Mid to late-Holocene vegetation dynamics on the Laikipia Plateau, Kenya

http://hol.sagepub.com

The Holocene

DOI: 10.1191/0959683605hl857ra 2005; 15; 837 The Holocene

D. Taylor, P. J. Lane, V. Muiruri, A. Ruttledge, R. Gaj McKeever, T. Nolan, P. Kenny and R. Goodhue Mid- to late-Holocene vegetation dynamics on the Laikipia Plateau, Kenya

http://hol.sagepub.com/cgi/content/abstract/15/6/837 The online version of this article can be found at:

Published by:

http://www.sagepublications.com

can be found at:The Holocene Additional services and information for

http://hol.sagepub.com/cgi/alerts Email Alerts:

http://hol.sagepub.com/subscriptions Subscriptions:

http://www.sagepub.com/journalsReprints.navReprints:

http://www.sagepub.com/journalsPermissions.navPermissions:

http://hol.sagepub.com/cgi/content/refs/15/6/837SAGE Journals Online and HighWire Press platforms):

(this article cites 23 articles hosted on the Citations

© 2005 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at RUTGERS UNIV on April 27, 2008 http://hol.sagepub.comDownloaded from

The Holocene 15,6 (2005) pp. 837-846

Mid- to late-Holocene vegetation dynamics

on the Laikipia Plateau, Kenya

D. Taylor,l* P.J. Lane,2 V. Muiruri,3 A. Ruttledge,1R. Gaj McKeever,l T. Nolan,1 P. Kennyl and R. Goodhue4('Department of Geography, Trinity College, University of Dublin, Dublin 2,Ireland; 2British Institute in Eastern Africa, PO. Box 30710, GPO 00100,Nairobi, Kenya; 3Department of Palynology and Palaeobotany, National MuseumsofKenya, Museum Hill, Nairobi, Kenya; 4Geochemistry Laboratory, Department ofGeology, Trinity College, University of Dublin, Dublin 2, Ireland)

Received 19 May 2004; revised manuscript accepted 1 December 2004

HOLOCENERESEARCHPAPER

Abstract: Interactions between Holocene environmental changes and human subsistence strategies in semi-arid parts of eastern Africa are relatively poorly understood because of a paucity of sites wherecontemporaneous archaeological and palaeoenvironmental records are preserved. This paper presents newAMS '4C-dated palaeoenvironmental evidence of mid- to late-Holocene vegetation changes, in the form ofpollen, charcoal, mineralogical and &'Cb,Ik data, from a floodplain location on the Laikipia Plateau,Kenya, and in the context of information on modern plant-environment relationships, existing and newarchaeological and historical data, and published palaeoclimatic records. Although relatively poorlyresolved, the evidence indicates more wooded and relatively humid conditions compared with the presentfrom before c. 6600 BP to c. 1900 BP. Evidence for vegetation changes over the last two millennia is morefinely resolved and indicates increased burning and the expansion of fire-modified Acacia bushland c. 1900BP and grassland c. 1700 BP. Burning to improve and extend pasture, and possibly to eradicate disease-prone habitats, may have been facilitated by prolonged periods of moisture deficits that might also havefacilitated the spread of food production technologies. Vegetation changes c. 700 BP are associated withevidence for the occurrence of fires locally and could represent the activities of people and their animalsclose to the study site.

Key words: C3-C4, charcoal,Holocene, Kenya.

Introduction

The Great Rift of eastern Africa and adjacent upland areas,including the Laikipia Plateau, Kenya, have been a focus ofarchaeological research into later prehistoric settlements andsubsistence strategies for more than four decades. Currentevidence suggests that the transition to food production, in theform of generalized, mixed cattle and sheep/goat pastoralism,had occurred in some parts of the northern lowlands borderingLake Turkana by c. 4000 BP, c. 3400-3000 BP in highlandcentral and southern Kenya (Marean, 1992; Marshall, 2000;Marshall and Hildebrand, 2002), and possibly as early as 3500BP in western Kenya (Karega-Mulnene, 2002) (dates referred toare based on uncalibrated radiocarbon dates, unless otherwisestated). Archaeological and genetic evidence indicates that theinitial introduction of livestock was due in part to local hunter-

*Author for correspondence (e-mail: [email protected])© 2005 Edward Amold (Publishers) Ltd

eastern Africa, fluvial sediments, isotope, Pastoral Neolithic, pollen,

gatherers coming into contact with small groups of herdersfrom Sudan, Ethiopia and possibly Somalia (Bower, 1991;Marshall, 2000). Marshall and Hildebrand (2002) propose thatincreased climatic aridification and unreliability of rainfall andthe benefits of being able to exert some control over theavailability of food were among a complex of factors thattriggered the domestication of cattle in northern Africa, andthe subsequent southward expansion of pastoralism. It is likelythat early herding communities continued to rely upon wildspecies during the initial phase of expansion of food produc-tion in eastern Africa because of the abundance of wild animalsand plants (Marshall and Hildebrand, 2002), and that this andthe incidence of livestock diseases (Gifford-Gonzalez, 1998,2000) contributed to an initially slow uptake of herding.

Pastoralism seems to have become widespread along the Riftand in adjacent areas as far south as the Serengeti and LakeEyasi in northern Tanzania by c. 3000-2000 BP. In severalcases, the faunal assemblages indicate a heavy reliance on the

10.1 191/0959683605hl857ra

© 2005 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at RUTGERS UNIV on April 27, 2008 http://hol.sagepub.comDownloaded from

838 The Holocene 15 (2005)

intensive use of domesticated cattle, sheep, goats and evendonkey, and there is a general increase in settlement size(Robertshaw, 1988; Gifford-Gonzalez, 1998; Marshall, 2000).Some former settlements are associated with ceramic wares andrelated material culture attributed to the Savanna PastoralNeolithic (SPN), while at others the associated ceramics, lithicsand other material culture is assigned to the Elementeitan.These two traditions were broadly contemporary, but generallydo not share the same spatial distribution - one exceptionbeing the Laikipia Plateau, where examples of Elmenteitan andseveral SPN variants have been reported to co-occur at thesame site (Siiriainen, 1977). At some sites, evidence suggeststhe presence of large, complex settlements and substantialshelter construction (Robertshaw, 1990), while at other,broadly contemporary, sites the evidence is more enigmatic.Bower (1991) suggests that a further shift occurred from

c. 1900 to 1300 BP, marked by a return to more highly mobilesettlement strategies and an increased importance of foragedwild resources, even among groups of previously specializedpastoralists. This later period of the Pastoral Neolithic (PN) isgenerally represented by the presence of Akira pottery. Theproduction and use of iron among early pastoralist commu-nities in eastern Africa appears to have begun relatively late.For instance, at the archaeological site at Deloraine Farm,which has yielded remains of cattle and cereals as well asevidence for the manufacture and use of iron implements, theshift to a Pastoral Iron Age (PIA) dates to around 1100 BP(Collett and Robertshaw, 1983: 71; cf. Bower and Nelson,1978), and at the archaeological site at Enkapune ya Muto toc. 1300 BP (Ambrose, 1998). The ensuing centuries witnessedthe consolidation of PIA economies along the Great Rift andadjacent highlands, out of which many of the ethnic identitiesand linguistic clusters that characterize these areas today wereformed.The research presented here aims to provide a mid- to late-

Holocene palaeoenvironmental context for the results of recentand ongoing archaeological research into the origins and

subsequent evolution of food production on and around thesemi-arid Laikipia Plateau. More specifically the research aimsto identify and account for significant vegetation impacts overthe last c. 6000 years.

Study site

The Laikipia Plateau ranges in altitude from 1750 to 2100 ma.s.l. A combination of climatic, geological and topographicalconditions means that surface water and relatively undisturbedorganic-rich sediments, and the proxies of former environ-mental conditions (such as pollen and plant macrofossils) theycontain, are rare. Consequently a floodplain site (1785 m a.s.l.,0044'16"N, 36027'56"E, Figure 1) is the focus of the presentresearch. The study site, known locally as Loitigon (Zebra), bor-ders the 01 Keju Losera river, which originates in the Aberd-ares and drains across the Laikipia Plateau to the Indian Ocean.The Great Rift and lakes Bogoria and- Baringo border the

Laikipia Plateau to the west, while the Aberdares and MountKenya form the southern boundary. To the north and east theLaikipia Plateau grades into low-lying plains. Flood phonolitelavas of Miocene age (Hackman, 1988) form the predominantgeology, although basement complex rocks of pre-Cambrianage outcrop in the east (Shackelton, 1946). Daily temperaturesvary with altitude and season; mean temperatures lie generallywithin the range 22-26°C and temperature minima andmaxima are, respectively, 6-14°C and 35°C (Siiriainen, 1984).Evapotranspiration is intense and moisture deficits are experi-enced widely in the majority of years. A gradient of declininglevels of annual precipitation spans the Laikipia Plateau, withannual average rainfall levels of 500-700 mm in the southwestand 300-500 mm in the central and northeastern parts(Siiriainen, 1984). The rains fall primarily in two seasons; themain wet season occurs during April-May, often accountingfor 80% of total annual rainfall, while a second wet seasonoccurs later in the year in October-November.

620N

N

I

II

)N I

V*l1;IIIIt

I 20

;38250 E i50 EJ-1....

Figure 1 Map showing the location of the study site (Loitigon) and soil pit 6/2 (NB vlei is an Afrikaans word, meaning floodplain). Thelocations of four I-M2 vegetation plots (15-18) are shown

-

I r

© 2005 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at RUTGERS UNIV on April 27, 2008 http://hol.sagepub.comDownloaded from

D. Taylor et a/.: Vegetation dynamics on the Laikipia Plateau, Kenya 839

Vegetation on the Laikipia Plateau generally reflects levels ofeffective precipitation and human modification: Afromontaneforest taxa (sensu White, 1983), including Croton megalocarpus,Juniperus procera and Olea africana, occur in the wettersouthern part as fragments of a relatively dry form of plateauforest (see Lind and Morrison, 1974: 43-45). Elsewhere Acaciabushland and fire-adapted grassland are common. The mostcommon species recorded at Loitigon during fieldwork in July2003 were the C4 grasses Cynodon dactylon, Echinochchloacolona and Pennisetum stramineum, a sedge (Cyperus sp.), andJusticia uncinulata, Indigofera spp. and Ipomea spp. The maintaxa recorded in vegetation on ferralsols bordering the flood-plain were the C4 grasses Eragrostis cilianensis and Pennisetumstramineum and scattered trees of Acacia drepanolobium, A.gerradii, A. nilotica, A. tortilis, Croton dichogamus, C. mega-

locarpus and Rhus nilotica. According to White (1983: 115),Acacia bushland with a composition similar to that aroundLoitigon represents a stage in the degradation of Afromontaneforest by frequent burning, overgrazing and the activities ofcharcoal producers. In the majority of cases, grassland in theregion is anthropogenic and is the end product of higherintensities of human activities and grazing pressure (Lind andMorrison, 1974).The pre-colonial history of human activities on the Laikipia

Plateau remains relatively poorly understood. Large cattle- andgame-ranches, many of which date to the early twentiethcentury, and more densely settled areas of communal farmlandcharacterize the Plateau today. Prior to the early twentiethcentury, pastoral Laikipiak Maasai communities co-existedwith scattered groups of other peoples that included hunter-gatherers (notably the Ndikirri Dorobo and Mukogodo).Precisely when the Laikipiak Maasai first occupied the Plateauis open to debate, although their presence may date to beforeAD 1600 (Jacobs, 1972a: 82; see also Lamprey and Waller,1990; Galaty, 1993). The pre-colonial occupants did, however,leave an imprint on the landscape that includes examples ofrock art, stone cairns and stone circles. Other archaeologicaltraces include a combination of rock shelters and open siteswith surface scatters of flaked obsidian and quartzite, PNpottery, iron slag and iron smelting furnaces. According torecent excavations at two rock shelters in the northeastern partof the Plateau, LSA hunter-gatherers possessed a complex andsophisticated lithic technology by 2000 BP (and probablyconsiderably earlier), were relatively residentially mobile withextended territories, and linked with other hunter-gatheringgroups through long-distance exchange (Gang, 2001; Dicksonand Gang, 2002: 19-20). Assemblages of PN and historicpottery associated with worked stone and faunal remains from

the two rock shelters have been dated to c. 1100 BP (Gang,2001: 14, 16). Detailed analysis of the faunal remains revealedthe relative abundance of wild animals, with only smallnumbers of domestic cattle and sheep/goats (Mutundu, 1999:50-56). Excavations at a limited number of other rockshelterson the Plateau by Jacobs (1972b) and Siiriainen (1977, 1984),as well as more recent work conducted as part of the currentresearch, suggest the first appearance of domestic stockoccurred sometime between 4000 and 3000 BRLand adjoining Loitigon is today sparsely populated.

Evidence of former occupation in the vicinity of the studysite is, however, relatively abundant and is the focus of ongoingarchaeological research. This evidence includes scatters ofworked obsidian and other lithic and ceramic materials, andnumerous stone cairns. Evidence from an area known locally as

Maasai Plains, around 1 km to the west of Loitigon, appears torepresent a substantial and prolonged settlement, or at leastrepeated reoccupation of the same site over a period of timesufficient for the accumulation of large middens: the siteextends over an area at least 750 m in diameter and consistsof several ashy mounds, some of which are up to 1 m high,arranged in three concentric semi-circles. The site resembles thePIA site at Lanet, some 130 km to the southwest of Loitigon.Detailed analyses of the fauna and artefact remains from a

10 m x 2 m trench cut into one of the mounds is ongoing,although provisional analysis of the pottery suggests that it isKisima ware, which Siiriainen (1984) tentatively dates to theperiod post-AD 1400. A sample of wood charcoal recoveredfrom the same level as Kisima ware potsherds during excava-

tions in 2004 yielded an AMS 14C age of 480+ 50 BP (Table 1).

MethodsVegetation survey and analysisIn order to provide information on plant-environment rela-tionships of use in the interpretation of palaeoenvironmentalproxies, representative areas of herbaceous vegetation atLoitigon and at six other locations on the Plateau were

sampled. The abundances of plants according to their percen-

tage cover were established in a total of 18 1-m2 plots, on lavaand on pre-Cambrian basement rock across an altitude rangefrom 1785 to 2035 m a.s.l. In addition, the percentage cover ofbare soil and dead vegetation and any plants (including trees)that were conspicuously present in the vicinity but not foundwithin a sample plot were recorded. A note was also made ofaltitude, moisture availability and the frequency of burning,and a single surface soil sample was collected from each plot.

Table 1 AMS radiocarbon dates for soil pit 6/2 and for charcoal within an ash mound at a former occupation site

Lab. ref. Sample depth Material dated Conventional Calibrated age ranges 613C(IC )no. (cm) 14C age (yr BP) (2 a calibration)*

Beta-188359 45-47 Bulk sample** 1420+50 cal. AD 538-689 - 14.3Beta-190490 85-87 Bulk sample** 1760+40 cal. AD 134-162 -15.2

cal. AD 168-198cal. AD 208-388

Beta-181349 115-117 Bulk sample** 2090+40 cal. 336-330 BC - 12.4cal. 202 BC-AD 2

Beta-181350 165-167 Bulk sample** 6630+40 cal. 5623-5507 BC -14.2cal. 5503-5483 BC

Beta-189982 Trench A, ash mound at Maasai Charcoal 480 +50 cal. AD 1323-1350 -25.2Plains, 96 cm < surface cal. AD 1390- 1513

cal. AD 1600-1615

* Intcal98.14C calibration data set used (Stuiver et al., 1998).** Obvious roots and root hairs removed before analysis.

© 2005 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at RUTGERS UNIV on April 27, 2008 http://hol.sagepub.comDownloaded from

840 The Holocene 15 (2005)

Vegetation and environmental data for the 18 1-mr2 plotswere subjected to canonical correspondence analysis (CCA)(ter Braak, 1986, 1987) using the computer programmeCANOCO 4.5 (ter Braak and Smilauer, 2002). CCA is anextension of correspondence analysis and aims to organize thecomposition of vegetation plot samples according to knownvariations in environment (in this case, field-based measure-ments of altitude and assessments of the availability ofmoisture and frequency of burning).

Sediment samplingSeven soil pits were excavated in histosols along two transectsrunning east to west from the main channel of the 01 KejuLosera across the western part of Loitigon. Soil pit 1/1involved cleaning-up the face of a bank being undercut bythe river; the other six test pits were excavated with amechanical excavator. Soil pit profiles were described in thefield. Because of the nature of the local sedimentary environ-ment, it is likely that rates of sedimentation will have variedgreatly, both spatially and temporally, and there will behiatuses. Soil pit 6/2 (236 cm deep) appeared to show themost complete sequence of sediments and was thereforeselected for further, more detailed study.

AMS radiocarbon (14C), Ai13CbUlk, organic carbonand XRD analysesSoil pit 6/2 yielded four samples for AMS 14C dates (Table 1),17 samples for 613Cbulk and organic carbon determinations andsix samples for XRD analysis. A further 18 surface soil samplesfrom the vegetation plots were analysed for 513Cbulk andabundances of organic carbon. 613Cbulk and abundances oforganic carbon were established using a Thermo DeltaPlusContinuous Flow Isotope Ratio Mass Spectrometer (CF-IRMS) interfaced via a Conflo III with a CE Instruments1112 Flash elemental analyser, following pretreatment withconcentrated sulphuric acid to remove inorganics. ReferenceCO2 was calibrated using International Atomic EnergyAuthority Reference Standards and crosschecked usingL-Alanine. The results from analysing up to four replicatesamples per depth indicate that the accuracy of 613Cbulk.estimates is higher than 0.6%.. An additional four estimatesof 613Cbulk content of samples from soil pit 6/2 are availablefrom the results of the AMS dating analysis. Mineralogy wasdetermined using XRD (Phillips PW1720 with a PhillipsPW1050 diffractometer and a Phillips PW3313/20 Cu k-alphaanode tube). All measurements were taken from 2° to 350 (20)at a step size of 0.20 per second.

Analyses of organic matter, subfossil pollen andspores, and charcoal contentsSediment samples from soil pit 6/2 were processed for theirorganic matter, subfossil pollen and spores and charcoalabundances. For estimates of organic matter content, oven-dried sediment samples were ignited in a furnace at 550°C(Bengtsson and Enell, 1986). Samples for subfossil pollen andspore analyses were prepared following the standard procedureof Faegri and Iversen (1989). A total of 39 samples wereinitially prepared and enumerated. In these, total countsranged from 0 (two samples, 161 cm and 171 cm, were foundto be completely lacking in pollen) to 4675 grains. Down-profile variations in microscopic charcoal abundances wereestablished in 15 1-cm3 samples using the point-count method(Clark, 1982). Samples for charcoal analysis were prepared asfor pollen and spore analysis, with the addition to each sampleof one tablet containing a known number of Lycopodiumspores. In order to provide a means of separating local fire

signatures from regional background burning, fragments thatwere recognizably charcoal were grouped according to thelength of their longest axis into two size classes: 2.5-100 gimand > 100-150 gm (the mesh size of the sieve used in samplepreparation). According to Thevenon et al. (2003), theabundant presence of microscopic charcoal fragments with along axis > 63 gm attests the occurrence of local fires. Thesmallest fragments of opaque material (long axis <2.5 gm)were ignored because these are the most difficult to identify ascharcoal to an acceptable level of certainty.

ResultsVegetation dataA total of 79 plant taxa was recognized in the vegetationsurvey: 19 of 24 grass taxa were identified to species level; threesedge and 52 forb taxa were also recorded (43 of the forbs wereidentified to genus level and below). Some grass taxa werecommon to many of the plots. For example, Pennisetumstramineum was abundant in areas of vegetation that hadbeen burnt in the last two decades, and in areas that had not,and was also the most common grass encountered in Acaciabushland. Cynodon dactylon was also recorded in the majorityof plots. Other grass taxa had a more restricted distribution.For example, Themeda triandra was common in well-drainedgrassland on soils derived from granite but not in Acaciabushland, meadows or soils derived from lava. T triandra wasalso recorded in vegetation that had been burnt in the last twodecades, but also reached its maximum abundance for thesurvey (80%) in vegetation that had remained unburnt. Setariasphacelata, a second member of the Pennisetum genus,P clandestinum and the sedge Cyperus rhotundrus wereabundant in well-grazed, damp grassland. Forb taxa weregenerally less ubiquitous than the common grasses. Pentanisiaouranogyna and Solanum incanum were both common in well-drained grassland on soils derived from granitic bedrock.Commelina petersii and C. reptans were locally common (theformer in association with Acacia bushland), as were Indigoferavolkensii, various Ipomoea spp and Leucas grandis. Justiciauncinulata was relatively abundant in recently burnt, lowAcacia bushland and on the grassy floodplain at Loitigon.

Figure 2 illustrates the results of CCA. Mean 613Cbu1k valuesfor surface soil samples from each of the 18 plots are alsoshown, as are the directions of change in the three environ-mental variables estimated in the field. Generally, plot sampleswith a similar floristic composition and in similar environ-ments are grouped close together on the plot; samples with adissimilar composition are located farther apart. No overallcorrelation is evident between any one of the three environ-mental variables and the mean 613Cbu1k values for surface soilsamples, presumably because the surface soil samples reflectthe combined effects of a complex of environmental conditionsover a longer period of time than the last two decades of firehistory considered here. Some relationship between the habitatlocation of the plots and the mean 613CbU,k values is apparent,however: mean 613Cbu1k values for surface soil samples fromgrassland plots within Acacia bushland are on the whole lower(- 17.0% +2.0%.) than those from grassland areas (- 14.4%. +2.0%0), possibly because of the incorporation of a greateramount of organic carbon from C3 sources (e.g., leaf litter fromAcacia) (Sage et al., 1999).

Sediment stratigraphy and chronologySediments exposed at the study site were generally brown-coloured silts and clays of fluvial origin. These grade withdepth into an evaporite-encrusted, lava gravel- and sand-rich

© 2005 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at RUTGERS UNIV on April 27, 2008 http://hol.sagepub.comDownloaded from

D. Taylor et al.: Vegetation dynamics on the Laikipia Plateau, Kenya 841

C'.' 16-14.7%c 16

Dry 0 14.9'l| lAcacia 15L2iIIbushland

18EI'-15.4%0 Damp

| -13.7%c |1 7 Frequently grassland

-137% -17% |1 burntL ostr

I7 Infrequeny

burnt-2.0 3.0

Figure 2 Ordination bi-plot for vegetation and environmental data. 613C (%/,O) values for surface soil samples from each 1-rn2 vegetation plotare shown but were not used in the ordination. 0, position of sample(s) according to CCA axes 1 and 2; numbers around the 0 arevegetation plot numbers. Plots 1-3: grassland, burnt in 1984, not in 1999, 1965 m a.s.l. Plots 4-6: grassland, not burnt in last two decades,1963 m a.s.l. Plots 7-9: grassland, burnt in 1984 and in 1999, 2000 m a.s.l. Plots 10, 11: grassland within Acacia bushland, not burnt in lasttwo decades, 1895 m a.s.l. Plot 12: damp grassland by perennial spring, 2030 m a.s.l. Plots 13, 14: grassland within low Acacia bushland,burnt in 1984 and 1999, 1859 m a.s.l. Plots 15, 16: burnt grassy vlei, 1785 m a.s.l. Plot 17, 18: burnt grassland within low Acacia bushland,1785 m a.s.l. Arrows with solid lines, the main direction of variation of the environmental gradients

horizon directly above bedrock. A layer of dark brown/black-coloured clay, up to 30-40 cm thick, was present in theotherwise paler brown-coloured silty clays in all but one of thesoil pits (7/2) and was a useful marker horizon.

Sediments exposed in soil pit 6/2 were predominantly offluvial origin. According to the AMS 14C dates obtained,which are stratigraphically consistent, the upper 170 cm or so

of the profile dates to the last 6000-7000 years. The uppermost85 cm of sediments were dark brown, organic, silty clays; thedark brown/black-coloured clay marker horizon occurred from96 to 126 cm, while a pale, sandy layer occurred from 85 to96 cm. Sediments from below 128 cm contained variableamounts of large fragments of lava and coarse gravel, coatedwith white evaporite, and sand. Large fragments, coated withwhite evaporite, were particularly common below 160 cm.

XRD analysis revealed marked similarities in mineralogydown the profile, suggesting little variation in the source ofsediments. Although distinctive inorganic provenance indica-tors were absent, the presence of alkali feldspars is consis-tent with the weathering of phonolitic volcanic rocks thatoccur close to the study site. Samples from 135-137 cm and215-217 cm contained significant gypsum peaks (hydratedcalcium sulphate and presumably the white evaporite noted inthe field).

813Cbulk and organic carbon613Cbulk values for surface soil samples ranged from -18.8&(grassland within Acacia bushland and not burnt within thelast two decades) to -12.3%o7 (grassland burnt twice within thelast two decades); the overall mean 613CbUIk value for the18 soil samples was -14.8%o (±2.1). Lower variation isevident in 6'3Cbulk values for samples from soil pit 6/2, whichranged from -17.8%. to -12.6%O (mean = -14.8%o+1.4)(Figure 3), thus falling within the range of values for thesurface soil samples. The four estimates of 6'3CbUlk associated

with the AMS dates fall within the range of 613Cbu1k values forsamples from soil pit 6/2.Lowest values were associated with sediments in the lower

part of the profile: samples from below 150 cm yielded a mean613Cbulk. value of - 15.9%7 (± 1.2), whereas those from abovethis depth yielded a mean of - 14.0%. (±0.9). Overall the613Cbulk analyses indicate a range of plant sources for organiccarbon, from a mixture of C4 and C3 to predominantly C4:according to Cerling (1999), 613C values for organic carbonderived from C3 plants generally range from - 35%° to- 20%0, whereas carbon from C4 plants generally rangesfrom -14%. to -10%7 (-16%. to -9%c,, according to deFreitas et al., 2001). Levels of organic carbon (%) for samplesfrom soil pit 6/2 were generally low, ranging from 0.1% to0.7% (mean = 0.4%+ 0.2). Values were slightly higher for thesurface soil samples analysed, ranging from 0.6% to 2.6%(mean = 1.5%+0.6).

Subfossil pollen and spores and charcoalDown-profile variations in pollen and spores data are shown inFigure 4. Generally samples below 106 cm were far lesspolliniferous than those above, and below 156 cm were toolacking in pollen and spores for reliable counts (the meanpollen count for the nine samples analysed from 161 cm andbelow in the profile was four grains), presumably because ofpoor conditions for preservation. Four pollen zones wererecognised: M4 (0-23.5 cm), M3 (23.5-73.5 cm), M2 (73.5-98.5 cm) and Ml (below 98.5 cm). The lowermost zone, Ml,was subdivided into two subzones. The lower of these (MIa,below 118.5 cm) was characterized by low total pollen andspore counts. According to the age-depth relationship(Figure 5), pollen zone boundaries have the following ages:Ml-M2, c. 1900 BP; M2-M3, c. 1700 BP; and M3-M4,c. 700 BP The estimated age for the Mla-lb subzoneboundary is c. 2300 BP.

© 2005 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at RUTGERS UNIV on April 27, 2008 http://hol.sagepub.comDownloaded from

842 The Holocene 15 (2005)

~~~~~A

1171

I 20 40 660 o 0(x 1000) (X 100)

0.2 0.4 OA0

I

. 1

lb

V-.I

700 BP

1700 BP

1900P

-18 -14 -10

* b'"C vOki tromAM '4C da* (%,)0 fC4t VbOADo %arbon

Figure 3 Stratigraphy and contents of organic matter (%), charcoal (number of particles per unit volume of sediment), 613C (%.) and organicC (%): soil pit 6/2. For basis for extrapolated/interpolated radiocarbon ages, see Figure 5

Down-profile variations in the abundances of microscopiccharcoal, expressed as numbers of fragments per unit volumeof sample, are illustrated in Figure 3. Two peaks are evident:

c. 80-120 cm, a result of large amounts of smaller charcoalfragments; and 0-c. 20 cm, comprising charcoal fragmentswith a long axis > 100 gm.

p Afrornontane

;I/~~~~4

20 40 6

I / Bushlond

Is,oSP/ nflcd1'*°P/oe '4,-- 4eDe 'A,

20 40 6d 20 4 2 20202~040~ 6080lO0

7 / Others

.,,0

F -~~~~~~~~~~~~~~~~~~~~~~~~~~~CAH--I0H-1

20 40 6 0 40 60 80100 20 40 60 8C

Figure 4 Subfossil pollen and spore data for soil pit 6/2. Afromontane (forest) and bushland taxa and Ricinus percentages shown arecalculated from a sum that does not include Cyperaceae, Poaceae, spores and damaged categories. The percentages of Cyperaceae andPoaceae are based upon a sum that includes all pollen and spore types, excluding damaged grains, while damaged grains are expressedaccording to a sum that includes all pollen and spore types, including the damaged category

Al4kto

0

14200

1760*5-40

209040 120'

6630. 1600-40

/

0

(%

ell

90

20

40

1420 +/- 501

60

80'E1760 +/- 401

2090 +/- 401

100

120

Zone,

M4

M3

M2

Mlb

Mla

140

160

6630 +/- 401

-_

iI

© 2005 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at RUTGERS UNIV on April 27, 2008 http://hol.sagepub.comDownloaded from

D. Taylor et a!.: Vegetation dynamics on the Laikipia Plateau, Kenya 843

0 -l

20 -

40 -

60 -

so-E

I.,6) 100 -

120 -

140

160 -

180 _-

+ = AMS 14C date

M3-M4 (c.700 BP)

1420+I-50 BP

M2-M3 (c. 1700 B

+ 1760 +/-40 BP

Ml -M2 (c. 1900

;i,. +/I40BP

6630 +/-40BP

1000 2000 3000 4000 5000 6000 7000

Radiocarbon years BP

Figure 5 Age-depth curve for soil pit 6/2. Data are from Table 1

Synthesis

The lowermost pollen zone, Pollen zone MI, from about165 cm to 98.5 cm (c. 6600-1900 BP), is divided into two sub-zones: Mla (165-118.5 cm, c. 6600-2300 BP) and Mlb (98.5-118.5 cm, c. 2300-1900 BP). Sediments dating to zone MIare generally inorganic and include sand- and gravel-sizedmaterial, indicating deposition in flowing water. Angular frag-ments of lava, coated with gypsum, are present below 128 cm.

Overall, pollen zone MI is characterized by low abundancesof pollen and by a low rate of sedimentation (1 cm per 90.6 yrfor 165-115 cm, as opposed to 1 cm per 18.2 yr for above115 cm), or possibly one or several hiatuses in sedimentation.The abundance of pollen is lowest in subzone Mla. The fewpollen grains present suggest the occurrence of Afromontaneforest (containing Podocarpus), and possibly also Acaciabushland (although the pollen indicators of the latter are

actually few, in keeping with the very low pollen counts overallin this subzone). Charcoal abundances are also relatively low.Some support for the occurrence of more woody vegetationlocally than is the case today is provided by the relatively low613Cbulk values, particularly in the lowermost part of theprofile, which indicate incorporation of organic carbon fromC3 plant sources. Pollen is more abundant in subzone MIb andindicates the continued presence of Afromontane forest,although because of the (fluvial) sediments and the pollentypes involved (high potential for wide dispersal) pollensources some distance from the study site cannot be dis-counted. A major peak in the total number of charcoalparticles in this subzone suggests fires were common (thedark-coloured horizon noted in the field in several of the soiltest pits and between 96 and 126 cm in 6/2 may thus owe itsdark colouration to the presence of abundant charcoal in thesediments). Furthermore, available 613Cbulk values are rela-tively high and similar to those established for surface soilsamples in C4 grassland. Taken together the data appear toindicate some disturbance of forest vegetation and the spreadof fire-adapted grassland.The disturbance of Afromontane forest by burning appears

to have preceded the spread of Acacia bushland, marked by the

pollen zone M -M2 boundary. Sediments dating to Pollenzone M2 (98.5-73.5 cm, c. 1900-1700 BP) are silty clays withsome sand and fine organic remains, suggesting the occurrenceat the sample site of a lower energy depositional environmentthan was previously the case. Pollen is abundant and therelatively low 613Cbu1k values indicate the incorporation oforganic carbon from C3 plant sources. Samples from this zone(and close to the M2-M3 zone boundary) contain moderatelyhigh levels of charcoal, and one sample contained relativelyhigh levels of the largest size fraction. The data thus indicate anexpansion of Acacia (C3) bushland locally, possibly as a resultof increased burning.

Pollen zone M3 is from 73.5-23.5 cm (c. 1700-700 BP).Sediments of zone M3 age are silty clays with fine organicfragments, indicating a relatively quiescent depositionalenvironment when compared with deeper in the profile.The charcoal data indicate the continued occurrence offires, while the pollen and 613Cbulk data appear to representthe replacement of Acacia bushland with fire-adaptedgrassland. Pollen from Justicia, a widely occurring genus ineastern Africa (Agnew and Agnew, 1994) and recorded atLoitigon in grassland and on adjacent ferralsols in recentlyburnt Acacia bushland, is common in this zone, while Ricinuspollen, presumably from R. communis (the castor plant), isalso present. According to Agnew and Agnew (1994),R. communis occurs naturally in woodland-grassland ecotonesin Kenya, but is also associated with disturbed and degradedsoils over a broad range of altitudes (500-2000 m). Vincenset al. (2003: 330) interpret the presence of Ricinus pollen,in association with forest disturbance indicators, in a sedi-ment sequence from southern Tanzania, as possibly represent-ing farming.The uppermost pollen zone, Pollen zone M4, is from 23.5 cm

to the surface (c. 700 BP to the present). The pollen and6 3Cbul data indicate a further reduction in the extent ofAcacia bushland in the catchment and an expansion of fire-adapted grassland, while a peak in the largest size fraction ofcharcoal analysed could indicate the local occurrence ofvegetation fires.

40

© 2005 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at RUTGERS UNIV on April 27, 2008 http://hol.sagepub.comDownloaded from

844 The Holocene 15 (2005)

Discussion

Data presented above provide evidence of significant mid- tolate-Holocene vegetation changes on the Laikipia Plateau.Although the pollen data are unable to provide detailedevidence of vegetation conditions for the earliest part of therecord from Loitigon, a greater contribution of C3 plantsources of organic carbon to sediments than was subsequentlythe case and the almost lack of the largest-sized fragments ofcharcoal suggest more wooded conditions and less frequentfires locally than today during the period c. 6600-2300 BP.Furthermore, the age-depth curve suggests a low rate ofsediment accumulation during this period, or perhaps more

likely the occurrence of one or several hiatuses in sedimenta-tion. Thus, although fragmentary and not highly resolved, theavailable data suggest that environmental conditions at Loiti-gon before c. 2300 BP were a response to higher levels ofeffective precipitation and lower levels of human activitythan today. Climatically relatively wet conditions are inkeeping with existing palaeoclimatic information from easternAfrica that suggests the early to mid Holocene was, on thewhole, more humid than the present (e.g., Barker et al., 2001;Stager et al., 2003), with climatic aridification and increasedseasonality setting in from around 6000-4000 BP (Nicholsonand Flohn, 1980; Gasse, 2000, 2002; de Menocal et al., 2000;Stager et al, 2003).More detailed information concerning past conditions at

Loitigon is available from c. 2300 BP. The available archae-ology suggests that the Laikipia Plateau was by this time a siteof food production involving domesticated animals. The pulseof burning and the replacement of Afromontane forest byAcacia bushland and fire-adapted grassland may represent theactivities of these early pastoralists. Climate change, inparticular reduced levels of effective precipitation and possiblyenhanced seasonality, may have facilitated burning and con-

tributed to more quiescent sedimentary conditions at the studysite; subzone Mlb (c. 2300-1900 BP) roughly coincides withpronounced low stands in Lake Tanganyika (dated c. 2200 BP,in Alin and Cohen, 2003) and Lake Edward (dated c. 2000 BP,in Russell et al., 2003). The palaeoclimate data are notunequivocal for this period, however; evidence from MountKenya indicates that the period from 2900 BP to 1900 BP was

one of heavy convective rainfall, enhanced soil erosion,Neoglacial ice advances and forest expansion on the mountain(Barker et al., 2001).

Increased burning of vegetation on the Plateau may havebeen linked to the desirability of clearing forest and woodlandto improve and extend grazing land, and possibly also toreduce the threat of insect-borne diseases such as Trypanoso-miasis (Gifford-Gonzalez, 1998). According to the charcoal,pollen and 6'3Cbulk data for zone M3 (c. 1700-700 BP),burning led to the replacement of Acacia bushland by fire-adapted, C4 grassland. Ricinus pollen may have been producedby plants associated with disturbed soils around settlements or

cultivated land as the period incorporates the later stages of thePN and the early part of the PIA in the region, and thereforethe use of iron implements and a greater importance of cereals(Bower, 1991). The earlier part of the period also coincideswith archaeological data that indicate greater settlementmobility, and with evidence for reduced levels of effectiveprecipitation (levels of water in Lake Tanganyika were lowAD 200-500, and again at AD 700-850 (Alin and Cohen,2003)); it is possible that prolonged phases of reduced levels ofeffective precipitation triggered reliance on more mobilesettlement strategies. However, the palaeoclimatic evidence isnot entirely harmonious. For example, variations in effective

precipitation evident in Lake Tanganyika sediments are less soin Lake Victoria (Stager et al., 2003) and are absent fromMount Kenya (Barker et al., 2001), although Karien et al.(1999) suggest that a phase of glacier re-advance on MountKenya around AD 700-800 may have been caused by lowertemperatures (associated with reduced convective rainfall).

Further replacement of Acacia bushland by grassland isevident from around 700 BP. The abundance of the largestsized fraction of charcoal in sediments suggests that fire was animportant factor driving vegetation change, and that at leastsome of the fires were located close to the study site. There is agood deal of controversy concerning climatic variations ineastern Africa during the early part of the second millenniumAD (see Alin and Cohen, 2003). Least controversy appears tosurround the period from the late AD 1 500s through to the late1700s, which appears to have been characterized by longintervals when levels of effective precipitation in eastern Africawere much reduced relative to the present and followed a phaseof climatically relatively humid conditions (Robertshaw andTaylor, 2000; Taylor et al., 2000; Verschuren et al., 2000; Alinand Cohen, 2003; Robertshaw et al., 2004). Within thiscontext, and bearing in mind the limitations of the evidenceso far yielded by archaeological excavations, the Maasai Plainssite potentially represents prolonged or repeated humansettlement and exploitation of resources close to Loitigonduring a period when levels of effective precipitation weresufficient to support relatively large, possibly semi-permanentencampments of herders and their animals.

Conclusions

(1) Analyses of sediments from the Laikipia Plateau, Kenya,have demonstrated the potential of river floodplain deposits assources of evidence of Holocene vegetation changes in semi-arid eastern Africa. XRD analysis of sediment samplessuggests that the source area for material accumulating at thesite has remained relatively constant; palaeoenvironmentalreconstructions have been facilitated by reference to contem-porary relationships between plants and environment, and toongoing archaeological investigations..(2) Although patchy, the available data indicate more woodedand relatively humid conditions compared with the presentfrom c. 6600-2300 BP.(3) Vegetation changes at Loitigon from c. 2300 BP areassociated with indicators of increased burning and involve thereplacement of Afromontane forest by fire-modified vegetationin the form of Acacia bushland and grassland. Burning ofvegetation, to extend and improve pasture and possibly toeradicate disease-prone habitats, may have been facilitated byprolonged periods of moisture deficits. The most recentsignificant vegetation change, in the form of an expansion offire-adapted grassland c. 700 BP, is associated with evidence forthe occurrence of fires locally. It could be that changes invegetation in the early part of the second millennium ADrepresent the impacts of sizeable encampments of people andtheir animals close to the study site.

AcknowledgementsResearch presented here forms part of a British Academy-funded regional interdisciplinary study, 'Landscape and envir-onmental change in semi-arid east and southern Africa;develpping interdisciplinary approaches'. Thanks are dueto the Government of Kenya for permission to carry out

© 2005 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at RUTGERS UNIV on April 27, 2008 http://hol.sagepub.comDownloaded from

D. Taylor et al.: Vegetation dynamics on the Laikipia Plateau, Kenya 845

fieldwork in 2003 and 2004, to the British Academy for corefunding and to the British Institute in Eastern Africa forlogistical support. Thanks are also due to the owners of MugieRanch, to Klaus and family at Mugie and to Fumi and Robertat Lolldaiga Hills; to Fiona, Max and Joseph for assistance inthe field; to Henry Lamb, Peter Robertshaw and Louis Scottfor their very helpful comments on earlier drafts; and finally toSheila McMorrow for Figures 1, 2 and 5. The Trinity CollegeResearch Maintenance fund covered the cost of 613C and XRFanalyses.

References

Agnew, A.D.Q. and Agnew, S. 1994: Upland Kenya wildflowers.Nairobi: East Africa Natural History Society.Alin, S.R. and Cohen, A.S. 2003: Lake-level history of LakeTanganyika, East Africa, for the past 2500 years based onostracode-inferred water-depth reconstruction. Palaeogeography,Palaeoclimatology, Palaeoecology 199, 31-49.Ambrose, S.H. 1998: Chronology of the Later Stone Age and foodproduction in East Africa. Journal of Archaeological Science 25,377-92.Barker, P.A., Street-Perrott, F.A., Leng, M.J., Greenwood, P.B.,Swain, D.L., Perrott, R.A., Telford, R.J. and Ficken, K.J. 2001: A14,000-year oxygen isotope record from diatom silica in two alpinelakes on Mt. Kenya. Science 292, 2307-10.Bengtsson, L. and Enell, M. 1986: Chemical analysis. In Berglund,B.E., editor, Handbook of Holocene palaeoecology and palaeo-hydrology. Chichester: Wiley, 423-50.Bower, J. 1991: The Pastoral Neolithic of East Africa. Journal ofWorld Prehistory 5, 49-82.Bower, J. and Nelson, C.M. 1978: Early pottery and pastoralcultures of the central Rift Valley, Kenya. Man 13, 554-66.Cerling, T.E. 1999: Paleorecords of C4 plants and ecosystems. InSage, R.F. and Monson, R.K., editors, C4 plant biology. London:Academic Press, 445-69.Clark, R.L. 1982: Point count estimation of charcoal in pollenpreparations and thin sections of sediment. Pollen et Spores XXIV,523-35.Collett, D.P. and Robertshaw, P.T. 1983: Problems in theinterpretation of radiocarbon dates: the Pastoral Neolithic ofEast Africa. African Archaeological Review 1, 57-74.de Freitas, H.A., Pessenda, L.C.R., Aravena, R., Gouveia, S.E.M.,Ribeiro, A.d-S. and Boulet, R. 2001: Late Quaternary vegetationdynamics in the southern Amazon basin inferred from carbonisotopes in soil organic matter. Quaternary Research 55, 39-46.de Menocal, P., Ortiz, J., Guilderson, T. and Sarnthein, M. 2000:Coherent high- and low-latitude climate variability during theHolocene Warm Period. Science 288, 2198-202.Dickson, D.B. and Gang, G-Y. 2002: Evidence of the emergence ofmodern behaviour in the Middle and Later Stone Age lithicassemblages at Shurmai rockshelter (GnJml) and Kakwa Lelashrockshelter (GnJm2) in the Mukogodo Hills of north-centralKenya. African Archaeological Review 19, 1-26.Faegri, K. and Iversen, J. 1989: Textbook of pollen analysis.Chichester: John Wiley and Sons.Galaty, J.G. 1993: Maasai expansion and the new East Africanpastoralism. In Spear, T. and Waller, R., editors, Being Maasai.Oxford: James Currey, 61-86.Gang, G-Y. 2001: The Middle and Later Stone Ages in theMukogodo Hills of Central Kenya. BAR International Series 964.Oxford: Archaeopress.Gasse, F. 2000: Hydrological changes in the African tropics sincethe Last Glacial Maximum. Quaternary Science Reviews 1 9' 347-61.

2002: Kilimanjaro's secrets revealed. Science 298, 548-49.Gifford-Gonzalez, D. 1998: Early pastoralists in East Africa:ecological and social dimensions. Journal of AnthropologicalArchaeology 17, 166-200.

2000: Animal disease challenges to the emergence ofpastoralism in sub-Saharan Africa. African ArchaeologicalReview 17, 95 -139.Hackman, B.D. 1988: Geology of the Baringo-Laikipia area. Report104. Nairobi; Ministry of Environment and Natural Resources,Mines and Geological Department.Jacobs, A. 1972a: Appendix A. The discovery and oral history ofthe Narosura site. In Odner, K., editor, Excavations at Narosura, aStone Bowl site in the southern Kenya Highlands. Azania 7, 79-87.

1 972b: New archaeological sites in the Mukogodo area,Laikipia District. Nairobi: University of Nairobi, Institute ofAfrican Studies.Karega-Milnene 2002: Holocene foragers, fishers and herders ofWestern Kenya. BAR International Series 1037. Oxford:Archaeopress.Karien, W., Fastook, J.L., Holmgren, K., Malmstr6m, M.,Matthews, J.A., Odada, E., Risberg, J., Rosqvist, G., Sandgren,P., Shemesh, A. and Westerberg, L.-O. 1999: Glacier fluctuationson Mount Kenya since 6000 cal. years BP: implications forHolocene climatic change in Africa. Ambio 28, 409-18.Lamprey, R. and Waller, R. 1990: The Loita-Mara region inhistorical times: patterns of subsistence, settlement and ecologicalchange. In Robertshaw, P.T., editor, Early pastoralists of south-western Kenya. Memoir 11. Nairobi: British Institute in EasternAfrica, 16-35.Lind, E.M. and Morrison, M.E.S. 1974: East African vegetation.London: Longman.Marean, C. 1992: Hunter to herder: large mammal remains fromthe hunter-gatherer occupation at Enkapune ya Muto rockshelter.African Archaeological Review 10, 65-127.Marshall, F. 2000: The origins and spread of domestic animals inEast Africa. In Blench, R.M. and MacDonald, K.C., editors, Theorigins and development of African livestock: archaeology, genetics,linguistics and ethnography. London: UCL Press, 191-221.Marshall, F. and Hildebrand, E. 2002: Cattle before crops: thebeginnings of food production in Africa. Journal of WorldPrehistory 16, 99-143.Mutundu, K.K. 1999: Ethnohistoric archaeology of the Mukogodo innorth central Kenya. BAR International Series 775. Oxford:Archaeopress.Nicholson, S.E. and Flohn, H. 1980: African environmental andclimatic changes and the general atmospheric circulation in latePleistocene and Holocene. Climatic Change 2, 313-48.Robertshaw, P.T. 1988: The Elmenteitan: an early food-producingculture in East Africa. World Archaeology 20, 57-69.

editor 1990: Early pastoralists of south-western Kenya.Memoir 11. Nairobi: British Institute in Eastern Africa.Robertshaw, P. and Taylor, D. 2000: Climate change and the rise ofpolitical complexity in western Uganda. Journal of African History41, 1-28.Robertshaw, P., Taylor, D., Doyle, S. and Marchant, R. 2004:Famine, climate and crisis in western Uganda. In Battarbee, R.,Gasse, F. and Stickley, C., editors, Past climate variability throughEurope and Africa. Amsterdam: Springer/Kluwer, 535-49.Russell, J.M., Johnson, T.C., Kelts, K.R., Lardal, T. and Talbot,M.R. 2003: An 11000-year lithostratigraphic and paleohydrologicrecord from equatorial Africa: Lake Edward, Uganda-Congo.Palaeogeography, Palaeoclimatology, Palaeoecology 193, 25-49.Sage, R.F., Wedin, D.A. and Li, M. 1999: The biogeography of C4photosynthesis: patterns and controlling factors. In Sage, R.F. andMonson, R.K., editors, C4 plant biology. London: Academic Press,313-73.Shackelton, R. 1946: The geology of the country between Nanyukiand Maralal. Geological Survey of Kenya, Report No. 11. Nairobi:Government Printer.Siiriainen, A. 1977: Later Stone Age investigation in the Laikipiahighlands, Kenya: a preliminary report. Azania 12, 161-86.

1984: Excavations in Laikipia: an archaeological study of therecent prehistory in the Eastern Highlands of Kenya. Helsinki:Suomen Muinaismuistoyhdistksen Aikakauskirja FinskaFornminnesforenengens Tidskrift.

© 2005 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at RUTGERS UNIV on April 27, 2008 http://hol.sagepub.comDownloaded from

846 7he Holocene 15 (2005)

Stager, J.C., Cumming, B.F. and Meeker, L.D. 2003: A 10,000-yearhigh-resolution diatom record from Pilkington Bay, Lake Victoria,East Africa. Quaternary Research 59, 172-81.Stuiver, M., Reimer, P.J., Bard, E., Beck, J.W., Burr, G.S., Hughen,K.A., Kromer, B., McCormac, F.G., v.d. Plicht, J. and Spurk, M.1998: INTCAL98 radiocarbon age calibration 24,000-0 cal BP.Radiocarbon 40, 1041-83.Taylor, D., Robertshaw, P. and Marchant, R.A. 2000: Environ-mental change and political-economic upheaval in precolonialwestern Uganda. The Holocene 10, 527-36.ter Braak, C.J.F. 1986: Canonical correspondence analysis: a new

eigenvector technique for multivariate direct gradient analysis.Ecology 67, 1167-79.

1987: The analysis of vegetation-environment relation-ships by canonical correspondence analysis. Vegetatio 69,69-77.

ter Braak, C.J.F. and Smilauer, P. 2002: CANOCO referencemanual and windows users guide: software for canonical commu-nity ordination (version 4.5). Ithaca NY: Microcomputer Power.Thevenon, F., Williamson, D., Vincens, A., Taieb, M., Merdaci, O.,Decobert, M. and Buchet, G. 2003: A late-Holocene charcoalrecord from Lake Masoko, SW Tanzania: climatic andanthropologic implications. The Holocene 13, 785-92.Verschuren, D., Laird, K.R. and Cumming, B.R. 2000: Rainfall anddrought in equatorial east Africa during the past 1,100 years.Nature 403, 410-14.Vincens, A., Williamson, D., Thevenon, F., Taieb, M., Buchet, G.,Decobert, M. and Thouveny, N. 2003: Pollen-based vegetationchanges in southern Tanzania during the last 4200 years: climatechange and/or human impact. Palaeogeography, Palaeoclimato-logy, Palaeoecology 198, 321-34.White, F. 1983: The vegetation of Africa. Switzerland: UNESCO.

© 2005 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at RUTGERS UNIV on April 27, 2008 http://hol.sagepub.comDownloaded from