Coexistence of forest and savanna in an Amazonian area from a geological perspective

13 Pages·2009·1.15 MB·English

JournalofVegetationScience21:120–132,2010
DOI:10.1111/j.1654-1103.2009.01129.x
&2009InternationalAssociationforVegetationScience
Coexistence of forest and savanna in an Amazonian area from a
geological perspective
D.F. Rossetti, S.Almeida, D.D. Amaral, C.M. Lima &L.C.R.Pessenda
Abstract Conclusion: Landforms in eastern Marajo´ Island
Question: How can the coexistence of savanna and reflect changes in the physical environment due to
forest in Amazonian areas with relatively uniform reactivation of tectonic faults during the latest
climatesbeexplained? Quaternary. This promoted a dynamic history of
Location:EasternMarajo´ Island,northeastAmazo- channel abandonment, which controlled a set of
nia,Brazil. interrelated parameters (soil type, topography, hy-
Methods: The study integrated floristic analysis, ter- drology)thatdeterminedspecieslocation.Inclusion
rain morphology, sedimentology and d13C of soil of a geological perspective for paleoenvironmental
organic matter. Floristic analysis involved rapid eco- reconstruction can increase understanding of plant
logical assessment of 33 sites, determination of distributioninAmazonia.
occurrence, specific richness, hierarchical distribution
Keywords:Amazonia;Geologicalhistory;Paleolands-
andmatrixoffloristicsimilaritybetweenpairedvege- cape;Plantdistribution;Quaternary;d13Cisotope.
tation types. Terrain characterization was based on
analysisofLandsatimagesusing4(R),5(G)and7(B) Nomenclature:Cronquist(1981).
composition and digital elevation model (DEM).
Sedimentology involved field descriptions of surface
Introduction
and core sediments. Finally, radiocarbon dating and
analysis of d13C of soil profile organic matter and
Many studies have highlighted the importance
naturalecotoneforest-savannawasundertaken.
of climate on species distribution (Pearson & Daw-
Results:SlighttectonicsubsidenceineasternMarajo´
son2003;terSteegeetal.2003;Huntleyetal.2004),
Islandfavoursseasonalflooding,makingitunsuita-
however,Amazonianspeciesdonotrespondonlyto
ble for forest growth. However, this area displays
climate changes (Haffer 1990; Ayres & Clutton-
slightlyconvex-up,sinuousmorphologiesrelatedto
Brock 1992), and a wider range of environmental
paleochannels, covered by forest. Terra-firme low-
parameters should therefore be considered (Bush
land forests are expanding from west to east,
1994; Tuomisto et al. 1995; Bonaccorso et al. 2006;
preferentially occupying paleochannels and repla-
Coudun et al. 2006). Numerous hypotheses have
cing savanna. Slack, running water during channel
been formulated in this regard, considering the in-
abandonment leads to disappearance of varzea/
fluence of topography (Tuomisto et al. 1995;
gallery forest at channel margins. Long-abandoned
Vormisto et al. 2004), soil (Tuomisto & Ruokolai-
channels sustain continuous terra-firme forests, be-
nen1994)andgeology(Ra¨sa¨nenetal.1990;Vander
cause of longer times for more species to establish.
Hammen et al. 1992). A crucial step to understand
Recentlyabandonedchannelshavehadlesstimeto
Amazonian biodiversity is to identify the environ-
become sites for widespread tree development, and
mental parameters that, in addition to climate,
areeithernotvegetatedorcoveredbysavanna.
might have determined the distribution of modern
species. Of particular interest is the coexistence of
forest and savanna in many Amazonian areas,
Rossetti, D.F. (corresponding author, rossetti@dsr.
which is unlikely to respond only to climatic chan-
inpe.br): Instituto Nacional de Pesquisas Espaciais-
ges.PatchesofsavannawithintheAmazoniandense
OBT/DSR,RuadosAstronautas1758CP515,12245-
forest have been interpreted as relics of increased
970Sa˜oJose´ dosCampos,SP,Brazil
Almeida, S. & Amaral, D.D.: Museu Paraense Emı´lio aridityinthelatePleistocene(Ledruetal.2006)and
Goeldi,Coordenac¸a˜odeBotaˆnica,Av.Perimetral1901 earlytomid-Holocene(Pessendaetal.1998;Freitas
CP399,66077-530Bele´m,PA-Brazil etal.2001).
Lima,C.M.&Pessenda,L.C.R.:UniversidadedeSa˜o The goal of this study is to introduce an alter-
Paulo, Laborato´rio de 14C, Avenida Centena´rio 303, native model based on landform evolution to
13416000Piracicaba,SP-Brazil. explain the coexistence of forest and savanna in
CoexistenceofforestandsavannainanAmazonianarea 121
Fig.1. LocationmapofthestudyareainMarajo´ Island,northernBrazil,withdistributionofvegetationtypes(basedon
SIVAM/IBGE).Insideboxlocatesthestudyarea.
an Amazonian area. The example discussed herein Physiography andGeologic Framework
derives from eastern Marajo´ Island, located at the
mouth of the Amazon River. The eastern side of The study area (Fig. 1) is characterized by
this island sustains a mosaic of savanna and a tropical climate, with mean annual temperature
forest,whiledenseforestprevailsonitswesternside. of281Candprecipitationof2500to3000mmyr(cid:2)1.
Thisworkfocusesontherelationshipbetweenthese Vegetation types consist of two sharply-bounded
contrasting vegetation types and the geological groups: savanna (or campo), which consists mostly
characteristics of the study area. The aim is to de- of grassland, cerrado (Miranda & Carneiro 1994)
termine the influence of geological history on plant and savanna woodland in the eastern side of the
distributionintheAmazonianlowland. island; and dense tropical Ombrophyla forest
122 Rossetti,D.F.etal.
(e.g. Bastos 1984; Pires & Prance 1985; Henderson phytophysiognomic types, following the macro-
et al. 1991;Fig. 1) including terra-firme and season- physionomiesofSIVAM/IBGE(Fig.1),withafew
allyfloodedvarzea/galleryforestsinthewesternside modification aiming to preserve terms commonly
of the island. The campos from eastern Marajo´, used by local researchers (e.g. Bastos 1984; Pires &
where the study area is located, are flooded during Prance 1985; Henderson et al. 1991). The phyto-
rainy seasons for a period of 3 to 4 months.A few, physiognomieswereassigneduppercasesforForest
usually elongated, areas above high-water level are (F) and Savanna (S), followed by low cases to spe-
sitessuitablefortreegrowth.Soilsaredominatedby cify vegetation type (e.g. Sw5woodland savanna).
poorly drained, hydromorphic gleysols (Aquept, Statisticalprocessing wasapplied withthesoftware
USDA Soil Taxonomy 1999), with a few podzols PC-ORDforwindows(McCune&Mefford1997).
(Udult) (RADAM 1974). Terra-firme forests have Terrain characterization was based on inter-
distinct oligotrophic, acidic,moderatelydrained la- pretation of Landsat 5-TM (Ref. 224-060 and 225-
tosols(Udox)withasandyorsandy-muddytexture, 061,INPE)andLandsat7-ETM1(Ref.223-060and
whilemosaicsofterra-firmeforestandsavannahave 223-061, GLCF) images, acquired in August 2001.
yellow latosols (Udox). Areas dominated by grass- A 4(R), 5(G) and 7(B) band composition scheme,
land savanna might lack soil horizon, when mud processed with the software SPRING, provided a
settlingisstilltakingplace,eitherduringfloodingor goodviewofthegeomorphicfeaturesofinterestfor
all year round in depressed areas where lakes of this study. The geomorphic characterization was
varioussizesarepresent. completed with application of digital elevation
Geologically,westernMarajo´ Islandisinserted model (DEM) derived from the Shuttle Radar To-
in the Limoeiro Sub-Basin of the Marajo´ Graben pographyMission(SRTM).
System, while its eastern side is within the Para´ The field study consisted of morphological
Platform. The surface of this island is covered characterization to confirm the landforms pre-
mostlybypos-BarreirasdepositsofPleistoceneand viously recognized on the images, added to
Holocene age, in addition to a narrow belt of Mio- descriptions of sediments at the surface of the var-
cene deposits of the Barreiras Formation in its ious phytophysiognomies. At three localities this
easternmostedge.Quaternarydepositswereformed analysis was accompanied by core descriptions ob-
within a tectonically controlled N/NW-S/SE or- tained with a shallow percussion drilling system
ientated paleovalley (Rossetti & Valeriano 2007; (RKS-RoboticKeySystem;COBRAmk1(COBRA
Rossetti et al. 2007). According to these authors, Directional Drilling Ltd, UK). Three continuous
thisvalleywasfedbyapaleo-TocantinsRiver.The coresof5-cmdiameterandupto18-mlong(Fig.2)
abandonment of this river left numerous pa- wereobtainedonthelarge(T1andT3)andnarrow
leochannels that are still preserved in Marajo´ (T2) paleochannels described herein, and several
landscape(e.g.Radam1974;Porsani1981). other drillings were made on nearby floodplain
areas. Descriptions included definition of sedimen-
taryfaciesbasedonlithology,texture,structureand
Methods verticalfaciesrelationships.
Carbon isotope (12C, 13C) analyses of organic
The present investigation integrated floristic matter were performed at three localities (see lo-
composition, terrain morphology, sedimentology calities 1, 3 and 4 in Fig. 2). Samples of sediment
and d13C of organic matter from soil profiles. The units(85)werecollectedalongshallowmanualdril-
floristicsurveyconsistedofarapidecologicevalua- lings and trenches at 10-cm intervals. Grain size
tion (REE) (Sayre et al. 2000). Thirty-three sites analyses were performed using dry sieving. After
were established to record the distribution of the pre-treatment (Pessenda et al. 1996), the samples
variousvegetationpatterns.Descriptionsweremade were analysed at the Stable Isotope Laboratory of
within a 50-m radius in each site, inventorying CENA/USP (Sa˜o Paulo, Brazil). d13C values were
all vascular plant species, and annotating species also obtained from 14 modern plant species cate-
occurrence as follows: (1) Abundant: large popula- gorized as abundant in each sampling site. Results
tions forming monospecific patches or aggregates; are expressed as d13C with respect to PDB (Peedee
(2) Common: numerousbut withoutpatches orag- Belemnite Standard) standard using the conven-
gregates; and 3. Occasional or rare for random or tional d(%) notation d13C (%)5[(R /
sample
infrequent species. The Cronquist (1981) classifica- R )(cid:2)1] 1000, where R and R are
standard sample standard
tion system was applied to define plant family the 13C/12C ratio of the sample and standard, re-
names. The studied sites were grouped into eight spectively.Analyticalprecisionis (cid:3)0.2%.
CoexistenceofforestandsavannainanAmazonianarea 123
Fig.2. (a) Map of paleochannels in eastern Marajo´ Island, with location of the sites where vegetation types were char-
acterized(1-33).NotelocationofdrillcoresT1toT4,illustratedinFigs.6and7(InsideboxlocatesFig.3).(b)Distribution
ofvegetationpatternswithrespecttopastandmoderndepositionalenvironments.TheYaxisrepresentsthepercentageof
vegetationtypeswithinthe33investigatedsites.
Tensamplesofsedimentunitswerecollectedfor Results
radiocarbon analyses at the Beta Analytic Radio-
carbon Dating Laboratory (Florida, USA), by
Floristicanalysisandterraincharacterization
accelerator mass spectrometer (AMS). Seven sam-
ples derive from the three sedimentological cores Characterization of vegetation types in the 33
and three from the two cores collected for carbon sites (Fig. 2) revealed periodically inundated areas
isotope studies. Samples were pre-treated with acid withsavanna,andnumerouselongatedandsinuous
to remove carbonates and weaken organic bonds, beltswitheithersavannaor varioustypes ofarbor-
washed with alkali to remove secondary organic eal vegetation. A total of 174 plant species were
acids,andthencombinedwithacidagaintoprovide recorded (Appendix S1) and grouped into the fol-
more accurate dating. Conventional 14C ages were lowing life forms: tree (81 spp., 46.55%), herb (59
calibratedtocalendaryearsusingthePretoriaCali- spp.,39.90%),liana(11spp.,6.32%),shrub(9spp.,
brationProcedureprogram,basedontree-ringdata 5.17%), palm (7 spp., 4.02%) and epiphyte(7 spp.,
ascalibrationcurves(Talma&Vogel1993). 4.02%)(Table1).
124 Rossetti,D.F.etal.
Table1. Frequency,absolute(n)andrelative(%)speciesrichnessofplantformsstudiedinMarajo´ Island.
Vegetationtype Numberof Shrub% Tree% Liana% Epiphyte Herb% Stem% Species
Species (n) (n) (n) %(n) (n) (n) Richness
Abs Rel
(n) (%)
Seasonallyfloodedvarzea/galleryforest(Fv) 53 18.84 13.71 22.22 100.00 18.29 7.69 53 16.77
(13) (17) (6) (1) (15) (1)
Anthropogenicallymodifiedterra-firme 73 23.19 27.42 18.52 (0) 15.85 38.46 73 23.10
(Ftfam). (16) (34) (5) 0.00 (13) (5)
Continuousterra-firmeforest(Ftf) 23 8.70 7.26 7.41 (0) 4.88 15.38 23 7.28
(6) (9) (2) 0.00 (4) (2)
Islandofterra-firmeforest(Ftfi) 64 20.29 27.42 14.81 (0) 12.20 15.38 64 20.25
(14) (34) (4) 0.00 (10) (2)
Mosaicofterra-firmeforestandwoodland 43 15.94 11.29 22.22 0,00 12.20 15.38 43 13.61
savanna(Ftf/Sw) (11) (14) (6) (0) (10) (2)
Woodland–shrublandsavannas(Sws) 13 1.45 1,61 0.00 0.00 12.20 0.00 13 4.11
(1) (2) (0) (0) (10) (0)
Grassland–shrublandsavanna(Sgs) 27 10.14 3.23 0.00 0.00 18.29 7.69 27 8.54
(7) (4) (0) (0) (15) (1)
Woodlandsavanna(Sw) 20 1.45 8.06 14.81 0.00 6.10 0.00 20 6.33
(1) (10) (4) (0) (5) (0)
Table2. Relationship between type of environment and plant physiognomy for each site, with indication of number of
species/site(seeFig.2bforlegendofplantcategories).
Environment Vegetationtype/site(numberofspecies)
Fv Ftfam Ftf Ftfi Ftf/Sw Sws Sgs Sw
Sandypaleochannelslightly 12(27) 1(38) 24(2) 14(17) 9(12)
higherthansurroundingplains(Holocene) 32(8) 4(15) 31(17) 16(18) 20(21)
13(25) 19(13) 21(5)
18(14) 30(24) 22(13)
33(14) 23(5)
29(4)
Sandymudplain(Holocene) 2(11) 17(4)
3(3)
Mudplain(Holocene) 5(8)
6(11)
8(13)
Iron-cementedsandonMiocenelateriticpaleosol 7(21)
Mudatthemarginofmodernactivechannel 10(19)11(17)
Heterolithic(sand/mud)paleochannelslightly 25(7) 15(22) 26(13)
higherthansurroundingplains(Holocene) 27(18) 28(10)
The highest absolute and relative species rich- paleochannel areas. Mosaics of terra-firme forest
ness (Table 1) was recorded for anthropogenically- and woodland savanna (Ftf/Sw)are found on both
modified terra-firme forest (Ftfam) (n573; deposits.Contactbetweenplanttypesiseithersharp
23.10%), followed by islands of terra-firme forest orprogressive.Paleochannelswitharborealvegeta-
(Ftfi) (n564; 20.25%). Vegetation types with the tion consist of elongated, sinuous to anastomosing
lowest species diversity are woodland-shrub sa- features up to 1.2-km wide (average 0.6 to 0.9km;
vanna(Sws)(n513;4.11%)andwoodlandsavanna Figs. 2, 3a and b). DEM analysis suggests that the
(Sw)(n520;6.33%).Themajorityofthebeltswith paleochannels stand slightly higher relative to ad-
forest(i.e.vegetationtypesFv,Ftfam,Ftf,Ftfi,Sws jacentareas(Fig.3bandc).Fieldworkrevealedthat
andSw)isrelatedtopaleochannels(Table2;Fig.3a thisismostlyduetocombinationoftreecanopyand
and b) and, to a lesser extent, abandoned bars. smoothterrainelevationofafewtensofcentimeters
Grassland-shrublandsavannas(Sgs)arewidespread upto2-3m.
on flat and slightly low-lying paleofloodplains, al- The paleochannels record different phases of
though they also occur discontinuously over development ranging from fully abandoned, with a
CoexistenceofforestandsavannainanAmazonianarea 125
Fig.3. CharacterizationofpaleochannelsineasternMarajo´ Island.(a)Landsat-TM,4(R),5(G)and7(B)compositionil-
lustratingamainwest-northwestoriented,meanderingpaleochannel.(b)Detailofapaleochannelareacoveredbytypical
wetlowlandforest.(c)Topographicprofiletransversetothepaleochannel(seeprofileI-IIinAforlocation)withdataob-
tained from SRTM-DEM. (d) Partly abandoned channel (Pac) with a slight concave profile. Vegetation at the channel
margin(Vcm)withseasonallyflooded,varzea/galleryforest(Fv)elementsisstressedanddyingduetothelackofpermanent
water.(e)Sharpcontactbetweennon-vegetatedsandypaleochanneldeposits(Chd)andmuddysurroundingflatgrassland
areas(Pd)(paleochanneldepositsdiagonaltothepicture).Notealsoanarrowandlongbeltoftreesdevelopedinaslightly
parallelpaleochannel(backgroundarrows).
typicalconvex-uptopography,topartlyabandoned, lost their leaves (Fig. 3d). This evidences stress due
with a noticeable smooth depression, although wa- to lack of permanent flows within channels, which
ter runoff was cut off. In the latter, relics of are now almost completely filled with sediments.
seasonallyflooded,varzea-gallerytreesarestillpre- Alternatively,channelmorphologyiseithercovered
sent, but they are sparsely distributed around the by vegetation (usually wet lowland forests and, less
channel margins, and many have partly or entirely commonly, savanna) or vegetation is absent, in
126 Rossetti,D.F.etal.
whichcasethetopmostsandysedimentisexposedto and Holocene. Non-forested floodplain areas are
the surface. A sharp contact exists between inner composed of mud packages up to 22 m thick
andouterchannelareas(Fig.3e). (Fig. 5a and b). These might be interbedded with
Cores from paleochannels display sand succes- streaks and discontinuous laminae or packages
sions that grade upward into sand and mud o0.5-m thick of very fine-grained sands. Radio-
interbeddings (Fig. 4), recording typical channel carbonages(Figs.4and5)rangefrom7900((cid:3)40)
fillingfollowedbyabandonment.Radiocarbonages 14C years BP at depths of 12m up to 3960 ((cid:3)40)
(Fig. 4) indicate deposition in the latest Pleistocene 14CyearsBPonly2mfromthesurface.
Fig.4. Lithostratigraphicprofilesfromcoresobtainedinpaleochannelareas(seeFig.2forlocationofdrillcoresT1toT3).
AfewradiocarbonagesalsoplottedintheseprofilesindicateanearlyHoloceneage.
CoexistenceofforestandsavannainanAmazonianarea 127
Fig.5. (a)Viewofatypicalnon-vegetatedmuddyareaaroundapaleochannel(vegetatedbackground)fromeasternMarajo´
Island.(b)Anexampleoflithostratigraphic profiledrilledinamuddyplainarea.Theradiocarbonageshowsdeposition
contemporaneoustothepaleochanneldepositsshowninFig.4(seeFig.4forlegend).
14Candd13Coforganicmatterfromsoilprofiles savanna area, there isan upwardenrichmentinthe
valuesupto1.0-mdepth,followedbydepletionnear
The selected ecotone forest-savanna used for the surface. Radiocarbon dating obtained from lo-
studying d13C of organic matter comes from the cality1providedonlylatePleistoceneandHolocene
westsideofthestudyarea(seelocations1,3and4in
ages.
Fig. 2). In forested paleochannel areas (localities 1
and4),11dominantplantspecieshadthefollowing
d13C values: Aniba citrifolia ((cid:2)33.1%), Cordia tet- Discussion
randra ((cid:2)29.2%), Desmoncus polyacanthos
((cid:2)36.5%),Maquiracoriaceous((cid:2)32.6%),Anemo- Oneofthemajorproblemstoevaluateenviron-
paegma chrysoleucum ((cid:2)31.5%), Parinari excelsa mental influence on the distribution of Amazonian
((cid:2)32.3%), Psychotria racemosa ((cid:2)31.7%), Pter- plant species is the lack of geological and geomor-
ocarpus amazonicus ((cid:2)32.2%) and Simaba phological information. Available low-resolution
multiflora((cid:2)29.9%),andtheherbsHeliconiapsita- maps, in general provided at a scale of 1:1000000,
corum (cid:2)32.9% and Bambusa sp ((cid:2)27.1%). The arefarfromadequatetoapproachthemainfactors
grass species Panicum laxum ((cid:2)27.7%) and Pani- controlling species distribution. Floristic survey in-
cum millegrana ((cid:2)30.4%) and the shrub Senna tegrated with terrain characterization helps to
reticulata ((cid:2)28.9%) were the dominant plants in discussthecontrolsonplantdistributioninMarajo´
the savanna vegetation located near the pa- Island. The coexistence of large areas of savanna
leochannelarea(locality3). and wet lowland forest has not yet been discussed.
The results of d13C of organic matter re- Other savanna patches in Amazonia (e.g. Takeuchi
presentative of forest (localities 1 and 4) and 1960; Sanaiotti et al. 2002) have been related to ei-
savanna(locality3),aswellastheradiocarbonages thermodernclimatechangesatalocalscaleorrelics
for these profiles, are shown in Fig. 6. The values of dry vegetation from the glacial episodes (Ledru
indicateanupwardisotopicdepletionintheforested etal.2006).
areas, with a gradual enrichment (i.e. up to An alternative model is proposed considering
(cid:2)18.5%) from 2.20 to 0.70m in locality 4. In the the geological history. An important point to be
128 Rossetti,D.F.etal.
Fig.6. d13Cvalues(%)oforganicmatterfromsoilsinrelationtodepth(correspondingtolocalities1,3and4inFig.2).
discussed is the sharp contrasts in vegetation form surrounding areas are mostly muddy. Taking into
from forestedtonon-forested areas inthe centre of account only these lithological compositions, more
the island. Eastern Marajo´ Island is dominated by fertilesoilscouldhavebeenfavouredonthemuddy
savanna because a great part of its surface remains areas due to their high chemical nutrient content
submerged during most of the year. The increased (i.e. Mg, K, Ca, Fe from clays) that contrasts with
wateraccumulationonthissideoftheislandreflects the Si-dominated quartzose sands from pa-
its lower elevation due to slight subsidence. The leochannel areas. However, there is an overall
sharp contact between forest and savanna occurs consensusamongpedologiststhatchemicalcompo-
along a major NW-SE to NNW-SSE fault zone re- sition is rarely the main factor determining soil
activated during the latest Quaternary (Rossetti fertility. In fact, the paleochannels have latosols,
etal.2007).Anumberofotherstudieshaveaddressed whilehydromorphicgleysolsandpodzolsdominate
theimportanceoftectonicsinthelatestTertiaryand thesurroundingareas. Thesearetypically nutrient-
Quaternary in Marajo´ Island (e.g. Costa et al. 2001; poor soils, with the two latter types been highly
Rossetti & Valeriano 2007). Following subsidence, restrictive to the development of deep-rooted
eastern Marajo´ Island progressively stabilized, plants due to their development on muddy, water-
promoting a complex history of channel/bar estab- saturatedenvironments.Ratherthanreflectingche-
lishmentandabandonment(Rossettietal.2008).As mical availability, these soil types respond to a
discussed below, and also documented elsewhere in combination of factors, which mostly include sedi-
South America (e.g. Sarmiento & Pinillos 2001), the ment texture, topography and water content. As
dynamicevolutionoftheselandformscontrolledaset discussed in detail below, these factors are highly
ofinterrelatedparameters(i.e.soiltype,topography, dependent on landform evolution through time in
hydrology) that guided plant distribution in eastern easternMarajo´.
Marajo´ Island. Topography,typeofsub-surface water andthe
There are several studies in the literature relat- history of drainage abandonment had crucial roles
ing contrasts in plant distribution to nutrient in tree growth in eastern Marajo´ Island, as also
content (e.g. Tuomisto et al. 1995; Coudun et al. documentedinotherAmazonianareas(e.g.Mertes
2006). Although this factor cannot be ruled out in etal.1995).Thisissuggestedbythefactthatforests
eastern Marajo´ Island without undertaking further are, in general, related to paleochannel and bar de-
studies focusing on soil minerals, base content at posits with smooth topographies. Abandoned bars
root reach and soluble nutrient concentrations, the naturallystandaboveadjacentareas.Paleochannels
distributionofforestandsavannaseemstorespond might have adopted a slightly positive topography
to other influences. Areas of abandoned channels due to marginal levees, added to the fact that in-
andbarshaveasandyquartzosecomposition,while channelsandsarelessliabletocompaction,relative
CoexistenceofforestandsavannainanAmazonianarea 129
Fig.7. ProposedmodelfortheoriginofvegetationtypesineasternMarajo´ Island.(a)Depositionofsandandmudwithin
channelsandsurroundingareas,respectively.(b)Asthechannelflowdecreasedinenergy,thechannelsbecameprogressively
shallower and eventually abandoned. Decreasing flows caused great stress on gallery vegetation, which was replaced by
savanna(notdepicted).(c)Savannaswereprogressivelyreplacedbyterra-firmeforest,withspeciesderivedfromforested
areas.d.Continuityofthisprocessmightamplifytheforestfrompaleochannelsoversurroundingareas.
to more malleable clay layers from nearby un- contributed to different plant groups in eastern
confined depositional settings (Porsani 1981). Marajo´ Island. The proposed model considers that
Although smooth, these topographies are re- the slack running water during channel abandon-
sponsible for protecting some areas from the effect ment led to the disappearance of varzea/gallery
of prolonged flooding, which favours terra-firme forest at channel margins. The occurrence of many
forests. Many previous works have highlighted the modern channels in process of abandonment with
importance of topography on the differentiation of decayinggalleryforests(Fig.7b)inspiredthismod-
plantspeciesinAmazonia(e.g.Tuomistoetal.1995; el. When channels are fully abandoned and filled
Vormistoetal.2004). with sediment, the channel surface is occupied first
The type of sub-surface water might have been by grasslands and ultimately by terra-firme forests
another factor influencing tree growth in the study (Fig. 7a–c). First, island and mosaics of terra-firme
area. A geophysical investigation indicates that and savanna are established over confined, sinuous
sands from paleochannel and bar deposits contain channel morphologies, then the forest expands be-
freshwater captured from either rain or running yond the channel limits through time, resulting in
flows during wet seasons, while mud sealing pre- more continuous areas with terra-firme forest
cludessurfacefreshwaterpenetrationdownthrough (Fig.7c).Thismodelisconsistentwiththefactthat
muddy deposits in the surrounding flat areas (Por- continuous terra-firme forest shares the highest
sani 1981). As a result, the latter have connate numberoftaxawithislandsofterra-firmeforestand
brackish to saline waters, constituting stressed en- mosaics of terra-firme forest and savanna, support-
vironments that are unsuitable for the growth of ingthatthelatterareintheprocessofdevelopment,
varzea and terra-firme forests. Similar controls on benefiting from species migration from adjacent
vegetation growth dictated by groundwater salinity forests. The transitional zone between forested
driven by river dynamics are recorded in the litera- paleochannelsandnearbygrasslandareasmight be
ture(e.g.Thevsetal.2008a). evidence that arboreal species present in the latter
In combinationwiththefactorsabove, the his- are migrating towards savanna areas (Fig. 7d).
tory of channel abandonment additionally Thus, channels abandoned for longer times sustain