A CatastrophicDestruction ofAfrican Forestsabout 2500Years Ago StillExerts a MajorInfluence onPresentVegetationFormationsJean Maley

IDS Bulletin Vol 33 No 1 2002

13

1 IntroductionExtensive research has shown that tropical forestsregressed considerably and fragmented during thelast glacial maximum dating from about 20 000 BP1

(Maley 1987 1996) which led to the extension ofice at high latitudes in both hemispheres Thesubsequent phase of maximal forest extensionbegan about 10 000 BP simultaneously with thephase of global warming corresponding to theHolocene period Until very recently numerousspecialists thought that the Central African forests(Figure 1) had not suffered major disturbanceduring the Holocene until the beginning of thetwentieth century and the current period ofintensive forest exploitation Yet during the lastdecade of the twentieth century geological andpalynological research conducted on lakesediments from several sites has shown that about2500 BP the Central African forest experiencedcatastrophic destruction with a major extension ofthe savannas (Maley and Brenac 1998a)

This article presents the principal data concerningthe event and considers its significance forvegetation development both during the Holoceneand up to the present day Then taking theexample of two species it will show how theirhistory has been profoundly affected by this eventFinally it will try to contextualise the event in theregional and global paleo-climatic framework bycomparing it with earlier phases of forestregression

2 Variations in the vegetation ofCentral Africa during theHoloceneOnly one site in Atlantic Central Africa lakeBarombi Mbo in west Cameroon provides adetailed pollen record which dates back beyondthe Holocene to about 28 000 years BP (Maley andBrenac 1998a) (Figure 2)

The records from other sites begin during theHolocene Only at Barombi Mbo has it been possibleto examine the installation of the forest environmentat the beginning of the Holocene and subsequentvariation in the principal tree taxa (Maley and Brenac1998a) This variation reveals pseudo-periods ofabout 2000 to 2500 years for several tree taxatypical of mature (or lsquoprimaryrsquo) forest (Figure 3a) A

14

Figure 1 Schematic map of present vegetation in Central Africa (Maley 1990)(1) Biafran and Gabonese evergreen forests with numerous Cesalpiniaceae (2) Atlantic coastal forests withSacoglottis gabonensis and Lophira alata in Cameroon with the addition of Aucoumea klaineana (Okoumeacute) inGabon (see 11) The boundary between 1 and 2 is irregular and often progressive (3) Forests of Congolesetype with mixture of evergreen and semi-deciduous formations (4) Open canopy forests with a dense groundcover dominated by large Monocotyledons belonging to Marantaceae and Zingiberaceae (5) (a) Mixture oftypes 4 and 5b (b) Evergreen forests with a high canopy dominance of Gilbertiodendron dewevrei(Caesalpiniaceae) (6) Zone flooded almost all the year with evergreen formations (eg 5b) raphiales and otherhygrophylous formations (7) Semi-deciduous forests (8) Mangroves (9) Various mountain formations (10) Savannas (11) Boundary of the area with Okoumeacute (from Brunck et al 1990 and unpublished data) (12) Current mean maximum extension of lsquocoolingrsquo influences (according to Saint-Vil 1984) (13) Boundariesbetween States

minimum for all these variations in tree taxaoccurred between 2500 and 2000 BP during themajor phase of destruction The inverse ndash peaks ofmaximum extension ndash are not synchronous betweentaxa with variations corresponding to long phases ofsylvigenesis and the progressive or abruptreplacement of taxonomic groups by others Cyclesof shorter duration (for example 1000 500 or 100years) have not been shown to date as the scale ofsampling is insufficiently fine2 The pseudo-periodsof 2000 to 2500 years can be related to climaticcycles of the same duration frequently observedacross the globe in particular during the Holoceneand to one of the dominant periodicities of solaractivity about 2300 years (Magny 1993 Maley andBrenac 1998a)

21 The phase of forest destructionculminating about 2500 years BPThis phase of forest destruction (Figure 4) wasgeneral and synchronous as can be deduced fromboth its abrupt occurrence between 3000 and2500 years BP and its repeated discovery at everysite studied in detail in the northern part of theforest domain in west and south Cameroon atBarombi Mbo (Maley and Brenac 1998a)Mboandong (Richards 1986) Njupi (Zogning etal 1997) and Ossa (Reynaud-Farrera et al 1996van Geel et al 1998) in the southern part of theforest domain and in the Mayombe of westernCongo at Kitina (Elenga et al 1996) Kakamoeacuteka(Maley and Giresse 1998) Songolo near to Pointe-Noire (Vincens et al 1999) and at Ngamakala on

15

Figure 2 Site of Lake Barombi Mbo altitude of ca 300 m in the forests of west Cameroon Syntheticpollen diagram (per cent) with an interpolated chronology based on 12 radiocarbon datesBlack dots represent total percentages of tree pollen open area total percentages of pollen fromherbaceous plants mainly Gramineae and light dots the pollen percentages of Cyperaceae hygrophilousplants linked to the lowering of the lake levelSource Maley and Brenac 1998a

16

Figure 3a Site of Lake Barombi Mbo west Cameroon percentages of diverse tree pollenTypical tree taxa from mature forestsSource Maley and Brenac 1998a

the Bateacutekeacute Plateau (Eastern Congo) (Elenga et al1994) Where dating can properly frame the eventit dates more precisely to between 2800 and 2500years BP

During this phase primary forest trees almostcompletely disappeared and in certain sites (inwest Cameroon at Barombi Mbo Mboandong andNjupi in south Congo and on the Bateacutekeacute Plateau)there was a major expansion of savannas Thisphase of forest destruction must have been verybrief in lacustrine strata where the abrupt andalmost complete disappearance of primary typetrees is observed an lsquoexplosionrsquo of pioneer woodyvegetation has been noted for some sites such as atlake Barombi Mbo in west Cameroon (Maley andBrenac 1998a) (Figure 3b) at lake Kitina in theMayombe western Congo (Elenga et al 1996) andat lake Ossa south Cameroon (Reynaud-Farrera etal 1996) The rapid expansion of pioneer treesmust have been the first phase of lsquoscabbingrsquo whichinitiated the reconstitution of the canopy The mostfrequent pollen taxa are the pioneers Alchorneacordifolia Musanga Trema Macaranga Elaeisguineensis etc Although brief destruction wasintense and several authors argue that the openingof the forest facilitated the major Bantu migrationfrom areas presently north of the forest to thosenow south of it (Schwartz 1992 Lavachery et al1996)

22 The reconstitution of the CentralAfrican forests during the last twomillennia

Humid conditions favourable for forest returnedearlier in west Cameroon in particular at BarombiMbo dating from 2000 years BP Yet thereconstitution especially of lsquoprimaryrsquo type forestswas not synchronous and probably reflects thedistribution of the residual forests (Figure 4) Thedelay can be understood in terms of lsquohysteresisrsquo(Maley and Brenac 1998a)

The reinvasion of savanna by forest continuedduring the twentieth century as observed overmuch of the forest-savanna transition region Thisis evidenced by the comparison of air and satellitephotographs of different dates from the 1950s tothe present (see Blanc-Pamard and Peltre 1984Maley 1990 1996 Fairhead and Leach 1995

17

Figure 3b Site of Lake Barombi Mbo westCameroon percentages of diverse tree pollenTypical tree taxa from pioneer forestsSource Maley and Brenac 1998a

18

Figure 4 Schematic drawing of the status of Central African vegetation at the phase of maximumperturbation and destruction of forest which culminated around 2500 BPLarge hatched areas represent the residual areas of the forest domain around 2500 BP which probablyconsisted of forest-savanna mosaic with predominantly larger forest islands Around these residual areasthe land would have been colonised by open vegetation above all savannas but also open canopy forestformations (see (4) in Figure 1) The base of this map presents the current status of Central Africanbiotopes as in Figure 1

1998 Servant 1996) For south Cameroon alone(south of the Adamaoua Plateau) forest gainedabout a million hectares during this period(Letouzey 1985 Youta Happi and Bonvallot 1996Achoundong et al 2000) Historical and palyno-logical research in Cocircte drsquoIvoire Nigeria Cameroonand Congo indicate that this reinvasion hascontinued over several centuries (Fairhead andLeach 1998 Maley 1999 Vincens et al 2000)Albeit with fluctuations and interludes it appearsto be the continuation of the forest reinvasionwhich began about 2000 years BP Currentreinvasion is thus the long-term result of themassive forest disturbance which happened around2500 years BP

A brief review of research findings concerning oilpalm and Okoumeacute can illustrate the history ofthese forest ecosystems over recent millennia

3 The history of oil palm and ofOkoumeacute in Central Africa duringthe Holocene impact of the forestdisturbance culminating about2500 years BP

31 The oil palm

Oil palm (Elaeis guineensis) has abundant andcharacteristic pollen and thus appears frequentlyin sedimented pollen profiles This palm is ofAfrican origin with its pollen and macro-remains(nuts) found in Equatorial African deposits datingback to the Tertiary (Zeven 1967 Dechamps et al1992 Maley 1996 1999 Maley and Brenac 1998aMaley and Chepstow-Lusty 2001)

The extent to which oil palm has been planted orits regeneration encouraged through land use hasbeen debated Certainly when people clear youngpioneer forests or fallows near to villages where oilpalm is naturally abundant they generallyconserve the palms effectively enriching themthrough fallow cycles (eg in Cameroon CentralAfrican Republic Cocircte drsquoIvoire Togo and Beninsee Maley 1999) Thus in 1937 the renownedbotanist Aubreacuteville characterised the lsquoimmense oilpalm grovesrsquo of southern Benin as lsquoa naturalformation simply improved by the inhabitantsover the centuriesrsquo Zevenrsquos research largely fromNigeria also concludes that with the exception of

modern industrial plantations most oil palms inAfrica are semi-wild and are not propagated bysowing or transplanting seedlings (see Zeven1972)

Ecologically oil palm is a pioneer that requires lightto complete the first stages of its growth so itdevelops naturally in windfall-clearings andespecially on the periphery of dense forest after thepassage of fire which it can tolerate (Swaine andHall 1986 Swaine 1992) The botanist Letouzey(1978 1985) has described a vast natural oil palmgrove in west Cameroon and near the north-west ofthe forest zone (Figure 5) A band of forest 10ndash20km wide dominated by large and numerous oilpalms extends over more than 150 km Thisfollows the boundary between forest and savannaat times from 5 to 30 km inside the forest From theabsence of trees classically found in anthropicplantations and other criteria Letouzey (19781985) concluded that this grove is a natural standA similar example has been described on theeastern flank of Mount Nimba (GuineaLiberianborder) by Schnell (1946) who considered that thedissemination of palm nuts was facilitated by bothtoucans and chimpanzees whose faeces oftencontain it abundantly These natural groves haveattracted migrant people (eg Guille-Escuret 1990Maley 1999) For example many Bamileke peoplemoved to the natural groves of western Cameroonin the mid nineteenth century (Barbier 1981Warnier 1985 Perrois and Notue 1997 Maley1999) In the vast lsquoGrassfieldsrsquo just north of westCameroonrsquos forest bloc the forest islands are still ina phase of expansion especially due to the oil palmwhich is one of the principal pioneer trees Close tothe Donga river the Wuli people colonised thesenatural stands Baeke (1996) explains how theWuli install their villages in the palm groves whereassorted social rules oppose all plantation of palmsMoreover she reports that a Wuli myth of originlsquoprovides evidence for the prior use of oil palms tothe working of land and clearly distinguishes twotypes of plant exploitation gathering andagriculturersquo (Baeke 1996)

As already noted during the major disturbanceabout 2500 BP oil palm was a principal pioneerthat subsequently developed strongly (Figure 3b)In west Cameroon the rapid forest recolonisationwhich occurred from 2000 years ago in the

19

vicinity of Barombi Mbo (Maley and Brenac1998a) was accompanied by a retreat of oil palmprobably because ecological and climatic factorsrendered other pioneers more active at this time InFigure 3 it is notable that in west Cameroonbetween 2000 and 1000 years BP the pioneer treeMilicia was in a phase of very strong expansionThen here and elsewhere in central Africa a secondphase of oil palm expansion began around 1400years BP This culminated towards 1000 BP at lakeOssa near to Edea (Reynaud-Farrera et al 1996) inthe south of the Central African Republic atNouabale-Ndoki (Fay 1997 Maley 1999) but also

further south in Gabon to La Lopeacute (White et al2000 Maley unpublished) and in western Congonear to lake Kitina (Elenga et al 1996) This phaseending about 700800 BP must have beenassociated with renewed forest disturbance butweaker than that 2500 BP This disturbance wasalso associated with a short phase of soil erosionand with a discontinuity in the deposits observedin assorted locations in Central Africa (Maley andBrenac 1998b) An increase in mineral fluxes hasbeen found between 1200ndash800 years BP for thesites of Kitina and Sinnda in western Congo(Bertaux et al 2000) From the thirteenth centurya phase of forest recovery has occurred in CentralAfrica with a new development of mature forestsand a generalised retreat of oil palm a recoverydespite increased human settlements (Alexandre1965 Vansina 1990)

At the western extension of the Central Africanforests lies a large area without forest which bio-geographers call the Dahomey Gap stretchingacross southern Togo and Benin Much of this iscovered by an immense palm forest of naturalorigin as it was described by Aubreacuteville (1937)and Mondjannagni (1969) Some pollen data showthat this area was covered by forests during theearly and mid Holocene (Dupont and Weinelt1996) and that the opening-up of the landscapeoccurred rather abruptly between 4000 and 3500 BPfollowing a climatic change well-documented forthe neighbouring region of Ghana at lakeBosumtwi (Maley 1991 1997 1999) Moreover itappears that the break in the forest bloc wasoriginally much wider than it is now perhapsextending across almost all of south-west Nigeriaup to the Niger river Indeed from the beginning ofthe last millennium to the sixteenth century ADsome data show that this vast sector was notforested but becoming so (Barber 1985) and thatthe vegetation was dominated by oil palmappearing similar to currently existing vegetationacross southern Benin and Togo Data indicatingthis exist for an area near to the ancient city ofBenin (Okomu Forest about 100 km west of theNiger river) suggesting also that the forestextension occurred only in recent centuries (Jones1956 White and Oates 1999) Further researchwill be necessary to clarify these various points andtheir precise chronology

20

Figure 5 An lsquooil palm beltrsquo near the edge of therain forest in west Cameroon

Belt with a high density of Elaeis guineensis(oil palm) reaching 20 to 25 m height inassociation with a mature forest of the semi-deciduous type

In the Grass Fields extensive savannasnorth of the forest pioneer forest islands withlarge concentrations of oil palms In italics thenames of some important groups of people(Bamileke Bamoun Tikar Wuli)Source adapted from Letouzey 1978 and 1985

32 OkoumeacuteOkoumeacute (Aucoumea Klaineana Burseraceae) iscurrently restricted to the west and centre ofGabon with a limited extension in the south-westof Congo and another towards the north inEquatorial Guinea (Brunck et al 1990) (Figure 1)Like oil palm Okoumeacute is a light demandingpioneer species (which for some ecological reasonappears unable to live outside of central AfricaBrunk et al 1990 Maley 1990 Nasi 1997) In itsarea this tree is perfectly adapted to colonisepioneer forest fronts as they progress into savannasand often dominates them (Nasi 1997 White et al2000) It is abundant in fallows and along foresttracks As pioneer fronts mature and age othertrees of shade-tolerant species appear (egCaesalpiniaceae) and the Okoumeacute which surviveare the dispersed individuals whose crowns havealready reached the canopy living to 100ndash150years and achieving 12 m diameters (Nasi 1997)It can be deduced that the large Okoumeacute currentlypresent in the forest interior in its natural rangewere born in contact with savannas

The endemism of Okoumeacute with its range limitedmainly to Gabon is linked to its history and itsecology It is intolerant of water deficit and thus ofthe elevated temperatures of the long dry seasons(about 3 months from December to February) oftropical climates Within its Gabonese range waterdeficit is less severe than in neighbouring southernCameroon because the equatorial climatersquos dryseason (mostly from June to August) retains theelevated air humidity and moderate temperatures(Brunck et al 1990 Maley 1990)

Several works bring important precision to thehistory of Okoumeacute First palynological dataobtained in south Cameroon at lake Ossa near toEdea (Reynaud-Farrera et al 1996) show thatbetween 7500 and 3000 BP (middle Holocene) thedistribution of Okoumeacute extended much furthernorth reaching the lake about 170 km north of thetreersquos present range and perhaps further east andfurther south in western Congo Second moleculargenetics show two completely distinct varieties ofOkoumeacute one to the north of its range and anotherto the south (Muloko et al 1998 2000) Theboundary between the ranges runs approximatelyeast-west a little to the south of the equatortowards 0ucirc40rsquoSouth

It can be deduced that the current range of Okoumeacutehas resulted from the fragmentation of a largeancient range dating back to the middle Holocenefragmented by the generalised forest disturbanceabout 2500 BP into two very dispersed sub-groupsone to the south and the other to the north Theextension of open landscapes during the period ofdestruction would have been favourable toOkoumeacutersquos new phase of development in a similarway to that evidenced at Barombi Mbo for severalother pioneer trees (Figure 3b) The general forestrecolonisation dating from about 2000 BP was inan extremely fragmented environment withnumerous ecotones favourable to the developmentof Okoumeacute It is thus likely that the current rangeof Okoumeacute is about 2000 years old at most

A dynamic model which nearly follows thischronological canvas has been established by White(1995) This is based on examination of the currentvegetation of the northern part of the La Lopeacutereserve which is situated at the heart of the Gabonrsquosforest and still holds large patches of residualsavanna (see Aubreacuteville 1967) White (1995) showshow the forest reconstituted itself progressivelythanks to the wide bands of pioneer forest rich inOkoumeacute which surround these savannas Theprogression of pioneer forest fronts can be blockedby savanna fires often of anthropic origin (Oslislyand White 2000) Even so over the long term fromabout 2000 years ago the continuous general trendhas been in favour of reforestation

That the current and expanding range of Okoumeacutehas not extended much into south Cameroon ashad been the case in the middle Holocene isprobably explained by a shift in climatic conditionsbetween this older period and the recent (post-2000 BP) Holocene This deduction is importantas given the ecological particularities of Okoumeacutewhich link it tightly to the equatorial climate onecan further deduce that equatorial climates musthave reigned over south Cameroon during themiddle Holocene The present tropical climatethere thus dates from only about 2000 BP All thisappears to translate for the middle Holocene intoan increase in lsquosouthernrsquo equatorial influencestowards the north associated with a reduction inlsquonorthernrsquo boreal influences and possibly linked toa reduction in the length of the dry seasonassociated with the boreal winter

21

4 The question of savanna firesand the extension of the forestdisturbance in eastern Congo

Observations and opinions have frequentlydiverged concerning the influence of savanna fireson contemporary forest regeneration Observersvariously report that fire can either preventregeneration as in the area of savanna inliers inLopeacute Gabon (Oslisly and White 2000) or merelydelay it as in other parts of the forest region ThusLetouzey (1968) notes for the forest-savannatransition zone of south Cameroon that lsquofire doesnot necessarily prevent the formation of forestrecoveryrsquo Similar conclusions have been drawn inCentral Africa by Sillans and even in westernCongo (ex-Zaire) for the south of the forest bloc(Letouzey 1968) On the same lines Youta Happiand Bonvallot (1996) report lsquothat on theperiphery of a town of 76000 people such asBertoua (east Cameroon) many small savannainlets into forest of two to ten hectares havebecome covered by forest regrowth despite theannual bush fires practiced by Bororopastoralistsrsquo A little further south in the OdzalaReserve situated at the northern extremity of theBateke savannas (north Congo) Dowsett-Lemairereports that lsquoWithout doubt one is in a phase offorest extension despite very frequent bush firesrsquo(1996) This author notes lsquothe fire-break role ofthe boundary thanks to groves dominated byGaertnera paniculata a species absent at Lopeacutewhere colonising species are apparently moresensitive to firersquo Thus the floristic composition ofthe boundary plays an important role in thisphenomenon This was shown in south-eastCameroon by the mid-twentieth century invasionof Chromolaena odorata which colonised theboundaries and which acts there at least as anefficacious fire break It also protects seeds ofwoody species which can germinate and developunder its cover (Youta Happi et al 2000) Otherobservations show how traditional agriculturalpractices and the movement of cattle can befavourable to forest recolonisation in particularby eliminating the grasses which facilitate thepropagation and intensity of fires These processeshave been well described in southern Cameroon(Letouzey 1968 1985) and in Cocircte drsquoIvoire(Spichiger and Blanc-Pamard 1973 Blanc-Pamardand Peltre 1984 Gautier 1990)

These observations indicate that in general andespecially over the long term forest colonisation atthe forest-savanna contact has been the broadlydominant process In certain exceptional yearswhen the dry season has been one or two monthslonger the savanna fires have penetrated moredeeply into the neighbouring forests In early 1983during a two-month longer dry season numerousobservers reported the strong invasion of fire intoforest in Cocircte drsquoIvoire (Bertault 1990) Ghana(Hawthorne 1991 Swaine 1992) and Cameroon(Amougou 1986) Equally fires in Indonesia thatyear were linked to an exceptional El Nintildeo event(Goldammer and Seibert 1990)

These observations indicate how the penetration offire into forest occurs only in very dry years Onemight thus think that fires could have played amajor role in the forest destruction 2500 BP Thishas indeed been shown in eastern BrazilianAmazonia during its major disturbance in themiddle Holocene That fires were frequent then isshown by the numerous charcoal fragmentsdeposited in the lake sediments at Carajas (Martinet al 1993 Servant et al 1993) Yet the lateHolocene lake deposits in Central Africa hardlycontain any charcoal fragments especially for theperiod 3000ndash2000 BP (Giresse et al 1994 Maleyand Brenac 1998a Bertaux et al 2000) It is thesame for the upper horizons of the numerous soilprofiles observed in this region (eg Valleacuterie 1973)In the thalwegs the coarse deposits in the lowerpart of the Lower Terrace do at times containcharcoal or fragments of fossil wood (Maley andBrenac 1998b) however Iron age archaeologicalsites dating from the beginning of the lateHolocene such as in the savanna areas of La Lopeacutealso present wood charcoal (Oslisly and White2000) One can conclude provisionally that thefires that existed were of limited extent

In particular charcoal associated with ironsmelting furnaces cannot have caused the forestdisturbance and cannot have been responsible forthe savanna extension about 2500 BP Goucher(1981) had advanced the hypothesis that ironsmelting could have been a major cause ofdeforestation in West Africa but Fairhead andLeach (1998) have demonstrated that thishypothesis had no serious foundation For CentralAfrica Pinccedilon (1990) also concludes that the

22

extraction of wood for iron metallurgy on theBateke Plateaux was minimal and did not explain(as had been suggested) the presence of theextensive savannas there Instead these savannasresult from an elevated edaphic drought caused byvery sandy soils in which rain infiltrates rapidly toa great depth (Laraque and Pandi 1996)

In contrast to observations from Atlantic CentralAfrica (south Cameroon Gabon and Congo) woodcharcoal has been found in abundance in Ituri nearto Epulu (c 1ucirc20rsquoNndash28ucirc35rsquoE) Here in the heart ofthe forests of the north-east Congo basin about 180km from the nearest savannas 416 soil profileshave been examined Almost all of them containnumerous wood charcoal in their upper levelsbetween the surface and 50 cm deep (Hart et al1996) Virtually all of the 1817 samples of woodcharcoal identified by Deschamps belong to treesfound in the regionrsquos diverse forest formationsexcept for a few pieces of charcoal from twoprofiles which belong to typical wooded savannaspecies The 28 dates obtained for the charcoalshow that fires occurred almost exclusively duringthe last three millennia with a phase of woodedsavanna expansion dating to about 2200 BP (Hartet al 1996) Moreover about 300 km south ofEpulu and 100 km from the forestsavanna limit aroad cutting near Osokari (1ucirc16rsquoSndash27ucirc48rsquoE)revealed a remarkable profile above the principalstone line that could be dated to about 11500years BP (Runge 1996 Maley 1996) A secondgravel level obliquely cutting the top of the soilprofile contains wood charcoal dating from about2200 to 1850 BP These results show that the forestdisturbance in Atlantic Central Africa extended toIturi and affected a large part of the eastern regionof the Congo river basin In contrast with regionsnearer the Atlantic fire had a large role

Also at Ituri (Epulu) Hart et al (1996) show thatcertain trees which are now found there are notfound in the wood charcoal in particularGilbertiodendron dewevrei (Caesalpiniaceae) which isnow abundant (almost mono-dominant) in certainparts of the forest The range of forests dominatedby Gilbertiodendron dewevrei extends to the east andnorth of the Congo basin as far as east Cameroon(Leacuteonard 1953) It is thus possible that thisCaesalpiniaceae has very recently colonised the Ituriforests and that the large area it now occupies is a

legacy of the disturbance about 2500 BP Ifconfirmed this area could correspond at leastpartly to an older forest area (of unknowncharacter) which was destroyed by the disturbanceA comparison with Okoumeacute can be suggestedwhich as indicated for Gabon above alsodominates areas which it has colonised since lessthan 2000 years ago However as the forest agesOkoumeacute loses its dominant character whereas inIturi Gilbertiodendron dewevrei remains dominantprobably because it can regenerate abundantly inthe shade of parent trees and also because itsreproductive character gives it an advantage overother trees in this formation (see Hart 2001)

5 Conclusion the major role ofclimatic conditions and theirhistoryThe present mosaic pattern of numerous forestscharacterised by a mix or juxtaposition ofgroupings of evergreen and deciduous species isprobably the consequence of long-term disturbanceswhich have affected the forest domain during thelast three millennia particularly the majordisturbance which culminated about 2500 years BPThe synchronism apparent in the disturbance fordifferent sites studied across Central Africa (southCameroon Gabon Congo as far east as the CongoRiver basin) and its association with a generalisederosive phase enables one to conclude that it wasthe result of a major climatic change

The climate at this time appears to have beenrelatively arid as it led to forest destruction and inplaces the extension of savannas Neverthelessother characteristics such as the strongsynchronous extension of diverse pioneer taxaindicate that in places there was not truly areduction in rainfall but rather a change in itsannual distribution The absence of a fall in annualrainfall has been demonstrated at lake BarombiMbo (Maley and Brenac 1998a) and at lake Ossawhere diatom studies show that there was even anet rise in this lake between 2500 and 2200 yearsBP with only a short fall between 2200 and 2000years BP (Nguetsop et al 2000) It is thereforepreferable to designate this catastrophic period asan lsquounfavourable climatic disturbancersquo caused by anincrease in seasonality and a lengthening of the dryseason (Maley 1997) The strong soil erosion of this

23

period suggests heavy rains but concentrated oversix or seven months of the year as is now found inthe peripheral savanna zones The growth inseasonality could have been associated with adominance of lsquosquall linesrsquo the typical cloudformation of savanna zones formed by northsouthalignments of cumuliform (convective storm-type)clouds They surplanted other cloud typesprincipally the stratiform types which giverelatively fine regular monsoon-type rains (seeMaley 1982) That these monsoon rains wouldhave dominated the earlier period (between 4000and 30002800 BP Maley 1997) is shown by therelative development of Caesalpinaceae in theforests of low altitude and in the mountains by theextension of Podocarpus which are trees typical ofcloud forests with stratiform clouds (Kerfoot1968 Maley 1996 1997 Maley and Brenac1998a)

Research in dynamic tropical climatology over thelast 20 years shows that Sea Surface Temperature(SST) on the regional and global scale stronglyinfluences climate (Fontaine and Bigot 1993

Moron et al 1995 Bigot et al 1997) It is possibleto characterise the relations between the monsoonrains and the SST more precisely Studies ofclimatic anomalies over tropical Africa occurringsince the beginning of the 1960s have evidencedtwo dominant modes of rainfall distribution eachassociated with a particular distribution of SST Thefirst mode is characterised on the one hand bywarmer than average temperatures in the Gulf ofGuinea and the southern Atlantic and on the otherhand by cooler temperatures on the north tropicalAtlantic offshore of West Africa The second modepresents an opposed distribution for the Africancontinent and the ocean A north-south tropicalAtlantic lsquodipolersquo has thus been found whichoscillates between one mode and the other(Fontaine and Bigot 1993 Wotling et al 1995Bigot et al 1997) An important research result hasbeen to show that over recent millennia the spatialdistribution of paleo-climatic anomalies overtropical Africa (more humid regions and more dryones Figure 6) and of SST on the neighbouringAtlantic is very similar to the spatial distributionobserved for the two dominant anomalies over thelast four decades (Maley 1997 Maley et al 2000)One can deduce that the climatic mechanisms arethe same at the timescale of years centuries ormillennia

So the variation of SST over the tropical Atlanticappears to play a major role in the variation ofclimate and rain types over Central Africa Thephase of forest disturbance that culminated about2500 years BP was associated with relatively lsquowarmrsquoSST over the Gulf of Guinea The earlier phasebetween about 40003800 and 2800 years BP wasassociated with relatively lsquocoolrsquo SST sharplyreduced from the early and middle Holocene (seeMorley and Dworetzky 1993 Figure 7) Over westand south Cameroon these SST allowed relativelyhigh rains and the evolution of stratiform cloudstowards rainy nimbostratus types The palaeo-vegetation data are presently lacking for GabonHowever new research is underway within thePALEOFORGA (Paleoenvironments of GabonForests) programme studying lakes MaridorNguegravene and Kamaleacuteteacute (Figure 1) In contrastfurther south in western Congo the periodstarting from 5000 years BP (Bertaux et al 2000)was marked by the development of semi-deciduousforests (Vincens et al 2000) linked to lower

24

Figure 6 Major rainfall trends in tropical Africa andSouth America for two opposed climatic periods c 3700ndash3000 BP and c 28002500ndash2000 BPPrecipitation trends (anomalies) abovenormalmean + below normalmean ndash P pointsto the sectors where a maximum of Podocarpuspollen intervenedSource Maley 1997 Maley et al 2000

rainfalls indicating that the stratiform clouds hadprobably evolved towards non-rainy clouds Thefour-month dry season here is characterised by thequasi-permanence of these cloud types (Saint Vil1984) which probably gained place progressivelybetween 5000 and 4000 years BP There wastherefore a reinforcing of southern influencesupporting conclusions concerning the history ofOkoumeacute It is notable that the vegetational impactsof the lsquowarmrsquo SST c 28002500 to 2000 years BPwere very different from those earlier c 9000 to65006000 years BP a difference which reflects abrutal strengthening over Central Africa of borealinfluences (Nguetsop et al 2000) to the detrimentof southern influences from about 28002500years BP

Before the major Holocene forest disturbanceculminating about 2500 BP other importantdisturbances occurred during the Quaternary(Maley 1996) The best documented is that betweenabout 20 000 and 15 000 years BP (Maley 19871996 Maley and Brenac 1998a Figure 2) In bothcases residual forest environments (refugia)

persisted apparently in similar locations butperhaps less as small blocks of relativelyhomogeneous forest than as a landscape of forest-savanna mosaic in which forest islands dominated(see Leal 2000) Yet two important characteristicsclearly differentiate these two phases of forestfragmentation First the older disturbance wasmuch longer Second they occurred under verydifferent (even opposed) general climatic conditionsThe disturbance between 20 000 and 15 000 BPcoincided with the development of glacialconditions in middle and high latitudes and coolerconditions by several degrees also affected the wholeregion of the Gulf of Guinea (Maley 1996 Maley andBrenac 1998a) The most recent disturbance bycontrast occurred during the Holocene interglacialwhich on a global scale is characterised by reducedglacial extension and relatively warmertemperatures The period between 2500 and 2000BP is associated with slightly warmer climaticconditions on the regional and global scale (Maley1997 Maley et al 2000) An important conclusion isthat the retreat and fragmentation of the Africanforests can be produced under very different

25

Figure 7 Variations in Sea Surface Temperature (SST) in the South Atlantic Ocean Benguela sector(22ucirc20rsquoSndash11ucirc12rsquoE) estimated with transfer functions based on the abundance of Radiolarian species in thecore RC13ndash228The open circles designate the 6000 and 9000 BP interpolated dates for the two curves one correspondingto the southern summer (February) and the other to the southern winter (August) modern SST areindicated by the solid squares above each curve The dashed lines represent some interpolated dates(italics) 18 000 BP based on the oxygen-isotope stratigraphy 38004000 BP 3000 BP and 2000 BP Theselate Holocene dates are confirmed by an independent study of shells from the same sector (Cohen et al1992) The coherence of the main SST anomalies throughout the Guinea Gulf permits the use of thesecurves in order to interpret the paleoclimatology of central Africa (Maley 1997)Source Morley and Dworetzky 1993

climates either relatively cool or relatively warmand thus with very different climatic situations(Maley 1996 1997)

Climatic models concerning lsquoglobal warmingrsquo inthe context of contemporary climate changeconcerns indicate that an average growth intemperature of about 4ucircC will lead also to anincrease in evaporation of about 30 per cent butonly 12 per cent more rain for tropical Africa (Rind

Notes The author is grateful to James Fairhead for his

translation of this article from the original Frenchand to Melissa Leach and James Fairhead for theirinvitation to the IDS workshop lsquoChangingPerspectives on Forestsrsquo in March 2001 This isInstitut des Sciences de lrsquoEvolution de Montpellier(ISEMCNRS) contribution No 125ndash2001

26

ReferencesAchoundong G Youta Happi J Guillet B

Bonvallot J and Kamgang Beyala V 2000lsquoFormation et eacutevolution des recrus sur savanes (ausud Cameroun)rsquo in M Servant and S Servant-Vildary (eds) Dynamique agrave long terme desEcosystegravemes forestiers intertropicaux MeacutemoireUNESCO Paris UNESCO 31ndash41

Alexandre P 1965 lsquoProto-histoire du groupe Beacuteti-Boulou-Fang essai de synthegravese provisoirersquo CahiersdrsquoEtudes Africaines Vol 20 503ndash60

Amougou Akoa 1986 lsquoEtude botanique eteacutecologique de la valleacutee inondable du Haut-Nyong(sud Cameroun) et de ses affluentsrsquo Thegravese SciencesUniv Yaoundeacute

Aubreacuteville A 1937 lsquoLes forecircts du Dahomey et duTogorsquo Bulletin du Comiteacute drsquoEtudes Historiques etScientifiques de lrsquoAfrique Occidentale Franccedilaises Vol20 112

Aubreacuteville A 1967 lsquoLes eacutetranges mosaiques forecirct-savane du sommet de la boucle de lrsquoOgooueacute auGabonrsquo Adansonia Vol 7 13ndash22

Baeke V 1996 lsquoLe temps des Rites Lrsquounivers magico-religieux des Wuli (Mfumte du Camerounoccidental)rsquo Thegravese Sc Sociale Univ Libre Bruxelles

Barber RJ 1985 lsquoLand snails and past environmentat the Igbo-Iwoto Esie site southwestern NigeriarsquoWest African Journal of Archaeology Vol 15 89ndash102

Barbier JC 1981 lsquoLe peuplement de la partiemeacuteridionale du plateau Bamileacutekeacutersquo in C Tardits

(ed) Contribution de la recherche ethnologique agravelrsquohistoire des civilisations du Cameroun ParisColloque International Centre National deRecherche Scientifique Paris Vol 2 331ndash53

Bertault JG 1990 lsquoComparaison drsquoeacutecosystegravemesforestiers naturels et modifieacutes apregraves incendie enCocircte drsquoIvoirersquo in H Puig (ed) Atelier surlrsquoAmeacutenagement et la Conservation de lrsquoEcosystegravemeForestier Tropical Humide Cayenne ParisUNESCO 6

Bertaux J Schwartz D Vincens A Sifeddine AElenga H Mansour M Mariotti A Fournier MMartin L Wirrmann D and Servant M 2000lsquoEnregistrement de la phase segraveche drsquoAfriqueCentrale vers 3000 ans BP par la spectromeacutetrie IRdans les lacs Sinnda et Kitina (sud-Congo)rsquo in MServant S Servant-Vildary (eds) Dynamique agrave longterme des Ecosystegravemes forestiers intertropicauxMeacutemoire UNESCO Paris UNESCO 43ndash9

Bigot S Camberlin P Moron V and Richard Y1997 lsquoStructures spatiales de la variabiliteacute despreacutecipitations en Afrique une transition climatiqueagrave la fin des anneacutees 1960rsquo Compte Rendu delrsquoAcademie des Sciences Paris 2a 324 181ndash88

Blanc-Pamard C and Peltre P 1984 lsquoDynamique despaysages preacuteforestiers et pratiques culturales enAfrique de lrsquoOuest (Cocircte drsquoIvoire centrale)rsquo LeDeacuteveloppement rural en question Paris MeacutemoireORSTOM No 106 55ndash74

Bonnefille R Chalieacute F Guiot J and Vincens A1992 lsquoQuantitative estimates of full glacial

1995) Though the causes of warming experiencedabout 2500 years BP were natural (Magny 1993)and therefore different from the warmingenvisaged for the twenty-first century it is possiblethat the catastrophic destruction of Central Africanforests which culminated about 2500 years BPcould be an lsquoanaloguersquo and an alarm signal forpotential developments during this current andfuture phase of human-induced global warming(Maley 1997 Maley et al 2000)

1 BP lsquoBefore the Presentrsquo according to standardisedC14 analysis

2 Only 4ndash5 samples have been studied for each 1000years during the Holocene To detect century-scaleperiodicity it would be necessary to study at least 10samples per millennium

temperatures in equatorial Africa frompalynological datarsquo Climate Dynamics Vol 6251ndash57

Brunck F Grison F and Maitre HF 1990Lrsquookoumeacute Aucoumea klaineana Pierre MonographieCentre Techn Forestier Tropical Nogent CIRAD 102

Cohen AI Perkington JE Brundrirt GB and vanDer Merwe NJ 1992 lsquoA Holocene marine climaterecord in Mollusc shells from the southwest Africancoastrsquo Quaternary Research Vol 38 379ndash85

Dechamps R Senut B and Pickford M 1992 Fruitsfossiles pliocegravenes et pleacuteistocegravenes du Rift occidentalougandais Signification paleacuteoenvironnementaleCompte Rendu de lrsquoAcadeacutemie des Sciences Parisseacuterie 2 Vol 314 325ndash331

Dowsett-Lemaire F 1996 lsquoComposition et eacutevolutionde la veacutegeacutetation forestiegravere au Parc NationaldrsquoOdzala Congorsquo Bulletin du Jardin BotaniqueNational de Belgique Vol 65 253ndash92

Dupont LM and Weinelt M 1996 lsquoVegetationhistory of the savanna corridor between theguinean and the Congolian rain forest during thelast 150000 yearsrsquo Vegetation History andArchaeobotany Vol 5 273ndash92

Elenga H Schwartz D and Vincens A 1994lsquoPollen evidence of late Quaternary vegetation andinferred climate changes in Congorsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 109345ndash56

Elenga H Schwartz D Vincens A Bertaux J DeNamur C Martin L Wirrmann D and ServantM 1996 lsquoDiagramme pollinique holocegravene du lacKitina (Congo) mise en eacutevidence de changementspaleacuteobotaniques et paleacuteoclimatiques dans le massifforestier du Mayombersquo Compte Rendu de lrsquoAcadeacutemiedes Sciences Paris Vol 323 No 2a 403ndash10

Elenga H Peyron O Bonnefille R Prentice ICJolly D Cheddadi R Guiot J Andrieu V DeBeaulieu JL Hamilton AC Maley J MarchantR Reille M Scott L Taylor D and Van CampoE 2000 lsquoPollen-based biome reconstruction forsouthern Europe and Africa 18000 years agorsquoJournal of Biogeography Vol 27 621ndash34

Fairhead J and Leach M 1995 lsquoFalse forest historycomplicit social analysis rethinking some westAfrican environmental narrativesrsquo WorldDevelopment Vol 23 1023ndash35

Fairhead J and Leach M 1996 lsquoEnriching thelandscape social history and the management oftransition ecology in the forest-savanna mosaic ofthe Republic of Guinearsquo Africa Vol 66 14ndash36

Fairhead J and Leach M 1998 lsquoReframingdeforestation Global analyses and local realitiesstudies in West Africarsquo Global EnvironmentalChanges Series London Routledge 238

Fay JM 1997 lsquoEvidence of forest degradation innorthern Congo and southwestern Central AfricanRepublic between 2340 and 990 BPrsquo in The ecologysocial organization populations habitat and historyof the western lowland Gorilla PhD ThesisWashington Univ Saint Louis Missouri Ch 6

Fontaine B and Bigot S 1993 lsquoWest African rainfalldeficit and sea surface temperaturesrsquo InternationalJournal of Climatology Vol 13 271ndash85

Gautier L 1990 lsquoContact forecirct-savane en CocirctedrsquoIvoire centrale eacutevolution du recouvrementligneux des savanes de la Reacuteserve de Lamto (sud duV-Baouleacute)rsquo Candollea Vol 45 627ndash41

Giresse P Maley J and Brenac P 1994 lsquoLateQuaternary palaeoenvironments in the lake BarombiMbo (Cameroon) deduced from pollen and carbonisotopes of organic matterrsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 107 65ndash78

Goldammer JG and Seibert B 1990 lsquoThe impact ofdroughts and forest fires on tropical lowland rainforest of East Kalimatanrsquo in JG Goldammer (ed)Fire in the Tropical Biota Ecological Studies BerlinSpringer-Verlag 84

Goucher CL 1981 lsquoIron is iron til it is rust tradeand ecology in the decline of West African iron-smeltingrsquo Journal of African History Vol 22 179ndash89

Guille-Escuret G 1990 lsquoPalmier agrave huile vin depalme et transformations sociales en Lobaye (ForecirctCentrafricaine)rsquo Information sur les Sciences SocialesVol 29 327ndash53

Hart TB 2001 lsquoForest dynamics in the Ituri basin(DR Congo) Dominance diversity andconservationrsquo in W Weber L White A Vedderand L Naughton-Treves (eds) African rain forestecology and conservation An interdisciplinaryperspective Yale Yale University Press 154ndash64

Hart TB Hart JA Dechamps R Fournier M andAtaholo M 1996 lsquoChanges in forest compositionover the last 4000 years in the Ituri basin Zairersquo inLJG Van der Maesen and XM Van der Burgt (eds)The Biodiversity of African plants Dordrecht Kluwer545ndash63

Hawthorne WD 1991 Fire damage and forestregeneration in Ghana Ghana Forestry Departmentamp ODA Report

Jolly D Harrison SP Damnati B and BonnefilleR 1998 lsquoSimulated climate and Biomes of Africa

27

14

Figure 1 Schematic map of present vegetation in Central Africa (Maley 1990)(1) Biafran and Gabonese evergreen forests with numerous Cesalpiniaceae (2) Atlantic coastal forests withSacoglottis gabonensis and Lophira alata in Cameroon with the addition of Aucoumea klaineana (Okoumeacute) inGabon (see 11) The boundary between 1 and 2 is irregular and often progressive (3) Forests of Congolesetype with mixture of evergreen and semi-deciduous formations (4) Open canopy forests with a dense groundcover dominated by large Monocotyledons belonging to Marantaceae and Zingiberaceae (5) (a) Mixture oftypes 4 and 5b (b) Evergreen forests with a high canopy dominance of Gilbertiodendron dewevrei(Caesalpiniaceae) (6) Zone flooded almost all the year with evergreen formations (eg 5b) raphiales and otherhygrophylous formations (7) Semi-deciduous forests (8) Mangroves (9) Various mountain formations (10) Savannas (11) Boundary of the area with Okoumeacute (from Brunck et al 1990 and unpublished data) (12) Current mean maximum extension of lsquocoolingrsquo influences (according to Saint-Vil 1984) (13) Boundariesbetween States

minimum for all these variations in tree taxaoccurred between 2500 and 2000 BP during themajor phase of destruction The inverse ndash peaks ofmaximum extension ndash are not synchronous betweentaxa with variations corresponding to long phases ofsylvigenesis and the progressive or abruptreplacement of taxonomic groups by others Cyclesof shorter duration (for example 1000 500 or 100years) have not been shown to date as the scale ofsampling is insufficiently fine2 The pseudo-periodsof 2000 to 2500 years can be related to climaticcycles of the same duration frequently observedacross the globe in particular during the Holoceneand to one of the dominant periodicities of solaractivity about 2300 years (Magny 1993 Maley andBrenac 1998a)

21 The phase of forest destructionculminating about 2500 years BPThis phase of forest destruction (Figure 4) wasgeneral and synchronous as can be deduced fromboth its abrupt occurrence between 3000 and2500 years BP and its repeated discovery at everysite studied in detail in the northern part of theforest domain in west and south Cameroon atBarombi Mbo (Maley and Brenac 1998a)Mboandong (Richards 1986) Njupi (Zogning etal 1997) and Ossa (Reynaud-Farrera et al 1996van Geel et al 1998) in the southern part of theforest domain and in the Mayombe of westernCongo at Kitina (Elenga et al 1996) Kakamoeacuteka(Maley and Giresse 1998) Songolo near to Pointe-Noire (Vincens et al 1999) and at Ngamakala on

15

Figure 2 Site of Lake Barombi Mbo altitude of ca 300 m in the forests of west Cameroon Syntheticpollen diagram (per cent) with an interpolated chronology based on 12 radiocarbon datesBlack dots represent total percentages of tree pollen open area total percentages of pollen fromherbaceous plants mainly Gramineae and light dots the pollen percentages of Cyperaceae hygrophilousplants linked to the lowering of the lake levelSource Maley and Brenac 1998a

16

Figure 3a Site of Lake Barombi Mbo west Cameroon percentages of diverse tree pollenTypical tree taxa from mature forestsSource Maley and Brenac 1998a

the Bateacutekeacute Plateau (Eastern Congo) (Elenga et al1994) Where dating can properly frame the eventit dates more precisely to between 2800 and 2500years BP

During this phase primary forest trees almostcompletely disappeared and in certain sites (inwest Cameroon at Barombi Mbo Mboandong andNjupi in south Congo and on the Bateacutekeacute Plateau)there was a major expansion of savannas Thisphase of forest destruction must have been verybrief in lacustrine strata where the abrupt andalmost complete disappearance of primary typetrees is observed an lsquoexplosionrsquo of pioneer woodyvegetation has been noted for some sites such as atlake Barombi Mbo in west Cameroon (Maley andBrenac 1998a) (Figure 3b) at lake Kitina in theMayombe western Congo (Elenga et al 1996) andat lake Ossa south Cameroon (Reynaud-Farrera etal 1996) The rapid expansion of pioneer treesmust have been the first phase of lsquoscabbingrsquo whichinitiated the reconstitution of the canopy The mostfrequent pollen taxa are the pioneers Alchorneacordifolia Musanga Trema Macaranga Elaeisguineensis etc Although brief destruction wasintense and several authors argue that the openingof the forest facilitated the major Bantu migrationfrom areas presently north of the forest to thosenow south of it (Schwartz 1992 Lavachery et al1996)

22 The reconstitution of the CentralAfrican forests during the last twomillennia

Humid conditions favourable for forest returnedearlier in west Cameroon in particular at BarombiMbo dating from 2000 years BP Yet thereconstitution especially of lsquoprimaryrsquo type forestswas not synchronous and probably reflects thedistribution of the residual forests (Figure 4) Thedelay can be understood in terms of lsquohysteresisrsquo(Maley and Brenac 1998a)

The reinvasion of savanna by forest continuedduring the twentieth century as observed overmuch of the forest-savanna transition region Thisis evidenced by the comparison of air and satellitephotographs of different dates from the 1950s tothe present (see Blanc-Pamard and Peltre 1984Maley 1990 1996 Fairhead and Leach 1995

17

Figure 3b Site of Lake Barombi Mbo westCameroon percentages of diverse tree pollenTypical tree taxa from pioneer forestsSource Maley and Brenac 1998a

18

Figure 4 Schematic drawing of the status of Central African vegetation at the phase of maximumperturbation and destruction of forest which culminated around 2500 BPLarge hatched areas represent the residual areas of the forest domain around 2500 BP which probablyconsisted of forest-savanna mosaic with predominantly larger forest islands Around these residual areasthe land would have been colonised by open vegetation above all savannas but also open canopy forestformations (see (4) in Figure 1) The base of this map presents the current status of Central Africanbiotopes as in Figure 1

1998 Servant 1996) For south Cameroon alone(south of the Adamaoua Plateau) forest gainedabout a million hectares during this period(Letouzey 1985 Youta Happi and Bonvallot 1996Achoundong et al 2000) Historical and palyno-logical research in Cocircte drsquoIvoire Nigeria Cameroonand Congo indicate that this reinvasion hascontinued over several centuries (Fairhead andLeach 1998 Maley 1999 Vincens et al 2000)Albeit with fluctuations and interludes it appearsto be the continuation of the forest reinvasionwhich began about 2000 years BP Currentreinvasion is thus the long-term result of themassive forest disturbance which happened around2500 years BP

A brief review of research findings concerning oilpalm and Okoumeacute can illustrate the history ofthese forest ecosystems over recent millennia

3 The history of oil palm and ofOkoumeacute in Central Africa duringthe Holocene impact of the forestdisturbance culminating about2500 years BP

31 The oil palm

Oil palm (Elaeis guineensis) has abundant andcharacteristic pollen and thus appears frequentlyin sedimented pollen profiles This palm is ofAfrican origin with its pollen and macro-remains(nuts) found in Equatorial African deposits datingback to the Tertiary (Zeven 1967 Dechamps et al1992 Maley 1996 1999 Maley and Brenac 1998aMaley and Chepstow-Lusty 2001)

The extent to which oil palm has been planted orits regeneration encouraged through land use hasbeen debated Certainly when people clear youngpioneer forests or fallows near to villages where oilpalm is naturally abundant they generallyconserve the palms effectively enriching themthrough fallow cycles (eg in Cameroon CentralAfrican Republic Cocircte drsquoIvoire Togo and Beninsee Maley 1999) Thus in 1937 the renownedbotanist Aubreacuteville characterised the lsquoimmense oilpalm grovesrsquo of southern Benin as lsquoa naturalformation simply improved by the inhabitantsover the centuriesrsquo Zevenrsquos research largely fromNigeria also concludes that with the exception of

modern industrial plantations most oil palms inAfrica are semi-wild and are not propagated bysowing or transplanting seedlings (see Zeven1972)

Ecologically oil palm is a pioneer that requires lightto complete the first stages of its growth so itdevelops naturally in windfall-clearings andespecially on the periphery of dense forest after thepassage of fire which it can tolerate (Swaine andHall 1986 Swaine 1992) The botanist Letouzey(1978 1985) has described a vast natural oil palmgrove in west Cameroon and near the north-west ofthe forest zone (Figure 5) A band of forest 10ndash20km wide dominated by large and numerous oilpalms extends over more than 150 km Thisfollows the boundary between forest and savannaat times from 5 to 30 km inside the forest From theabsence of trees classically found in anthropicplantations and other criteria Letouzey (19781985) concluded that this grove is a natural standA similar example has been described on theeastern flank of Mount Nimba (GuineaLiberianborder) by Schnell (1946) who considered that thedissemination of palm nuts was facilitated by bothtoucans and chimpanzees whose faeces oftencontain it abundantly These natural groves haveattracted migrant people (eg Guille-Escuret 1990Maley 1999) For example many Bamileke peoplemoved to the natural groves of western Cameroonin the mid nineteenth century (Barbier 1981Warnier 1985 Perrois and Notue 1997 Maley1999) In the vast lsquoGrassfieldsrsquo just north of westCameroonrsquos forest bloc the forest islands are still ina phase of expansion especially due to the oil palmwhich is one of the principal pioneer trees Close tothe Donga river the Wuli people colonised thesenatural stands Baeke (1996) explains how theWuli install their villages in the palm groves whereassorted social rules oppose all plantation of palmsMoreover she reports that a Wuli myth of originlsquoprovides evidence for the prior use of oil palms tothe working of land and clearly distinguishes twotypes of plant exploitation gathering andagriculturersquo (Baeke 1996)

As already noted during the major disturbanceabout 2500 BP oil palm was a principal pioneerthat subsequently developed strongly (Figure 3b)In west Cameroon the rapid forest recolonisationwhich occurred from 2000 years ago in the

19

vicinity of Barombi Mbo (Maley and Brenac1998a) was accompanied by a retreat of oil palmprobably because ecological and climatic factorsrendered other pioneers more active at this time InFigure 3 it is notable that in west Cameroonbetween 2000 and 1000 years BP the pioneer treeMilicia was in a phase of very strong expansionThen here and elsewhere in central Africa a secondphase of oil palm expansion began around 1400years BP This culminated towards 1000 BP at lakeOssa near to Edea (Reynaud-Farrera et al 1996) inthe south of the Central African Republic atNouabale-Ndoki (Fay 1997 Maley 1999) but also

further south in Gabon to La Lopeacute (White et al2000 Maley unpublished) and in western Congonear to lake Kitina (Elenga et al 1996) This phaseending about 700800 BP must have beenassociated with renewed forest disturbance butweaker than that 2500 BP This disturbance wasalso associated with a short phase of soil erosionand with a discontinuity in the deposits observedin assorted locations in Central Africa (Maley andBrenac 1998b) An increase in mineral fluxes hasbeen found between 1200ndash800 years BP for thesites of Kitina and Sinnda in western Congo(Bertaux et al 2000) From the thirteenth centurya phase of forest recovery has occurred in CentralAfrica with a new development of mature forestsand a generalised retreat of oil palm a recoverydespite increased human settlements (Alexandre1965 Vansina 1990)

At the western extension of the Central Africanforests lies a large area without forest which bio-geographers call the Dahomey Gap stretchingacross southern Togo and Benin Much of this iscovered by an immense palm forest of naturalorigin as it was described by Aubreacuteville (1937)and Mondjannagni (1969) Some pollen data showthat this area was covered by forests during theearly and mid Holocene (Dupont and Weinelt1996) and that the opening-up of the landscapeoccurred rather abruptly between 4000 and 3500 BPfollowing a climatic change well-documented forthe neighbouring region of Ghana at lakeBosumtwi (Maley 1991 1997 1999) Moreover itappears that the break in the forest bloc wasoriginally much wider than it is now perhapsextending across almost all of south-west Nigeriaup to the Niger river Indeed from the beginning ofthe last millennium to the sixteenth century ADsome data show that this vast sector was notforested but becoming so (Barber 1985) and thatthe vegetation was dominated by oil palmappearing similar to currently existing vegetationacross southern Benin and Togo Data indicatingthis exist for an area near to the ancient city ofBenin (Okomu Forest about 100 km west of theNiger river) suggesting also that the forestextension occurred only in recent centuries (Jones1956 White and Oates 1999) Further researchwill be necessary to clarify these various points andtheir precise chronology

20

Figure 5 An lsquooil palm beltrsquo near the edge of therain forest in west Cameroon

Belt with a high density of Elaeis guineensis(oil palm) reaching 20 to 25 m height inassociation with a mature forest of the semi-deciduous type

In the Grass Fields extensive savannasnorth of the forest pioneer forest islands withlarge concentrations of oil palms In italics thenames of some important groups of people(Bamileke Bamoun Tikar Wuli)Source adapted from Letouzey 1978 and 1985

32 OkoumeacuteOkoumeacute (Aucoumea Klaineana Burseraceae) iscurrently restricted to the west and centre ofGabon with a limited extension in the south-westof Congo and another towards the north inEquatorial Guinea (Brunck et al 1990) (Figure 1)Like oil palm Okoumeacute is a light demandingpioneer species (which for some ecological reasonappears unable to live outside of central AfricaBrunk et al 1990 Maley 1990 Nasi 1997) In itsarea this tree is perfectly adapted to colonisepioneer forest fronts as they progress into savannasand often dominates them (Nasi 1997 White et al2000) It is abundant in fallows and along foresttracks As pioneer fronts mature and age othertrees of shade-tolerant species appear (egCaesalpiniaceae) and the Okoumeacute which surviveare the dispersed individuals whose crowns havealready reached the canopy living to 100ndash150years and achieving 12 m diameters (Nasi 1997)It can be deduced that the large Okoumeacute currentlypresent in the forest interior in its natural rangewere born in contact with savannas

The endemism of Okoumeacute with its range limitedmainly to Gabon is linked to its history and itsecology It is intolerant of water deficit and thus ofthe elevated temperatures of the long dry seasons(about 3 months from December to February) oftropical climates Within its Gabonese range waterdeficit is less severe than in neighbouring southernCameroon because the equatorial climatersquos dryseason (mostly from June to August) retains theelevated air humidity and moderate temperatures(Brunck et al 1990 Maley 1990)

Several works bring important precision to thehistory of Okoumeacute First palynological dataobtained in south Cameroon at lake Ossa near toEdea (Reynaud-Farrera et al 1996) show thatbetween 7500 and 3000 BP (middle Holocene) thedistribution of Okoumeacute extended much furthernorth reaching the lake about 170 km north of thetreersquos present range and perhaps further east andfurther south in western Congo Second moleculargenetics show two completely distinct varieties ofOkoumeacute one to the north of its range and anotherto the south (Muloko et al 1998 2000) Theboundary between the ranges runs approximatelyeast-west a little to the south of the equatortowards 0ucirc40rsquoSouth

It can be deduced that the current range of Okoumeacutehas resulted from the fragmentation of a largeancient range dating back to the middle Holocenefragmented by the generalised forest disturbanceabout 2500 BP into two very dispersed sub-groupsone to the south and the other to the north Theextension of open landscapes during the period ofdestruction would have been favourable toOkoumeacutersquos new phase of development in a similarway to that evidenced at Barombi Mbo for severalother pioneer trees (Figure 3b) The general forestrecolonisation dating from about 2000 BP was inan extremely fragmented environment withnumerous ecotones favourable to the developmentof Okoumeacute It is thus likely that the current rangeof Okoumeacute is about 2000 years old at most

A dynamic model which nearly follows thischronological canvas has been established by White(1995) This is based on examination of the currentvegetation of the northern part of the La Lopeacutereserve which is situated at the heart of the Gabonrsquosforest and still holds large patches of residualsavanna (see Aubreacuteville 1967) White (1995) showshow the forest reconstituted itself progressivelythanks to the wide bands of pioneer forest rich inOkoumeacute which surround these savannas Theprogression of pioneer forest fronts can be blockedby savanna fires often of anthropic origin (Oslislyand White 2000) Even so over the long term fromabout 2000 years ago the continuous general trendhas been in favour of reforestation

That the current and expanding range of Okoumeacutehas not extended much into south Cameroon ashad been the case in the middle Holocene isprobably explained by a shift in climatic conditionsbetween this older period and the recent (post-2000 BP) Holocene This deduction is importantas given the ecological particularities of Okoumeacutewhich link it tightly to the equatorial climate onecan further deduce that equatorial climates musthave reigned over south Cameroon during themiddle Holocene The present tropical climatethere thus dates from only about 2000 BP All thisappears to translate for the middle Holocene intoan increase in lsquosouthernrsquo equatorial influencestowards the north associated with a reduction inlsquonorthernrsquo boreal influences and possibly linked toa reduction in the length of the dry seasonassociated with the boreal winter

21

4 The question of savanna firesand the extension of the forestdisturbance in eastern Congo

Observations and opinions have frequentlydiverged concerning the influence of savanna fireson contemporary forest regeneration Observersvariously report that fire can either preventregeneration as in the area of savanna inliers inLopeacute Gabon (Oslisly and White 2000) or merelydelay it as in other parts of the forest region ThusLetouzey (1968) notes for the forest-savannatransition zone of south Cameroon that lsquofire doesnot necessarily prevent the formation of forestrecoveryrsquo Similar conclusions have been drawn inCentral Africa by Sillans and even in westernCongo (ex-Zaire) for the south of the forest bloc(Letouzey 1968) On the same lines Youta Happiand Bonvallot (1996) report lsquothat on theperiphery of a town of 76000 people such asBertoua (east Cameroon) many small savannainlets into forest of two to ten hectares havebecome covered by forest regrowth despite theannual bush fires practiced by Bororopastoralistsrsquo A little further south in the OdzalaReserve situated at the northern extremity of theBateke savannas (north Congo) Dowsett-Lemairereports that lsquoWithout doubt one is in a phase offorest extension despite very frequent bush firesrsquo(1996) This author notes lsquothe fire-break role ofthe boundary thanks to groves dominated byGaertnera paniculata a species absent at Lopeacutewhere colonising species are apparently moresensitive to firersquo Thus the floristic composition ofthe boundary plays an important role in thisphenomenon This was shown in south-eastCameroon by the mid-twentieth century invasionof Chromolaena odorata which colonised theboundaries and which acts there at least as anefficacious fire break It also protects seeds ofwoody species which can germinate and developunder its cover (Youta Happi et al 2000) Otherobservations show how traditional agriculturalpractices and the movement of cattle can befavourable to forest recolonisation in particularby eliminating the grasses which facilitate thepropagation and intensity of fires These processeshave been well described in southern Cameroon(Letouzey 1968 1985) and in Cocircte drsquoIvoire(Spichiger and Blanc-Pamard 1973 Blanc-Pamardand Peltre 1984 Gautier 1990)

These observations indicate that in general andespecially over the long term forest colonisation atthe forest-savanna contact has been the broadlydominant process In certain exceptional yearswhen the dry season has been one or two monthslonger the savanna fires have penetrated moredeeply into the neighbouring forests In early 1983during a two-month longer dry season numerousobservers reported the strong invasion of fire intoforest in Cocircte drsquoIvoire (Bertault 1990) Ghana(Hawthorne 1991 Swaine 1992) and Cameroon(Amougou 1986) Equally fires in Indonesia thatyear were linked to an exceptional El Nintildeo event(Goldammer and Seibert 1990)

These observations indicate how the penetration offire into forest occurs only in very dry years Onemight thus think that fires could have played amajor role in the forest destruction 2500 BP Thishas indeed been shown in eastern BrazilianAmazonia during its major disturbance in themiddle Holocene That fires were frequent then isshown by the numerous charcoal fragmentsdeposited in the lake sediments at Carajas (Martinet al 1993 Servant et al 1993) Yet the lateHolocene lake deposits in Central Africa hardlycontain any charcoal fragments especially for theperiod 3000ndash2000 BP (Giresse et al 1994 Maleyand Brenac 1998a Bertaux et al 2000) It is thesame for the upper horizons of the numerous soilprofiles observed in this region (eg Valleacuterie 1973)In the thalwegs the coarse deposits in the lowerpart of the Lower Terrace do at times containcharcoal or fragments of fossil wood (Maley andBrenac 1998b) however Iron age archaeologicalsites dating from the beginning of the lateHolocene such as in the savanna areas of La Lopeacutealso present wood charcoal (Oslisly and White2000) One can conclude provisionally that thefires that existed were of limited extent

In particular charcoal associated with ironsmelting furnaces cannot have caused the forestdisturbance and cannot have been responsible forthe savanna extension about 2500 BP Goucher(1981) had advanced the hypothesis that ironsmelting could have been a major cause ofdeforestation in West Africa but Fairhead andLeach (1998) have demonstrated that thishypothesis had no serious foundation For CentralAfrica Pinccedilon (1990) also concludes that the

22

extraction of wood for iron metallurgy on theBateke Plateaux was minimal and did not explain(as had been suggested) the presence of theextensive savannas there Instead these savannasresult from an elevated edaphic drought caused byvery sandy soils in which rain infiltrates rapidly toa great depth (Laraque and Pandi 1996)

In contrast to observations from Atlantic CentralAfrica (south Cameroon Gabon and Congo) woodcharcoal has been found in abundance in Ituri nearto Epulu (c 1ucirc20rsquoNndash28ucirc35rsquoE) Here in the heart ofthe forests of the north-east Congo basin about 180km from the nearest savannas 416 soil profileshave been examined Almost all of them containnumerous wood charcoal in their upper levelsbetween the surface and 50 cm deep (Hart et al1996) Virtually all of the 1817 samples of woodcharcoal identified by Deschamps belong to treesfound in the regionrsquos diverse forest formationsexcept for a few pieces of charcoal from twoprofiles which belong to typical wooded savannaspecies The 28 dates obtained for the charcoalshow that fires occurred almost exclusively duringthe last three millennia with a phase of woodedsavanna expansion dating to about 2200 BP (Hartet al 1996) Moreover about 300 km south ofEpulu and 100 km from the forestsavanna limit aroad cutting near Osokari (1ucirc16rsquoSndash27ucirc48rsquoE)revealed a remarkable profile above the principalstone line that could be dated to about 11500years BP (Runge 1996 Maley 1996) A secondgravel level obliquely cutting the top of the soilprofile contains wood charcoal dating from about2200 to 1850 BP These results show that the forestdisturbance in Atlantic Central Africa extended toIturi and affected a large part of the eastern regionof the Congo river basin In contrast with regionsnearer the Atlantic fire had a large role

Also at Ituri (Epulu) Hart et al (1996) show thatcertain trees which are now found there are notfound in the wood charcoal in particularGilbertiodendron dewevrei (Caesalpiniaceae) which isnow abundant (almost mono-dominant) in certainparts of the forest The range of forests dominatedby Gilbertiodendron dewevrei extends to the east andnorth of the Congo basin as far as east Cameroon(Leacuteonard 1953) It is thus possible that thisCaesalpiniaceae has very recently colonised the Ituriforests and that the large area it now occupies is a

legacy of the disturbance about 2500 BP Ifconfirmed this area could correspond at leastpartly to an older forest area (of unknowncharacter) which was destroyed by the disturbanceA comparison with Okoumeacute can be suggestedwhich as indicated for Gabon above alsodominates areas which it has colonised since lessthan 2000 years ago However as the forest agesOkoumeacute loses its dominant character whereas inIturi Gilbertiodendron dewevrei remains dominantprobably because it can regenerate abundantly inthe shade of parent trees and also because itsreproductive character gives it an advantage overother trees in this formation (see Hart 2001)

5 Conclusion the major role ofclimatic conditions and theirhistoryThe present mosaic pattern of numerous forestscharacterised by a mix or juxtaposition ofgroupings of evergreen and deciduous species isprobably the consequence of long-term disturbanceswhich have affected the forest domain during thelast three millennia particularly the majordisturbance which culminated about 2500 years BPThe synchronism apparent in the disturbance fordifferent sites studied across Central Africa (southCameroon Gabon Congo as far east as the CongoRiver basin) and its association with a generalisederosive phase enables one to conclude that it wasthe result of a major climatic change

The climate at this time appears to have beenrelatively arid as it led to forest destruction and inplaces the extension of savannas Neverthelessother characteristics such as the strongsynchronous extension of diverse pioneer taxaindicate that in places there was not truly areduction in rainfall but rather a change in itsannual distribution The absence of a fall in annualrainfall has been demonstrated at lake BarombiMbo (Maley and Brenac 1998a) and at lake Ossawhere diatom studies show that there was even anet rise in this lake between 2500 and 2200 yearsBP with only a short fall between 2200 and 2000years BP (Nguetsop et al 2000) It is thereforepreferable to designate this catastrophic period asan lsquounfavourable climatic disturbancersquo caused by anincrease in seasonality and a lengthening of the dryseason (Maley 1997) The strong soil erosion of this

23

period suggests heavy rains but concentrated oversix or seven months of the year as is now found inthe peripheral savanna zones The growth inseasonality could have been associated with adominance of lsquosquall linesrsquo the typical cloudformation of savanna zones formed by northsouthalignments of cumuliform (convective storm-type)clouds They surplanted other cloud typesprincipally the stratiform types which giverelatively fine regular monsoon-type rains (seeMaley 1982) That these monsoon rains wouldhave dominated the earlier period (between 4000and 30002800 BP Maley 1997) is shown by therelative development of Caesalpinaceae in theforests of low altitude and in the mountains by theextension of Podocarpus which are trees typical ofcloud forests with stratiform clouds (Kerfoot1968 Maley 1996 1997 Maley and Brenac1998a)

Research in dynamic tropical climatology over thelast 20 years shows that Sea Surface Temperature(SST) on the regional and global scale stronglyinfluences climate (Fontaine and Bigot 1993

Moron et al 1995 Bigot et al 1997) It is possibleto characterise the relations between the monsoonrains and the SST more precisely Studies ofclimatic anomalies over tropical Africa occurringsince the beginning of the 1960s have evidencedtwo dominant modes of rainfall distribution eachassociated with a particular distribution of SST Thefirst mode is characterised on the one hand bywarmer than average temperatures in the Gulf ofGuinea and the southern Atlantic and on the otherhand by cooler temperatures on the north tropicalAtlantic offshore of West Africa The second modepresents an opposed distribution for the Africancontinent and the ocean A north-south tropicalAtlantic lsquodipolersquo has thus been found whichoscillates between one mode and the other(Fontaine and Bigot 1993 Wotling et al 1995Bigot et al 1997) An important research result hasbeen to show that over recent millennia the spatialdistribution of paleo-climatic anomalies overtropical Africa (more humid regions and more dryones Figure 6) and of SST on the neighbouringAtlantic is very similar to the spatial distributionobserved for the two dominant anomalies over thelast four decades (Maley 1997 Maley et al 2000)One can deduce that the climatic mechanisms arethe same at the timescale of years centuries ormillennia

So the variation of SST over the tropical Atlanticappears to play a major role in the variation ofclimate and rain types over Central Africa Thephase of forest disturbance that culminated about2500 years BP was associated with relatively lsquowarmrsquoSST over the Gulf of Guinea The earlier phasebetween about 40003800 and 2800 years BP wasassociated with relatively lsquocoolrsquo SST sharplyreduced from the early and middle Holocene (seeMorley and Dworetzky 1993 Figure 7) Over westand south Cameroon these SST allowed relativelyhigh rains and the evolution of stratiform cloudstowards rainy nimbostratus types The palaeo-vegetation data are presently lacking for GabonHowever new research is underway within thePALEOFORGA (Paleoenvironments of GabonForests) programme studying lakes MaridorNguegravene and Kamaleacuteteacute (Figure 1) In contrastfurther south in western Congo the periodstarting from 5000 years BP (Bertaux et al 2000)was marked by the development of semi-deciduousforests (Vincens et al 2000) linked to lower

24

Figure 6 Major rainfall trends in tropical Africa andSouth America for two opposed climatic periods c 3700ndash3000 BP and c 28002500ndash2000 BPPrecipitation trends (anomalies) abovenormalmean + below normalmean ndash P pointsto the sectors where a maximum of Podocarpuspollen intervenedSource Maley 1997 Maley et al 2000

rainfalls indicating that the stratiform clouds hadprobably evolved towards non-rainy clouds Thefour-month dry season here is characterised by thequasi-permanence of these cloud types (Saint Vil1984) which probably gained place progressivelybetween 5000 and 4000 years BP There wastherefore a reinforcing of southern influencesupporting conclusions concerning the history ofOkoumeacute It is notable that the vegetational impactsof the lsquowarmrsquo SST c 28002500 to 2000 years BPwere very different from those earlier c 9000 to65006000 years BP a difference which reflects abrutal strengthening over Central Africa of borealinfluences (Nguetsop et al 2000) to the detrimentof southern influences from about 28002500years BP

Before the major Holocene forest disturbanceculminating about 2500 BP other importantdisturbances occurred during the Quaternary(Maley 1996) The best documented is that betweenabout 20 000 and 15 000 years BP (Maley 19871996 Maley and Brenac 1998a Figure 2) In bothcases residual forest environments (refugia)

persisted apparently in similar locations butperhaps less as small blocks of relativelyhomogeneous forest than as a landscape of forest-savanna mosaic in which forest islands dominated(see Leal 2000) Yet two important characteristicsclearly differentiate these two phases of forestfragmentation First the older disturbance wasmuch longer Second they occurred under verydifferent (even opposed) general climatic conditionsThe disturbance between 20 000 and 15 000 BPcoincided with the development of glacialconditions in middle and high latitudes and coolerconditions by several degrees also affected the wholeregion of the Gulf of Guinea (Maley 1996 Maley andBrenac 1998a) The most recent disturbance bycontrast occurred during the Holocene interglacialwhich on a global scale is characterised by reducedglacial extension and relatively warmertemperatures The period between 2500 and 2000BP is associated with slightly warmer climaticconditions on the regional and global scale (Maley1997 Maley et al 2000) An important conclusion isthat the retreat and fragmentation of the Africanforests can be produced under very different

25

Figure 7 Variations in Sea Surface Temperature (SST) in the South Atlantic Ocean Benguela sector(22ucirc20rsquoSndash11ucirc12rsquoE) estimated with transfer functions based on the abundance of Radiolarian species in thecore RC13ndash228The open circles designate the 6000 and 9000 BP interpolated dates for the two curves one correspondingto the southern summer (February) and the other to the southern winter (August) modern SST areindicated by the solid squares above each curve The dashed lines represent some interpolated dates(italics) 18 000 BP based on the oxygen-isotope stratigraphy 38004000 BP 3000 BP and 2000 BP Theselate Holocene dates are confirmed by an independent study of shells from the same sector (Cohen et al1992) The coherence of the main SST anomalies throughout the Guinea Gulf permits the use of thesecurves in order to interpret the paleoclimatology of central Africa (Maley 1997)Source Morley and Dworetzky 1993

climates either relatively cool or relatively warmand thus with very different climatic situations(Maley 1996 1997)

Climatic models concerning lsquoglobal warmingrsquo inthe context of contemporary climate changeconcerns indicate that an average growth intemperature of about 4ucircC will lead also to anincrease in evaporation of about 30 per cent butonly 12 per cent more rain for tropical Africa (Rind

Notes The author is grateful to James Fairhead for his

translation of this article from the original Frenchand to Melissa Leach and James Fairhead for theirinvitation to the IDS workshop lsquoChangingPerspectives on Forestsrsquo in March 2001 This isInstitut des Sciences de lrsquoEvolution de Montpellier(ISEMCNRS) contribution No 125ndash2001

26

ReferencesAchoundong G Youta Happi J Guillet B

Bonvallot J and Kamgang Beyala V 2000lsquoFormation et eacutevolution des recrus sur savanes (ausud Cameroun)rsquo in M Servant and S Servant-Vildary (eds) Dynamique agrave long terme desEcosystegravemes forestiers intertropicaux MeacutemoireUNESCO Paris UNESCO 31ndash41

Alexandre P 1965 lsquoProto-histoire du groupe Beacuteti-Boulou-Fang essai de synthegravese provisoirersquo CahiersdrsquoEtudes Africaines Vol 20 503ndash60

Amougou Akoa 1986 lsquoEtude botanique eteacutecologique de la valleacutee inondable du Haut-Nyong(sud Cameroun) et de ses affluentsrsquo Thegravese SciencesUniv Yaoundeacute

Aubreacuteville A 1937 lsquoLes forecircts du Dahomey et duTogorsquo Bulletin du Comiteacute drsquoEtudes Historiques etScientifiques de lrsquoAfrique Occidentale Franccedilaises Vol20 112

Aubreacuteville A 1967 lsquoLes eacutetranges mosaiques forecirct-savane du sommet de la boucle de lrsquoOgooueacute auGabonrsquo Adansonia Vol 7 13ndash22

Baeke V 1996 lsquoLe temps des Rites Lrsquounivers magico-religieux des Wuli (Mfumte du Camerounoccidental)rsquo Thegravese Sc Sociale Univ Libre Bruxelles

Barber RJ 1985 lsquoLand snails and past environmentat the Igbo-Iwoto Esie site southwestern NigeriarsquoWest African Journal of Archaeology Vol 15 89ndash102

Barbier JC 1981 lsquoLe peuplement de la partiemeacuteridionale du plateau Bamileacutekeacutersquo in C Tardits

(ed) Contribution de la recherche ethnologique agravelrsquohistoire des civilisations du Cameroun ParisColloque International Centre National deRecherche Scientifique Paris Vol 2 331ndash53

Bertault JG 1990 lsquoComparaison drsquoeacutecosystegravemesforestiers naturels et modifieacutes apregraves incendie enCocircte drsquoIvoirersquo in H Puig (ed) Atelier surlrsquoAmeacutenagement et la Conservation de lrsquoEcosystegravemeForestier Tropical Humide Cayenne ParisUNESCO 6

Bertaux J Schwartz D Vincens A Sifeddine AElenga H Mansour M Mariotti A Fournier MMartin L Wirrmann D and Servant M 2000lsquoEnregistrement de la phase segraveche drsquoAfriqueCentrale vers 3000 ans BP par la spectromeacutetrie IRdans les lacs Sinnda et Kitina (sud-Congo)rsquo in MServant S Servant-Vildary (eds) Dynamique agrave longterme des Ecosystegravemes forestiers intertropicauxMeacutemoire UNESCO Paris UNESCO 43ndash9

Bigot S Camberlin P Moron V and Richard Y1997 lsquoStructures spatiales de la variabiliteacute despreacutecipitations en Afrique une transition climatiqueagrave la fin des anneacutees 1960rsquo Compte Rendu delrsquoAcademie des Sciences Paris 2a 324 181ndash88

Blanc-Pamard C and Peltre P 1984 lsquoDynamique despaysages preacuteforestiers et pratiques culturales enAfrique de lrsquoOuest (Cocircte drsquoIvoire centrale)rsquo LeDeacuteveloppement rural en question Paris MeacutemoireORSTOM No 106 55ndash74

Bonnefille R Chalieacute F Guiot J and Vincens A1992 lsquoQuantitative estimates of full glacial

1995) Though the causes of warming experiencedabout 2500 years BP were natural (Magny 1993)and therefore different from the warmingenvisaged for the twenty-first century it is possiblethat the catastrophic destruction of Central Africanforests which culminated about 2500 years BPcould be an lsquoanaloguersquo and an alarm signal forpotential developments during this current andfuture phase of human-induced global warming(Maley 1997 Maley et al 2000)

1 BP lsquoBefore the Presentrsquo according to standardisedC14 analysis

2 Only 4ndash5 samples have been studied for each 1000years during the Holocene To detect century-scaleperiodicity it would be necessary to study at least 10samples per millennium

temperatures in equatorial Africa frompalynological datarsquo Climate Dynamics Vol 6251ndash57

Brunck F Grison F and Maitre HF 1990Lrsquookoumeacute Aucoumea klaineana Pierre MonographieCentre Techn Forestier Tropical Nogent CIRAD 102

Cohen AI Perkington JE Brundrirt GB and vanDer Merwe NJ 1992 lsquoA Holocene marine climaterecord in Mollusc shells from the southwest Africancoastrsquo Quaternary Research Vol 38 379ndash85

Dechamps R Senut B and Pickford M 1992 Fruitsfossiles pliocegravenes et pleacuteistocegravenes du Rift occidentalougandais Signification paleacuteoenvironnementaleCompte Rendu de lrsquoAcadeacutemie des Sciences Parisseacuterie 2 Vol 314 325ndash331

Dowsett-Lemaire F 1996 lsquoComposition et eacutevolutionde la veacutegeacutetation forestiegravere au Parc NationaldrsquoOdzala Congorsquo Bulletin du Jardin BotaniqueNational de Belgique Vol 65 253ndash92

Dupont LM and Weinelt M 1996 lsquoVegetationhistory of the savanna corridor between theguinean and the Congolian rain forest during thelast 150000 yearsrsquo Vegetation History andArchaeobotany Vol 5 273ndash92

Elenga H Schwartz D and Vincens A 1994lsquoPollen evidence of late Quaternary vegetation andinferred climate changes in Congorsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 109345ndash56

Elenga H Schwartz D Vincens A Bertaux J DeNamur C Martin L Wirrmann D and ServantM 1996 lsquoDiagramme pollinique holocegravene du lacKitina (Congo) mise en eacutevidence de changementspaleacuteobotaniques et paleacuteoclimatiques dans le massifforestier du Mayombersquo Compte Rendu de lrsquoAcadeacutemiedes Sciences Paris Vol 323 No 2a 403ndash10

Elenga H Peyron O Bonnefille R Prentice ICJolly D Cheddadi R Guiot J Andrieu V DeBeaulieu JL Hamilton AC Maley J MarchantR Reille M Scott L Taylor D and Van CampoE 2000 lsquoPollen-based biome reconstruction forsouthern Europe and Africa 18000 years agorsquoJournal of Biogeography Vol 27 621ndash34

Fairhead J and Leach M 1995 lsquoFalse forest historycomplicit social analysis rethinking some westAfrican environmental narrativesrsquo WorldDevelopment Vol 23 1023ndash35

Fairhead J and Leach M 1996 lsquoEnriching thelandscape social history and the management oftransition ecology in the forest-savanna mosaic ofthe Republic of Guinearsquo Africa Vol 66 14ndash36

Fairhead J and Leach M 1998 lsquoReframingdeforestation Global analyses and local realitiesstudies in West Africarsquo Global EnvironmentalChanges Series London Routledge 238

Fay JM 1997 lsquoEvidence of forest degradation innorthern Congo and southwestern Central AfricanRepublic between 2340 and 990 BPrsquo in The ecologysocial organization populations habitat and historyof the western lowland Gorilla PhD ThesisWashington Univ Saint Louis Missouri Ch 6

Fontaine B and Bigot S 1993 lsquoWest African rainfalldeficit and sea surface temperaturesrsquo InternationalJournal of Climatology Vol 13 271ndash85

Gautier L 1990 lsquoContact forecirct-savane en CocirctedrsquoIvoire centrale eacutevolution du recouvrementligneux des savanes de la Reacuteserve de Lamto (sud duV-Baouleacute)rsquo Candollea Vol 45 627ndash41

Giresse P Maley J and Brenac P 1994 lsquoLateQuaternary palaeoenvironments in the lake BarombiMbo (Cameroon) deduced from pollen and carbonisotopes of organic matterrsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 107 65ndash78

Goldammer JG and Seibert B 1990 lsquoThe impact ofdroughts and forest fires on tropical lowland rainforest of East Kalimatanrsquo in JG Goldammer (ed)Fire in the Tropical Biota Ecological Studies BerlinSpringer-Verlag 84

Goucher CL 1981 lsquoIron is iron til it is rust tradeand ecology in the decline of West African iron-smeltingrsquo Journal of African History Vol 22 179ndash89

Guille-Escuret G 1990 lsquoPalmier agrave huile vin depalme et transformations sociales en Lobaye (ForecirctCentrafricaine)rsquo Information sur les Sciences SocialesVol 29 327ndash53

Hart TB 2001 lsquoForest dynamics in the Ituri basin(DR Congo) Dominance diversity andconservationrsquo in W Weber L White A Vedderand L Naughton-Treves (eds) African rain forestecology and conservation An interdisciplinaryperspective Yale Yale University Press 154ndash64

Hart TB Hart JA Dechamps R Fournier M andAtaholo M 1996 lsquoChanges in forest compositionover the last 4000 years in the Ituri basin Zairersquo inLJG Van der Maesen and XM Van der Burgt (eds)The Biodiversity of African plants Dordrecht Kluwer545ndash63

Hawthorne WD 1991 Fire damage and forestregeneration in Ghana Ghana Forestry Departmentamp ODA Report

Jolly D Harrison SP Damnati B and BonnefilleR 1998 lsquoSimulated climate and Biomes of Africa

27

minimum for all these variations in tree taxaoccurred between 2500 and 2000 BP during themajor phase of destruction The inverse ndash peaks ofmaximum extension ndash are not synchronous betweentaxa with variations corresponding to long phases ofsylvigenesis and the progressive or abruptreplacement of taxonomic groups by others Cyclesof shorter duration (for example 1000 500 or 100years) have not been shown to date as the scale ofsampling is insufficiently fine2 The pseudo-periodsof 2000 to 2500 years can be related to climaticcycles of the same duration frequently observedacross the globe in particular during the Holoceneand to one of the dominant periodicities of solaractivity about 2300 years (Magny 1993 Maley andBrenac 1998a)

21 The phase of forest destructionculminating about 2500 years BPThis phase of forest destruction (Figure 4) wasgeneral and synchronous as can be deduced fromboth its abrupt occurrence between 3000 and2500 years BP and its repeated discovery at everysite studied in detail in the northern part of theforest domain in west and south Cameroon atBarombi Mbo (Maley and Brenac 1998a)Mboandong (Richards 1986) Njupi (Zogning etal 1997) and Ossa (Reynaud-Farrera et al 1996van Geel et al 1998) in the southern part of theforest domain and in the Mayombe of westernCongo at Kitina (Elenga et al 1996) Kakamoeacuteka(Maley and Giresse 1998) Songolo near to Pointe-Noire (Vincens et al 1999) and at Ngamakala on

15

Figure 2 Site of Lake Barombi Mbo altitude of ca 300 m in the forests of west Cameroon Syntheticpollen diagram (per cent) with an interpolated chronology based on 12 radiocarbon datesBlack dots represent total percentages of tree pollen open area total percentages of pollen fromherbaceous plants mainly Gramineae and light dots the pollen percentages of Cyperaceae hygrophilousplants linked to the lowering of the lake levelSource Maley and Brenac 1998a

16

Figure 3a Site of Lake Barombi Mbo west Cameroon percentages of diverse tree pollenTypical tree taxa from mature forestsSource Maley and Brenac 1998a

the Bateacutekeacute Plateau (Eastern Congo) (Elenga et al1994) Where dating can properly frame the eventit dates more precisely to between 2800 and 2500years BP

During this phase primary forest trees almostcompletely disappeared and in certain sites (inwest Cameroon at Barombi Mbo Mboandong andNjupi in south Congo and on the Bateacutekeacute Plateau)there was a major expansion of savannas Thisphase of forest destruction must have been verybrief in lacustrine strata where the abrupt andalmost complete disappearance of primary typetrees is observed an lsquoexplosionrsquo of pioneer woodyvegetation has been noted for some sites such as atlake Barombi Mbo in west Cameroon (Maley andBrenac 1998a) (Figure 3b) at lake Kitina in theMayombe western Congo (Elenga et al 1996) andat lake Ossa south Cameroon (Reynaud-Farrera etal 1996) The rapid expansion of pioneer treesmust have been the first phase of lsquoscabbingrsquo whichinitiated the reconstitution of the canopy The mostfrequent pollen taxa are the pioneers Alchorneacordifolia Musanga Trema Macaranga Elaeisguineensis etc Although brief destruction wasintense and several authors argue that the openingof the forest facilitated the major Bantu migrationfrom areas presently north of the forest to thosenow south of it (Schwartz 1992 Lavachery et al1996)

22 The reconstitution of the CentralAfrican forests during the last twomillennia

Humid conditions favourable for forest returnedearlier in west Cameroon in particular at BarombiMbo dating from 2000 years BP Yet thereconstitution especially of lsquoprimaryrsquo type forestswas not synchronous and probably reflects thedistribution of the residual forests (Figure 4) Thedelay can be understood in terms of lsquohysteresisrsquo(Maley and Brenac 1998a)

The reinvasion of savanna by forest continuedduring the twentieth century as observed overmuch of the forest-savanna transition region Thisis evidenced by the comparison of air and satellitephotographs of different dates from the 1950s tothe present (see Blanc-Pamard and Peltre 1984Maley 1990 1996 Fairhead and Leach 1995

17

Figure 3b Site of Lake Barombi Mbo westCameroon percentages of diverse tree pollenTypical tree taxa from pioneer forestsSource Maley and Brenac 1998a

18

Figure 4 Schematic drawing of the status of Central African vegetation at the phase of maximumperturbation and destruction of forest which culminated around 2500 BPLarge hatched areas represent the residual areas of the forest domain around 2500 BP which probablyconsisted of forest-savanna mosaic with predominantly larger forest islands Around these residual areasthe land would have been colonised by open vegetation above all savannas but also open canopy forestformations (see (4) in Figure 1) The base of this map presents the current status of Central Africanbiotopes as in Figure 1

1998 Servant 1996) For south Cameroon alone(south of the Adamaoua Plateau) forest gainedabout a million hectares during this period(Letouzey 1985 Youta Happi and Bonvallot 1996Achoundong et al 2000) Historical and palyno-logical research in Cocircte drsquoIvoire Nigeria Cameroonand Congo indicate that this reinvasion hascontinued over several centuries (Fairhead andLeach 1998 Maley 1999 Vincens et al 2000)Albeit with fluctuations and interludes it appearsto be the continuation of the forest reinvasionwhich began about 2000 years BP Currentreinvasion is thus the long-term result of themassive forest disturbance which happened around2500 years BP

A brief review of research findings concerning oilpalm and Okoumeacute can illustrate the history ofthese forest ecosystems over recent millennia

3 The history of oil palm and ofOkoumeacute in Central Africa duringthe Holocene impact of the forestdisturbance culminating about2500 years BP

31 The oil palm

Oil palm (Elaeis guineensis) has abundant andcharacteristic pollen and thus appears frequentlyin sedimented pollen profiles This palm is ofAfrican origin with its pollen and macro-remains(nuts) found in Equatorial African deposits datingback to the Tertiary (Zeven 1967 Dechamps et al1992 Maley 1996 1999 Maley and Brenac 1998aMaley and Chepstow-Lusty 2001)

The extent to which oil palm has been planted orits regeneration encouraged through land use hasbeen debated Certainly when people clear youngpioneer forests or fallows near to villages where oilpalm is naturally abundant they generallyconserve the palms effectively enriching themthrough fallow cycles (eg in Cameroon CentralAfrican Republic Cocircte drsquoIvoire Togo and Beninsee Maley 1999) Thus in 1937 the renownedbotanist Aubreacuteville characterised the lsquoimmense oilpalm grovesrsquo of southern Benin as lsquoa naturalformation simply improved by the inhabitantsover the centuriesrsquo Zevenrsquos research largely fromNigeria also concludes that with the exception of

modern industrial plantations most oil palms inAfrica are semi-wild and are not propagated bysowing or transplanting seedlings (see Zeven1972)

Ecologically oil palm is a pioneer that requires lightto complete the first stages of its growth so itdevelops naturally in windfall-clearings andespecially on the periphery of dense forest after thepassage of fire which it can tolerate (Swaine andHall 1986 Swaine 1992) The botanist Letouzey(1978 1985) has described a vast natural oil palmgrove in west Cameroon and near the north-west ofthe forest zone (Figure 5) A band of forest 10ndash20km wide dominated by large and numerous oilpalms extends over more than 150 km Thisfollows the boundary between forest and savannaat times from 5 to 30 km inside the forest From theabsence of trees classically found in anthropicplantations and other criteria Letouzey (19781985) concluded that this grove is a natural standA similar example has been described on theeastern flank of Mount Nimba (GuineaLiberianborder) by Schnell (1946) who considered that thedissemination of palm nuts was facilitated by bothtoucans and chimpanzees whose faeces oftencontain it abundantly These natural groves haveattracted migrant people (eg Guille-Escuret 1990Maley 1999) For example many Bamileke peoplemoved to the natural groves of western Cameroonin the mid nineteenth century (Barbier 1981Warnier 1985 Perrois and Notue 1997 Maley1999) In the vast lsquoGrassfieldsrsquo just north of westCameroonrsquos forest bloc the forest islands are still ina phase of expansion especially due to the oil palmwhich is one of the principal pioneer trees Close tothe Donga river the Wuli people colonised thesenatural stands Baeke (1996) explains how theWuli install their villages in the palm groves whereassorted social rules oppose all plantation of palmsMoreover she reports that a Wuli myth of originlsquoprovides evidence for the prior use of oil palms tothe working of land and clearly distinguishes twotypes of plant exploitation gathering andagriculturersquo (Baeke 1996)

As already noted during the major disturbanceabout 2500 BP oil palm was a principal pioneerthat subsequently developed strongly (Figure 3b)In west Cameroon the rapid forest recolonisationwhich occurred from 2000 years ago in the

19

vicinity of Barombi Mbo (Maley and Brenac1998a) was accompanied by a retreat of oil palmprobably because ecological and climatic factorsrendered other pioneers more active at this time InFigure 3 it is notable that in west Cameroonbetween 2000 and 1000 years BP the pioneer treeMilicia was in a phase of very strong expansionThen here and elsewhere in central Africa a secondphase of oil palm expansion began around 1400years BP This culminated towards 1000 BP at lakeOssa near to Edea (Reynaud-Farrera et al 1996) inthe south of the Central African Republic atNouabale-Ndoki (Fay 1997 Maley 1999) but also

further south in Gabon to La Lopeacute (White et al2000 Maley unpublished) and in western Congonear to lake Kitina (Elenga et al 1996) This phaseending about 700800 BP must have beenassociated with renewed forest disturbance butweaker than that 2500 BP This disturbance wasalso associated with a short phase of soil erosionand with a discontinuity in the deposits observedin assorted locations in Central Africa (Maley andBrenac 1998b) An increase in mineral fluxes hasbeen found between 1200ndash800 years BP for thesites of Kitina and Sinnda in western Congo(Bertaux et al 2000) From the thirteenth centurya phase of forest recovery has occurred in CentralAfrica with a new development of mature forestsand a generalised retreat of oil palm a recoverydespite increased human settlements (Alexandre1965 Vansina 1990)

At the western extension of the Central Africanforests lies a large area without forest which bio-geographers call the Dahomey Gap stretchingacross southern Togo and Benin Much of this iscovered by an immense palm forest of naturalorigin as it was described by Aubreacuteville (1937)and Mondjannagni (1969) Some pollen data showthat this area was covered by forests during theearly and mid Holocene (Dupont and Weinelt1996) and that the opening-up of the landscapeoccurred rather abruptly between 4000 and 3500 BPfollowing a climatic change well-documented forthe neighbouring region of Ghana at lakeBosumtwi (Maley 1991 1997 1999) Moreover itappears that the break in the forest bloc wasoriginally much wider than it is now perhapsextending across almost all of south-west Nigeriaup to the Niger river Indeed from the beginning ofthe last millennium to the sixteenth century ADsome data show that this vast sector was notforested but becoming so (Barber 1985) and thatthe vegetation was dominated by oil palmappearing similar to currently existing vegetationacross southern Benin and Togo Data indicatingthis exist for an area near to the ancient city ofBenin (Okomu Forest about 100 km west of theNiger river) suggesting also that the forestextension occurred only in recent centuries (Jones1956 White and Oates 1999) Further researchwill be necessary to clarify these various points andtheir precise chronology

20

Figure 5 An lsquooil palm beltrsquo near the edge of therain forest in west Cameroon

Belt with a high density of Elaeis guineensis(oil palm) reaching 20 to 25 m height inassociation with a mature forest of the semi-deciduous type

In the Grass Fields extensive savannasnorth of the forest pioneer forest islands withlarge concentrations of oil palms In italics thenames of some important groups of people(Bamileke Bamoun Tikar Wuli)Source adapted from Letouzey 1978 and 1985

32 OkoumeacuteOkoumeacute (Aucoumea Klaineana Burseraceae) iscurrently restricted to the west and centre ofGabon with a limited extension in the south-westof Congo and another towards the north inEquatorial Guinea (Brunck et al 1990) (Figure 1)Like oil palm Okoumeacute is a light demandingpioneer species (which for some ecological reasonappears unable to live outside of central AfricaBrunk et al 1990 Maley 1990 Nasi 1997) In itsarea this tree is perfectly adapted to colonisepioneer forest fronts as they progress into savannasand often dominates them (Nasi 1997 White et al2000) It is abundant in fallows and along foresttracks As pioneer fronts mature and age othertrees of shade-tolerant species appear (egCaesalpiniaceae) and the Okoumeacute which surviveare the dispersed individuals whose crowns havealready reached the canopy living to 100ndash150years and achieving 12 m diameters (Nasi 1997)It can be deduced that the large Okoumeacute currentlypresent in the forest interior in its natural rangewere born in contact with savannas

The endemism of Okoumeacute with its range limitedmainly to Gabon is linked to its history and itsecology It is intolerant of water deficit and thus ofthe elevated temperatures of the long dry seasons(about 3 months from December to February) oftropical climates Within its Gabonese range waterdeficit is less severe than in neighbouring southernCameroon because the equatorial climatersquos dryseason (mostly from June to August) retains theelevated air humidity and moderate temperatures(Brunck et al 1990 Maley 1990)

Several works bring important precision to thehistory of Okoumeacute First palynological dataobtained in south Cameroon at lake Ossa near toEdea (Reynaud-Farrera et al 1996) show thatbetween 7500 and 3000 BP (middle Holocene) thedistribution of Okoumeacute extended much furthernorth reaching the lake about 170 km north of thetreersquos present range and perhaps further east andfurther south in western Congo Second moleculargenetics show two completely distinct varieties ofOkoumeacute one to the north of its range and anotherto the south (Muloko et al 1998 2000) Theboundary between the ranges runs approximatelyeast-west a little to the south of the equatortowards 0ucirc40rsquoSouth

It can be deduced that the current range of Okoumeacutehas resulted from the fragmentation of a largeancient range dating back to the middle Holocenefragmented by the generalised forest disturbanceabout 2500 BP into two very dispersed sub-groupsone to the south and the other to the north Theextension of open landscapes during the period ofdestruction would have been favourable toOkoumeacutersquos new phase of development in a similarway to that evidenced at Barombi Mbo for severalother pioneer trees (Figure 3b) The general forestrecolonisation dating from about 2000 BP was inan extremely fragmented environment withnumerous ecotones favourable to the developmentof Okoumeacute It is thus likely that the current rangeof Okoumeacute is about 2000 years old at most

A dynamic model which nearly follows thischronological canvas has been established by White(1995) This is based on examination of the currentvegetation of the northern part of the La Lopeacutereserve which is situated at the heart of the Gabonrsquosforest and still holds large patches of residualsavanna (see Aubreacuteville 1967) White (1995) showshow the forest reconstituted itself progressivelythanks to the wide bands of pioneer forest rich inOkoumeacute which surround these savannas Theprogression of pioneer forest fronts can be blockedby savanna fires often of anthropic origin (Oslislyand White 2000) Even so over the long term fromabout 2000 years ago the continuous general trendhas been in favour of reforestation

That the current and expanding range of Okoumeacutehas not extended much into south Cameroon ashad been the case in the middle Holocene isprobably explained by a shift in climatic conditionsbetween this older period and the recent (post-2000 BP) Holocene This deduction is importantas given the ecological particularities of Okoumeacutewhich link it tightly to the equatorial climate onecan further deduce that equatorial climates musthave reigned over south Cameroon during themiddle Holocene The present tropical climatethere thus dates from only about 2000 BP All thisappears to translate for the middle Holocene intoan increase in lsquosouthernrsquo equatorial influencestowards the north associated with a reduction inlsquonorthernrsquo boreal influences and possibly linked toa reduction in the length of the dry seasonassociated with the boreal winter

21

4 The question of savanna firesand the extension of the forestdisturbance in eastern Congo

Observations and opinions have frequentlydiverged concerning the influence of savanna fireson contemporary forest regeneration Observersvariously report that fire can either preventregeneration as in the area of savanna inliers inLopeacute Gabon (Oslisly and White 2000) or merelydelay it as in other parts of the forest region ThusLetouzey (1968) notes for the forest-savannatransition zone of south Cameroon that lsquofire doesnot necessarily prevent the formation of forestrecoveryrsquo Similar conclusions have been drawn inCentral Africa by Sillans and even in westernCongo (ex-Zaire) for the south of the forest bloc(Letouzey 1968) On the same lines Youta Happiand Bonvallot (1996) report lsquothat on theperiphery of a town of 76000 people such asBertoua (east Cameroon) many small savannainlets into forest of two to ten hectares havebecome covered by forest regrowth despite theannual bush fires practiced by Bororopastoralistsrsquo A little further south in the OdzalaReserve situated at the northern extremity of theBateke savannas (north Congo) Dowsett-Lemairereports that lsquoWithout doubt one is in a phase offorest extension despite very frequent bush firesrsquo(1996) This author notes lsquothe fire-break role ofthe boundary thanks to groves dominated byGaertnera paniculata a species absent at Lopeacutewhere colonising species are apparently moresensitive to firersquo Thus the floristic composition ofthe boundary plays an important role in thisphenomenon This was shown in south-eastCameroon by the mid-twentieth century invasionof Chromolaena odorata which colonised theboundaries and which acts there at least as anefficacious fire break It also protects seeds ofwoody species which can germinate and developunder its cover (Youta Happi et al 2000) Otherobservations show how traditional agriculturalpractices and the movement of cattle can befavourable to forest recolonisation in particularby eliminating the grasses which facilitate thepropagation and intensity of fires These processeshave been well described in southern Cameroon(Letouzey 1968 1985) and in Cocircte drsquoIvoire(Spichiger and Blanc-Pamard 1973 Blanc-Pamardand Peltre 1984 Gautier 1990)

These observations indicate that in general andespecially over the long term forest colonisation atthe forest-savanna contact has been the broadlydominant process In certain exceptional yearswhen the dry season has been one or two monthslonger the savanna fires have penetrated moredeeply into the neighbouring forests In early 1983during a two-month longer dry season numerousobservers reported the strong invasion of fire intoforest in Cocircte drsquoIvoire (Bertault 1990) Ghana(Hawthorne 1991 Swaine 1992) and Cameroon(Amougou 1986) Equally fires in Indonesia thatyear were linked to an exceptional El Nintildeo event(Goldammer and Seibert 1990)

These observations indicate how the penetration offire into forest occurs only in very dry years Onemight thus think that fires could have played amajor role in the forest destruction 2500 BP Thishas indeed been shown in eastern BrazilianAmazonia during its major disturbance in themiddle Holocene That fires were frequent then isshown by the numerous charcoal fragmentsdeposited in the lake sediments at Carajas (Martinet al 1993 Servant et al 1993) Yet the lateHolocene lake deposits in Central Africa hardlycontain any charcoal fragments especially for theperiod 3000ndash2000 BP (Giresse et al 1994 Maleyand Brenac 1998a Bertaux et al 2000) It is thesame for the upper horizons of the numerous soilprofiles observed in this region (eg Valleacuterie 1973)In the thalwegs the coarse deposits in the lowerpart of the Lower Terrace do at times containcharcoal or fragments of fossil wood (Maley andBrenac 1998b) however Iron age archaeologicalsites dating from the beginning of the lateHolocene such as in the savanna areas of La Lopeacutealso present wood charcoal (Oslisly and White2000) One can conclude provisionally that thefires that existed were of limited extent

In particular charcoal associated with ironsmelting furnaces cannot have caused the forestdisturbance and cannot have been responsible forthe savanna extension about 2500 BP Goucher(1981) had advanced the hypothesis that ironsmelting could have been a major cause ofdeforestation in West Africa but Fairhead andLeach (1998) have demonstrated that thishypothesis had no serious foundation For CentralAfrica Pinccedilon (1990) also concludes that the

22

extraction of wood for iron metallurgy on theBateke Plateaux was minimal and did not explain(as had been suggested) the presence of theextensive savannas there Instead these savannasresult from an elevated edaphic drought caused byvery sandy soils in which rain infiltrates rapidly toa great depth (Laraque and Pandi 1996)

In contrast to observations from Atlantic CentralAfrica (south Cameroon Gabon and Congo) woodcharcoal has been found in abundance in Ituri nearto Epulu (c 1ucirc20rsquoNndash28ucirc35rsquoE) Here in the heart ofthe forests of the north-east Congo basin about 180km from the nearest savannas 416 soil profileshave been examined Almost all of them containnumerous wood charcoal in their upper levelsbetween the surface and 50 cm deep (Hart et al1996) Virtually all of the 1817 samples of woodcharcoal identified by Deschamps belong to treesfound in the regionrsquos diverse forest formationsexcept for a few pieces of charcoal from twoprofiles which belong to typical wooded savannaspecies The 28 dates obtained for the charcoalshow that fires occurred almost exclusively duringthe last three millennia with a phase of woodedsavanna expansion dating to about 2200 BP (Hartet al 1996) Moreover about 300 km south ofEpulu and 100 km from the forestsavanna limit aroad cutting near Osokari (1ucirc16rsquoSndash27ucirc48rsquoE)revealed a remarkable profile above the principalstone line that could be dated to about 11500years BP (Runge 1996 Maley 1996) A secondgravel level obliquely cutting the top of the soilprofile contains wood charcoal dating from about2200 to 1850 BP These results show that the forestdisturbance in Atlantic Central Africa extended toIturi and affected a large part of the eastern regionof the Congo river basin In contrast with regionsnearer the Atlantic fire had a large role

Also at Ituri (Epulu) Hart et al (1996) show thatcertain trees which are now found there are notfound in the wood charcoal in particularGilbertiodendron dewevrei (Caesalpiniaceae) which isnow abundant (almost mono-dominant) in certainparts of the forest The range of forests dominatedby Gilbertiodendron dewevrei extends to the east andnorth of the Congo basin as far as east Cameroon(Leacuteonard 1953) It is thus possible that thisCaesalpiniaceae has very recently colonised the Ituriforests and that the large area it now occupies is a

legacy of the disturbance about 2500 BP Ifconfirmed this area could correspond at leastpartly to an older forest area (of unknowncharacter) which was destroyed by the disturbanceA comparison with Okoumeacute can be suggestedwhich as indicated for Gabon above alsodominates areas which it has colonised since lessthan 2000 years ago However as the forest agesOkoumeacute loses its dominant character whereas inIturi Gilbertiodendron dewevrei remains dominantprobably because it can regenerate abundantly inthe shade of parent trees and also because itsreproductive character gives it an advantage overother trees in this formation (see Hart 2001)

5 Conclusion the major role ofclimatic conditions and theirhistoryThe present mosaic pattern of numerous forestscharacterised by a mix or juxtaposition ofgroupings of evergreen and deciduous species isprobably the consequence of long-term disturbanceswhich have affected the forest domain during thelast three millennia particularly the majordisturbance which culminated about 2500 years BPThe synchronism apparent in the disturbance fordifferent sites studied across Central Africa (southCameroon Gabon Congo as far east as the CongoRiver basin) and its association with a generalisederosive phase enables one to conclude that it wasthe result of a major climatic change

The climate at this time appears to have beenrelatively arid as it led to forest destruction and inplaces the extension of savannas Neverthelessother characteristics such as the strongsynchronous extension of diverse pioneer taxaindicate that in places there was not truly areduction in rainfall but rather a change in itsannual distribution The absence of a fall in annualrainfall has been demonstrated at lake BarombiMbo (Maley and Brenac 1998a) and at lake Ossawhere diatom studies show that there was even anet rise in this lake between 2500 and 2200 yearsBP with only a short fall between 2200 and 2000years BP (Nguetsop et al 2000) It is thereforepreferable to designate this catastrophic period asan lsquounfavourable climatic disturbancersquo caused by anincrease in seasonality and a lengthening of the dryseason (Maley 1997) The strong soil erosion of this

23

period suggests heavy rains but concentrated oversix or seven months of the year as is now found inthe peripheral savanna zones The growth inseasonality could have been associated with adominance of lsquosquall linesrsquo the typical cloudformation of savanna zones formed by northsouthalignments of cumuliform (convective storm-type)clouds They surplanted other cloud typesprincipally the stratiform types which giverelatively fine regular monsoon-type rains (seeMaley 1982) That these monsoon rains wouldhave dominated the earlier period (between 4000and 30002800 BP Maley 1997) is shown by therelative development of Caesalpinaceae in theforests of low altitude and in the mountains by theextension of Podocarpus which are trees typical ofcloud forests with stratiform clouds (Kerfoot1968 Maley 1996 1997 Maley and Brenac1998a)

Research in dynamic tropical climatology over thelast 20 years shows that Sea Surface Temperature(SST) on the regional and global scale stronglyinfluences climate (Fontaine and Bigot 1993

Moron et al 1995 Bigot et al 1997) It is possibleto characterise the relations between the monsoonrains and the SST more precisely Studies ofclimatic anomalies over tropical Africa occurringsince the beginning of the 1960s have evidencedtwo dominant modes of rainfall distribution eachassociated with a particular distribution of SST Thefirst mode is characterised on the one hand bywarmer than average temperatures in the Gulf ofGuinea and the southern Atlantic and on the otherhand by cooler temperatures on the north tropicalAtlantic offshore of West Africa The second modepresents an opposed distribution for the Africancontinent and the ocean A north-south tropicalAtlantic lsquodipolersquo has thus been found whichoscillates between one mode and the other(Fontaine and Bigot 1993 Wotling et al 1995Bigot et al 1997) An important research result hasbeen to show that over recent millennia the spatialdistribution of paleo-climatic anomalies overtropical Africa (more humid regions and more dryones Figure 6) and of SST on the neighbouringAtlantic is very similar to the spatial distributionobserved for the two dominant anomalies over thelast four decades (Maley 1997 Maley et al 2000)One can deduce that the climatic mechanisms arethe same at the timescale of years centuries ormillennia

So the variation of SST over the tropical Atlanticappears to play a major role in the variation ofclimate and rain types over Central Africa Thephase of forest disturbance that culminated about2500 years BP was associated with relatively lsquowarmrsquoSST over the Gulf of Guinea The earlier phasebetween about 40003800 and 2800 years BP wasassociated with relatively lsquocoolrsquo SST sharplyreduced from the early and middle Holocene (seeMorley and Dworetzky 1993 Figure 7) Over westand south Cameroon these SST allowed relativelyhigh rains and the evolution of stratiform cloudstowards rainy nimbostratus types The palaeo-vegetation data are presently lacking for GabonHowever new research is underway within thePALEOFORGA (Paleoenvironments of GabonForests) programme studying lakes MaridorNguegravene and Kamaleacuteteacute (Figure 1) In contrastfurther south in western Congo the periodstarting from 5000 years BP (Bertaux et al 2000)was marked by the development of semi-deciduousforests (Vincens et al 2000) linked to lower

24

Figure 6 Major rainfall trends in tropical Africa andSouth America for two opposed climatic periods c 3700ndash3000 BP and c 28002500ndash2000 BPPrecipitation trends (anomalies) abovenormalmean + below normalmean ndash P pointsto the sectors where a maximum of Podocarpuspollen intervenedSource Maley 1997 Maley et al 2000

rainfalls indicating that the stratiform clouds hadprobably evolved towards non-rainy clouds Thefour-month dry season here is characterised by thequasi-permanence of these cloud types (Saint Vil1984) which probably gained place progressivelybetween 5000 and 4000 years BP There wastherefore a reinforcing of southern influencesupporting conclusions concerning the history ofOkoumeacute It is notable that the vegetational impactsof the lsquowarmrsquo SST c 28002500 to 2000 years BPwere very different from those earlier c 9000 to65006000 years BP a difference which reflects abrutal strengthening over Central Africa of borealinfluences (Nguetsop et al 2000) to the detrimentof southern influences from about 28002500years BP

Before the major Holocene forest disturbanceculminating about 2500 BP other importantdisturbances occurred during the Quaternary(Maley 1996) The best documented is that betweenabout 20 000 and 15 000 years BP (Maley 19871996 Maley and Brenac 1998a Figure 2) In bothcases residual forest environments (refugia)

persisted apparently in similar locations butperhaps less as small blocks of relativelyhomogeneous forest than as a landscape of forest-savanna mosaic in which forest islands dominated(see Leal 2000) Yet two important characteristicsclearly differentiate these two phases of forestfragmentation First the older disturbance wasmuch longer Second they occurred under verydifferent (even opposed) general climatic conditionsThe disturbance between 20 000 and 15 000 BPcoincided with the development of glacialconditions in middle and high latitudes and coolerconditions by several degrees also affected the wholeregion of the Gulf of Guinea (Maley 1996 Maley andBrenac 1998a) The most recent disturbance bycontrast occurred during the Holocene interglacialwhich on a global scale is characterised by reducedglacial extension and relatively warmertemperatures The period between 2500 and 2000BP is associated with slightly warmer climaticconditions on the regional and global scale (Maley1997 Maley et al 2000) An important conclusion isthat the retreat and fragmentation of the Africanforests can be produced under very different

25

Figure 7 Variations in Sea Surface Temperature (SST) in the South Atlantic Ocean Benguela sector(22ucirc20rsquoSndash11ucirc12rsquoE) estimated with transfer functions based on the abundance of Radiolarian species in thecore RC13ndash228The open circles designate the 6000 and 9000 BP interpolated dates for the two curves one correspondingto the southern summer (February) and the other to the southern winter (August) modern SST areindicated by the solid squares above each curve The dashed lines represent some interpolated dates(italics) 18 000 BP based on the oxygen-isotope stratigraphy 38004000 BP 3000 BP and 2000 BP Theselate Holocene dates are confirmed by an independent study of shells from the same sector (Cohen et al1992) The coherence of the main SST anomalies throughout the Guinea Gulf permits the use of thesecurves in order to interpret the paleoclimatology of central Africa (Maley 1997)Source Morley and Dworetzky 1993

climates either relatively cool or relatively warmand thus with very different climatic situations(Maley 1996 1997)

Climatic models concerning lsquoglobal warmingrsquo inthe context of contemporary climate changeconcerns indicate that an average growth intemperature of about 4ucircC will lead also to anincrease in evaporation of about 30 per cent butonly 12 per cent more rain for tropical Africa (Rind

Notes The author is grateful to James Fairhead for his

translation of this article from the original Frenchand to Melissa Leach and James Fairhead for theirinvitation to the IDS workshop lsquoChangingPerspectives on Forestsrsquo in March 2001 This isInstitut des Sciences de lrsquoEvolution de Montpellier(ISEMCNRS) contribution No 125ndash2001

26

ReferencesAchoundong G Youta Happi J Guillet B

Bonvallot J and Kamgang Beyala V 2000lsquoFormation et eacutevolution des recrus sur savanes (ausud Cameroun)rsquo in M Servant and S Servant-Vildary (eds) Dynamique agrave long terme desEcosystegravemes forestiers intertropicaux MeacutemoireUNESCO Paris UNESCO 31ndash41

Alexandre P 1965 lsquoProto-histoire du groupe Beacuteti-Boulou-Fang essai de synthegravese provisoirersquo CahiersdrsquoEtudes Africaines Vol 20 503ndash60

Amougou Akoa 1986 lsquoEtude botanique eteacutecologique de la valleacutee inondable du Haut-Nyong(sud Cameroun) et de ses affluentsrsquo Thegravese SciencesUniv Yaoundeacute

Aubreacuteville A 1937 lsquoLes forecircts du Dahomey et duTogorsquo Bulletin du Comiteacute drsquoEtudes Historiques etScientifiques de lrsquoAfrique Occidentale Franccedilaises Vol20 112

Aubreacuteville A 1967 lsquoLes eacutetranges mosaiques forecirct-savane du sommet de la boucle de lrsquoOgooueacute auGabonrsquo Adansonia Vol 7 13ndash22

Baeke V 1996 lsquoLe temps des Rites Lrsquounivers magico-religieux des Wuli (Mfumte du Camerounoccidental)rsquo Thegravese Sc Sociale Univ Libre Bruxelles

Barber RJ 1985 lsquoLand snails and past environmentat the Igbo-Iwoto Esie site southwestern NigeriarsquoWest African Journal of Archaeology Vol 15 89ndash102

Barbier JC 1981 lsquoLe peuplement de la partiemeacuteridionale du plateau Bamileacutekeacutersquo in C Tardits

(ed) Contribution de la recherche ethnologique agravelrsquohistoire des civilisations du Cameroun ParisColloque International Centre National deRecherche Scientifique Paris Vol 2 331ndash53

Bertault JG 1990 lsquoComparaison drsquoeacutecosystegravemesforestiers naturels et modifieacutes apregraves incendie enCocircte drsquoIvoirersquo in H Puig (ed) Atelier surlrsquoAmeacutenagement et la Conservation de lrsquoEcosystegravemeForestier Tropical Humide Cayenne ParisUNESCO 6

Bertaux J Schwartz D Vincens A Sifeddine AElenga H Mansour M Mariotti A Fournier MMartin L Wirrmann D and Servant M 2000lsquoEnregistrement de la phase segraveche drsquoAfriqueCentrale vers 3000 ans BP par la spectromeacutetrie IRdans les lacs Sinnda et Kitina (sud-Congo)rsquo in MServant S Servant-Vildary (eds) Dynamique agrave longterme des Ecosystegravemes forestiers intertropicauxMeacutemoire UNESCO Paris UNESCO 43ndash9

Bigot S Camberlin P Moron V and Richard Y1997 lsquoStructures spatiales de la variabiliteacute despreacutecipitations en Afrique une transition climatiqueagrave la fin des anneacutees 1960rsquo Compte Rendu delrsquoAcademie des Sciences Paris 2a 324 181ndash88

Blanc-Pamard C and Peltre P 1984 lsquoDynamique despaysages preacuteforestiers et pratiques culturales enAfrique de lrsquoOuest (Cocircte drsquoIvoire centrale)rsquo LeDeacuteveloppement rural en question Paris MeacutemoireORSTOM No 106 55ndash74

Bonnefille R Chalieacute F Guiot J and Vincens A1992 lsquoQuantitative estimates of full glacial

1995) Though the causes of warming experiencedabout 2500 years BP were natural (Magny 1993)and therefore different from the warmingenvisaged for the twenty-first century it is possiblethat the catastrophic destruction of Central Africanforests which culminated about 2500 years BPcould be an lsquoanaloguersquo and an alarm signal forpotential developments during this current andfuture phase of human-induced global warming(Maley 1997 Maley et al 2000)

1 BP lsquoBefore the Presentrsquo according to standardisedC14 analysis

2 Only 4ndash5 samples have been studied for each 1000years during the Holocene To detect century-scaleperiodicity it would be necessary to study at least 10samples per millennium

temperatures in equatorial Africa frompalynological datarsquo Climate Dynamics Vol 6251ndash57

Brunck F Grison F and Maitre HF 1990Lrsquookoumeacute Aucoumea klaineana Pierre MonographieCentre Techn Forestier Tropical Nogent CIRAD 102

Cohen AI Perkington JE Brundrirt GB and vanDer Merwe NJ 1992 lsquoA Holocene marine climaterecord in Mollusc shells from the southwest Africancoastrsquo Quaternary Research Vol 38 379ndash85

Dechamps R Senut B and Pickford M 1992 Fruitsfossiles pliocegravenes et pleacuteistocegravenes du Rift occidentalougandais Signification paleacuteoenvironnementaleCompte Rendu de lrsquoAcadeacutemie des Sciences Parisseacuterie 2 Vol 314 325ndash331

Dowsett-Lemaire F 1996 lsquoComposition et eacutevolutionde la veacutegeacutetation forestiegravere au Parc NationaldrsquoOdzala Congorsquo Bulletin du Jardin BotaniqueNational de Belgique Vol 65 253ndash92

Dupont LM and Weinelt M 1996 lsquoVegetationhistory of the savanna corridor between theguinean and the Congolian rain forest during thelast 150000 yearsrsquo Vegetation History andArchaeobotany Vol 5 273ndash92

Elenga H Schwartz D and Vincens A 1994lsquoPollen evidence of late Quaternary vegetation andinferred climate changes in Congorsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 109345ndash56

Elenga H Schwartz D Vincens A Bertaux J DeNamur C Martin L Wirrmann D and ServantM 1996 lsquoDiagramme pollinique holocegravene du lacKitina (Congo) mise en eacutevidence de changementspaleacuteobotaniques et paleacuteoclimatiques dans le massifforestier du Mayombersquo Compte Rendu de lrsquoAcadeacutemiedes Sciences Paris Vol 323 No 2a 403ndash10

Elenga H Peyron O Bonnefille R Prentice ICJolly D Cheddadi R Guiot J Andrieu V DeBeaulieu JL Hamilton AC Maley J MarchantR Reille M Scott L Taylor D and Van CampoE 2000 lsquoPollen-based biome reconstruction forsouthern Europe and Africa 18000 years agorsquoJournal of Biogeography Vol 27 621ndash34

Fairhead J and Leach M 1995 lsquoFalse forest historycomplicit social analysis rethinking some westAfrican environmental narrativesrsquo WorldDevelopment Vol 23 1023ndash35

Fairhead J and Leach M 1996 lsquoEnriching thelandscape social history and the management oftransition ecology in the forest-savanna mosaic ofthe Republic of Guinearsquo Africa Vol 66 14ndash36

Fairhead J and Leach M 1998 lsquoReframingdeforestation Global analyses and local realitiesstudies in West Africarsquo Global EnvironmentalChanges Series London Routledge 238

Fay JM 1997 lsquoEvidence of forest degradation innorthern Congo and southwestern Central AfricanRepublic between 2340 and 990 BPrsquo in The ecologysocial organization populations habitat and historyof the western lowland Gorilla PhD ThesisWashington Univ Saint Louis Missouri Ch 6

Fontaine B and Bigot S 1993 lsquoWest African rainfalldeficit and sea surface temperaturesrsquo InternationalJournal of Climatology Vol 13 271ndash85

Gautier L 1990 lsquoContact forecirct-savane en CocirctedrsquoIvoire centrale eacutevolution du recouvrementligneux des savanes de la Reacuteserve de Lamto (sud duV-Baouleacute)rsquo Candollea Vol 45 627ndash41

Giresse P Maley J and Brenac P 1994 lsquoLateQuaternary palaeoenvironments in the lake BarombiMbo (Cameroon) deduced from pollen and carbonisotopes of organic matterrsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 107 65ndash78

Goldammer JG and Seibert B 1990 lsquoThe impact ofdroughts and forest fires on tropical lowland rainforest of East Kalimatanrsquo in JG Goldammer (ed)Fire in the Tropical Biota Ecological Studies BerlinSpringer-Verlag 84

Goucher CL 1981 lsquoIron is iron til it is rust tradeand ecology in the decline of West African iron-smeltingrsquo Journal of African History Vol 22 179ndash89

Guille-Escuret G 1990 lsquoPalmier agrave huile vin depalme et transformations sociales en Lobaye (ForecirctCentrafricaine)rsquo Information sur les Sciences SocialesVol 29 327ndash53

Hart TB 2001 lsquoForest dynamics in the Ituri basin(DR Congo) Dominance diversity andconservationrsquo in W Weber L White A Vedderand L Naughton-Treves (eds) African rain forestecology and conservation An interdisciplinaryperspective Yale Yale University Press 154ndash64

Hart TB Hart JA Dechamps R Fournier M andAtaholo M 1996 lsquoChanges in forest compositionover the last 4000 years in the Ituri basin Zairersquo inLJG Van der Maesen and XM Van der Burgt (eds)The Biodiversity of African plants Dordrecht Kluwer545ndash63

Hawthorne WD 1991 Fire damage and forestregeneration in Ghana Ghana Forestry Departmentamp ODA Report

Jolly D Harrison SP Damnati B and BonnefilleR 1998 lsquoSimulated climate and Biomes of Africa

27

16

Figure 3a Site of Lake Barombi Mbo west Cameroon percentages of diverse tree pollenTypical tree taxa from mature forestsSource Maley and Brenac 1998a

the Bateacutekeacute Plateau (Eastern Congo) (Elenga et al1994) Where dating can properly frame the eventit dates more precisely to between 2800 and 2500years BP

During this phase primary forest trees almostcompletely disappeared and in certain sites (inwest Cameroon at Barombi Mbo Mboandong andNjupi in south Congo and on the Bateacutekeacute Plateau)there was a major expansion of savannas Thisphase of forest destruction must have been verybrief in lacustrine strata where the abrupt andalmost complete disappearance of primary typetrees is observed an lsquoexplosionrsquo of pioneer woodyvegetation has been noted for some sites such as atlake Barombi Mbo in west Cameroon (Maley andBrenac 1998a) (Figure 3b) at lake Kitina in theMayombe western Congo (Elenga et al 1996) andat lake Ossa south Cameroon (Reynaud-Farrera etal 1996) The rapid expansion of pioneer treesmust have been the first phase of lsquoscabbingrsquo whichinitiated the reconstitution of the canopy The mostfrequent pollen taxa are the pioneers Alchorneacordifolia Musanga Trema Macaranga Elaeisguineensis etc Although brief destruction wasintense and several authors argue that the openingof the forest facilitated the major Bantu migrationfrom areas presently north of the forest to thosenow south of it (Schwartz 1992 Lavachery et al1996)

22 The reconstitution of the CentralAfrican forests during the last twomillennia

Humid conditions favourable for forest returnedearlier in west Cameroon in particular at BarombiMbo dating from 2000 years BP Yet thereconstitution especially of lsquoprimaryrsquo type forestswas not synchronous and probably reflects thedistribution of the residual forests (Figure 4) Thedelay can be understood in terms of lsquohysteresisrsquo(Maley and Brenac 1998a)

The reinvasion of savanna by forest continuedduring the twentieth century as observed overmuch of the forest-savanna transition region Thisis evidenced by the comparison of air and satellitephotographs of different dates from the 1950s tothe present (see Blanc-Pamard and Peltre 1984Maley 1990 1996 Fairhead and Leach 1995

17

Figure 3b Site of Lake Barombi Mbo westCameroon percentages of diverse tree pollenTypical tree taxa from pioneer forestsSource Maley and Brenac 1998a

18

Figure 4 Schematic drawing of the status of Central African vegetation at the phase of maximumperturbation and destruction of forest which culminated around 2500 BPLarge hatched areas represent the residual areas of the forest domain around 2500 BP which probablyconsisted of forest-savanna mosaic with predominantly larger forest islands Around these residual areasthe land would have been colonised by open vegetation above all savannas but also open canopy forestformations (see (4) in Figure 1) The base of this map presents the current status of Central Africanbiotopes as in Figure 1

1998 Servant 1996) For south Cameroon alone(south of the Adamaoua Plateau) forest gainedabout a million hectares during this period(Letouzey 1985 Youta Happi and Bonvallot 1996Achoundong et al 2000) Historical and palyno-logical research in Cocircte drsquoIvoire Nigeria Cameroonand Congo indicate that this reinvasion hascontinued over several centuries (Fairhead andLeach 1998 Maley 1999 Vincens et al 2000)Albeit with fluctuations and interludes it appearsto be the continuation of the forest reinvasionwhich began about 2000 years BP Currentreinvasion is thus the long-term result of themassive forest disturbance which happened around2500 years BP

A brief review of research findings concerning oilpalm and Okoumeacute can illustrate the history ofthese forest ecosystems over recent millennia

3 The history of oil palm and ofOkoumeacute in Central Africa duringthe Holocene impact of the forestdisturbance culminating about2500 years BP

31 The oil palm

Oil palm (Elaeis guineensis) has abundant andcharacteristic pollen and thus appears frequentlyin sedimented pollen profiles This palm is ofAfrican origin with its pollen and macro-remains(nuts) found in Equatorial African deposits datingback to the Tertiary (Zeven 1967 Dechamps et al1992 Maley 1996 1999 Maley and Brenac 1998aMaley and Chepstow-Lusty 2001)

The extent to which oil palm has been planted orits regeneration encouraged through land use hasbeen debated Certainly when people clear youngpioneer forests or fallows near to villages where oilpalm is naturally abundant they generallyconserve the palms effectively enriching themthrough fallow cycles (eg in Cameroon CentralAfrican Republic Cocircte drsquoIvoire Togo and Beninsee Maley 1999) Thus in 1937 the renownedbotanist Aubreacuteville characterised the lsquoimmense oilpalm grovesrsquo of southern Benin as lsquoa naturalformation simply improved by the inhabitantsover the centuriesrsquo Zevenrsquos research largely fromNigeria also concludes that with the exception of

modern industrial plantations most oil palms inAfrica are semi-wild and are not propagated bysowing or transplanting seedlings (see Zeven1972)

Ecologically oil palm is a pioneer that requires lightto complete the first stages of its growth so itdevelops naturally in windfall-clearings andespecially on the periphery of dense forest after thepassage of fire which it can tolerate (Swaine andHall 1986 Swaine 1992) The botanist Letouzey(1978 1985) has described a vast natural oil palmgrove in west Cameroon and near the north-west ofthe forest zone (Figure 5) A band of forest 10ndash20km wide dominated by large and numerous oilpalms extends over more than 150 km Thisfollows the boundary between forest and savannaat times from 5 to 30 km inside the forest From theabsence of trees classically found in anthropicplantations and other criteria Letouzey (19781985) concluded that this grove is a natural standA similar example has been described on theeastern flank of Mount Nimba (GuineaLiberianborder) by Schnell (1946) who considered that thedissemination of palm nuts was facilitated by bothtoucans and chimpanzees whose faeces oftencontain it abundantly These natural groves haveattracted migrant people (eg Guille-Escuret 1990Maley 1999) For example many Bamileke peoplemoved to the natural groves of western Cameroonin the mid nineteenth century (Barbier 1981Warnier 1985 Perrois and Notue 1997 Maley1999) In the vast lsquoGrassfieldsrsquo just north of westCameroonrsquos forest bloc the forest islands are still ina phase of expansion especially due to the oil palmwhich is one of the principal pioneer trees Close tothe Donga river the Wuli people colonised thesenatural stands Baeke (1996) explains how theWuli install their villages in the palm groves whereassorted social rules oppose all plantation of palmsMoreover she reports that a Wuli myth of originlsquoprovides evidence for the prior use of oil palms tothe working of land and clearly distinguishes twotypes of plant exploitation gathering andagriculturersquo (Baeke 1996)

As already noted during the major disturbanceabout 2500 BP oil palm was a principal pioneerthat subsequently developed strongly (Figure 3b)In west Cameroon the rapid forest recolonisationwhich occurred from 2000 years ago in the

19

vicinity of Barombi Mbo (Maley and Brenac1998a) was accompanied by a retreat of oil palmprobably because ecological and climatic factorsrendered other pioneers more active at this time InFigure 3 it is notable that in west Cameroonbetween 2000 and 1000 years BP the pioneer treeMilicia was in a phase of very strong expansionThen here and elsewhere in central Africa a secondphase of oil palm expansion began around 1400years BP This culminated towards 1000 BP at lakeOssa near to Edea (Reynaud-Farrera et al 1996) inthe south of the Central African Republic atNouabale-Ndoki (Fay 1997 Maley 1999) but also

further south in Gabon to La Lopeacute (White et al2000 Maley unpublished) and in western Congonear to lake Kitina (Elenga et al 1996) This phaseending about 700800 BP must have beenassociated with renewed forest disturbance butweaker than that 2500 BP This disturbance wasalso associated with a short phase of soil erosionand with a discontinuity in the deposits observedin assorted locations in Central Africa (Maley andBrenac 1998b) An increase in mineral fluxes hasbeen found between 1200ndash800 years BP for thesites of Kitina and Sinnda in western Congo(Bertaux et al 2000) From the thirteenth centurya phase of forest recovery has occurred in CentralAfrica with a new development of mature forestsand a generalised retreat of oil palm a recoverydespite increased human settlements (Alexandre1965 Vansina 1990)

At the western extension of the Central Africanforests lies a large area without forest which bio-geographers call the Dahomey Gap stretchingacross southern Togo and Benin Much of this iscovered by an immense palm forest of naturalorigin as it was described by Aubreacuteville (1937)and Mondjannagni (1969) Some pollen data showthat this area was covered by forests during theearly and mid Holocene (Dupont and Weinelt1996) and that the opening-up of the landscapeoccurred rather abruptly between 4000 and 3500 BPfollowing a climatic change well-documented forthe neighbouring region of Ghana at lakeBosumtwi (Maley 1991 1997 1999) Moreover itappears that the break in the forest bloc wasoriginally much wider than it is now perhapsextending across almost all of south-west Nigeriaup to the Niger river Indeed from the beginning ofthe last millennium to the sixteenth century ADsome data show that this vast sector was notforested but becoming so (Barber 1985) and thatthe vegetation was dominated by oil palmappearing similar to currently existing vegetationacross southern Benin and Togo Data indicatingthis exist for an area near to the ancient city ofBenin (Okomu Forest about 100 km west of theNiger river) suggesting also that the forestextension occurred only in recent centuries (Jones1956 White and Oates 1999) Further researchwill be necessary to clarify these various points andtheir precise chronology

20

Figure 5 An lsquooil palm beltrsquo near the edge of therain forest in west Cameroon

Belt with a high density of Elaeis guineensis(oil palm) reaching 20 to 25 m height inassociation with a mature forest of the semi-deciduous type

In the Grass Fields extensive savannasnorth of the forest pioneer forest islands withlarge concentrations of oil palms In italics thenames of some important groups of people(Bamileke Bamoun Tikar Wuli)Source adapted from Letouzey 1978 and 1985

32 OkoumeacuteOkoumeacute (Aucoumea Klaineana Burseraceae) iscurrently restricted to the west and centre ofGabon with a limited extension in the south-westof Congo and another towards the north inEquatorial Guinea (Brunck et al 1990) (Figure 1)Like oil palm Okoumeacute is a light demandingpioneer species (which for some ecological reasonappears unable to live outside of central AfricaBrunk et al 1990 Maley 1990 Nasi 1997) In itsarea this tree is perfectly adapted to colonisepioneer forest fronts as they progress into savannasand often dominates them (Nasi 1997 White et al2000) It is abundant in fallows and along foresttracks As pioneer fronts mature and age othertrees of shade-tolerant species appear (egCaesalpiniaceae) and the Okoumeacute which surviveare the dispersed individuals whose crowns havealready reached the canopy living to 100ndash150years and achieving 12 m diameters (Nasi 1997)It can be deduced that the large Okoumeacute currentlypresent in the forest interior in its natural rangewere born in contact with savannas

The endemism of Okoumeacute with its range limitedmainly to Gabon is linked to its history and itsecology It is intolerant of water deficit and thus ofthe elevated temperatures of the long dry seasons(about 3 months from December to February) oftropical climates Within its Gabonese range waterdeficit is less severe than in neighbouring southernCameroon because the equatorial climatersquos dryseason (mostly from June to August) retains theelevated air humidity and moderate temperatures(Brunck et al 1990 Maley 1990)

Several works bring important precision to thehistory of Okoumeacute First palynological dataobtained in south Cameroon at lake Ossa near toEdea (Reynaud-Farrera et al 1996) show thatbetween 7500 and 3000 BP (middle Holocene) thedistribution of Okoumeacute extended much furthernorth reaching the lake about 170 km north of thetreersquos present range and perhaps further east andfurther south in western Congo Second moleculargenetics show two completely distinct varieties ofOkoumeacute one to the north of its range and anotherto the south (Muloko et al 1998 2000) Theboundary between the ranges runs approximatelyeast-west a little to the south of the equatortowards 0ucirc40rsquoSouth

It can be deduced that the current range of Okoumeacutehas resulted from the fragmentation of a largeancient range dating back to the middle Holocenefragmented by the generalised forest disturbanceabout 2500 BP into two very dispersed sub-groupsone to the south and the other to the north Theextension of open landscapes during the period ofdestruction would have been favourable toOkoumeacutersquos new phase of development in a similarway to that evidenced at Barombi Mbo for severalother pioneer trees (Figure 3b) The general forestrecolonisation dating from about 2000 BP was inan extremely fragmented environment withnumerous ecotones favourable to the developmentof Okoumeacute It is thus likely that the current rangeof Okoumeacute is about 2000 years old at most

A dynamic model which nearly follows thischronological canvas has been established by White(1995) This is based on examination of the currentvegetation of the northern part of the La Lopeacutereserve which is situated at the heart of the Gabonrsquosforest and still holds large patches of residualsavanna (see Aubreacuteville 1967) White (1995) showshow the forest reconstituted itself progressivelythanks to the wide bands of pioneer forest rich inOkoumeacute which surround these savannas Theprogression of pioneer forest fronts can be blockedby savanna fires often of anthropic origin (Oslislyand White 2000) Even so over the long term fromabout 2000 years ago the continuous general trendhas been in favour of reforestation

That the current and expanding range of Okoumeacutehas not extended much into south Cameroon ashad been the case in the middle Holocene isprobably explained by a shift in climatic conditionsbetween this older period and the recent (post-2000 BP) Holocene This deduction is importantas given the ecological particularities of Okoumeacutewhich link it tightly to the equatorial climate onecan further deduce that equatorial climates musthave reigned over south Cameroon during themiddle Holocene The present tropical climatethere thus dates from only about 2000 BP All thisappears to translate for the middle Holocene intoan increase in lsquosouthernrsquo equatorial influencestowards the north associated with a reduction inlsquonorthernrsquo boreal influences and possibly linked toa reduction in the length of the dry seasonassociated with the boreal winter

21

4 The question of savanna firesand the extension of the forestdisturbance in eastern Congo

Observations and opinions have frequentlydiverged concerning the influence of savanna fireson contemporary forest regeneration Observersvariously report that fire can either preventregeneration as in the area of savanna inliers inLopeacute Gabon (Oslisly and White 2000) or merelydelay it as in other parts of the forest region ThusLetouzey (1968) notes for the forest-savannatransition zone of south Cameroon that lsquofire doesnot necessarily prevent the formation of forestrecoveryrsquo Similar conclusions have been drawn inCentral Africa by Sillans and even in westernCongo (ex-Zaire) for the south of the forest bloc(Letouzey 1968) On the same lines Youta Happiand Bonvallot (1996) report lsquothat on theperiphery of a town of 76000 people such asBertoua (east Cameroon) many small savannainlets into forest of two to ten hectares havebecome covered by forest regrowth despite theannual bush fires practiced by Bororopastoralistsrsquo A little further south in the OdzalaReserve situated at the northern extremity of theBateke savannas (north Congo) Dowsett-Lemairereports that lsquoWithout doubt one is in a phase offorest extension despite very frequent bush firesrsquo(1996) This author notes lsquothe fire-break role ofthe boundary thanks to groves dominated byGaertnera paniculata a species absent at Lopeacutewhere colonising species are apparently moresensitive to firersquo Thus the floristic composition ofthe boundary plays an important role in thisphenomenon This was shown in south-eastCameroon by the mid-twentieth century invasionof Chromolaena odorata which colonised theboundaries and which acts there at least as anefficacious fire break It also protects seeds ofwoody species which can germinate and developunder its cover (Youta Happi et al 2000) Otherobservations show how traditional agriculturalpractices and the movement of cattle can befavourable to forest recolonisation in particularby eliminating the grasses which facilitate thepropagation and intensity of fires These processeshave been well described in southern Cameroon(Letouzey 1968 1985) and in Cocircte drsquoIvoire(Spichiger and Blanc-Pamard 1973 Blanc-Pamardand Peltre 1984 Gautier 1990)

These observations indicate that in general andespecially over the long term forest colonisation atthe forest-savanna contact has been the broadlydominant process In certain exceptional yearswhen the dry season has been one or two monthslonger the savanna fires have penetrated moredeeply into the neighbouring forests In early 1983during a two-month longer dry season numerousobservers reported the strong invasion of fire intoforest in Cocircte drsquoIvoire (Bertault 1990) Ghana(Hawthorne 1991 Swaine 1992) and Cameroon(Amougou 1986) Equally fires in Indonesia thatyear were linked to an exceptional El Nintildeo event(Goldammer and Seibert 1990)

These observations indicate how the penetration offire into forest occurs only in very dry years Onemight thus think that fires could have played amajor role in the forest destruction 2500 BP Thishas indeed been shown in eastern BrazilianAmazonia during its major disturbance in themiddle Holocene That fires were frequent then isshown by the numerous charcoal fragmentsdeposited in the lake sediments at Carajas (Martinet al 1993 Servant et al 1993) Yet the lateHolocene lake deposits in Central Africa hardlycontain any charcoal fragments especially for theperiod 3000ndash2000 BP (Giresse et al 1994 Maleyand Brenac 1998a Bertaux et al 2000) It is thesame for the upper horizons of the numerous soilprofiles observed in this region (eg Valleacuterie 1973)In the thalwegs the coarse deposits in the lowerpart of the Lower Terrace do at times containcharcoal or fragments of fossil wood (Maley andBrenac 1998b) however Iron age archaeologicalsites dating from the beginning of the lateHolocene such as in the savanna areas of La Lopeacutealso present wood charcoal (Oslisly and White2000) One can conclude provisionally that thefires that existed were of limited extent

In particular charcoal associated with ironsmelting furnaces cannot have caused the forestdisturbance and cannot have been responsible forthe savanna extension about 2500 BP Goucher(1981) had advanced the hypothesis that ironsmelting could have been a major cause ofdeforestation in West Africa but Fairhead andLeach (1998) have demonstrated that thishypothesis had no serious foundation For CentralAfrica Pinccedilon (1990) also concludes that the

22

extraction of wood for iron metallurgy on theBateke Plateaux was minimal and did not explain(as had been suggested) the presence of theextensive savannas there Instead these savannasresult from an elevated edaphic drought caused byvery sandy soils in which rain infiltrates rapidly toa great depth (Laraque and Pandi 1996)

In contrast to observations from Atlantic CentralAfrica (south Cameroon Gabon and Congo) woodcharcoal has been found in abundance in Ituri nearto Epulu (c 1ucirc20rsquoNndash28ucirc35rsquoE) Here in the heart ofthe forests of the north-east Congo basin about 180km from the nearest savannas 416 soil profileshave been examined Almost all of them containnumerous wood charcoal in their upper levelsbetween the surface and 50 cm deep (Hart et al1996) Virtually all of the 1817 samples of woodcharcoal identified by Deschamps belong to treesfound in the regionrsquos diverse forest formationsexcept for a few pieces of charcoal from twoprofiles which belong to typical wooded savannaspecies The 28 dates obtained for the charcoalshow that fires occurred almost exclusively duringthe last three millennia with a phase of woodedsavanna expansion dating to about 2200 BP (Hartet al 1996) Moreover about 300 km south ofEpulu and 100 km from the forestsavanna limit aroad cutting near Osokari (1ucirc16rsquoSndash27ucirc48rsquoE)revealed a remarkable profile above the principalstone line that could be dated to about 11500years BP (Runge 1996 Maley 1996) A secondgravel level obliquely cutting the top of the soilprofile contains wood charcoal dating from about2200 to 1850 BP These results show that the forestdisturbance in Atlantic Central Africa extended toIturi and affected a large part of the eastern regionof the Congo river basin In contrast with regionsnearer the Atlantic fire had a large role

Also at Ituri (Epulu) Hart et al (1996) show thatcertain trees which are now found there are notfound in the wood charcoal in particularGilbertiodendron dewevrei (Caesalpiniaceae) which isnow abundant (almost mono-dominant) in certainparts of the forest The range of forests dominatedby Gilbertiodendron dewevrei extends to the east andnorth of the Congo basin as far as east Cameroon(Leacuteonard 1953) It is thus possible that thisCaesalpiniaceae has very recently colonised the Ituriforests and that the large area it now occupies is a

legacy of the disturbance about 2500 BP Ifconfirmed this area could correspond at leastpartly to an older forest area (of unknowncharacter) which was destroyed by the disturbanceA comparison with Okoumeacute can be suggestedwhich as indicated for Gabon above alsodominates areas which it has colonised since lessthan 2000 years ago However as the forest agesOkoumeacute loses its dominant character whereas inIturi Gilbertiodendron dewevrei remains dominantprobably because it can regenerate abundantly inthe shade of parent trees and also because itsreproductive character gives it an advantage overother trees in this formation (see Hart 2001)

5 Conclusion the major role ofclimatic conditions and theirhistoryThe present mosaic pattern of numerous forestscharacterised by a mix or juxtaposition ofgroupings of evergreen and deciduous species isprobably the consequence of long-term disturbanceswhich have affected the forest domain during thelast three millennia particularly the majordisturbance which culminated about 2500 years BPThe synchronism apparent in the disturbance fordifferent sites studied across Central Africa (southCameroon Gabon Congo as far east as the CongoRiver basin) and its association with a generalisederosive phase enables one to conclude that it wasthe result of a major climatic change

The climate at this time appears to have beenrelatively arid as it led to forest destruction and inplaces the extension of savannas Neverthelessother characteristics such as the strongsynchronous extension of diverse pioneer taxaindicate that in places there was not truly areduction in rainfall but rather a change in itsannual distribution The absence of a fall in annualrainfall has been demonstrated at lake BarombiMbo (Maley and Brenac 1998a) and at lake Ossawhere diatom studies show that there was even anet rise in this lake between 2500 and 2200 yearsBP with only a short fall between 2200 and 2000years BP (Nguetsop et al 2000) It is thereforepreferable to designate this catastrophic period asan lsquounfavourable climatic disturbancersquo caused by anincrease in seasonality and a lengthening of the dryseason (Maley 1997) The strong soil erosion of this

23

period suggests heavy rains but concentrated oversix or seven months of the year as is now found inthe peripheral savanna zones The growth inseasonality could have been associated with adominance of lsquosquall linesrsquo the typical cloudformation of savanna zones formed by northsouthalignments of cumuliform (convective storm-type)clouds They surplanted other cloud typesprincipally the stratiform types which giverelatively fine regular monsoon-type rains (seeMaley 1982) That these monsoon rains wouldhave dominated the earlier period (between 4000and 30002800 BP Maley 1997) is shown by therelative development of Caesalpinaceae in theforests of low altitude and in the mountains by theextension of Podocarpus which are trees typical ofcloud forests with stratiform clouds (Kerfoot1968 Maley 1996 1997 Maley and Brenac1998a)

Research in dynamic tropical climatology over thelast 20 years shows that Sea Surface Temperature(SST) on the regional and global scale stronglyinfluences climate (Fontaine and Bigot 1993

Moron et al 1995 Bigot et al 1997) It is possibleto characterise the relations between the monsoonrains and the SST more precisely Studies ofclimatic anomalies over tropical Africa occurringsince the beginning of the 1960s have evidencedtwo dominant modes of rainfall distribution eachassociated with a particular distribution of SST Thefirst mode is characterised on the one hand bywarmer than average temperatures in the Gulf ofGuinea and the southern Atlantic and on the otherhand by cooler temperatures on the north tropicalAtlantic offshore of West Africa The second modepresents an opposed distribution for the Africancontinent and the ocean A north-south tropicalAtlantic lsquodipolersquo has thus been found whichoscillates between one mode and the other(Fontaine and Bigot 1993 Wotling et al 1995Bigot et al 1997) An important research result hasbeen to show that over recent millennia the spatialdistribution of paleo-climatic anomalies overtropical Africa (more humid regions and more dryones Figure 6) and of SST on the neighbouringAtlantic is very similar to the spatial distributionobserved for the two dominant anomalies over thelast four decades (Maley 1997 Maley et al 2000)One can deduce that the climatic mechanisms arethe same at the timescale of years centuries ormillennia

So the variation of SST over the tropical Atlanticappears to play a major role in the variation ofclimate and rain types over Central Africa Thephase of forest disturbance that culminated about2500 years BP was associated with relatively lsquowarmrsquoSST over the Gulf of Guinea The earlier phasebetween about 40003800 and 2800 years BP wasassociated with relatively lsquocoolrsquo SST sharplyreduced from the early and middle Holocene (seeMorley and Dworetzky 1993 Figure 7) Over westand south Cameroon these SST allowed relativelyhigh rains and the evolution of stratiform cloudstowards rainy nimbostratus types The palaeo-vegetation data are presently lacking for GabonHowever new research is underway within thePALEOFORGA (Paleoenvironments of GabonForests) programme studying lakes MaridorNguegravene and Kamaleacuteteacute (Figure 1) In contrastfurther south in western Congo the periodstarting from 5000 years BP (Bertaux et al 2000)was marked by the development of semi-deciduousforests (Vincens et al 2000) linked to lower

24

Figure 6 Major rainfall trends in tropical Africa andSouth America for two opposed climatic periods c 3700ndash3000 BP and c 28002500ndash2000 BPPrecipitation trends (anomalies) abovenormalmean + below normalmean ndash P pointsto the sectors where a maximum of Podocarpuspollen intervenedSource Maley 1997 Maley et al 2000

rainfalls indicating that the stratiform clouds hadprobably evolved towards non-rainy clouds Thefour-month dry season here is characterised by thequasi-permanence of these cloud types (Saint Vil1984) which probably gained place progressivelybetween 5000 and 4000 years BP There wastherefore a reinforcing of southern influencesupporting conclusions concerning the history ofOkoumeacute It is notable that the vegetational impactsof the lsquowarmrsquo SST c 28002500 to 2000 years BPwere very different from those earlier c 9000 to65006000 years BP a difference which reflects abrutal strengthening over Central Africa of borealinfluences (Nguetsop et al 2000) to the detrimentof southern influences from about 28002500years BP

Before the major Holocene forest disturbanceculminating about 2500 BP other importantdisturbances occurred during the Quaternary(Maley 1996) The best documented is that betweenabout 20 000 and 15 000 years BP (Maley 19871996 Maley and Brenac 1998a Figure 2) In bothcases residual forest environments (refugia)

persisted apparently in similar locations butperhaps less as small blocks of relativelyhomogeneous forest than as a landscape of forest-savanna mosaic in which forest islands dominated(see Leal 2000) Yet two important characteristicsclearly differentiate these two phases of forestfragmentation First the older disturbance wasmuch longer Second they occurred under verydifferent (even opposed) general climatic conditionsThe disturbance between 20 000 and 15 000 BPcoincided with the development of glacialconditions in middle and high latitudes and coolerconditions by several degrees also affected the wholeregion of the Gulf of Guinea (Maley 1996 Maley andBrenac 1998a) The most recent disturbance bycontrast occurred during the Holocene interglacialwhich on a global scale is characterised by reducedglacial extension and relatively warmertemperatures The period between 2500 and 2000BP is associated with slightly warmer climaticconditions on the regional and global scale (Maley1997 Maley et al 2000) An important conclusion isthat the retreat and fragmentation of the Africanforests can be produced under very different

25

Figure 7 Variations in Sea Surface Temperature (SST) in the South Atlantic Ocean Benguela sector(22ucirc20rsquoSndash11ucirc12rsquoE) estimated with transfer functions based on the abundance of Radiolarian species in thecore RC13ndash228The open circles designate the 6000 and 9000 BP interpolated dates for the two curves one correspondingto the southern summer (February) and the other to the southern winter (August) modern SST areindicated by the solid squares above each curve The dashed lines represent some interpolated dates(italics) 18 000 BP based on the oxygen-isotope stratigraphy 38004000 BP 3000 BP and 2000 BP Theselate Holocene dates are confirmed by an independent study of shells from the same sector (Cohen et al1992) The coherence of the main SST anomalies throughout the Guinea Gulf permits the use of thesecurves in order to interpret the paleoclimatology of central Africa (Maley 1997)Source Morley and Dworetzky 1993

climates either relatively cool or relatively warmand thus with very different climatic situations(Maley 1996 1997)

Climatic models concerning lsquoglobal warmingrsquo inthe context of contemporary climate changeconcerns indicate that an average growth intemperature of about 4ucircC will lead also to anincrease in evaporation of about 30 per cent butonly 12 per cent more rain for tropical Africa (Rind

Notes The author is grateful to James Fairhead for his

translation of this article from the original Frenchand to Melissa Leach and James Fairhead for theirinvitation to the IDS workshop lsquoChangingPerspectives on Forestsrsquo in March 2001 This isInstitut des Sciences de lrsquoEvolution de Montpellier(ISEMCNRS) contribution No 125ndash2001

26

ReferencesAchoundong G Youta Happi J Guillet B

Bonvallot J and Kamgang Beyala V 2000lsquoFormation et eacutevolution des recrus sur savanes (ausud Cameroun)rsquo in M Servant and S Servant-Vildary (eds) Dynamique agrave long terme desEcosystegravemes forestiers intertropicaux MeacutemoireUNESCO Paris UNESCO 31ndash41

Alexandre P 1965 lsquoProto-histoire du groupe Beacuteti-Boulou-Fang essai de synthegravese provisoirersquo CahiersdrsquoEtudes Africaines Vol 20 503ndash60

Amougou Akoa 1986 lsquoEtude botanique eteacutecologique de la valleacutee inondable du Haut-Nyong(sud Cameroun) et de ses affluentsrsquo Thegravese SciencesUniv Yaoundeacute

Aubreacuteville A 1937 lsquoLes forecircts du Dahomey et duTogorsquo Bulletin du Comiteacute drsquoEtudes Historiques etScientifiques de lrsquoAfrique Occidentale Franccedilaises Vol20 112

Aubreacuteville A 1967 lsquoLes eacutetranges mosaiques forecirct-savane du sommet de la boucle de lrsquoOgooueacute auGabonrsquo Adansonia Vol 7 13ndash22

Baeke V 1996 lsquoLe temps des Rites Lrsquounivers magico-religieux des Wuli (Mfumte du Camerounoccidental)rsquo Thegravese Sc Sociale Univ Libre Bruxelles

Barber RJ 1985 lsquoLand snails and past environmentat the Igbo-Iwoto Esie site southwestern NigeriarsquoWest African Journal of Archaeology Vol 15 89ndash102

Barbier JC 1981 lsquoLe peuplement de la partiemeacuteridionale du plateau Bamileacutekeacutersquo in C Tardits

(ed) Contribution de la recherche ethnologique agravelrsquohistoire des civilisations du Cameroun ParisColloque International Centre National deRecherche Scientifique Paris Vol 2 331ndash53

Bertault JG 1990 lsquoComparaison drsquoeacutecosystegravemesforestiers naturels et modifieacutes apregraves incendie enCocircte drsquoIvoirersquo in H Puig (ed) Atelier surlrsquoAmeacutenagement et la Conservation de lrsquoEcosystegravemeForestier Tropical Humide Cayenne ParisUNESCO 6

Bertaux J Schwartz D Vincens A Sifeddine AElenga H Mansour M Mariotti A Fournier MMartin L Wirrmann D and Servant M 2000lsquoEnregistrement de la phase segraveche drsquoAfriqueCentrale vers 3000 ans BP par la spectromeacutetrie IRdans les lacs Sinnda et Kitina (sud-Congo)rsquo in MServant S Servant-Vildary (eds) Dynamique agrave longterme des Ecosystegravemes forestiers intertropicauxMeacutemoire UNESCO Paris UNESCO 43ndash9

Bigot S Camberlin P Moron V and Richard Y1997 lsquoStructures spatiales de la variabiliteacute despreacutecipitations en Afrique une transition climatiqueagrave la fin des anneacutees 1960rsquo Compte Rendu delrsquoAcademie des Sciences Paris 2a 324 181ndash88

Blanc-Pamard C and Peltre P 1984 lsquoDynamique despaysages preacuteforestiers et pratiques culturales enAfrique de lrsquoOuest (Cocircte drsquoIvoire centrale)rsquo LeDeacuteveloppement rural en question Paris MeacutemoireORSTOM No 106 55ndash74

Bonnefille R Chalieacute F Guiot J and Vincens A1992 lsquoQuantitative estimates of full glacial

1995) Though the causes of warming experiencedabout 2500 years BP were natural (Magny 1993)and therefore different from the warmingenvisaged for the twenty-first century it is possiblethat the catastrophic destruction of Central Africanforests which culminated about 2500 years BPcould be an lsquoanaloguersquo and an alarm signal forpotential developments during this current andfuture phase of human-induced global warming(Maley 1997 Maley et al 2000)

1 BP lsquoBefore the Presentrsquo according to standardisedC14 analysis

2 Only 4ndash5 samples have been studied for each 1000years during the Holocene To detect century-scaleperiodicity it would be necessary to study at least 10samples per millennium

temperatures in equatorial Africa frompalynological datarsquo Climate Dynamics Vol 6251ndash57

Brunck F Grison F and Maitre HF 1990Lrsquookoumeacute Aucoumea klaineana Pierre MonographieCentre Techn Forestier Tropical Nogent CIRAD 102

Cohen AI Perkington JE Brundrirt GB and vanDer Merwe NJ 1992 lsquoA Holocene marine climaterecord in Mollusc shells from the southwest Africancoastrsquo Quaternary Research Vol 38 379ndash85

Dechamps R Senut B and Pickford M 1992 Fruitsfossiles pliocegravenes et pleacuteistocegravenes du Rift occidentalougandais Signification paleacuteoenvironnementaleCompte Rendu de lrsquoAcadeacutemie des Sciences Parisseacuterie 2 Vol 314 325ndash331

Dowsett-Lemaire F 1996 lsquoComposition et eacutevolutionde la veacutegeacutetation forestiegravere au Parc NationaldrsquoOdzala Congorsquo Bulletin du Jardin BotaniqueNational de Belgique Vol 65 253ndash92

Dupont LM and Weinelt M 1996 lsquoVegetationhistory of the savanna corridor between theguinean and the Congolian rain forest during thelast 150000 yearsrsquo Vegetation History andArchaeobotany Vol 5 273ndash92

Elenga H Schwartz D and Vincens A 1994lsquoPollen evidence of late Quaternary vegetation andinferred climate changes in Congorsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 109345ndash56

Elenga H Schwartz D Vincens A Bertaux J DeNamur C Martin L Wirrmann D and ServantM 1996 lsquoDiagramme pollinique holocegravene du lacKitina (Congo) mise en eacutevidence de changementspaleacuteobotaniques et paleacuteoclimatiques dans le massifforestier du Mayombersquo Compte Rendu de lrsquoAcadeacutemiedes Sciences Paris Vol 323 No 2a 403ndash10

Elenga H Peyron O Bonnefille R Prentice ICJolly D Cheddadi R Guiot J Andrieu V DeBeaulieu JL Hamilton AC Maley J MarchantR Reille M Scott L Taylor D and Van CampoE 2000 lsquoPollen-based biome reconstruction forsouthern Europe and Africa 18000 years agorsquoJournal of Biogeography Vol 27 621ndash34

Fairhead J and Leach M 1995 lsquoFalse forest historycomplicit social analysis rethinking some westAfrican environmental narrativesrsquo WorldDevelopment Vol 23 1023ndash35

Fairhead J and Leach M 1996 lsquoEnriching thelandscape social history and the management oftransition ecology in the forest-savanna mosaic ofthe Republic of Guinearsquo Africa Vol 66 14ndash36

Fairhead J and Leach M 1998 lsquoReframingdeforestation Global analyses and local realitiesstudies in West Africarsquo Global EnvironmentalChanges Series London Routledge 238

Fay JM 1997 lsquoEvidence of forest degradation innorthern Congo and southwestern Central AfricanRepublic between 2340 and 990 BPrsquo in The ecologysocial organization populations habitat and historyof the western lowland Gorilla PhD ThesisWashington Univ Saint Louis Missouri Ch 6

Fontaine B and Bigot S 1993 lsquoWest African rainfalldeficit and sea surface temperaturesrsquo InternationalJournal of Climatology Vol 13 271ndash85

Gautier L 1990 lsquoContact forecirct-savane en CocirctedrsquoIvoire centrale eacutevolution du recouvrementligneux des savanes de la Reacuteserve de Lamto (sud duV-Baouleacute)rsquo Candollea Vol 45 627ndash41

Giresse P Maley J and Brenac P 1994 lsquoLateQuaternary palaeoenvironments in the lake BarombiMbo (Cameroon) deduced from pollen and carbonisotopes of organic matterrsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 107 65ndash78

Goldammer JG and Seibert B 1990 lsquoThe impact ofdroughts and forest fires on tropical lowland rainforest of East Kalimatanrsquo in JG Goldammer (ed)Fire in the Tropical Biota Ecological Studies BerlinSpringer-Verlag 84

Goucher CL 1981 lsquoIron is iron til it is rust tradeand ecology in the decline of West African iron-smeltingrsquo Journal of African History Vol 22 179ndash89

Guille-Escuret G 1990 lsquoPalmier agrave huile vin depalme et transformations sociales en Lobaye (ForecirctCentrafricaine)rsquo Information sur les Sciences SocialesVol 29 327ndash53

Hart TB 2001 lsquoForest dynamics in the Ituri basin(DR Congo) Dominance diversity andconservationrsquo in W Weber L White A Vedderand L Naughton-Treves (eds) African rain forestecology and conservation An interdisciplinaryperspective Yale Yale University Press 154ndash64

Hart TB Hart JA Dechamps R Fournier M andAtaholo M 1996 lsquoChanges in forest compositionover the last 4000 years in the Ituri basin Zairersquo inLJG Van der Maesen and XM Van der Burgt (eds)The Biodiversity of African plants Dordrecht Kluwer545ndash63

Hawthorne WD 1991 Fire damage and forestregeneration in Ghana Ghana Forestry Departmentamp ODA Report

Jolly D Harrison SP Damnati B and BonnefilleR 1998 lsquoSimulated climate and Biomes of Africa

27

the Bateacutekeacute Plateau (Eastern Congo) (Elenga et al1994) Where dating can properly frame the eventit dates more precisely to between 2800 and 2500years BP

During this phase primary forest trees almostcompletely disappeared and in certain sites (inwest Cameroon at Barombi Mbo Mboandong andNjupi in south Congo and on the Bateacutekeacute Plateau)there was a major expansion of savannas Thisphase of forest destruction must have been verybrief in lacustrine strata where the abrupt andalmost complete disappearance of primary typetrees is observed an lsquoexplosionrsquo of pioneer woodyvegetation has been noted for some sites such as atlake Barombi Mbo in west Cameroon (Maley andBrenac 1998a) (Figure 3b) at lake Kitina in theMayombe western Congo (Elenga et al 1996) andat lake Ossa south Cameroon (Reynaud-Farrera etal 1996) The rapid expansion of pioneer treesmust have been the first phase of lsquoscabbingrsquo whichinitiated the reconstitution of the canopy The mostfrequent pollen taxa are the pioneers Alchorneacordifolia Musanga Trema Macaranga Elaeisguineensis etc Although brief destruction wasintense and several authors argue that the openingof the forest facilitated the major Bantu migrationfrom areas presently north of the forest to thosenow south of it (Schwartz 1992 Lavachery et al1996)

22 The reconstitution of the CentralAfrican forests during the last twomillennia

Humid conditions favourable for forest returnedearlier in west Cameroon in particular at BarombiMbo dating from 2000 years BP Yet thereconstitution especially of lsquoprimaryrsquo type forestswas not synchronous and probably reflects thedistribution of the residual forests (Figure 4) Thedelay can be understood in terms of lsquohysteresisrsquo(Maley and Brenac 1998a)

The reinvasion of savanna by forest continuedduring the twentieth century as observed overmuch of the forest-savanna transition region Thisis evidenced by the comparison of air and satellitephotographs of different dates from the 1950s tothe present (see Blanc-Pamard and Peltre 1984Maley 1990 1996 Fairhead and Leach 1995

17

Figure 3b Site of Lake Barombi Mbo westCameroon percentages of diverse tree pollenTypical tree taxa from pioneer forestsSource Maley and Brenac 1998a

18

Figure 4 Schematic drawing of the status of Central African vegetation at the phase of maximumperturbation and destruction of forest which culminated around 2500 BPLarge hatched areas represent the residual areas of the forest domain around 2500 BP which probablyconsisted of forest-savanna mosaic with predominantly larger forest islands Around these residual areasthe land would have been colonised by open vegetation above all savannas but also open canopy forestformations (see (4) in Figure 1) The base of this map presents the current status of Central Africanbiotopes as in Figure 1

1998 Servant 1996) For south Cameroon alone(south of the Adamaoua Plateau) forest gainedabout a million hectares during this period(Letouzey 1985 Youta Happi and Bonvallot 1996Achoundong et al 2000) Historical and palyno-logical research in Cocircte drsquoIvoire Nigeria Cameroonand Congo indicate that this reinvasion hascontinued over several centuries (Fairhead andLeach 1998 Maley 1999 Vincens et al 2000)Albeit with fluctuations and interludes it appearsto be the continuation of the forest reinvasionwhich began about 2000 years BP Currentreinvasion is thus the long-term result of themassive forest disturbance which happened around2500 years BP

A brief review of research findings concerning oilpalm and Okoumeacute can illustrate the history ofthese forest ecosystems over recent millennia

3 The history of oil palm and ofOkoumeacute in Central Africa duringthe Holocene impact of the forestdisturbance culminating about2500 years BP

31 The oil palm

Oil palm (Elaeis guineensis) has abundant andcharacteristic pollen and thus appears frequentlyin sedimented pollen profiles This palm is ofAfrican origin with its pollen and macro-remains(nuts) found in Equatorial African deposits datingback to the Tertiary (Zeven 1967 Dechamps et al1992 Maley 1996 1999 Maley and Brenac 1998aMaley and Chepstow-Lusty 2001)

The extent to which oil palm has been planted orits regeneration encouraged through land use hasbeen debated Certainly when people clear youngpioneer forests or fallows near to villages where oilpalm is naturally abundant they generallyconserve the palms effectively enriching themthrough fallow cycles (eg in Cameroon CentralAfrican Republic Cocircte drsquoIvoire Togo and Beninsee Maley 1999) Thus in 1937 the renownedbotanist Aubreacuteville characterised the lsquoimmense oilpalm grovesrsquo of southern Benin as lsquoa naturalformation simply improved by the inhabitantsover the centuriesrsquo Zevenrsquos research largely fromNigeria also concludes that with the exception of

modern industrial plantations most oil palms inAfrica are semi-wild and are not propagated bysowing or transplanting seedlings (see Zeven1972)

Ecologically oil palm is a pioneer that requires lightto complete the first stages of its growth so itdevelops naturally in windfall-clearings andespecially on the periphery of dense forest after thepassage of fire which it can tolerate (Swaine andHall 1986 Swaine 1992) The botanist Letouzey(1978 1985) has described a vast natural oil palmgrove in west Cameroon and near the north-west ofthe forest zone (Figure 5) A band of forest 10ndash20km wide dominated by large and numerous oilpalms extends over more than 150 km Thisfollows the boundary between forest and savannaat times from 5 to 30 km inside the forest From theabsence of trees classically found in anthropicplantations and other criteria Letouzey (19781985) concluded that this grove is a natural standA similar example has been described on theeastern flank of Mount Nimba (GuineaLiberianborder) by Schnell (1946) who considered that thedissemination of palm nuts was facilitated by bothtoucans and chimpanzees whose faeces oftencontain it abundantly These natural groves haveattracted migrant people (eg Guille-Escuret 1990Maley 1999) For example many Bamileke peoplemoved to the natural groves of western Cameroonin the mid nineteenth century (Barbier 1981Warnier 1985 Perrois and Notue 1997 Maley1999) In the vast lsquoGrassfieldsrsquo just north of westCameroonrsquos forest bloc the forest islands are still ina phase of expansion especially due to the oil palmwhich is one of the principal pioneer trees Close tothe Donga river the Wuli people colonised thesenatural stands Baeke (1996) explains how theWuli install their villages in the palm groves whereassorted social rules oppose all plantation of palmsMoreover she reports that a Wuli myth of originlsquoprovides evidence for the prior use of oil palms tothe working of land and clearly distinguishes twotypes of plant exploitation gathering andagriculturersquo (Baeke 1996)

As already noted during the major disturbanceabout 2500 BP oil palm was a principal pioneerthat subsequently developed strongly (Figure 3b)In west Cameroon the rapid forest recolonisationwhich occurred from 2000 years ago in the

19

vicinity of Barombi Mbo (Maley and Brenac1998a) was accompanied by a retreat of oil palmprobably because ecological and climatic factorsrendered other pioneers more active at this time InFigure 3 it is notable that in west Cameroonbetween 2000 and 1000 years BP the pioneer treeMilicia was in a phase of very strong expansionThen here and elsewhere in central Africa a secondphase of oil palm expansion began around 1400years BP This culminated towards 1000 BP at lakeOssa near to Edea (Reynaud-Farrera et al 1996) inthe south of the Central African Republic atNouabale-Ndoki (Fay 1997 Maley 1999) but also

further south in Gabon to La Lopeacute (White et al2000 Maley unpublished) and in western Congonear to lake Kitina (Elenga et al 1996) This phaseending about 700800 BP must have beenassociated with renewed forest disturbance butweaker than that 2500 BP This disturbance wasalso associated with a short phase of soil erosionand with a discontinuity in the deposits observedin assorted locations in Central Africa (Maley andBrenac 1998b) An increase in mineral fluxes hasbeen found between 1200ndash800 years BP for thesites of Kitina and Sinnda in western Congo(Bertaux et al 2000) From the thirteenth centurya phase of forest recovery has occurred in CentralAfrica with a new development of mature forestsand a generalised retreat of oil palm a recoverydespite increased human settlements (Alexandre1965 Vansina 1990)

At the western extension of the Central Africanforests lies a large area without forest which bio-geographers call the Dahomey Gap stretchingacross southern Togo and Benin Much of this iscovered by an immense palm forest of naturalorigin as it was described by Aubreacuteville (1937)and Mondjannagni (1969) Some pollen data showthat this area was covered by forests during theearly and mid Holocene (Dupont and Weinelt1996) and that the opening-up of the landscapeoccurred rather abruptly between 4000 and 3500 BPfollowing a climatic change well-documented forthe neighbouring region of Ghana at lakeBosumtwi (Maley 1991 1997 1999) Moreover itappears that the break in the forest bloc wasoriginally much wider than it is now perhapsextending across almost all of south-west Nigeriaup to the Niger river Indeed from the beginning ofthe last millennium to the sixteenth century ADsome data show that this vast sector was notforested but becoming so (Barber 1985) and thatthe vegetation was dominated by oil palmappearing similar to currently existing vegetationacross southern Benin and Togo Data indicatingthis exist for an area near to the ancient city ofBenin (Okomu Forest about 100 km west of theNiger river) suggesting also that the forestextension occurred only in recent centuries (Jones1956 White and Oates 1999) Further researchwill be necessary to clarify these various points andtheir precise chronology

20

Figure 5 An lsquooil palm beltrsquo near the edge of therain forest in west Cameroon

Belt with a high density of Elaeis guineensis(oil palm) reaching 20 to 25 m height inassociation with a mature forest of the semi-deciduous type

In the Grass Fields extensive savannasnorth of the forest pioneer forest islands withlarge concentrations of oil palms In italics thenames of some important groups of people(Bamileke Bamoun Tikar Wuli)Source adapted from Letouzey 1978 and 1985

32 OkoumeacuteOkoumeacute (Aucoumea Klaineana Burseraceae) iscurrently restricted to the west and centre ofGabon with a limited extension in the south-westof Congo and another towards the north inEquatorial Guinea (Brunck et al 1990) (Figure 1)Like oil palm Okoumeacute is a light demandingpioneer species (which for some ecological reasonappears unable to live outside of central AfricaBrunk et al 1990 Maley 1990 Nasi 1997) In itsarea this tree is perfectly adapted to colonisepioneer forest fronts as they progress into savannasand often dominates them (Nasi 1997 White et al2000) It is abundant in fallows and along foresttracks As pioneer fronts mature and age othertrees of shade-tolerant species appear (egCaesalpiniaceae) and the Okoumeacute which surviveare the dispersed individuals whose crowns havealready reached the canopy living to 100ndash150years and achieving 12 m diameters (Nasi 1997)It can be deduced that the large Okoumeacute currentlypresent in the forest interior in its natural rangewere born in contact with savannas

The endemism of Okoumeacute with its range limitedmainly to Gabon is linked to its history and itsecology It is intolerant of water deficit and thus ofthe elevated temperatures of the long dry seasons(about 3 months from December to February) oftropical climates Within its Gabonese range waterdeficit is less severe than in neighbouring southernCameroon because the equatorial climatersquos dryseason (mostly from June to August) retains theelevated air humidity and moderate temperatures(Brunck et al 1990 Maley 1990)

Several works bring important precision to thehistory of Okoumeacute First palynological dataobtained in south Cameroon at lake Ossa near toEdea (Reynaud-Farrera et al 1996) show thatbetween 7500 and 3000 BP (middle Holocene) thedistribution of Okoumeacute extended much furthernorth reaching the lake about 170 km north of thetreersquos present range and perhaps further east andfurther south in western Congo Second moleculargenetics show two completely distinct varieties ofOkoumeacute one to the north of its range and anotherto the south (Muloko et al 1998 2000) Theboundary between the ranges runs approximatelyeast-west a little to the south of the equatortowards 0ucirc40rsquoSouth

It can be deduced that the current range of Okoumeacutehas resulted from the fragmentation of a largeancient range dating back to the middle Holocenefragmented by the generalised forest disturbanceabout 2500 BP into two very dispersed sub-groupsone to the south and the other to the north Theextension of open landscapes during the period ofdestruction would have been favourable toOkoumeacutersquos new phase of development in a similarway to that evidenced at Barombi Mbo for severalother pioneer trees (Figure 3b) The general forestrecolonisation dating from about 2000 BP was inan extremely fragmented environment withnumerous ecotones favourable to the developmentof Okoumeacute It is thus likely that the current rangeof Okoumeacute is about 2000 years old at most

A dynamic model which nearly follows thischronological canvas has been established by White(1995) This is based on examination of the currentvegetation of the northern part of the La Lopeacutereserve which is situated at the heart of the Gabonrsquosforest and still holds large patches of residualsavanna (see Aubreacuteville 1967) White (1995) showshow the forest reconstituted itself progressivelythanks to the wide bands of pioneer forest rich inOkoumeacute which surround these savannas Theprogression of pioneer forest fronts can be blockedby savanna fires often of anthropic origin (Oslislyand White 2000) Even so over the long term fromabout 2000 years ago the continuous general trendhas been in favour of reforestation

That the current and expanding range of Okoumeacutehas not extended much into south Cameroon ashad been the case in the middle Holocene isprobably explained by a shift in climatic conditionsbetween this older period and the recent (post-2000 BP) Holocene This deduction is importantas given the ecological particularities of Okoumeacutewhich link it tightly to the equatorial climate onecan further deduce that equatorial climates musthave reigned over south Cameroon during themiddle Holocene The present tropical climatethere thus dates from only about 2000 BP All thisappears to translate for the middle Holocene intoan increase in lsquosouthernrsquo equatorial influencestowards the north associated with a reduction inlsquonorthernrsquo boreal influences and possibly linked toa reduction in the length of the dry seasonassociated with the boreal winter

21

4 The question of savanna firesand the extension of the forestdisturbance in eastern Congo

Observations and opinions have frequentlydiverged concerning the influence of savanna fireson contemporary forest regeneration Observersvariously report that fire can either preventregeneration as in the area of savanna inliers inLopeacute Gabon (Oslisly and White 2000) or merelydelay it as in other parts of the forest region ThusLetouzey (1968) notes for the forest-savannatransition zone of south Cameroon that lsquofire doesnot necessarily prevent the formation of forestrecoveryrsquo Similar conclusions have been drawn inCentral Africa by Sillans and even in westernCongo (ex-Zaire) for the south of the forest bloc(Letouzey 1968) On the same lines Youta Happiand Bonvallot (1996) report lsquothat on theperiphery of a town of 76000 people such asBertoua (east Cameroon) many small savannainlets into forest of two to ten hectares havebecome covered by forest regrowth despite theannual bush fires practiced by Bororopastoralistsrsquo A little further south in the OdzalaReserve situated at the northern extremity of theBateke savannas (north Congo) Dowsett-Lemairereports that lsquoWithout doubt one is in a phase offorest extension despite very frequent bush firesrsquo(1996) This author notes lsquothe fire-break role ofthe boundary thanks to groves dominated byGaertnera paniculata a species absent at Lopeacutewhere colonising species are apparently moresensitive to firersquo Thus the floristic composition ofthe boundary plays an important role in thisphenomenon This was shown in south-eastCameroon by the mid-twentieth century invasionof Chromolaena odorata which colonised theboundaries and which acts there at least as anefficacious fire break It also protects seeds ofwoody species which can germinate and developunder its cover (Youta Happi et al 2000) Otherobservations show how traditional agriculturalpractices and the movement of cattle can befavourable to forest recolonisation in particularby eliminating the grasses which facilitate thepropagation and intensity of fires These processeshave been well described in southern Cameroon(Letouzey 1968 1985) and in Cocircte drsquoIvoire(Spichiger and Blanc-Pamard 1973 Blanc-Pamardand Peltre 1984 Gautier 1990)

These observations indicate that in general andespecially over the long term forest colonisation atthe forest-savanna contact has been the broadlydominant process In certain exceptional yearswhen the dry season has been one or two monthslonger the savanna fires have penetrated moredeeply into the neighbouring forests In early 1983during a two-month longer dry season numerousobservers reported the strong invasion of fire intoforest in Cocircte drsquoIvoire (Bertault 1990) Ghana(Hawthorne 1991 Swaine 1992) and Cameroon(Amougou 1986) Equally fires in Indonesia thatyear were linked to an exceptional El Nintildeo event(Goldammer and Seibert 1990)

These observations indicate how the penetration offire into forest occurs only in very dry years Onemight thus think that fires could have played amajor role in the forest destruction 2500 BP Thishas indeed been shown in eastern BrazilianAmazonia during its major disturbance in themiddle Holocene That fires were frequent then isshown by the numerous charcoal fragmentsdeposited in the lake sediments at Carajas (Martinet al 1993 Servant et al 1993) Yet the lateHolocene lake deposits in Central Africa hardlycontain any charcoal fragments especially for theperiod 3000ndash2000 BP (Giresse et al 1994 Maleyand Brenac 1998a Bertaux et al 2000) It is thesame for the upper horizons of the numerous soilprofiles observed in this region (eg Valleacuterie 1973)In the thalwegs the coarse deposits in the lowerpart of the Lower Terrace do at times containcharcoal or fragments of fossil wood (Maley andBrenac 1998b) however Iron age archaeologicalsites dating from the beginning of the lateHolocene such as in the savanna areas of La Lopeacutealso present wood charcoal (Oslisly and White2000) One can conclude provisionally that thefires that existed were of limited extent

In particular charcoal associated with ironsmelting furnaces cannot have caused the forestdisturbance and cannot have been responsible forthe savanna extension about 2500 BP Goucher(1981) had advanced the hypothesis that ironsmelting could have been a major cause ofdeforestation in West Africa but Fairhead andLeach (1998) have demonstrated that thishypothesis had no serious foundation For CentralAfrica Pinccedilon (1990) also concludes that the

22

extraction of wood for iron metallurgy on theBateke Plateaux was minimal and did not explain(as had been suggested) the presence of theextensive savannas there Instead these savannasresult from an elevated edaphic drought caused byvery sandy soils in which rain infiltrates rapidly toa great depth (Laraque and Pandi 1996)

In contrast to observations from Atlantic CentralAfrica (south Cameroon Gabon and Congo) woodcharcoal has been found in abundance in Ituri nearto Epulu (c 1ucirc20rsquoNndash28ucirc35rsquoE) Here in the heart ofthe forests of the north-east Congo basin about 180km from the nearest savannas 416 soil profileshave been examined Almost all of them containnumerous wood charcoal in their upper levelsbetween the surface and 50 cm deep (Hart et al1996) Virtually all of the 1817 samples of woodcharcoal identified by Deschamps belong to treesfound in the regionrsquos diverse forest formationsexcept for a few pieces of charcoal from twoprofiles which belong to typical wooded savannaspecies The 28 dates obtained for the charcoalshow that fires occurred almost exclusively duringthe last three millennia with a phase of woodedsavanna expansion dating to about 2200 BP (Hartet al 1996) Moreover about 300 km south ofEpulu and 100 km from the forestsavanna limit aroad cutting near Osokari (1ucirc16rsquoSndash27ucirc48rsquoE)revealed a remarkable profile above the principalstone line that could be dated to about 11500years BP (Runge 1996 Maley 1996) A secondgravel level obliquely cutting the top of the soilprofile contains wood charcoal dating from about2200 to 1850 BP These results show that the forestdisturbance in Atlantic Central Africa extended toIturi and affected a large part of the eastern regionof the Congo river basin In contrast with regionsnearer the Atlantic fire had a large role

Also at Ituri (Epulu) Hart et al (1996) show thatcertain trees which are now found there are notfound in the wood charcoal in particularGilbertiodendron dewevrei (Caesalpiniaceae) which isnow abundant (almost mono-dominant) in certainparts of the forest The range of forests dominatedby Gilbertiodendron dewevrei extends to the east andnorth of the Congo basin as far as east Cameroon(Leacuteonard 1953) It is thus possible that thisCaesalpiniaceae has very recently colonised the Ituriforests and that the large area it now occupies is a

legacy of the disturbance about 2500 BP Ifconfirmed this area could correspond at leastpartly to an older forest area (of unknowncharacter) which was destroyed by the disturbanceA comparison with Okoumeacute can be suggestedwhich as indicated for Gabon above alsodominates areas which it has colonised since lessthan 2000 years ago However as the forest agesOkoumeacute loses its dominant character whereas inIturi Gilbertiodendron dewevrei remains dominantprobably because it can regenerate abundantly inthe shade of parent trees and also because itsreproductive character gives it an advantage overother trees in this formation (see Hart 2001)

5 Conclusion the major role ofclimatic conditions and theirhistoryThe present mosaic pattern of numerous forestscharacterised by a mix or juxtaposition ofgroupings of evergreen and deciduous species isprobably the consequence of long-term disturbanceswhich have affected the forest domain during thelast three millennia particularly the majordisturbance which culminated about 2500 years BPThe synchronism apparent in the disturbance fordifferent sites studied across Central Africa (southCameroon Gabon Congo as far east as the CongoRiver basin) and its association with a generalisederosive phase enables one to conclude that it wasthe result of a major climatic change

The climate at this time appears to have beenrelatively arid as it led to forest destruction and inplaces the extension of savannas Neverthelessother characteristics such as the strongsynchronous extension of diverse pioneer taxaindicate that in places there was not truly areduction in rainfall but rather a change in itsannual distribution The absence of a fall in annualrainfall has been demonstrated at lake BarombiMbo (Maley and Brenac 1998a) and at lake Ossawhere diatom studies show that there was even anet rise in this lake between 2500 and 2200 yearsBP with only a short fall between 2200 and 2000years BP (Nguetsop et al 2000) It is thereforepreferable to designate this catastrophic period asan lsquounfavourable climatic disturbancersquo caused by anincrease in seasonality and a lengthening of the dryseason (Maley 1997) The strong soil erosion of this

23

period suggests heavy rains but concentrated oversix or seven months of the year as is now found inthe peripheral savanna zones The growth inseasonality could have been associated with adominance of lsquosquall linesrsquo the typical cloudformation of savanna zones formed by northsouthalignments of cumuliform (convective storm-type)clouds They surplanted other cloud typesprincipally the stratiform types which giverelatively fine regular monsoon-type rains (seeMaley 1982) That these monsoon rains wouldhave dominated the earlier period (between 4000and 30002800 BP Maley 1997) is shown by therelative development of Caesalpinaceae in theforests of low altitude and in the mountains by theextension of Podocarpus which are trees typical ofcloud forests with stratiform clouds (Kerfoot1968 Maley 1996 1997 Maley and Brenac1998a)

Research in dynamic tropical climatology over thelast 20 years shows that Sea Surface Temperature(SST) on the regional and global scale stronglyinfluences climate (Fontaine and Bigot 1993

Moron et al 1995 Bigot et al 1997) It is possibleto characterise the relations between the monsoonrains and the SST more precisely Studies ofclimatic anomalies over tropical Africa occurringsince the beginning of the 1960s have evidencedtwo dominant modes of rainfall distribution eachassociated with a particular distribution of SST Thefirst mode is characterised on the one hand bywarmer than average temperatures in the Gulf ofGuinea and the southern Atlantic and on the otherhand by cooler temperatures on the north tropicalAtlantic offshore of West Africa The second modepresents an opposed distribution for the Africancontinent and the ocean A north-south tropicalAtlantic lsquodipolersquo has thus been found whichoscillates between one mode and the other(Fontaine and Bigot 1993 Wotling et al 1995Bigot et al 1997) An important research result hasbeen to show that over recent millennia the spatialdistribution of paleo-climatic anomalies overtropical Africa (more humid regions and more dryones Figure 6) and of SST on the neighbouringAtlantic is very similar to the spatial distributionobserved for the two dominant anomalies over thelast four decades (Maley 1997 Maley et al 2000)One can deduce that the climatic mechanisms arethe same at the timescale of years centuries ormillennia

So the variation of SST over the tropical Atlanticappears to play a major role in the variation ofclimate and rain types over Central Africa Thephase of forest disturbance that culminated about2500 years BP was associated with relatively lsquowarmrsquoSST over the Gulf of Guinea The earlier phasebetween about 40003800 and 2800 years BP wasassociated with relatively lsquocoolrsquo SST sharplyreduced from the early and middle Holocene (seeMorley and Dworetzky 1993 Figure 7) Over westand south Cameroon these SST allowed relativelyhigh rains and the evolution of stratiform cloudstowards rainy nimbostratus types The palaeo-vegetation data are presently lacking for GabonHowever new research is underway within thePALEOFORGA (Paleoenvironments of GabonForests) programme studying lakes MaridorNguegravene and Kamaleacuteteacute (Figure 1) In contrastfurther south in western Congo the periodstarting from 5000 years BP (Bertaux et al 2000)was marked by the development of semi-deciduousforests (Vincens et al 2000) linked to lower

24

Figure 6 Major rainfall trends in tropical Africa andSouth America for two opposed climatic periods c 3700ndash3000 BP and c 28002500ndash2000 BPPrecipitation trends (anomalies) abovenormalmean + below normalmean ndash P pointsto the sectors where a maximum of Podocarpuspollen intervenedSource Maley 1997 Maley et al 2000

rainfalls indicating that the stratiform clouds hadprobably evolved towards non-rainy clouds Thefour-month dry season here is characterised by thequasi-permanence of these cloud types (Saint Vil1984) which probably gained place progressivelybetween 5000 and 4000 years BP There wastherefore a reinforcing of southern influencesupporting conclusions concerning the history ofOkoumeacute It is notable that the vegetational impactsof the lsquowarmrsquo SST c 28002500 to 2000 years BPwere very different from those earlier c 9000 to65006000 years BP a difference which reflects abrutal strengthening over Central Africa of borealinfluences (Nguetsop et al 2000) to the detrimentof southern influences from about 28002500years BP

Before the major Holocene forest disturbanceculminating about 2500 BP other importantdisturbances occurred during the Quaternary(Maley 1996) The best documented is that betweenabout 20 000 and 15 000 years BP (Maley 19871996 Maley and Brenac 1998a Figure 2) In bothcases residual forest environments (refugia)

persisted apparently in similar locations butperhaps less as small blocks of relativelyhomogeneous forest than as a landscape of forest-savanna mosaic in which forest islands dominated(see Leal 2000) Yet two important characteristicsclearly differentiate these two phases of forestfragmentation First the older disturbance wasmuch longer Second they occurred under verydifferent (even opposed) general climatic conditionsThe disturbance between 20 000 and 15 000 BPcoincided with the development of glacialconditions in middle and high latitudes and coolerconditions by several degrees also affected the wholeregion of the Gulf of Guinea (Maley 1996 Maley andBrenac 1998a) The most recent disturbance bycontrast occurred during the Holocene interglacialwhich on a global scale is characterised by reducedglacial extension and relatively warmertemperatures The period between 2500 and 2000BP is associated with slightly warmer climaticconditions on the regional and global scale (Maley1997 Maley et al 2000) An important conclusion isthat the retreat and fragmentation of the Africanforests can be produced under very different

25

Figure 7 Variations in Sea Surface Temperature (SST) in the South Atlantic Ocean Benguela sector(22ucirc20rsquoSndash11ucirc12rsquoE) estimated with transfer functions based on the abundance of Radiolarian species in thecore RC13ndash228The open circles designate the 6000 and 9000 BP interpolated dates for the two curves one correspondingto the southern summer (February) and the other to the southern winter (August) modern SST areindicated by the solid squares above each curve The dashed lines represent some interpolated dates(italics) 18 000 BP based on the oxygen-isotope stratigraphy 38004000 BP 3000 BP and 2000 BP Theselate Holocene dates are confirmed by an independent study of shells from the same sector (Cohen et al1992) The coherence of the main SST anomalies throughout the Guinea Gulf permits the use of thesecurves in order to interpret the paleoclimatology of central Africa (Maley 1997)Source Morley and Dworetzky 1993

climates either relatively cool or relatively warmand thus with very different climatic situations(Maley 1996 1997)

Climatic models concerning lsquoglobal warmingrsquo inthe context of contemporary climate changeconcerns indicate that an average growth intemperature of about 4ucircC will lead also to anincrease in evaporation of about 30 per cent butonly 12 per cent more rain for tropical Africa (Rind

Notes The author is grateful to James Fairhead for his

translation of this article from the original Frenchand to Melissa Leach and James Fairhead for theirinvitation to the IDS workshop lsquoChangingPerspectives on Forestsrsquo in March 2001 This isInstitut des Sciences de lrsquoEvolution de Montpellier(ISEMCNRS) contribution No 125ndash2001

26

ReferencesAchoundong G Youta Happi J Guillet B

Bonvallot J and Kamgang Beyala V 2000lsquoFormation et eacutevolution des recrus sur savanes (ausud Cameroun)rsquo in M Servant and S Servant-Vildary (eds) Dynamique agrave long terme desEcosystegravemes forestiers intertropicaux MeacutemoireUNESCO Paris UNESCO 31ndash41

Alexandre P 1965 lsquoProto-histoire du groupe Beacuteti-Boulou-Fang essai de synthegravese provisoirersquo CahiersdrsquoEtudes Africaines Vol 20 503ndash60

Amougou Akoa 1986 lsquoEtude botanique eteacutecologique de la valleacutee inondable du Haut-Nyong(sud Cameroun) et de ses affluentsrsquo Thegravese SciencesUniv Yaoundeacute

Aubreacuteville A 1937 lsquoLes forecircts du Dahomey et duTogorsquo Bulletin du Comiteacute drsquoEtudes Historiques etScientifiques de lrsquoAfrique Occidentale Franccedilaises Vol20 112

Aubreacuteville A 1967 lsquoLes eacutetranges mosaiques forecirct-savane du sommet de la boucle de lrsquoOgooueacute auGabonrsquo Adansonia Vol 7 13ndash22

Baeke V 1996 lsquoLe temps des Rites Lrsquounivers magico-religieux des Wuli (Mfumte du Camerounoccidental)rsquo Thegravese Sc Sociale Univ Libre Bruxelles

Barber RJ 1985 lsquoLand snails and past environmentat the Igbo-Iwoto Esie site southwestern NigeriarsquoWest African Journal of Archaeology Vol 15 89ndash102

Barbier JC 1981 lsquoLe peuplement de la partiemeacuteridionale du plateau Bamileacutekeacutersquo in C Tardits

(ed) Contribution de la recherche ethnologique agravelrsquohistoire des civilisations du Cameroun ParisColloque International Centre National deRecherche Scientifique Paris Vol 2 331ndash53

Bertault JG 1990 lsquoComparaison drsquoeacutecosystegravemesforestiers naturels et modifieacutes apregraves incendie enCocircte drsquoIvoirersquo in H Puig (ed) Atelier surlrsquoAmeacutenagement et la Conservation de lrsquoEcosystegravemeForestier Tropical Humide Cayenne ParisUNESCO 6

Bertaux J Schwartz D Vincens A Sifeddine AElenga H Mansour M Mariotti A Fournier MMartin L Wirrmann D and Servant M 2000lsquoEnregistrement de la phase segraveche drsquoAfriqueCentrale vers 3000 ans BP par la spectromeacutetrie IRdans les lacs Sinnda et Kitina (sud-Congo)rsquo in MServant S Servant-Vildary (eds) Dynamique agrave longterme des Ecosystegravemes forestiers intertropicauxMeacutemoire UNESCO Paris UNESCO 43ndash9

Bigot S Camberlin P Moron V and Richard Y1997 lsquoStructures spatiales de la variabiliteacute despreacutecipitations en Afrique une transition climatiqueagrave la fin des anneacutees 1960rsquo Compte Rendu delrsquoAcademie des Sciences Paris 2a 324 181ndash88

Blanc-Pamard C and Peltre P 1984 lsquoDynamique despaysages preacuteforestiers et pratiques culturales enAfrique de lrsquoOuest (Cocircte drsquoIvoire centrale)rsquo LeDeacuteveloppement rural en question Paris MeacutemoireORSTOM No 106 55ndash74

Bonnefille R Chalieacute F Guiot J and Vincens A1992 lsquoQuantitative estimates of full glacial

1995) Though the causes of warming experiencedabout 2500 years BP were natural (Magny 1993)and therefore different from the warmingenvisaged for the twenty-first century it is possiblethat the catastrophic destruction of Central Africanforests which culminated about 2500 years BPcould be an lsquoanaloguersquo and an alarm signal forpotential developments during this current andfuture phase of human-induced global warming(Maley 1997 Maley et al 2000)

1 BP lsquoBefore the Presentrsquo according to standardisedC14 analysis

2 Only 4ndash5 samples have been studied for each 1000years during the Holocene To detect century-scaleperiodicity it would be necessary to study at least 10samples per millennium

temperatures in equatorial Africa frompalynological datarsquo Climate Dynamics Vol 6251ndash57

Brunck F Grison F and Maitre HF 1990Lrsquookoumeacute Aucoumea klaineana Pierre MonographieCentre Techn Forestier Tropical Nogent CIRAD 102

Cohen AI Perkington JE Brundrirt GB and vanDer Merwe NJ 1992 lsquoA Holocene marine climaterecord in Mollusc shells from the southwest Africancoastrsquo Quaternary Research Vol 38 379ndash85

Dechamps R Senut B and Pickford M 1992 Fruitsfossiles pliocegravenes et pleacuteistocegravenes du Rift occidentalougandais Signification paleacuteoenvironnementaleCompte Rendu de lrsquoAcadeacutemie des Sciences Parisseacuterie 2 Vol 314 325ndash331

Dowsett-Lemaire F 1996 lsquoComposition et eacutevolutionde la veacutegeacutetation forestiegravere au Parc NationaldrsquoOdzala Congorsquo Bulletin du Jardin BotaniqueNational de Belgique Vol 65 253ndash92

Dupont LM and Weinelt M 1996 lsquoVegetationhistory of the savanna corridor between theguinean and the Congolian rain forest during thelast 150000 yearsrsquo Vegetation History andArchaeobotany Vol 5 273ndash92

Elenga H Schwartz D and Vincens A 1994lsquoPollen evidence of late Quaternary vegetation andinferred climate changes in Congorsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 109345ndash56

Elenga H Schwartz D Vincens A Bertaux J DeNamur C Martin L Wirrmann D and ServantM 1996 lsquoDiagramme pollinique holocegravene du lacKitina (Congo) mise en eacutevidence de changementspaleacuteobotaniques et paleacuteoclimatiques dans le massifforestier du Mayombersquo Compte Rendu de lrsquoAcadeacutemiedes Sciences Paris Vol 323 No 2a 403ndash10

Elenga H Peyron O Bonnefille R Prentice ICJolly D Cheddadi R Guiot J Andrieu V DeBeaulieu JL Hamilton AC Maley J MarchantR Reille M Scott L Taylor D and Van CampoE 2000 lsquoPollen-based biome reconstruction forsouthern Europe and Africa 18000 years agorsquoJournal of Biogeography Vol 27 621ndash34

Fairhead J and Leach M 1995 lsquoFalse forest historycomplicit social analysis rethinking some westAfrican environmental narrativesrsquo WorldDevelopment Vol 23 1023ndash35

Fairhead J and Leach M 1996 lsquoEnriching thelandscape social history and the management oftransition ecology in the forest-savanna mosaic ofthe Republic of Guinearsquo Africa Vol 66 14ndash36

Fairhead J and Leach M 1998 lsquoReframingdeforestation Global analyses and local realitiesstudies in West Africarsquo Global EnvironmentalChanges Series London Routledge 238

Fay JM 1997 lsquoEvidence of forest degradation innorthern Congo and southwestern Central AfricanRepublic between 2340 and 990 BPrsquo in The ecologysocial organization populations habitat and historyof the western lowland Gorilla PhD ThesisWashington Univ Saint Louis Missouri Ch 6

Fontaine B and Bigot S 1993 lsquoWest African rainfalldeficit and sea surface temperaturesrsquo InternationalJournal of Climatology Vol 13 271ndash85

Gautier L 1990 lsquoContact forecirct-savane en CocirctedrsquoIvoire centrale eacutevolution du recouvrementligneux des savanes de la Reacuteserve de Lamto (sud duV-Baouleacute)rsquo Candollea Vol 45 627ndash41

Giresse P Maley J and Brenac P 1994 lsquoLateQuaternary palaeoenvironments in the lake BarombiMbo (Cameroon) deduced from pollen and carbonisotopes of organic matterrsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 107 65ndash78

Goldammer JG and Seibert B 1990 lsquoThe impact ofdroughts and forest fires on tropical lowland rainforest of East Kalimatanrsquo in JG Goldammer (ed)Fire in the Tropical Biota Ecological Studies BerlinSpringer-Verlag 84

Goucher CL 1981 lsquoIron is iron til it is rust tradeand ecology in the decline of West African iron-smeltingrsquo Journal of African History Vol 22 179ndash89

Guille-Escuret G 1990 lsquoPalmier agrave huile vin depalme et transformations sociales en Lobaye (ForecirctCentrafricaine)rsquo Information sur les Sciences SocialesVol 29 327ndash53

Hart TB 2001 lsquoForest dynamics in the Ituri basin(DR Congo) Dominance diversity andconservationrsquo in W Weber L White A Vedderand L Naughton-Treves (eds) African rain forestecology and conservation An interdisciplinaryperspective Yale Yale University Press 154ndash64

Hart TB Hart JA Dechamps R Fournier M andAtaholo M 1996 lsquoChanges in forest compositionover the last 4000 years in the Ituri basin Zairersquo inLJG Van der Maesen and XM Van der Burgt (eds)The Biodiversity of African plants Dordrecht Kluwer545ndash63

Hawthorne WD 1991 Fire damage and forestregeneration in Ghana Ghana Forestry Departmentamp ODA Report

Jolly D Harrison SP Damnati B and BonnefilleR 1998 lsquoSimulated climate and Biomes of Africa

27

18

Figure 4 Schematic drawing of the status of Central African vegetation at the phase of maximumperturbation and destruction of forest which culminated around 2500 BPLarge hatched areas represent the residual areas of the forest domain around 2500 BP which probablyconsisted of forest-savanna mosaic with predominantly larger forest islands Around these residual areasthe land would have been colonised by open vegetation above all savannas but also open canopy forestformations (see (4) in Figure 1) The base of this map presents the current status of Central Africanbiotopes as in Figure 1

1998 Servant 1996) For south Cameroon alone(south of the Adamaoua Plateau) forest gainedabout a million hectares during this period(Letouzey 1985 Youta Happi and Bonvallot 1996Achoundong et al 2000) Historical and palyno-logical research in Cocircte drsquoIvoire Nigeria Cameroonand Congo indicate that this reinvasion hascontinued over several centuries (Fairhead andLeach 1998 Maley 1999 Vincens et al 2000)Albeit with fluctuations and interludes it appearsto be the continuation of the forest reinvasionwhich began about 2000 years BP Currentreinvasion is thus the long-term result of themassive forest disturbance which happened around2500 years BP

A brief review of research findings concerning oilpalm and Okoumeacute can illustrate the history ofthese forest ecosystems over recent millennia

3 The history of oil palm and ofOkoumeacute in Central Africa duringthe Holocene impact of the forestdisturbance culminating about2500 years BP

31 The oil palm

Oil palm (Elaeis guineensis) has abundant andcharacteristic pollen and thus appears frequentlyin sedimented pollen profiles This palm is ofAfrican origin with its pollen and macro-remains(nuts) found in Equatorial African deposits datingback to the Tertiary (Zeven 1967 Dechamps et al1992 Maley 1996 1999 Maley and Brenac 1998aMaley and Chepstow-Lusty 2001)

The extent to which oil palm has been planted orits regeneration encouraged through land use hasbeen debated Certainly when people clear youngpioneer forests or fallows near to villages where oilpalm is naturally abundant they generallyconserve the palms effectively enriching themthrough fallow cycles (eg in Cameroon CentralAfrican Republic Cocircte drsquoIvoire Togo and Beninsee Maley 1999) Thus in 1937 the renownedbotanist Aubreacuteville characterised the lsquoimmense oilpalm grovesrsquo of southern Benin as lsquoa naturalformation simply improved by the inhabitantsover the centuriesrsquo Zevenrsquos research largely fromNigeria also concludes that with the exception of

modern industrial plantations most oil palms inAfrica are semi-wild and are not propagated bysowing or transplanting seedlings (see Zeven1972)

Ecologically oil palm is a pioneer that requires lightto complete the first stages of its growth so itdevelops naturally in windfall-clearings andespecially on the periphery of dense forest after thepassage of fire which it can tolerate (Swaine andHall 1986 Swaine 1992) The botanist Letouzey(1978 1985) has described a vast natural oil palmgrove in west Cameroon and near the north-west ofthe forest zone (Figure 5) A band of forest 10ndash20km wide dominated by large and numerous oilpalms extends over more than 150 km Thisfollows the boundary between forest and savannaat times from 5 to 30 km inside the forest From theabsence of trees classically found in anthropicplantations and other criteria Letouzey (19781985) concluded that this grove is a natural standA similar example has been described on theeastern flank of Mount Nimba (GuineaLiberianborder) by Schnell (1946) who considered that thedissemination of palm nuts was facilitated by bothtoucans and chimpanzees whose faeces oftencontain it abundantly These natural groves haveattracted migrant people (eg Guille-Escuret 1990Maley 1999) For example many Bamileke peoplemoved to the natural groves of western Cameroonin the mid nineteenth century (Barbier 1981Warnier 1985 Perrois and Notue 1997 Maley1999) In the vast lsquoGrassfieldsrsquo just north of westCameroonrsquos forest bloc the forest islands are still ina phase of expansion especially due to the oil palmwhich is one of the principal pioneer trees Close tothe Donga river the Wuli people colonised thesenatural stands Baeke (1996) explains how theWuli install their villages in the palm groves whereassorted social rules oppose all plantation of palmsMoreover she reports that a Wuli myth of originlsquoprovides evidence for the prior use of oil palms tothe working of land and clearly distinguishes twotypes of plant exploitation gathering andagriculturersquo (Baeke 1996)

As already noted during the major disturbanceabout 2500 BP oil palm was a principal pioneerthat subsequently developed strongly (Figure 3b)In west Cameroon the rapid forest recolonisationwhich occurred from 2000 years ago in the

19

vicinity of Barombi Mbo (Maley and Brenac1998a) was accompanied by a retreat of oil palmprobably because ecological and climatic factorsrendered other pioneers more active at this time InFigure 3 it is notable that in west Cameroonbetween 2000 and 1000 years BP the pioneer treeMilicia was in a phase of very strong expansionThen here and elsewhere in central Africa a secondphase of oil palm expansion began around 1400years BP This culminated towards 1000 BP at lakeOssa near to Edea (Reynaud-Farrera et al 1996) inthe south of the Central African Republic atNouabale-Ndoki (Fay 1997 Maley 1999) but also

further south in Gabon to La Lopeacute (White et al2000 Maley unpublished) and in western Congonear to lake Kitina (Elenga et al 1996) This phaseending about 700800 BP must have beenassociated with renewed forest disturbance butweaker than that 2500 BP This disturbance wasalso associated with a short phase of soil erosionand with a discontinuity in the deposits observedin assorted locations in Central Africa (Maley andBrenac 1998b) An increase in mineral fluxes hasbeen found between 1200ndash800 years BP for thesites of Kitina and Sinnda in western Congo(Bertaux et al 2000) From the thirteenth centurya phase of forest recovery has occurred in CentralAfrica with a new development of mature forestsand a generalised retreat of oil palm a recoverydespite increased human settlements (Alexandre1965 Vansina 1990)

At the western extension of the Central Africanforests lies a large area without forest which bio-geographers call the Dahomey Gap stretchingacross southern Togo and Benin Much of this iscovered by an immense palm forest of naturalorigin as it was described by Aubreacuteville (1937)and Mondjannagni (1969) Some pollen data showthat this area was covered by forests during theearly and mid Holocene (Dupont and Weinelt1996) and that the opening-up of the landscapeoccurred rather abruptly between 4000 and 3500 BPfollowing a climatic change well-documented forthe neighbouring region of Ghana at lakeBosumtwi (Maley 1991 1997 1999) Moreover itappears that the break in the forest bloc wasoriginally much wider than it is now perhapsextending across almost all of south-west Nigeriaup to the Niger river Indeed from the beginning ofthe last millennium to the sixteenth century ADsome data show that this vast sector was notforested but becoming so (Barber 1985) and thatthe vegetation was dominated by oil palmappearing similar to currently existing vegetationacross southern Benin and Togo Data indicatingthis exist for an area near to the ancient city ofBenin (Okomu Forest about 100 km west of theNiger river) suggesting also that the forestextension occurred only in recent centuries (Jones1956 White and Oates 1999) Further researchwill be necessary to clarify these various points andtheir precise chronology

20

Figure 5 An lsquooil palm beltrsquo near the edge of therain forest in west Cameroon

Belt with a high density of Elaeis guineensis(oil palm) reaching 20 to 25 m height inassociation with a mature forest of the semi-deciduous type

In the Grass Fields extensive savannasnorth of the forest pioneer forest islands withlarge concentrations of oil palms In italics thenames of some important groups of people(Bamileke Bamoun Tikar Wuli)Source adapted from Letouzey 1978 and 1985

32 OkoumeacuteOkoumeacute (Aucoumea Klaineana Burseraceae) iscurrently restricted to the west and centre ofGabon with a limited extension in the south-westof Congo and another towards the north inEquatorial Guinea (Brunck et al 1990) (Figure 1)Like oil palm Okoumeacute is a light demandingpioneer species (which for some ecological reasonappears unable to live outside of central AfricaBrunk et al 1990 Maley 1990 Nasi 1997) In itsarea this tree is perfectly adapted to colonisepioneer forest fronts as they progress into savannasand often dominates them (Nasi 1997 White et al2000) It is abundant in fallows and along foresttracks As pioneer fronts mature and age othertrees of shade-tolerant species appear (egCaesalpiniaceae) and the Okoumeacute which surviveare the dispersed individuals whose crowns havealready reached the canopy living to 100ndash150years and achieving 12 m diameters (Nasi 1997)It can be deduced that the large Okoumeacute currentlypresent in the forest interior in its natural rangewere born in contact with savannas

The endemism of Okoumeacute with its range limitedmainly to Gabon is linked to its history and itsecology It is intolerant of water deficit and thus ofthe elevated temperatures of the long dry seasons(about 3 months from December to February) oftropical climates Within its Gabonese range waterdeficit is less severe than in neighbouring southernCameroon because the equatorial climatersquos dryseason (mostly from June to August) retains theelevated air humidity and moderate temperatures(Brunck et al 1990 Maley 1990)

Several works bring important precision to thehistory of Okoumeacute First palynological dataobtained in south Cameroon at lake Ossa near toEdea (Reynaud-Farrera et al 1996) show thatbetween 7500 and 3000 BP (middle Holocene) thedistribution of Okoumeacute extended much furthernorth reaching the lake about 170 km north of thetreersquos present range and perhaps further east andfurther south in western Congo Second moleculargenetics show two completely distinct varieties ofOkoumeacute one to the north of its range and anotherto the south (Muloko et al 1998 2000) Theboundary between the ranges runs approximatelyeast-west a little to the south of the equatortowards 0ucirc40rsquoSouth

It can be deduced that the current range of Okoumeacutehas resulted from the fragmentation of a largeancient range dating back to the middle Holocenefragmented by the generalised forest disturbanceabout 2500 BP into two very dispersed sub-groupsone to the south and the other to the north Theextension of open landscapes during the period ofdestruction would have been favourable toOkoumeacutersquos new phase of development in a similarway to that evidenced at Barombi Mbo for severalother pioneer trees (Figure 3b) The general forestrecolonisation dating from about 2000 BP was inan extremely fragmented environment withnumerous ecotones favourable to the developmentof Okoumeacute It is thus likely that the current rangeof Okoumeacute is about 2000 years old at most

A dynamic model which nearly follows thischronological canvas has been established by White(1995) This is based on examination of the currentvegetation of the northern part of the La Lopeacutereserve which is situated at the heart of the Gabonrsquosforest and still holds large patches of residualsavanna (see Aubreacuteville 1967) White (1995) showshow the forest reconstituted itself progressivelythanks to the wide bands of pioneer forest rich inOkoumeacute which surround these savannas Theprogression of pioneer forest fronts can be blockedby savanna fires often of anthropic origin (Oslislyand White 2000) Even so over the long term fromabout 2000 years ago the continuous general trendhas been in favour of reforestation

That the current and expanding range of Okoumeacutehas not extended much into south Cameroon ashad been the case in the middle Holocene isprobably explained by a shift in climatic conditionsbetween this older period and the recent (post-2000 BP) Holocene This deduction is importantas given the ecological particularities of Okoumeacutewhich link it tightly to the equatorial climate onecan further deduce that equatorial climates musthave reigned over south Cameroon during themiddle Holocene The present tropical climatethere thus dates from only about 2000 BP All thisappears to translate for the middle Holocene intoan increase in lsquosouthernrsquo equatorial influencestowards the north associated with a reduction inlsquonorthernrsquo boreal influences and possibly linked toa reduction in the length of the dry seasonassociated with the boreal winter

21

4 The question of savanna firesand the extension of the forestdisturbance in eastern Congo

Observations and opinions have frequentlydiverged concerning the influence of savanna fireson contemporary forest regeneration Observersvariously report that fire can either preventregeneration as in the area of savanna inliers inLopeacute Gabon (Oslisly and White 2000) or merelydelay it as in other parts of the forest region ThusLetouzey (1968) notes for the forest-savannatransition zone of south Cameroon that lsquofire doesnot necessarily prevent the formation of forestrecoveryrsquo Similar conclusions have been drawn inCentral Africa by Sillans and even in westernCongo (ex-Zaire) for the south of the forest bloc(Letouzey 1968) On the same lines Youta Happiand Bonvallot (1996) report lsquothat on theperiphery of a town of 76000 people such asBertoua (east Cameroon) many small savannainlets into forest of two to ten hectares havebecome covered by forest regrowth despite theannual bush fires practiced by Bororopastoralistsrsquo A little further south in the OdzalaReserve situated at the northern extremity of theBateke savannas (north Congo) Dowsett-Lemairereports that lsquoWithout doubt one is in a phase offorest extension despite very frequent bush firesrsquo(1996) This author notes lsquothe fire-break role ofthe boundary thanks to groves dominated byGaertnera paniculata a species absent at Lopeacutewhere colonising species are apparently moresensitive to firersquo Thus the floristic composition ofthe boundary plays an important role in thisphenomenon This was shown in south-eastCameroon by the mid-twentieth century invasionof Chromolaena odorata which colonised theboundaries and which acts there at least as anefficacious fire break It also protects seeds ofwoody species which can germinate and developunder its cover (Youta Happi et al 2000) Otherobservations show how traditional agriculturalpractices and the movement of cattle can befavourable to forest recolonisation in particularby eliminating the grasses which facilitate thepropagation and intensity of fires These processeshave been well described in southern Cameroon(Letouzey 1968 1985) and in Cocircte drsquoIvoire(Spichiger and Blanc-Pamard 1973 Blanc-Pamardand Peltre 1984 Gautier 1990)

These observations indicate that in general andespecially over the long term forest colonisation atthe forest-savanna contact has been the broadlydominant process In certain exceptional yearswhen the dry season has been one or two monthslonger the savanna fires have penetrated moredeeply into the neighbouring forests In early 1983during a two-month longer dry season numerousobservers reported the strong invasion of fire intoforest in Cocircte drsquoIvoire (Bertault 1990) Ghana(Hawthorne 1991 Swaine 1992) and Cameroon(Amougou 1986) Equally fires in Indonesia thatyear were linked to an exceptional El Nintildeo event(Goldammer and Seibert 1990)

These observations indicate how the penetration offire into forest occurs only in very dry years Onemight thus think that fires could have played amajor role in the forest destruction 2500 BP Thishas indeed been shown in eastern BrazilianAmazonia during its major disturbance in themiddle Holocene That fires were frequent then isshown by the numerous charcoal fragmentsdeposited in the lake sediments at Carajas (Martinet al 1993 Servant et al 1993) Yet the lateHolocene lake deposits in Central Africa hardlycontain any charcoal fragments especially for theperiod 3000ndash2000 BP (Giresse et al 1994 Maleyand Brenac 1998a Bertaux et al 2000) It is thesame for the upper horizons of the numerous soilprofiles observed in this region (eg Valleacuterie 1973)In the thalwegs the coarse deposits in the lowerpart of the Lower Terrace do at times containcharcoal or fragments of fossil wood (Maley andBrenac 1998b) however Iron age archaeologicalsites dating from the beginning of the lateHolocene such as in the savanna areas of La Lopeacutealso present wood charcoal (Oslisly and White2000) One can conclude provisionally that thefires that existed were of limited extent

In particular charcoal associated with ironsmelting furnaces cannot have caused the forestdisturbance and cannot have been responsible forthe savanna extension about 2500 BP Goucher(1981) had advanced the hypothesis that ironsmelting could have been a major cause ofdeforestation in West Africa but Fairhead andLeach (1998) have demonstrated that thishypothesis had no serious foundation For CentralAfrica Pinccedilon (1990) also concludes that the

22

extraction of wood for iron metallurgy on theBateke Plateaux was minimal and did not explain(as had been suggested) the presence of theextensive savannas there Instead these savannasresult from an elevated edaphic drought caused byvery sandy soils in which rain infiltrates rapidly toa great depth (Laraque and Pandi 1996)

In contrast to observations from Atlantic CentralAfrica (south Cameroon Gabon and Congo) woodcharcoal has been found in abundance in Ituri nearto Epulu (c 1ucirc20rsquoNndash28ucirc35rsquoE) Here in the heart ofthe forests of the north-east Congo basin about 180km from the nearest savannas 416 soil profileshave been examined Almost all of them containnumerous wood charcoal in their upper levelsbetween the surface and 50 cm deep (Hart et al1996) Virtually all of the 1817 samples of woodcharcoal identified by Deschamps belong to treesfound in the regionrsquos diverse forest formationsexcept for a few pieces of charcoal from twoprofiles which belong to typical wooded savannaspecies The 28 dates obtained for the charcoalshow that fires occurred almost exclusively duringthe last three millennia with a phase of woodedsavanna expansion dating to about 2200 BP (Hartet al 1996) Moreover about 300 km south ofEpulu and 100 km from the forestsavanna limit aroad cutting near Osokari (1ucirc16rsquoSndash27ucirc48rsquoE)revealed a remarkable profile above the principalstone line that could be dated to about 11500years BP (Runge 1996 Maley 1996) A secondgravel level obliquely cutting the top of the soilprofile contains wood charcoal dating from about2200 to 1850 BP These results show that the forestdisturbance in Atlantic Central Africa extended toIturi and affected a large part of the eastern regionof the Congo river basin In contrast with regionsnearer the Atlantic fire had a large role

Also at Ituri (Epulu) Hart et al (1996) show thatcertain trees which are now found there are notfound in the wood charcoal in particularGilbertiodendron dewevrei (Caesalpiniaceae) which isnow abundant (almost mono-dominant) in certainparts of the forest The range of forests dominatedby Gilbertiodendron dewevrei extends to the east andnorth of the Congo basin as far as east Cameroon(Leacuteonard 1953) It is thus possible that thisCaesalpiniaceae has very recently colonised the Ituriforests and that the large area it now occupies is a

legacy of the disturbance about 2500 BP Ifconfirmed this area could correspond at leastpartly to an older forest area (of unknowncharacter) which was destroyed by the disturbanceA comparison with Okoumeacute can be suggestedwhich as indicated for Gabon above alsodominates areas which it has colonised since lessthan 2000 years ago However as the forest agesOkoumeacute loses its dominant character whereas inIturi Gilbertiodendron dewevrei remains dominantprobably because it can regenerate abundantly inthe shade of parent trees and also because itsreproductive character gives it an advantage overother trees in this formation (see Hart 2001)

5 Conclusion the major role ofclimatic conditions and theirhistoryThe present mosaic pattern of numerous forestscharacterised by a mix or juxtaposition ofgroupings of evergreen and deciduous species isprobably the consequence of long-term disturbanceswhich have affected the forest domain during thelast three millennia particularly the majordisturbance which culminated about 2500 years BPThe synchronism apparent in the disturbance fordifferent sites studied across Central Africa (southCameroon Gabon Congo as far east as the CongoRiver basin) and its association with a generalisederosive phase enables one to conclude that it wasthe result of a major climatic change

The climate at this time appears to have beenrelatively arid as it led to forest destruction and inplaces the extension of savannas Neverthelessother characteristics such as the strongsynchronous extension of diverse pioneer taxaindicate that in places there was not truly areduction in rainfall but rather a change in itsannual distribution The absence of a fall in annualrainfall has been demonstrated at lake BarombiMbo (Maley and Brenac 1998a) and at lake Ossawhere diatom studies show that there was even anet rise in this lake between 2500 and 2200 yearsBP with only a short fall between 2200 and 2000years BP (Nguetsop et al 2000) It is thereforepreferable to designate this catastrophic period asan lsquounfavourable climatic disturbancersquo caused by anincrease in seasonality and a lengthening of the dryseason (Maley 1997) The strong soil erosion of this

23

period suggests heavy rains but concentrated oversix or seven months of the year as is now found inthe peripheral savanna zones The growth inseasonality could have been associated with adominance of lsquosquall linesrsquo the typical cloudformation of savanna zones formed by northsouthalignments of cumuliform (convective storm-type)clouds They surplanted other cloud typesprincipally the stratiform types which giverelatively fine regular monsoon-type rains (seeMaley 1982) That these monsoon rains wouldhave dominated the earlier period (between 4000and 30002800 BP Maley 1997) is shown by therelative development of Caesalpinaceae in theforests of low altitude and in the mountains by theextension of Podocarpus which are trees typical ofcloud forests with stratiform clouds (Kerfoot1968 Maley 1996 1997 Maley and Brenac1998a)

Research in dynamic tropical climatology over thelast 20 years shows that Sea Surface Temperature(SST) on the regional and global scale stronglyinfluences climate (Fontaine and Bigot 1993

Moron et al 1995 Bigot et al 1997) It is possibleto characterise the relations between the monsoonrains and the SST more precisely Studies ofclimatic anomalies over tropical Africa occurringsince the beginning of the 1960s have evidencedtwo dominant modes of rainfall distribution eachassociated with a particular distribution of SST Thefirst mode is characterised on the one hand bywarmer than average temperatures in the Gulf ofGuinea and the southern Atlantic and on the otherhand by cooler temperatures on the north tropicalAtlantic offshore of West Africa The second modepresents an opposed distribution for the Africancontinent and the ocean A north-south tropicalAtlantic lsquodipolersquo has thus been found whichoscillates between one mode and the other(Fontaine and Bigot 1993 Wotling et al 1995Bigot et al 1997) An important research result hasbeen to show that over recent millennia the spatialdistribution of paleo-climatic anomalies overtropical Africa (more humid regions and more dryones Figure 6) and of SST on the neighbouringAtlantic is very similar to the spatial distributionobserved for the two dominant anomalies over thelast four decades (Maley 1997 Maley et al 2000)One can deduce that the climatic mechanisms arethe same at the timescale of years centuries ormillennia

So the variation of SST over the tropical Atlanticappears to play a major role in the variation ofclimate and rain types over Central Africa Thephase of forest disturbance that culminated about2500 years BP was associated with relatively lsquowarmrsquoSST over the Gulf of Guinea The earlier phasebetween about 40003800 and 2800 years BP wasassociated with relatively lsquocoolrsquo SST sharplyreduced from the early and middle Holocene (seeMorley and Dworetzky 1993 Figure 7) Over westand south Cameroon these SST allowed relativelyhigh rains and the evolution of stratiform cloudstowards rainy nimbostratus types The palaeo-vegetation data are presently lacking for GabonHowever new research is underway within thePALEOFORGA (Paleoenvironments of GabonForests) programme studying lakes MaridorNguegravene and Kamaleacuteteacute (Figure 1) In contrastfurther south in western Congo the periodstarting from 5000 years BP (Bertaux et al 2000)was marked by the development of semi-deciduousforests (Vincens et al 2000) linked to lower

24

Figure 6 Major rainfall trends in tropical Africa andSouth America for two opposed climatic periods c 3700ndash3000 BP and c 28002500ndash2000 BPPrecipitation trends (anomalies) abovenormalmean + below normalmean ndash P pointsto the sectors where a maximum of Podocarpuspollen intervenedSource Maley 1997 Maley et al 2000

rainfalls indicating that the stratiform clouds hadprobably evolved towards non-rainy clouds Thefour-month dry season here is characterised by thequasi-permanence of these cloud types (Saint Vil1984) which probably gained place progressivelybetween 5000 and 4000 years BP There wastherefore a reinforcing of southern influencesupporting conclusions concerning the history ofOkoumeacute It is notable that the vegetational impactsof the lsquowarmrsquo SST c 28002500 to 2000 years BPwere very different from those earlier c 9000 to65006000 years BP a difference which reflects abrutal strengthening over Central Africa of borealinfluences (Nguetsop et al 2000) to the detrimentof southern influences from about 28002500years BP

Before the major Holocene forest disturbanceculminating about 2500 BP other importantdisturbances occurred during the Quaternary(Maley 1996) The best documented is that betweenabout 20 000 and 15 000 years BP (Maley 19871996 Maley and Brenac 1998a Figure 2) In bothcases residual forest environments (refugia)

persisted apparently in similar locations butperhaps less as small blocks of relativelyhomogeneous forest than as a landscape of forest-savanna mosaic in which forest islands dominated(see Leal 2000) Yet two important characteristicsclearly differentiate these two phases of forestfragmentation First the older disturbance wasmuch longer Second they occurred under verydifferent (even opposed) general climatic conditionsThe disturbance between 20 000 and 15 000 BPcoincided with the development of glacialconditions in middle and high latitudes and coolerconditions by several degrees also affected the wholeregion of the Gulf of Guinea (Maley 1996 Maley andBrenac 1998a) The most recent disturbance bycontrast occurred during the Holocene interglacialwhich on a global scale is characterised by reducedglacial extension and relatively warmertemperatures The period between 2500 and 2000BP is associated with slightly warmer climaticconditions on the regional and global scale (Maley1997 Maley et al 2000) An important conclusion isthat the retreat and fragmentation of the Africanforests can be produced under very different

25

Figure 7 Variations in Sea Surface Temperature (SST) in the South Atlantic Ocean Benguela sector(22ucirc20rsquoSndash11ucirc12rsquoE) estimated with transfer functions based on the abundance of Radiolarian species in thecore RC13ndash228The open circles designate the 6000 and 9000 BP interpolated dates for the two curves one correspondingto the southern summer (February) and the other to the southern winter (August) modern SST areindicated by the solid squares above each curve The dashed lines represent some interpolated dates(italics) 18 000 BP based on the oxygen-isotope stratigraphy 38004000 BP 3000 BP and 2000 BP Theselate Holocene dates are confirmed by an independent study of shells from the same sector (Cohen et al1992) The coherence of the main SST anomalies throughout the Guinea Gulf permits the use of thesecurves in order to interpret the paleoclimatology of central Africa (Maley 1997)Source Morley and Dworetzky 1993

climates either relatively cool or relatively warmand thus with very different climatic situations(Maley 1996 1997)

Climatic models concerning lsquoglobal warmingrsquo inthe context of contemporary climate changeconcerns indicate that an average growth intemperature of about 4ucircC will lead also to anincrease in evaporation of about 30 per cent butonly 12 per cent more rain for tropical Africa (Rind

Notes The author is grateful to James Fairhead for his

translation of this article from the original Frenchand to Melissa Leach and James Fairhead for theirinvitation to the IDS workshop lsquoChangingPerspectives on Forestsrsquo in March 2001 This isInstitut des Sciences de lrsquoEvolution de Montpellier(ISEMCNRS) contribution No 125ndash2001

26

ReferencesAchoundong G Youta Happi J Guillet B

Bonvallot J and Kamgang Beyala V 2000lsquoFormation et eacutevolution des recrus sur savanes (ausud Cameroun)rsquo in M Servant and S Servant-Vildary (eds) Dynamique agrave long terme desEcosystegravemes forestiers intertropicaux MeacutemoireUNESCO Paris UNESCO 31ndash41

Alexandre P 1965 lsquoProto-histoire du groupe Beacuteti-Boulou-Fang essai de synthegravese provisoirersquo CahiersdrsquoEtudes Africaines Vol 20 503ndash60

Amougou Akoa 1986 lsquoEtude botanique eteacutecologique de la valleacutee inondable du Haut-Nyong(sud Cameroun) et de ses affluentsrsquo Thegravese SciencesUniv Yaoundeacute

Aubreacuteville A 1937 lsquoLes forecircts du Dahomey et duTogorsquo Bulletin du Comiteacute drsquoEtudes Historiques etScientifiques de lrsquoAfrique Occidentale Franccedilaises Vol20 112

Aubreacuteville A 1967 lsquoLes eacutetranges mosaiques forecirct-savane du sommet de la boucle de lrsquoOgooueacute auGabonrsquo Adansonia Vol 7 13ndash22

Baeke V 1996 lsquoLe temps des Rites Lrsquounivers magico-religieux des Wuli (Mfumte du Camerounoccidental)rsquo Thegravese Sc Sociale Univ Libre Bruxelles

Barber RJ 1985 lsquoLand snails and past environmentat the Igbo-Iwoto Esie site southwestern NigeriarsquoWest African Journal of Archaeology Vol 15 89ndash102

Barbier JC 1981 lsquoLe peuplement de la partiemeacuteridionale du plateau Bamileacutekeacutersquo in C Tardits

(ed) Contribution de la recherche ethnologique agravelrsquohistoire des civilisations du Cameroun ParisColloque International Centre National deRecherche Scientifique Paris Vol 2 331ndash53

Bertault JG 1990 lsquoComparaison drsquoeacutecosystegravemesforestiers naturels et modifieacutes apregraves incendie enCocircte drsquoIvoirersquo in H Puig (ed) Atelier surlrsquoAmeacutenagement et la Conservation de lrsquoEcosystegravemeForestier Tropical Humide Cayenne ParisUNESCO 6

Bertaux J Schwartz D Vincens A Sifeddine AElenga H Mansour M Mariotti A Fournier MMartin L Wirrmann D and Servant M 2000lsquoEnregistrement de la phase segraveche drsquoAfriqueCentrale vers 3000 ans BP par la spectromeacutetrie IRdans les lacs Sinnda et Kitina (sud-Congo)rsquo in MServant S Servant-Vildary (eds) Dynamique agrave longterme des Ecosystegravemes forestiers intertropicauxMeacutemoire UNESCO Paris UNESCO 43ndash9

Bigot S Camberlin P Moron V and Richard Y1997 lsquoStructures spatiales de la variabiliteacute despreacutecipitations en Afrique une transition climatiqueagrave la fin des anneacutees 1960rsquo Compte Rendu delrsquoAcademie des Sciences Paris 2a 324 181ndash88

Blanc-Pamard C and Peltre P 1984 lsquoDynamique despaysages preacuteforestiers et pratiques culturales enAfrique de lrsquoOuest (Cocircte drsquoIvoire centrale)rsquo LeDeacuteveloppement rural en question Paris MeacutemoireORSTOM No 106 55ndash74

Bonnefille R Chalieacute F Guiot J and Vincens A1992 lsquoQuantitative estimates of full glacial

1995) Though the causes of warming experiencedabout 2500 years BP were natural (Magny 1993)and therefore different from the warmingenvisaged for the twenty-first century it is possiblethat the catastrophic destruction of Central Africanforests which culminated about 2500 years BPcould be an lsquoanaloguersquo and an alarm signal forpotential developments during this current andfuture phase of human-induced global warming(Maley 1997 Maley et al 2000)

1 BP lsquoBefore the Presentrsquo according to standardisedC14 analysis

2 Only 4ndash5 samples have been studied for each 1000years during the Holocene To detect century-scaleperiodicity it would be necessary to study at least 10samples per millennium

temperatures in equatorial Africa frompalynological datarsquo Climate Dynamics Vol 6251ndash57

Brunck F Grison F and Maitre HF 1990Lrsquookoumeacute Aucoumea klaineana Pierre MonographieCentre Techn Forestier Tropical Nogent CIRAD 102

Cohen AI Perkington JE Brundrirt GB and vanDer Merwe NJ 1992 lsquoA Holocene marine climaterecord in Mollusc shells from the southwest Africancoastrsquo Quaternary Research Vol 38 379ndash85

Dechamps R Senut B and Pickford M 1992 Fruitsfossiles pliocegravenes et pleacuteistocegravenes du Rift occidentalougandais Signification paleacuteoenvironnementaleCompte Rendu de lrsquoAcadeacutemie des Sciences Parisseacuterie 2 Vol 314 325ndash331

Dowsett-Lemaire F 1996 lsquoComposition et eacutevolutionde la veacutegeacutetation forestiegravere au Parc NationaldrsquoOdzala Congorsquo Bulletin du Jardin BotaniqueNational de Belgique Vol 65 253ndash92

Dupont LM and Weinelt M 1996 lsquoVegetationhistory of the savanna corridor between theguinean and the Congolian rain forest during thelast 150000 yearsrsquo Vegetation History andArchaeobotany Vol 5 273ndash92

Elenga H Schwartz D and Vincens A 1994lsquoPollen evidence of late Quaternary vegetation andinferred climate changes in Congorsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 109345ndash56

Elenga H Schwartz D Vincens A Bertaux J DeNamur C Martin L Wirrmann D and ServantM 1996 lsquoDiagramme pollinique holocegravene du lacKitina (Congo) mise en eacutevidence de changementspaleacuteobotaniques et paleacuteoclimatiques dans le massifforestier du Mayombersquo Compte Rendu de lrsquoAcadeacutemiedes Sciences Paris Vol 323 No 2a 403ndash10

Elenga H Peyron O Bonnefille R Prentice ICJolly D Cheddadi R Guiot J Andrieu V DeBeaulieu JL Hamilton AC Maley J MarchantR Reille M Scott L Taylor D and Van CampoE 2000 lsquoPollen-based biome reconstruction forsouthern Europe and Africa 18000 years agorsquoJournal of Biogeography Vol 27 621ndash34

Fairhead J and Leach M 1995 lsquoFalse forest historycomplicit social analysis rethinking some westAfrican environmental narrativesrsquo WorldDevelopment Vol 23 1023ndash35

Fairhead J and Leach M 1996 lsquoEnriching thelandscape social history and the management oftransition ecology in the forest-savanna mosaic ofthe Republic of Guinearsquo Africa Vol 66 14ndash36

Fairhead J and Leach M 1998 lsquoReframingdeforestation Global analyses and local realitiesstudies in West Africarsquo Global EnvironmentalChanges Series London Routledge 238

Fay JM 1997 lsquoEvidence of forest degradation innorthern Congo and southwestern Central AfricanRepublic between 2340 and 990 BPrsquo in The ecologysocial organization populations habitat and historyof the western lowland Gorilla PhD ThesisWashington Univ Saint Louis Missouri Ch 6

Fontaine B and Bigot S 1993 lsquoWest African rainfalldeficit and sea surface temperaturesrsquo InternationalJournal of Climatology Vol 13 271ndash85

Gautier L 1990 lsquoContact forecirct-savane en CocirctedrsquoIvoire centrale eacutevolution du recouvrementligneux des savanes de la Reacuteserve de Lamto (sud duV-Baouleacute)rsquo Candollea Vol 45 627ndash41

Giresse P Maley J and Brenac P 1994 lsquoLateQuaternary palaeoenvironments in the lake BarombiMbo (Cameroon) deduced from pollen and carbonisotopes of organic matterrsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 107 65ndash78

Goldammer JG and Seibert B 1990 lsquoThe impact ofdroughts and forest fires on tropical lowland rainforest of East Kalimatanrsquo in JG Goldammer (ed)Fire in the Tropical Biota Ecological Studies BerlinSpringer-Verlag 84

Goucher CL 1981 lsquoIron is iron til it is rust tradeand ecology in the decline of West African iron-smeltingrsquo Journal of African History Vol 22 179ndash89

Guille-Escuret G 1990 lsquoPalmier agrave huile vin depalme et transformations sociales en Lobaye (ForecirctCentrafricaine)rsquo Information sur les Sciences SocialesVol 29 327ndash53

Hart TB 2001 lsquoForest dynamics in the Ituri basin(DR Congo) Dominance diversity andconservationrsquo in W Weber L White A Vedderand L Naughton-Treves (eds) African rain forestecology and conservation An interdisciplinaryperspective Yale Yale University Press 154ndash64

Hart TB Hart JA Dechamps R Fournier M andAtaholo M 1996 lsquoChanges in forest compositionover the last 4000 years in the Ituri basin Zairersquo inLJG Van der Maesen and XM Van der Burgt (eds)The Biodiversity of African plants Dordrecht Kluwer545ndash63

Hawthorne WD 1991 Fire damage and forestregeneration in Ghana Ghana Forestry Departmentamp ODA Report

Jolly D Harrison SP Damnati B and BonnefilleR 1998 lsquoSimulated climate and Biomes of Africa

27

1998 Servant 1996) For south Cameroon alone(south of the Adamaoua Plateau) forest gainedabout a million hectares during this period(Letouzey 1985 Youta Happi and Bonvallot 1996Achoundong et al 2000) Historical and palyno-logical research in Cocircte drsquoIvoire Nigeria Cameroonand Congo indicate that this reinvasion hascontinued over several centuries (Fairhead andLeach 1998 Maley 1999 Vincens et al 2000)Albeit with fluctuations and interludes it appearsto be the continuation of the forest reinvasionwhich began about 2000 years BP Currentreinvasion is thus the long-term result of themassive forest disturbance which happened around2500 years BP

A brief review of research findings concerning oilpalm and Okoumeacute can illustrate the history ofthese forest ecosystems over recent millennia

3 The history of oil palm and ofOkoumeacute in Central Africa duringthe Holocene impact of the forestdisturbance culminating about2500 years BP

31 The oil palm

Oil palm (Elaeis guineensis) has abundant andcharacteristic pollen and thus appears frequentlyin sedimented pollen profiles This palm is ofAfrican origin with its pollen and macro-remains(nuts) found in Equatorial African deposits datingback to the Tertiary (Zeven 1967 Dechamps et al1992 Maley 1996 1999 Maley and Brenac 1998aMaley and Chepstow-Lusty 2001)

The extent to which oil palm has been planted orits regeneration encouraged through land use hasbeen debated Certainly when people clear youngpioneer forests or fallows near to villages where oilpalm is naturally abundant they generallyconserve the palms effectively enriching themthrough fallow cycles (eg in Cameroon CentralAfrican Republic Cocircte drsquoIvoire Togo and Beninsee Maley 1999) Thus in 1937 the renownedbotanist Aubreacuteville characterised the lsquoimmense oilpalm grovesrsquo of southern Benin as lsquoa naturalformation simply improved by the inhabitantsover the centuriesrsquo Zevenrsquos research largely fromNigeria also concludes that with the exception of

modern industrial plantations most oil palms inAfrica are semi-wild and are not propagated bysowing or transplanting seedlings (see Zeven1972)

Ecologically oil palm is a pioneer that requires lightto complete the first stages of its growth so itdevelops naturally in windfall-clearings andespecially on the periphery of dense forest after thepassage of fire which it can tolerate (Swaine andHall 1986 Swaine 1992) The botanist Letouzey(1978 1985) has described a vast natural oil palmgrove in west Cameroon and near the north-west ofthe forest zone (Figure 5) A band of forest 10ndash20km wide dominated by large and numerous oilpalms extends over more than 150 km Thisfollows the boundary between forest and savannaat times from 5 to 30 km inside the forest From theabsence of trees classically found in anthropicplantations and other criteria Letouzey (19781985) concluded that this grove is a natural standA similar example has been described on theeastern flank of Mount Nimba (GuineaLiberianborder) by Schnell (1946) who considered that thedissemination of palm nuts was facilitated by bothtoucans and chimpanzees whose faeces oftencontain it abundantly These natural groves haveattracted migrant people (eg Guille-Escuret 1990Maley 1999) For example many Bamileke peoplemoved to the natural groves of western Cameroonin the mid nineteenth century (Barbier 1981Warnier 1985 Perrois and Notue 1997 Maley1999) In the vast lsquoGrassfieldsrsquo just north of westCameroonrsquos forest bloc the forest islands are still ina phase of expansion especially due to the oil palmwhich is one of the principal pioneer trees Close tothe Donga river the Wuli people colonised thesenatural stands Baeke (1996) explains how theWuli install their villages in the palm groves whereassorted social rules oppose all plantation of palmsMoreover she reports that a Wuli myth of originlsquoprovides evidence for the prior use of oil palms tothe working of land and clearly distinguishes twotypes of plant exploitation gathering andagriculturersquo (Baeke 1996)

As already noted during the major disturbanceabout 2500 BP oil palm was a principal pioneerthat subsequently developed strongly (Figure 3b)In west Cameroon the rapid forest recolonisationwhich occurred from 2000 years ago in the

19

vicinity of Barombi Mbo (Maley and Brenac1998a) was accompanied by a retreat of oil palmprobably because ecological and climatic factorsrendered other pioneers more active at this time InFigure 3 it is notable that in west Cameroonbetween 2000 and 1000 years BP the pioneer treeMilicia was in a phase of very strong expansionThen here and elsewhere in central Africa a secondphase of oil palm expansion began around 1400years BP This culminated towards 1000 BP at lakeOssa near to Edea (Reynaud-Farrera et al 1996) inthe south of the Central African Republic atNouabale-Ndoki (Fay 1997 Maley 1999) but also

further south in Gabon to La Lopeacute (White et al2000 Maley unpublished) and in western Congonear to lake Kitina (Elenga et al 1996) This phaseending about 700800 BP must have beenassociated with renewed forest disturbance butweaker than that 2500 BP This disturbance wasalso associated with a short phase of soil erosionand with a discontinuity in the deposits observedin assorted locations in Central Africa (Maley andBrenac 1998b) An increase in mineral fluxes hasbeen found between 1200ndash800 years BP for thesites of Kitina and Sinnda in western Congo(Bertaux et al 2000) From the thirteenth centurya phase of forest recovery has occurred in CentralAfrica with a new development of mature forestsand a generalised retreat of oil palm a recoverydespite increased human settlements (Alexandre1965 Vansina 1990)

At the western extension of the Central Africanforests lies a large area without forest which bio-geographers call the Dahomey Gap stretchingacross southern Togo and Benin Much of this iscovered by an immense palm forest of naturalorigin as it was described by Aubreacuteville (1937)and Mondjannagni (1969) Some pollen data showthat this area was covered by forests during theearly and mid Holocene (Dupont and Weinelt1996) and that the opening-up of the landscapeoccurred rather abruptly between 4000 and 3500 BPfollowing a climatic change well-documented forthe neighbouring region of Ghana at lakeBosumtwi (Maley 1991 1997 1999) Moreover itappears that the break in the forest bloc wasoriginally much wider than it is now perhapsextending across almost all of south-west Nigeriaup to the Niger river Indeed from the beginning ofthe last millennium to the sixteenth century ADsome data show that this vast sector was notforested but becoming so (Barber 1985) and thatthe vegetation was dominated by oil palmappearing similar to currently existing vegetationacross southern Benin and Togo Data indicatingthis exist for an area near to the ancient city ofBenin (Okomu Forest about 100 km west of theNiger river) suggesting also that the forestextension occurred only in recent centuries (Jones1956 White and Oates 1999) Further researchwill be necessary to clarify these various points andtheir precise chronology

20

Figure 5 An lsquooil palm beltrsquo near the edge of therain forest in west Cameroon

Belt with a high density of Elaeis guineensis(oil palm) reaching 20 to 25 m height inassociation with a mature forest of the semi-deciduous type

In the Grass Fields extensive savannasnorth of the forest pioneer forest islands withlarge concentrations of oil palms In italics thenames of some important groups of people(Bamileke Bamoun Tikar Wuli)Source adapted from Letouzey 1978 and 1985

32 OkoumeacuteOkoumeacute (Aucoumea Klaineana Burseraceae) iscurrently restricted to the west and centre ofGabon with a limited extension in the south-westof Congo and another towards the north inEquatorial Guinea (Brunck et al 1990) (Figure 1)Like oil palm Okoumeacute is a light demandingpioneer species (which for some ecological reasonappears unable to live outside of central AfricaBrunk et al 1990 Maley 1990 Nasi 1997) In itsarea this tree is perfectly adapted to colonisepioneer forest fronts as they progress into savannasand often dominates them (Nasi 1997 White et al2000) It is abundant in fallows and along foresttracks As pioneer fronts mature and age othertrees of shade-tolerant species appear (egCaesalpiniaceae) and the Okoumeacute which surviveare the dispersed individuals whose crowns havealready reached the canopy living to 100ndash150years and achieving 12 m diameters (Nasi 1997)It can be deduced that the large Okoumeacute currentlypresent in the forest interior in its natural rangewere born in contact with savannas

The endemism of Okoumeacute with its range limitedmainly to Gabon is linked to its history and itsecology It is intolerant of water deficit and thus ofthe elevated temperatures of the long dry seasons(about 3 months from December to February) oftropical climates Within its Gabonese range waterdeficit is less severe than in neighbouring southernCameroon because the equatorial climatersquos dryseason (mostly from June to August) retains theelevated air humidity and moderate temperatures(Brunck et al 1990 Maley 1990)

Several works bring important precision to thehistory of Okoumeacute First palynological dataobtained in south Cameroon at lake Ossa near toEdea (Reynaud-Farrera et al 1996) show thatbetween 7500 and 3000 BP (middle Holocene) thedistribution of Okoumeacute extended much furthernorth reaching the lake about 170 km north of thetreersquos present range and perhaps further east andfurther south in western Congo Second moleculargenetics show two completely distinct varieties ofOkoumeacute one to the north of its range and anotherto the south (Muloko et al 1998 2000) Theboundary between the ranges runs approximatelyeast-west a little to the south of the equatortowards 0ucirc40rsquoSouth

It can be deduced that the current range of Okoumeacutehas resulted from the fragmentation of a largeancient range dating back to the middle Holocenefragmented by the generalised forest disturbanceabout 2500 BP into two very dispersed sub-groupsone to the south and the other to the north Theextension of open landscapes during the period ofdestruction would have been favourable toOkoumeacutersquos new phase of development in a similarway to that evidenced at Barombi Mbo for severalother pioneer trees (Figure 3b) The general forestrecolonisation dating from about 2000 BP was inan extremely fragmented environment withnumerous ecotones favourable to the developmentof Okoumeacute It is thus likely that the current rangeof Okoumeacute is about 2000 years old at most

A dynamic model which nearly follows thischronological canvas has been established by White(1995) This is based on examination of the currentvegetation of the northern part of the La Lopeacutereserve which is situated at the heart of the Gabonrsquosforest and still holds large patches of residualsavanna (see Aubreacuteville 1967) White (1995) showshow the forest reconstituted itself progressivelythanks to the wide bands of pioneer forest rich inOkoumeacute which surround these savannas Theprogression of pioneer forest fronts can be blockedby savanna fires often of anthropic origin (Oslislyand White 2000) Even so over the long term fromabout 2000 years ago the continuous general trendhas been in favour of reforestation

That the current and expanding range of Okoumeacutehas not extended much into south Cameroon ashad been the case in the middle Holocene isprobably explained by a shift in climatic conditionsbetween this older period and the recent (post-2000 BP) Holocene This deduction is importantas given the ecological particularities of Okoumeacutewhich link it tightly to the equatorial climate onecan further deduce that equatorial climates musthave reigned over south Cameroon during themiddle Holocene The present tropical climatethere thus dates from only about 2000 BP All thisappears to translate for the middle Holocene intoan increase in lsquosouthernrsquo equatorial influencestowards the north associated with a reduction inlsquonorthernrsquo boreal influences and possibly linked toa reduction in the length of the dry seasonassociated with the boreal winter

21

4 The question of savanna firesand the extension of the forestdisturbance in eastern Congo

Observations and opinions have frequentlydiverged concerning the influence of savanna fireson contemporary forest regeneration Observersvariously report that fire can either preventregeneration as in the area of savanna inliers inLopeacute Gabon (Oslisly and White 2000) or merelydelay it as in other parts of the forest region ThusLetouzey (1968) notes for the forest-savannatransition zone of south Cameroon that lsquofire doesnot necessarily prevent the formation of forestrecoveryrsquo Similar conclusions have been drawn inCentral Africa by Sillans and even in westernCongo (ex-Zaire) for the south of the forest bloc(Letouzey 1968) On the same lines Youta Happiand Bonvallot (1996) report lsquothat on theperiphery of a town of 76000 people such asBertoua (east Cameroon) many small savannainlets into forest of two to ten hectares havebecome covered by forest regrowth despite theannual bush fires practiced by Bororopastoralistsrsquo A little further south in the OdzalaReserve situated at the northern extremity of theBateke savannas (north Congo) Dowsett-Lemairereports that lsquoWithout doubt one is in a phase offorest extension despite very frequent bush firesrsquo(1996) This author notes lsquothe fire-break role ofthe boundary thanks to groves dominated byGaertnera paniculata a species absent at Lopeacutewhere colonising species are apparently moresensitive to firersquo Thus the floristic composition ofthe boundary plays an important role in thisphenomenon This was shown in south-eastCameroon by the mid-twentieth century invasionof Chromolaena odorata which colonised theboundaries and which acts there at least as anefficacious fire break It also protects seeds ofwoody species which can germinate and developunder its cover (Youta Happi et al 2000) Otherobservations show how traditional agriculturalpractices and the movement of cattle can befavourable to forest recolonisation in particularby eliminating the grasses which facilitate thepropagation and intensity of fires These processeshave been well described in southern Cameroon(Letouzey 1968 1985) and in Cocircte drsquoIvoire(Spichiger and Blanc-Pamard 1973 Blanc-Pamardand Peltre 1984 Gautier 1990)

These observations indicate that in general andespecially over the long term forest colonisation atthe forest-savanna contact has been the broadlydominant process In certain exceptional yearswhen the dry season has been one or two monthslonger the savanna fires have penetrated moredeeply into the neighbouring forests In early 1983during a two-month longer dry season numerousobservers reported the strong invasion of fire intoforest in Cocircte drsquoIvoire (Bertault 1990) Ghana(Hawthorne 1991 Swaine 1992) and Cameroon(Amougou 1986) Equally fires in Indonesia thatyear were linked to an exceptional El Nintildeo event(Goldammer and Seibert 1990)

These observations indicate how the penetration offire into forest occurs only in very dry years Onemight thus think that fires could have played amajor role in the forest destruction 2500 BP Thishas indeed been shown in eastern BrazilianAmazonia during its major disturbance in themiddle Holocene That fires were frequent then isshown by the numerous charcoal fragmentsdeposited in the lake sediments at Carajas (Martinet al 1993 Servant et al 1993) Yet the lateHolocene lake deposits in Central Africa hardlycontain any charcoal fragments especially for theperiod 3000ndash2000 BP (Giresse et al 1994 Maleyand Brenac 1998a Bertaux et al 2000) It is thesame for the upper horizons of the numerous soilprofiles observed in this region (eg Valleacuterie 1973)In the thalwegs the coarse deposits in the lowerpart of the Lower Terrace do at times containcharcoal or fragments of fossil wood (Maley andBrenac 1998b) however Iron age archaeologicalsites dating from the beginning of the lateHolocene such as in the savanna areas of La Lopeacutealso present wood charcoal (Oslisly and White2000) One can conclude provisionally that thefires that existed were of limited extent

In particular charcoal associated with ironsmelting furnaces cannot have caused the forestdisturbance and cannot have been responsible forthe savanna extension about 2500 BP Goucher(1981) had advanced the hypothesis that ironsmelting could have been a major cause ofdeforestation in West Africa but Fairhead andLeach (1998) have demonstrated that thishypothesis had no serious foundation For CentralAfrica Pinccedilon (1990) also concludes that the

22

extraction of wood for iron metallurgy on theBateke Plateaux was minimal and did not explain(as had been suggested) the presence of theextensive savannas there Instead these savannasresult from an elevated edaphic drought caused byvery sandy soils in which rain infiltrates rapidly toa great depth (Laraque and Pandi 1996)

In contrast to observations from Atlantic CentralAfrica (south Cameroon Gabon and Congo) woodcharcoal has been found in abundance in Ituri nearto Epulu (c 1ucirc20rsquoNndash28ucirc35rsquoE) Here in the heart ofthe forests of the north-east Congo basin about 180km from the nearest savannas 416 soil profileshave been examined Almost all of them containnumerous wood charcoal in their upper levelsbetween the surface and 50 cm deep (Hart et al1996) Virtually all of the 1817 samples of woodcharcoal identified by Deschamps belong to treesfound in the regionrsquos diverse forest formationsexcept for a few pieces of charcoal from twoprofiles which belong to typical wooded savannaspecies The 28 dates obtained for the charcoalshow that fires occurred almost exclusively duringthe last three millennia with a phase of woodedsavanna expansion dating to about 2200 BP (Hartet al 1996) Moreover about 300 km south ofEpulu and 100 km from the forestsavanna limit aroad cutting near Osokari (1ucirc16rsquoSndash27ucirc48rsquoE)revealed a remarkable profile above the principalstone line that could be dated to about 11500years BP (Runge 1996 Maley 1996) A secondgravel level obliquely cutting the top of the soilprofile contains wood charcoal dating from about2200 to 1850 BP These results show that the forestdisturbance in Atlantic Central Africa extended toIturi and affected a large part of the eastern regionof the Congo river basin In contrast with regionsnearer the Atlantic fire had a large role

Also at Ituri (Epulu) Hart et al (1996) show thatcertain trees which are now found there are notfound in the wood charcoal in particularGilbertiodendron dewevrei (Caesalpiniaceae) which isnow abundant (almost mono-dominant) in certainparts of the forest The range of forests dominatedby Gilbertiodendron dewevrei extends to the east andnorth of the Congo basin as far as east Cameroon(Leacuteonard 1953) It is thus possible that thisCaesalpiniaceae has very recently colonised the Ituriforests and that the large area it now occupies is a

legacy of the disturbance about 2500 BP Ifconfirmed this area could correspond at leastpartly to an older forest area (of unknowncharacter) which was destroyed by the disturbanceA comparison with Okoumeacute can be suggestedwhich as indicated for Gabon above alsodominates areas which it has colonised since lessthan 2000 years ago However as the forest agesOkoumeacute loses its dominant character whereas inIturi Gilbertiodendron dewevrei remains dominantprobably because it can regenerate abundantly inthe shade of parent trees and also because itsreproductive character gives it an advantage overother trees in this formation (see Hart 2001)

5 Conclusion the major role ofclimatic conditions and theirhistoryThe present mosaic pattern of numerous forestscharacterised by a mix or juxtaposition ofgroupings of evergreen and deciduous species isprobably the consequence of long-term disturbanceswhich have affected the forest domain during thelast three millennia particularly the majordisturbance which culminated about 2500 years BPThe synchronism apparent in the disturbance fordifferent sites studied across Central Africa (southCameroon Gabon Congo as far east as the CongoRiver basin) and its association with a generalisederosive phase enables one to conclude that it wasthe result of a major climatic change

The climate at this time appears to have beenrelatively arid as it led to forest destruction and inplaces the extension of savannas Neverthelessother characteristics such as the strongsynchronous extension of diverse pioneer taxaindicate that in places there was not truly areduction in rainfall but rather a change in itsannual distribution The absence of a fall in annualrainfall has been demonstrated at lake BarombiMbo (Maley and Brenac 1998a) and at lake Ossawhere diatom studies show that there was even anet rise in this lake between 2500 and 2200 yearsBP with only a short fall between 2200 and 2000years BP (Nguetsop et al 2000) It is thereforepreferable to designate this catastrophic period asan lsquounfavourable climatic disturbancersquo caused by anincrease in seasonality and a lengthening of the dryseason (Maley 1997) The strong soil erosion of this

23

period suggests heavy rains but concentrated oversix or seven months of the year as is now found inthe peripheral savanna zones The growth inseasonality could have been associated with adominance of lsquosquall linesrsquo the typical cloudformation of savanna zones formed by northsouthalignments of cumuliform (convective storm-type)clouds They surplanted other cloud typesprincipally the stratiform types which giverelatively fine regular monsoon-type rains (seeMaley 1982) That these monsoon rains wouldhave dominated the earlier period (between 4000and 30002800 BP Maley 1997) is shown by therelative development of Caesalpinaceae in theforests of low altitude and in the mountains by theextension of Podocarpus which are trees typical ofcloud forests with stratiform clouds (Kerfoot1968 Maley 1996 1997 Maley and Brenac1998a)

Research in dynamic tropical climatology over thelast 20 years shows that Sea Surface Temperature(SST) on the regional and global scale stronglyinfluences climate (Fontaine and Bigot 1993

Moron et al 1995 Bigot et al 1997) It is possibleto characterise the relations between the monsoonrains and the SST more precisely Studies ofclimatic anomalies over tropical Africa occurringsince the beginning of the 1960s have evidencedtwo dominant modes of rainfall distribution eachassociated with a particular distribution of SST Thefirst mode is characterised on the one hand bywarmer than average temperatures in the Gulf ofGuinea and the southern Atlantic and on the otherhand by cooler temperatures on the north tropicalAtlantic offshore of West Africa The second modepresents an opposed distribution for the Africancontinent and the ocean A north-south tropicalAtlantic lsquodipolersquo has thus been found whichoscillates between one mode and the other(Fontaine and Bigot 1993 Wotling et al 1995Bigot et al 1997) An important research result hasbeen to show that over recent millennia the spatialdistribution of paleo-climatic anomalies overtropical Africa (more humid regions and more dryones Figure 6) and of SST on the neighbouringAtlantic is very similar to the spatial distributionobserved for the two dominant anomalies over thelast four decades (Maley 1997 Maley et al 2000)One can deduce that the climatic mechanisms arethe same at the timescale of years centuries ormillennia

So the variation of SST over the tropical Atlanticappears to play a major role in the variation ofclimate and rain types over Central Africa Thephase of forest disturbance that culminated about2500 years BP was associated with relatively lsquowarmrsquoSST over the Gulf of Guinea The earlier phasebetween about 40003800 and 2800 years BP wasassociated with relatively lsquocoolrsquo SST sharplyreduced from the early and middle Holocene (seeMorley and Dworetzky 1993 Figure 7) Over westand south Cameroon these SST allowed relativelyhigh rains and the evolution of stratiform cloudstowards rainy nimbostratus types The palaeo-vegetation data are presently lacking for GabonHowever new research is underway within thePALEOFORGA (Paleoenvironments of GabonForests) programme studying lakes MaridorNguegravene and Kamaleacuteteacute (Figure 1) In contrastfurther south in western Congo the periodstarting from 5000 years BP (Bertaux et al 2000)was marked by the development of semi-deciduousforests (Vincens et al 2000) linked to lower

24

Figure 6 Major rainfall trends in tropical Africa andSouth America for two opposed climatic periods c 3700ndash3000 BP and c 28002500ndash2000 BPPrecipitation trends (anomalies) abovenormalmean + below normalmean ndash P pointsto the sectors where a maximum of Podocarpuspollen intervenedSource Maley 1997 Maley et al 2000

rainfalls indicating that the stratiform clouds hadprobably evolved towards non-rainy clouds Thefour-month dry season here is characterised by thequasi-permanence of these cloud types (Saint Vil1984) which probably gained place progressivelybetween 5000 and 4000 years BP There wastherefore a reinforcing of southern influencesupporting conclusions concerning the history ofOkoumeacute It is notable that the vegetational impactsof the lsquowarmrsquo SST c 28002500 to 2000 years BPwere very different from those earlier c 9000 to65006000 years BP a difference which reflects abrutal strengthening over Central Africa of borealinfluences (Nguetsop et al 2000) to the detrimentof southern influences from about 28002500years BP

Before the major Holocene forest disturbanceculminating about 2500 BP other importantdisturbances occurred during the Quaternary(Maley 1996) The best documented is that betweenabout 20 000 and 15 000 years BP (Maley 19871996 Maley and Brenac 1998a Figure 2) In bothcases residual forest environments (refugia)

persisted apparently in similar locations butperhaps less as small blocks of relativelyhomogeneous forest than as a landscape of forest-savanna mosaic in which forest islands dominated(see Leal 2000) Yet two important characteristicsclearly differentiate these two phases of forestfragmentation First the older disturbance wasmuch longer Second they occurred under verydifferent (even opposed) general climatic conditionsThe disturbance between 20 000 and 15 000 BPcoincided with the development of glacialconditions in middle and high latitudes and coolerconditions by several degrees also affected the wholeregion of the Gulf of Guinea (Maley 1996 Maley andBrenac 1998a) The most recent disturbance bycontrast occurred during the Holocene interglacialwhich on a global scale is characterised by reducedglacial extension and relatively warmertemperatures The period between 2500 and 2000BP is associated with slightly warmer climaticconditions on the regional and global scale (Maley1997 Maley et al 2000) An important conclusion isthat the retreat and fragmentation of the Africanforests can be produced under very different

25

Figure 7 Variations in Sea Surface Temperature (SST) in the South Atlantic Ocean Benguela sector(22ucirc20rsquoSndash11ucirc12rsquoE) estimated with transfer functions based on the abundance of Radiolarian species in thecore RC13ndash228The open circles designate the 6000 and 9000 BP interpolated dates for the two curves one correspondingto the southern summer (February) and the other to the southern winter (August) modern SST areindicated by the solid squares above each curve The dashed lines represent some interpolated dates(italics) 18 000 BP based on the oxygen-isotope stratigraphy 38004000 BP 3000 BP and 2000 BP Theselate Holocene dates are confirmed by an independent study of shells from the same sector (Cohen et al1992) The coherence of the main SST anomalies throughout the Guinea Gulf permits the use of thesecurves in order to interpret the paleoclimatology of central Africa (Maley 1997)Source Morley and Dworetzky 1993

climates either relatively cool or relatively warmand thus with very different climatic situations(Maley 1996 1997)

Climatic models concerning lsquoglobal warmingrsquo inthe context of contemporary climate changeconcerns indicate that an average growth intemperature of about 4ucircC will lead also to anincrease in evaporation of about 30 per cent butonly 12 per cent more rain for tropical Africa (Rind

Notes The author is grateful to James Fairhead for his

translation of this article from the original Frenchand to Melissa Leach and James Fairhead for theirinvitation to the IDS workshop lsquoChangingPerspectives on Forestsrsquo in March 2001 This isInstitut des Sciences de lrsquoEvolution de Montpellier(ISEMCNRS) contribution No 125ndash2001

26

ReferencesAchoundong G Youta Happi J Guillet B

Bonvallot J and Kamgang Beyala V 2000lsquoFormation et eacutevolution des recrus sur savanes (ausud Cameroun)rsquo in M Servant and S Servant-Vildary (eds) Dynamique agrave long terme desEcosystegravemes forestiers intertropicaux MeacutemoireUNESCO Paris UNESCO 31ndash41

Alexandre P 1965 lsquoProto-histoire du groupe Beacuteti-Boulou-Fang essai de synthegravese provisoirersquo CahiersdrsquoEtudes Africaines Vol 20 503ndash60

Amougou Akoa 1986 lsquoEtude botanique eteacutecologique de la valleacutee inondable du Haut-Nyong(sud Cameroun) et de ses affluentsrsquo Thegravese SciencesUniv Yaoundeacute

Aubreacuteville A 1937 lsquoLes forecircts du Dahomey et duTogorsquo Bulletin du Comiteacute drsquoEtudes Historiques etScientifiques de lrsquoAfrique Occidentale Franccedilaises Vol20 112

Aubreacuteville A 1967 lsquoLes eacutetranges mosaiques forecirct-savane du sommet de la boucle de lrsquoOgooueacute auGabonrsquo Adansonia Vol 7 13ndash22

Baeke V 1996 lsquoLe temps des Rites Lrsquounivers magico-religieux des Wuli (Mfumte du Camerounoccidental)rsquo Thegravese Sc Sociale Univ Libre Bruxelles

Barber RJ 1985 lsquoLand snails and past environmentat the Igbo-Iwoto Esie site southwestern NigeriarsquoWest African Journal of Archaeology Vol 15 89ndash102

Barbier JC 1981 lsquoLe peuplement de la partiemeacuteridionale du plateau Bamileacutekeacutersquo in C Tardits

(ed) Contribution de la recherche ethnologique agravelrsquohistoire des civilisations du Cameroun ParisColloque International Centre National deRecherche Scientifique Paris Vol 2 331ndash53

Bertault JG 1990 lsquoComparaison drsquoeacutecosystegravemesforestiers naturels et modifieacutes apregraves incendie enCocircte drsquoIvoirersquo in H Puig (ed) Atelier surlrsquoAmeacutenagement et la Conservation de lrsquoEcosystegravemeForestier Tropical Humide Cayenne ParisUNESCO 6

Bertaux J Schwartz D Vincens A Sifeddine AElenga H Mansour M Mariotti A Fournier MMartin L Wirrmann D and Servant M 2000lsquoEnregistrement de la phase segraveche drsquoAfriqueCentrale vers 3000 ans BP par la spectromeacutetrie IRdans les lacs Sinnda et Kitina (sud-Congo)rsquo in MServant S Servant-Vildary (eds) Dynamique agrave longterme des Ecosystegravemes forestiers intertropicauxMeacutemoire UNESCO Paris UNESCO 43ndash9

Bigot S Camberlin P Moron V and Richard Y1997 lsquoStructures spatiales de la variabiliteacute despreacutecipitations en Afrique une transition climatiqueagrave la fin des anneacutees 1960rsquo Compte Rendu delrsquoAcademie des Sciences Paris 2a 324 181ndash88

Blanc-Pamard C and Peltre P 1984 lsquoDynamique despaysages preacuteforestiers et pratiques culturales enAfrique de lrsquoOuest (Cocircte drsquoIvoire centrale)rsquo LeDeacuteveloppement rural en question Paris MeacutemoireORSTOM No 106 55ndash74

Bonnefille R Chalieacute F Guiot J and Vincens A1992 lsquoQuantitative estimates of full glacial

1995) Though the causes of warming experiencedabout 2500 years BP were natural (Magny 1993)and therefore different from the warmingenvisaged for the twenty-first century it is possiblethat the catastrophic destruction of Central Africanforests which culminated about 2500 years BPcould be an lsquoanaloguersquo and an alarm signal forpotential developments during this current andfuture phase of human-induced global warming(Maley 1997 Maley et al 2000)

1 BP lsquoBefore the Presentrsquo according to standardisedC14 analysis

2 Only 4ndash5 samples have been studied for each 1000years during the Holocene To detect century-scaleperiodicity it would be necessary to study at least 10samples per millennium

temperatures in equatorial Africa frompalynological datarsquo Climate Dynamics Vol 6251ndash57

Brunck F Grison F and Maitre HF 1990Lrsquookoumeacute Aucoumea klaineana Pierre MonographieCentre Techn Forestier Tropical Nogent CIRAD 102

Cohen AI Perkington JE Brundrirt GB and vanDer Merwe NJ 1992 lsquoA Holocene marine climaterecord in Mollusc shells from the southwest Africancoastrsquo Quaternary Research Vol 38 379ndash85

Dechamps R Senut B and Pickford M 1992 Fruitsfossiles pliocegravenes et pleacuteistocegravenes du Rift occidentalougandais Signification paleacuteoenvironnementaleCompte Rendu de lrsquoAcadeacutemie des Sciences Parisseacuterie 2 Vol 314 325ndash331

Dowsett-Lemaire F 1996 lsquoComposition et eacutevolutionde la veacutegeacutetation forestiegravere au Parc NationaldrsquoOdzala Congorsquo Bulletin du Jardin BotaniqueNational de Belgique Vol 65 253ndash92

Dupont LM and Weinelt M 1996 lsquoVegetationhistory of the savanna corridor between theguinean and the Congolian rain forest during thelast 150000 yearsrsquo Vegetation History andArchaeobotany Vol 5 273ndash92

Elenga H Schwartz D and Vincens A 1994lsquoPollen evidence of late Quaternary vegetation andinferred climate changes in Congorsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 109345ndash56

Elenga H Schwartz D Vincens A Bertaux J DeNamur C Martin L Wirrmann D and ServantM 1996 lsquoDiagramme pollinique holocegravene du lacKitina (Congo) mise en eacutevidence de changementspaleacuteobotaniques et paleacuteoclimatiques dans le massifforestier du Mayombersquo Compte Rendu de lrsquoAcadeacutemiedes Sciences Paris Vol 323 No 2a 403ndash10

Elenga H Peyron O Bonnefille R Prentice ICJolly D Cheddadi R Guiot J Andrieu V DeBeaulieu JL Hamilton AC Maley J MarchantR Reille M Scott L Taylor D and Van CampoE 2000 lsquoPollen-based biome reconstruction forsouthern Europe and Africa 18000 years agorsquoJournal of Biogeography Vol 27 621ndash34

Fairhead J and Leach M 1995 lsquoFalse forest historycomplicit social analysis rethinking some westAfrican environmental narrativesrsquo WorldDevelopment Vol 23 1023ndash35

Fairhead J and Leach M 1996 lsquoEnriching thelandscape social history and the management oftransition ecology in the forest-savanna mosaic ofthe Republic of Guinearsquo Africa Vol 66 14ndash36

Fairhead J and Leach M 1998 lsquoReframingdeforestation Global analyses and local realitiesstudies in West Africarsquo Global EnvironmentalChanges Series London Routledge 238

Fay JM 1997 lsquoEvidence of forest degradation innorthern Congo and southwestern Central AfricanRepublic between 2340 and 990 BPrsquo in The ecologysocial organization populations habitat and historyof the western lowland Gorilla PhD ThesisWashington Univ Saint Louis Missouri Ch 6

Fontaine B and Bigot S 1993 lsquoWest African rainfalldeficit and sea surface temperaturesrsquo InternationalJournal of Climatology Vol 13 271ndash85

Gautier L 1990 lsquoContact forecirct-savane en CocirctedrsquoIvoire centrale eacutevolution du recouvrementligneux des savanes de la Reacuteserve de Lamto (sud duV-Baouleacute)rsquo Candollea Vol 45 627ndash41

Giresse P Maley J and Brenac P 1994 lsquoLateQuaternary palaeoenvironments in the lake BarombiMbo (Cameroon) deduced from pollen and carbonisotopes of organic matterrsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 107 65ndash78

Goldammer JG and Seibert B 1990 lsquoThe impact ofdroughts and forest fires on tropical lowland rainforest of East Kalimatanrsquo in JG Goldammer (ed)Fire in the Tropical Biota Ecological Studies BerlinSpringer-Verlag 84

Goucher CL 1981 lsquoIron is iron til it is rust tradeand ecology in the decline of West African iron-smeltingrsquo Journal of African History Vol 22 179ndash89

Guille-Escuret G 1990 lsquoPalmier agrave huile vin depalme et transformations sociales en Lobaye (ForecirctCentrafricaine)rsquo Information sur les Sciences SocialesVol 29 327ndash53

Hart TB 2001 lsquoForest dynamics in the Ituri basin(DR Congo) Dominance diversity andconservationrsquo in W Weber L White A Vedderand L Naughton-Treves (eds) African rain forestecology and conservation An interdisciplinaryperspective Yale Yale University Press 154ndash64

Hart TB Hart JA Dechamps R Fournier M andAtaholo M 1996 lsquoChanges in forest compositionover the last 4000 years in the Ituri basin Zairersquo inLJG Van der Maesen and XM Van der Burgt (eds)The Biodiversity of African plants Dordrecht Kluwer545ndash63

Hawthorne WD 1991 Fire damage and forestregeneration in Ghana Ghana Forestry Departmentamp ODA Report

Jolly D Harrison SP Damnati B and BonnefilleR 1998 lsquoSimulated climate and Biomes of Africa

27

vicinity of Barombi Mbo (Maley and Brenac1998a) was accompanied by a retreat of oil palmprobably because ecological and climatic factorsrendered other pioneers more active at this time InFigure 3 it is notable that in west Cameroonbetween 2000 and 1000 years BP the pioneer treeMilicia was in a phase of very strong expansionThen here and elsewhere in central Africa a secondphase of oil palm expansion began around 1400years BP This culminated towards 1000 BP at lakeOssa near to Edea (Reynaud-Farrera et al 1996) inthe south of the Central African Republic atNouabale-Ndoki (Fay 1997 Maley 1999) but also

further south in Gabon to La Lopeacute (White et al2000 Maley unpublished) and in western Congonear to lake Kitina (Elenga et al 1996) This phaseending about 700800 BP must have beenassociated with renewed forest disturbance butweaker than that 2500 BP This disturbance wasalso associated with a short phase of soil erosionand with a discontinuity in the deposits observedin assorted locations in Central Africa (Maley andBrenac 1998b) An increase in mineral fluxes hasbeen found between 1200ndash800 years BP for thesites of Kitina and Sinnda in western Congo(Bertaux et al 2000) From the thirteenth centurya phase of forest recovery has occurred in CentralAfrica with a new development of mature forestsand a generalised retreat of oil palm a recoverydespite increased human settlements (Alexandre1965 Vansina 1990)

At the western extension of the Central Africanforests lies a large area without forest which bio-geographers call the Dahomey Gap stretchingacross southern Togo and Benin Much of this iscovered by an immense palm forest of naturalorigin as it was described by Aubreacuteville (1937)and Mondjannagni (1969) Some pollen data showthat this area was covered by forests during theearly and mid Holocene (Dupont and Weinelt1996) and that the opening-up of the landscapeoccurred rather abruptly between 4000 and 3500 BPfollowing a climatic change well-documented forthe neighbouring region of Ghana at lakeBosumtwi (Maley 1991 1997 1999) Moreover itappears that the break in the forest bloc wasoriginally much wider than it is now perhapsextending across almost all of south-west Nigeriaup to the Niger river Indeed from the beginning ofthe last millennium to the sixteenth century ADsome data show that this vast sector was notforested but becoming so (Barber 1985) and thatthe vegetation was dominated by oil palmappearing similar to currently existing vegetationacross southern Benin and Togo Data indicatingthis exist for an area near to the ancient city ofBenin (Okomu Forest about 100 km west of theNiger river) suggesting also that the forestextension occurred only in recent centuries (Jones1956 White and Oates 1999) Further researchwill be necessary to clarify these various points andtheir precise chronology

20

Figure 5 An lsquooil palm beltrsquo near the edge of therain forest in west Cameroon

Belt with a high density of Elaeis guineensis(oil palm) reaching 20 to 25 m height inassociation with a mature forest of the semi-deciduous type

In the Grass Fields extensive savannasnorth of the forest pioneer forest islands withlarge concentrations of oil palms In italics thenames of some important groups of people(Bamileke Bamoun Tikar Wuli)Source adapted from Letouzey 1978 and 1985

32 OkoumeacuteOkoumeacute (Aucoumea Klaineana Burseraceae) iscurrently restricted to the west and centre ofGabon with a limited extension in the south-westof Congo and another towards the north inEquatorial Guinea (Brunck et al 1990) (Figure 1)Like oil palm Okoumeacute is a light demandingpioneer species (which for some ecological reasonappears unable to live outside of central AfricaBrunk et al 1990 Maley 1990 Nasi 1997) In itsarea this tree is perfectly adapted to colonisepioneer forest fronts as they progress into savannasand often dominates them (Nasi 1997 White et al2000) It is abundant in fallows and along foresttracks As pioneer fronts mature and age othertrees of shade-tolerant species appear (egCaesalpiniaceae) and the Okoumeacute which surviveare the dispersed individuals whose crowns havealready reached the canopy living to 100ndash150years and achieving 12 m diameters (Nasi 1997)It can be deduced that the large Okoumeacute currentlypresent in the forest interior in its natural rangewere born in contact with savannas

The endemism of Okoumeacute with its range limitedmainly to Gabon is linked to its history and itsecology It is intolerant of water deficit and thus ofthe elevated temperatures of the long dry seasons(about 3 months from December to February) oftropical climates Within its Gabonese range waterdeficit is less severe than in neighbouring southernCameroon because the equatorial climatersquos dryseason (mostly from June to August) retains theelevated air humidity and moderate temperatures(Brunck et al 1990 Maley 1990)

Several works bring important precision to thehistory of Okoumeacute First palynological dataobtained in south Cameroon at lake Ossa near toEdea (Reynaud-Farrera et al 1996) show thatbetween 7500 and 3000 BP (middle Holocene) thedistribution of Okoumeacute extended much furthernorth reaching the lake about 170 km north of thetreersquos present range and perhaps further east andfurther south in western Congo Second moleculargenetics show two completely distinct varieties ofOkoumeacute one to the north of its range and anotherto the south (Muloko et al 1998 2000) Theboundary between the ranges runs approximatelyeast-west a little to the south of the equatortowards 0ucirc40rsquoSouth

It can be deduced that the current range of Okoumeacutehas resulted from the fragmentation of a largeancient range dating back to the middle Holocenefragmented by the generalised forest disturbanceabout 2500 BP into two very dispersed sub-groupsone to the south and the other to the north Theextension of open landscapes during the period ofdestruction would have been favourable toOkoumeacutersquos new phase of development in a similarway to that evidenced at Barombi Mbo for severalother pioneer trees (Figure 3b) The general forestrecolonisation dating from about 2000 BP was inan extremely fragmented environment withnumerous ecotones favourable to the developmentof Okoumeacute It is thus likely that the current rangeof Okoumeacute is about 2000 years old at most

A dynamic model which nearly follows thischronological canvas has been established by White(1995) This is based on examination of the currentvegetation of the northern part of the La Lopeacutereserve which is situated at the heart of the Gabonrsquosforest and still holds large patches of residualsavanna (see Aubreacuteville 1967) White (1995) showshow the forest reconstituted itself progressivelythanks to the wide bands of pioneer forest rich inOkoumeacute which surround these savannas Theprogression of pioneer forest fronts can be blockedby savanna fires often of anthropic origin (Oslislyand White 2000) Even so over the long term fromabout 2000 years ago the continuous general trendhas been in favour of reforestation

That the current and expanding range of Okoumeacutehas not extended much into south Cameroon ashad been the case in the middle Holocene isprobably explained by a shift in climatic conditionsbetween this older period and the recent (post-2000 BP) Holocene This deduction is importantas given the ecological particularities of Okoumeacutewhich link it tightly to the equatorial climate onecan further deduce that equatorial climates musthave reigned over south Cameroon during themiddle Holocene The present tropical climatethere thus dates from only about 2000 BP All thisappears to translate for the middle Holocene intoan increase in lsquosouthernrsquo equatorial influencestowards the north associated with a reduction inlsquonorthernrsquo boreal influences and possibly linked toa reduction in the length of the dry seasonassociated with the boreal winter

21

4 The question of savanna firesand the extension of the forestdisturbance in eastern Congo

Observations and opinions have frequentlydiverged concerning the influence of savanna fireson contemporary forest regeneration Observersvariously report that fire can either preventregeneration as in the area of savanna inliers inLopeacute Gabon (Oslisly and White 2000) or merelydelay it as in other parts of the forest region ThusLetouzey (1968) notes for the forest-savannatransition zone of south Cameroon that lsquofire doesnot necessarily prevent the formation of forestrecoveryrsquo Similar conclusions have been drawn inCentral Africa by Sillans and even in westernCongo (ex-Zaire) for the south of the forest bloc(Letouzey 1968) On the same lines Youta Happiand Bonvallot (1996) report lsquothat on theperiphery of a town of 76000 people such asBertoua (east Cameroon) many small savannainlets into forest of two to ten hectares havebecome covered by forest regrowth despite theannual bush fires practiced by Bororopastoralistsrsquo A little further south in the OdzalaReserve situated at the northern extremity of theBateke savannas (north Congo) Dowsett-Lemairereports that lsquoWithout doubt one is in a phase offorest extension despite very frequent bush firesrsquo(1996) This author notes lsquothe fire-break role ofthe boundary thanks to groves dominated byGaertnera paniculata a species absent at Lopeacutewhere colonising species are apparently moresensitive to firersquo Thus the floristic composition ofthe boundary plays an important role in thisphenomenon This was shown in south-eastCameroon by the mid-twentieth century invasionof Chromolaena odorata which colonised theboundaries and which acts there at least as anefficacious fire break It also protects seeds ofwoody species which can germinate and developunder its cover (Youta Happi et al 2000) Otherobservations show how traditional agriculturalpractices and the movement of cattle can befavourable to forest recolonisation in particularby eliminating the grasses which facilitate thepropagation and intensity of fires These processeshave been well described in southern Cameroon(Letouzey 1968 1985) and in Cocircte drsquoIvoire(Spichiger and Blanc-Pamard 1973 Blanc-Pamardand Peltre 1984 Gautier 1990)

These observations indicate that in general andespecially over the long term forest colonisation atthe forest-savanna contact has been the broadlydominant process In certain exceptional yearswhen the dry season has been one or two monthslonger the savanna fires have penetrated moredeeply into the neighbouring forests In early 1983during a two-month longer dry season numerousobservers reported the strong invasion of fire intoforest in Cocircte drsquoIvoire (Bertault 1990) Ghana(Hawthorne 1991 Swaine 1992) and Cameroon(Amougou 1986) Equally fires in Indonesia thatyear were linked to an exceptional El Nintildeo event(Goldammer and Seibert 1990)

These observations indicate how the penetration offire into forest occurs only in very dry years Onemight thus think that fires could have played amajor role in the forest destruction 2500 BP Thishas indeed been shown in eastern BrazilianAmazonia during its major disturbance in themiddle Holocene That fires were frequent then isshown by the numerous charcoal fragmentsdeposited in the lake sediments at Carajas (Martinet al 1993 Servant et al 1993) Yet the lateHolocene lake deposits in Central Africa hardlycontain any charcoal fragments especially for theperiod 3000ndash2000 BP (Giresse et al 1994 Maleyand Brenac 1998a Bertaux et al 2000) It is thesame for the upper horizons of the numerous soilprofiles observed in this region (eg Valleacuterie 1973)In the thalwegs the coarse deposits in the lowerpart of the Lower Terrace do at times containcharcoal or fragments of fossil wood (Maley andBrenac 1998b) however Iron age archaeologicalsites dating from the beginning of the lateHolocene such as in the savanna areas of La Lopeacutealso present wood charcoal (Oslisly and White2000) One can conclude provisionally that thefires that existed were of limited extent

In particular charcoal associated with ironsmelting furnaces cannot have caused the forestdisturbance and cannot have been responsible forthe savanna extension about 2500 BP Goucher(1981) had advanced the hypothesis that ironsmelting could have been a major cause ofdeforestation in West Africa but Fairhead andLeach (1998) have demonstrated that thishypothesis had no serious foundation For CentralAfrica Pinccedilon (1990) also concludes that the

22

extraction of wood for iron metallurgy on theBateke Plateaux was minimal and did not explain(as had been suggested) the presence of theextensive savannas there Instead these savannasresult from an elevated edaphic drought caused byvery sandy soils in which rain infiltrates rapidly toa great depth (Laraque and Pandi 1996)

In contrast to observations from Atlantic CentralAfrica (south Cameroon Gabon and Congo) woodcharcoal has been found in abundance in Ituri nearto Epulu (c 1ucirc20rsquoNndash28ucirc35rsquoE) Here in the heart ofthe forests of the north-east Congo basin about 180km from the nearest savannas 416 soil profileshave been examined Almost all of them containnumerous wood charcoal in their upper levelsbetween the surface and 50 cm deep (Hart et al1996) Virtually all of the 1817 samples of woodcharcoal identified by Deschamps belong to treesfound in the regionrsquos diverse forest formationsexcept for a few pieces of charcoal from twoprofiles which belong to typical wooded savannaspecies The 28 dates obtained for the charcoalshow that fires occurred almost exclusively duringthe last three millennia with a phase of woodedsavanna expansion dating to about 2200 BP (Hartet al 1996) Moreover about 300 km south ofEpulu and 100 km from the forestsavanna limit aroad cutting near Osokari (1ucirc16rsquoSndash27ucirc48rsquoE)revealed a remarkable profile above the principalstone line that could be dated to about 11500years BP (Runge 1996 Maley 1996) A secondgravel level obliquely cutting the top of the soilprofile contains wood charcoal dating from about2200 to 1850 BP These results show that the forestdisturbance in Atlantic Central Africa extended toIturi and affected a large part of the eastern regionof the Congo river basin In contrast with regionsnearer the Atlantic fire had a large role

Also at Ituri (Epulu) Hart et al (1996) show thatcertain trees which are now found there are notfound in the wood charcoal in particularGilbertiodendron dewevrei (Caesalpiniaceae) which isnow abundant (almost mono-dominant) in certainparts of the forest The range of forests dominatedby Gilbertiodendron dewevrei extends to the east andnorth of the Congo basin as far as east Cameroon(Leacuteonard 1953) It is thus possible that thisCaesalpiniaceae has very recently colonised the Ituriforests and that the large area it now occupies is a

legacy of the disturbance about 2500 BP Ifconfirmed this area could correspond at leastpartly to an older forest area (of unknowncharacter) which was destroyed by the disturbanceA comparison with Okoumeacute can be suggestedwhich as indicated for Gabon above alsodominates areas which it has colonised since lessthan 2000 years ago However as the forest agesOkoumeacute loses its dominant character whereas inIturi Gilbertiodendron dewevrei remains dominantprobably because it can regenerate abundantly inthe shade of parent trees and also because itsreproductive character gives it an advantage overother trees in this formation (see Hart 2001)

5 Conclusion the major role ofclimatic conditions and theirhistoryThe present mosaic pattern of numerous forestscharacterised by a mix or juxtaposition ofgroupings of evergreen and deciduous species isprobably the consequence of long-term disturbanceswhich have affected the forest domain during thelast three millennia particularly the majordisturbance which culminated about 2500 years BPThe synchronism apparent in the disturbance fordifferent sites studied across Central Africa (southCameroon Gabon Congo as far east as the CongoRiver basin) and its association with a generalisederosive phase enables one to conclude that it wasthe result of a major climatic change

The climate at this time appears to have beenrelatively arid as it led to forest destruction and inplaces the extension of savannas Neverthelessother characteristics such as the strongsynchronous extension of diverse pioneer taxaindicate that in places there was not truly areduction in rainfall but rather a change in itsannual distribution The absence of a fall in annualrainfall has been demonstrated at lake BarombiMbo (Maley and Brenac 1998a) and at lake Ossawhere diatom studies show that there was even anet rise in this lake between 2500 and 2200 yearsBP with only a short fall between 2200 and 2000years BP (Nguetsop et al 2000) It is thereforepreferable to designate this catastrophic period asan lsquounfavourable climatic disturbancersquo caused by anincrease in seasonality and a lengthening of the dryseason (Maley 1997) The strong soil erosion of this

23

period suggests heavy rains but concentrated oversix or seven months of the year as is now found inthe peripheral savanna zones The growth inseasonality could have been associated with adominance of lsquosquall linesrsquo the typical cloudformation of savanna zones formed by northsouthalignments of cumuliform (convective storm-type)clouds They surplanted other cloud typesprincipally the stratiform types which giverelatively fine regular monsoon-type rains (seeMaley 1982) That these monsoon rains wouldhave dominated the earlier period (between 4000and 30002800 BP Maley 1997) is shown by therelative development of Caesalpinaceae in theforests of low altitude and in the mountains by theextension of Podocarpus which are trees typical ofcloud forests with stratiform clouds (Kerfoot1968 Maley 1996 1997 Maley and Brenac1998a)

Research in dynamic tropical climatology over thelast 20 years shows that Sea Surface Temperature(SST) on the regional and global scale stronglyinfluences climate (Fontaine and Bigot 1993

Moron et al 1995 Bigot et al 1997) It is possibleto characterise the relations between the monsoonrains and the SST more precisely Studies ofclimatic anomalies over tropical Africa occurringsince the beginning of the 1960s have evidencedtwo dominant modes of rainfall distribution eachassociated with a particular distribution of SST Thefirst mode is characterised on the one hand bywarmer than average temperatures in the Gulf ofGuinea and the southern Atlantic and on the otherhand by cooler temperatures on the north tropicalAtlantic offshore of West Africa The second modepresents an opposed distribution for the Africancontinent and the ocean A north-south tropicalAtlantic lsquodipolersquo has thus been found whichoscillates between one mode and the other(Fontaine and Bigot 1993 Wotling et al 1995Bigot et al 1997) An important research result hasbeen to show that over recent millennia the spatialdistribution of paleo-climatic anomalies overtropical Africa (more humid regions and more dryones Figure 6) and of SST on the neighbouringAtlantic is very similar to the spatial distributionobserved for the two dominant anomalies over thelast four decades (Maley 1997 Maley et al 2000)One can deduce that the climatic mechanisms arethe same at the timescale of years centuries ormillennia

So the variation of SST over the tropical Atlanticappears to play a major role in the variation ofclimate and rain types over Central Africa Thephase of forest disturbance that culminated about2500 years BP was associated with relatively lsquowarmrsquoSST over the Gulf of Guinea The earlier phasebetween about 40003800 and 2800 years BP wasassociated with relatively lsquocoolrsquo SST sharplyreduced from the early and middle Holocene (seeMorley and Dworetzky 1993 Figure 7) Over westand south Cameroon these SST allowed relativelyhigh rains and the evolution of stratiform cloudstowards rainy nimbostratus types The palaeo-vegetation data are presently lacking for GabonHowever new research is underway within thePALEOFORGA (Paleoenvironments of GabonForests) programme studying lakes MaridorNguegravene and Kamaleacuteteacute (Figure 1) In contrastfurther south in western Congo the periodstarting from 5000 years BP (Bertaux et al 2000)was marked by the development of semi-deciduousforests (Vincens et al 2000) linked to lower

24

Figure 6 Major rainfall trends in tropical Africa andSouth America for two opposed climatic periods c 3700ndash3000 BP and c 28002500ndash2000 BPPrecipitation trends (anomalies) abovenormalmean + below normalmean ndash P pointsto the sectors where a maximum of Podocarpuspollen intervenedSource Maley 1997 Maley et al 2000

rainfalls indicating that the stratiform clouds hadprobably evolved towards non-rainy clouds Thefour-month dry season here is characterised by thequasi-permanence of these cloud types (Saint Vil1984) which probably gained place progressivelybetween 5000 and 4000 years BP There wastherefore a reinforcing of southern influencesupporting conclusions concerning the history ofOkoumeacute It is notable that the vegetational impactsof the lsquowarmrsquo SST c 28002500 to 2000 years BPwere very different from those earlier c 9000 to65006000 years BP a difference which reflects abrutal strengthening over Central Africa of borealinfluences (Nguetsop et al 2000) to the detrimentof southern influences from about 28002500years BP

Before the major Holocene forest disturbanceculminating about 2500 BP other importantdisturbances occurred during the Quaternary(Maley 1996) The best documented is that betweenabout 20 000 and 15 000 years BP (Maley 19871996 Maley and Brenac 1998a Figure 2) In bothcases residual forest environments (refugia)

persisted apparently in similar locations butperhaps less as small blocks of relativelyhomogeneous forest than as a landscape of forest-savanna mosaic in which forest islands dominated(see Leal 2000) Yet two important characteristicsclearly differentiate these two phases of forestfragmentation First the older disturbance wasmuch longer Second they occurred under verydifferent (even opposed) general climatic conditionsThe disturbance between 20 000 and 15 000 BPcoincided with the development of glacialconditions in middle and high latitudes and coolerconditions by several degrees also affected the wholeregion of the Gulf of Guinea (Maley 1996 Maley andBrenac 1998a) The most recent disturbance bycontrast occurred during the Holocene interglacialwhich on a global scale is characterised by reducedglacial extension and relatively warmertemperatures The period between 2500 and 2000BP is associated with slightly warmer climaticconditions on the regional and global scale (Maley1997 Maley et al 2000) An important conclusion isthat the retreat and fragmentation of the Africanforests can be produced under very different

25

Figure 7 Variations in Sea Surface Temperature (SST) in the South Atlantic Ocean Benguela sector(22ucirc20rsquoSndash11ucirc12rsquoE) estimated with transfer functions based on the abundance of Radiolarian species in thecore RC13ndash228The open circles designate the 6000 and 9000 BP interpolated dates for the two curves one correspondingto the southern summer (February) and the other to the southern winter (August) modern SST areindicated by the solid squares above each curve The dashed lines represent some interpolated dates(italics) 18 000 BP based on the oxygen-isotope stratigraphy 38004000 BP 3000 BP and 2000 BP Theselate Holocene dates are confirmed by an independent study of shells from the same sector (Cohen et al1992) The coherence of the main SST anomalies throughout the Guinea Gulf permits the use of thesecurves in order to interpret the paleoclimatology of central Africa (Maley 1997)Source Morley and Dworetzky 1993

climates either relatively cool or relatively warmand thus with very different climatic situations(Maley 1996 1997)

Climatic models concerning lsquoglobal warmingrsquo inthe context of contemporary climate changeconcerns indicate that an average growth intemperature of about 4ucircC will lead also to anincrease in evaporation of about 30 per cent butonly 12 per cent more rain for tropical Africa (Rind

Notes The author is grateful to James Fairhead for his

translation of this article from the original Frenchand to Melissa Leach and James Fairhead for theirinvitation to the IDS workshop lsquoChangingPerspectives on Forestsrsquo in March 2001 This isInstitut des Sciences de lrsquoEvolution de Montpellier(ISEMCNRS) contribution No 125ndash2001

26

ReferencesAchoundong G Youta Happi J Guillet B

Bonvallot J and Kamgang Beyala V 2000lsquoFormation et eacutevolution des recrus sur savanes (ausud Cameroun)rsquo in M Servant and S Servant-Vildary (eds) Dynamique agrave long terme desEcosystegravemes forestiers intertropicaux MeacutemoireUNESCO Paris UNESCO 31ndash41

Alexandre P 1965 lsquoProto-histoire du groupe Beacuteti-Boulou-Fang essai de synthegravese provisoirersquo CahiersdrsquoEtudes Africaines Vol 20 503ndash60

Amougou Akoa 1986 lsquoEtude botanique eteacutecologique de la valleacutee inondable du Haut-Nyong(sud Cameroun) et de ses affluentsrsquo Thegravese SciencesUniv Yaoundeacute

Aubreacuteville A 1937 lsquoLes forecircts du Dahomey et duTogorsquo Bulletin du Comiteacute drsquoEtudes Historiques etScientifiques de lrsquoAfrique Occidentale Franccedilaises Vol20 112

Aubreacuteville A 1967 lsquoLes eacutetranges mosaiques forecirct-savane du sommet de la boucle de lrsquoOgooueacute auGabonrsquo Adansonia Vol 7 13ndash22

Baeke V 1996 lsquoLe temps des Rites Lrsquounivers magico-religieux des Wuli (Mfumte du Camerounoccidental)rsquo Thegravese Sc Sociale Univ Libre Bruxelles

Barber RJ 1985 lsquoLand snails and past environmentat the Igbo-Iwoto Esie site southwestern NigeriarsquoWest African Journal of Archaeology Vol 15 89ndash102

Barbier JC 1981 lsquoLe peuplement de la partiemeacuteridionale du plateau Bamileacutekeacutersquo in C Tardits

(ed) Contribution de la recherche ethnologique agravelrsquohistoire des civilisations du Cameroun ParisColloque International Centre National deRecherche Scientifique Paris Vol 2 331ndash53

Bertault JG 1990 lsquoComparaison drsquoeacutecosystegravemesforestiers naturels et modifieacutes apregraves incendie enCocircte drsquoIvoirersquo in H Puig (ed) Atelier surlrsquoAmeacutenagement et la Conservation de lrsquoEcosystegravemeForestier Tropical Humide Cayenne ParisUNESCO 6

Bertaux J Schwartz D Vincens A Sifeddine AElenga H Mansour M Mariotti A Fournier MMartin L Wirrmann D and Servant M 2000lsquoEnregistrement de la phase segraveche drsquoAfriqueCentrale vers 3000 ans BP par la spectromeacutetrie IRdans les lacs Sinnda et Kitina (sud-Congo)rsquo in MServant S Servant-Vildary (eds) Dynamique agrave longterme des Ecosystegravemes forestiers intertropicauxMeacutemoire UNESCO Paris UNESCO 43ndash9

Bigot S Camberlin P Moron V and Richard Y1997 lsquoStructures spatiales de la variabiliteacute despreacutecipitations en Afrique une transition climatiqueagrave la fin des anneacutees 1960rsquo Compte Rendu delrsquoAcademie des Sciences Paris 2a 324 181ndash88

Blanc-Pamard C and Peltre P 1984 lsquoDynamique despaysages preacuteforestiers et pratiques culturales enAfrique de lrsquoOuest (Cocircte drsquoIvoire centrale)rsquo LeDeacuteveloppement rural en question Paris MeacutemoireORSTOM No 106 55ndash74

Bonnefille R Chalieacute F Guiot J and Vincens A1992 lsquoQuantitative estimates of full glacial

1995) Though the causes of warming experiencedabout 2500 years BP were natural (Magny 1993)and therefore different from the warmingenvisaged for the twenty-first century it is possiblethat the catastrophic destruction of Central Africanforests which culminated about 2500 years BPcould be an lsquoanaloguersquo and an alarm signal forpotential developments during this current andfuture phase of human-induced global warming(Maley 1997 Maley et al 2000)

1 BP lsquoBefore the Presentrsquo according to standardisedC14 analysis

2 Only 4ndash5 samples have been studied for each 1000years during the Holocene To detect century-scaleperiodicity it would be necessary to study at least 10samples per millennium

temperatures in equatorial Africa frompalynological datarsquo Climate Dynamics Vol 6251ndash57

Brunck F Grison F and Maitre HF 1990Lrsquookoumeacute Aucoumea klaineana Pierre MonographieCentre Techn Forestier Tropical Nogent CIRAD 102

Cohen AI Perkington JE Brundrirt GB and vanDer Merwe NJ 1992 lsquoA Holocene marine climaterecord in Mollusc shells from the southwest Africancoastrsquo Quaternary Research Vol 38 379ndash85

Dechamps R Senut B and Pickford M 1992 Fruitsfossiles pliocegravenes et pleacuteistocegravenes du Rift occidentalougandais Signification paleacuteoenvironnementaleCompte Rendu de lrsquoAcadeacutemie des Sciences Parisseacuterie 2 Vol 314 325ndash331

Dowsett-Lemaire F 1996 lsquoComposition et eacutevolutionde la veacutegeacutetation forestiegravere au Parc NationaldrsquoOdzala Congorsquo Bulletin du Jardin BotaniqueNational de Belgique Vol 65 253ndash92

Dupont LM and Weinelt M 1996 lsquoVegetationhistory of the savanna corridor between theguinean and the Congolian rain forest during thelast 150000 yearsrsquo Vegetation History andArchaeobotany Vol 5 273ndash92

Elenga H Schwartz D and Vincens A 1994lsquoPollen evidence of late Quaternary vegetation andinferred climate changes in Congorsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 109345ndash56

Elenga H Schwartz D Vincens A Bertaux J DeNamur C Martin L Wirrmann D and ServantM 1996 lsquoDiagramme pollinique holocegravene du lacKitina (Congo) mise en eacutevidence de changementspaleacuteobotaniques et paleacuteoclimatiques dans le massifforestier du Mayombersquo Compte Rendu de lrsquoAcadeacutemiedes Sciences Paris Vol 323 No 2a 403ndash10

Elenga H Peyron O Bonnefille R Prentice ICJolly D Cheddadi R Guiot J Andrieu V DeBeaulieu JL Hamilton AC Maley J MarchantR Reille M Scott L Taylor D and Van CampoE 2000 lsquoPollen-based biome reconstruction forsouthern Europe and Africa 18000 years agorsquoJournal of Biogeography Vol 27 621ndash34

Fairhead J and Leach M 1995 lsquoFalse forest historycomplicit social analysis rethinking some westAfrican environmental narrativesrsquo WorldDevelopment Vol 23 1023ndash35

Fairhead J and Leach M 1996 lsquoEnriching thelandscape social history and the management oftransition ecology in the forest-savanna mosaic ofthe Republic of Guinearsquo Africa Vol 66 14ndash36

Fairhead J and Leach M 1998 lsquoReframingdeforestation Global analyses and local realitiesstudies in West Africarsquo Global EnvironmentalChanges Series London Routledge 238

Fay JM 1997 lsquoEvidence of forest degradation innorthern Congo and southwestern Central AfricanRepublic between 2340 and 990 BPrsquo in The ecologysocial organization populations habitat and historyof the western lowland Gorilla PhD ThesisWashington Univ Saint Louis Missouri Ch 6

Fontaine B and Bigot S 1993 lsquoWest African rainfalldeficit and sea surface temperaturesrsquo InternationalJournal of Climatology Vol 13 271ndash85

Gautier L 1990 lsquoContact forecirct-savane en CocirctedrsquoIvoire centrale eacutevolution du recouvrementligneux des savanes de la Reacuteserve de Lamto (sud duV-Baouleacute)rsquo Candollea Vol 45 627ndash41

Giresse P Maley J and Brenac P 1994 lsquoLateQuaternary palaeoenvironments in the lake BarombiMbo (Cameroon) deduced from pollen and carbonisotopes of organic matterrsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 107 65ndash78

Goldammer JG and Seibert B 1990 lsquoThe impact ofdroughts and forest fires on tropical lowland rainforest of East Kalimatanrsquo in JG Goldammer (ed)Fire in the Tropical Biota Ecological Studies BerlinSpringer-Verlag 84

Goucher CL 1981 lsquoIron is iron til it is rust tradeand ecology in the decline of West African iron-smeltingrsquo Journal of African History Vol 22 179ndash89

Guille-Escuret G 1990 lsquoPalmier agrave huile vin depalme et transformations sociales en Lobaye (ForecirctCentrafricaine)rsquo Information sur les Sciences SocialesVol 29 327ndash53

Hart TB 2001 lsquoForest dynamics in the Ituri basin(DR Congo) Dominance diversity andconservationrsquo in W Weber L White A Vedderand L Naughton-Treves (eds) African rain forestecology and conservation An interdisciplinaryperspective Yale Yale University Press 154ndash64

Hart TB Hart JA Dechamps R Fournier M andAtaholo M 1996 lsquoChanges in forest compositionover the last 4000 years in the Ituri basin Zairersquo inLJG Van der Maesen and XM Van der Burgt (eds)The Biodiversity of African plants Dordrecht Kluwer545ndash63

Hawthorne WD 1991 Fire damage and forestregeneration in Ghana Ghana Forestry Departmentamp ODA Report

Jolly D Harrison SP Damnati B and BonnefilleR 1998 lsquoSimulated climate and Biomes of Africa

27

32 OkoumeacuteOkoumeacute (Aucoumea Klaineana Burseraceae) iscurrently restricted to the west and centre ofGabon with a limited extension in the south-westof Congo and another towards the north inEquatorial Guinea (Brunck et al 1990) (Figure 1)Like oil palm Okoumeacute is a light demandingpioneer species (which for some ecological reasonappears unable to live outside of central AfricaBrunk et al 1990 Maley 1990 Nasi 1997) In itsarea this tree is perfectly adapted to colonisepioneer forest fronts as they progress into savannasand often dominates them (Nasi 1997 White et al2000) It is abundant in fallows and along foresttracks As pioneer fronts mature and age othertrees of shade-tolerant species appear (egCaesalpiniaceae) and the Okoumeacute which surviveare the dispersed individuals whose crowns havealready reached the canopy living to 100ndash150years and achieving 12 m diameters (Nasi 1997)It can be deduced that the large Okoumeacute currentlypresent in the forest interior in its natural rangewere born in contact with savannas

The endemism of Okoumeacute with its range limitedmainly to Gabon is linked to its history and itsecology It is intolerant of water deficit and thus ofthe elevated temperatures of the long dry seasons(about 3 months from December to February) oftropical climates Within its Gabonese range waterdeficit is less severe than in neighbouring southernCameroon because the equatorial climatersquos dryseason (mostly from June to August) retains theelevated air humidity and moderate temperatures(Brunck et al 1990 Maley 1990)

Several works bring important precision to thehistory of Okoumeacute First palynological dataobtained in south Cameroon at lake Ossa near toEdea (Reynaud-Farrera et al 1996) show thatbetween 7500 and 3000 BP (middle Holocene) thedistribution of Okoumeacute extended much furthernorth reaching the lake about 170 km north of thetreersquos present range and perhaps further east andfurther south in western Congo Second moleculargenetics show two completely distinct varieties ofOkoumeacute one to the north of its range and anotherto the south (Muloko et al 1998 2000) Theboundary between the ranges runs approximatelyeast-west a little to the south of the equatortowards 0ucirc40rsquoSouth

It can be deduced that the current range of Okoumeacutehas resulted from the fragmentation of a largeancient range dating back to the middle Holocenefragmented by the generalised forest disturbanceabout 2500 BP into two very dispersed sub-groupsone to the south and the other to the north Theextension of open landscapes during the period ofdestruction would have been favourable toOkoumeacutersquos new phase of development in a similarway to that evidenced at Barombi Mbo for severalother pioneer trees (Figure 3b) The general forestrecolonisation dating from about 2000 BP was inan extremely fragmented environment withnumerous ecotones favourable to the developmentof Okoumeacute It is thus likely that the current rangeof Okoumeacute is about 2000 years old at most

A dynamic model which nearly follows thischronological canvas has been established by White(1995) This is based on examination of the currentvegetation of the northern part of the La Lopeacutereserve which is situated at the heart of the Gabonrsquosforest and still holds large patches of residualsavanna (see Aubreacuteville 1967) White (1995) showshow the forest reconstituted itself progressivelythanks to the wide bands of pioneer forest rich inOkoumeacute which surround these savannas Theprogression of pioneer forest fronts can be blockedby savanna fires often of anthropic origin (Oslislyand White 2000) Even so over the long term fromabout 2000 years ago the continuous general trendhas been in favour of reforestation

That the current and expanding range of Okoumeacutehas not extended much into south Cameroon ashad been the case in the middle Holocene isprobably explained by a shift in climatic conditionsbetween this older period and the recent (post-2000 BP) Holocene This deduction is importantas given the ecological particularities of Okoumeacutewhich link it tightly to the equatorial climate onecan further deduce that equatorial climates musthave reigned over south Cameroon during themiddle Holocene The present tropical climatethere thus dates from only about 2000 BP All thisappears to translate for the middle Holocene intoan increase in lsquosouthernrsquo equatorial influencestowards the north associated with a reduction inlsquonorthernrsquo boreal influences and possibly linked toa reduction in the length of the dry seasonassociated with the boreal winter

21

4 The question of savanna firesand the extension of the forestdisturbance in eastern Congo

Observations and opinions have frequentlydiverged concerning the influence of savanna fireson contemporary forest regeneration Observersvariously report that fire can either preventregeneration as in the area of savanna inliers inLopeacute Gabon (Oslisly and White 2000) or merelydelay it as in other parts of the forest region ThusLetouzey (1968) notes for the forest-savannatransition zone of south Cameroon that lsquofire doesnot necessarily prevent the formation of forestrecoveryrsquo Similar conclusions have been drawn inCentral Africa by Sillans and even in westernCongo (ex-Zaire) for the south of the forest bloc(Letouzey 1968) On the same lines Youta Happiand Bonvallot (1996) report lsquothat on theperiphery of a town of 76000 people such asBertoua (east Cameroon) many small savannainlets into forest of two to ten hectares havebecome covered by forest regrowth despite theannual bush fires practiced by Bororopastoralistsrsquo A little further south in the OdzalaReserve situated at the northern extremity of theBateke savannas (north Congo) Dowsett-Lemairereports that lsquoWithout doubt one is in a phase offorest extension despite very frequent bush firesrsquo(1996) This author notes lsquothe fire-break role ofthe boundary thanks to groves dominated byGaertnera paniculata a species absent at Lopeacutewhere colonising species are apparently moresensitive to firersquo Thus the floristic composition ofthe boundary plays an important role in thisphenomenon This was shown in south-eastCameroon by the mid-twentieth century invasionof Chromolaena odorata which colonised theboundaries and which acts there at least as anefficacious fire break It also protects seeds ofwoody species which can germinate and developunder its cover (Youta Happi et al 2000) Otherobservations show how traditional agriculturalpractices and the movement of cattle can befavourable to forest recolonisation in particularby eliminating the grasses which facilitate thepropagation and intensity of fires These processeshave been well described in southern Cameroon(Letouzey 1968 1985) and in Cocircte drsquoIvoire(Spichiger and Blanc-Pamard 1973 Blanc-Pamardand Peltre 1984 Gautier 1990)

These observations indicate that in general andespecially over the long term forest colonisation atthe forest-savanna contact has been the broadlydominant process In certain exceptional yearswhen the dry season has been one or two monthslonger the savanna fires have penetrated moredeeply into the neighbouring forests In early 1983during a two-month longer dry season numerousobservers reported the strong invasion of fire intoforest in Cocircte drsquoIvoire (Bertault 1990) Ghana(Hawthorne 1991 Swaine 1992) and Cameroon(Amougou 1986) Equally fires in Indonesia thatyear were linked to an exceptional El Nintildeo event(Goldammer and Seibert 1990)

These observations indicate how the penetration offire into forest occurs only in very dry years Onemight thus think that fires could have played amajor role in the forest destruction 2500 BP Thishas indeed been shown in eastern BrazilianAmazonia during its major disturbance in themiddle Holocene That fires were frequent then isshown by the numerous charcoal fragmentsdeposited in the lake sediments at Carajas (Martinet al 1993 Servant et al 1993) Yet the lateHolocene lake deposits in Central Africa hardlycontain any charcoal fragments especially for theperiod 3000ndash2000 BP (Giresse et al 1994 Maleyand Brenac 1998a Bertaux et al 2000) It is thesame for the upper horizons of the numerous soilprofiles observed in this region (eg Valleacuterie 1973)In the thalwegs the coarse deposits in the lowerpart of the Lower Terrace do at times containcharcoal or fragments of fossil wood (Maley andBrenac 1998b) however Iron age archaeologicalsites dating from the beginning of the lateHolocene such as in the savanna areas of La Lopeacutealso present wood charcoal (Oslisly and White2000) One can conclude provisionally that thefires that existed were of limited extent

In particular charcoal associated with ironsmelting furnaces cannot have caused the forestdisturbance and cannot have been responsible forthe savanna extension about 2500 BP Goucher(1981) had advanced the hypothesis that ironsmelting could have been a major cause ofdeforestation in West Africa but Fairhead andLeach (1998) have demonstrated that thishypothesis had no serious foundation For CentralAfrica Pinccedilon (1990) also concludes that the

22

extraction of wood for iron metallurgy on theBateke Plateaux was minimal and did not explain(as had been suggested) the presence of theextensive savannas there Instead these savannasresult from an elevated edaphic drought caused byvery sandy soils in which rain infiltrates rapidly toa great depth (Laraque and Pandi 1996)

In contrast to observations from Atlantic CentralAfrica (south Cameroon Gabon and Congo) woodcharcoal has been found in abundance in Ituri nearto Epulu (c 1ucirc20rsquoNndash28ucirc35rsquoE) Here in the heart ofthe forests of the north-east Congo basin about 180km from the nearest savannas 416 soil profileshave been examined Almost all of them containnumerous wood charcoal in their upper levelsbetween the surface and 50 cm deep (Hart et al1996) Virtually all of the 1817 samples of woodcharcoal identified by Deschamps belong to treesfound in the regionrsquos diverse forest formationsexcept for a few pieces of charcoal from twoprofiles which belong to typical wooded savannaspecies The 28 dates obtained for the charcoalshow that fires occurred almost exclusively duringthe last three millennia with a phase of woodedsavanna expansion dating to about 2200 BP (Hartet al 1996) Moreover about 300 km south ofEpulu and 100 km from the forestsavanna limit aroad cutting near Osokari (1ucirc16rsquoSndash27ucirc48rsquoE)revealed a remarkable profile above the principalstone line that could be dated to about 11500years BP (Runge 1996 Maley 1996) A secondgravel level obliquely cutting the top of the soilprofile contains wood charcoal dating from about2200 to 1850 BP These results show that the forestdisturbance in Atlantic Central Africa extended toIturi and affected a large part of the eastern regionof the Congo river basin In contrast with regionsnearer the Atlantic fire had a large role

Also at Ituri (Epulu) Hart et al (1996) show thatcertain trees which are now found there are notfound in the wood charcoal in particularGilbertiodendron dewevrei (Caesalpiniaceae) which isnow abundant (almost mono-dominant) in certainparts of the forest The range of forests dominatedby Gilbertiodendron dewevrei extends to the east andnorth of the Congo basin as far as east Cameroon(Leacuteonard 1953) It is thus possible that thisCaesalpiniaceae has very recently colonised the Ituriforests and that the large area it now occupies is a

legacy of the disturbance about 2500 BP Ifconfirmed this area could correspond at leastpartly to an older forest area (of unknowncharacter) which was destroyed by the disturbanceA comparison with Okoumeacute can be suggestedwhich as indicated for Gabon above alsodominates areas which it has colonised since lessthan 2000 years ago However as the forest agesOkoumeacute loses its dominant character whereas inIturi Gilbertiodendron dewevrei remains dominantprobably because it can regenerate abundantly inthe shade of parent trees and also because itsreproductive character gives it an advantage overother trees in this formation (see Hart 2001)

5 Conclusion the major role ofclimatic conditions and theirhistoryThe present mosaic pattern of numerous forestscharacterised by a mix or juxtaposition ofgroupings of evergreen and deciduous species isprobably the consequence of long-term disturbanceswhich have affected the forest domain during thelast three millennia particularly the majordisturbance which culminated about 2500 years BPThe synchronism apparent in the disturbance fordifferent sites studied across Central Africa (southCameroon Gabon Congo as far east as the CongoRiver basin) and its association with a generalisederosive phase enables one to conclude that it wasthe result of a major climatic change

The climate at this time appears to have beenrelatively arid as it led to forest destruction and inplaces the extension of savannas Neverthelessother characteristics such as the strongsynchronous extension of diverse pioneer taxaindicate that in places there was not truly areduction in rainfall but rather a change in itsannual distribution The absence of a fall in annualrainfall has been demonstrated at lake BarombiMbo (Maley and Brenac 1998a) and at lake Ossawhere diatom studies show that there was even anet rise in this lake between 2500 and 2200 yearsBP with only a short fall between 2200 and 2000years BP (Nguetsop et al 2000) It is thereforepreferable to designate this catastrophic period asan lsquounfavourable climatic disturbancersquo caused by anincrease in seasonality and a lengthening of the dryseason (Maley 1997) The strong soil erosion of this

23

period suggests heavy rains but concentrated oversix or seven months of the year as is now found inthe peripheral savanna zones The growth inseasonality could have been associated with adominance of lsquosquall linesrsquo the typical cloudformation of savanna zones formed by northsouthalignments of cumuliform (convective storm-type)clouds They surplanted other cloud typesprincipally the stratiform types which giverelatively fine regular monsoon-type rains (seeMaley 1982) That these monsoon rains wouldhave dominated the earlier period (between 4000and 30002800 BP Maley 1997) is shown by therelative development of Caesalpinaceae in theforests of low altitude and in the mountains by theextension of Podocarpus which are trees typical ofcloud forests with stratiform clouds (Kerfoot1968 Maley 1996 1997 Maley and Brenac1998a)

Research in dynamic tropical climatology over thelast 20 years shows that Sea Surface Temperature(SST) on the regional and global scale stronglyinfluences climate (Fontaine and Bigot 1993

Moron et al 1995 Bigot et al 1997) It is possibleto characterise the relations between the monsoonrains and the SST more precisely Studies ofclimatic anomalies over tropical Africa occurringsince the beginning of the 1960s have evidencedtwo dominant modes of rainfall distribution eachassociated with a particular distribution of SST Thefirst mode is characterised on the one hand bywarmer than average temperatures in the Gulf ofGuinea and the southern Atlantic and on the otherhand by cooler temperatures on the north tropicalAtlantic offshore of West Africa The second modepresents an opposed distribution for the Africancontinent and the ocean A north-south tropicalAtlantic lsquodipolersquo has thus been found whichoscillates between one mode and the other(Fontaine and Bigot 1993 Wotling et al 1995Bigot et al 1997) An important research result hasbeen to show that over recent millennia the spatialdistribution of paleo-climatic anomalies overtropical Africa (more humid regions and more dryones Figure 6) and of SST on the neighbouringAtlantic is very similar to the spatial distributionobserved for the two dominant anomalies over thelast four decades (Maley 1997 Maley et al 2000)One can deduce that the climatic mechanisms arethe same at the timescale of years centuries ormillennia

So the variation of SST over the tropical Atlanticappears to play a major role in the variation ofclimate and rain types over Central Africa Thephase of forest disturbance that culminated about2500 years BP was associated with relatively lsquowarmrsquoSST over the Gulf of Guinea The earlier phasebetween about 40003800 and 2800 years BP wasassociated with relatively lsquocoolrsquo SST sharplyreduced from the early and middle Holocene (seeMorley and Dworetzky 1993 Figure 7) Over westand south Cameroon these SST allowed relativelyhigh rains and the evolution of stratiform cloudstowards rainy nimbostratus types The palaeo-vegetation data are presently lacking for GabonHowever new research is underway within thePALEOFORGA (Paleoenvironments of GabonForests) programme studying lakes MaridorNguegravene and Kamaleacuteteacute (Figure 1) In contrastfurther south in western Congo the periodstarting from 5000 years BP (Bertaux et al 2000)was marked by the development of semi-deciduousforests (Vincens et al 2000) linked to lower

24

Figure 6 Major rainfall trends in tropical Africa andSouth America for two opposed climatic periods c 3700ndash3000 BP and c 28002500ndash2000 BPPrecipitation trends (anomalies) abovenormalmean + below normalmean ndash P pointsto the sectors where a maximum of Podocarpuspollen intervenedSource Maley 1997 Maley et al 2000

rainfalls indicating that the stratiform clouds hadprobably evolved towards non-rainy clouds Thefour-month dry season here is characterised by thequasi-permanence of these cloud types (Saint Vil1984) which probably gained place progressivelybetween 5000 and 4000 years BP There wastherefore a reinforcing of southern influencesupporting conclusions concerning the history ofOkoumeacute It is notable that the vegetational impactsof the lsquowarmrsquo SST c 28002500 to 2000 years BPwere very different from those earlier c 9000 to65006000 years BP a difference which reflects abrutal strengthening over Central Africa of borealinfluences (Nguetsop et al 2000) to the detrimentof southern influences from about 28002500years BP

Before the major Holocene forest disturbanceculminating about 2500 BP other importantdisturbances occurred during the Quaternary(Maley 1996) The best documented is that betweenabout 20 000 and 15 000 years BP (Maley 19871996 Maley and Brenac 1998a Figure 2) In bothcases residual forest environments (refugia)

persisted apparently in similar locations butperhaps less as small blocks of relativelyhomogeneous forest than as a landscape of forest-savanna mosaic in which forest islands dominated(see Leal 2000) Yet two important characteristicsclearly differentiate these two phases of forestfragmentation First the older disturbance wasmuch longer Second they occurred under verydifferent (even opposed) general climatic conditionsThe disturbance between 20 000 and 15 000 BPcoincided with the development of glacialconditions in middle and high latitudes and coolerconditions by several degrees also affected the wholeregion of the Gulf of Guinea (Maley 1996 Maley andBrenac 1998a) The most recent disturbance bycontrast occurred during the Holocene interglacialwhich on a global scale is characterised by reducedglacial extension and relatively warmertemperatures The period between 2500 and 2000BP is associated with slightly warmer climaticconditions on the regional and global scale (Maley1997 Maley et al 2000) An important conclusion isthat the retreat and fragmentation of the Africanforests can be produced under very different

25

Figure 7 Variations in Sea Surface Temperature (SST) in the South Atlantic Ocean Benguela sector(22ucirc20rsquoSndash11ucirc12rsquoE) estimated with transfer functions based on the abundance of Radiolarian species in thecore RC13ndash228The open circles designate the 6000 and 9000 BP interpolated dates for the two curves one correspondingto the southern summer (February) and the other to the southern winter (August) modern SST areindicated by the solid squares above each curve The dashed lines represent some interpolated dates(italics) 18 000 BP based on the oxygen-isotope stratigraphy 38004000 BP 3000 BP and 2000 BP Theselate Holocene dates are confirmed by an independent study of shells from the same sector (Cohen et al1992) The coherence of the main SST anomalies throughout the Guinea Gulf permits the use of thesecurves in order to interpret the paleoclimatology of central Africa (Maley 1997)Source Morley and Dworetzky 1993

climates either relatively cool or relatively warmand thus with very different climatic situations(Maley 1996 1997)

Climatic models concerning lsquoglobal warmingrsquo inthe context of contemporary climate changeconcerns indicate that an average growth intemperature of about 4ucircC will lead also to anincrease in evaporation of about 30 per cent butonly 12 per cent more rain for tropical Africa (Rind

Notes The author is grateful to James Fairhead for his

translation of this article from the original Frenchand to Melissa Leach and James Fairhead for theirinvitation to the IDS workshop lsquoChangingPerspectives on Forestsrsquo in March 2001 This isInstitut des Sciences de lrsquoEvolution de Montpellier(ISEMCNRS) contribution No 125ndash2001

26

ReferencesAchoundong G Youta Happi J Guillet B

Bonvallot J and Kamgang Beyala V 2000lsquoFormation et eacutevolution des recrus sur savanes (ausud Cameroun)rsquo in M Servant and S Servant-Vildary (eds) Dynamique agrave long terme desEcosystegravemes forestiers intertropicaux MeacutemoireUNESCO Paris UNESCO 31ndash41

Alexandre P 1965 lsquoProto-histoire du groupe Beacuteti-Boulou-Fang essai de synthegravese provisoirersquo CahiersdrsquoEtudes Africaines Vol 20 503ndash60

Amougou Akoa 1986 lsquoEtude botanique eteacutecologique de la valleacutee inondable du Haut-Nyong(sud Cameroun) et de ses affluentsrsquo Thegravese SciencesUniv Yaoundeacute

Aubreacuteville A 1937 lsquoLes forecircts du Dahomey et duTogorsquo Bulletin du Comiteacute drsquoEtudes Historiques etScientifiques de lrsquoAfrique Occidentale Franccedilaises Vol20 112

Aubreacuteville A 1967 lsquoLes eacutetranges mosaiques forecirct-savane du sommet de la boucle de lrsquoOgooueacute auGabonrsquo Adansonia Vol 7 13ndash22

Baeke V 1996 lsquoLe temps des Rites Lrsquounivers magico-religieux des Wuli (Mfumte du Camerounoccidental)rsquo Thegravese Sc Sociale Univ Libre Bruxelles

Barber RJ 1985 lsquoLand snails and past environmentat the Igbo-Iwoto Esie site southwestern NigeriarsquoWest African Journal of Archaeology Vol 15 89ndash102

Barbier JC 1981 lsquoLe peuplement de la partiemeacuteridionale du plateau Bamileacutekeacutersquo in C Tardits

(ed) Contribution de la recherche ethnologique agravelrsquohistoire des civilisations du Cameroun ParisColloque International Centre National deRecherche Scientifique Paris Vol 2 331ndash53

Bertault JG 1990 lsquoComparaison drsquoeacutecosystegravemesforestiers naturels et modifieacutes apregraves incendie enCocircte drsquoIvoirersquo in H Puig (ed) Atelier surlrsquoAmeacutenagement et la Conservation de lrsquoEcosystegravemeForestier Tropical Humide Cayenne ParisUNESCO 6

Bertaux J Schwartz D Vincens A Sifeddine AElenga H Mansour M Mariotti A Fournier MMartin L Wirrmann D and Servant M 2000lsquoEnregistrement de la phase segraveche drsquoAfriqueCentrale vers 3000 ans BP par la spectromeacutetrie IRdans les lacs Sinnda et Kitina (sud-Congo)rsquo in MServant S Servant-Vildary (eds) Dynamique agrave longterme des Ecosystegravemes forestiers intertropicauxMeacutemoire UNESCO Paris UNESCO 43ndash9

Bigot S Camberlin P Moron V and Richard Y1997 lsquoStructures spatiales de la variabiliteacute despreacutecipitations en Afrique une transition climatiqueagrave la fin des anneacutees 1960rsquo Compte Rendu delrsquoAcademie des Sciences Paris 2a 324 181ndash88

Blanc-Pamard C and Peltre P 1984 lsquoDynamique despaysages preacuteforestiers et pratiques culturales enAfrique de lrsquoOuest (Cocircte drsquoIvoire centrale)rsquo LeDeacuteveloppement rural en question Paris MeacutemoireORSTOM No 106 55ndash74

Bonnefille R Chalieacute F Guiot J and Vincens A1992 lsquoQuantitative estimates of full glacial

1995) Though the causes of warming experiencedabout 2500 years BP were natural (Magny 1993)and therefore different from the warmingenvisaged for the twenty-first century it is possiblethat the catastrophic destruction of Central Africanforests which culminated about 2500 years BPcould be an lsquoanaloguersquo and an alarm signal forpotential developments during this current andfuture phase of human-induced global warming(Maley 1997 Maley et al 2000)

1 BP lsquoBefore the Presentrsquo according to standardisedC14 analysis

2 Only 4ndash5 samples have been studied for each 1000years during the Holocene To detect century-scaleperiodicity it would be necessary to study at least 10samples per millennium

temperatures in equatorial Africa frompalynological datarsquo Climate Dynamics Vol 6251ndash57

Brunck F Grison F and Maitre HF 1990Lrsquookoumeacute Aucoumea klaineana Pierre MonographieCentre Techn Forestier Tropical Nogent CIRAD 102

Cohen AI Perkington JE Brundrirt GB and vanDer Merwe NJ 1992 lsquoA Holocene marine climaterecord in Mollusc shells from the southwest Africancoastrsquo Quaternary Research Vol 38 379ndash85

Dechamps R Senut B and Pickford M 1992 Fruitsfossiles pliocegravenes et pleacuteistocegravenes du Rift occidentalougandais Signification paleacuteoenvironnementaleCompte Rendu de lrsquoAcadeacutemie des Sciences Parisseacuterie 2 Vol 314 325ndash331

Dowsett-Lemaire F 1996 lsquoComposition et eacutevolutionde la veacutegeacutetation forestiegravere au Parc NationaldrsquoOdzala Congorsquo Bulletin du Jardin BotaniqueNational de Belgique Vol 65 253ndash92

Dupont LM and Weinelt M 1996 lsquoVegetationhistory of the savanna corridor between theguinean and the Congolian rain forest during thelast 150000 yearsrsquo Vegetation History andArchaeobotany Vol 5 273ndash92

Elenga H Schwartz D and Vincens A 1994lsquoPollen evidence of late Quaternary vegetation andinferred climate changes in Congorsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 109345ndash56

Elenga H Schwartz D Vincens A Bertaux J DeNamur C Martin L Wirrmann D and ServantM 1996 lsquoDiagramme pollinique holocegravene du lacKitina (Congo) mise en eacutevidence de changementspaleacuteobotaniques et paleacuteoclimatiques dans le massifforestier du Mayombersquo Compte Rendu de lrsquoAcadeacutemiedes Sciences Paris Vol 323 No 2a 403ndash10

Elenga H Peyron O Bonnefille R Prentice ICJolly D Cheddadi R Guiot J Andrieu V DeBeaulieu JL Hamilton AC Maley J MarchantR Reille M Scott L Taylor D and Van CampoE 2000 lsquoPollen-based biome reconstruction forsouthern Europe and Africa 18000 years agorsquoJournal of Biogeography Vol 27 621ndash34

Fairhead J and Leach M 1995 lsquoFalse forest historycomplicit social analysis rethinking some westAfrican environmental narrativesrsquo WorldDevelopment Vol 23 1023ndash35

Fairhead J and Leach M 1996 lsquoEnriching thelandscape social history and the management oftransition ecology in the forest-savanna mosaic ofthe Republic of Guinearsquo Africa Vol 66 14ndash36

Fairhead J and Leach M 1998 lsquoReframingdeforestation Global analyses and local realitiesstudies in West Africarsquo Global EnvironmentalChanges Series London Routledge 238

Fay JM 1997 lsquoEvidence of forest degradation innorthern Congo and southwestern Central AfricanRepublic between 2340 and 990 BPrsquo in The ecologysocial organization populations habitat and historyof the western lowland Gorilla PhD ThesisWashington Univ Saint Louis Missouri Ch 6

Fontaine B and Bigot S 1993 lsquoWest African rainfalldeficit and sea surface temperaturesrsquo InternationalJournal of Climatology Vol 13 271ndash85

Gautier L 1990 lsquoContact forecirct-savane en CocirctedrsquoIvoire centrale eacutevolution du recouvrementligneux des savanes de la Reacuteserve de Lamto (sud duV-Baouleacute)rsquo Candollea Vol 45 627ndash41

Giresse P Maley J and Brenac P 1994 lsquoLateQuaternary palaeoenvironments in the lake BarombiMbo (Cameroon) deduced from pollen and carbonisotopes of organic matterrsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 107 65ndash78

Goldammer JG and Seibert B 1990 lsquoThe impact ofdroughts and forest fires on tropical lowland rainforest of East Kalimatanrsquo in JG Goldammer (ed)Fire in the Tropical Biota Ecological Studies BerlinSpringer-Verlag 84

Goucher CL 1981 lsquoIron is iron til it is rust tradeand ecology in the decline of West African iron-smeltingrsquo Journal of African History Vol 22 179ndash89

Guille-Escuret G 1990 lsquoPalmier agrave huile vin depalme et transformations sociales en Lobaye (ForecirctCentrafricaine)rsquo Information sur les Sciences SocialesVol 29 327ndash53

Hart TB 2001 lsquoForest dynamics in the Ituri basin(DR Congo) Dominance diversity andconservationrsquo in W Weber L White A Vedderand L Naughton-Treves (eds) African rain forestecology and conservation An interdisciplinaryperspective Yale Yale University Press 154ndash64

Hart TB Hart JA Dechamps R Fournier M andAtaholo M 1996 lsquoChanges in forest compositionover the last 4000 years in the Ituri basin Zairersquo inLJG Van der Maesen and XM Van der Burgt (eds)The Biodiversity of African plants Dordrecht Kluwer545ndash63

Hawthorne WD 1991 Fire damage and forestregeneration in Ghana Ghana Forestry Departmentamp ODA Report

Jolly D Harrison SP Damnati B and BonnefilleR 1998 lsquoSimulated climate and Biomes of Africa

27

4 The question of savanna firesand the extension of the forestdisturbance in eastern Congo

Observations and opinions have frequentlydiverged concerning the influence of savanna fireson contemporary forest regeneration Observersvariously report that fire can either preventregeneration as in the area of savanna inliers inLopeacute Gabon (Oslisly and White 2000) or merelydelay it as in other parts of the forest region ThusLetouzey (1968) notes for the forest-savannatransition zone of south Cameroon that lsquofire doesnot necessarily prevent the formation of forestrecoveryrsquo Similar conclusions have been drawn inCentral Africa by Sillans and even in westernCongo (ex-Zaire) for the south of the forest bloc(Letouzey 1968) On the same lines Youta Happiand Bonvallot (1996) report lsquothat on theperiphery of a town of 76000 people such asBertoua (east Cameroon) many small savannainlets into forest of two to ten hectares havebecome covered by forest regrowth despite theannual bush fires practiced by Bororopastoralistsrsquo A little further south in the OdzalaReserve situated at the northern extremity of theBateke savannas (north Congo) Dowsett-Lemairereports that lsquoWithout doubt one is in a phase offorest extension despite very frequent bush firesrsquo(1996) This author notes lsquothe fire-break role ofthe boundary thanks to groves dominated byGaertnera paniculata a species absent at Lopeacutewhere colonising species are apparently moresensitive to firersquo Thus the floristic composition ofthe boundary plays an important role in thisphenomenon This was shown in south-eastCameroon by the mid-twentieth century invasionof Chromolaena odorata which colonised theboundaries and which acts there at least as anefficacious fire break It also protects seeds ofwoody species which can germinate and developunder its cover (Youta Happi et al 2000) Otherobservations show how traditional agriculturalpractices and the movement of cattle can befavourable to forest recolonisation in particularby eliminating the grasses which facilitate thepropagation and intensity of fires These processeshave been well described in southern Cameroon(Letouzey 1968 1985) and in Cocircte drsquoIvoire(Spichiger and Blanc-Pamard 1973 Blanc-Pamardand Peltre 1984 Gautier 1990)

These observations indicate that in general andespecially over the long term forest colonisation atthe forest-savanna contact has been the broadlydominant process In certain exceptional yearswhen the dry season has been one or two monthslonger the savanna fires have penetrated moredeeply into the neighbouring forests In early 1983during a two-month longer dry season numerousobservers reported the strong invasion of fire intoforest in Cocircte drsquoIvoire (Bertault 1990) Ghana(Hawthorne 1991 Swaine 1992) and Cameroon(Amougou 1986) Equally fires in Indonesia thatyear were linked to an exceptional El Nintildeo event(Goldammer and Seibert 1990)

These observations indicate how the penetration offire into forest occurs only in very dry years Onemight thus think that fires could have played amajor role in the forest destruction 2500 BP Thishas indeed been shown in eastern BrazilianAmazonia during its major disturbance in themiddle Holocene That fires were frequent then isshown by the numerous charcoal fragmentsdeposited in the lake sediments at Carajas (Martinet al 1993 Servant et al 1993) Yet the lateHolocene lake deposits in Central Africa hardlycontain any charcoal fragments especially for theperiod 3000ndash2000 BP (Giresse et al 1994 Maleyand Brenac 1998a Bertaux et al 2000) It is thesame for the upper horizons of the numerous soilprofiles observed in this region (eg Valleacuterie 1973)In the thalwegs the coarse deposits in the lowerpart of the Lower Terrace do at times containcharcoal or fragments of fossil wood (Maley andBrenac 1998b) however Iron age archaeologicalsites dating from the beginning of the lateHolocene such as in the savanna areas of La Lopeacutealso present wood charcoal (Oslisly and White2000) One can conclude provisionally that thefires that existed were of limited extent

In particular charcoal associated with ironsmelting furnaces cannot have caused the forestdisturbance and cannot have been responsible forthe savanna extension about 2500 BP Goucher(1981) had advanced the hypothesis that ironsmelting could have been a major cause ofdeforestation in West Africa but Fairhead andLeach (1998) have demonstrated that thishypothesis had no serious foundation For CentralAfrica Pinccedilon (1990) also concludes that the

22

extraction of wood for iron metallurgy on theBateke Plateaux was minimal and did not explain(as had been suggested) the presence of theextensive savannas there Instead these savannasresult from an elevated edaphic drought caused byvery sandy soils in which rain infiltrates rapidly toa great depth (Laraque and Pandi 1996)

In contrast to observations from Atlantic CentralAfrica (south Cameroon Gabon and Congo) woodcharcoal has been found in abundance in Ituri nearto Epulu (c 1ucirc20rsquoNndash28ucirc35rsquoE) Here in the heart ofthe forests of the north-east Congo basin about 180km from the nearest savannas 416 soil profileshave been examined Almost all of them containnumerous wood charcoal in their upper levelsbetween the surface and 50 cm deep (Hart et al1996) Virtually all of the 1817 samples of woodcharcoal identified by Deschamps belong to treesfound in the regionrsquos diverse forest formationsexcept for a few pieces of charcoal from twoprofiles which belong to typical wooded savannaspecies The 28 dates obtained for the charcoalshow that fires occurred almost exclusively duringthe last three millennia with a phase of woodedsavanna expansion dating to about 2200 BP (Hartet al 1996) Moreover about 300 km south ofEpulu and 100 km from the forestsavanna limit aroad cutting near Osokari (1ucirc16rsquoSndash27ucirc48rsquoE)revealed a remarkable profile above the principalstone line that could be dated to about 11500years BP (Runge 1996 Maley 1996) A secondgravel level obliquely cutting the top of the soilprofile contains wood charcoal dating from about2200 to 1850 BP These results show that the forestdisturbance in Atlantic Central Africa extended toIturi and affected a large part of the eastern regionof the Congo river basin In contrast with regionsnearer the Atlantic fire had a large role

Also at Ituri (Epulu) Hart et al (1996) show thatcertain trees which are now found there are notfound in the wood charcoal in particularGilbertiodendron dewevrei (Caesalpiniaceae) which isnow abundant (almost mono-dominant) in certainparts of the forest The range of forests dominatedby Gilbertiodendron dewevrei extends to the east andnorth of the Congo basin as far as east Cameroon(Leacuteonard 1953) It is thus possible that thisCaesalpiniaceae has very recently colonised the Ituriforests and that the large area it now occupies is a

legacy of the disturbance about 2500 BP Ifconfirmed this area could correspond at leastpartly to an older forest area (of unknowncharacter) which was destroyed by the disturbanceA comparison with Okoumeacute can be suggestedwhich as indicated for Gabon above alsodominates areas which it has colonised since lessthan 2000 years ago However as the forest agesOkoumeacute loses its dominant character whereas inIturi Gilbertiodendron dewevrei remains dominantprobably because it can regenerate abundantly inthe shade of parent trees and also because itsreproductive character gives it an advantage overother trees in this formation (see Hart 2001)

5 Conclusion the major role ofclimatic conditions and theirhistoryThe present mosaic pattern of numerous forestscharacterised by a mix or juxtaposition ofgroupings of evergreen and deciduous species isprobably the consequence of long-term disturbanceswhich have affected the forest domain during thelast three millennia particularly the majordisturbance which culminated about 2500 years BPThe synchronism apparent in the disturbance fordifferent sites studied across Central Africa (southCameroon Gabon Congo as far east as the CongoRiver basin) and its association with a generalisederosive phase enables one to conclude that it wasthe result of a major climatic change

The climate at this time appears to have beenrelatively arid as it led to forest destruction and inplaces the extension of savannas Neverthelessother characteristics such as the strongsynchronous extension of diverse pioneer taxaindicate that in places there was not truly areduction in rainfall but rather a change in itsannual distribution The absence of a fall in annualrainfall has been demonstrated at lake BarombiMbo (Maley and Brenac 1998a) and at lake Ossawhere diatom studies show that there was even anet rise in this lake between 2500 and 2200 yearsBP with only a short fall between 2200 and 2000years BP (Nguetsop et al 2000) It is thereforepreferable to designate this catastrophic period asan lsquounfavourable climatic disturbancersquo caused by anincrease in seasonality and a lengthening of the dryseason (Maley 1997) The strong soil erosion of this

23

period suggests heavy rains but concentrated oversix or seven months of the year as is now found inthe peripheral savanna zones The growth inseasonality could have been associated with adominance of lsquosquall linesrsquo the typical cloudformation of savanna zones formed by northsouthalignments of cumuliform (convective storm-type)clouds They surplanted other cloud typesprincipally the stratiform types which giverelatively fine regular monsoon-type rains (seeMaley 1982) That these monsoon rains wouldhave dominated the earlier period (between 4000and 30002800 BP Maley 1997) is shown by therelative development of Caesalpinaceae in theforests of low altitude and in the mountains by theextension of Podocarpus which are trees typical ofcloud forests with stratiform clouds (Kerfoot1968 Maley 1996 1997 Maley and Brenac1998a)

Research in dynamic tropical climatology over thelast 20 years shows that Sea Surface Temperature(SST) on the regional and global scale stronglyinfluences climate (Fontaine and Bigot 1993

Moron et al 1995 Bigot et al 1997) It is possibleto characterise the relations between the monsoonrains and the SST more precisely Studies ofclimatic anomalies over tropical Africa occurringsince the beginning of the 1960s have evidencedtwo dominant modes of rainfall distribution eachassociated with a particular distribution of SST Thefirst mode is characterised on the one hand bywarmer than average temperatures in the Gulf ofGuinea and the southern Atlantic and on the otherhand by cooler temperatures on the north tropicalAtlantic offshore of West Africa The second modepresents an opposed distribution for the Africancontinent and the ocean A north-south tropicalAtlantic lsquodipolersquo has thus been found whichoscillates between one mode and the other(Fontaine and Bigot 1993 Wotling et al 1995Bigot et al 1997) An important research result hasbeen to show that over recent millennia the spatialdistribution of paleo-climatic anomalies overtropical Africa (more humid regions and more dryones Figure 6) and of SST on the neighbouringAtlantic is very similar to the spatial distributionobserved for the two dominant anomalies over thelast four decades (Maley 1997 Maley et al 2000)One can deduce that the climatic mechanisms arethe same at the timescale of years centuries ormillennia

So the variation of SST over the tropical Atlanticappears to play a major role in the variation ofclimate and rain types over Central Africa Thephase of forest disturbance that culminated about2500 years BP was associated with relatively lsquowarmrsquoSST over the Gulf of Guinea The earlier phasebetween about 40003800 and 2800 years BP wasassociated with relatively lsquocoolrsquo SST sharplyreduced from the early and middle Holocene (seeMorley and Dworetzky 1993 Figure 7) Over westand south Cameroon these SST allowed relativelyhigh rains and the evolution of stratiform cloudstowards rainy nimbostratus types The palaeo-vegetation data are presently lacking for GabonHowever new research is underway within thePALEOFORGA (Paleoenvironments of GabonForests) programme studying lakes MaridorNguegravene and Kamaleacuteteacute (Figure 1) In contrastfurther south in western Congo the periodstarting from 5000 years BP (Bertaux et al 2000)was marked by the development of semi-deciduousforests (Vincens et al 2000) linked to lower

24

Figure 6 Major rainfall trends in tropical Africa andSouth America for two opposed climatic periods c 3700ndash3000 BP and c 28002500ndash2000 BPPrecipitation trends (anomalies) abovenormalmean + below normalmean ndash P pointsto the sectors where a maximum of Podocarpuspollen intervenedSource Maley 1997 Maley et al 2000

rainfalls indicating that the stratiform clouds hadprobably evolved towards non-rainy clouds Thefour-month dry season here is characterised by thequasi-permanence of these cloud types (Saint Vil1984) which probably gained place progressivelybetween 5000 and 4000 years BP There wastherefore a reinforcing of southern influencesupporting conclusions concerning the history ofOkoumeacute It is notable that the vegetational impactsof the lsquowarmrsquo SST c 28002500 to 2000 years BPwere very different from those earlier c 9000 to65006000 years BP a difference which reflects abrutal strengthening over Central Africa of borealinfluences (Nguetsop et al 2000) to the detrimentof southern influences from about 28002500years BP

Before the major Holocene forest disturbanceculminating about 2500 BP other importantdisturbances occurred during the Quaternary(Maley 1996) The best documented is that betweenabout 20 000 and 15 000 years BP (Maley 19871996 Maley and Brenac 1998a Figure 2) In bothcases residual forest environments (refugia)

persisted apparently in similar locations butperhaps less as small blocks of relativelyhomogeneous forest than as a landscape of forest-savanna mosaic in which forest islands dominated(see Leal 2000) Yet two important characteristicsclearly differentiate these two phases of forestfragmentation First the older disturbance wasmuch longer Second they occurred under verydifferent (even opposed) general climatic conditionsThe disturbance between 20 000 and 15 000 BPcoincided with the development of glacialconditions in middle and high latitudes and coolerconditions by several degrees also affected the wholeregion of the Gulf of Guinea (Maley 1996 Maley andBrenac 1998a) The most recent disturbance bycontrast occurred during the Holocene interglacialwhich on a global scale is characterised by reducedglacial extension and relatively warmertemperatures The period between 2500 and 2000BP is associated with slightly warmer climaticconditions on the regional and global scale (Maley1997 Maley et al 2000) An important conclusion isthat the retreat and fragmentation of the Africanforests can be produced under very different

25

Figure 7 Variations in Sea Surface Temperature (SST) in the South Atlantic Ocean Benguela sector(22ucirc20rsquoSndash11ucirc12rsquoE) estimated with transfer functions based on the abundance of Radiolarian species in thecore RC13ndash228The open circles designate the 6000 and 9000 BP interpolated dates for the two curves one correspondingto the southern summer (February) and the other to the southern winter (August) modern SST areindicated by the solid squares above each curve The dashed lines represent some interpolated dates(italics) 18 000 BP based on the oxygen-isotope stratigraphy 38004000 BP 3000 BP and 2000 BP Theselate Holocene dates are confirmed by an independent study of shells from the same sector (Cohen et al1992) The coherence of the main SST anomalies throughout the Guinea Gulf permits the use of thesecurves in order to interpret the paleoclimatology of central Africa (Maley 1997)Source Morley and Dworetzky 1993

climates either relatively cool or relatively warmand thus with very different climatic situations(Maley 1996 1997)

Climatic models concerning lsquoglobal warmingrsquo inthe context of contemporary climate changeconcerns indicate that an average growth intemperature of about 4ucircC will lead also to anincrease in evaporation of about 30 per cent butonly 12 per cent more rain for tropical Africa (Rind

Notes The author is grateful to James Fairhead for his

translation of this article from the original Frenchand to Melissa Leach and James Fairhead for theirinvitation to the IDS workshop lsquoChangingPerspectives on Forestsrsquo in March 2001 This isInstitut des Sciences de lrsquoEvolution de Montpellier(ISEMCNRS) contribution No 125ndash2001

26

ReferencesAchoundong G Youta Happi J Guillet B

Bonvallot J and Kamgang Beyala V 2000lsquoFormation et eacutevolution des recrus sur savanes (ausud Cameroun)rsquo in M Servant and S Servant-Vildary (eds) Dynamique agrave long terme desEcosystegravemes forestiers intertropicaux MeacutemoireUNESCO Paris UNESCO 31ndash41

Alexandre P 1965 lsquoProto-histoire du groupe Beacuteti-Boulou-Fang essai de synthegravese provisoirersquo CahiersdrsquoEtudes Africaines Vol 20 503ndash60

Amougou Akoa 1986 lsquoEtude botanique eteacutecologique de la valleacutee inondable du Haut-Nyong(sud Cameroun) et de ses affluentsrsquo Thegravese SciencesUniv Yaoundeacute

Aubreacuteville A 1937 lsquoLes forecircts du Dahomey et duTogorsquo Bulletin du Comiteacute drsquoEtudes Historiques etScientifiques de lrsquoAfrique Occidentale Franccedilaises Vol20 112

Aubreacuteville A 1967 lsquoLes eacutetranges mosaiques forecirct-savane du sommet de la boucle de lrsquoOgooueacute auGabonrsquo Adansonia Vol 7 13ndash22

Baeke V 1996 lsquoLe temps des Rites Lrsquounivers magico-religieux des Wuli (Mfumte du Camerounoccidental)rsquo Thegravese Sc Sociale Univ Libre Bruxelles

Barber RJ 1985 lsquoLand snails and past environmentat the Igbo-Iwoto Esie site southwestern NigeriarsquoWest African Journal of Archaeology Vol 15 89ndash102

Barbier JC 1981 lsquoLe peuplement de la partiemeacuteridionale du plateau Bamileacutekeacutersquo in C Tardits

(ed) Contribution de la recherche ethnologique agravelrsquohistoire des civilisations du Cameroun ParisColloque International Centre National deRecherche Scientifique Paris Vol 2 331ndash53

Bertault JG 1990 lsquoComparaison drsquoeacutecosystegravemesforestiers naturels et modifieacutes apregraves incendie enCocircte drsquoIvoirersquo in H Puig (ed) Atelier surlrsquoAmeacutenagement et la Conservation de lrsquoEcosystegravemeForestier Tropical Humide Cayenne ParisUNESCO 6

Bertaux J Schwartz D Vincens A Sifeddine AElenga H Mansour M Mariotti A Fournier MMartin L Wirrmann D and Servant M 2000lsquoEnregistrement de la phase segraveche drsquoAfriqueCentrale vers 3000 ans BP par la spectromeacutetrie IRdans les lacs Sinnda et Kitina (sud-Congo)rsquo in MServant S Servant-Vildary (eds) Dynamique agrave longterme des Ecosystegravemes forestiers intertropicauxMeacutemoire UNESCO Paris UNESCO 43ndash9

Bigot S Camberlin P Moron V and Richard Y1997 lsquoStructures spatiales de la variabiliteacute despreacutecipitations en Afrique une transition climatiqueagrave la fin des anneacutees 1960rsquo Compte Rendu delrsquoAcademie des Sciences Paris 2a 324 181ndash88

Blanc-Pamard C and Peltre P 1984 lsquoDynamique despaysages preacuteforestiers et pratiques culturales enAfrique de lrsquoOuest (Cocircte drsquoIvoire centrale)rsquo LeDeacuteveloppement rural en question Paris MeacutemoireORSTOM No 106 55ndash74

Bonnefille R Chalieacute F Guiot J and Vincens A1992 lsquoQuantitative estimates of full glacial

1995) Though the causes of warming experiencedabout 2500 years BP were natural (Magny 1993)and therefore different from the warmingenvisaged for the twenty-first century it is possiblethat the catastrophic destruction of Central Africanforests which culminated about 2500 years BPcould be an lsquoanaloguersquo and an alarm signal forpotential developments during this current andfuture phase of human-induced global warming(Maley 1997 Maley et al 2000)

1 BP lsquoBefore the Presentrsquo according to standardisedC14 analysis

2 Only 4ndash5 samples have been studied for each 1000years during the Holocene To detect century-scaleperiodicity it would be necessary to study at least 10samples per millennium

temperatures in equatorial Africa frompalynological datarsquo Climate Dynamics Vol 6251ndash57

Brunck F Grison F and Maitre HF 1990Lrsquookoumeacute Aucoumea klaineana Pierre MonographieCentre Techn Forestier Tropical Nogent CIRAD 102

Cohen AI Perkington JE Brundrirt GB and vanDer Merwe NJ 1992 lsquoA Holocene marine climaterecord in Mollusc shells from the southwest Africancoastrsquo Quaternary Research Vol 38 379ndash85

Dechamps R Senut B and Pickford M 1992 Fruitsfossiles pliocegravenes et pleacuteistocegravenes du Rift occidentalougandais Signification paleacuteoenvironnementaleCompte Rendu de lrsquoAcadeacutemie des Sciences Parisseacuterie 2 Vol 314 325ndash331

Dowsett-Lemaire F 1996 lsquoComposition et eacutevolutionde la veacutegeacutetation forestiegravere au Parc NationaldrsquoOdzala Congorsquo Bulletin du Jardin BotaniqueNational de Belgique Vol 65 253ndash92

Dupont LM and Weinelt M 1996 lsquoVegetationhistory of the savanna corridor between theguinean and the Congolian rain forest during thelast 150000 yearsrsquo Vegetation History andArchaeobotany Vol 5 273ndash92

Elenga H Schwartz D and Vincens A 1994lsquoPollen evidence of late Quaternary vegetation andinferred climate changes in Congorsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 109345ndash56

Elenga H Schwartz D Vincens A Bertaux J DeNamur C Martin L Wirrmann D and ServantM 1996 lsquoDiagramme pollinique holocegravene du lacKitina (Congo) mise en eacutevidence de changementspaleacuteobotaniques et paleacuteoclimatiques dans le massifforestier du Mayombersquo Compte Rendu de lrsquoAcadeacutemiedes Sciences Paris Vol 323 No 2a 403ndash10

Elenga H Peyron O Bonnefille R Prentice ICJolly D Cheddadi R Guiot J Andrieu V DeBeaulieu JL Hamilton AC Maley J MarchantR Reille M Scott L Taylor D and Van CampoE 2000 lsquoPollen-based biome reconstruction forsouthern Europe and Africa 18000 years agorsquoJournal of Biogeography Vol 27 621ndash34

Fairhead J and Leach M 1995 lsquoFalse forest historycomplicit social analysis rethinking some westAfrican environmental narrativesrsquo WorldDevelopment Vol 23 1023ndash35

Fairhead J and Leach M 1996 lsquoEnriching thelandscape social history and the management oftransition ecology in the forest-savanna mosaic ofthe Republic of Guinearsquo Africa Vol 66 14ndash36

Fairhead J and Leach M 1998 lsquoReframingdeforestation Global analyses and local realitiesstudies in West Africarsquo Global EnvironmentalChanges Series London Routledge 238

Fay JM 1997 lsquoEvidence of forest degradation innorthern Congo and southwestern Central AfricanRepublic between 2340 and 990 BPrsquo in The ecologysocial organization populations habitat and historyof the western lowland Gorilla PhD ThesisWashington Univ Saint Louis Missouri Ch 6

Fontaine B and Bigot S 1993 lsquoWest African rainfalldeficit and sea surface temperaturesrsquo InternationalJournal of Climatology Vol 13 271ndash85

Gautier L 1990 lsquoContact forecirct-savane en CocirctedrsquoIvoire centrale eacutevolution du recouvrementligneux des savanes de la Reacuteserve de Lamto (sud duV-Baouleacute)rsquo Candollea Vol 45 627ndash41

Giresse P Maley J and Brenac P 1994 lsquoLateQuaternary palaeoenvironments in the lake BarombiMbo (Cameroon) deduced from pollen and carbonisotopes of organic matterrsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 107 65ndash78

Goldammer JG and Seibert B 1990 lsquoThe impact ofdroughts and forest fires on tropical lowland rainforest of East Kalimatanrsquo in JG Goldammer (ed)Fire in the Tropical Biota Ecological Studies BerlinSpringer-Verlag 84

Goucher CL 1981 lsquoIron is iron til it is rust tradeand ecology in the decline of West African iron-smeltingrsquo Journal of African History Vol 22 179ndash89

Guille-Escuret G 1990 lsquoPalmier agrave huile vin depalme et transformations sociales en Lobaye (ForecirctCentrafricaine)rsquo Information sur les Sciences SocialesVol 29 327ndash53

Hart TB 2001 lsquoForest dynamics in the Ituri basin(DR Congo) Dominance diversity andconservationrsquo in W Weber L White A Vedderand L Naughton-Treves (eds) African rain forestecology and conservation An interdisciplinaryperspective Yale Yale University Press 154ndash64

Hart TB Hart JA Dechamps R Fournier M andAtaholo M 1996 lsquoChanges in forest compositionover the last 4000 years in the Ituri basin Zairersquo inLJG Van der Maesen and XM Van der Burgt (eds)The Biodiversity of African plants Dordrecht Kluwer545ndash63

Hawthorne WD 1991 Fire damage and forestregeneration in Ghana Ghana Forestry Departmentamp ODA Report

Jolly D Harrison SP Damnati B and BonnefilleR 1998 lsquoSimulated climate and Biomes of Africa

27

extraction of wood for iron metallurgy on theBateke Plateaux was minimal and did not explain(as had been suggested) the presence of theextensive savannas there Instead these savannasresult from an elevated edaphic drought caused byvery sandy soils in which rain infiltrates rapidly toa great depth (Laraque and Pandi 1996)

In contrast to observations from Atlantic CentralAfrica (south Cameroon Gabon and Congo) woodcharcoal has been found in abundance in Ituri nearto Epulu (c 1ucirc20rsquoNndash28ucirc35rsquoE) Here in the heart ofthe forests of the north-east Congo basin about 180km from the nearest savannas 416 soil profileshave been examined Almost all of them containnumerous wood charcoal in their upper levelsbetween the surface and 50 cm deep (Hart et al1996) Virtually all of the 1817 samples of woodcharcoal identified by Deschamps belong to treesfound in the regionrsquos diverse forest formationsexcept for a few pieces of charcoal from twoprofiles which belong to typical wooded savannaspecies The 28 dates obtained for the charcoalshow that fires occurred almost exclusively duringthe last three millennia with a phase of woodedsavanna expansion dating to about 2200 BP (Hartet al 1996) Moreover about 300 km south ofEpulu and 100 km from the forestsavanna limit aroad cutting near Osokari (1ucirc16rsquoSndash27ucirc48rsquoE)revealed a remarkable profile above the principalstone line that could be dated to about 11500years BP (Runge 1996 Maley 1996) A secondgravel level obliquely cutting the top of the soilprofile contains wood charcoal dating from about2200 to 1850 BP These results show that the forestdisturbance in Atlantic Central Africa extended toIturi and affected a large part of the eastern regionof the Congo river basin In contrast with regionsnearer the Atlantic fire had a large role

Also at Ituri (Epulu) Hart et al (1996) show thatcertain trees which are now found there are notfound in the wood charcoal in particularGilbertiodendron dewevrei (Caesalpiniaceae) which isnow abundant (almost mono-dominant) in certainparts of the forest The range of forests dominatedby Gilbertiodendron dewevrei extends to the east andnorth of the Congo basin as far as east Cameroon(Leacuteonard 1953) It is thus possible that thisCaesalpiniaceae has very recently colonised the Ituriforests and that the large area it now occupies is a

legacy of the disturbance about 2500 BP Ifconfirmed this area could correspond at leastpartly to an older forest area (of unknowncharacter) which was destroyed by the disturbanceA comparison with Okoumeacute can be suggestedwhich as indicated for Gabon above alsodominates areas which it has colonised since lessthan 2000 years ago However as the forest agesOkoumeacute loses its dominant character whereas inIturi Gilbertiodendron dewevrei remains dominantprobably because it can regenerate abundantly inthe shade of parent trees and also because itsreproductive character gives it an advantage overother trees in this formation (see Hart 2001)

5 Conclusion the major role ofclimatic conditions and theirhistoryThe present mosaic pattern of numerous forestscharacterised by a mix or juxtaposition ofgroupings of evergreen and deciduous species isprobably the consequence of long-term disturbanceswhich have affected the forest domain during thelast three millennia particularly the majordisturbance which culminated about 2500 years BPThe synchronism apparent in the disturbance fordifferent sites studied across Central Africa (southCameroon Gabon Congo as far east as the CongoRiver basin) and its association with a generalisederosive phase enables one to conclude that it wasthe result of a major climatic change

The climate at this time appears to have beenrelatively arid as it led to forest destruction and inplaces the extension of savannas Neverthelessother characteristics such as the strongsynchronous extension of diverse pioneer taxaindicate that in places there was not truly areduction in rainfall but rather a change in itsannual distribution The absence of a fall in annualrainfall has been demonstrated at lake BarombiMbo (Maley and Brenac 1998a) and at lake Ossawhere diatom studies show that there was even anet rise in this lake between 2500 and 2200 yearsBP with only a short fall between 2200 and 2000years BP (Nguetsop et al 2000) It is thereforepreferable to designate this catastrophic period asan lsquounfavourable climatic disturbancersquo caused by anincrease in seasonality and a lengthening of the dryseason (Maley 1997) The strong soil erosion of this

23

period suggests heavy rains but concentrated oversix or seven months of the year as is now found inthe peripheral savanna zones The growth inseasonality could have been associated with adominance of lsquosquall linesrsquo the typical cloudformation of savanna zones formed by northsouthalignments of cumuliform (convective storm-type)clouds They surplanted other cloud typesprincipally the stratiform types which giverelatively fine regular monsoon-type rains (seeMaley 1982) That these monsoon rains wouldhave dominated the earlier period (between 4000and 30002800 BP Maley 1997) is shown by therelative development of Caesalpinaceae in theforests of low altitude and in the mountains by theextension of Podocarpus which are trees typical ofcloud forests with stratiform clouds (Kerfoot1968 Maley 1996 1997 Maley and Brenac1998a)

Research in dynamic tropical climatology over thelast 20 years shows that Sea Surface Temperature(SST) on the regional and global scale stronglyinfluences climate (Fontaine and Bigot 1993

Moron et al 1995 Bigot et al 1997) It is possibleto characterise the relations between the monsoonrains and the SST more precisely Studies ofclimatic anomalies over tropical Africa occurringsince the beginning of the 1960s have evidencedtwo dominant modes of rainfall distribution eachassociated with a particular distribution of SST Thefirst mode is characterised on the one hand bywarmer than average temperatures in the Gulf ofGuinea and the southern Atlantic and on the otherhand by cooler temperatures on the north tropicalAtlantic offshore of West Africa The second modepresents an opposed distribution for the Africancontinent and the ocean A north-south tropicalAtlantic lsquodipolersquo has thus been found whichoscillates between one mode and the other(Fontaine and Bigot 1993 Wotling et al 1995Bigot et al 1997) An important research result hasbeen to show that over recent millennia the spatialdistribution of paleo-climatic anomalies overtropical Africa (more humid regions and more dryones Figure 6) and of SST on the neighbouringAtlantic is very similar to the spatial distributionobserved for the two dominant anomalies over thelast four decades (Maley 1997 Maley et al 2000)One can deduce that the climatic mechanisms arethe same at the timescale of years centuries ormillennia

So the variation of SST over the tropical Atlanticappears to play a major role in the variation ofclimate and rain types over Central Africa Thephase of forest disturbance that culminated about2500 years BP was associated with relatively lsquowarmrsquoSST over the Gulf of Guinea The earlier phasebetween about 40003800 and 2800 years BP wasassociated with relatively lsquocoolrsquo SST sharplyreduced from the early and middle Holocene (seeMorley and Dworetzky 1993 Figure 7) Over westand south Cameroon these SST allowed relativelyhigh rains and the evolution of stratiform cloudstowards rainy nimbostratus types The palaeo-vegetation data are presently lacking for GabonHowever new research is underway within thePALEOFORGA (Paleoenvironments of GabonForests) programme studying lakes MaridorNguegravene and Kamaleacuteteacute (Figure 1) In contrastfurther south in western Congo the periodstarting from 5000 years BP (Bertaux et al 2000)was marked by the development of semi-deciduousforests (Vincens et al 2000) linked to lower

24

Figure 6 Major rainfall trends in tropical Africa andSouth America for two opposed climatic periods c 3700ndash3000 BP and c 28002500ndash2000 BPPrecipitation trends (anomalies) abovenormalmean + below normalmean ndash P pointsto the sectors where a maximum of Podocarpuspollen intervenedSource Maley 1997 Maley et al 2000

rainfalls indicating that the stratiform clouds hadprobably evolved towards non-rainy clouds Thefour-month dry season here is characterised by thequasi-permanence of these cloud types (Saint Vil1984) which probably gained place progressivelybetween 5000 and 4000 years BP There wastherefore a reinforcing of southern influencesupporting conclusions concerning the history ofOkoumeacute It is notable that the vegetational impactsof the lsquowarmrsquo SST c 28002500 to 2000 years BPwere very different from those earlier c 9000 to65006000 years BP a difference which reflects abrutal strengthening over Central Africa of borealinfluences (Nguetsop et al 2000) to the detrimentof southern influences from about 28002500years BP

Before the major Holocene forest disturbanceculminating about 2500 BP other importantdisturbances occurred during the Quaternary(Maley 1996) The best documented is that betweenabout 20 000 and 15 000 years BP (Maley 19871996 Maley and Brenac 1998a Figure 2) In bothcases residual forest environments (refugia)

persisted apparently in similar locations butperhaps less as small blocks of relativelyhomogeneous forest than as a landscape of forest-savanna mosaic in which forest islands dominated(see Leal 2000) Yet two important characteristicsclearly differentiate these two phases of forestfragmentation First the older disturbance wasmuch longer Second they occurred under verydifferent (even opposed) general climatic conditionsThe disturbance between 20 000 and 15 000 BPcoincided with the development of glacialconditions in middle and high latitudes and coolerconditions by several degrees also affected the wholeregion of the Gulf of Guinea (Maley 1996 Maley andBrenac 1998a) The most recent disturbance bycontrast occurred during the Holocene interglacialwhich on a global scale is characterised by reducedglacial extension and relatively warmertemperatures The period between 2500 and 2000BP is associated with slightly warmer climaticconditions on the regional and global scale (Maley1997 Maley et al 2000) An important conclusion isthat the retreat and fragmentation of the Africanforests can be produced under very different

25

Figure 7 Variations in Sea Surface Temperature (SST) in the South Atlantic Ocean Benguela sector(22ucirc20rsquoSndash11ucirc12rsquoE) estimated with transfer functions based on the abundance of Radiolarian species in thecore RC13ndash228The open circles designate the 6000 and 9000 BP interpolated dates for the two curves one correspondingto the southern summer (February) and the other to the southern winter (August) modern SST areindicated by the solid squares above each curve The dashed lines represent some interpolated dates(italics) 18 000 BP based on the oxygen-isotope stratigraphy 38004000 BP 3000 BP and 2000 BP Theselate Holocene dates are confirmed by an independent study of shells from the same sector (Cohen et al1992) The coherence of the main SST anomalies throughout the Guinea Gulf permits the use of thesecurves in order to interpret the paleoclimatology of central Africa (Maley 1997)Source Morley and Dworetzky 1993

climates either relatively cool or relatively warmand thus with very different climatic situations(Maley 1996 1997)

Climatic models concerning lsquoglobal warmingrsquo inthe context of contemporary climate changeconcerns indicate that an average growth intemperature of about 4ucircC will lead also to anincrease in evaporation of about 30 per cent butonly 12 per cent more rain for tropical Africa (Rind

Notes The author is grateful to James Fairhead for his

translation of this article from the original Frenchand to Melissa Leach and James Fairhead for theirinvitation to the IDS workshop lsquoChangingPerspectives on Forestsrsquo in March 2001 This isInstitut des Sciences de lrsquoEvolution de Montpellier(ISEMCNRS) contribution No 125ndash2001

26

ReferencesAchoundong G Youta Happi J Guillet B

Bonvallot J and Kamgang Beyala V 2000lsquoFormation et eacutevolution des recrus sur savanes (ausud Cameroun)rsquo in M Servant and S Servant-Vildary (eds) Dynamique agrave long terme desEcosystegravemes forestiers intertropicaux MeacutemoireUNESCO Paris UNESCO 31ndash41

Alexandre P 1965 lsquoProto-histoire du groupe Beacuteti-Boulou-Fang essai de synthegravese provisoirersquo CahiersdrsquoEtudes Africaines Vol 20 503ndash60

Amougou Akoa 1986 lsquoEtude botanique eteacutecologique de la valleacutee inondable du Haut-Nyong(sud Cameroun) et de ses affluentsrsquo Thegravese SciencesUniv Yaoundeacute

Aubreacuteville A 1937 lsquoLes forecircts du Dahomey et duTogorsquo Bulletin du Comiteacute drsquoEtudes Historiques etScientifiques de lrsquoAfrique Occidentale Franccedilaises Vol20 112

Aubreacuteville A 1967 lsquoLes eacutetranges mosaiques forecirct-savane du sommet de la boucle de lrsquoOgooueacute auGabonrsquo Adansonia Vol 7 13ndash22

Baeke V 1996 lsquoLe temps des Rites Lrsquounivers magico-religieux des Wuli (Mfumte du Camerounoccidental)rsquo Thegravese Sc Sociale Univ Libre Bruxelles

Barber RJ 1985 lsquoLand snails and past environmentat the Igbo-Iwoto Esie site southwestern NigeriarsquoWest African Journal of Archaeology Vol 15 89ndash102

Barbier JC 1981 lsquoLe peuplement de la partiemeacuteridionale du plateau Bamileacutekeacutersquo in C Tardits

(ed) Contribution de la recherche ethnologique agravelrsquohistoire des civilisations du Cameroun ParisColloque International Centre National deRecherche Scientifique Paris Vol 2 331ndash53

Bertault JG 1990 lsquoComparaison drsquoeacutecosystegravemesforestiers naturels et modifieacutes apregraves incendie enCocircte drsquoIvoirersquo in H Puig (ed) Atelier surlrsquoAmeacutenagement et la Conservation de lrsquoEcosystegravemeForestier Tropical Humide Cayenne ParisUNESCO 6

Bertaux J Schwartz D Vincens A Sifeddine AElenga H Mansour M Mariotti A Fournier MMartin L Wirrmann D and Servant M 2000lsquoEnregistrement de la phase segraveche drsquoAfriqueCentrale vers 3000 ans BP par la spectromeacutetrie IRdans les lacs Sinnda et Kitina (sud-Congo)rsquo in MServant S Servant-Vildary (eds) Dynamique agrave longterme des Ecosystegravemes forestiers intertropicauxMeacutemoire UNESCO Paris UNESCO 43ndash9

Bigot S Camberlin P Moron V and Richard Y1997 lsquoStructures spatiales de la variabiliteacute despreacutecipitations en Afrique une transition climatiqueagrave la fin des anneacutees 1960rsquo Compte Rendu delrsquoAcademie des Sciences Paris 2a 324 181ndash88

Blanc-Pamard C and Peltre P 1984 lsquoDynamique despaysages preacuteforestiers et pratiques culturales enAfrique de lrsquoOuest (Cocircte drsquoIvoire centrale)rsquo LeDeacuteveloppement rural en question Paris MeacutemoireORSTOM No 106 55ndash74

Bonnefille R Chalieacute F Guiot J and Vincens A1992 lsquoQuantitative estimates of full glacial

1995) Though the causes of warming experiencedabout 2500 years BP were natural (Magny 1993)and therefore different from the warmingenvisaged for the twenty-first century it is possiblethat the catastrophic destruction of Central Africanforests which culminated about 2500 years BPcould be an lsquoanaloguersquo and an alarm signal forpotential developments during this current andfuture phase of human-induced global warming(Maley 1997 Maley et al 2000)

1 BP lsquoBefore the Presentrsquo according to standardisedC14 analysis

2 Only 4ndash5 samples have been studied for each 1000years during the Holocene To detect century-scaleperiodicity it would be necessary to study at least 10samples per millennium

temperatures in equatorial Africa frompalynological datarsquo Climate Dynamics Vol 6251ndash57

Brunck F Grison F and Maitre HF 1990Lrsquookoumeacute Aucoumea klaineana Pierre MonographieCentre Techn Forestier Tropical Nogent CIRAD 102

Cohen AI Perkington JE Brundrirt GB and vanDer Merwe NJ 1992 lsquoA Holocene marine climaterecord in Mollusc shells from the southwest Africancoastrsquo Quaternary Research Vol 38 379ndash85

Dechamps R Senut B and Pickford M 1992 Fruitsfossiles pliocegravenes et pleacuteistocegravenes du Rift occidentalougandais Signification paleacuteoenvironnementaleCompte Rendu de lrsquoAcadeacutemie des Sciences Parisseacuterie 2 Vol 314 325ndash331

Dowsett-Lemaire F 1996 lsquoComposition et eacutevolutionde la veacutegeacutetation forestiegravere au Parc NationaldrsquoOdzala Congorsquo Bulletin du Jardin BotaniqueNational de Belgique Vol 65 253ndash92

Dupont LM and Weinelt M 1996 lsquoVegetationhistory of the savanna corridor between theguinean and the Congolian rain forest during thelast 150000 yearsrsquo Vegetation History andArchaeobotany Vol 5 273ndash92

Elenga H Schwartz D and Vincens A 1994lsquoPollen evidence of late Quaternary vegetation andinferred climate changes in Congorsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 109345ndash56

Elenga H Schwartz D Vincens A Bertaux J DeNamur C Martin L Wirrmann D and ServantM 1996 lsquoDiagramme pollinique holocegravene du lacKitina (Congo) mise en eacutevidence de changementspaleacuteobotaniques et paleacuteoclimatiques dans le massifforestier du Mayombersquo Compte Rendu de lrsquoAcadeacutemiedes Sciences Paris Vol 323 No 2a 403ndash10

Elenga H Peyron O Bonnefille R Prentice ICJolly D Cheddadi R Guiot J Andrieu V DeBeaulieu JL Hamilton AC Maley J MarchantR Reille M Scott L Taylor D and Van CampoE 2000 lsquoPollen-based biome reconstruction forsouthern Europe and Africa 18000 years agorsquoJournal of Biogeography Vol 27 621ndash34

Fairhead J and Leach M 1995 lsquoFalse forest historycomplicit social analysis rethinking some westAfrican environmental narrativesrsquo WorldDevelopment Vol 23 1023ndash35

Fairhead J and Leach M 1996 lsquoEnriching thelandscape social history and the management oftransition ecology in the forest-savanna mosaic ofthe Republic of Guinearsquo Africa Vol 66 14ndash36

Fairhead J and Leach M 1998 lsquoReframingdeforestation Global analyses and local realitiesstudies in West Africarsquo Global EnvironmentalChanges Series London Routledge 238

Fay JM 1997 lsquoEvidence of forest degradation innorthern Congo and southwestern Central AfricanRepublic between 2340 and 990 BPrsquo in The ecologysocial organization populations habitat and historyof the western lowland Gorilla PhD ThesisWashington Univ Saint Louis Missouri Ch 6

Fontaine B and Bigot S 1993 lsquoWest African rainfalldeficit and sea surface temperaturesrsquo InternationalJournal of Climatology Vol 13 271ndash85

Gautier L 1990 lsquoContact forecirct-savane en CocirctedrsquoIvoire centrale eacutevolution du recouvrementligneux des savanes de la Reacuteserve de Lamto (sud duV-Baouleacute)rsquo Candollea Vol 45 627ndash41

Giresse P Maley J and Brenac P 1994 lsquoLateQuaternary palaeoenvironments in the lake BarombiMbo (Cameroon) deduced from pollen and carbonisotopes of organic matterrsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 107 65ndash78

Goldammer JG and Seibert B 1990 lsquoThe impact ofdroughts and forest fires on tropical lowland rainforest of East Kalimatanrsquo in JG Goldammer (ed)Fire in the Tropical Biota Ecological Studies BerlinSpringer-Verlag 84

Goucher CL 1981 lsquoIron is iron til it is rust tradeand ecology in the decline of West African iron-smeltingrsquo Journal of African History Vol 22 179ndash89

Guille-Escuret G 1990 lsquoPalmier agrave huile vin depalme et transformations sociales en Lobaye (ForecirctCentrafricaine)rsquo Information sur les Sciences SocialesVol 29 327ndash53

Hart TB 2001 lsquoForest dynamics in the Ituri basin(DR Congo) Dominance diversity andconservationrsquo in W Weber L White A Vedderand L Naughton-Treves (eds) African rain forestecology and conservation An interdisciplinaryperspective Yale Yale University Press 154ndash64

Hart TB Hart JA Dechamps R Fournier M andAtaholo M 1996 lsquoChanges in forest compositionover the last 4000 years in the Ituri basin Zairersquo inLJG Van der Maesen and XM Van der Burgt (eds)The Biodiversity of African plants Dordrecht Kluwer545ndash63

Hawthorne WD 1991 Fire damage and forestregeneration in Ghana Ghana Forestry Departmentamp ODA Report

Jolly D Harrison SP Damnati B and BonnefilleR 1998 lsquoSimulated climate and Biomes of Africa

27

period suggests heavy rains but concentrated oversix or seven months of the year as is now found inthe peripheral savanna zones The growth inseasonality could have been associated with adominance of lsquosquall linesrsquo the typical cloudformation of savanna zones formed by northsouthalignments of cumuliform (convective storm-type)clouds They surplanted other cloud typesprincipally the stratiform types which giverelatively fine regular monsoon-type rains (seeMaley 1982) That these monsoon rains wouldhave dominated the earlier period (between 4000and 30002800 BP Maley 1997) is shown by therelative development of Caesalpinaceae in theforests of low altitude and in the mountains by theextension of Podocarpus which are trees typical ofcloud forests with stratiform clouds (Kerfoot1968 Maley 1996 1997 Maley and Brenac1998a)

Research in dynamic tropical climatology over thelast 20 years shows that Sea Surface Temperature(SST) on the regional and global scale stronglyinfluences climate (Fontaine and Bigot 1993

Moron et al 1995 Bigot et al 1997) It is possibleto characterise the relations between the monsoonrains and the SST more precisely Studies ofclimatic anomalies over tropical Africa occurringsince the beginning of the 1960s have evidencedtwo dominant modes of rainfall distribution eachassociated with a particular distribution of SST Thefirst mode is characterised on the one hand bywarmer than average temperatures in the Gulf ofGuinea and the southern Atlantic and on the otherhand by cooler temperatures on the north tropicalAtlantic offshore of West Africa The second modepresents an opposed distribution for the Africancontinent and the ocean A north-south tropicalAtlantic lsquodipolersquo has thus been found whichoscillates between one mode and the other(Fontaine and Bigot 1993 Wotling et al 1995Bigot et al 1997) An important research result hasbeen to show that over recent millennia the spatialdistribution of paleo-climatic anomalies overtropical Africa (more humid regions and more dryones Figure 6) and of SST on the neighbouringAtlantic is very similar to the spatial distributionobserved for the two dominant anomalies over thelast four decades (Maley 1997 Maley et al 2000)One can deduce that the climatic mechanisms arethe same at the timescale of years centuries ormillennia

So the variation of SST over the tropical Atlanticappears to play a major role in the variation ofclimate and rain types over Central Africa Thephase of forest disturbance that culminated about2500 years BP was associated with relatively lsquowarmrsquoSST over the Gulf of Guinea The earlier phasebetween about 40003800 and 2800 years BP wasassociated with relatively lsquocoolrsquo SST sharplyreduced from the early and middle Holocene (seeMorley and Dworetzky 1993 Figure 7) Over westand south Cameroon these SST allowed relativelyhigh rains and the evolution of stratiform cloudstowards rainy nimbostratus types The palaeo-vegetation data are presently lacking for GabonHowever new research is underway within thePALEOFORGA (Paleoenvironments of GabonForests) programme studying lakes MaridorNguegravene and Kamaleacuteteacute (Figure 1) In contrastfurther south in western Congo the periodstarting from 5000 years BP (Bertaux et al 2000)was marked by the development of semi-deciduousforests (Vincens et al 2000) linked to lower

24

Figure 6 Major rainfall trends in tropical Africa andSouth America for two opposed climatic periods c 3700ndash3000 BP and c 28002500ndash2000 BPPrecipitation trends (anomalies) abovenormalmean + below normalmean ndash P pointsto the sectors where a maximum of Podocarpuspollen intervenedSource Maley 1997 Maley et al 2000

rainfalls indicating that the stratiform clouds hadprobably evolved towards non-rainy clouds Thefour-month dry season here is characterised by thequasi-permanence of these cloud types (Saint Vil1984) which probably gained place progressivelybetween 5000 and 4000 years BP There wastherefore a reinforcing of southern influencesupporting conclusions concerning the history ofOkoumeacute It is notable that the vegetational impactsof the lsquowarmrsquo SST c 28002500 to 2000 years BPwere very different from those earlier c 9000 to65006000 years BP a difference which reflects abrutal strengthening over Central Africa of borealinfluences (Nguetsop et al 2000) to the detrimentof southern influences from about 28002500years BP

Before the major Holocene forest disturbanceculminating about 2500 BP other importantdisturbances occurred during the Quaternary(Maley 1996) The best documented is that betweenabout 20 000 and 15 000 years BP (Maley 19871996 Maley and Brenac 1998a Figure 2) In bothcases residual forest environments (refugia)

persisted apparently in similar locations butperhaps less as small blocks of relativelyhomogeneous forest than as a landscape of forest-savanna mosaic in which forest islands dominated(see Leal 2000) Yet two important characteristicsclearly differentiate these two phases of forestfragmentation First the older disturbance wasmuch longer Second they occurred under verydifferent (even opposed) general climatic conditionsThe disturbance between 20 000 and 15 000 BPcoincided with the development of glacialconditions in middle and high latitudes and coolerconditions by several degrees also affected the wholeregion of the Gulf of Guinea (Maley 1996 Maley andBrenac 1998a) The most recent disturbance bycontrast occurred during the Holocene interglacialwhich on a global scale is characterised by reducedglacial extension and relatively warmertemperatures The period between 2500 and 2000BP is associated with slightly warmer climaticconditions on the regional and global scale (Maley1997 Maley et al 2000) An important conclusion isthat the retreat and fragmentation of the Africanforests can be produced under very different

25

Figure 7 Variations in Sea Surface Temperature (SST) in the South Atlantic Ocean Benguela sector(22ucirc20rsquoSndash11ucirc12rsquoE) estimated with transfer functions based on the abundance of Radiolarian species in thecore RC13ndash228The open circles designate the 6000 and 9000 BP interpolated dates for the two curves one correspondingto the southern summer (February) and the other to the southern winter (August) modern SST areindicated by the solid squares above each curve The dashed lines represent some interpolated dates(italics) 18 000 BP based on the oxygen-isotope stratigraphy 38004000 BP 3000 BP and 2000 BP Theselate Holocene dates are confirmed by an independent study of shells from the same sector (Cohen et al1992) The coherence of the main SST anomalies throughout the Guinea Gulf permits the use of thesecurves in order to interpret the paleoclimatology of central Africa (Maley 1997)Source Morley and Dworetzky 1993

climates either relatively cool or relatively warmand thus with very different climatic situations(Maley 1996 1997)

Climatic models concerning lsquoglobal warmingrsquo inthe context of contemporary climate changeconcerns indicate that an average growth intemperature of about 4ucircC will lead also to anincrease in evaporation of about 30 per cent butonly 12 per cent more rain for tropical Africa (Rind

Notes The author is grateful to James Fairhead for his

translation of this article from the original Frenchand to Melissa Leach and James Fairhead for theirinvitation to the IDS workshop lsquoChangingPerspectives on Forestsrsquo in March 2001 This isInstitut des Sciences de lrsquoEvolution de Montpellier(ISEMCNRS) contribution No 125ndash2001

26

ReferencesAchoundong G Youta Happi J Guillet B

Bonvallot J and Kamgang Beyala V 2000lsquoFormation et eacutevolution des recrus sur savanes (ausud Cameroun)rsquo in M Servant and S Servant-Vildary (eds) Dynamique agrave long terme desEcosystegravemes forestiers intertropicaux MeacutemoireUNESCO Paris UNESCO 31ndash41

Alexandre P 1965 lsquoProto-histoire du groupe Beacuteti-Boulou-Fang essai de synthegravese provisoirersquo CahiersdrsquoEtudes Africaines Vol 20 503ndash60

Amougou Akoa 1986 lsquoEtude botanique eteacutecologique de la valleacutee inondable du Haut-Nyong(sud Cameroun) et de ses affluentsrsquo Thegravese SciencesUniv Yaoundeacute

Aubreacuteville A 1937 lsquoLes forecircts du Dahomey et duTogorsquo Bulletin du Comiteacute drsquoEtudes Historiques etScientifiques de lrsquoAfrique Occidentale Franccedilaises Vol20 112

Aubreacuteville A 1967 lsquoLes eacutetranges mosaiques forecirct-savane du sommet de la boucle de lrsquoOgooueacute auGabonrsquo Adansonia Vol 7 13ndash22

Baeke V 1996 lsquoLe temps des Rites Lrsquounivers magico-religieux des Wuli (Mfumte du Camerounoccidental)rsquo Thegravese Sc Sociale Univ Libre Bruxelles

Barber RJ 1985 lsquoLand snails and past environmentat the Igbo-Iwoto Esie site southwestern NigeriarsquoWest African Journal of Archaeology Vol 15 89ndash102

Barbier JC 1981 lsquoLe peuplement de la partiemeacuteridionale du plateau Bamileacutekeacutersquo in C Tardits

(ed) Contribution de la recherche ethnologique agravelrsquohistoire des civilisations du Cameroun ParisColloque International Centre National deRecherche Scientifique Paris Vol 2 331ndash53

Bertault JG 1990 lsquoComparaison drsquoeacutecosystegravemesforestiers naturels et modifieacutes apregraves incendie enCocircte drsquoIvoirersquo in H Puig (ed) Atelier surlrsquoAmeacutenagement et la Conservation de lrsquoEcosystegravemeForestier Tropical Humide Cayenne ParisUNESCO 6

Bertaux J Schwartz D Vincens A Sifeddine AElenga H Mansour M Mariotti A Fournier MMartin L Wirrmann D and Servant M 2000lsquoEnregistrement de la phase segraveche drsquoAfriqueCentrale vers 3000 ans BP par la spectromeacutetrie IRdans les lacs Sinnda et Kitina (sud-Congo)rsquo in MServant S Servant-Vildary (eds) Dynamique agrave longterme des Ecosystegravemes forestiers intertropicauxMeacutemoire UNESCO Paris UNESCO 43ndash9

Bigot S Camberlin P Moron V and Richard Y1997 lsquoStructures spatiales de la variabiliteacute despreacutecipitations en Afrique une transition climatiqueagrave la fin des anneacutees 1960rsquo Compte Rendu delrsquoAcademie des Sciences Paris 2a 324 181ndash88

Blanc-Pamard C and Peltre P 1984 lsquoDynamique despaysages preacuteforestiers et pratiques culturales enAfrique de lrsquoOuest (Cocircte drsquoIvoire centrale)rsquo LeDeacuteveloppement rural en question Paris MeacutemoireORSTOM No 106 55ndash74

Bonnefille R Chalieacute F Guiot J and Vincens A1992 lsquoQuantitative estimates of full glacial

1995) Though the causes of warming experiencedabout 2500 years BP were natural (Magny 1993)and therefore different from the warmingenvisaged for the twenty-first century it is possiblethat the catastrophic destruction of Central Africanforests which culminated about 2500 years BPcould be an lsquoanaloguersquo and an alarm signal forpotential developments during this current andfuture phase of human-induced global warming(Maley 1997 Maley et al 2000)

1 BP lsquoBefore the Presentrsquo according to standardisedC14 analysis

2 Only 4ndash5 samples have been studied for each 1000years during the Holocene To detect century-scaleperiodicity it would be necessary to study at least 10samples per millennium

temperatures in equatorial Africa frompalynological datarsquo Climate Dynamics Vol 6251ndash57

Brunck F Grison F and Maitre HF 1990Lrsquookoumeacute Aucoumea klaineana Pierre MonographieCentre Techn Forestier Tropical Nogent CIRAD 102

Cohen AI Perkington JE Brundrirt GB and vanDer Merwe NJ 1992 lsquoA Holocene marine climaterecord in Mollusc shells from the southwest Africancoastrsquo Quaternary Research Vol 38 379ndash85

Dechamps R Senut B and Pickford M 1992 Fruitsfossiles pliocegravenes et pleacuteistocegravenes du Rift occidentalougandais Signification paleacuteoenvironnementaleCompte Rendu de lrsquoAcadeacutemie des Sciences Parisseacuterie 2 Vol 314 325ndash331

Dowsett-Lemaire F 1996 lsquoComposition et eacutevolutionde la veacutegeacutetation forestiegravere au Parc NationaldrsquoOdzala Congorsquo Bulletin du Jardin BotaniqueNational de Belgique Vol 65 253ndash92

Dupont LM and Weinelt M 1996 lsquoVegetationhistory of the savanna corridor between theguinean and the Congolian rain forest during thelast 150000 yearsrsquo Vegetation History andArchaeobotany Vol 5 273ndash92

Elenga H Schwartz D and Vincens A 1994lsquoPollen evidence of late Quaternary vegetation andinferred climate changes in Congorsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 109345ndash56

Elenga H Schwartz D Vincens A Bertaux J DeNamur C Martin L Wirrmann D and ServantM 1996 lsquoDiagramme pollinique holocegravene du lacKitina (Congo) mise en eacutevidence de changementspaleacuteobotaniques et paleacuteoclimatiques dans le massifforestier du Mayombersquo Compte Rendu de lrsquoAcadeacutemiedes Sciences Paris Vol 323 No 2a 403ndash10

Elenga H Peyron O Bonnefille R Prentice ICJolly D Cheddadi R Guiot J Andrieu V DeBeaulieu JL Hamilton AC Maley J MarchantR Reille M Scott L Taylor D and Van CampoE 2000 lsquoPollen-based biome reconstruction forsouthern Europe and Africa 18000 years agorsquoJournal of Biogeography Vol 27 621ndash34

Fairhead J and Leach M 1995 lsquoFalse forest historycomplicit social analysis rethinking some westAfrican environmental narrativesrsquo WorldDevelopment Vol 23 1023ndash35

Fairhead J and Leach M 1996 lsquoEnriching thelandscape social history and the management oftransition ecology in the forest-savanna mosaic ofthe Republic of Guinearsquo Africa Vol 66 14ndash36

Fairhead J and Leach M 1998 lsquoReframingdeforestation Global analyses and local realitiesstudies in West Africarsquo Global EnvironmentalChanges Series London Routledge 238

Fay JM 1997 lsquoEvidence of forest degradation innorthern Congo and southwestern Central AfricanRepublic between 2340 and 990 BPrsquo in The ecologysocial organization populations habitat and historyof the western lowland Gorilla PhD ThesisWashington Univ Saint Louis Missouri Ch 6

Fontaine B and Bigot S 1993 lsquoWest African rainfalldeficit and sea surface temperaturesrsquo InternationalJournal of Climatology Vol 13 271ndash85

Gautier L 1990 lsquoContact forecirct-savane en CocirctedrsquoIvoire centrale eacutevolution du recouvrementligneux des savanes de la Reacuteserve de Lamto (sud duV-Baouleacute)rsquo Candollea Vol 45 627ndash41

Giresse P Maley J and Brenac P 1994 lsquoLateQuaternary palaeoenvironments in the lake BarombiMbo (Cameroon) deduced from pollen and carbonisotopes of organic matterrsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 107 65ndash78

Goldammer JG and Seibert B 1990 lsquoThe impact ofdroughts and forest fires on tropical lowland rainforest of East Kalimatanrsquo in JG Goldammer (ed)Fire in the Tropical Biota Ecological Studies BerlinSpringer-Verlag 84

Goucher CL 1981 lsquoIron is iron til it is rust tradeand ecology in the decline of West African iron-smeltingrsquo Journal of African History Vol 22 179ndash89

Guille-Escuret G 1990 lsquoPalmier agrave huile vin depalme et transformations sociales en Lobaye (ForecirctCentrafricaine)rsquo Information sur les Sciences SocialesVol 29 327ndash53

Hart TB 2001 lsquoForest dynamics in the Ituri basin(DR Congo) Dominance diversity andconservationrsquo in W Weber L White A Vedderand L Naughton-Treves (eds) African rain forestecology and conservation An interdisciplinaryperspective Yale Yale University Press 154ndash64

Hart TB Hart JA Dechamps R Fournier M andAtaholo M 1996 lsquoChanges in forest compositionover the last 4000 years in the Ituri basin Zairersquo inLJG Van der Maesen and XM Van der Burgt (eds)The Biodiversity of African plants Dordrecht Kluwer545ndash63

Hawthorne WD 1991 Fire damage and forestregeneration in Ghana Ghana Forestry Departmentamp ODA Report

Jolly D Harrison SP Damnati B and BonnefilleR 1998 lsquoSimulated climate and Biomes of Africa

27

rainfalls indicating that the stratiform clouds hadprobably evolved towards non-rainy clouds Thefour-month dry season here is characterised by thequasi-permanence of these cloud types (Saint Vil1984) which probably gained place progressivelybetween 5000 and 4000 years BP There wastherefore a reinforcing of southern influencesupporting conclusions concerning the history ofOkoumeacute It is notable that the vegetational impactsof the lsquowarmrsquo SST c 28002500 to 2000 years BPwere very different from those earlier c 9000 to65006000 years BP a difference which reflects abrutal strengthening over Central Africa of borealinfluences (Nguetsop et al 2000) to the detrimentof southern influences from about 28002500years BP

Before the major Holocene forest disturbanceculminating about 2500 BP other importantdisturbances occurred during the Quaternary(Maley 1996) The best documented is that betweenabout 20 000 and 15 000 years BP (Maley 19871996 Maley and Brenac 1998a Figure 2) In bothcases residual forest environments (refugia)

persisted apparently in similar locations butperhaps less as small blocks of relativelyhomogeneous forest than as a landscape of forest-savanna mosaic in which forest islands dominated(see Leal 2000) Yet two important characteristicsclearly differentiate these two phases of forestfragmentation First the older disturbance wasmuch longer Second they occurred under verydifferent (even opposed) general climatic conditionsThe disturbance between 20 000 and 15 000 BPcoincided with the development of glacialconditions in middle and high latitudes and coolerconditions by several degrees also affected the wholeregion of the Gulf of Guinea (Maley 1996 Maley andBrenac 1998a) The most recent disturbance bycontrast occurred during the Holocene interglacialwhich on a global scale is characterised by reducedglacial extension and relatively warmertemperatures The period between 2500 and 2000BP is associated with slightly warmer climaticconditions on the regional and global scale (Maley1997 Maley et al 2000) An important conclusion isthat the retreat and fragmentation of the Africanforests can be produced under very different

25

Figure 7 Variations in Sea Surface Temperature (SST) in the South Atlantic Ocean Benguela sector(22ucirc20rsquoSndash11ucirc12rsquoE) estimated with transfer functions based on the abundance of Radiolarian species in thecore RC13ndash228The open circles designate the 6000 and 9000 BP interpolated dates for the two curves one correspondingto the southern summer (February) and the other to the southern winter (August) modern SST areindicated by the solid squares above each curve The dashed lines represent some interpolated dates(italics) 18 000 BP based on the oxygen-isotope stratigraphy 38004000 BP 3000 BP and 2000 BP Theselate Holocene dates are confirmed by an independent study of shells from the same sector (Cohen et al1992) The coherence of the main SST anomalies throughout the Guinea Gulf permits the use of thesecurves in order to interpret the paleoclimatology of central Africa (Maley 1997)Source Morley and Dworetzky 1993

climates either relatively cool or relatively warmand thus with very different climatic situations(Maley 1996 1997)

Climatic models concerning lsquoglobal warmingrsquo inthe context of contemporary climate changeconcerns indicate that an average growth intemperature of about 4ucircC will lead also to anincrease in evaporation of about 30 per cent butonly 12 per cent more rain for tropical Africa (Rind

Notes The author is grateful to James Fairhead for his

translation of this article from the original Frenchand to Melissa Leach and James Fairhead for theirinvitation to the IDS workshop lsquoChangingPerspectives on Forestsrsquo in March 2001 This isInstitut des Sciences de lrsquoEvolution de Montpellier(ISEMCNRS) contribution No 125ndash2001

26

ReferencesAchoundong G Youta Happi J Guillet B

Bonvallot J and Kamgang Beyala V 2000lsquoFormation et eacutevolution des recrus sur savanes (ausud Cameroun)rsquo in M Servant and S Servant-Vildary (eds) Dynamique agrave long terme desEcosystegravemes forestiers intertropicaux MeacutemoireUNESCO Paris UNESCO 31ndash41

Alexandre P 1965 lsquoProto-histoire du groupe Beacuteti-Boulou-Fang essai de synthegravese provisoirersquo CahiersdrsquoEtudes Africaines Vol 20 503ndash60

Amougou Akoa 1986 lsquoEtude botanique eteacutecologique de la valleacutee inondable du Haut-Nyong(sud Cameroun) et de ses affluentsrsquo Thegravese SciencesUniv Yaoundeacute

Aubreacuteville A 1937 lsquoLes forecircts du Dahomey et duTogorsquo Bulletin du Comiteacute drsquoEtudes Historiques etScientifiques de lrsquoAfrique Occidentale Franccedilaises Vol20 112

Aubreacuteville A 1967 lsquoLes eacutetranges mosaiques forecirct-savane du sommet de la boucle de lrsquoOgooueacute auGabonrsquo Adansonia Vol 7 13ndash22

Baeke V 1996 lsquoLe temps des Rites Lrsquounivers magico-religieux des Wuli (Mfumte du Camerounoccidental)rsquo Thegravese Sc Sociale Univ Libre Bruxelles

Barber RJ 1985 lsquoLand snails and past environmentat the Igbo-Iwoto Esie site southwestern NigeriarsquoWest African Journal of Archaeology Vol 15 89ndash102

Barbier JC 1981 lsquoLe peuplement de la partiemeacuteridionale du plateau Bamileacutekeacutersquo in C Tardits

(ed) Contribution de la recherche ethnologique agravelrsquohistoire des civilisations du Cameroun ParisColloque International Centre National deRecherche Scientifique Paris Vol 2 331ndash53

Bertault JG 1990 lsquoComparaison drsquoeacutecosystegravemesforestiers naturels et modifieacutes apregraves incendie enCocircte drsquoIvoirersquo in H Puig (ed) Atelier surlrsquoAmeacutenagement et la Conservation de lrsquoEcosystegravemeForestier Tropical Humide Cayenne ParisUNESCO 6

Bertaux J Schwartz D Vincens A Sifeddine AElenga H Mansour M Mariotti A Fournier MMartin L Wirrmann D and Servant M 2000lsquoEnregistrement de la phase segraveche drsquoAfriqueCentrale vers 3000 ans BP par la spectromeacutetrie IRdans les lacs Sinnda et Kitina (sud-Congo)rsquo in MServant S Servant-Vildary (eds) Dynamique agrave longterme des Ecosystegravemes forestiers intertropicauxMeacutemoire UNESCO Paris UNESCO 43ndash9

Bigot S Camberlin P Moron V and Richard Y1997 lsquoStructures spatiales de la variabiliteacute despreacutecipitations en Afrique une transition climatiqueagrave la fin des anneacutees 1960rsquo Compte Rendu delrsquoAcademie des Sciences Paris 2a 324 181ndash88

Blanc-Pamard C and Peltre P 1984 lsquoDynamique despaysages preacuteforestiers et pratiques culturales enAfrique de lrsquoOuest (Cocircte drsquoIvoire centrale)rsquo LeDeacuteveloppement rural en question Paris MeacutemoireORSTOM No 106 55ndash74

Bonnefille R Chalieacute F Guiot J and Vincens A1992 lsquoQuantitative estimates of full glacial

1995) Though the causes of warming experiencedabout 2500 years BP were natural (Magny 1993)and therefore different from the warmingenvisaged for the twenty-first century it is possiblethat the catastrophic destruction of Central Africanforests which culminated about 2500 years BPcould be an lsquoanaloguersquo and an alarm signal forpotential developments during this current andfuture phase of human-induced global warming(Maley 1997 Maley et al 2000)

1 BP lsquoBefore the Presentrsquo according to standardisedC14 analysis

2 Only 4ndash5 samples have been studied for each 1000years during the Holocene To detect century-scaleperiodicity it would be necessary to study at least 10samples per millennium

temperatures in equatorial Africa frompalynological datarsquo Climate Dynamics Vol 6251ndash57

Brunck F Grison F and Maitre HF 1990Lrsquookoumeacute Aucoumea klaineana Pierre MonographieCentre Techn Forestier Tropical Nogent CIRAD 102

Cohen AI Perkington JE Brundrirt GB and vanDer Merwe NJ 1992 lsquoA Holocene marine climaterecord in Mollusc shells from the southwest Africancoastrsquo Quaternary Research Vol 38 379ndash85

Dechamps R Senut B and Pickford M 1992 Fruitsfossiles pliocegravenes et pleacuteistocegravenes du Rift occidentalougandais Signification paleacuteoenvironnementaleCompte Rendu de lrsquoAcadeacutemie des Sciences Parisseacuterie 2 Vol 314 325ndash331

Dowsett-Lemaire F 1996 lsquoComposition et eacutevolutionde la veacutegeacutetation forestiegravere au Parc NationaldrsquoOdzala Congorsquo Bulletin du Jardin BotaniqueNational de Belgique Vol 65 253ndash92

Dupont LM and Weinelt M 1996 lsquoVegetationhistory of the savanna corridor between theguinean and the Congolian rain forest during thelast 150000 yearsrsquo Vegetation History andArchaeobotany Vol 5 273ndash92

Elenga H Schwartz D and Vincens A 1994lsquoPollen evidence of late Quaternary vegetation andinferred climate changes in Congorsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 109345ndash56

Elenga H Schwartz D Vincens A Bertaux J DeNamur C Martin L Wirrmann D and ServantM 1996 lsquoDiagramme pollinique holocegravene du lacKitina (Congo) mise en eacutevidence de changementspaleacuteobotaniques et paleacuteoclimatiques dans le massifforestier du Mayombersquo Compte Rendu de lrsquoAcadeacutemiedes Sciences Paris Vol 323 No 2a 403ndash10

Elenga H Peyron O Bonnefille R Prentice ICJolly D Cheddadi R Guiot J Andrieu V DeBeaulieu JL Hamilton AC Maley J MarchantR Reille M Scott L Taylor D and Van CampoE 2000 lsquoPollen-based biome reconstruction forsouthern Europe and Africa 18000 years agorsquoJournal of Biogeography Vol 27 621ndash34

Fairhead J and Leach M 1995 lsquoFalse forest historycomplicit social analysis rethinking some westAfrican environmental narrativesrsquo WorldDevelopment Vol 23 1023ndash35

Fairhead J and Leach M 1996 lsquoEnriching thelandscape social history and the management oftransition ecology in the forest-savanna mosaic ofthe Republic of Guinearsquo Africa Vol 66 14ndash36

Fairhead J and Leach M 1998 lsquoReframingdeforestation Global analyses and local realitiesstudies in West Africarsquo Global EnvironmentalChanges Series London Routledge 238

Fay JM 1997 lsquoEvidence of forest degradation innorthern Congo and southwestern Central AfricanRepublic between 2340 and 990 BPrsquo in The ecologysocial organization populations habitat and historyof the western lowland Gorilla PhD ThesisWashington Univ Saint Louis Missouri Ch 6

Fontaine B and Bigot S 1993 lsquoWest African rainfalldeficit and sea surface temperaturesrsquo InternationalJournal of Climatology Vol 13 271ndash85

Gautier L 1990 lsquoContact forecirct-savane en CocirctedrsquoIvoire centrale eacutevolution du recouvrementligneux des savanes de la Reacuteserve de Lamto (sud duV-Baouleacute)rsquo Candollea Vol 45 627ndash41

Giresse P Maley J and Brenac P 1994 lsquoLateQuaternary palaeoenvironments in the lake BarombiMbo (Cameroon) deduced from pollen and carbonisotopes of organic matterrsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 107 65ndash78

Goldammer JG and Seibert B 1990 lsquoThe impact ofdroughts and forest fires on tropical lowland rainforest of East Kalimatanrsquo in JG Goldammer (ed)Fire in the Tropical Biota Ecological Studies BerlinSpringer-Verlag 84

Goucher CL 1981 lsquoIron is iron til it is rust tradeand ecology in the decline of West African iron-smeltingrsquo Journal of African History Vol 22 179ndash89

Guille-Escuret G 1990 lsquoPalmier agrave huile vin depalme et transformations sociales en Lobaye (ForecirctCentrafricaine)rsquo Information sur les Sciences SocialesVol 29 327ndash53

Hart TB 2001 lsquoForest dynamics in the Ituri basin(DR Congo) Dominance diversity andconservationrsquo in W Weber L White A Vedderand L Naughton-Treves (eds) African rain forestecology and conservation An interdisciplinaryperspective Yale Yale University Press 154ndash64

Hart TB Hart JA Dechamps R Fournier M andAtaholo M 1996 lsquoChanges in forest compositionover the last 4000 years in the Ituri basin Zairersquo inLJG Van der Maesen and XM Van der Burgt (eds)The Biodiversity of African plants Dordrecht Kluwer545ndash63

Hawthorne WD 1991 Fire damage and forestregeneration in Ghana Ghana Forestry Departmentamp ODA Report

Jolly D Harrison SP Damnati B and BonnefilleR 1998 lsquoSimulated climate and Biomes of Africa

27

climates either relatively cool or relatively warmand thus with very different climatic situations(Maley 1996 1997)

Climatic models concerning lsquoglobal warmingrsquo inthe context of contemporary climate changeconcerns indicate that an average growth intemperature of about 4ucircC will lead also to anincrease in evaporation of about 30 per cent butonly 12 per cent more rain for tropical Africa (Rind

Notes The author is grateful to James Fairhead for his

translation of this article from the original Frenchand to Melissa Leach and James Fairhead for theirinvitation to the IDS workshop lsquoChangingPerspectives on Forestsrsquo in March 2001 This isInstitut des Sciences de lrsquoEvolution de Montpellier(ISEMCNRS) contribution No 125ndash2001

26

ReferencesAchoundong G Youta Happi J Guillet B

Bonvallot J and Kamgang Beyala V 2000lsquoFormation et eacutevolution des recrus sur savanes (ausud Cameroun)rsquo in M Servant and S Servant-Vildary (eds) Dynamique agrave long terme desEcosystegravemes forestiers intertropicaux MeacutemoireUNESCO Paris UNESCO 31ndash41

Alexandre P 1965 lsquoProto-histoire du groupe Beacuteti-Boulou-Fang essai de synthegravese provisoirersquo CahiersdrsquoEtudes Africaines Vol 20 503ndash60

Amougou Akoa 1986 lsquoEtude botanique eteacutecologique de la valleacutee inondable du Haut-Nyong(sud Cameroun) et de ses affluentsrsquo Thegravese SciencesUniv Yaoundeacute

Aubreacuteville A 1937 lsquoLes forecircts du Dahomey et duTogorsquo Bulletin du Comiteacute drsquoEtudes Historiques etScientifiques de lrsquoAfrique Occidentale Franccedilaises Vol20 112

Aubreacuteville A 1967 lsquoLes eacutetranges mosaiques forecirct-savane du sommet de la boucle de lrsquoOgooueacute auGabonrsquo Adansonia Vol 7 13ndash22

Baeke V 1996 lsquoLe temps des Rites Lrsquounivers magico-religieux des Wuli (Mfumte du Camerounoccidental)rsquo Thegravese Sc Sociale Univ Libre Bruxelles

Barber RJ 1985 lsquoLand snails and past environmentat the Igbo-Iwoto Esie site southwestern NigeriarsquoWest African Journal of Archaeology Vol 15 89ndash102

Barbier JC 1981 lsquoLe peuplement de la partiemeacuteridionale du plateau Bamileacutekeacutersquo in C Tardits

(ed) Contribution de la recherche ethnologique agravelrsquohistoire des civilisations du Cameroun ParisColloque International Centre National deRecherche Scientifique Paris Vol 2 331ndash53

Bertault JG 1990 lsquoComparaison drsquoeacutecosystegravemesforestiers naturels et modifieacutes apregraves incendie enCocircte drsquoIvoirersquo in H Puig (ed) Atelier surlrsquoAmeacutenagement et la Conservation de lrsquoEcosystegravemeForestier Tropical Humide Cayenne ParisUNESCO 6

Bertaux J Schwartz D Vincens A Sifeddine AElenga H Mansour M Mariotti A Fournier MMartin L Wirrmann D and Servant M 2000lsquoEnregistrement de la phase segraveche drsquoAfriqueCentrale vers 3000 ans BP par la spectromeacutetrie IRdans les lacs Sinnda et Kitina (sud-Congo)rsquo in MServant S Servant-Vildary (eds) Dynamique agrave longterme des Ecosystegravemes forestiers intertropicauxMeacutemoire UNESCO Paris UNESCO 43ndash9

Bigot S Camberlin P Moron V and Richard Y1997 lsquoStructures spatiales de la variabiliteacute despreacutecipitations en Afrique une transition climatiqueagrave la fin des anneacutees 1960rsquo Compte Rendu delrsquoAcademie des Sciences Paris 2a 324 181ndash88

Blanc-Pamard C and Peltre P 1984 lsquoDynamique despaysages preacuteforestiers et pratiques culturales enAfrique de lrsquoOuest (Cocircte drsquoIvoire centrale)rsquo LeDeacuteveloppement rural en question Paris MeacutemoireORSTOM No 106 55ndash74

Bonnefille R Chalieacute F Guiot J and Vincens A1992 lsquoQuantitative estimates of full glacial

1995) Though the causes of warming experiencedabout 2500 years BP were natural (Magny 1993)and therefore different from the warmingenvisaged for the twenty-first century it is possiblethat the catastrophic destruction of Central Africanforests which culminated about 2500 years BPcould be an lsquoanaloguersquo and an alarm signal forpotential developments during this current andfuture phase of human-induced global warming(Maley 1997 Maley et al 2000)

1 BP lsquoBefore the Presentrsquo according to standardisedC14 analysis

2 Only 4ndash5 samples have been studied for each 1000years during the Holocene To detect century-scaleperiodicity it would be necessary to study at least 10samples per millennium

temperatures in equatorial Africa frompalynological datarsquo Climate Dynamics Vol 6251ndash57

Brunck F Grison F and Maitre HF 1990Lrsquookoumeacute Aucoumea klaineana Pierre MonographieCentre Techn Forestier Tropical Nogent CIRAD 102

Cohen AI Perkington JE Brundrirt GB and vanDer Merwe NJ 1992 lsquoA Holocene marine climaterecord in Mollusc shells from the southwest Africancoastrsquo Quaternary Research Vol 38 379ndash85

Dechamps R Senut B and Pickford M 1992 Fruitsfossiles pliocegravenes et pleacuteistocegravenes du Rift occidentalougandais Signification paleacuteoenvironnementaleCompte Rendu de lrsquoAcadeacutemie des Sciences Parisseacuterie 2 Vol 314 325ndash331

Dowsett-Lemaire F 1996 lsquoComposition et eacutevolutionde la veacutegeacutetation forestiegravere au Parc NationaldrsquoOdzala Congorsquo Bulletin du Jardin BotaniqueNational de Belgique Vol 65 253ndash92

Dupont LM and Weinelt M 1996 lsquoVegetationhistory of the savanna corridor between theguinean and the Congolian rain forest during thelast 150000 yearsrsquo Vegetation History andArchaeobotany Vol 5 273ndash92

Elenga H Schwartz D and Vincens A 1994lsquoPollen evidence of late Quaternary vegetation andinferred climate changes in Congorsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 109345ndash56

Elenga H Schwartz D Vincens A Bertaux J DeNamur C Martin L Wirrmann D and ServantM 1996 lsquoDiagramme pollinique holocegravene du lacKitina (Congo) mise en eacutevidence de changementspaleacuteobotaniques et paleacuteoclimatiques dans le massifforestier du Mayombersquo Compte Rendu de lrsquoAcadeacutemiedes Sciences Paris Vol 323 No 2a 403ndash10

Elenga H Peyron O Bonnefille R Prentice ICJolly D Cheddadi R Guiot J Andrieu V DeBeaulieu JL Hamilton AC Maley J MarchantR Reille M Scott L Taylor D and Van CampoE 2000 lsquoPollen-based biome reconstruction forsouthern Europe and Africa 18000 years agorsquoJournal of Biogeography Vol 27 621ndash34

Fairhead J and Leach M 1995 lsquoFalse forest historycomplicit social analysis rethinking some westAfrican environmental narrativesrsquo WorldDevelopment Vol 23 1023ndash35

Fairhead J and Leach M 1996 lsquoEnriching thelandscape social history and the management oftransition ecology in the forest-savanna mosaic ofthe Republic of Guinearsquo Africa Vol 66 14ndash36

Fairhead J and Leach M 1998 lsquoReframingdeforestation Global analyses and local realitiesstudies in West Africarsquo Global EnvironmentalChanges Series London Routledge 238

Fay JM 1997 lsquoEvidence of forest degradation innorthern Congo and southwestern Central AfricanRepublic between 2340 and 990 BPrsquo in The ecologysocial organization populations habitat and historyof the western lowland Gorilla PhD ThesisWashington Univ Saint Louis Missouri Ch 6

Fontaine B and Bigot S 1993 lsquoWest African rainfalldeficit and sea surface temperaturesrsquo InternationalJournal of Climatology Vol 13 271ndash85

Gautier L 1990 lsquoContact forecirct-savane en CocirctedrsquoIvoire centrale eacutevolution du recouvrementligneux des savanes de la Reacuteserve de Lamto (sud duV-Baouleacute)rsquo Candollea Vol 45 627ndash41

Giresse P Maley J and Brenac P 1994 lsquoLateQuaternary palaeoenvironments in the lake BarombiMbo (Cameroon) deduced from pollen and carbonisotopes of organic matterrsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 107 65ndash78

Goldammer JG and Seibert B 1990 lsquoThe impact ofdroughts and forest fires on tropical lowland rainforest of East Kalimatanrsquo in JG Goldammer (ed)Fire in the Tropical Biota Ecological Studies BerlinSpringer-Verlag 84

Goucher CL 1981 lsquoIron is iron til it is rust tradeand ecology in the decline of West African iron-smeltingrsquo Journal of African History Vol 22 179ndash89

Guille-Escuret G 1990 lsquoPalmier agrave huile vin depalme et transformations sociales en Lobaye (ForecirctCentrafricaine)rsquo Information sur les Sciences SocialesVol 29 327ndash53

Hart TB 2001 lsquoForest dynamics in the Ituri basin(DR Congo) Dominance diversity andconservationrsquo in W Weber L White A Vedderand L Naughton-Treves (eds) African rain forestecology and conservation An interdisciplinaryperspective Yale Yale University Press 154ndash64

Hart TB Hart JA Dechamps R Fournier M andAtaholo M 1996 lsquoChanges in forest compositionover the last 4000 years in the Ituri basin Zairersquo inLJG Van der Maesen and XM Van der Burgt (eds)The Biodiversity of African plants Dordrecht Kluwer545ndash63

Hawthorne WD 1991 Fire damage and forestregeneration in Ghana Ghana Forestry Departmentamp ODA Report

Jolly D Harrison SP Damnati B and BonnefilleR 1998 lsquoSimulated climate and Biomes of Africa

27

temperatures in equatorial Africa frompalynological datarsquo Climate Dynamics Vol 6251ndash57

Brunck F Grison F and Maitre HF 1990Lrsquookoumeacute Aucoumea klaineana Pierre MonographieCentre Techn Forestier Tropical Nogent CIRAD 102

Cohen AI Perkington JE Brundrirt GB and vanDer Merwe NJ 1992 lsquoA Holocene marine climaterecord in Mollusc shells from the southwest Africancoastrsquo Quaternary Research Vol 38 379ndash85

Dechamps R Senut B and Pickford M 1992 Fruitsfossiles pliocegravenes et pleacuteistocegravenes du Rift occidentalougandais Signification paleacuteoenvironnementaleCompte Rendu de lrsquoAcadeacutemie des Sciences Parisseacuterie 2 Vol 314 325ndash331

Dowsett-Lemaire F 1996 lsquoComposition et eacutevolutionde la veacutegeacutetation forestiegravere au Parc NationaldrsquoOdzala Congorsquo Bulletin du Jardin BotaniqueNational de Belgique Vol 65 253ndash92

Dupont LM and Weinelt M 1996 lsquoVegetationhistory of the savanna corridor between theguinean and the Congolian rain forest during thelast 150000 yearsrsquo Vegetation History andArchaeobotany Vol 5 273ndash92

Elenga H Schwartz D and Vincens A 1994lsquoPollen evidence of late Quaternary vegetation andinferred climate changes in Congorsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 109345ndash56

Elenga H Schwartz D Vincens A Bertaux J DeNamur C Martin L Wirrmann D and ServantM 1996 lsquoDiagramme pollinique holocegravene du lacKitina (Congo) mise en eacutevidence de changementspaleacuteobotaniques et paleacuteoclimatiques dans le massifforestier du Mayombersquo Compte Rendu de lrsquoAcadeacutemiedes Sciences Paris Vol 323 No 2a 403ndash10

Elenga H Peyron O Bonnefille R Prentice ICJolly D Cheddadi R Guiot J Andrieu V DeBeaulieu JL Hamilton AC Maley J MarchantR Reille M Scott L Taylor D and Van CampoE 2000 lsquoPollen-based biome reconstruction forsouthern Europe and Africa 18000 years agorsquoJournal of Biogeography Vol 27 621ndash34

Fairhead J and Leach M 1995 lsquoFalse forest historycomplicit social analysis rethinking some westAfrican environmental narrativesrsquo WorldDevelopment Vol 23 1023ndash35

Fairhead J and Leach M 1996 lsquoEnriching thelandscape social history and the management oftransition ecology in the forest-savanna mosaic ofthe Republic of Guinearsquo Africa Vol 66 14ndash36

Fairhead J and Leach M 1998 lsquoReframingdeforestation Global analyses and local realitiesstudies in West Africarsquo Global EnvironmentalChanges Series London Routledge 238

Fay JM 1997 lsquoEvidence of forest degradation innorthern Congo and southwestern Central AfricanRepublic between 2340 and 990 BPrsquo in The ecologysocial organization populations habitat and historyof the western lowland Gorilla PhD ThesisWashington Univ Saint Louis Missouri Ch 6

Fontaine B and Bigot S 1993 lsquoWest African rainfalldeficit and sea surface temperaturesrsquo InternationalJournal of Climatology Vol 13 271ndash85

Gautier L 1990 lsquoContact forecirct-savane en CocirctedrsquoIvoire centrale eacutevolution du recouvrementligneux des savanes de la Reacuteserve de Lamto (sud duV-Baouleacute)rsquo Candollea Vol 45 627ndash41

Giresse P Maley J and Brenac P 1994 lsquoLateQuaternary palaeoenvironments in the lake BarombiMbo (Cameroon) deduced from pollen and carbonisotopes of organic matterrsquo PalaeogeographyPalaeoclimatology and Palaeoecology Vol 107 65ndash78

Goldammer JG and Seibert B 1990 lsquoThe impact ofdroughts and forest fires on tropical lowland rainforest of East Kalimatanrsquo in JG Goldammer (ed)Fire in the Tropical Biota Ecological Studies BerlinSpringer-Verlag 84

Goucher CL 1981 lsquoIron is iron til it is rust tradeand ecology in the decline of West African iron-smeltingrsquo Journal of African History Vol 22 179ndash89

Guille-Escuret G 1990 lsquoPalmier agrave huile vin depalme et transformations sociales en Lobaye (ForecirctCentrafricaine)rsquo Information sur les Sciences SocialesVol 29 327ndash53

Hart TB 2001 lsquoForest dynamics in the Ituri basin(DR Congo) Dominance diversity andconservationrsquo in W Weber L White A Vedderand L Naughton-Treves (eds) African rain forestecology and conservation An interdisciplinaryperspective Yale Yale University Press 154ndash64

Hart TB Hart JA Dechamps R Fournier M andAtaholo M 1996 lsquoChanges in forest compositionover the last 4000 years in the Ituri basin Zairersquo inLJG Van der Maesen and XM Van der Burgt (eds)The Biodiversity of African plants Dordrecht Kluwer545ndash63

Hawthorne WD 1991 Fire damage and forestregeneration in Ghana Ghana Forestry Departmentamp ODA Report

Jolly D Harrison SP Damnati B and BonnefilleR 1998 lsquoSimulated climate and Biomes of Africa

27

during the Late Quaternary comparison withpollen and lake status datarsquo Quaternary ScienceReview Vol 17 629ndash57

Jones EW 1956 lsquoEcological studies on the rainforest of southern Nigeria IV The Plateau forest ofthe Okomu Forest Reserversquo Journal of Ecology Vol44 83ndash117

Kerfoot O 1968 lsquoMist precipitation on vegetationrsquoForestry Abstracts Vol 29 8ndash20

Laraque A and Pandi A 1996 lsquoRocircle des donneacuteesphysiographiques dans la classification hydrologiquedes affluents congolais du fleuve Congo-ZaiumlrersquoCompte Rendu de lrsquoAcadeacutemie des Sciences ParisVol 323 No 2a 855ndash58

Lavachery P Cornelissen E Moeyersons J and DeMaret P 1996 lsquo30000 ans drsquooccupation 6 mois defouilles Shum Laka un site exceptionnel en Afriquecentralersquo Anthropologie et Preacutehistoire Vol 107 197ndash211

Leal M 2000 lsquoMicrorefugia small scale ice age forestremnantsrsquo Scripta Botanica Belgica Vol 20 AbstractCongres AETFAT Meise appendix 3

Leacuteonard J 1953 lsquoLes forecircts du Congo belgersquo LesNaturalistes Belges Vol 34 53ndash65

Letouzey R 1978 lsquoNotes phytogeacuteographiques sur lesPalmiers du Camerounrsquo Adansonia Vol 18 293ndash325

Letouzey R 1985 lsquoNotice de la cartephytogeacuteographique du Cameroun au 1500000rsquoInstitut de la Carte Internationale de la VeacutegeacutetationToulouse and Institut de la RechercheAgronomique Yaoundeacute

Magny M 1993 lsquoSolar influences on Holoceneclimatic changes illustrated by correlations betweenpast lake-level fluctuations and the atmospheric14C recordrsquo Quaternary Research Vol 40 1ndash9

Maley J 1982 lsquoDust clouds rain types and climaticvariations in tropical North Africarsquo QuaternaryResearch 18 1ndash16

Maley J 1987 lsquoFragmentation de la Forecirct DenseHumide Africaine et extension des biotopesmontagnards au Quaternaire reacutecent nouvellesdonneacutees polliniques et chronologiques Implicationspaleacuteoclimatiques et biogeacuteographiquesrsquo Palaeoecologyof Africa Vol 18 307ndash34

Maley J 1990 lsquoHistoire reacutecente de la forecirct densehumide africaine essai sur le dynamisme dequelques formations forestiegraveresrsquo in R Lanfranchiand D Schwartz (eds) Paysages Quaternaires delrsquoAfrique centrale Atlantique Meacutemoire ORSTOMParis ORSTOM 367ndash82

Maley J 1996 lsquoThe African rain forest maincharacteristics of changes in vegetation and climate

from the upper Cretaceous to the QuaternaryrsquoProceedings of the Royal Society of Edinburgh BiolScVol 104B 31ndash73

Maley J 1997 lsquoMiddle to late Holocene changes intropical Africa and other continents Paleomonsoonand sea surface temperature variationsrsquo in HNDalfes G Kukla and H Weiss (eds) Third milleniumBC climate change and old world collapse NATOAdvanced Sciences Institut Series Global EnvironmentalChange Berlin Springer 611ndash40

Maley J 1999 lsquoLrsquoexpansion du palmier agrave huile (Elaeisguineensis) en Afrique Centrale au cours des troisderniers milleacutenaires nouvelles donneacutees etinterpreacutetationsrsquo in Bahuchet S Bley D and PagezyH (eds) Lrsquohomme et la Forecirct Tropicale Bergier ParisTravaux de la Sociegraveteacute drsquoEcologie Humaine 237ndash54

Maley J and Brenac P 1998a lsquoVegetation dynamicspalaeoenvironments and climatic changes in theforests of West Cameroon during the last 28000yearsrsquo Review of Palaeobotany and Palynology Vol99 157ndash88

Maley J and Brenac P 1998b lsquoLes variations de laveacutegeacutetation et des paleacuteoenvironnements du sudCameroun au cours des derniers milleacutenaires Etudede lrsquoexpansion du Palmier agrave huilersquo in P Bilong JPVicat (eds) Geacuteosciences au Cameroun GEOCAMCameroun Yaoundeacute Presses Univ No 1 85ndash97

Maley J Brenac P Bigot S and Moron V 2000lsquoVariations de la veacutegeacutetation et despaleacuteoenvironnements en forecirct dense africaine aucours de lrsquoHolocegravene Impact de la variation destempeacuteratures marinesrsquo in M Servant S Servant-Vildary (eds) Dynamique agrave long terme desEcosystegravemes forestiers intertropicaux MeacutemoireUNESCO Paris UNESCO 205ndash20

Maley J and Chepstow-Lusty A 2001 lsquoElaeisguineensis Jacq (oil palm) fluctuations in centralAfrica during the late Holocene climate or humandriving forces for this pioneering speciesrsquo VegetationHistory and Archaeobotany Vol 10 117ndash20

Maley J and Giresse P 1998 lsquoEtude drsquoun niveauargileux organique du Mayombe (Congooccidental) riche en pollens drsquoElaeis guineensis etdateacute drsquoenviron 2800 ans BP Implications pour lespaleacuteoenvironnements de lrsquoAfrique Centralersquo in PBilong and JP Vicat (eds) Geacuteosciences au CamerounGEOCAM Cameroun Yaoundeacute Presses Univ No1 77ndash84

Martin L Fournier M Mourguiart P Sifeddine ATurcq B Absy ML and Flexor JM 1993lsquoSouthern Oscillation signal in South Americanpalaeoclimatic data of the last 7000 yearsrsquoQuaternary Research Vol 39 338ndash46

28

Mondjannagni A 1969 lsquoContribution agrave lrsquoeacutetude despaysages veacutegeacutetaux du Bas ndash Dahomeyrsquo Annales delrsquoUniversiteacute drsquoAbidjan seacuterie Geacuteographie Vol 1 187

Morley JJ and Dworetzky BA 1993 lsquoHolocenetemperature patterns in the South AtlanticSouthern and Pacific Oceansrsquo in Wright HFKutzbach JE Webb T Ruddiman WF Street-Perrott FA and Bartlein PJ (eds) Global climatessince the Last Glacial Maximum University ofMinnesota Press 125ndash35

Moron V Bigot S and Roucou P 1995 Rainfallvariability in subequatorial America and Africa andrelationships with the main sea surface temperaturemodes (1951ndash1990) International Journal ofClimatology 15 1297ndash1322

Muloko N Abernethy K White L Petit R andMaley J 1998 lsquoUtilisation des marqueursmoleacuteculaires dans la reconstitution de lrsquohistoire dela forecirct tropicale humide gabonaise le modegraveleAucoumea klaineanarsquo in R Nasi I Amsallem and SDrouineau (eds) Actes Seacuteminaire FORAFRILibreville Montpellier CIRAD-Forecirct 8

Muloko-Ntoutoumeacute N Petit R White L andAbernethy K 2000 lsquoChloroplast DNA variation ina rainforest tree (Aucoumea klaineana Burseraceae)rsquoin Gabon Molecular Ecology Vol 9 359ndash63

Nasi R 1997 lsquoLes peuplements drsquoOkoumeacutes auGabon Leur dynamique et croissance en zonecocirctiegraverersquo Bois amp Forecircts des Tropiques Vol 251 5ndash27

Nguetsop F Servant-Vildary S Roux M Reynaud-Farrera I Servant M and Wirrmann D 2000lsquoLac Ossa Cameroun relations statistiquesdiatomeacuteesmilieux aquatiques application agravelrsquoestimation des paleo-niveaux lacustres durant les5000 derniegraveres anneacutees Comparisons avec leschangements de la veacutegeacutetationrsquo in M Servant and SServant-Vildary (eds) Dynamique agrave long terme desEcosystegravemes forestiers intertropicaux MeacutemoireUNESCO Paris UNESCO 233ndash9

Oslisly R and White L 2000 lsquoLa relationhommemilieu dans la reacuteserve de La Lopeacute (Gabon)au cours de lrsquoHolocegravene les implications surlrsquoenvironnementrsquo in M Servant and S Servant-Vildary (eds) Dynamique agrave long terme desEcosystegravemes forestiers intertropicaux MeacutemoireUNESCO Paris UNESCO241ndash250

Perrois L and Notueacute JP 1997 Rois et Sculpteurs delrsquoOuest Cameroun La panthegravere et la mygale ParisKarthala and ORSTOM 387

Pinccedilon B 1990 lsquoLa meacutetallurgie du fer sur lesplateaux teacutekeacute (Congo) Quelle influence surlrsquoeacutevolution des paysages au cours des deux derniers

milleacutenairesrsquo in R Lanfranchi and D Schwartz (eds)Paysages Quaternaires de lrsquoAfrique centrale AtlantiqueMeacutemoire ORSTOM Paris ORSTOM 479ndash92

Reynaud-Farrera I Maley J and Wirrmann D1996 lsquoVeacutegeacutetation et climat dans les forecircts du sud-est Cameroun depuis 4770 ans BP Analysepollinique des seacutediments du Lac Ossarsquo CompteRendu de lrsquoAcadeacutemie des Sciences Paris Vol 322No 2a 749ndash55

Richards K 1986 lsquoPreliminary results of pollenanalysis of a 6000 year core from Mboandong acrater lake in Camerounrsquo Hull University GeographyDepartment Miscellaneous Serie Vol 32 14ndash28

Rind D 1995 lsquoDrying out the Tropicsrsquo New ScientistVol 5 36ndash40

Runge J 1996 lsquoPalaeoenvironmental interpretationof geomorphological and pedological studies in therain forest ldquocore-areasrdquo of eastern Zairersquo SouthAfrican Geographical Journal Vol 78 91ndash7

Saint-Vil J 1984 lsquoLa grande saison segraveche au GabonrsquoAnnales de lrsquoUniversiteacute Nationale du Gabon Vol 5107ndash19

Schnell R 1946 lsquoNote sur le Palmier agrave huile sareacutepartition et sa disseacutemination dans la reacutegionforestiegraverersquo Notes Africaines IFAN Dakar Vol 31 30ndash1

Servant M 1996 Dynamique agrave long terme desEcosystegravemes Forestiers Intertropicaux SymposiumECOFIT Bondy Reacutesumeacutes Paris ORSTOM amp CNRS335

Servant M Maley J Turcq B Absy ML Brenac PFournier M and Ledru MP 1993 lsquoTropical forestchanges during the late Quaternary in African andSouth American lowlandsrsquo Global amp PlanetaryChange Vol 7 25ndash40

Schwartz D 1992 lsquoAssegravechement climatique vers3000 BP et expansion Bantu en Afrique centraleatlantique quelques reacuteflexionsrsquo Bulletin de la SociegraveteacuteGeacuteologique de France Vol 163 353ndash61

Spichiger R and Blanc-Pamard C 1973lsquoRecherches sur le contact forecirct-savane en CocirctedrsquoIvoire Etude du recru forestier sur des parcellescultiveacutees en lisiegravere drsquoun ilot forestier dans le sud dupays Baouleacutersquo Candollea Vol 28 21ndash37

Swaine MD 1992 lsquoCharacteristics of dry forest inWest Africa and the influence of firersquo Journal ofVegetation Science Vol 3 365ndash74

Swaine MD and Hall JB 1986 lsquoForest structureand dynamicsrsquo in GW Lawson (ed) Plant Ecologyin West Africa The Hague Wiley 47ndash93

Tardy C Vernet JL Servant M Fournier MLeprun JC Pessenda LC Sifeddine A Solari

29

ME Soubiegraves F Turcq B Wengler L Vacher SJeacutereacutemie S Ceccantini G Cordeiro R and ScheelR 2000 Feux sols et eacutecosystegravemes tropicaux in MServant and S Servant-Vildary (eds) Dynamique agravelong terme des Ecosystegravemes forestiers intertropicauxMeacutemoire UNESCO Paris UNESCO 343ndash8

Valleacuterie M 1973 lsquoContribution agrave lrsquoeacutetude des sols ducentre sud Cameroun Types de diffeacuterenciationmorphologique et peacutedogeacuteneacutetique du sous-climatsubeacutequatorialrsquo Travaux amp Documents ParisORSTOM Vol 29 111

van der Hammen T and Hooghiemstra H 2000lsquoNeogene and Quaternary history of vegetationclimate and plant diversity in AmazoniarsquoQuaternary Sciences Review Vol 19 725ndash42

van Geel B Van der Plicht J Kilian MR KlaverER Kouwenberg JH Renssen H Reynaud-Farrera I and Waterbolk HT 1998 lsquoThe sharprise of 14C ca 800 cal BC possible causes relatedclimatic teleconnections and the impact on humanenvironmentsrsquo in WG Mook and J Van der Plicht(eds) Radiocarbon Vol 40 535ndash50

Vansina J 1990 Paths in the Rainforests Towards ahistory of political tradition in equatorial AfricaWisconsin University Press

Vincens A Schwartz D Elenga H Reynaud-Farrera I Servant M and Wirrmann D 1999lsquoForest response to climate changes in AtlanticEquatorial Africa during the last 4000 years BP andinheritance on the modern landscapesrsquo Journal ofBiogeography Vol 26 879ndash85

Vincens A Elenga H Reynaud-Farrera I SchwartzD Servant M and Wirrmann D 2000 lsquoReacuteponsedes forecircts aux changements du climat en AfriqueAtlantique Equatoriale durant les derniers 4000 anset heacuteritage sur les paysages veacutegeacutetaux actuelsrsquo in MServant and S Servant-Vildary (eds) Dynamique agravelong terme des Ecosystegravemes forestiers intertropicauxMeacutemoire UNESCO Paris UNESCO 381ndash7

Warnier JP 1985 Echanges deacuteveloppement ethieacuterarchies dans le Bamenda preacute-colonial(Cameroun) Studien zur Kulturkunde 76 FWiesbaden Steiner Verlag 323

White L 1995 Etude de la veacutegeacutetation de la Reacuteserve deLa Lopeacute Rapport Final Prog ECOFAC Gabon 140

White L and Oates JF 1999 lsquoNew data on thehistory of the plateau forest of Okomu southernNigeria an insight into how human disturbancehas shaped the African rain forestrsquo Global Ecology ampBiodiversity Letters Vol 8 355ndash61

White L Oslisly R Abernethy K and Maley J2000 lsquoLrsquoOkoumeacute (Aucoumea klaineana) expansionet deacuteclin drsquoun arbre pionnier en Afrique CentraleAtlantique au cours de lrsquoHolocegravenersquo in M Servantand S Servant-Vildary (eds) Dynamique agrave long termedes Ecosystegravemes forestiers intertropicaux MeacutemoireUNESCO Paris UNESCO 399ndash411

Wotling G Maheacute G Lebarbe L and LrsquoHote Y1995 lsquoAnalyse par les vecteurs reacutegionaux de lavariabiliteacute spatio-temporelle des preacutecipitationsannuelles lieacutees agrave la mousson africainersquo VeilleClimatique Satellitaire Lannion Vol 52 58ndash73

Youta Happi J and Bonvallot J 1996 lsquoLa disparitiondes savanes au centre Cameroun entre 1950 et1990rsquo in M Servant (ed) Dynamique agrave long termedes Ecosystegravemes Forestiers Intertropicaux SymposiumECOFIT Paris ORSTOM amp CNRS Bondy199ndash200 Reacutesumeacutes

Youta Happi J Hotyat M and Bonvallot J 2000lsquoLa colonisation des savanes par la forecirct agrave lrsquoest duCamerounrsquo in M Servant and S Servant-Vildary(eds) Dynamique agrave long terme des Ecosystegravemesforestiers intertropicaux Meacutemoire UNESCO ParisUNESCO 423ndash7

Zeven AC 1967 The semi-wild oil palm and itsindustry in Africa Agricultural Research ReportWageningen Univ 689 178

Zeven AC 1972 lsquoThe partial and completedomestication of the oil palmrsquo (Elaeis guineensis)Economic Botany 26 274ndash9

Zogning A Giresse P Maley J and Gadel F 1997lsquoThe Late Holocene palaeoenvironment in the LakeNjupi area west Cameroon implications regardingthe history of Lake Nyosrsquo Journal of African EarthSciences Vol 24 285ndash300

30