibrated radiocarbonage estimates on the shells of intrusive gastropods (Limicolaria cf.L. martensiana) from both the Wasiriya and Waware Beds provideminimum ages ofw33e45 ka for the deposits. These age estimatesdemonstrate the pre-LGM age of the Pleistocene sediments onRusinga and Mfangano Islands. The position of these sedimentaryarchives on islands within Lake Victoria make them importantpieces of evidence to study changes in the lake margin as a result ofexpansion or contraction over the last w100 kyr.

2.3. Pleistocene archaeological sites on Rusinga and Mfanganoislands

In situ Stone Age archaeological sites within the Wasiriya Bedsand Waware Beds demonstrate hominin occupation of Rusinga Is-land and Mfangano Island during one or more intervals betweenw33 ka andw100 ka. A comprehensive survey of all Wasiriya Bedsoutcrops has shown that artifacts on Rusinga Island are largelyconcentrated at three localities (Nyamita, Wakondo, and Nyam-singula), with a single locality (Kakrigu) identified during a simi-larly detailed pedestrian survey of the Waware Beds on MfanganoIsland (Fig. 1). As indicated by surface collections and limited testexcavations (Table 1), artifacts at these open-air sites are neitherabundant (n ¼ 351) nor dense (see Tryon et al., 2010, 2012 forcollection methodology). On Rusinga (n ¼ 339) they include typo-logically Middle Stone Age (MSA) artifacts such as unifacially andbifacially flaked points, recurrent and preferential Levallois cores,

Waware Beds, Mfangano Island

msingula Other Total Kakrigu Kaswanga Total

2 0.20 8 0 0 0

0 10 1 0 10 52 3 0 30 1 0 0 0

1 13 0 0 00 91 2 0 20 1 0 0 01 124 1 0 1

3 10 0 0 00 5 0 0 00 5 0 0 00 11 2 1 30 1 0 0 00 1 0 0 0

1 2 1 0 10 4 1 0 10 3 0 0 00 5 0 0 06 339 11 1 12


Fig. 2. Lithic artifacts from the Wasirya and Waware Beds of Rusinga and MfanganoIslands. (a) Chert overshot flake from a Levallois core; striking platform preserved onproximal, distal and right margins. (b) Quartz bifacial piece, point (c) Chert bifacialpoint. (d) Lava Levallois core (Nubian Type I). (e) Scraper on lava Levallois flake frag-ment. All from Rusinga Island except (b).


and Levallois flakes, some retouched as scrapers, among other flake,tool, and core types (Fig. 2). The Mfangano collections are smaller(n ¼ 12) and contain bifacial points, but lack evidence for Levalloistechnology (Fig. 2). TheMfangano sample can likely be attributed tothe MSA as well, but a larger sample size would aid attribution.

Stone artifacts co-occur with fossil fauna at all localities onRusinga and Mfangano Islands, and our surveys suggest that thisassociation is significant, as concentrations of only artifacts or faunaare rare. The association of stones and bones on the landscape maysignal the presence of a resource, such as water, that attractedhominins and other animals to these areas. AtWakondo on RusingaIsland, cut-marked fossils demonstrate direct hominin involvementin the taphonomic history of some of the fauna (Tryon et al., 2010;Jenkins et al., 2012).

3. Modeling past lake conditions using terrestrial fauna andbathymetry

Fossil fauna are both relatively abundant and well preserved onRusinga Island and to a lesser extent, Mfangano Island (NISP>450),and provide some of the best evidence for past environmentaldifferences. We combine evidence from the fauna for conditionsthat are drier, more open, and grassland-dominated than the pre-sent with detailed bathymetric studies of Lake Victoria tomodel theeffect of increased aridity on lake level to better understand thenature of the landscape available to MSA hominins.

3.1. Faunal evidence for reduced lake levels

The fossil mammals from Rusinga Island and Mfangano Islandindicatemore arid conditions than the present, and in particular, anexpansion of dry grassland habitats during the Late Pleistocene(Faith et al., 2011; Tryon et al., 2012). Alcelaphine bovids dominateall fossil assemblages from Rusinga and Mfangano, indicating thepresence of open, dry grasslands (e.g., Vrba, 1980). A grassland


ecosystem is further supported by the locally abundant and well-preserved sample of microfauna from Rusinga Island (NISP >50),which includes Tachyoryctes (African Mole Rat) and Otomys(Groove-Toothed Rat), with extant representatives of Tachyoryctescharacteristics of open grasslands or thinly treed savannas(Kingdon, 1974; Nowak, 1999). Some extant taxa from Rusinga Is-land, such as the oryx (Oryx beisa) and Grevy’s zebra (Equus grevyi),are distributed well beyond their modern ranges, suggesting rangeexpansions or shifts during one or more arid intervals of thePleistocene (Tryon et al., 2010; Faith et al., 2013). In light of the�3 m variation in lake level since the late 19th century (Nicholson,1998), the shallow (w5e10 m) and narrow (250 m) channel sepa-rating Rusinga Island is unlikely to have served as a long-termbarrier to the movement of land animals between the island andthe mainland. However, Mfangano is w10 km from the shore, andseparated from it by a w25-m-deep channel (Fig. 1). Despite this,fossils of several species of large gregarious grazers are present inMfangano Island’s Pleistocene sediments (Tryon et al., 2012),including plains zebra (Equus quagga), wildebeest (Connochaetestaurinus), and African buffalo (Syncerus caffer), in addition to extinctlong-horn buffalo (Syncerus antiquus) and two arid-adapted alce-laphine antelopes (Rusingoryx atopocranion and Damaliscus hyp-sodon) (Faith et al., 2011, 2012). The migratory habits of gregariousgrazers such as wildebeest e driven by seasonal variation inresource availability e are well-documented (Kingdon, 1997;Skinner and Chimimba, 2005) and similar migrationsmay have alsocharacterized D. hypsodon (Faith et al., 2012). It is doubtful thatmigratory species would establish island populations that are un-able to migrate in response to seasonal resource variation.Furthermore, the landmass of Mfangano Island encompasses amere 64 km2, much of which is dominated by the steep topographyof Mount Kwitutu, a size insufficient to support viable populationsof gregarious migratory game.

Because Lake Victoria is largely rainfall-fed, the presence of arid-adapted species and migratory grazers in the fossil fauna imply asubstantial reduction in lake level and precipitation. The WasiriyaBeds and Waware Beds thus preserve evidence of hominin occu-pation of the areawhen Lake Victoriawas smaller than present, andpossibly absent.

3.2. GIS reconstructions

In order to model the effects of lake level reduction on thelandscape implied by the fossil fauna, we used geographic infor-mation systems (GIS) analysis to reconstruct the bathymetry ofLake Victoria in the vicinity of Rusinga and Mfangano Islands. GISanalysis uses precise data that conform to established mathemat-ical models of the shape of the earth and, where elevation is aconcern, to a known zero elevation point. The choices of spheroid(model of the ellipsoid shape of the earth) and datum (model of theelevation irregularities of the earth’s surface) can significantlyaffect the accuracy of measurements, especially if data from mul-tiple sources using different models are used.

We used three sources of data on the shoreline, islands, anddepths of Lake Victoria. (1) The elevation of the current shorelineand nearby features (including islands) is based on scanned anddigitized Survey of Kenya (1979) topographic maps of the lake areathat have a known spheroid and datum. (2) Topographic andbathymetric data were collected from British Admiralty maps fromthe early twentieth century (UK Hydrographic Office, 1902e1956,1911e1956). The spheroid and datum for these maps are also areknown, with lake surface reported as measured by a tide gauge atPort Florence in Kisumu, Kenya as 3726 ft (1135 m) above LWOS atMombasa. LWOS refers to “Low Water Ordinary Springs,” an esti-mate of sea level based on observations taken of low tides during



non-equinoctial periods in the spring (United States HydrographicOffice, 1915). The depth values provided by the Admiralty charts,like most bathymetry prior to the development of global posi-tioning systems (GPS), represent minimum values of a group ofsoundings, because the primary users of these charts were mari-ners wishing to avoid running aground (Monmonier, 2008). (3)Additional bathymetric data collected using a hydroacousticsounder and GPS are derived from the Lake Victoria FisheriesResearch Project (LVFRP) described by Silsbe (2003). Althoughdepth recordings are dense (approximately 15 million points), theutility of these data are restricted because the bathymetry pointscollected by different methods could not be distinguished, theinterpolation method, datum, and spheroid used are unknown, andthe elevation zero point used is not stated. These uncertaintiesintroduce unknown systematic errors into all elevation calcula-tions, and the LVFRP data set was used to supplement the moreaccurate and precise bathymetric data from the Admiralty Charts.

The Admiralty and Survey of Kenya maps were georeferenced tothe New Arc 60 datum (meters) in ArcGIS 9.3, and the bathymetricand topographic data they contain were digitized as points andlines and converted into meters. The LVFRP bathymetric data werealso projected into New Arc 60 in ArcGIS 9.3, using an assumptionthat its original projection was WGS84; the resulting spatial matchbetween the two data sets was good enough to proceed. We thenedited the data set to eliminate land areas and anomalous datapoints, and rounded all the remaining values to the nearest meterto simplify computation. Because the bathymetric data set was stillvery large, we imported it into a different GIS program, IDRISI(produced by Clark University, USA), which converted the data fromvector format (points) into a raster format, and further examinedand edited them for anomalous points and errors. The projectedbathymetric data consisted of a set of negative numbers repre-senting distance from the lake surface, to which we added 1135 m

Fig. 3. Topographic and bathymetric map of the region near Rusinga and Mfangano Islands10 m intervals, reported in m above sea level (asl). The present shoreline at 1135 m asl is sh1130 m asl); Mfangano Island connects to the mainland with a 25 m drop in lake level (toMfangano Island; mainland topography is simplified.


to each (negative) bathymetric value using IDRISI’s map algebrafunction, yielding an elevation relative to the lake surface level. Thedata were then exported back into ArcGIS as vector point datawhere the points of the bathymetric data set and the lines of thetopographic data set were combined into a single triangulatedirregular network (TIN) to permit areal analysis and the display ofareas that would be above water at varying lake levels.

Mapping results are shown in Fig. 3, and suggest that a drop inwater level of�5m (contour interval 1130m) is required to connectRusinga Island to the mainland, a reduction just beyond the �3 mvariation recorded since the late 19th century (Nicholson, 1998). Adecline of 25 m (contour interval 1110 m) is sufficient to linkMfangano with the mainland (Fig. 3), providing a landmass ofsuitable size for large herds of migratory game implied by theobserved fossil faunal assemblage. These values represent mini-mum estimates of lake margin location and lake size during hom-inin occupation of Rusinga and Mfangano Islands. Whether LakeVictoria was absent or if a reduced version of it was nearby duringhominin occupation is presently unknown, but estimates of pastrainfall budgets suggest that massive depth declines (e.g., �25 m)would have set in motion a series of feedback mechanisms thatcould have led to a significantly reduced or even completelydesiccated the lake (Broeker et al., 1998; Milly, 1999). Either way,the environment of the area surrounding the present-day islandswould have been substantially different from the moderncondition.

3.3. A landscape without Lake Victoria: lowlands, uplands, andgrazing lawns

Lake level decline or desiccation would have substantiallyaltered the landscape and resources available to hominins in thevicinity of Rusinga and Mfangano Islands. Rather than islands,

, showing gentle offshore slopes exposed with lowered lake level. Contours shown atown in bold. Rusinga Island connects to the mainland with a 5 m drop in lake level (to1100 m asl). Maximum elevation on Rusinga Island is 1432 m asl and 1676 m asl on



Rusinga and Mfangano would have represented one of severaltopographic high points on the landscape of the Lake Victoria basin.With the reduction of 25 m required to connect Mfangano andRusinga to the mainland, all of the other islands in Lake Victoriathat have published evidence of MSA hominin occupation wouldalso become connected to the mainland (Fig. 4) and similarlyrepresent local topographically high areas. These islands includeLolui, Buvuma/Bugaia, and the Sese Islands in Uganda (Fagan andLofgren, 1966; Nenquin, 1971; Posnansky et al., 2005). This sug-gests that (1) models developed for Rusinga and Mfangano mayalso be applied elsewhere in the Lake Victoria basin, and (2) MSAdeposits on these widespread islands near the margin of LakeVictoria around much of its circumference may record the samearid interval represented by Rusinga and Mfangano. Additionalfieldwork and chronometric data are needed to test the latterhypothesis.

As shown in Fig. 3, the land near Rusinga and Mfangano Islandsexposed with lake level decline is characterized by a shallowtopographic gradient. This trend is seen at a larger scale in Fig. 4,which suggests that much of the center of the basin is relatively flat,with the�60 m isobath spanning an area of>100 km across. As thelake is �68 m at its deepest, this implies a grade of 0.01% acrossmuch of the center of the basin. Although possibly traversed byseasonal slow-flowing streams or rivers when the lake was absent,much of the center of the basin was probably relatively flat low-lands with present-day islands occurring as isolated areas of higherelevation terrain near the basin margin. The central part of thebasin was likely a grassland-dominated lowland area during lakedesiccation. This is suggested by (1) seismic reflection data thatindicates a relatively smooth topography underlying the HeinrichEvent 1/LGM desiccation surface that is in places traversed bystreams and rivers, (2) the abundance of hypsodont alcelaphinebovids including wildebeest and a variety of extinct forms such as

Fig. 4. Schematic bathymetric map of Lake Victoria showing water levelat þ18 m, �20 m, and �60 m, after Temple (1966), Stager et al. (1997), and Verschurenet al. (2000). Also shown are the locations of other islands with Middle Stone Ageartifacts.


Rusingoryx atopocranion and D. hypsodon from the Wasiriya andWaware Beds on Rusinga and Mfangano Islands, and (3) compari-son with data from lake cores dating to the LGM and Holocenesuggesting a general trend of increasing grass cover with lake leveldecline in the vicinity of Lake Victoria (Kendall, 1969; Talbot andLivingstone, 1989; Scholz et al., 1998; Talbot and Laerdal, 2000,Talbot et al., 2006).

Large mammalian herbivores have numerous effects on plantcommunities. One of the more significant effects of herd-forminggrazers (e.g., wildebeest) in African grasslands is the establish-ment of grazing lawns (McNaughton, 1984; Cromsigt and Olff,2008), which are intensely grazed patches of grazing-tolerantgrass species. Feedback mechanisms result in grazers enhancingboth the quality (nutrient content) and quantity (primary produc-tivity) of the grazed vegetation, which in turn promotes grazinglawn formation. Grazing lawns are characteristic of East Africangrasslands where large gregarious grazers are present (e.g., theSerengeti) (McNaughton, 1985).

Grazing lawns are not ubiquitous across the landscape becausethe distribution of grazing species is limited by both abiotic andbiotic factors. Among abiotic factors, topography plays an impor-tant role in mediating the distribution of grazing herbivores (Bell,1971; Coughener, 1991; Bailey et al., 1996; Fowler, 2002), withtopographically rugged/steep areas serving as a barrier to largegrazers. Grazing lawns cannot be established in such areas oflimited grazing pressure, which in turn allows for the developmentof woody vegetation types. This can be seen here in Fig. 5, whichshows the distribution of grazing lawns and woody vegetation inrelation to topographic breaks in the greater Serengeti ecosystem(Ngorongoro Conservation Area, Tanzania). There is a clearboundary between the expansive grazing lawns on the flatlandsand the woody vegetation on the topographically steep/ruggedhighlands formed by the Oldonyo Gol Hills.

Similar vegetation heterogeneity would have also characterizedthe late Pleistocene ecosystems in the vicinity of Rusinga andMfangano Islands, and the fossil bovid assemblages of Rusinga andMfangano are very similar to extant bovid communities from theSerengeti (Tryon et al., 2010, 2012). The dominance of grazing un-gulates (14 species in Tryon et al., 2012) suggests abundant grass-land habitats, and the presence of gregarious species such aswildebeest, zebra, and buffalo would have contributed to theestablishment of expansive grazing lawns in the flatlands. At thesame time, however, the numerous topographically steep volcaniccones in the region (Pickford, 1972; see also Fig. 3) and parts ofRusinga and Mfangano Islands likely supported densely vegetatedwoody habitats, perhaps similar to those in Fig. 5. The combinedevidence suggests that what are now Rusinga and MfanganoIslands were during the Late Pleistocene local topographic highpoints that were more densely vegetated than an adjacent grassyplain. The Pleistocene Wasiriya Beds and Waware Beds samplesediments near the foot of these upland areas.

3.4. Implications for hominin mobility, group size, and landscapeuse

Ethnographic observations demonstrate that environmentalfactors, namely resource predictability and density, play a centralrole in the socio-territorial organization of hunter-gatherers (e.g.,Dyson-Hudson and Smith, 1978; Ambrose and Lorenz, 1990; Kelly,1995; Marean, 1997). These observations provide a framework forpredicting broader patterns of MSA hominin mobility, group size,and landscape use during the Late Pleistocene within the LakeVictoria basin.

Multiple lines of evidence suggest that the landscapes exploitedbyMSAhomininswould have been characterized by highly seasonal


Fig. 5. Distribution of grassland vegetation (grazing lawns) and wooded vegetation in relation to marked shifts in topography at Ngorongoro Conservation Area, Tanzania.


arid or semi-arid grasslands that were seasonally populated byherds of large migratory game, including zebras, wildebeest, andseveral extinct species. Such nomadic herds of gregarious gamerepresent high-density resource patches, but their dailymovementstogether with seasonal migrations render them unpredictable inspace and time. The ethnographic record suggests that the combi-nation of dense but unpredictable resource patches should promotean adaptive strategy characterized by large home ranges lackingfixed boundaries, with high residential mobility and opportunisticsite relocation (Ambrose and Lorenz, 1990). Human populationdensity should vary from low tomoderate, depending on terrestrialprimary productivity. The presence of arid-adapted ungulates onRusinga and Mfangano may signal the presence of arid grasslands(<500 mm annual rainfall), in which case both terrestrial produc-tivity andhumanpopulationdensitieswould bequite low.However,this conflicts with the impressive diversity of the fossil ungulatecommunity (24 species) fromRusinga andMfangano Islands, whichmay signal a more productive semi-arid grassland (>500 mmannual rainfall) (Faith, 2013) and perhaps moderate human popu-lation densities.

From an archaeological perspective, highly mobile and wide-ranging MSA hominin populations at low-to-moderate popula-tion densities should be associated with a diffuse archaeologicalrecord characterized by sites reflecting ephemeral and low-intensity occupation. Following Ambrose and Lorenz (1990), interms of raw material exploitation, this adaptive strategy shouldbe associated with the use of both local and exotic raw materialsources.

4. Resource magnets on Rusinga and Mfangano

One or more natural resources served as ‘magnets’ to repeatedlydraw hominins to Rusinga and Mfangano Islands and thus producethe recovered archaeological evidence. The presence of spring de-posits on Rusinga suggests that water was one. Topographic effectsprovided several others. Topography plays a key role in the distri-bution of grazing lawns noted above, and, like Blumenschine et al.(2012), we emphasize the effects that even modest topographic


change can have on site suitability for hominin use. The higherelevation parts of both islands would have provided (1) good visi-bility of nearby animal migration corridors, (2) high-quality stonefor tool manufacture, and (3) access to a variety of vegetation typesnot found in nearby grassy lowlands.

4.1. Springs and fresh water sources

Tufas are sedimentary deposits formed by the precipitation ofcalcium carbonate minerals. In cool water systems, tufas typicallycontain the remains of microphytes, macrophytes, invertebrates,and bacteria that form in a persistent aqueous environment(Pedley, 1990; Ford and Pedley, 1996; Pedley et al., 2003). At Nya-mita, on Rusinga Island (Fig. 1), multiple tufa deposits include alarge fluvial barrage (Fig. 6A) and in situ stromatolites (Fig. 6B).These deposits indicate the presence of a w2 km2 spring-fed freshwater marsh and affiliated fresh water stream system. The tufadeposits are interbedded with tephra that identify them ascontemporary with the fossil- and artifact-bearing strata at Nya-mita and other localities on the island. Tufa fragments are presentin Waware Beds conglomerates of Mfangano Island, but no in situdeposits were identified. The Nyamita spring deposits are adjacentto modern spring deposits, suggesting the long-term persistence offresh water at this locality. Conglomerates and channels in theWasiriya Beds (at Nyamita, Nyamsingula, and Wakondo) and theWaware Beds (at Kakrigu) indicate the presence of at leastseasonally available water at each of these localities. Particularlyduring periods of increased aridity and/or lake margin retreat, afresh water source would have been a key resource for human andnon-human foragers.

The presence of at least seasonally available water at the spring(and perhaps stream channels) is consistent with elements of thefaunal assemblages such as reedbuck (Redunca redunca or R. ful-vorufula), kob or southern reedbuck (Kobus kob or R. arundinum),and hippo (Hippopotamus cf. amphibius), each of which indicatesthe presence of standing water. Although much of the fauna in-dicates a general dry grassland environment, the combined evi-dence suggests that the archaeological and paleontological sites


Fig. 6. Photographs from Nyamita on Rusinga Island of (A) barrage tufa deposit outcrop, (B) thin section of in situ stromatolite (width of section ¼ 5 cm), and (C) detail of fossil leaffragment preserved within tufa (scale bar ¼ 1 cm).


investigated here are probably wetter, more densely vegetatedareas within this setting. Rare megafossil leaves (Fig. 6C) furthersupport the presence of more dense vegetation. These include atleast three species of relatively large dicot angiosperm leaves thatdemonstrate the presence of trees and/or shrubs surrounding thespring. Additionally, monocot fossil leaves, which are mostly likelya species of Typha, are clear indications of a wetland habitat with atleast periodic standing water. Together, the paleobotanical evi-dence, though relatively sparse, suggests the tufa deposits at Nya-mita were likely areas with standing water surrounded by apotentially dense stand of trees and/or shrubs, an interpretationalso supported by preliminary isotopic analyses of soil carbonatesand bulk organics that indicate the presence of relatively closed C3-dominated vegetation at Nyamita and other Rusinga and Mfanganolocalities (Garrett et al., 2010; Peppe et al., 2011).

The artifact density at the Nyamita spring and other sites is low,in apparent contradiction to the inferred importance of fresh waterin a relatively dry landscape. However, this pattern is consistentwith our expectations of ephemeral and low-intensity site occu-pations in general, and specifically for springs, where the danger ofcarnivore competition or the possibility of scaring away game ishigh (Brooks, 1978; Sampson, 2001). Although cores and core-


trimming elements have been recovered (Fig. 2), extensive refit-ting efforts suggest that complete or extensive reduction sequencesare not present at any of the sites. This is seen most clearly in thediversity of stone raw materials. Although only 339 stone artifactshave been recovered from Rusinga Island’s Wasiriya Beds, >30different raw material types are present. These data suggest thatthe sites we study are locations of artifact use and discard ratherthan places of tool manufacture. Whereas the reduction of even asingle cobble can produce hundreds of flakes in matter of minutes,sites where primarily finished tools are used and occasionally dis-carded will have a low artifact density regardless of the frequencyof its use by hominins (e.g., Roebroeks, 1988).

4.2. Game visibility and acquisition

Their position as topographic high points overlooking grassylowlands would have made both Rusinga and Mfangano Islandsnatural vantage points for tracking herds of migratory game, manyspecies of which are found among the recovered fauna. A similarargument has beenmade for MSA sites at Porc Epic Cave in Ethiopia(Clark, 2001), Nasera rockshelter in Tanzania (Mehlman, 1989), andLukenya Hill in Kenya (Marean, 1992, 1997).



Streams draining the highlands on Rusinga and Mfangano mayhave also provided terrain suitable for driving, corralling, or hin-dering game, as has been argued for GvJm-46 at Lukenya Hill(Miller, 1979; Marean, 1997). At Wakondo on Rusinga, our ongoingresearch focuses on an apparent mass-death assemblage of theextinct wildebeest-like Rusingoryx atopocranion. The fossils occur inone or more related high-energy, shallow stream channels at thebase of the uplands. The assemblage includes multiple cut- andchop-marked specimens and an unusual large blade-based stonetool assemblage (Jenkins et al., 2012), but whether this representsopportunistic scavenging or specialized tactical hunting (sensuMarean, 1997) is currently under investigation.

4.3. Stone raw material availability

Despite the diversity of raw material types, the lithic assem-blages from Rusinga Island and Mfangano Island are dominated bylocally available raw materials. From Rusinga, where sample size islarger, the majority of artifacts are made of lava (65%) and chert(29%) (Tryon et al., 2012). Of the lava artifacts, 52% are made from ofLunene Lava (Tryon et al., 2012), a fine-grained nephelinite that ispart of a series of Miocene lava flows that cap the Miocene volca-niclastic sedimentary rocks and create prominent topographichighs in the center of the island (Fig. 1)(Shackleton, 1950; VanCouvering, 1972). Other raw material types (many represented byonly a single artifact) such as quartz, quartzite, and some varietiesof lava and chert, have no known source on the island, and implythe use of more distant sources indicated by our models of homininland use. The presence of obsidian (n ¼ 1) from sources >200 kmaway in particular suggests contact with distant regions eitherthrough extensive seasonal or annual territorial movement, directprocurement, or trade with other hominin groups (Tryon et al.,2012). Further support for the acquisition of resources furtherafield can be seen in the sample of points from Rusinga andMfangano, the majority of which (5/6) are of raw materials notfound on either island, suggesting that artifact deposition on theislands represents the end of use cycles that began elsewhere.

Several varieties of lava as well as chert are available from pri-mary deposits (outcrops) or secondary deposits (as stream clasts)on Rusinga Island, Mfangano Island or the mainland to the east.Sampling of Pleistocene stream deposits at multiple outcropsacross Rusinga Island has demonstrated that average maximumdimension of clasts for all raw materials is <7 cm. While lavarepresents between 55% and 83% of the clasts available in thesedeposits, many of these are significantly weathered and thus highlyfriable, making them unsuitable as material for stone tools.Furthermore, the complete absence of chert clasts in the sampleddeposits strongly suggest that most if not all of the chert artifactsmust have been quarried from the local outcrop or transportedfrom sources further afield. These observations suggest that lavaand chert outcrops (rather than stream gravels) may have served asa further resource that attracted hominins to Rusinga and Mfan-gano, with initial raw material acquisition and reduction occurringcloser to the outcrops rather than from the sites currently understudy.

The importance of the lava and chert outcrops on Rusinga andMfangano is underscored by the rarity of high quality stone rawmaterials elsewhere in the Lake Victoria basin, particularly west ofthe islands. The deposits from the Miocene Kisingiri Volcano thatform the bedrock of Rusinga, Mfangano, and parts of the mainlandare the easternmost source of lava suitable for stone tool produc-tion in the immediate vicinity. The surrounding bedrock, and thatwhich probably underlies most of Lake Victoria, is metamorphosedcoarsely crystalline bedrock (Schlüter, 2008), generally poorlysuited for tool manufacture. Archaeological sites on Lolui and


Buvuma Islands in Uganda (Fig. 4) indicate that quartz is availablein some areas (e.g., Van Noten, 1971; Posnansky et al., 2005). Theabundance of quartz outcrops in the central part of the basin isunknown, but may have been sparse if buried by alluvium, oldlacustrine sediments, or soil developed on either of these. Althoughinselbergs of basement rock in the central part of the basin mayhave provided stone raw materials when lake level was reduced,few are observed in the seismic survey data of the lake basin(Scholz et al., 1998), suggesting limited availability to hominins.

In addition to the probable rarity of outcrops of suitable stoneraw materials west of Rusinga and Mfangano, secondary depositswere also likely absent in the central part of the Lake Victoria basin.Channel gradients suggest that streams draining the mainland orisland highlands would not have transported suitably sized rockclasts far beyond Rusinga or Mfangano. As noted above, in theabsence of Lake Victoria, any rivers crossing the central part of thebasin would have traversed a shallow gradient, meaning lowcompetence and the transport of small (i.e. sand sized and smaller)clasts. Thus, for any hominin groups operating further west towardthe center of the Lake Victoria basin, Rusinga and Mfangano wouldhave been important areas for raw material acquisition and stonetool production.

4.4. Plant resources

As discussed above, the effects of topography and foragingpreferences of grazing ungulate herds play an important role inshaping the vegetation of eastern African savannas. In contrast tothe grassy lowlands, the topographically high and rough areas, suchas Rusinga and Mfangano Islands, would have been avoided bygrazers and are likely to have supported a variety of plant resourcesincluding shrubs and trees. Such a flora would have providedseveral resources for hominins that may have been more predict-ably available than mobile or migratory game (see comparablediscussion in Marean, 1997). Multiple ethnobotanical studies ofmodern populations in arid eastern African savanna environmentsindicate that these plant resources are important sources of food,medicine, fuel, and for tool construction and other purposes (seebelow). Although adjacent grazing lawns provided excellent fodderfor migratory game herds, vegetation on the uplands on RusingaIsland, Mfangano Island, and on the present day mainland mayhave been particularly important to hominin foragers.

Different parts of various shrubs and trees are useful for anumber of purposes. Woody parts would have been important fuelfor making fire, hafting tools, and making digging sticks or huntingweapons. Some woody stems are chewed for water content, andstems, leaves, piths, roots, bark, and fruits are commonly used asfood and medicine (Ichikawa, 1987; Maundu and Berger, 2001;Bussman, 2006; Bussman et al., 2006). In eastern African savannas,wooded patches are likely to occur in topographically rough areasand in areas with higher moisture availability, conditions found inthe uplands of Rusinga and Mfangano and at each of the archaeo-logical and paleontological sites discussed here. The presence of atleast some tree cover is implied by the fossil leaves from Nyamita(Fig. 6C). Direct evidence of the use of plants by hominins is pres-ently lacking. However, the ethnobotanical data emphasize theimportance of shrubs and woody vegetation, and the role theylikely played in shaping hominin foraging decisions and settlementlocations.

5. Discussion and conclusions

The Lake Victoria basin is a dynamic environment. Although thelargest lake in Africa today in terms of surface area, this importantbody of water expanded and contracted in response to climatic



changes, in particular moisture availability. The history of thechanges in Lake Victoria’s shape, size, and surrounding plant andanimal communities, including early populations of Homo sapiens,remains largely unknown beyond the Last Glacial Maximum.Terrestrial sediments dated to w33e100 ka from the near-shoreislands of Rusinga and Mfangano record one or more pre-LGMarid intervals, as indicated by a suite of arid-adapted extinct andextant ungulate taxa.

During arid intervals, including the period of hominin occupa-tion studied here, lake level dropped at least 25 m (based on ourbathymetric reconstructions) and grasslands expanded. Underthese conditions, Rusinga andMfangano likely represented denselyvegetated local topographic highpoints on the landscape, over-looking grassy lowlands. These topographic high points would havebeen important to hominin foragers for several reasons: (1) theyprovided vantage points to observe migratory game, (2) they con-tained outcrops of high quality raw material, and (3) they includeda range of potentially useful plants not found in the lowlands, and(4) water, probably rare in the surrounding environment, wasavailable at a spring or in seasonal streams. The archaeologicalrecord of Rusinga and Mfangano Islands is sparse, but lithic rawmaterial diversity in particular suggests repeated visits to theselocalities.

The effects of even minor changes in topography can haveimportant effects in savanna environments in eastern Africa. Wehavemodeled some of these for the easternmargin of Lake Victoria,but similar factors likely explain hominin use of other open air sites,as has already been suggested for Oldowan sites at Olduvai Gorge,Tanzania (Blumenschine et al., 2012).Wewill continue to refine andtest the hypotheses presented here through our ongoing fieldworkon Pleistocene sediments on Rusinga and Mfangano Islands as wellas the adjacent mainland.

Acknowledgments

Research presented here was conducted under research permitNCST/5/002/R/576 issued to Tryon by the government of the Re-public of Kenya and an Exploration and Excavation License issuedby the National Museums of Kenya. Funding was provided by theNational Science Foundation (BCS-0841530, BCS-1013199, and BCS-1013108), the National Geographic Society’s Committee forResearch and Exploration (8762-10), the Leakey Foundation, theGeological Society of America, New York University, Baylor Uni-versity, and the University of Queensland. We thank Drs. EmmaMbua, Kieran McNulty, William Harcourt-Smith, Holly Dunsworth,and Thomas Lehmann for assistance in Kenya, and Dr. Erella Hoversfor her help as a sounding board for some of the ideas presentedhere, which were further refined and substantially improved bycomments from Stanley Ambrose and an anonymous reviewer.Tryon thanks Erella Hovers, Gonen Sharon and Yosi Zaidner for theinvitation to participate in this conference, and the wonderful op-portunity to see the Middle Paleolithic of Israel firsthand. Tryon,Johnson, and Bill Keegan would like to dedicate this paper to thememory of Robert E. Dewar, who first got them thinking aboutlandscapes and settlement patterns.

References

Ambrose, S.H., Lorenz, K.G., 1990. Social and ecological models for the Middle StoneAge in Southern Africa. In: Mellars, P. (Ed.), The Emergence of Modern Humans:an Archaeological Perspective. Cornell University Press, Ithaca, New York,pp. 3e33.

Ambrose, S.H., 2001. Middle and Later Stone Age settlement patterns in the centralrift valley, Kenya: Comparisons and contrasts. In: Conard, N.J. (Ed.), SettlementDynamics in the Middle Paleolithic and Middle Stone Age. Kerns-Verlag,Tübingen, pp. 21e44.


Andrews, P., 1973. Vegetation of Rusinga Island. Journal of the East African NaturalHistory Society and National Museum 142, 1e8.

Assefa, Z., Yirga, S., Reed, K.E., 2008. The large-mammal fauna from the KibishFormation. Journal of Human Evolution 55, 501e512.

Bailey, D.W., Gross, J.E., Laca, E.A., Rittenhouse, L.R., Coughener, M.B., Swift, D.M.,Sims, P.L., 1996. Mechanisms that result in large herbivore grazing-distributionpatterns. Journal of Range Management 49, 386e400.

Barboni, D., Bonnefille, R., Alexandre, A., Meunier, J.D., 1999. Phytoliths as paleo-environmental indicators, west side middle Awash valley, Ethiopia. Palae-ogeography, Palaeoclimatology, Palaeoecology 152, 87e100.

Bell, R.H.V., 1971. A grazing ecosystem in the Serengeti. Scientific American 225,86e93.

Beverly, E., Faith, J.T., Peppe, D.J., Tryon, C.A., Driese, S.G., Blegen, N., Patterson, D.,Horner, W.H., 2012. Paleoenvironmental context of Pleistocene archaeologicaland paleontological sites in Kenya’s Lake Victoria basin. Geological Society ofAmerica Abstracts with Programs 44, 415.

Bishop, W.W., Trendall, A.F., 1967. Erosion-surfaces, tectonics and volcanic activity inUganda. Quaterly Journal of the Geological Society of London 122, 385e420.

Blumenschine, R.J., Masao, F.T., Stollhofen, H., Stanistreet, I.G., Bamford, M.K.,Albert, R.M., Njau, J.K., Prassack, K.A., 2012. Landscape distribution of Oldowanstone artifact assemblages across the fault compartments of the eastern OlduvaiLake Basin during early lowermost Bed II times. Journal of Human Evolution 63,384e394.

Braun, D.R., Rogers, M.J., Harris, J.W.K., Walker, S.J., 2008. Landscape-scale variationin hominin tool use: evidence from the developed Oldowan. Journal of HumanEvolution 55, 1053e1063.

Broecker, W.C., Peteet, D., Hajdas, I., Lin, J., 1998. Antiphasing between rainfall inAfrica’s rift valley and North America’s Great basin. Quaternary Research 50,12e20.

Brooks, A.S., 1978. A note on the LSA features at sGi: Analogies from historic Sanhunting practices. Botswana Notes and Records 10, 1e3.

Brown, F.H., McDougall, I., Fleagle, J.G., 2012. Correlation of the KHS Tuff of theKibish Formation to volcanic ash layers at other sites, and the age of early Homosapiens (Omo I and Omo II). Journal of Human Evolution 63, 577e585.

Buol, S.W., Southard, R.J., Graham, R.C., McDaniel, P.A., 2003. Vertisols, Shrinkingand Swelling Dark Clay Soils: Soil Genesis and Classification. Iowa State Uni-versity Press, Ames.

Bussman, R.W., 2006. Ethnobotany of the Samburu of Mt. Nyiru, South Turkana,Kenya. Journal of Ethnobiology and Ethnomedicine 2, 35.

Bussman, R.W., Gilbreath, G.G., Solio, J., Manja, L., Lutuluo, R., Kunguru, K., Wood, N.,Mathenge, S.G., 2006. Plant use of the Maasai of Sekenani valley, Maasai Mara,Kenya. Journal of Ethnobiology and Ethnomedicine 2, 22.

Clark, J.D., 2001. Ecological and behavioral implications of the siting of Middle StoneAge rockshelter and cave settlements in Africa. In: Conard, N.J. (Ed.), SettlementDynamics of the Middle Paleolithic and Middle Stone Age. Kerns Verlag,Tübingen, pp. 91e98.

Clark, P.U., Dyke, A.S., Shakun, J.D., Carlson, A.E., Clark, J., Wohlfarth, B.,Mitrovica, J.X., Hostetler, S.W., McCabe, A.M., 2009. The last glacial maximum.Science 325, 710e714.

Coughener, M.B., 1991. Spatial components of plant-herbivore interactions in pas-toral, ranching, and native ungulate ecosystems. Journal of Range Management44, 530e542.

Coulombe, C.E., Dixon, J.B., Wilding, L.P., 1996. Mineralogy and chemistry of Vertisols.In: Ahmad, N., Mermut, A. (Eds.), Vertisols and Technologies for Their Manage-ment. Developments in Soil Science 24. Elsevier, New York, pp. 115e200.

Cromsigt, J.P.G.M., Olff, H., 2008. Dynamics of grazing lawn formation: an experi-mental test of the role of scale-dependent processes. Oikos 117, 1444e1452.

deMenocal, P., Ortiz, J., Guilderson, T., Adkins, J., Sarnthein, M., Baker, L.,Yarusinsky, M., 2000. Abrupt onset and termination of the African HumidPeriod: rapid climate responses to gradual insolation forcing. Quaternary Sci-ence Reviews 19, 347e361.

Drake, R.E., Van Couvering, J.A., Pickford, M.H., Curtis, G.H., Harris, J.A., 1988. Newchronology for the Early Miocene mammalian faunas of Kisingiri, WesternKenya. Journal of the Geological Society London 145, 479e491.

Dyson-Hudson, R., Smith, E.A., 1978. Human territoriality: an ecological reassess-ment. American Anthropologist 80, 21e41.

Fagan, B.M., Lofgren, L., 1966. Archaeological reconnaissance on the Sese Islands.Uganda Journal 30, 81e86.

Faith, J.T., 2013. Ungulate diversity and precipitation history since the Last GlacialMaximum in the Western Cape, South Africa. Quaternary Science Reviews 68,191e199.

Faith, J.T., Choiniere, J.N., Tryon, C.A., Peppe, D.J., Fox, D.L., 2011. Taxonomic statusand paleoecology of Rusingoryx atopocranion (Mammalia, Artiodactyla), anextinct Pleistocene bovid from Rusinga Island, Kenya. Quaternary Research 75,697e707.

Faith, J.T., Potts, R., Plummer, T.W., Bishop, L.C., Marean, C.W., Tryon, C.A., 2012. Newperspectives on middle Pleistocene change in the large mammal faunas of EastAfrica: Damaliscus hypsodon sp. nov. (Mammalia, Artiodactyla) from Lainyamok,Kenya. Palaeogeography, Palaeoclimatology, Palaeoecology 361e362, 84e93.

Faith, J.T., Tryon, C.A., Peppe, D.J., Fox, D.L., 2013. The fossil history of Grévy’s zebra(Equus grevyi) in equatorial East Africa. Journal of Biogeography 40, 359e369.

Ford, T.D., Pedley, H.M., 1996. A review of tufa and travertine deposits of the world.Earth-Science Reviews 41, 117e175.

Fowler, N.L., 2002. The joint effects of grazing, competition, and topographic po-sition on six savanna grasses. Ecology 83, 2477e2488.


http://refhub.elsevier.com/S1040-6182(13)00301-7/sref1




























































































































Gaffney, V., Stan�ci�c, Z., 1996. GIS Approaches to Regional Analysis: a Case Study ofthe Island of Hvar. Znanstveni in�stitut Filozofske fakultete, Ljubljana.

Garrett, N., McNulty, K., Fox, D., Peppe, D., Tryon, C.A., 2010. Isotope paleoecology ofthe PleistoceneWasiriyabeds of Rusinga Island, Kenya. In: ProgramandAbstractsof the Annual Meetings of the Society of Vertebrate Paleontology, vol. 94A.

Hemming, S.R., 2004. Heinrich events: massive Late Pleistocene detritus layers ofthe North Atlantic and their global climate imprint. Review of Geophysics 42,RG1005.

Ichikawa, M., 1987. A preliminary report on the ethnobotany of the Suiei Dorobo inNorthern Kenya. African Study Monographs Supplement 7, 1e52.

Jarman, M.R., Vita-Finzi, C., Higgs, E.S., 1972. Site catchment analysis in archaeology.In: Ucko, P.J., Tringham, R., Dimbleby, G.W. (Eds.), Man, Settlement, and Ur-banism. Schenkman, Cambridge, MA, pp. 61e66.

Jenkins, K., Faith, J.T., Tryon, C., Peppe, D., Nightingale, S., Ogondo, J., Johnson, C.R.,Driese, S., 2012. New excavations of a Late Pleistocene bonebed and associatedMSA artifacts from Rusinga Island, Kenya. PaleoAnthropology 2012, A17.

Johnson, T.C., Scholz, C.A., Talbot, M.R., Kelts, K., Ricketts, R.D., Ngobi, G., Beuning, K.,Ssemmanda, I., McGill, J.W., 1996. Late Pleistocene desiccation of Lake Victoriaand rapid evolution of cichlid fishes. Science 273, 1091e1093.

Kelly, R.L., 1995. The Foraging Spectrum. Smithsonian Institution Press, Washington,D.C.

Kendall, R.L., 1969. An ecological history of the Lake Victoria basin. EcologicalMonographs 39, 121e176.

Kingdon, J., 1974. East African Mammals: an Atlas of Evolution. vol. II. In: Hares andRodents. Academic Press, New York.

Kingdon, J., 1997. The Kingdon Field Guide to African Mammals. Academic Press,San Diego.

Marean, C.W., 1992. Implications of late Quaternary mammalian fauna from Luke-nya Hill (south-central Kenya) for paleoenvironmental change and faunal ex-tinctions. Quaternary Research 37, 239e255.

Marean, C.W., 1997. Hunter-gatherer foraging strategies in tropical grasslands:model-building and testing in the East African Middle and Later Stone Age.Journal of Anthropological Archaeology 16, 189e225.

Maundu, P., Berger, D., ole Saitabau, C., Nasieku, J., Lipelian, M., Mathenge, S.G.,Morimoto, Y., Höft, R., 2001. Ethnobotany of the Loita Maasai. UNESCO Peopleand Plants Working Paper 8, Paris.

McNaughton, S.J., 1984. Grazing lawns: animals in herds, plant form, and coevo-lution. American Naturalist 124, 863e886.

McNaughton, S.J., 1985. Ecology of a grazing ecosystem: the Serengeti. EcologicalMonographs 55, 259e294.

Mehlman, M.J., 1989. Late Quaternary Archaeological Sequences in NorthernTanzania. University of Illinois, Urbana.

Miller, S.F., 1979. Lukenya Hill, GvJm 46, excavation report. Nyame Akuma 14, 31e34.Milly, P.C.D., 1999. Comment on “Antiphasing between rainfall in Africa’s rift valley

and North America’s Great Basin”. Quaternary Research 51, 104e107.Monmonier, M., 2008. Coast Lines: How Mapmakers Frame the World and Chart

Environmental Change. University of Chicago Press, Chicago.Nenquin, J., 1971. Archaeological prospections on the islands of Buvuma and Bugaia,

LakeVictoriaNyanza (Uganda). Proceedingsof thePrehistoricSociety37, 381e418.Nicholson, S.E., 1998. Historical fluctuations of Lake Victoria and other lakes in the

northern rift valley of East Africa. In: Lehman, J.T. (Ed.), Environmental Changeand Response in East African Lakes. Kluwer Academic Publishers, Dordtrecht,pp. 7e35.

Nowak, R.M., 1999. Walker’s Mammals of the World, sixth ed., vol. 2. Johns HopkinsUniversity Press, Baltimore.

Office, U.K.H., 1907e1956. Central Africa: Victoria Nyanza (Northern Portion). Sur-veyed by Commander B. Whitehouse, R.N., and Mr. C. S. Hunter, Uganda RailwaySurveys, 1900-01. Published 1902, updated 1907, 1911, 1929, 1932, 1956. Hy-drographic Office, Admiralty Charts and Publications No. 3252, Taunton, Som-erset, United Kingdom.

Office, U.K.H., 1911e1956. Central Africa: Victoria Nyanaza: Plans in Kavirondo Gulfand Entrance. Surveyed by Mr. C. S. Hunter, C.E., Uganda Railway, 1908-1910,published 1911, updated 1929, 1932, 1956. Hydrographic Office, AdmiraltyCharts and Publications No. 1120, Taunton, Somerset, United Kingdom.

Office, U.S.H., 1915. North Sea Pilot (Eastern Shores): From Dunkerque to the Skaw.In: Hydrographic Office Publication No. 135. Government Publication Office,Washington, D.C.

Pedley, H.M., 1990. Classification and environmental models of cool freshwatertufas. Sedimentary Geology 68, 143e154.

Pedley, H.M., Martin, J.A.G., Ordonez, D., del Curas, A.G., 2003. Sedimentology ofQuaternary perched spring- line and paludal tufas: criteria for recognition, withexamples from Guadalajara Province, Spain. Sedimentology 50, 23e44.

Peppe, D.J., Faith, J.T., Tryon, C.A., Fox, D.L., Garrett, N., Van Plantinga, A., Keegan, W.,Nightingale, S., Ogondo, J., Zaarur, S., 2011. Terrestrial evidence for late Qua-ternary lake level changes from the Lake Victoria basin, Equatorial Africa.Geological Society of America Abstracts with Programs 43, 335.

Peppe, D.J., McNulty, K.P., Cote, S.M., Harcourt-Smith, W.E.H., Dunsworth, H.M., VanCouvering, J.A., 2009. Stratigraphic interpretation of the Kulu formation (EarlyMiocene, Rusinga Island, Kenya) and its implications for primate evolution.Journal of Human Evolution 56, 447e461.

Pickford, M., 1972. The tectonics, volcanics and sedments of the Nyanza Rift Valley.Zeitschrift für Geomorphology 42, 1e33.

Pickford, M., 1986. Cainozoic paleontological sites of Western Kenya. MünchnerGeowissenschaftlich Abhandlungen 8, 1e151.


Piper, B.S., Plinston, D.T., Sutcliffe, J.V., 1986. The water balance of Lake Victoria.Hydrological Sciences Journal 31, 25e37.

Posnansky, M., Reid, A., Ashley, C., 2005. Archaeology on Lolui Island, Uganda 1964-5. Anazia 40, 73e100.

Potts, R., Behrensmeyer, A.K., Ditchfield, P., 1999. Paleolandscape variation and earlyPleistocene hominid activities: members 1 and 7, Olorgesailie formation, Kenya.Journal of Human Evolution 37, 747e788.

Roebroeks, W., 1988. From Find Scatters to Early Hominid Behavior: a Study ofMiddle Palaeolithic Riverside Settlements at Maastricht-belvédère (TheNetherlands). In: Analecta Praehistorica Leidensia 21. University of Ledien,Leiden.

Sampson, C.G., 2001. An Acheulian settlement pattern in the Upper Karoo region ofSouth Africa. In: Milliken, S., Cook, J. (Eds.), A Very Remote Period Indeed:Papers on the Palaeolithic Presented to Derek Roe. Oxbow Books, Oxford,pp. 28e36.

Schlüter, T., 2008. Geological Atlas of Africa, second ed. Springer, Berlin.Scholz, C.A., Johnson, T.C., Cattaneo, P., Malinga, H., Shana, S., 1998. Initial results of

1995 IDEAL seismic reflection survey of Lake Victoria, Uganda and Tanzania. In:Lehman, J.T. (Ed.), Environmental Change and Response in East African Lakes.Kluwer Academic Publishers, Dordrecht, pp. 47e58.

Sene, K.J., Plinston, D.T., 1994. A review and update of the hydrology of Lake Victoriain East Africa. Hydrological Sciences Journal 39, 47e63.

Shackleton, R.M., 1950. A contribution to the geology of the Kavirondo Rift Valley.Quarterly Journal of the Geological Society 106, 345e392.

Silsbe, G., 2003. A new digital map for Lake Victoria, East Africa. IDEAL Bulletin:Bulletin of the International Decade for the East African Lakes Summer, 6e7.

Skinner, J.D., Chimimba, C.T., 2005. The Mammals of the Southern African Subre-gion. Cambridge University Press, Cambridge.

Southard, R.J., Driese, S.G., Nordt, L.C., 2011. Vertisols. In: Huang, P.M., Li, Y.,Sumner, M.E. (Eds.), Handbook of Soil Science, second ed. CRC Press, Boca Raton,FL, pp. 33e97.

Stager, J.C., Cumming, B., Meeker, L., 1997. A high-resolution 11,400-yr diatom re-cord from Lake Victoria, East Africa. Quaternary Research 47, 81e89.

Stager, J.C., Mayewski, P.A., Meeker, D.L., 2002. Cooling cycles, Heinrich Event 1, andthe desiccation of Lake Victoria. Palaeogeography, Palaeoclimatology, Palae-oecology 183, 169e178.

Stager, J.C., Johnson, T.C., 2008. The Late Pleistocene desiccation of Lake Victoria andthe origin of its endemic fauna. Hydrobiologia 596, 5e16.

Stager, J.C., Ryves, D.B., Chase, B.M., Pausata, F.S.R., 2011. Catastrophic Drought in theAfro-Asian Monsoon region during Heinrich Event 1. Science 331, 1299e1302.

Survey of Kenya, 1979. Rusinga, Topographic Series East Africa 1:50,000 (Kenya),Sheet 115/3. Survey of Kenya, Nairobi.

Talbot, M.R., Jensen, N.B., Laerdal, T., Filippi, M.L., 2006. Geochemical responses to amajor transgression in giant African lakes. Journal of Paleolimnology 35, 467e489.

Talbot, M.R., Laerdal, T., 2000. The Late Pleistocene-Holocene paleolimnology ofLake Victoria, East Africa, based upon elemental and isotopic analyses of sedi-mentary organic matter. Journal of Paleolimnology 23, 141e164.

Talbot, M.R., Livingstone, D.A., 1989. Hydrogen index and carbon isotopes oflacustrine organic matter as lake level indicators. Palaeogeography, Palae-oclimatology, Palaeoecology 70, 121e137.

Temple, P.H., 1966. Evidence of Changes in the Level of Lake Victoria and theirSignificance. Ph.D. Thesis, University of London.

Temple, P.H., 1967. Causes of intermittent decline of the level of Lake Victoria duringthe Late Pleistocene and Holocene. In: Steel, R.W., Lawton, R. (Eds.), LiverpoolEssays in Geography. Longmans, London.

Tryon, C.A., Faith, J.T., Peppe, D.J., Fox, D.L., McNulty, K.P., Jenkins, K., Dunsworth, H.,Harcourt-Smith, W., 2010. The Pleistocene archaeology and environments of theWasiriya Beds, Rusinga Island, Kenya. Journal of Human Evolution 59, 657e671.

Tryon, C.A., Peppe, D.J., Faith, J.T., Van Plantinga, A., Nightengale, S., Ogondo, J., 2012.Late Pleistocene artefacts and fauna from Rusinga and Mfangano islands, LakeVictoria, Kenya. Azania: Archaeological Research in Africa 47, 14e38.

Van Couvering, J., 1972. Geology of Rusinga Island and Correlation of the Kenya Mid-tertiary Fauna. Cambridge University.

Van Noten, F., 1971. Excavations at Munyama cave. Antiquity 45, 56e58.Verschuren, D., Johnson, T.C., Kling, H.J., Edgington, D.N., Leavitt, P.R., Brown, E.T.,

Talbot, M.R., Hecky, R.E., 2000. History and timing of human impact on LakeVictoria, East Africa. Proceedings of the Royal Society of London B 269, 289e294.

Vrba, E.S., 1980. The significance of bovid remains as an indicator of environmentand predation patterns. In: Behrensmeyer, A.K., Hill, A.P. (Eds.), Fossils in theMaking. University of Chicago Press, Chicago, pp. 247e272.

Westcott, K.L., Brandon, R.J., 2000. Practical Applications of GIS for Archaeologists: aPredictive Modeling Kit. Taylor & Francis, London.

Whitworth, T., 1961. The geology of Mfwanganu island, western Kenya. OverseasGeology and Mineral Resources 8, 150e192.

Wilding, L.P., Tessier, D., 1988. Genesis of Vertisols: Shrink-swell phenomena. In:Wilding, L.P., Puentes, R. (Eds.), Vertisols: Their Distribution, Properties, Clas-sification and Management. Texas A&M University Printing Center, CollegeStation, pp. 55e79.

Yellen, J.E., Brooks, A.S., Helgren, D.M., Tappen, M., Ambrose, S.H., Bonnefille, R.,Feathers, J.K., Goodfriend, G., Ludwig, K., Renne, P., Stewart, K., 2005. Thearchaeology of Aduma middle stone age sites in the Awash valley, Ethiopia.PaleoAnthropology 10, 25e100.









































































































































































































sites on the landscape: paleoenvironmental context of late ... · sediments are shown in gray;...

Documents