HOLOCENE VEGETATION AND CLIMATE HISTORY OF SAVANNA-FOREST ECOTONES IN NORTHEASTERN AMAZONIA

Abstract

This is my doctoral dissertation.

Full text

HOLOCENE VEGETATION AND CLIMATE HISTORY OF
SAVANNA-FOREST ECOTONES IN
NORTHEASTERN AMAZONIA
by
MAURO BEVILACQUA DE TOLEDO
B.S., Instituto de Ciências e Tecnologia Maria Thereza
M.S., Universidade Federal Fluminense
A dissertation submitted to the Department of Biological Sciences
of Florida Institute of Technology in partial fulfillment
of the requirements for the degree of
DOCTOR OF PHILOSOPHY
in
BIOLOGICAL SCIENCES
Melbourne, Florida
May 2004

HOLOCENE VEGETATION AND CLIMATE HISTORY OF
SAVANNA-FOREST ECOTONES IN
NORTHEASTERN AMAZONIA
A DISSERTATION
by
MAURO BEVILACQUA DE TOLEDO
Approved as to style and content by:
______________________________________
Mark B. Bush, Ph.D., Chairperson
Associate Professor
Department of Biological Sciences
______________________________________
John G. Morris, Ph.D., Member
Associate Professor
Department of Biological Sciences
______________________________________
Robert van Woesik, Ph.D., Member
Associate Professor
Department of Biological Sciences
______________________________________
Elizabeth A. Irlandi, Ph.D., Member
Assistant Professor
Oceanography Department
______________________________________
Gary N. Wells, Ph.D.
Professor and Head
Department of Biological Sciences
May 2004

iii
ABSTRACT
HOLOCENE VEGETATION AND CLIMATE HISTORY OF
SAVANNA-FOREST ECOTONES IN
NORTHEASTERN AMAZONIA
By Mauro Bevilacqua de Toledo,
B.S., Faculdades Integradas Maria Thereza;
M.S., Universidade Federal Fluminense
Chairperson of Advisory Committee: Mark B. Bush, Ph.D.
The main goal of this study was to investigate how climate and human
activity may have influenced ecotonal areas of disjunct savannas within Brazilian
Amazonia. The fossil pollen records of 6 lakes were used to provide a regional
Holocene paleoecological history of northeastern Amazonia: Lakes Marcio and
Tapera from Amapá; Lakes Santa Maria, Geral, and Saracuri from Prainha, near
Monte Alegre (Pará); and Lake Jacaré from northeastern Roraima. The fossil pollen
and charcoal analyses performed on these cores identified vegetation changes and
human occupation near the lakes. Although each individual record indicated local
variation, a general pattern was recognized. All the lakes are Holocene in age, and
exhibited rising water levels until ca. 6500 – 5300 years BP, which was documented

iv
by an expansion of wetland/swamp taxa. Even though a sedimentary hiatus was
recognized in the sediment cores from Lakes Marcio and Tapera, a marked
vegetation change was evident from the pollen record. Because the timing of the
hiatus overlapped with the highest sea-level of the Holocene, which would have
increased the local water table, the sedimentary gap was probably caused by reduced
Amazon River discharge, due to decreased precipitation in the Andes between 8000
and 5000 years BP. Even though the resumption of sedimentation in Lake Marcio is
contemporaneous with increasingly wet conditions in the Andes after 5000 years BP,
Lake Tapera remained dry until 2000 years BP, which coincided with another sea-
level rise. Both pollen records documented markedly increased fire frequency
following the replacement of the swamp forest by flooded savanna when
sedimentation resumed. The lakes from Prainha provided a strong record of human
impacts, which made the climatic influences on the vegetation changes very difficult
to infer. As the pattern of charcoal accumulation displays peaks of fire frequency
alternating between the lakes through time, it is probable that this region was
intensively occupied by pre-Columbian peoples. A short period of forest re-growth
recorded between 5000 and 3200 years BP is coincident with both low fire frequency
and wet conditions. The sediments from Lake Jacaré recorded the presence of
savanna throughout the history of the lake. A period of forest regrowth was also
recorded at Jacaré between 5700 and 4700 years BP. That the timing of forest
expansion recorded by Lakes Jacaré, Santa Maria and Saracuri overlaps, suggests an
underlying climatic forcing, although decreased human impacts cannot be totally

v
dismissed. It is possible that a dry event at ca. 4700 years BP in Roraima caused the
Lake Jacaré to dry out until 1700 years BP. However, additional radiocarbon dates
are necessary to confirm this sedimentary hiatus. The fossil pollen records suggested
that the proportions of savanna and forest fluctuated throughout the Holocene. Even
though no apparent synchronism was found between the study sites, they showed an
overall pattern of increasingly wet conditions, which did not support a widespread
mid-Holocene dry event. Although all the study sites recorded long-term occupation
by pre-Columbian peoples, it is still unclear if these disjunct savannas have an
anthropogenic origin. The main conclusion that can be drawn from this study is that
Amazonian savannas have complex and independent evolutionary histories.
Therefore, no unifying hypothesis can provide for the formation of all Amazonian
savannas.

vi
ACKNOWLEDGEMENTS
Many people and institutions contributed to this work. First of all, I would
like to thank my advisor, Dr. Mark Bush, for offering me the great opportunity of
working with him in this exciting project. He may not have taught me “everything I
know about Palynology”, but I definitely owe to him everything I know about
Paleoecology, not to mention my world famous driving skills. Thanks are due to Dr.
Bruno Turcq (IRD-France), for the mineralogical analysis of Lake Jacaré. I would
also like to thank all my friends from the laboratory, especially Dr. Chengyu Weng,
Claudia Listopad, Dunia Urrego, Jen Hanselman, John Shepherd, and of course the
new arrivals, Alex Correa and William Gosling for their constant encouragement,
support and friendship. I am very grateful to my committee members, Dr. Elizabeth
Irlandi, Dr. John Morris, and Dr. Robert van Woesik for their patience, flexibility
and trust. Thanks to Dr. Gary Wells for his interest and advice, and Carolyn Sorrell
for her never-ending patience and assistance. Many thanks are due to several people
from Brazil and their institutions: Dr. Paulo de Oliveira, Dr. Odete F. da Silveira
(IEPA), Marco Antonio Chagas and Odécio (SEMA), Neiva Lúcia (IBAMA), Dr.
Reinaldo Imbrozio (INPA-Roraima), Ronaldo Justo, Ana Amélia Lavenere (UESC),
LAGEMAR-UFF. I would like to thank all my family, especially Maria da Graça S.
Bevilaqua, Gabriel B. de Toledo, Sérgio Luiz B. de Toledo, and Débora Alves Silva
for their unconditional love and support. Financial support provided by CNPq
(200065/99-8) and NSF (DEB-9732951)

vii
TABLE OF CONTENTS
ABSTRACT ............................................................................................................... iii
ACKNOWLEDGEMENTS ....................................................................................... vi
TABLE OF CONTENTS ...........................................Error! Bookmark not defined.
LIST OF TABLES .......................................................................................................x
LIST OF FIGURES.................................................................................................... xi
LIST OF FIGURES.................................................................................................... xi
CHAPTER I..................................................................................................................1
THE AMAZONIAN REGION.....................................................................................1
INTRODUCTION........................................................................................................1
STUDY GOALS ......................................................................................................1
VEGETATION.............................................................................................................2
RAINFORESTS .......................................................................................................2
GALLERY FORESTS .............................................................................................4
SAVANNAS ............................................................................................................4
WHITE-SAND SAVANNAS ..................................................................................6
AMAZONIAN SAVANNAS.......................................................................................7
ENVIRONMENTAL REQUIREMENTS................................................................7
SAVANNA DISTRIBUTION..................................................................................8
SAVANNA ORIGINS ...........................................................................................11
CLIMATIC ORIGINS.........................................................................................11
ANTHROPOGENIC ORIGINS ..........................................................................12
USING POACEAE AS A SAVANNA INDICATOR...........................................12
AMAZONIAN ARIDITY - THE REFUGIA HYPOTHESIS ...............................14
THE ARCHAEOLOGICAL RECORD .................................................................18
CHARCOAL IN THE SEDIMENTS.........................................................................22
HOLOCENE HISTORY OF AMAZONIAN SAVANNA-FOREST ECOTONES..23
DESIGN OF STUDY.................................................................................................26
TESTING HYPOTHESES.........................................................................................28
CHAPTER II ..............................................................................................................29
PALEOECOLOGICAL RECORD OF AMAPÁ (BRAZIL).....................................29
INTRODUCTION......................................................................................................29
STUDY SITE .........................................................................................................31
HYPOTHESES.......................................................................................................32
METHODS.................................................................................................................34
RESULTS...................................................................................................................37
STRATIGRAPHY..................................................................................................37
RADIOCARBON DATES.....................................................................................39
PALEOECOLOGICAL RECORD – LAKE MARCIO.........................................42
M1A (470 cm – 320 cm)......................................................................................42
M1B (320 cm – 200 cm)......................................................................................44
M1C (200 cm – 120 cm)......................................................................................44
M2 (120 cm – 0 cm) ............................................................................................45

viii
PALEOECOLOGICAL RECORD – LAKE TAPERA .........................................47
T1A (200 cm – 110 cm).......................................................................................47
T1B (110 cm – 75 cm).........................................................................................49
T2 (75 cm – 0 cm)................................................................................................49
MULTIVARIATE ANALYSIS – LAKE MARCIO .............................................52
AXIS 1 vs. AXIS 2 ..............................................................................................52
AXIS 1 vs. DEPTH..............................................................................................52
MULTIVARIATE ANALYSIS – LAKE TAPERA..............................................54
AXIS 1 vs. AXIS 2 ..............................................................................................54
AXIS 1 vs. DEPTH..............................................................................................54
AXIS 2 vs. DEPTH..............................................................................................55
DISCUSSION.............................................................................................................58
CHAPTER III.............................................................................................................66
PALEOECOLOGICAL RECORD OF PRAINHA REGION (BRAZIL) .................66
INTRODUCTION......................................................................................................66
STUDY SITE .........................................................................................................69
HYPOTHESES.......................................................................................................71
METHODS.................................................................................................................72
RESULTS...................................................................................................................75
STRATIGRAPHY..................................................................................................75
RADIOCARBON DATES.....................................................................................77
PALEOECOLOGICAL RECORD – LAKE SANTA MARIA .............................80
SM1 (650 cm – 594 cm; ca. 8300 yrs BP – 7800 yrs BP)...................................83
SM2 (594 cm – 440 cm; 7800 yrs BP – 5800 yrs BP).........................................83
SM3 (440 cm – 297 cm; 5800 yrs BP – 4290 yrs BP).........................................84
SM4 (297 cm – 180 cm; 4290 yrs BP – 3160 yrs BP).........................................84
SM5 (180 cm – 0 cm; 3160 yrs BP until present) ...............................................85
PALEOECOLOGICAL RECORD – LAKE GERAL ...........................................85
G1 (596 cm – 520 cm; 8260 yrs BP – 8000 yrs BP) ...........................................85
G2 (520 cm – 390 cm; 8000 yrs BP – 6700 yrs BP) ...........................................86
G3 (390 cm – 130 cm; 6700 yrs BP – 2390 yrs BP) ...........................................86
G4 (130 cm – 0 cm; 2390 yrs BP until present) ..................................................86
PALEOECOLOGICAL RECORD – LAKE SARACURI.....................................88
S1 (860 cm – 740 cm; 8480 yrs BP – 7490 yrs BP) ............................................88
S2 (740 cm – 470 cm; 7490 yrs BP – 4960 yrs BP) ............................................88
S3 (470 cm – 190 cm; 4960 yrs BP – 2100 yrs BP) ............................................88
S4 (190 cm – 0 cm; 2100 yrs BP until present)...................................................90
MULTIVARIATE ANALYSIS – LAKE SANTA MARIA..................................91
AXIS 1 vs. AXIS 2 ..............................................................................................91
AXIS 1 vs. DEPTH..............................................................................................92
AXIS 2 vs. DEPTH..............................................................................................92
MULTIVARIATE ANALYSIS – LAKE GERAL ................................................94
AXIS 1 vs. AXIS 2 ..............................................................................................94
AXIS 1 vs. DEPTH..............................................................................................96

ix
MULTIVARIATE ANALYSIS – LAKE SARACURI .........................................96
AXIS 1 vs. AXIS 2 ..............................................................................................96
AXIS 1 vs. DEPTH..............................................................................................96
AXIS 2 vs. DEPTH..............................................................................................99
DISCUSSION...........................................................................................................100
PALEOECOLOGICAL RECORD OF RORAIMA (BRAZIL) ..............................108
INTRODUCTION....................................................................................................108
STUDY SITE .......................................................................................................110
HYPOTHESES.....................................................................................................112
METHODS...............................................................................................................113
RESULTS.................................................................................................................116
RADIOCARBON DATES...................................................................................116
STRATIGRAPHY................................................................................................117
PALEOECOLOGICAL RECORD – LAKE JACARÉ........................................118
J1 (132 cm – 65 cm; ca. 9040 yrs BP – 6500 yrs BP) .......................................119
J2 (65 cm – 50 cm; 6500 yrs BP – 5620 yrs BP)...............................................120
J3 (50 cm – 30 cm; 5620 yrs BP – 3400 yrs BP)...............................................120
J4 (30 cm – 0 cm; 3400 yrs BP until present)....................................................123
MULTIVARIATE ANALYSIS – LAKE JACARÉ ............................................123
AXIS 1 vs. AXIS 2 ............................................................................................123
AXIS 1 vs. DEPTH............................................................................................125
DISCUSSION...........................................................................................................126
CONCLUSIONS ......................................................................................................134
CHAPTER V............................................................................................................135
HOLOCENE VEGETATION AND CLIMATE HISTORY OF SAVANNA-
FOREST ECOTONES IN NORTHEASTERN AMAZONIA.................................135
SOLAR FORCING ..............................................................................................138
ITCZ MIGRATION .............................................................................................140
ENSO EVENTS ...................................................................................................144
RELATIVE SEA-LEVEL CHANGE ..................................................................147
HUMAN OCCUPATION ....................................................................................149
INTEGRATION OF ALL THE FACTORS ............................................................151
THE AMAPÁ LAKES.........................................................................................151
THE PRAINHA LAKES......................................................................................157
LAKE JACARÉ ...................................................................................................159
CONCLUSIONS ......................................................................................................164
LITERATURE CITED.............................................................................................166

x
LIST OF TABLES
Table 2. 1. Sediment description of core from Lake Marcio (Amapá - Brazil). ........37
Table 2. 2. Sediment description of core from Lake Tapera (Amapá – Brazil).........39
Table 2. 3. Radiocarbon dates from Lake Marcio sediment core...............................40
Table 2. 4. Radiocarbon dates from Lake Tapera sediment core. ..............................41
Table 3. 1. Sediment description of core from Lake Santa Maria (Pará - Brazil)......75
Table 3. 2. Sediment description of core from Lake Geral (Pará – Brazil)................75
Table 3. 3. Sediment description of core from Lake Saracuri (Pará – Brazil). ..........77
Table 3. 4. Radiocarbon dates from Lake Santa Maria sediment core.......................79
Table 3. 5. Radiocarbon dates from Lake Geral sediment core (Bush et al. 2000)....80
Table 3. 6. Radiocarbon dates from Lake Saracuri sediment core.............................80
Table 4. 1. Radiocarbon dates from Lake Jacaré sediment core. .............................116
Table 4. 2. Sediment description of core from Lake Jacaré (Roraima - Brazil).......117

xi
LIST OF FIGURES
Figure 1. 1. Present vegetation distribution of northern and central Brazil. White dots
represent study sites (modified from IBGE 1992).................................................3
Figure 1. 2. Climate and precipitation patterns of northern Brazil. Red dots represent
study sites (modified from IBGE 1990). ...............................................................8
Figure 1. 3. False color Landsat TM image showing bands 7, 4, and 2. Pink
represents highly reflective surfaces such as savannas, urban areas and bare
ground, whereas green represents the more complex structure of forests. Disjunct
savanna in Amapá (red) surrounded by tropical rainforest (green) at one of the
studied sites............................................................................................................9
Figure 1. 4. False color Landsat TM image showing bands 7, 4, and 2. Pink
represents highly reflective surfaces such as savannas, urban areas and bare
ground, whereas green represents the more complex structure of forests. Disjunct
savanna (red) on eastern Marajó Island, and tropical rainforest (green) on
western Marajó Island, at the mouth of the Amazon River...................................9
Figure 1. 5. False color Landsat TM image showing bands 7, 4, and 2. Pink
represents highly reflective surfaces such as savannas, urban areas and bare
ground, whereas green represents the more complex structure of forests. Disjunct
savannas (red) around Alter do Chão, Santarém, and Monte Alegre surrounded
by tropical rainforest (green), and one of the studied sites..................................10
Figure 1. 6. False color Landsat TM image showing bands 7, 4, and 2. Pink
represents highly reflective surfaces such as savannas, urban areas and bare
ground, whereas green represents the more complex structure of forests. Disjunct
savannas (red) of northern Roraima at one of the studies sites and at lower Rio
Negro surrounded by tropical rainforest (green). ................................................10
Figure 1. 7. Sketch map of vegetation distribution in Amazonia during the Last
Glacial Maximum according to the Refuge Hypothesis. Precipitation is reduced
by up to 70%, resulting in c. 80% savanna cover (modified from Haberle and
Maslin 1999)........................................................................................................14

xii
Figure 1. 8. Paintings from Caverna da Pedra Pintada (Monte Alegre), where the
earliest record of human occupation in Amazonia was found (Roosevelt et al.
1996). ...................................................................................................................20
Figure 2. 1. Map of the study area, showing the location of Lakes Marcio and Tapera
(studied lakes), as well as Lake Curiaú, the City of Macapá, and the Amazon
River.....................................................................................................................33
Figure 2. 2. Lithology of sediment cores from Lakes Marcio and Tapera, Amapá
(Brazil). Also showing location of dated samples (cal years BP). ......................38
Figure 2. 3. Radiocarbon ages from Lakes Marcio (diamonds) and Tapera (circles)
plotted versus depth (cm). The Lake Tapera radiocarbon age that is represented
by an open circle (11,750 years BP) was rejected as an outlier. The dashed lines
represent the extrapolated ages toward the present. ............................................41
Figure 2. 4. Pollen concentration diagram of Lake Marcio, Amapá (Brazil), showing
selected taxa (scale is x103 grains/cm3 of sediment), lithology, and radiocarbon
dates (cal years BP). The hollow curves are exaggerated 5 times.......................43
Figure 2. 5. Pollen percentage diagram of Lake Marcio, Amapá (Brazil), showing the
most representative taxa and area of charcoal particles (mm2/cm3 of sediment).
Hollow curves are exaggerated 5 times. The pollen record allowed the
recognition of 2 main zones, and 3 sub-zones.....................................................46
Figure 2. 6. Pollen concentration diagram of Lake Tapera, Amapá (Brazil), showing
selected taxa (scale is x100 grains/cm3 of sediment), lithology, and radiocarbon
dates (cal years BP). The hollow curves are exaggerated 5 times.......................48
Figure 2. 7. Pollen percentage diagram of Lake Tapera, Amapá (Brazil), showing the
most representative taxa and area of charcoal particles (mm2/cm3 of sediment).
Hollow curves are exaggerated 5 times. The pollen record allowed the
recognition of 2 main zones, and 2 sub-zones.....................................................51

xiii
Figure 2. 8. DCA scores from Lake Marcio. Top graph shows axis 1 vs. axis 2.
Samples are grouped according to zones interpreted from pollen diagrams.
Bottom graph displays resulting DCA scores of axis 1 plotted vs. depth (cm)
showing a strong polarization of samples into two main groups representing
zones M1(ABC) and M2......................................................................................53
Figure 2. 9. DCA scores from Lake Tapera showing axis 1 vs. axis 2. Samples are
grouped according to zones interpreted from pollen diagrams. Axis 1 polarizes
samples into two main groups that represent zones T1 (AB) and T2. Axis 2
polarizes bottom samples into two subgroups, allowing further division of zone
T1 into T1A and T1B...........................................................................................56
Figure 2. 10. DCA scores from Lake Tapera, axes 1 and 2 plotted vs. depth (cm), and
respective calibrated ages. Scores from axis 1 show a marked division into two
main zones. Scores from axis 2 show further subdivision of zone T1. ...............57
Figure 3. 1. Map of South America showing Prainha. The inset is a false color
Landsat TM image showing bands 7, 4, and 2. Pink represents highly reflective
surfaces such as savannas, urban areas and bare ground, whereas green
represents the more complex structure of forests ................................................70
Figure 3. 2. Lithology of sediment cores from Lakes Santa Maria, Geral (modified
from Bush et al. 2000), and Saracuri, Monte Alegre-Prainha region, Pará
(Brazil). Also showing location of dated samples in calibrated years Before
Present (BP). Ages inside ellipse were rejected. .................................................76
Figure 3. 3. Radiocarbon ages from Lakes Santa Maria (diamonds), Geral (squares),
and Saracuri (triangles). The two radiocarbon dates represented by an open
diamond and an open triangle were rejected as outliers. .....................................78
Figure 3. 4. Pollen concentration diagram of Lake Santa Maria, Pará (Brazil),
showing selected taxa (scale is x 103 grains/cm3 of sediment), lithology, and
radiocarbon dates (in calibrated years BP). The hollow curves are exaggerated 5
times.....................................................................................................................81

xiv
Figure 3. 5. Pollen percentage diagram of Lake Santa Maria, Pará (Brazil), showing
the most representative taxa and area of charcoal particles (mm2/cm3 of
sediment). Hollow curves are exaggerated 5 times. ............................................82
Figure 3. 6. Pollen percentage diagram of Lake Geral, Pará (Brazil), showing the
most representative taxa, lithology, and radiocarbon dates (calibrated years BP).
Hollow curves are exaggerated 5 times, and black dots identify presence of Zea
(modified from Bush et al. 2000).........................................................................87
Figure 3. 7. Pollen concentration diagram of Lake Saracuri, Pará (Brazil), showing
selected taxa (scale is x 103 grains/cm3 of sediment), lithology, and radiocarbon
dates (calibrated years BP). The hollow curves are exaggerated 5 times............89
Figure 3. 8. Pollen percentage diagram of Lake Saracuri, Pará (Brazil), showing the
most representative taxa and area of charcoal particles (mm2/cm3 of sediment).
Hollow curves are exaggerated 5 times)..............................................................90
Figure 3. 9. DCA scores from Lake Santa Maria showing axis 1 vs. axis 2. Samples
are grouped according to interpreted phases. Axis 1 polarizes samples into 3
groups that represent phases SM1, SM3 and SM4, phase SM2, and phase SM5.
Axis 2 separates phases SM1 from SM3 and SM4..............................................92
Figure 3. 10. DCA scores from Lake Santa Maria, axes 1 and 2 plotted vs. depth
(cm). Also showing dates in cal years BP, the 5 most important species
associated with each axis of DCA, and interpreted zones. ..................................93
Figure 3. 11. DCA scores from Lake Geral showing axis 1 vs. axis 2. Samples are
grouped according to phases interpreted. Axis 1 polarizes samples into 2 main
groups that represent phases G1 and G2 (bottom of the core), and G3 and G4
(top of the core). Axis 2 polarizes bottom samples into 2 groups that represent
phases G1 and G2. Samples G-410 and G-390 appear to be outliers. Distinction
between phases G3 and G4 is nearly impossible.................................................94
Figure 3. 12. DCA scores from Lake Geral, axis 1 vs. depth (cm). Also showing
dates in cal years BP, the 5 most important species associated with each end of
the axis of DCA, and interpreted zones. ..............................................................95

xv
Figure 3. 13. DCA scores from Lake Saracuri showing axis 1 vs. axis 2. Samples are
grouped according to phases interpreted. Axis 1 displays a sample distribution
that is relatively coincident with phase depths. Bottom samples (S1) yielded high
scores (right), top samples (S4) yielded low scores (left), and intermediate
samples were placed in the middle. Axis 2 polarizes samples from phases S2
(bottom) and S3 (top)...........................................................................................97
Figure 3. 14. DCA scores from Lake Saracuri, axes 1 and 2 vs. depth (cm). Also
showing dates in cal years BP, the 5 most important species associated with each
axis of DCA, and interpreted zones.....................................................................98
Figure 3. 15. Phases of environmental changes interpreted from the paleoecological
records of Lakes Santa Maria, Geral, and Saracuri. Also showing chronology
and pollen zones of each record. Human impacts are inferred from high fire
frequency and agriculture. .................................................................................100
Figure 4. 1. Map of South America showing Roraima. The inset is a false color
Landsat TM image showing bands 7, 4, and 2. Pink represents highly reflective
surfaces such as savannas, urban areas and bare ground, whereas green
represents the more complex structure of forests. .............................................111
Figure 4. 2. Radiocarbon ages from Lake Jacaré plotted with depths (cm). Dashed
line represents extrapolated ages. ......................................................................117
Figure 4. 3. Lithology of sediment core from Lake Jacaré, Roraima (Brazil). Also
showing location of dated samples in calibrated years BP................................118
Figure 4. 4. Diagram showing mineral fractions, lithology, and radiocarbon dates (cal
years BP)............................................................................................................119
Figure 4. 5. Pollen concentration diagram of Lake Jacaré, Roraima (Brazil), showing
selected taxa (scale is x 100 grains/cm3 of sediment), lithology, radiocarbon
dates (cal years BP), and charcoal particles (mm2/cm3). The hollow curves are
exaggerated 5 times. ..........................................................................................121

xvi
Figure 4. 6. Pollen percentage diagram of Lake Jacaré, Roraima (Brazil), showing
the most representative taxa, lithology, and radiocarbon dates (cal years BP).
The hollow curves are exaggerated 5 times.......................................................122
Figure 4. 7. DCA scores from Lake Jacaré showing Axis 1 vs. Axis 2. Samples are
grouped according to interpreted phases. ..........................................................124
Figure 4. 8. DCA scores from Lake Jacaré, Roraima (Brazil). Axis 1 plotted with
depth (cm). Also showing dates (cal years BP), the interpreted pollen zones, and
the five most characteristic taxa of each axis extreme.......................................125
Figure 4. 9. Estimated sedimentation rates through rudimentary linear interpolation
reveal a sedimentary discontinuity at 30 cm. The duration of the hiatus is
estimated to be of c. 3050 years, assuming constant sedimentation rates. ........132
Figure 5. 1. Environmental changes interpreted from the paleoecological records of
Lakes Santa Maria, Geral, and Saracuri (ca. 1°S), Lakes Marcio and Tapera (ca.
0°N), and Lake Jacaré (ca. 3°N). Also showing chronology (cal years BP) and
human impacts, as suggested by large charcoal deposits and agriculture
indicators (e.g. maize pollen grains)..................................................................137
Figure 5. 2. Mean insolation (W/m2) during the months of June and December at 0°
latitude for the last 20,000 years (Laskar 1990). Between 7000 and 6000 years
BP the difference between June and December mean insolation was the smallest
within the Holocene...........................................................................................140
Figure 5. 3. Schematic map of South America showing the study sites and the
present mean positions of the ITCZ (white clouds) during the Austral summer
(December-February) on the left, and during the Austral winter (June-August) on
the right. The pink clouds over the continent represent increased convective
activity. ..............................................................................................................142
Figure 5. 4. Variation in event frequency of ENSO during the Holocene (modified
from Moy et al. 2002)........................................................................................146

xvii
Figure 5. 5. Schematic representation of stratigraphic profiles of Lakes Tapera and
Marcio. Blue represents the water column, black represents the layer of gyttja,
and gray represents the basal clays. Notice that Lake Marcio’s basin is at least
185 cm deeper than that of Lake Tapera............................................................152

1
CHAPTER I
THE AMAZONIAN REGION
INTRODUCTION
The Amazonian region accommodates the largest rainforest ecosystem on
Earth, representing approximately 50% of all such forests. Despite playing a
significant role in the global carbon cycle, biodiversity and climate, the long-term
dynamics of this system remain poorly understood. Tropical savannas cover a total
of ca. 24 x 106 km2 in Africa, Asia, and South America. In South America, the
largest savanna area is located in central Brazil, where it is known as cerrado (Silva
and Bates 2002). Despite its high biodiversity, only recently have conservation
efforts focused on cerrado vegetation (Myers and others 2000). The origins of that
diversity remain unclear, although the importance of ecotonal areas in generating and
maintaining diversity has been suggested (Smith and others 1997). Ecotonal
migration due to climate change has also been suggested as a mechanism to induce
local isolation and speciation (Bush 1994).
STUDY GOALS
The overall goal of this study is to investigate how climate and human
activity have influenced ecotonal areas of savanna enclaves within Amazonia. This is
the first paleoecological study based on fossil pollen and charcoal analyses

2
specifically designed to reveal the history of Amazonian savanna-forest ecotones in
Brazil during the Holocene. The paleoecological data gathered from the studied sites
are representative of the northeastern region of Brazilian Amazonia.
VEGETATION
To help evaluate the impacts of past climate changes and pre-Columbian
human populations on Amazonian ecosystems, a brief outline of major vegetation
types is provided (Fig 1.1). For more detailed descriptions see Brown and Prance
(1987), Salgado-Labouriau (1997), and Oliveira-Filho and Ratter (2002).
RAINFORESTS
Amazonia contains the largest rainforest system in the world. These forests
can be further divided into 3 main categories: terra firme forest (never flooded),
várzea forest and igapó forest (both subject to flooding). Terra firme forests alone
cover ca. 51% of the Amazonian region (Brown and Prance 1987), and are
characterized by tall trees (25-35 m) that provide a closed canopy. Species diversity
in these forests is extremely high, with more than 200 tree species per hectare. In
western Amazonia tree species diversity is even higher, reaching ca. 700 species/ha
in Ecuadorian and Peruvian forests (Pitman and others 2001).

3
Figure 1. 1. Present vegetation distribution of northern and central Brazil. Cerrado
and Caatinga are equivalents of savanna vegetation. White dots represent
study sites (modified from IBGE 1992).
The inundated forests occupy areas subject to flooding (periodical and
permanent), and are known as várzea or igapó depending on what kind of river
floods the area. The igapó forests occur in areas flooded by rivers that carry very
little Andean sediments (loaded with nutrients) but are rich in humic acids. Because
of the humic acids coloring these waters of dark brown, these are known as black-
water rivers. The trees are shorter than the ones from terra firme and várzea forests,
and the species diversity is also lower than these other forest types. The rivers that
flood the areas occupied by várzea forests are know as white-water rivers because of
their sediment load (eroded from the Andes). The trees are as tall as the trees from

4
terra firme forests, but there are more lianas in várzea forests. The species diversity
is higher than that of igapó forests but still lower than that of terra firme forests.
Even though both igapó and várzea forests are subject to similar flooding regimes,
their species composition is different, as several tree species from igapó forests
present scleromorphic adaptations (Brown and Prance 1987).
GALLERY FORESTS
Gallery forests, also known as riverine forests, are usually present along the
watercourses within major savanna areas such as the Colombian llanos, Roraima-
Rupununi, and cerrado in central Brazil (Brown and Prance 1987, Salgado-
Labouriau 1997, Behling and Hooghiemstra 1999, Behling and Hooghiemstra 2000,
Berrio et al. 2000, Oliveira-Filho and Ratter 2002). Gallery forests are present within
savanna and cerrado areas (dry vegetation types) because of local soil moisture
availability throughout the year (Brown and Prance 1987, Oliveira-Filho and Ratter
2002).
SAVANNAS
Despite having the second largest vegetation cover in South America, the
savannas are still poorly understood. In central Brazil alone (the core area), savannas
cover an area of ca. 2 million km2, although disjunct areas also occur scattered
throughout Amazonia.

5
While Amazonian savannas and Central Brazilian cerrados have some
floristic and physiognomic similarities they can still be distinguished from each
other. Cerrado is a woody type of savanna, varying from nearly treeless grasslands to
a woodland of semideciduous trees with a herbaceous ground cover (Furley 1999).
Cerado can be further subdivided into 3 main types: cerradão, with a high density of
tree cover (5-15 m tall) with a herbaceous component; campo cerrado with a low
density of tree cover (2-5 m tall) evenly scattered; and campo limpo, a grassland with
very few trees (Brown and Prance 1987). The most characteristic tree species are
Byrsonima verbascifolia and Curatella americana, but other taxa can also be present
at relatively high abundances, e.g. Bowdichia virgilioides, Caryocar brasiliensis,
Dimorphandra mollis, Hancornia speciosa, Qualea grandiflora, and
Stryphnodendron barbatima (Brown and Prance 1987, Furley 1999). Precipitation in
the cerrado region (central Brazil) ranges from 1500 to 2000 mm per year with a
strong seasonality creating an uneven rainfall distribution (Brown and Prance 1987)
(Fig 1.6). Cerrados occur above 500 m a.s.l. (e.g. highlands of central Brazil), but are
also found at lower altitudes in some areas (e.g. in Mato Grosso), where they flank
the southern boundary of the Amazonian forest and the northwestern boundary of the
Atlantic forest (Fig 1.7).
The Amazonian savannas are present mostly in regions with low precipitation
(under 2000 mm per year) and strong seasonality (Fig 1.6). Even though Amazonian
savannas have some species in common with cerrado, especially campo cerrado (e.g.
Byrsonima verbascifolia, Curatella americana, Hancornia speciosa, Palicourea

6
rigida, Qualea grandiflora, and Salvertia convalariaeodora) they occur at lower
altitudes and have a much lower species diversity and endemism than cerrados
(Brown and Prance 1987, Salgado-Labouriau 1997). Haffer (1969) used the present
distribution of a few plant taxa characteristic of cerrados to support a more
continuous geographical distribution of savanna-like vegetation in the past.
Nevertheless, an increasing number of studies also suggested that a forest corridor
connected the Amazonian and the Atlantic forests in the past (Costa 2003). Whether
this connection was a continuous forest corridor or a series of forest patches is still
unclear.
WHITE-SAND SAVANNAS
Patches of sandy soils are scattered throughout Amazonia, but especially in
the Rio Negro region, in Roraima. These areas are occupied by a very distinct
vegetation type ranging in structure from open savanna (mostly grasses) to woodland
(Cooper 1979, Anderson 1981). The vegetation on white-sand soils is differentiated
from the regular forests and savannas by its high species endemism and low species
diversity. Additionally, a large number of species present scleromorphic adaptations
(Brown and Prance 1987). The white-sand savannas are sometimes referred to as
campinas, which would be the equivalent to woodland savannas. Some of the white-
sand savannas, especially the ones near Rio Negro (Fig 1.5), were suggested to have
been formed as a result of intense land-use (Prance and Schubart 1977). The human
impacts were so severe that the soil became depleted of nutrients. As a consequence,

7
only savanna vegetation was able to colonize the areas, and the rainforest could not
grow back even after the areas were abandoned.
AMAZONIAN SAVANNAS
ENVIRONMENTAL REQUIREMENTS
A simple climatic characterization of savanna versus forest habitat is not
possible as both can occur under the same climatic regime. As the present climate
regime (wet conditions) would favor forest dominance over savanna, Oliveira-Filho
and Ratter (2002), and Pennington et al. (2000) suggest that neither soil fertility nor
precipitation patterns (Fig 1.2) alone can explain modern savanna distribution (Fig
1.1). Therefore, the missing variable determining the presence of savanna could well
be disturbance by fire. Several taxa from savannas and especially from cerrados in
central Brazil are fire-adapted, being not only fire-tolerant but also fire-dependent
(Miranda and others 2002).

8
Figure 1. 2. Climate and precipitation patterns of northern Brazil. Red dots represent
study sites (modified from IBGE 1990).
SAVANNA DISTRIBUTION
Even though the core area of savannas is in central Brazil, where they are
known as cerrado, some of the disjunct savannas in the Brazilian Amazonia are
located in Amapá (Fig 1.3), Marajó Island (Fig 1.4), Alter do Chão (Fig 1.5), and
Roraima (Fig 1.6) (Salgado-Labouriau 1997). However, the largest continuous area
of natural savannas within the Brazilian Amazonia is located in Roraima (Miranda
and Absy 1997), where they occupy an area of ca. 40,000 km2 (ca. 16% of the State)
(Silva 1997).

9
Figure 1. 3. False color Landsat TM image showing bands 7, 4, and 2. Pink
represents highly reflective surfaces such as savannas, urban areas and
bare ground, whereas green represents the more complex structure of
forests. Disjunct savanna in Amapá (red) surrounded by tropical
rainforest (green) at one of the studied sites.
Figure 1. 4. False color Landsat TM image showing bands 7, 4, and 2. Pink
represents highly reflective surfaces such as savannas, urban areas and
bare ground, whereas green represents the more complex structure of
forests. Disjunct savanna (red) on eastern Marajó Island, and tropical
rainforest (green) on western Marajó Island, at the mouth of the Amazon
River.

10
Figure 1. 5. False color Landsat TM image showing bands 7, 4, and 2. Pink
represents highly reflective surfaces such as savannas, urban areas and
bare ground, whereas green represents the more complex structure of
forests. Disjunct savannas (red) around Alter do Chão, Santarém, and
Monte Alegre surrounded by tropical rainforest (green), and one of the
studied sites.
Figure 1. 6. False color Landsat TM image showing bands 7, 4, and 2. Pink
represents highly reflective surfaces such as savannas, urban areas and
bare ground, whereas green represents the more complex structure of
forests. Disjunct savannas (red) of northern Roraima at one of the studies
sites and at lower Rio Negro surrounded by tropical rainforest (green).

11
SAVANNA ORIGINS
Even though there are more hypotheses about the formation and prior extent
of Amazonian savannas than available data, the ideas regarding the origins of
Amazonian savannas can be basically grouped around two major hypotheses:
climatic and anthropogenic origins.
CLIMATIC ORIGINS. The supporters of the climatic origin hypothesis can
be further subdivided into two opposing groups. The first group, led by supporters of
the Refugia Hypothesis (Haffer 1969), propose that savannas expanded and replaced
Amazonian rainforests during the last Ice Age. With the end of the Ice Age
(interglacial period), savannas were replaced by forests and became isolated in
central Brazil and in a few relatively small scattered areas (e.g. Figs 1.3, 1.4, 1.5, and
1.6). A further savanna expansion within Amazonia may have occurred in the middle
of the Holocene due to a dry episode (Absy et al. 1991, Ledru 1993). Whether or not
there was a mid-Holocene dry period and to what extent it affected the vegetation in
Amazonia is still unresolved (Bush and Colinvaux 1988, Absy et al. 1991, Salgado-
Labouriau 1997, Behling et al. 1999). If these hypotheses are true, the disjunction of
savannas scattered across Amazonia region would be relatively young, i.e. < 30,000
years. Because the Refugia Hypothesis, as formerly proposed, had a tremendous
impact on paleoecological studies in Amazonia, it will be discussed in more detail in
another section.
The second group, led mainly by Colinvaux and Bush, defend a much older
age for isolation of Amazonian savannas. Based on well-dated paleoecological

12
records, this group proposes that savannas may have had insignificant expansions
during drier climates but never replaced forests, at least in the last 170,000 years
(Bush et al. 2002, Colinvaux et al. 1996, Colinvaux et al. 2000, and Haberle and
Maslin 1999). Therefore, Amazonian savannas would have independent origins and,
if climatically derived, could be up to 8 million years old.
ANTHROPOGENIC ORIGINS. This hypothesis was proposed based on
observations of fire-adapted species from savannas and cerrados (in central Brazil),
which suggest that savannas are the result of extensive human impacts (mostly
cutting and burning) on dry forests (Eiten 1972, Prance and Schubart 1977).
Consequently, the Amazonian savannas would be fairly young (late Pleistocene), as
the oldest record of human occupation in Amazonia dated between 14,200 and
10,500 cal years BP (Roosevelt et al. 1996).
USING POACEAE AS A SAVANNA INDICATOR
The majority of claims of Amazonian aridity are based on a relative increase
of Poaceae (grasses) pollen and a few other taxa. Despite Poaceae pollen being easily
identifiable to family level, it is extremely difficult to distinguish them to species or
even genus level under light microscopy.

13
The importance of Poaceae in savanna ecosystems is clear, as 50-90% of the
pollen rain in the cerrados is made up of Poaceae (Salgado-Labouriau 1979).
However these pollen types can also be found in almost every habitat, and
interpretation of the Poaceae pollen signature must be conducted with caution (Bush
2002).
An increase in Poaceae pollen percentages could reflect transition from forest
to a grassland landscape. However, it could also reflect hydrologic changes. Poaceae
can colonize exposed lake shores during lower lake stands, and their representation
in the sediments can increase even though there was no regional vegetation change.
An example of such an interpretation is given by Behling et al. (2001), who
correlated Poaceae pollen abundance in the sediments of Lake Calado (central
Amazon) with the area of exposed lands suitable for colonization during low stand
events.
The risk of error in interpretations based on Poaceae pollen is magnified
when using only percentage data (Ferraz-Vincentini and Salgado-Labouriau 1996).
Because palynological analyses are based on a fixed sample size of 300-500 pollen
grains, when the number of grains of a certain taxon (e.g. Poaceae) increases in the
record, it will cause contribution of other pollen taxa to diminish, or even disappear.
In other words, because Poaceae is producing a greater amount of pollen grains, all
the other taxa will show a decrease in their percentages even if their pollen
production also increases, but by a lesser amount. The best approach is to use both
pollen percentage and concentration data (grains deposited per cm3 of sediment).

14
However, for even more reliable interpretations of fossil pollen records, the pollen
influx data (grains deposited per cm2 of sediment per year) should always be used if
a robust chronology is available.
AMAZONIAN ARIDITY - THE REFUGIA HYPOTHESIS
The extremely high biodiversity found in the Amazonia (especially western
Amazonia) has always intrigued biologists. Trying to explain the geographic
distribution of endemic species, Haffer (1969) suggested that during the last
glaciation the rainforest was fragmented, being almost entirely replaced by savanna
vegetation because of existing arid conditions (Fig 1.7).
Figure 1. 7. Sketch map of vegetation distribution in Amazonia during the Last
Glacial Maximum according to the Refuge Hypothesis. Precipitation is
reduced by up to 70%, resulting in ca. 80% savanna cover (modified
from Haberle and Maslin 1999).

15
The original hypothesis stated that temperature in Amazonia remained
constant during the last glacial time; however precipitation was reduced by up to
70%. With the reduced rainfall, Amazonia became arid, savannas covered most of
the Amazon basin (ca. 80%), and the rainforest was confined to hills, high enough to
concentrate orographic rains. These areas of local forest persistence were named
“refuges”. Haffer (1969) argued that such geographic isolation promoted allopatric
speciation. With the end of the ice age and beginning of the Holocene, precipitation
increased to modern levels, and forests replaced the savanna vegetation over most of
the Amazon basin. It is important to note that this hypothesis was solely based on the
modern distribution of birds. Subsequently, many authors following Haffer (e.g.
Vanzolini 1970, Prance 1973, and Brown Jr. 1977) supported his theory with
biogeographic data for lizards, plants, and butterflies.
Because this was such an elegant hypothesis, several paleoecological studies
were interpreted to corroborate Amazonian aridity. The main supporters of this
hypothesis in the field of Paleoecology were Absy and Van der Hammen (Absy et al.
1991, van der Hammen and Absy 1994). They have interpreted increased
percentages in Poaceae pollen as an indication of savannas replacing forests in the
Serra dos Carajás and Rondônia (respectively) during glacial times. Pollen data from
Guyana and Surinam show mangroves being replaced by dry savannas (van der
Hammen 1963, Wijmstra and van der Hammen 1966). Stone lines and dune
formations that are supposedly unlikely to exist in the modern moist conditions were
found in the middle of the rainforest (Servant and others 1981). These geological

16
data were also attributed to Last Glacial Maximum (LGM) aridity, even without
being independently dated.
The Refuge Hypothesis has also been criticized. According to Nelson et al.
(1990), maps of endemic plant distributions used to support the existence of refuges
during the Pleistocene are coincident with maps of botanical collection density.
These data gathered in several herbaria indicate that supposed centers of plant
endemism are just a sampling artefact, weakening the botanical argument for
Pleistocene aridity.
Regional paleoecological data have also been used to refute the Refuge
Hypothesis. The Amazon River carries sediments throughout an immense
geographic area (Andes and Amazon basin). These sediments are deposited offshore
in the Amazon fan, and provide a regional record of climate change. Even though
results of δ18O analysis of fossil planktonic foraminifera were interpreted to indicate
that the Amazon basin was drier than the present during the Younger Dryas (ca.
12,000-13,000 years BP), and the Amazon River discharge was reduced by at least
40% (Maslin and Burns 2000), fossil pollen data are not in accord with these
findings. Pollen analyses conducted on sediments from the Amazon fan by Haberle
(1997) and Haberle and Maslin (1999) have shown that vegetation changes related to
the presence of cold-adapted taxa (e.g. Alnus, Hedyosmum, and Podocarpus) were
more significant than any variation in savanna-forest expansions observed during
glacial times. Their results are supported by analyses of organic matter composition

17
of sediments from the Amazon fan (Kastner and Goñi 2003). The authors found no
evidence of savannas replacing forests during the LGM.
The pollen data from Lake Pata (Colinvaux et al. 1996, Colinvaux et al.
2000) seem to agree with the Amazon fan record. In no period of time (in the last
50,000 years) does the record suggest the expansion of open vegetation (e.g.
savannas) at the expense of rain forest over a regional area. Paleolimnological data
from Lakes Pata and Verde suggest a wet Last Glacial Maximum and the driest
period of the last 170,000 years between 35,000 and 27,000 years BP (Bush and
others 2002). Even though Milankovitch forcing was thought to have influenced the
phases of low lake levels, no evidence of a biome replacement was found, as
arboreal pollen abundances were about 80% of the pollen sums throughout the entire
record. Other records with findings contrary to the Refuge Hypothesis exist for other
parts of Brazil and South America (e.g. De Oliveira 1992).
In conclusion, there are just a few paleoecological records spanning the last
glaciation in the Amazonian lowlands, and none forms a continuous record, as gaps
in sedimentation punctuate them all (Ledru and others 1998a). It is evident that more
data points are needed in order to have a reliable reconstruction of the climate and
vegetation history of such a large region.
The number of paleoecological studies falsifying the Refuge Hypothesis
continues to grow. Geological studies formerly used as evidence supporting the
Refuge Hypothesis (e.g. stone-lines and paleodunes) have been re-interpreted and
show no relationship with glacial aridity (Colinvaux et al. 2001). Nevertheless, the

18
latest reviews of the Refugia Hypothesis indicate that the debate has not yet reached
an agreement (Colinvaux et al. 2000, van der Hammen and Hooghiemstra 2000, and
Colinvaux et al. 2001). However, in its latest manifestation the Refugialists
suggested that the forest fragmentation that led to allopatric speciation took place at
anytime during the Cenozoic Era (from 65 million years ago to the present) (Haffer
and Prance 2001), making it practically impossible to test.
THE ARCHAEOLOGICAL RECORD
Holocene climatic changes have been proposed to explain variations in the
savanna coverage. However, human activities are also known to have altered
Amazonian landscapes. Archaeological studies have provided strong evidence that
human occupation of the Amazonian lowlands is more ancient than previously
thought. The first humans arrived in South America probably between ca. 20,000 and
14,000 years BP (Meggers 1987), and despite the romanticized early views of
Amazonia being pristine and untouched, humans have lived in this area for more
than 11,000 years (Roosevelt et al. 1991, Roosevelt et al. 1996, Athens and Ward
1999, Roosevelt 2000).
It was initially postulated that rainforest ecosystems could not provide
enough resources to support early hunters and gatherers (Bailey and others 1989).
Therefore, humans could not survive in the rainforest before the development of
slash and burn cultivation, and were environmentally determined to be culturally
backwards (Meggers 1971). An ecological analysis conducted by Colinvaux and

19
Bush (1991) showed that productivity in the diverse Amazonian rain forests is high
enough to support hunting and gathering for early humans.
Rather than having an Andean origin, the evidence for agricultural invention
has, so far, been found in the lowlands. In coastal Ecuador, phytoliths of
domesticated Cucurbita (squash) were deposited in a midden between 11,000 and
10,000 years BP (Piperno and Stothert 2003). Additional evidence for early
agriculture in the wet lowland tropics comes from lowland Panama (Piperno and
others 2000), where manioc starch grains indicate domestication and dispersal by
prehistoric societies around 7000-6000 cal years BP. which shows an early transition
from foraging to agriculture in Neotropical forests.
Furthermore, the environmental determinism was challenged also by
archaeological data from the Monte Alegre region, at the mouth of the Tapajós River
in eastern Brazilian Amazonia (Roosevelt et al. 1991, Roosevelt et al. 1996,
Roosevelt 2000). Pottery remains found in the Village of Taperinha are ca. 3000
years older than the artifacts found in the Andes, which indicates that pottery did not
spread from the Andes to the lowlands. Archaeological remains excavated from
“Caverna da Pedra Pintada”, a painted sandstone cave near Monte Alegre (Fig 1.8),
suggest a long-term occupation by foragers since the Late Pleistocene in a terra firme
forest ecosystem (Roosevelt et al. 1996). Additionally, that this Amazonian
paleoindian tradition is contemporary with, but distinct from, the Clovis tradition in
North America implies that the specialized big-game hunters from North America
were not the only source of migration into South America (Roosevelt et al. 1996).

20
Figure 1. 8. Paintings from Caverna da Pedra Pintada (Monte Alegre), where the
earliest record of human occupation in Amazonia was found (Roosevelt
and others 1996).
Further evidence of human activities in this region is provided by the
paleoecological records of 2 lakes located 50 km away from Monte Alegre, near the
village of Prainha (Bush and others 2000). Human disturbance is indicated by
continuous charcoal accumulation since 6600 cal years BP and fossil pollen evidence
of maize cultivation since ca. 4300 cal years BP (Bush and others 2000). Near the
top of the sequence, there is a decrease in the frequency of disturbance indicators
(e.g. phosphate influx, Cecropia pollen, and charcoal) and an increase of forest taxa
that indicate reduced human activity. Such changes could well reflect site
abandonment coincident with the arrival of Europeans or their transmitted diseases.
Additional evidence of human activity as early as 10,700 cal years BP is
suggested based on charcoal deposited in sediments from a lake near the mouth of

21
the Amazon River (Behling 1996). As natural fires would not be consistent with the
wet climate and vegetation (rainforest) of the study site, the charcoal particles were
thought to be anthropogenic in origin. A single pollen grain of Manihot (manioc)
was also found in the first peak of charcoal, but its origin was not identified (Behling
1996).
Later versions of environmental determinism (neo-determinism),
emphasizing the impact of climate change on cultural development, i.e. the collapse
of the Tiwanaku civilization in the Andes (Binford et al. 1997), have also been
challenged by Erickson (1999).
Archaeological data from savannas in the Bolivian Amazonia suggest that
these neotropical landscapes have been intensively used and modified by pre-
Columbian human populations. These peoples built elevated roads that cut through
the savannas in Baures (Bolivia) connecting forest islands within the savanna
(Erickson 2000, Erickson 2001). By ca. 1490 cal years AD, hydraulic networks of
constructed canals and ponds with fish weirs allowed dense, large populations to be
sustained in a savanna environment (Erickson 2000, Erickson 2001). Similar patterns
of large scale landscape modification by pre-Columbian human populations are
found in the upper Xingu in Brazil (Heckenberger and others 2003).
Even though several records indicate that human disturbance in Amazonia
was intense and widespread, that may not be true for all regions in the lowlands.
Several cores collected in the Lowlands of Ecuador suggest that human impacts can
be a very local phenomenon. Athens and Ward (1999) and Weng et al. (2002)

22
analyzed cores Maxus 5, and Maxus 1 and 4 (respectively). While Maxus 5 records
human impact during the Holocene based on the charcoal record, Maxus 4 shows no
sign of disturbance, despite both cores being collected from sites only 25 km apart.
Another paleoecological record from western Amazonia provides further evidence
that human occupation could be very local in some regions. Similarly, Lakes Gentry
and Werth are just 20 km apart in southeastern Peru, but have very different
histories. Charcoal particles were continuously recorded in the sediments of Lake
Gentry, while insignificant quantities of charcoal were found in the sediments of
Lake Werth, suggesting distinct and very local land use histories (Listopad 2001).
CHARCOAL IN THE SEDIMENTS
Associated either with human activity or climatic change, fires can be an
important element in shaping the environment and vegetation of a region (Cochrane
and Schulze 1999), and in determining the community structure of some ecosystems
(e.g. cerrados in central Brazil). As fire frequency is at least partially climatically
controlled, charcoal analyses of sediments that quantify past fire frequency may
provide a proxy for past climate changes (Patterson et al. 1987, Clark 1988, Behling
1996, Clark et al. 1996, Kennedy and Horn 1997, Bush et al. 2000). However, as
humans have used fire to modify landscapes throughout human history, charcoal can
also be an indicator of anthropogenic activities.

23
Because the size of charcoal particles may indicate the distance from the fires
to the study site, charcoal particles in lake sediments can provide information about
fire frequency on a small or large geographic scale (local or regional fires
respectively). Small particles can be transported further than larger particles, and
therefore represent regional fires (large geographic scale). Whereas large particles of
charcoal reflect fires that burned close to the site.
According to Clark et al. (1996), larger quantities of charcoal may not
necessarily indicate a higher fire frequency as high charcoal accumulation rates are
sometimes just the result of bulk sediment accumulation, and may have nothing to do
with higher charcoal production or higher fire frequency. Independent measures of
sediment and charcoal accumulation are the best approach to reliable interpretations
of paleoecological records.
HOLOCENE HISTORY OF AMAZONIAN SAVANNA-FOREST ECOTONES
Fossil pollen data from 8 savannas, seven located in Brazil and one in
Venezuela, were reviewed and published by Salgado-Labouriau (1997). The records
that extend further than 10,000 years BP (Carajás, Águas Emendadas, Cromínia, and
Serra Negra) show the driest period being recorded between 16,700 and 12,490 cal
years BP. Precipitation increased between 7800 and 6800 cal years BP and marshes
and swamps began to form in central Brazil. However, arboreal pollen increased at
4470-3170 cal years BP suggesting that more dense vegetation was growing in the

24
savannas. Even though cerrados are known to burn naturally, the presence of
charcoal in the Holocene sequences could be also anthropogenic in origin, which
would be in agreement with records of human occupation from central Brazil (Kipnis
1998).
A further comparison of paleoecological data collected from several other
savanna sites was carried out by Behling and Hooghiemstra (2001). Sites located in
the Llanos Orientales of Colombia show progressively humid conditions from the
late-Pleistocene to the early- and mid-Holocene. Increased human impact after ca.
4190 cal years BP was inferred from higher pollen percentages of the palm Mauritia
and/or Mauritiella as these trees have edible fruits and palm fronds widely used for
thatch. An expansion of forests was recorded after ca. 3900 cal years BP. In the last
2320 years representation of forest taxa decreased while Mauritia pollen increased in
abundance, suggesting human impacts in this region (Behling and Hooghiemstra
2000). A pollen profile from Laguna Carimagua (Colombia) suggested an expansion
of forest vegetation after ca. 4960 cal years BP (Behling and Hooghiemstra 1999).
Another pollen record from the Carimagua area showed no significant change
between forest and savanna proportions. To reconcile this data set with others from
the Colombian savannas, it was suggested that the pollen signature of savanna
vegetation was masked by a local gallery forest in the last 1300 years (Berrio and
others 2000).
Data from Rupununi Savanna of Suriname show that woodland and grassland
savannas constantly alternated from ca. 11,420 until ca. 5730 cal years BP, but after

25
this period, especially during the last 3100 years, grassland savanna dominated
(Behling and Hooghiemstra (2001).
For the Roraima-Rupununi Savanna region of Northern Brazil a few records
are available. Simões Filho et al. (1997) interpreted sedimentary changes in the Lake
Caracaranã (northern Roraima) core as dry events of short-duration between 10,200
and 8170 cal years BP. Even though the oldest fossil pollen record comes from Lake
Fazenda São Joaquim (Absy and others 1997), it spans only the late Holocene (last
4000 years), and no significant vegetation changes were recorded. Desjardins et al.
(1997) found pieces of charcoal derived from forest taxa in the bottom of soil
profiles within modern savannas near Boa Vista. The ages yielded by the charcoal
pieces (8460 and 7470 cal years BP) provide an approximate age for the on set of
fires and savanna expansion near Boa Vista (Desjardins and others 1997).
In the coastal savannas of Guyana, Suriname, and French Guyana (northern
hemisphere), records indicate that swamp savannas, mainly represented by Poaceae
and Cyperaceae, have been present in these regions for the last 5700 years. A pollen
profile from Marajó Island, in the mouth of the Amazon River (Pará State), indicates
a change from a more closed to open swamp savanna at ca. 7460 cal years BP (Absy
1985). The pollen record from Lagoa da Curuça in Pará State shows peaks in the
Poaceae abundance that coincide with charcoal peaks during the Holocene (Behling
1996).
The pollen data from Carajás (Absy and others 1991) show arboreal and
herbaceous taxa intercalating during the last 60,000 years, suggesting shifts between

26
forest and edaphic savanna vegetation, being interpreted as dry and wet periods.
Markedly dry periods associated with savanna expansion in the Holocene are
recorded between 8300 and 3100 cal years BP.
Paleoecological data from Noel Kempff Mercado National Park in Bolivia
show that humid evergreen forest expanded in the last 3000 years (Mayle and others
2000). The ecotone migrated ca. 100 km southward over the savanna vegetation.
In conclusion, several records show a savanna expansion from the early
Holocene until the middle Holocene (from ca. 11,000 until 5000 cal years BP),
which is followed by a forest re-growth and increased anthropogenic activities,
especially in Colombian sites, after ca. 4500 cal years BP.
DESIGN OF STUDY
The dynamics of Amazonian savanna-forest ecotones were studied by
collecting sediment cores from three sites in eastern Amazonia (Fig 1.1). All the sites
lie within the modern boundary of savanna-forest ecosystems. The first site is located
in Amapá State, a few kilometers north of Macapá City (the state capital). Two lakes
from the same hydrological system were chosen (Fig 1.2). The second site is in the
Prainha region, ca. 80 km away from Monte Alegre (Pará State). Two lakes were
chosen because they form a N-S transect of lakes having Lake Geral (Bush and
others 2000) in the middle (Fig 1.4). The third site is located in the northern portion

27
of Roraima State, 7 km east of Lake Caracaranã, and 8 km west of the Guyana (Fig
1.5).
Fossil pollen analyses were carried out in the sediment cores. The chosen
lakes have basins large enough to provide a regional pollen signal that allowed the
reconstruction of the past vegetation changes revealing the history of savanna-forest
dynamics.
The analysis of charcoal particles in the sediments was used as a proxy for
past fire frequency. As early human populations modified landscapes through fire,
charcoal analysis provide indication of human presence. Agricultural activities can
also be estimated through the presence of pollen of cultivated taxa (e.g. maize and
manioc).

28
TESTING HYPOTHESES
As the overall pattern indicated by several studies in the literature suggests a
savanna expansion from ca. 8000 until 5000 years BP, and a savanna contraction
with associated forest expansion in the last 4000 years, 3 main hypotheses were
tested. The first 2 hypotheses were tested in each of the data chapters (Chapters II,
III, and IV). The third hypothesis was tested only in the final chapter.
H1: Amazonian savannas expanded between 8000 and 5000 years BP.
H0: No savanna expansion was recorded between 8000 and 5000 years BP.
H2: Forests expanded and savanna vegetation contracted in the last 4000 years.
H0: No forest expansion was recorded in the last 4000 years.
H3: If vegetation changes are recorded, they will be synchronous between the study
sites.
H0: If vegetation changes are recorded, there will be no synchronism between the
study sites.

29
CHAPTER II
PALEOECOLOGICAL RECORD OF AMAPÁ (BRAZIL)
INTRODUCTION
This is the first paleoecological study based on fossil pollen and charcoal
analyses specifically designed to reveal the history of Amazonian forest-savanna
ecotonal regions in Brazil during the Holocene.
In South America, savanna vegetation occupies a discontinuous area.
Similarly, within Brazilian Amazonia, savannas occupy large areas of Amapá and
Roraima states, and are also found in Humaitá (Amazon State) and Alter do Chão
(Pará State) (Salgado-Labouriau 1997). Even though savannas may be found under
similar precipitation regimes to some Neotropical dry forests (1600 mm/year, with 5-
6 months generally receiving less than 1000 mm/year), they tend to occupy areas
with poorer soils (Pennington and others 2000). In dry forests, grasses are a minor
component, while in savannas they account for a much greater proportion.
The romanticized early view of Amazonia being pristine and untouched has
been revised in the last 20 years. Archeological studies provided evidence that
humans have lived in Amazonian lowlands for at least 11,000 years (Roosevelt et al.
1991, Roosevelt et al. 1996). Further evidence that these Neotropical landscapes
have been intensely used and modified by pre-Columbian human populations comes

30
from savannas in the Bolivian Amazonia and Upper Xingu, in Southern Brazilian
Amazonia (Erickson 2000, Erickson 2001, Heckenberger et al. 2003).
Although human activities are known to have altered Amazonian
environments, Holocene climatic changes have also been invoked to account for
variation in the range of savannas. A paleoecological study conducted on a mosaic
landscape on the forest-savanna ecotone in Bolivia revealed a forest expansion of ca.
100 km in the last 3000 years (Mayle and others 2000). Similarly, paleoecological
records from Colombian savannas also suggest a forest expansion in the last 4000
years (Behling and Hooghiemstra 1998, Behling and Hooghiemstra 1999, Berrio et
al. 2002).
To study the effects of climate and human occupation on the forest-savanna
ecotones in Eastern Amazonia, paleorecords from two lakes (Marcio and Tapera)
that lie in the modern savanna within 5 km of forest were analyzed. Lakes Marcio
and Tapera have basins large enough to provide a regional pollen signal and should
provide records sensitive to ecotonal shifts.
Associated either with human activity or climatic change, fire can be an
important element reshaping the environment and vegetation of a region (Cochrane
and Schulze 1999), and in determining the community structure of some ecosystems
(e.g. cerrados in central Brazil). Charcoal analyses of sediments can be used to assess
past fire frequency in the environment (Patterson et al. 1987, Clark 1988, Behling
1996, Clark et al. 1996, Kennedy and Horn 1997, Bush et al. 2000).

31
STUDY SITE
Lakes Marcio and Tapera (ca. 0°07’40.9” N 51°04’47.8” W) are only 7.5 km
from each other, and ca. 10 km from the Amazon River. Both lakes are shallow (1.5-
2.5 m deep) and lie at less than 10 m above sea level (Fig 2.1). Lake Marcio is part
of a much larger hydrologic system known as Curiaú. Lake Curiaú is an L-shaped
basin measuring ca. 6.5 km long on its longest axis, 3.5 km on the shortest axis, and
is 1.2 km wide. The watershed of the lake is estimated to be at least 150 km2. Lake
Marcio occupies an area of about 300x350 m on the northern tip of Lake Curiaú, and
although Lake Curiaú may dry out almost completely in the dry season, the sub-
basin of Lake Marcio is said by local inhabitants to be a permanent water body. Lake
Tapera is 1 km long and about 700 m wide, and drains into Lake Curiaú. As Tapera
lies upstream of Lake Curiaú, and flows into the larger lake, it must lie at a slightly
higher elevation than Marcio.
Both lakes are located in an area that is presently occupied by small
settlements. The area around Lake Curiaú is occupied by a community founded by
former slaves who escaped captivity and built a village during the XVIII century.
About 1500 people, all descendants of the former slaves live today in the village.
The vegetation of the study site is a mosaic of savannas and patches of
secondary and dry forests. Gallery forests and palm swamps are also present along
the water drainages (locally called “igarapés”). While Lake Marcio is surrounded by
savanna, with a fringe of Mauritia palms on the shoreline, Lake Tapera is surrounded
by semi-deciduous and gallery forests, with Mauritia palms growing on the margins

32
as well. The climate in this region is tropical humid, with mean annual temperatures
of 25-27 °C. Regional weather stations document ca. 2500 mm of precipitation,
falling mainly between December-August with a dry season from September to
November (IBGE 1990).
HYPOTHESES
H1: Amazonian savannas expanded between 8000 and 5000 years BP.
H0: No savanna expansion was recorded between 8000 and 5000 years BP.
H2: Forests expanded and savanna vegetation contracted in the last 4000 years.
H0: No forest expansion was recorded in the last 4000 years.

33
N
Lake Tapera
Lake Curiaú
Amazon River
City of Macapá
Lake Marcio
10 Km
Figure 2. 1. Map of the study area, showing the location of Lakes Marcio and Tapera
(studied lakes), as well as Lake Curiaú, the City of Macapá, and the
Amazon River.

34
METHODS
The two sediment cores were raised using a Colinvaux-Vohnout piston corer
from a platform attached to inflatable boats (Colinvaux and others 1999). The
platform was anchored in the middle of each lake at the deepest point. The sealed
core tubes were transported unopened to the laboratory and stored in a dark cold
room until opened and the sediments described. A total of 68 samples (0.5cm3), 29
for pollen and 39 for charcoal, were collected from the Lake Marcio sediment core,
while 79 samples (0.5 cm3), 39 for pollen and 40 for charcoal, were collected from
the Lake Tapera sediment core. Samples for C14 AMS dating were sent to the
INSTAAR – AMS Radiocarbon Laboratory at the University of Colorado at
Boulder, and the resulting dates were calibrated to calendar years using the software
CALIB 4.0 (Stuiver and Reimer 1993).
Standard pollen extraction procedures with KOH, HF, and acetolysis
followed Faegri and Iversen (1989) and Stockmarr (1971) for the addition of tablets
of Lycopodium spores to calculate pollen concentrations. Sodium pyrophosphate was
also used on clay-rich samples. Pollen samples were mounted in glycerol and counts
of at least 300 grains were conducted at 400x and 1000x magnification on a Zeiss
Axioskop equipped with a digital camera. Pollen grains were identified using the
Florida Institute of Technology reference collection of modern pollen and published
catalogs with photographs and morphological descriptions of pollen types
(Colinvaux et al. 1999, Hooghiemstra 1984, Roubik and Moreno 1991).

35
Samples for charcoal analysis were disaggregated in 10% KOH, and the
resulting slurry washed through a 170 µm sieve (particles >170 µm were retained).
The material in the sieve was transferred to a Petri dish and charcoal counts
performed under an Olympus dissection microscope (20x magnification) equipped
with a video camera (Clark and Patterson 1997). Charcoal particles were manually
identified and then digital measurements were made via the video system and image
recognition (NIH-IMAGE) software. This software provides the area of charcoal
particles according to the number of pixels occupied by the fragment on the screen.
The pollen sum, percentage, and concentration were all calculated in TILIA (Grimm
1992), and plotted in C2 1.3 (Juggins 2003) and Corel Draw 8.
Detrended Correspondence Analysis – DCA (Hill and Gauch 1980) was
performed using PC-ORD 4.0 for Windows (McCune and Mefford 1999). This
technique was chosen because of 2 main reasons: The first one is that it prevents the
problems caused by the arch effect, which are a constant on other techniques such as
Principal Component analysis (PCA) and Correspondence Analysis (CA) (McCune
and Grace 2002). The second reason is that DCA allows the axes to be re-scaled and
shown as Standard Deviation units, which can be used to estimate the degree of
species turnover (McCune and Grace 2002). As very complex data sets with many
rare taxa can hinder DCA, the data set was reduced to include only pollen taxa with
percentage values > 1% an present at least in 3 samples throughout the sediment
core. The resulting matrices from Lakes Marcio and Tapera had 35 and 42 pollen
types respectively. The percentage values were then transformed using square root

36
transformation, as this technique is less drastic than log transformation (Kovach
1992). The resulting sample scores from DCA were plotted against sample depths for
axes 1 and 2 in EXCEL.

37
RESULTS
STRATIGRAPHY
The 470 cm-long sediment core from Lake Marcio (Fig 2.2) shows a strong
change in sediment type at 118 cm. Below 118 cm the core is composed of blue-gray
clay with wood and charcoal fragments, while above 118 cm sediments are
composed of gyttja rich in plant remains (Table 2.1). A log of wood at 347-385 cm
deep in the core was pierced longitudinally.
The 200 cm-long sediment core from Lake Tapera (Fig 2.2) is mostly
composed of blue-gray clay rich in plant fragments in the bottom (200-65 cm) and
black silt with no plant remains on the top 65 cm. In the bottom of the core (197-200
cm) there is a layer of dark-blue clay rich in black organic fragments, and between
69-76 cm depth the clay is very sandy (Table 2.2).
Table 2. 1. Sediment description of core from Lake Marcio (Amapá - Brazil).
Depth (cm) Sediment description
0 - 118 Gyttja – black peaty mud, very organic rich in plant remains
118 - 347 Blue-gray clay, with wood and charcoal fragments at different
concentrations throughout the core
347 - 385 Piece of a log that was cored through longitudinally
385 - 470 Gray clay with pieces of charcoal

38
Lake Marcio Lake Taper
a
Depth (cm)
Depth (cm)
4590 cal years BP
6100 cal years BP
6900 cal years BP
7220 cal years BP
7250 cal years BP
0
20
40
60
80
100
120
140
160
180
200
220
240
260
280
300
320
340
360
380
400
420
440
460
480
Gyttja Blue-gray clay Piece of a log Dark gray claySandy clay
Calibrated radiocarbon dates Calibrated radiocarbon dates
1260 cal years BP
1620 cal years BP
7600 cal years BP
13,690 cal years BP
7760 cal years BP
8060 cal years BP
7600 cal years BP
20
40
60
80
100
120
140
160
180
200
0
Figure 2. 2. Lithology of sediment cores from Lakes Marcio and Tapera, Amapá
(Brazil). Also showing location of dated samples (cal years BP).

39
Table 2. 2. Sediment description of core from Lake Tapera (Amapá – Brazil).
Depth (cm) Sediment description
0 - 65 Black organic silt, no plant remains present
65 - 69 Blue-gray clay rich in plant remains
69 - 76 Very sandy blue-gray clay
76 - 197 Blue-gray clay with some plant remains
197 - 200 Dark clay rich in charcoal fragments
RADIOCARBON DATES
Sedimentary chronologies for lakes Marcio and Tapera are derived from 12
AMS radiocarbon dates (Tables 2.3 and 2.4). For each calibrated age a maximum
probability solution (Calib 4.3 type 2 solution) was adopted. All ages used
henceforth will be interpolated (rounded up) calibrated years before present (BP),
unless noted otherwise. The basal ages of 7250 cal yrs BP and 8060 cal yrs BP for
lakes Marcio and Tapera respectively, indicate that the sediments provide Mid- and
Late-Holocene histories. However, the age vs. depth relationship (Fig 2.3) shows that
sedimentation rates were not constant through time.
At Lake Marcio, the bottom 354 cm of sediments was deposited in ca. 2660
years, while the top 116 cm was deposited in ca. 4590 years, showing a slowing
sedimentation rate. The transition between depositional environments is reflected in
a major sedimentary change at 118 cm. In the Lake Tapera sediments, the ages are

40
more confusing, but the chronology is still reliable if the radiocarbon age of 11,750 +
55 years BP (uncalibrated) at 132 cm is rejected as an outlier (Fig 2.3). Nevertheless,
the bottom 92 cm of sediments was deposited in only ca. 460 years. The relatively
rapid sedimentation of the upper and lower portions of the core contrasts markedly
with the section between 108 cm and 65 cm when little net sediment accumulation
occurred as the bounding ages are 7600 years BP and 1620 years BP respectively. A
likely explanation is that a sedimentary hiatus of about 6000 years occurred in this
section of the core. Thus, the basic chronology established treats sedimentation in
Lake Marcio as continuous, but identifies a ca. 6000-year hiatus in Tapera.
Table 2. 3. Radiocarbon dates from Lake Marcio sediment core.
Sample Depth (cm) 14C yr BP 13C/12C ratio Age (cal years BP)
OS-24117 116 4100 + 35 -27.30 4520 – 4660
OS-24118 218 5330 + 45 -25.60 5990 – 6210
OS-24119 317 6050 + 40 -29.30 6790 – 7000
OS-24120 432 6280 + 40 -27.30 7160 – 7280
OS-24121 470 6320 + 45 -27.80 7160 – 7330

41
Table 2. 4. Radiocarbon dates from Lake Tapera sediment core.
Sample Depth (cm) 14C yr BP 13C/12C ratio Age (cal years BP)
OS-38388 35 1320 + 30 -27.93 1230 – 1300
OS-38389 65 1710 + 35 -27.37 1540 – 1700
OS-40060 108 6720 + 30 -26.72 7570 – 7630
OS-38390 132 11,750 + 55 -25.37 *13,480 – 13,890
OS-40061 150 6730 + 35 -27.86 7570 – 7630
OS-38391 182 6930 + 45 -27.03 7670 – 7850
OS-38392 197 7260 + 35 -23.39 8000 – 8120
* Date rejected as an outlier.
11,7 50 y ears B P
0
20
40
60
80
100
120
140
160
180
200
220
240
260
280
300
320
340
360
380
400
420
440
460
480
500
0 1,000 2,000 3,000 4,000 5, 000 6,000 7, 000 8,000 9,000 10,000 11,000 12,000 13,000
Age ( 14
C years BP)
Depth (cm)
Lak e M arcio
Lak e Tapera
Rejected
Figure 2. 3. Radiocarbon ages from Lakes Marcio (diamonds) and Tapera (circles)
plotted versus depth (cm). The Lake Tapera radiocarbon age that is
represented by an open circle (11,750 years BP) was rejected as an
outlier. The dashed lines represent the extrapolated ages toward the
present.

42
PALEOECOLOGICAL RECORD – LAKE MARCIO
The pollen and charcoal records from Lake Marcio (Figs 2.4 and 2.5) allow
the distinction of 2 main zones: zone M1, which lasts from ca. 7250 cal years BP
until 4590 cal years BP, and is subdivided into zones M1A, M1B, and M1C, and
zone M2 lasting from ca. 4,590 cal years BP until the present.
M1A (470 cm – 320 cm). Sediments are composed of blue-gray clay with
wood and charcoal fragments. Pollen concentration increases from ca. 50,000 to ca.
170,000 grains/cm3 of sediment upwards (Fig 2.4). This zone lasts from ca. 7250 cal
years BP until ca. 6910 cal years BP, and is characterized by the presence of Alseis
and Apeiba at low percentages; this is the only zone in which these 2 taxa can be
found. Arecaceae and Zonorate type show their highest values in this zone.
Apocynaceae pollen, as well as Araliaceae, Asteraceae, Astronium, Bignoniaceae,
Cassia, Cupania, Dalbergia, Malpighiaceae, Mauritia, Poaceae, Protium,
Symphonia, and Virola are present at low percentage values. Cecropia pollen, a well
known disturbance indicator, can be found in this zone, although at low frequency.
Papillionaceae shows percent values of 3-5% in this zone. Podocarpus pollen is
present but very rare in this zone (Fig 2.5).
The amount of charcoal particles present is negligible (ca. 0.5 mm2/cm3 of
sediment).

43
x 10 grains/cm of sediment
33
010010010010010010010200100100100100100100100100100100100100102001020
Alch
ornea
Alseis
Ama/Che
Anacardiaceae
Apocynaceae
Araliaceae
Arecaceae
Asteraceae
Astronium
Bignoniaceae
Cassia
Cecr
opia
Cupania
Dalbergia
Euterpe
Ludwigia
Macrolobium
Malpighiaceae
Mel/Comb
Poaceae
Virola
M1A
M1B
M1C
M2
0
20
40
60
80
100
120
140
160
180
200
220
240
260
280
300
320
340
360
380
400
420
440
460
480
500
Depth (cm)
Ca l years BP
4590
6100
6900
7220
7250
x 10 grains/cm of sediment
33
010
Mauritia
01001001001001001001020304050600100100100100100 50 100 150 200
Papillionaceae
Podocarpus
Polygalaceae
Polygonum
Protium
Symphonia
Zonorate
Cyperaceae
Sagittaria
Typha
Utricularia
Spores
Total
Concentration
M1A
M1B
M1C
M2
0
20
40
60
80
100
120
140
160
180
200
220
240
260
280
300
320
340
360
380
400
420
440
460
480
500
Depth (cm)
Cal year s BP
4590
6100
6900
7220
7250
Figure 2. 4. Pollen concentration diagram of Lake Marcio, Amapá (Brazil), showing
selected taxa (scale is x103 grains/cm3 of sediment), lithology, and
radiocarbon dates (cal years BP). The hollow curves are exaggerated 5
times.

44
M1B (320 cm – 200 cm). Sediments are composed of blue-gray clay with
wood and charcoal fragments. Pollen concentration decreases from ca. 150,000 to ca.
50,000-80,000 grains/cm3 of sediment upwards (Fig 2.4). This zone lasts from ca.
6910 cal years BP until 5840 cal years BP, and is characterized by still low
percentage values of Alchornea and Poaceae. Anacardiaceae pollen starts being
deposited but at low percentages. Pollen of Apocynaceae, Araliaceae, Arecaceae,
Asteraceae, Cupania, Dalbergia, Papillionaceae and Zonorate type all show
decreasing percentage values. Conversely, pollen from Astronium, Bignoniaceae,
Cassia, Cecropia, Euterpe, Malpighiaceae, Podocarpus, Protium, Symphonia, and
Virola show increased percentages in this zone. Pollen from aquatic plants such as
Cyperaceae, Sagittaria and Typha start being deposited (Fig 2.5).
There is a peak in the accumulation of charcoal particles (ca. 16 mm2/cm3 of
sediment) in the beginning of this zone.
M1C (200 cm – 120 cm). Sediments are composed of blue-gray clay with
wood and charcoal fragments. Pollen concentration decreases from ca. 80,000 to ca.
15,000 grains/cm3 of sediment upwards (Fig 2.4). This zone lasts from ca. 5840 until
4750-4650 cal years BP, and is characterized by an increased percentage of
Alchornea, Apocynaceae, Araliaceae, Arecaceae, Asteraceae, Bignoniaceae,
Dalbergia, Papillionaceae, Melastomataceae/Combretaceae type (Mel/Comb), and
Podocarpus. Percentage of Poaceae pollen shows a 7-fold increase in abundance (ca.
5-35%) on the top part of this zone. Pollen from Cassia, Cecropia, Malpighiaceae,
Protium, Virola and Zonorate type show slightly decreased percentages. Pollen from

45
Amaranthus/Chenopodiaceae type is recorded for the first time, but still at low
percentage values (Fig 2.5).
There is another peak in the accumulation of charcoal particles (ca. 13
mm2/cm3 of sediment) in the beginning of this zone.
M2 (120 cm – 0 cm). Sediments are composed of gyttja rich in plant remains.
Pollen concentration fluctuates around ca. 30,000 grains/cm3 of sediment (Fig 2.4).
This zone lasts from ca. 4750-4650 cal years BP until the present, and is
characterized by low percentages of Alchornea, Apocynaceae, Arecaceae,
Bignoniaceae, Cassia, Dalbergia, Papillionaceae, Malpighiaceae, Protium,
Symphonia, Virola, and Zonorate type. Podocarpus is not recorded in this zone.
Pollen from Ludwigia, Macrolobium, Polygalaceae, Polygonum, and Utricularia are
recorded only in this top zone with percentage values varying from 2-3%
(Polygalaceae) to 20% (Polygonum). Amaranthus/Chenopodiaceae pollen shows
slightly higher percentage values. Percentage of Poaceae pollen displays some
fluctuation, but remains high until the end of the zone (Fig 2.5).
Accumulation of charcoal particles is greatly increased in this zone. There are
at least 5 peaks of charcoal varying in area from 50-140 mm2/cm3 of sediment.

46
0
100
200
300
400
500
Alchornea
010
Alseis
010
Ama/Che
010
Anacardiaceae
010
Apocynaceae
010
Araliaceae
010
Arecaceae
010 20
Astronium
010
Asteraceae
010
Bignoniaceae
010
Cassia
010
Cecropia
010
Cupania
010
Dalbergia
010
Euterpe
010
Papillionaceae
010
Ludwigia
010
Macrolobium
010
Malpighiaceae
010
Mauritia
010 20
Mel/Comb
010
Poaceae
01020 30 40 50 6 0
Podocarpus
010
Apeiba
Depth (cm)
0
100
200
300
400
500
Polygalaceae
010
Polygonum
0102030
Protium
010
Symphonia
010
Virola
01020
Zonorate
01020304050 60 70 80 90
Cyperaceae
0102030
40 50 60 70 80
Sagittaria
010
Typha
010
Utricularia
01020
Spores
0102030
Charcoal
(mm /cm )
23
020 40 60 80 100 120 140 160 180
Depth (cm)
M1A
M1B
M1C
M2
M1A
M1B
M1C
M2
Figure 2. 5. Pollen percentage diagram of Lake Marcio, Amapá (Brazil), showing the
most representative taxa and area of charcoal particles (mm2/cm3 of
sediment). Hollow curves are exaggerated 5 times. The pollen record
allowed the recognition of 2 main zones, and 3 sub-zones.

47
PALEOECOLOGICAL RECORD – LAKE TAPERA
The pollen and charcoal records from Lake Tapera (Figs 2.6 and 2.7) also
allow the distinction of 2 main zones: zone T1, which lasts from ca. 8060 until
before 1620 cal years BP, and is subdivided into zones T1A and T1B, and zone T2
lasting from before 1620 cal years BP until the present.
T1A (200 cm – 110 cm). Sediments are composed of dark clay rich in
charcoal fragments on the bottom, and blue gray clay with plant fragments. Pollen
concentration displays strong fluctuations, peaking at 190 cm (ca. 32,500 grains/cm3
of sediment), 165 cm and 150 cm (around 15,000 grains/cm3 of sediment) (Fig 2.6).
This zone lasts from 8060 cal years BP until ca. 7610 cal years BP, and is
characterized by low percentages of Asteraceae, Bignoniaceae, Cassia, Cecropia,
Dilleniaceae, Euterpe, Mimosaceae, Macrolobium, Mauritia, Combretaceae type
(Mel/Comb), Moraceae/Urticaceae 2-porate and 3-porate types (Mor/Urt2,
Mor/Urt3), Poaceae, Polygonum, Spondias, and Cyperaceae. On the other hand,
pollen from Coccoloba, Desmodium, Didymopanax, Lecythidaceae, Caesalpiniaceae,
Papillionaceae, Machaerium, Malpighiaceae, Protium, Symphonia, Virola, and
Zonorate (types 1 and 2) are relatively abundant in this zone (Fig 2.7). No charcoal
particles were found in this zone.

48
x 100 grains/cm of sediment
3
020010 010 010 010 010 010 02040010 010 010 010 010 010 010 01002002040010 010 010 020030010 010 010 010
Al
chornea
Ama/Che
Anacardiaceae
Asteraceae
Bignoniaceae
Cass
ia
Caesalpiniaceae
Cecropia
Coccoloba
Desmodium
Didymopanax
Dilleniaceae
Euphorbiaceae
Euterpe
Lecythidaceae
Machaerium
Macrolobium
Malpighiaceae
Mauritia
Mel/Comb
Mimosaceae
Mor/Urt2
Mor/Urt
3
Mor/Urt4
Myrtaceae
Pachira
Papillionaceae
Depth (cm)
Cal years BP
T1A
T1B
T2
1270
1620
7600
7760
8060
7600
0
10
20
30
40
50
60
70
80
90
100
110
120
130
140
150
160
170
180
190
200
x 100 grains/cm of sediment
3
T1A
T1B
T2
020406080010 020010 010 010 010 0204060800 204060800204060020020010 010 0200 50 100150200250300350
Poaceae
Polygonum
Pr
otium
Rubiaceae
Sapindaceae
Solanaceae
Symphonia
Virola
Zonorate 1
Zonorate 2
Cyperaceae
Sagittaria
T
ypha
Utricularia
Spores
Tot
al
Concentration
Depth (cm)
Cal years BP
1270
1620
7600
7760
8060
7600
0
10
20
30
40
50
60
70
80
90
100
110
120
130
140
150
160
170
180
190
200
Figure 2. 6. Pollen concentration diagram of Lake Tapera, Amapá (Brazil), showing
selected taxa (scale is x100 grains/cm3 of sediment), lithology, and
radiocarbon dates (cal years BP). The hollow curves are exaggerated 5
times.

49
T1B (110 cm – 75 cm). Sediments are composed of blue-gray clay with plant
fragments. Pollen concentration displays its lowest values (ca. 1500 grains/cm3 of
sediment) (Fig 2.6). This zone lasts from 7610 cal years BP until before 1620 cal
years BP, and is differentiated from the previous zone by higher percentage values of
pollen from Asteraceae, Bignoniaceae, Cassia, Cecropia, Caesalpiniaceae,
Mor/Urt3, Myrtaceae, Rubiaceae, and Cyperaceae. Nevertheless, a few taxa such as
Didymopanax, Machaerium, Malpighiaceae, Pachira, Protium, and Zonorate type 2
show decreased percentage values. Pollen of Coccoloba, Desmodium,
Euphorbiaceae, Moraceae/Urticaceae 4porate type (Mor/Urt4), and Spondias are not
recorded in this zone. Symphonia and Virola present lower percentages in the
beginning of this zone, but show a recovery closer to the end (Fig 2.7).
No charcoal particles were recorded in this zone.
T2 (75 cm – 0 cm). Sediments are composed of a layer of sandy blue-gray
clay on the bottom, a transition from blue-gray clay rich in plant fragments to black
organic silt with no plant fragments. Pollen concentration increases sharply and
fluctuates around 10,000 and 20,000 grains/cm3 of sediment, displaying values
comparable to the zone T1A (Fig 2.6). This zone lasts from before 1620 cal years BP
until the present, and can be characterized by decreased percentage values of
Asteraceae, Bignoniaceae, Cassia, Coccoloba, Desmodium, Didymopanax,
Lecythidaceae, Caesalpiniaceae, Machaerium, Malpighiaceae, Protium, Rubiaceae,
Symphonia, Virola, and Zonorate type 1. Other taxa display increased percentages,
such as Euterpe, Macrolobium, Mauritia, Mel/Comb, Polygonum, Spondias,

50
Zonorate type 2, Sagittaria, Typha, and Utricularia. The percentage of Poaceae
pollen exhibits a 5-fold increase in this zone, shows some fluctuation but remains
high until the end. Cecropia pollen reaches its highest values at 55 cm (ca. 17%), but
does not increase above 10% displaying only small variation throughout this zone
(Fig 2.7).
Significant amounts of charcoal particles occur for the first time at 65 cm
(1620 cal years BP), and display a peak of ca. 15 mm2/cm3 of sediment at 60 cm. At
the depth of 35 cm (1258 cal years BP) charcoal area begins to decrease, and values
remain small until the end of the zone.

51
0
20
40
60
80
100
120
140
160
180
200
010
Alchornea
010
Ama/ Che
010
Anacardiaceae
01020
Aster ace
ae
01020
Bignoniaceae
010
Cassia
01020
Cecropia
010
Coccoloba
010
De
smodium
010
Didymopanax
010
Dilleniaceae
010
Euphorbia
ceae
010
Euter pe
01020
Lecythidacea e
010 20
Caesalpiniacea
e
010
Mimosaceae
010
Papillionaceae
010
Machaerium
01020
Macrolobium
0102030
Malpighiaceae
010
Mauritia
010
Mel/Comb
010
Mor/Urt2
01020
Mor/Urt3
Depth (cm)
T1A
T1B
T2
0
20
40
60
80
10 0
12 0
14 0
16 0
18 0
20 0
010
Mor/Urt4
010
Myrtaceae
010
Pachira
0204060
Poaceae
010
Polygonum
010
Protium
010
Rubiaceae
010
Sapin
daceae
010
Solanaceae
010
Spondias
0102030
Symphonia
02040
Virola
01020 30
Zono rate 1
0102030
Zonorate 2
01020
Cyperaceae
010
Sagittaria
010
Typha
010
Utricularia
0150
Spores
01020
Char coal
Depth (cm)
(mm / c m )
23
T1A
T1B
T2
Figure 2. 7. Pollen percentage diagram of Lake Tapera, Amapá (Brazil), showing the
most representative taxa and area of charcoal particles (mm2/cm3 of
sediment). Hollow curves are exaggerated 5 times. The pollen record
allowed the recognition of 2 main zones, and 2 sub-zones.

52
MULTIVARIATE ANALYSIS – LAKE MARCIO
The resulting DCA scores were plotted in two different ways. First, Axis 1
vs. Axis 2, then the sample scores from Axis 1 were plotted with the corresponding
depths (Fig 2.8).
AXIS 1 vs. AXIS 2. The DCA scores from axes 1 and 2 showed a strong
polarization of samples on Axis 1 that divided them into two main groups: the
bottom samples of the core on the right and the top samples of the core on the left
representing zones M1 (ABC) and M2 (respectively). The species that scored highest
and therefore were the most characteristic of samples being placed on the positive
side of Axis 1 were Alseis, Podocarpus, Zonorate type 1, Symphonia, and Arecaceae.
The lowest scores were yielded by Polygala, Utricularia, Polygonum, Ludwigia, and
Macrolobium, which brought the samples to the negative side of Axis 1 in zone M2.
A relatively weaker polarization of samples that corresponded to zones M1A and
M1C was also observed on Axis 1. Axis 2 displayed a very weak polarization of
samples into groups subdividing zone M1, therefore only Axis 1 was plotted against
depth.
AXIS 1 vs. DEPTH. The contrast in the DCA scores of the samples in groups
M1 (ABC) and M2 was especially pronounced when plotted against depth. Within
their respective groups the samples showed little variation, however the transition
from zone M1 (ABC) to zone M2 is seen to occur abruptly at 120 cm (between ca.
6100 cal years BP and 4590 cal years BP).

53
Axis 1
80
80 10 0
60
60
40
40
20
20
0
0
Axis 2
M-5
M-25
M-85
M-55
M-6 5
M-75
M-95
M-0
M-17 0
M-390
M-415
M-380
M-347
M -340
M-135
M -125
M-200
M-320
M-255
M-2 1 3
M-400
M-4 72
M-425
M-290
M-3 5
M-15
M-45
M-120
M-1 10
Phase M2
Phase M1C
Phase M1A
Phase M1B
Phase M1(ABC) Phase M2
4590 cal y ears BP
6100 cal years BP
6900 cal years BP
7250 cal years BP
7220 cal years BP
0
20
40
60
80
100
120
140
160
180
200
220
240
260
280
300
320
340
360
380
400
420
440
460
480
500
-40 -20 0 20 40 60 80 100 120 140 160
DCA scores - axis 1
Depth (cm)
Figure 2. 8. DCA scores from Lake Marcio. Top graph shows Axis 1 vs. Axis 2.
Samples are grouped according to zones interpreted from pollen
diagrams. Bottom graph displays resulting DCA scores of Axis 1 plotted
vs. depth (cm) showing a strong polarization of samples into two main
groups representing zones M1(ABC) and M2.

54
MULTIVARIATE ANALYSIS – LAKE TAPERA
The resulting DCA scores from Lake Tapera were also plotted in two
different ways. First, Axis 1 vs. Axis 2 (Fig 2.9), then the sample scores from axes 1
and 2 were plotted with the corresponding depths (Fig 2.10).
AXIS 1 vs. AXIS 2. The DCA polarized samples on axes 1 and 2 so that the
bottom samples of the core are on the positive side of Axis 1, representing zone T1
(AB), and the top samples of the core on the left side representing zone T2.
The highest scores for Axis 1 were yielded by Pachira, Bignoniaceae,
Malpighiaceae, Symphonia, and Euphorbiaceae. The species that scored lowest in
this axis, and therefore were responsible for samples being placed on the more
negative side of Axis 1 on zone T2, were Utricularia, Polygonum, Sagittaria,
Mimosaceae, and Macrolobium.
On Axis 2 samples from zone T1 (bottom of the core) were subdivided into
two groups, T1A and T1B. The species that scored highest in this axis and were
indicators of T1A were Coccoloba, Typha, Desmodium, Didymopanax, and Virola.
The lowest species scores (associated with T1B) were yielded by Solanaceae,
Bignoniaceae, Rubiaceae, Lecythidaceae, Papillionaceae, Asteraceae, Cyperaceae,
Caesalpiniaceae, Myrtaceae, and Symphonia.
AXIS 1 vs. DEPTH. Plotting Axis 1 DCA sample scores against sample
depth revealed a striking pattern, similar to the one observed in Lake Marcio, in
which a transition between zones T1 (AB) and T2 occurred at 75 cm (between ca.
7591 cal years BP and 1618 cal years BP). In zone T1 samples were placed on

55
positive side of Axis 1 and showed small oscillation, with the exception of sample T-
100 (100 cm) that scored much lower values. In zone T2, from 75 cm until 0 cm
(before 1618 cal years BP until the present), samples present smaller range of
variation, and were placed on the negative side of the axis.
AXIS 2 vs. DEPTH. The contrast in the DCA scores of the samples from
zones T1A and T1B was well-defined when plotted against depth. The transition
from zone T1A to T1B occurred at 110 cm, and was marked by the displacement of
samples from the positive to the negative side of the axis. The transition from T1B to
T2 occurred at 75 cm, and was characterized by the displacement of samples toward
relatively more positive values.

56
0
0
40 80
40
80
T-20
T-40
T-10 0
T-25
T-0
T-60
T- 5
T- 1 0
T-15
T-35 T- 5 0
T- 4 5
T- 5 5
T- 20 0
T-19 7 T-195
T-17 0
T- 1 55
T- 1 65
T- 1 50 T-14 5
T-160
T- 1 25 T-140
T-115 T- 1 30
T- 1 20
T- 1 3 5 T-80
T- 1 8 0
T-90
T-95
T-85
T- 7 5
T- 11 0
T-185
T- 1 75
T-30
T-65
Axis 2
Axis 1
Phase T1A
Phase T1B
Phase T2
Figure 2. 9. DCA scores from Lake Tapera showing Axis 1 vs. Axis 2. Samples are
grouped according to zones interpreted from pollen diagrams. Axis 1
polarizes samples into two main groups that represent zones T1 (AB)
and T2. Axis 2 polarizes bottom samples into two subgroups, allowing
further division of zone T1 into T1A and T1B.

57
1270 cal years BP
1620 cal years BP
7600 ca l years BP
8060 cal years BP
Phase T1(AB) Phase T2
7600 cal years BP
7760 cal years BP
0
10
20
30
40
50
60
70
80
90
100
110
120
130
140
150
160
170
180
190
200
-2 0 -1 0 0 10 2 0 30 40 50 60 7 0 8 0 90 100 1 10 120 13 0 14 0 150
DCA scores - Axis 1
Dept h ( cm)
Phase T1A Phase T1B Phase T2
1270 cal years BP
1620 cal years BP
7600 cal years BP
8060 cal years BP
7600 cal years BP
7760 cal years BP
0
10
20
30
40
50
60
70
80
90
100
110
120
130
140
150
160
170
180
190
200
-7 0 -60 -50 - 40 - 30 -2 0 -1 0 0 10 20 30 4 0 5 0 6 0 70 80 90
DC
A
sco res -
A
xis 2
Dept h (cm)
Figure 2. 10. DCA scores from Lake Tapera, axes 1 and 2 plotted vs. depth (cm), and
respective calibrated ages. Scores from Axis 1 show a marked division
into two main zones. Scores from Axis 2 show further subdivision of
zone T1.

58
DISCUSSION
Looking at the stratigraphy alone it is apparent that the cores from Lakes
Marcio and Tapera provide a record of an environmental change, since both cores
showed blue-gray clay being replaced by black-organic sediments (Fig 2.2). The
sharp transition of the sediments implied that this change was at least locally
profound and maybe even abrupt. The blue-gray clay that constituted the bottom of
both sediment cores was full of plant remains, wood and charcoal fragments,
indicating a shallow and unstable depositional environment. The coring location may
have supported a swamp forest subject to periodic flooding, which may explain the
presence of wood and plant fragments, especially the log that was pierced during
coring of Lake Marcio. The sandy layer at 69-76 cm in the Lake Tapera core was
another indication of environmental instability, suggesting drought and maybe
associated erosion.
The radiocarbon dates provided further evidence of an alteration in the
environment (Fig 2.3). At Lake Marcio, the basal 354cm of clay had an average
sedimentation rate of ~ 0.13 cm yr-1, but sedimentation in the upper 116 cm of the
core was at just 0.025 cm yr-1, a five-fold decrease. The obvious slowing coincided
with the transition from a core dominated by allochthonous clays to autochthonous
organic material.
A sharp break in sediment type, a rapid change in microfossil composition
(Fig 2.8), and different sediment deposition rates above and below the boundary

59
suggested the possible presence of a sedimentary hiatus. Given the above, a simple
projection of sedimentation rates above and below the sediment transition to the
point of the transition at 118 cm revealed a temporal mismatch, i.e. an apparent gap.
Applying the sedimentation rate of 0.123 cm yr -1 (between 317 and 218 cm), the
interpolated age for the clay at 120 cm (immediately beneath the transition) would be
5300 cal years BP. Similarly, by projecting the sedimentation rate for the sediments
between 0 and 116 cm downward, an interpolated age for onset of the organic
sediment deposition at 120cm would be 4750 cal years BP. The most parsimonious
explanation of all the observed sedimentary and biological changes is a ~110-year
gap in sedimentation (5300-4750 = 550). Such breaks in sedimentation are common
in shallow tropical lakes (e.g. Listopad 2001) and are generally taken to indicate
conditions in which lake level fell.
Lake Tapera, which is slightly more isolated from floodwaters than Marcio,
revealed a record that is different from, but consistent with, that of Marcio. The
basal age of Tapera is 8060 cal years BP, suggesting a similar initial timing of
ponding to that of Marcio (~7200 BP). As in Marcio, sedimentary and biotic changes
suggested a strong discontinuity. The duration of the sedimentary hiatus at Tapera,
based on interpolated sedimentation rates above and below the horizon is ca. 5840
years (7510-1670=5840 years). From this observation we inferred that lake level fell
during this interval and a permanent water body was reestablished only at ca. 1670
cal years BP.

60
The re-flooding of Marcio and Tapera both initiated deposition of an organic
rich mud. That the hiatus at Tapera started 2200 years earlier and finished almost
3000 years later than at Marcio suggested that the environmental conditions that
caused the sedimentary hiatuses had stronger and more prolonged effects on Lake
Tapera than on Lake Marcio.
The pollen and charcoal records from Lakes Marcio and Tapera revealed two
main contrasting zones. The most obvious biological distinction between these zones
in both lakes was the increase in Poaceae pollen, rising from < 10% in M1 and T1 to
almost 60% (30-55%) in M2 and T2. These percentage values are comparable with
the proportion of Poaceae pollen (50-90%) that is found in the pollen rain of cerrados
(Salgado-Labouriau 1973), suggesting a savanna expansion in M2 and T2.
Even though comparable percentages of Poaceae pollen are also observed in
records from Colombian savannas such as Lagunas Carimagua and El Piñal (Behling
and Hooghiemstra 1999), Lagunas Chenevo and Mozambique (Berrio and others
2002), and Lagunas Angel and Sardinas (Behling and Hooghiemstra 1998), an
increase in Poaceae, as marked as the one observed in Lakes Marcio and Tapera,
could be found only in the Lake Crispim (northern coast of Pará) pollen record
(Behling and Costa 2001).
A further indication of a drier environment was the loss of the swamp
Zonorate type pollen in zone 2. Unfortunately, the identification of Zonorate pollen
types was uncertain, though they closely resemble Cestrum in the Solanaceae. From
observing the co-occurring species in these and other lakes, we infer that this pollen

61
type is derived from a woody species of swamp forest. Consistent with a drier habitat
at this time, is the reduced occurrence of forest taxa, e.g. Alseis, Apocynaceae,
Bignoniaceae, Caesalpiniaceae, Cassia, Lecythidaceae, Machaerium, Malpighiaceae,
Papillionaceae, Podocarpus, Protium, Symphonia, and Virola (Marchant and others
2002) in M2 and T2. Taxa representative of flooded savanna and swamp vegetation,
such as Amaranthus/Chenopodiaceae, Asteraceae, Ludwigia, Macrolobium,
Mauritia, Polygalaceae, Polygonum, Cyperaceae, and Utricularia display increased
percentages in M2 and T2 compared with M1 and T1. Significantly, some swamp
indicator taxa such as Ludwigia, Macrolobium, Polygalaceae, and Polygonum were
present only in M2 and T2, reinforcing the suggestion of a general trend toward
flooded savanna vegetation.
Fluctuations of pollen concentration can sometimes be interpreted as changes
in the sedimentation rate. At Lake Marcio, slower sedimentation rates were inferred
from chronology. Therefore, the decreasing pollen concentration upcore was
probably the result of the replacement of a swamp forest (M1), a more closed
vegetation type which would provide a more direct input of pollen, by a savanna
(M2), a more open vegetation type, which would provide relatively smaller
quantities of pollen.
In Lake Tapera however, the decreasing pollen concentration toward T1B
was consistent with the observed faster sedimentation rates. At the end of T1B, the
lake dried out, which would explain the lowest concentration values. Within T2,

62
when the lake started depositing sediments again, fluctuations in pollen
concentration were minor.
The quantity of charcoal particles present in the sediments from Lakes
Marcio and Tapera, measured here as charcoal area (mm2) per volume of sediment
(cm3), was used as proxy for past fire frequency. In Lake Marcio, fires were recorded
since 6900 cal years BP, but at insignificant levels. However, they became much
more frequent after 4590 cal years BP. In Lake Tapera, charcoal was recorded only
after 1620 cal years BP. As fire is most unlikely to have occurred in the mesic forests
present around the sites at ca. 7000 BP, the most probable cause of fire is human
activity. Hence, these charcoal records provide evidence of human impacts in the
area since ca. 6900 cal years BP. Their effects were probably stronger at Lake
Marcio, since fires are recorded at much higher frequencies than in Lake Tapera,
though there was a sedimentary hiatus at Tapera from 7520 to 1620 cal years BP.
The sample scores derived from DCA when plotted with core depths
provided an illustrative way to demonstrate the impact of environmental changes that
took place in those systems. The Axis 1 of both lakes suggested the same general
pattern. A system more influenced by forest vegetation in M1 and T1 changing into a
flooded savanna in M2 and T2. The timing of environmental change was relatively
coincident with the on set of fire. The replacement of Symphonia and Pachira in M1
and T1 by Polygonum and Polygalaceae in M2 and T2 suggested subtle changes
within the wetland plant community.

63
The ordination results indicated that zone T1B represents a relatively more
impacted environment than zones T1A and T2.
Even though Lake Marcio and Lake Tapera revealed somewhat independent
histories, a similar pattern emerged from their pollen data. In both records, a general
environmental trend from forest (in M1 and T1) to a flooded savanna (M2 and T2)
was apparent. Anthropogenic impacts on the environment were suggested by large
quantities of charcoal particles especially after ca. 4590 cal years BP in Lake Marcio
and after ca. 1600 cal years BP in Lake Tapera. Furthermore, dry environmental
conditions were inferred from both records. In Lake Marcio these conditions were
severe enough to interrupt deposition of sediments between 5300 cal years BP
(interpolated) and 4750 cal years BP, lasting ca. 550 years. In Lake Tapera however,
the dry conditions had more profound impacts, stopping deposition of sediments for
a much longer period of time, from 7520 cal years BP (interpolated) to 1670 cal
years BP.
Based on the Paleoecological records provided by Lakes Marcio and Tapera,
it is clear that no savanna expansion was recorded between 8000 and 5000 years BP,
therefore hypothesis 1 was rejected.
H1: Amazonian savannas expanded between 8000 and 5000 years BP.
As the records indicate that savanna vegetation expanded only after ca. 4500
years BP, and did not contract in the last 4000 years, hypothesis 2 was also rejected.
H2: Forests expanded and savanna vegetation contracted in the last 4000 years.

64
In the early Holocene, sea-level started to rise and reached its modern level at
approximately 7000 cal years BP in the northern coast of Brazil (Bezerra and others
2003). The impacts of higher sea-level on the water table changed the dynamics of
rivers and formed lakes in the Amazonian lowlands (Behling 1996; Behling and Da
Costa 2000, Behling and Costa 2001, Behling et al. 2001, Behling 2002). The rising
water is a probable cause for the initial flooding of the site.
At ca. 5000 cal years BP the relative sea-level was 2-3 m higher than modern
levels (Suguio et al. 1985, Bezerra et al. 2003), though some of this apparent
highstand is attributable to isostatic change. Nevertheless sea-level was somewhat
higher at this time, yet lake level was falling. Therefore, it is parsimonious to
conclude that the dry environmental conditions were caused by precipitation or
evaporative change, rather than falling sea-level. From the hiatus in Marcio between
5300-4750 cal years BP we inferred that this period represents the peak of the dry
event. That Lake Tapera was more isolated from seasonal floodwater may have made
it more susceptible to evaporative loss than Marcio. Although a number of factors
can contribute to lowered lake level, the falling lake level would be consistent with
evaporation exceeding precipitation input. Lake Marcio is part of the larger seasonal
wetland, Lake Curiaú. The larger catchment and seasonal floodwater received by
Marcio may have allowed it to continue to hold water even though evaporation
exceeded precipitation between ca. 7500 BP and 1620 BP reaching its peak between
5300-4750 cal years BP.

65
When relative sea level rose for a second time, at 2000 cal years BP (Bezerra
and others 2003), it probably caused the water table to rise again. This time however,
climate was wetter, and sediment deposition in Lake Tapera resumed.
Identifying the direct cause of environmental changes observed in the
paleoecological records is often difficult, especially if the studied site records human
impacts. However, at the Amapá sites at least three driving forces could be
recognized as possible sources for the observed mid- and late-Holocene
environmental changes: relative sea-level change, human impacts, and climate
change. Although initial ponding is attributed to rising sea-level in the Holocene, a
lack of synchrony in the phasing of lake and ocean highstands suggests that relative
sea-level changes can be eliminated as a cause of lake level fluctuations after ca.
7000 cal BP. Drawing the distinction between the climate and anthropogenic signals
is more complex. However, that the rise in Poaceae pollen abundance (120 cm) at
Lake Marcio clearly precedes the increase in charcoal representation suggests that
vegetation change took place before the onset of fire, and therefore is unlikely to
result from human impacts on the environment. Thus, the transition to a drier middle
Holocene landscape, which originated a savanna expansion at the Amapá site, was
probably climatic rather than anthropogenic in origin.

66
CHAPTER III
PALEOECOLOGICAL RECORD OF PRAINHA REGION (BRAZIL)
INTRODUCTION
This study aims to reveal the paleoecological history of Amazonian forest-
savanna ecotones in Brazil during the Holocene. Despite having the second largest
vegetation cover in South America, savannas are still poorly known. In central Brazil
(the core area), savannas cover an area of ca. 2 million km2, although disjunct areas
occur within Amazonia, especially in Amapá and Roraima states (Salgado-Labouriau
1997). According to Pennington et al. (2000), savannas and dry forests can occur
under similar precipitation regimes, however, savannas are mostly found in areas
with poorer soils.
The most recent paleoecological data from forest-savanna ecotones indicate
that these landscapes have been very dynamic during the Holocene. Fossil pollen
records from Colombian savannas suggest a savanna contraction with an associated
forest expansion after ca. 4000 cal years BP (Behling and Hooghiemstra 1998;
Behling and Hooghiemstra 1999; Berrio and others 2002). Conversely, another study
conducted on the savanna-forest ecotone in Bolivia indicated a 100 km migration of
the forest edge into the savanna in the last 3000 years (Mayle and others 2000).
Even though Holocene climate changes have been proposed to explain
variations in the savanna coverage, human activities are also known to have altered

67
Amazonian landscapes. Indeed, over the last 20 years several archaeological studies
have provided strong evidence that Amazonian lowlands have been occupied by
humans for much longer than previously thought, contesting the early view of a
virgin, untouched Amazonia (Athens and Ward 1999; Roosevelt 2000; Roosevelt
and others 1991; Roosevelt and others 1996). Early humans in Bolivia, not only built
elevated roads that cut through savannas in Baures, but also constructed hydraulic
networks (e.g. canals and ponds) where they used fish weirs to increase their animal
protein intake (Erickson 2000; Erickson 2001). Such modifications allowed dense,
large populations to be sustained in a savanna environment. Further evidence of Pre-
Columbian large scale transformation of local landscapes associated with
anthropogenic impacts comes from a study conducted in the Upper Xingu
(Heckenberger and others 2003).
Near the study site (Monte Alegre and Taperinha), archaeological studies
provide further evidence of extensive human occupation in the Amazonian lowlands
(Roosevelt 2000; Roosevelt and others 1991; Roosevelt and others 1996).
Archaeological remains excavated from “Caverna da Pedra Pintada”, a painted
sandstone cave near Monte Alegre, suggest a long-term occupation by foragers since
the Late Pleistocene. These Paleoindians represent the earliest record of human
occupation in the Amazon region.
Past fires in the catchment are manifested in sedimentary sequences by the
presence of particulate carbon. Charcoal analyses of sediments are performed to
quantify past fire frequency and its role during past climate changes or as a tool of

68
land clearance (Behling 1996; Bush and others 2000; Clark 1988; Clark and others
1996; Kennedy and Horn 1997; Patterson and others 1987). The size of charcoal
particles may indicate the distance from the fire to the study site. Small particles can
be transported further than larger particles, suggesting more regional fires (large
geographic scale). Whereas large particles of charcoal reflect fires that burned close
to the water’s edge.
Paleoecological data from two lakes ca. 80 km from Monte Alegre (Lakes
Geral and Comprida) suggest that human disturbance started as early as ca. 6600 cal
years BP in this region. The consistent presence of abundant charcoal and crop
pollen suggest agriculture from 4300 cal yrs BP until ca. 1000 cal yrs BP at Geral
(Bush and others 2000). Near the top of the sequence, disturbance indicators
decrease, suggesting a reduction in the intensity of human activity.
To study the environmental changes of forest-savanna ecotones in Eastern
Amazonia, paleoecological records from two lakes (Santa Maria and Saracuri), that
lie in this mosaic landscape, were analyzed. Because Lake Geral lies between these
two lakes, its paleoecological data were re-analyzed taking into account the most
recent records. The resulting data yielded by the three lakes together provided the
paleoecological history of the Prainha region at both large and small geographical
scales.

69
STUDY SITE
Lakes Santa Maria (1°35’S, 53°35’W), Geral (1°38’S, 53°36’W), and
Saracuri (1°40’S, 53°34’W) lie between 40 and 70 m elevation, on a N-S transect
perpendicular to the Amazon River. Lake Santa Maria (the northernmost and
highest) is ca. 24 km away from the river, whereas Saracuri (the southernmost and
lowest) is only ca. 16 km from the Amazon. Lake Geral lies centrally between Santa
Maria and Saracuri (Fig 3.1). The study sites are about 25 km away from the village
of Prainha, the closest relatively large human settlement, and about 80 km from
Monte Alegre.
Lake Santa Maria is a small, very eutrophic lake with about 400 m x 100 m
of open water, ca. 2.5 m deep. The open water is surrounded by a ca. 40 m wide
fringing floating mat of swamp taxa (e.g. Pontederia, Utricularia, Montricardia,
Pachira, and Poaceae). A marked beach suggests a paleo-shoreline ca. 3 m above the
present level. Around the lake, some forest patches remain, though they are all
disturbed, and large areas have been cleared for cattle (water buffalos) and small-
scale cultivation. Lake Geral is about 3 km long, 1 km wide, and ca. 6 m deep. This
lake has been largely used as a fishery and resort by the people of Prainha. At the
east side there is an extensive wooded back-swamp, which is probably related to the
large wetland area located east of Geral and Saracuri. For a further description of
Geral see Bush et al. (2000). Lake Saracuri is ca. 1.8 km long, 0.5 km wide and ca. 4
m deep, with a sandy shoreline. The highest lake level is estimated to have been 0.5

70
m above the present level. However, as the fieldwork took place in the dry season,
this could be the water level reached during the wet season.
N
Lake Santa Maria
Amazon River
4 km
Lake Gera l
Lake Saracuri
Figure 3. 1. Map of South America showing Prainha. The inset is a false color
Landsat TM image showing bands 7, 4, and 2. Pink represents highly
reflective surfaces such as savannas, urban areas and bare ground,
whereas green represents the more complex structure of forests

71
The three lakes are located within a mosaic of semi-deciduous forest and dry
forest, although large areas of scrub savanna lie on nearby higher ground. Gallery
forests and swamps are also present along the water drainages (locally called
“igarapés”) and in a wetland located east of the sites.
The climate in this region is tropical humid, with mean annual temperatures
of 27 °C and 2200 mm of precipitation, with a dry season 3-4 months long (IBGE
1990). Because this region is wet enough to support at least semi-deciduous rain
forest, the abundance of savanna and dry forest is either due to edaphic factors,
which implies that the ecotone should not migrate much through time, or human
activities. Prance and Schubart (1977) suggested that savannas on white sand soils,
especially near Rio Negro, were formed as a result of intense land-use (e.g. cutting
and burning). As a result of the impacts of human activities, the soil became so
nutrient-poor that the forest cannot recover, and the only vegetation supported is
savanna.
HYPOTHESES
H1: Amazonian savannas expanded between 8000 and 5000 years BP.
H0: No savanna expansion was recorded between 8000 and 5000 years BP.
H2: Forests expanded and savanna vegetation contracted in the last 4000 years.
H0: No forest expansion was recorded in the last 4000 years.

72
METHODS
The 3 sediment cores were raised using a Colinvaux-Vohnout piston corer
from a platform attached to inflatable boats (Colinvaux and others 1999). The
platform was anchored in the middle of each lake at the deepest point. The sealed
core tubes were transported unopened to the laboratory and stored in a dark cold
room until opened and the sediments described. A total of 134 samples (0.5cm3), 45
for pollen and 89 for charcoal, were collected from Lake Santa Maria sediment core,
while 131 samples (0.5 cm3), 45 for pollen and 86 for charcoal, were collected from
Lake Saracuri sediment core. The sediment core from Lake Geral yielded 36 samples
for pollen and charcoal analysis (Bush and others 2000). Samples for C14 AMS
dating were sent to the INSTAAR – AMS Radiocarbon Laboratory at the University
of Colorado at Boulder, and the resulting dates were calibrated to calendar years
using the software CALIB 4.0 (Stuiver and Reimer 1993).
Standard pollen extraction procedures with KOH, HF, and acetolysis
followed Faegri and Iversen (1989) and Stockmarr (1971) for the addition of tablets
of Lycopodium spores to calculate pollen concentrations. Sodium pyrophosphate was
also used on clay-rich samples. Pollen samples were mounted in glycerol and counts
of at least 300 grains were conducted at 400x and 1000x magnification on a Zeiss
Axioskop equipped with a digital camera. Pollen grains were identified using the
Florida Institute of Technology reference collection of modern pollen, and published

73
catalogs with photographs and morphological descriptions of pollen types
(Colinvaux et al. 1999, Hooghiemstra 1984, Roubik and Moreno 1991).
Samples for charcoal analysis were disaggregated in 10% KOH, and the
resulting slurry washed through a 170 µm sieve (particles >170 µm were retained).
The material in the sieve was transferred to a Petri dish and charcoal counts
performed under an Olympus dissection microscope (20x magnification) equipped
with a video camera (Clark and Patterson 1997). Charcoal particles were manually
identified and then digital measurements were made via the video system and image
recognition (NIH-IMAGE) software. This software provides the area of charcoal
particles according to the number of pixels occupied by the fragment on the screen.
The pollen sum, percentage, and concentration were all calculated in TILIA (Grimm
1992), and plotted in C2 1.3 (Juggins 2003) and Corel Draw 8.
Detrended Correspondence Analysis – DCA (Hill and Gauch 1980) was
performed using PC-ORD 4.0 for Windows (McCune and Mefford 1999). This
technique was chosen because of 2 main reasons: The first one is that it prevents the
problems caused by the arch effect, which are a constant on other techniques such as
Principal Component analysis (PCA) and Correspondence Analysis (CA) (McCune
and Grace 2002). The second reason is that DCA allows the axes to be re-scaled and
shown as Standard Deviation units, which can be used to estimate the degree of
species turnover (McCune and Grace 2002). As very complex data sets with many
rare taxa can hinder DCA, the data set was reduced to include only pollen taxa with
percentage values > 1% and present at least in 3 samples throughout the sediment

74
core. The resulting matrices from Lakes Santa Maria, Geral and Saracuri had 43, 30
and 42 pollen types respectively. The percentage values were then transformed using
square root transformation, as this technique is less drastic than log transformation
(Kovach 1992). The resulting sample scores from DCA were plotted against sample
depths for axes 1 and 2 in EXCEL.

75
RESULTS
STRATIGRAPHY
The 856 cm-long sediment core from Lake Santa Maria (Fig 3.2) is mostly
composed of clays in the basal 156 cm, overlain by 700 cm of gray gyttja (Table
3.1). The sediment core from Lake Geral is 570 cm long (Fig 3.2), with basal sands
overlain by light gray clay and gyttja (Table 3.2). The 886 cm-long sediment core
from Lake Saracuri (Fig 3.2) is mostly composed of clays and gyttja (Table 3.3).
Table 3. 1. Sediment description of core from Lake Santa Maria (Pará - Brazil).
Depth (cm) Sediment description
700 - 0 Gray gyttja rich in charcoal and plant fragments
800 - 700 Green-yellowish clay
856 - 800 Gray clay with fine sands
Table 3. 2. Sediment description of core from Lake Geral (Pará – Brazil).
Depth (cm) Sediment description
400 - 0 Gray gyttja
450 - 400 Dark clay rich in wood fragments
560 - 450 Light-gray clay with wood fragments
570 - 560 Sand

76
Lak e Ger a
l
Depth (cm)
0
50
100
150
200
250
300
350
400
450
500
550
600
Light gray clay with wood remains
Gray gyttja
Dark clay rich in wood remains
Sand
6570 cal years BP
8260 cal years BP
Lake Santa Maria
Depth (cm)
Gray gyttja with charcoal
and plant fragm ents
Green-yellowish clay
Gray c lay with sand
3120 cal years BP
3710 cal years BP
5390 cal years BP
7060 cal years BP
7630 cal years BP
7620 cal years BP
0
50
100
150
200
250
300
350
400
450
500
550
600
650
700
750
800
850
900
0
50
100
150
200
250
300
350
400
450
500
550
600
650
700
750
800
850
900
Lake Saracur
i
Depth (cm)
Dark br own gyttja
Gray gyttja with clay
Dark-gray clay
Black clay
700 cal years BP
4170 cal years BP
4620 cal years BP
4960cal years BP
6920 cal years BP
8480 cal years BP
Figure 3. 2. Lithology of sediment cores from Lakes Santa Maria, Geral (modified
from Bush et al. 2000), and Saracuri, Monte Alegre-Prainha region, Pará
(Brazil). Also showing location of dated samples in calibrated years
Before Present (BP). Ages inside ellipse were rejected.

77
Table 3. 3. Sediment description of core from Lake Saracuri (Pará – Brazil).
Depth (cm) Sediment description
100 - 0 Dark-brown gyttja
300 - 100 Gray gyttja with clay
735 - 300 Dark-gray clay
800 - 735 Gray clay with fine sands
840 - 800 Dark-gray clay with a few pieces of charcoal
860 - 840 Black clay with a few pieces of charcoal
886 - 860 Sand
RADIOCARBON DATES
Sedimentary chronologies for Lakes Santa Maria, Geral and Saracuri are
derived from 14 AMS radiocarbon dates (Tables 3.4, 3.5 and 3.6). For each
calibrated age a maximum probability solution (Calib 4.3 type 2 solution) was
adopted. All ages used henceforth will be interpolated calibrated years before present
(BP), unless noted otherwise.
The basal ages of 7620 cal yrs BP, 8260 cal years BP and 8480 cal yrs BP for
Lakes Santa Maria, Geral and Saracuri respectively, indicate that the lakes formed at
approximately the same time and that their sediments provide Mid- and Late-
Holocene histories. However, the age vs. depth relationships (Fig 3.3) show that
sedimentation rates in the three lakes have changed through time. All three lakes
reveal fairly constant rates of sedimentation, although Geral accumulated sediments
the fastest and Saracuri the slowest.

78
0
50
100
150
200
250
300
350
400
450
500
550
600
650
700
750
800
850
900
0 1 00 0 2000 3000 4000 5000 600 0 7000 8000 9000
Age (
14
C ye ars BP)
Depth (cm)
Santa Mar ia
Geral
Saracu ri
Figure 3. 3. Radiocarbon ages from Lakes Santa Maria (diamonds), Geral (squares),
and Saracuri (triangles). The two radiocarbon dates represented by an
open diamond and an open triangle were rejected as outliers.
The two radiocarbon dates from the bottom of Lake Santa Maria core, 806
cm and 570 cm, provided very similar ages (7620 cal years BP and 7630 cal years
BP). The bottommost age will be rejected for several reasons. First, the sample at
806 cm is not only in a different sedimentary unit, but also in a sandy layer,
suggesting very high sedimentation rates and probably erosion. Second, this age was
obtained from bulk sediments, while the date from the sediment immediately
overlying the sand was from a piece of wood, which is more reliable. Finally, the
sedimentation rates, calculated for the sections 570 cm – 528 cm, and 528 cm – 407
cm are exactly the same, 0.07 cm yr -1, suggesting a more or less uniform sediment
deposition, which is more plausible as the lithology appears unchanged. Assuming

79
this constant rate of deposition, the basal age of the gray gyttja would be ca. 9400 cal
years BP. However, as no reliable date is available for the lowest section of the core,
an age of lake formation similar to that of the overlapping ages of the neighboring
lakes is considered probable, e.g. ca. 8300 cal years BP. Tentative ages for the
sediments occurring between 700 cm and 570 cm are based on an assumed age of
8300 years BP for the beginning of lake sediment deposition at 700 cm.
The chronology of Lake Geral is dependent upon only two radiocarbon dates
(Table 3.5), making estimates of sedimentation rates somewhat tenuous.
Nevertheless, sedimentation rates decrease from 0.10 cm yr -1 between 546 cm and
381 cm depth, to 0.06 cm yr -1 between 381 cm and 0 cm depth.
Table 3. 4. Radiocarbon dates from Lake Santa Maria sediment core.
Sample Depth (cm) 14C yr BP 13C/12C ratio Age (cal years BP)
OS-24122 172 2960 + 45 -26.90 2967 – 3264
OS-24123 240 3450 + 35 -28.80 3633 – 3779
OS-24124 407 4660 + 40 -28.80 5307 – 5472
OS-24125 528 6180 + 40 -28.00 6948 – 7164
OS-24126 570 6770 + 45 -27.10 7568 – 7680
* CURL-5386 806 6740 + 45 -26.50 7563 – 7673
* date rejected as outlier
In Lake Saracuri (Table 3.6), depth-age relationship appears to be almost
linear (Fig 3.3), suggesting relatively constant sedimentation rates. As the age
yielded by the sample from 240 cm (4620 cal years BP) deviates markedly from the

80
general pattern, this date will be rejected. Sedimentation rates display minor
fluctuations around 0.08 – 0.13 cm yr-1, with the lowest values (0.08 cm yr -1)
occurring between 355 cm and 70 cm depth.
Table 3. 5. Radiocarbon dates from Lake Geral sediment core (Bush et al. 2000).
Sample Depth (cm) 14C yr BP 13C/12C ratio Age (cal years BP)
Β-41654 3.75-3.87 5820 + 90 -28.90 6440 – 6700
Β-39702 5.42-5.51 7550 + 100 -28.20 8140 – 8370
Table 3. 6. Radiocarbon dates from Lake Saracuri sediment core.
Sample Depth (cm) 14C yr BP 13C/12C ratio Age (cal years BP)
OS-38383 70 785 + 25 -28.50 668 – 733
*OS-38384 263 4130 + 30 -24.86 4567 – 4661
OS-38385 355 3780 + 30 -28.60 4083 – 4243
OS-38386 462 4390 + 35 -29.34 4859 – 5047
OS-38387 666 5980 + 45 -27.40 6909 – 6925
CURL-5385 859 7690 + 75 -28.70 8355 – 8602
* date rejected as outlier
PALEOECOLOGICAL RECORD – LAKE SANTA MARIA
The pollen and charcoal records from Lake Santa Maria (Figs 3.4 and 3.5)
allow the distinction of 5 zones: SM1, SM2, SM3, SM4 and SM5. Samples from
below 650 cm depth contained no pollen.

81
010
Alchornea
010
Alibertia
010
Anacardiaceae
010
Anthurium
010
Apeiba
010
Arecaceae
01020
Asteraceae
010
Bignoniaceae
010
Caesalpiniaceae
010
Cassia
0102030
Cecropia
010
C
eltis
010
Cor
dia
010
C
roton
010
D
esmodium
010
D
ioclea
010
Euphorbiaceae
010
H
ymenaea
01020
H
yptis
010
Ludwigia
010
Mabea
010
Machaerium
010
Macrolobium
050100150
Mauritia
010
Mel/Comb
Depth (cm)
010
Mor/Urt2
010
Mor/Urt3
010
Mor/Urt4
010
Papillionaceae
0 204060
Poaceae
010
Polygala
010
Polygonum
010
Protium
010
Rubiaceae
010
Sapindaceae
010
Sapium
010
Solanaceae
010
Spermacoce
010
Symplocos
0204060
Cyperaceae
010
Pontederia
010
Sagittaria
0204060
Spores
0 50 100 150 2 00
Total concentration
Cal years BP
x 1000
x 1000
3120
3710
5390
7060
7630
0
50
100
150
200
250
300
350
400
450
500
550
600
650
700
750
800
850
900
Depth (cm)
Cal years BP
312 0
371 0
539 0
706 0
763 0
0
50
100
150
200
250
300
350
400
450
500
550
600
650
700
750
800
850
900
SM1
SM2
SM3
SM4
SM5
SM1
SM2
SM3
SM4
SM5
Figure 3. 4. Pollen concentration diagram of Lake Santa Maria, Pará (Brazil),
showing selected taxa (scale is x 103 grains/cm3 of sediment), lithology,
and radiocarbon dates (in calibrated years BP). The hollow curves are
exaggerated 5 times.

82
Depth (cm)
0
50
100
150
200
250
300
350
400
450
500
550
600
650
030
Mor/Urt2
020
Mor/Urt3
010
Mor/Urt4
010
Papillionaceae
020406080100
Poaceae
010
Polygala
020
Polygonum
010
Protium
010
Rubiaceae
020
Sapindaceae
010
Sapium
010
Solanaceae
010
Spermacoce
010
Symplocos
020406080100
Cyperaceae
010
Pontederia
030
Sagittaria
020406080100
Spores
030
Char
coal
SM1
SM2
SM3
SM4
SM5
0
50
100
150
200
250
300
350
400
450
500
550
600
650
010
Alchornea
010
Alibertia
010
Anacardiaceae
010
Anthurium
030
Apeiba
010
Arecaceae
030
Asteraceae
010
Bignoniaceae
010
Caesalpiniaceae
010
Cassia
0204060
Cecr
opia
010
Celtis
020
Cordia
010
Croton
010
Desmodium
010
D
ioclea
010
Euphorbiaceae
010
Hymenaea
02040
Hyptis
010
Ludwigia
010
Mabea
010
Machaerium
010
Macrolobium
020406080100
Mauritia
Depth (cm)
010
Mel/Comb
SM1
SM2
SM 3
SM 4
SM5
Figure 3. 5. Pollen percentage diagram of Lake Santa Maria, Pará (Brazil), showing
the most representative taxa and area of charcoal particles (mm2/cm3 of
sediment). Hollow curves are exaggerated 5 times.

83
SM1 (650 cm – 594 cm; ca. 8300 yrs BP – 7800 yrs BP). Sediments are
composed of gray gyttja with charcoal and plant fragments. Pollen concentration
increases from ca. 10,000 to ca. 80,000 grains/cm3 of sediment upward (Fig 3.4),
with as much as ca. 52,000 grains/cm3 coming from Poaceae (87%). Triporate and 4-
porate pollen grains of Moraceae/Urticaceae (Mor/Urt3 and Mor/Urt4 respectively)
are present at frequencies of 20% and 9%, respectively (Fig 3.5). Croton pollen is
only found in this zone (ca. 4.5%), and Desmodium pollen display its highest
frequency within the core (ca. 6%). The representation of Cyperaceae increases
reaching ca. 25% of total pollen and 19,000 grains/cm3 of sediment. Cecropia pollen
is also abundant near the top of this zone. The remaining pollen taxa are relatively
scarce. Charcoal accumulation is relatively high, and fluctuates between 2-9
mm2/cm3 of sediment (Fig 3.5).
SM2 (594 cm – 440 cm; 7800 yrs BP – 5800 yrs BP). Sediments are
indistinguishable from those of SM1. Pollen concentrations peak between 574 cm
and 554 cm, and at 460 cm reaching ca. 135,000 and 140,000 grains/cm3,
respectively (Fig 3.4). This zone is characterized by extremely high concentration
and percentage values of Mauritia pollen (124,000 grains/cm3 and 89%,
respectively). Percentage representation and concentrations of Poaceae and
Cyperaceae pollen decrease, while Alchornea and Cecropia increase. However,
Poaceae and Cyperaceae pollen increase in abundance again at the end of this zone.
Although percentages of Mor/Urt3 and Mor/Urt4 pollen display lower values than in
SM1, their pollen concentration increases. Pollen grains of Hyptis, Ludwigia,

84
Macrolobium and Polygonum are recorded for the first time and are present at low
percentages. Charcoal particles are less abundant with the exception of samples 554
cm and 544 cm that had peaks of ca. 18 and 7 mm2/cm3, respectively.
SM3 (440 cm – 297 cm; 5800 yrs BP – 4290 yrs BP). Sediments are gray
gyttja with charcoal and plant fragments. Pollen concentration is very low and
fluctuates between 15,000 and 38,000 grains/cm3. Percentages of Cecropia pollen
increase up to ca. 29%, while concentration values decrease (ca. 5000 grains/cm3).
Pollen of Hyptis and Sagittaria are abundant in the middle of this zone, whereas that
of Poaceae oscillates around 50%. Percentage and concentration of Cyperaceae and
Mauritia decrease and reach their lowest values of the core. Mor/Urt2 and Mor/Urt3
pollen are more abundant (ca. 10-15%) than in SM1 and SM2. Charcoal is especially
abundant in this zone with a peak at 420 cm of ca. 25 mm2/cm3.
SM4 (297 cm – 180 cm; 4290 yrs BP – 3160 yrs BP). Sediments are
composed of gray gyttja with charcoal and plant fragments. Pollen concentrations
fluctuate around ca. 30,000 grains/cm3 increasing upward. Cecropia shows its
highest percentage values (57.5%) and second highest concentration (ca. 20,000
grains/cm3 of sediment) at 260 cm depth. Even though percentage and concentration
values of Poaceae are generally low (<20%), a single sample at 240 cm contained
51%. Concentrations and percentages of Mor/Urt3 and Mor/Urt4 pollen continue to
increase. Pollen of Apeiba is singularly abundant at 280 cm depth (29.2%), followed
by high percentages of Alibertia, Celtis, Cordia and other forest taxa. Cyperaceae

85
percentages increase upward, and charcoal accumulation is greatly diminished with
the exception of a peak at 280 cm (ca. 10 mm2/cm3).
SM5 (180 cm – 0 cm; 3160 yrs BP until present). Sediments are gray gyttja
with charcoal and plant fragments. Pollen concentrations are consistently high,
peaking at ca. 163,000 grains/cm3. Percentage and concentration values of Poaceae
pollen are around 45% and 20,000 grains/cm3, respectively. Abundance of
Cyperaceae pollen increases markedly with peaks of 55% and 80% in this zone.
Pollen percentages of Asteraceae, Polygonum, Rubiaceae, Sapindaceae, Spermacoce,
Symplocos and Pontederia increased. However, percentages of Cecropia, Mor/Urt2,
Mor/Urt3, Mor/Urt4, and Polygala pollen decreased. Spores have their only major
peak abundance peak of 95%. Charcoal particles are scarce, except for peaks at 150
cm (ca. 6 mm2/cm3) and 3 cm (ca. 9 mm2/cm3).
PALEOECOLOGICAL RECORD – LAKE GERAL
The pollen record from Lake Geral allowed the recognition of 4 main zones
(Fig 3.6). The accumulation of charcoal was recorded as concentration of
particles/cm3 of sediment (Bush et al. 2000).
G1 (596 cm – 520 cm; 8260 yrs BP – 8000 yrs BP). This zone is composed
of 3 sediment units starting as sands, changing to gray gyttja and changing again to
light-gray clay. Pollen concentrations fluctuate around 50,000 grains/cm3 (Bush et al.
2000). This zone is characterized by moderate to high percentage values of forest
taxa and Poaceae pollen. Sapotaceae pollen abundance peaks at 545 cm depth

86
(19.8%), and Symphonia pollen increases from ca. 7% to ca. 29% at the top of this
zone. Charcoal particles are effectively absent (Bush et al. 2000).
G2 (520 cm – 390 cm; 8000 yrs BP – 6700 yrs BP). Three sedimentary units
are found in this zone, e.g. light-gray clay, dark clay, and gray gyttja (from the
bottom to the top). Pollen concentrations display minor variations around 50,000
grains/cm3 (Bush et al. 2000), and wet forest taxa (e.g. Amanoa, Symphonia, and
Mauritia) dominate the zone. Percentages of other taxa display minor fluctuations.
Accumulation of charcoal particles is negligible (Bush et al. 2000).
G3 (390 cm – 130 cm; 6700 yrs BP – 2390 yrs BP). Sediments are composed
of gray gyttja, and pollen concentrations are slightly higher (ca. 52,000 grains/cm3).
Wet forest taxa other than Mauritia decline to background levels as Cecropia,
Poaceae, Cyperaceae and Mauritia dominate the spectrum. Byrsonima occurs
regularly in this zone. Pollen grains of Zea mays (maize) are present from 250 cm
until ca. 40 cm, and are coincident with the increased charcoal accumulation (Bush et
al. 2000).
G4 (130 cm – 0 cm; 2390 yrs BP until present). Sediments deposited are
composed of gray gyttja, and changes in pollen concentration are insignificant.
Cecropia and Mabea pollen have percentage peaks at the bottom of the zone, and
decline thereafter. Percentages of Mauritia, Poaceae, and Cyperaceae diminish
upward, though there is an isolated peak in Mauritia percentage at 1 cm depth.
Pollen of Euterpe, Mel/Comb, Mor/Urt3, and Myrtaceae display increased

87
abundances. Charcoal accumulation though still high, presents a declining pattern
upward (Bush et al. 2000).
01001020010010010203001001001001001020010203040500102030010
Depth (cm)
Cal years BP
0
50
100
150
200
250
300
350
400
450
500
550
600
6570
8260
G1
G2
G3
G4
010203040500100100100102001001001020300100102030405060010010010010010010203040010
Depth (cm)
Cal years BP
0
50
100
150
200
250
300
350
400
450
500
550
600
6570
8260
G1
G2
G3
G4
Other Forest taxa
Acanthaceae
Mor/Urt2
Anacardiaceae
Mor/Urt4
Arecaceae
Poaceae
Asteraceae
Proteaceae
Byrsonima
Rubiaceae
Am
anoa
Sapindaceae
Cassia
Sapium
Cecropia
Sapotaceae
Didymopanax
Myrtaceae
Euterpe
Symphonia
Fabaceae
Cyperaceae
Ilex
Spores
Mabea
Zea
Mauritia
Mel/Comb
Alchornea
Mor/Urt3
Figure 3. 6. Pollen percentage diagram of Lake Geral, Pará (Brazil), showing the
most representative taxa, lithology, and radiocarbon dates (calibrated
years BP). Hollow curves are exaggerated 5 times, and black dots
identify presence of Zea (modified from Bush et al. 2000).

88
PALEOECOLOGICAL RECORD – LAKE SARACURI
The pollen zones S1, S2, S3, and S4 are documented at Lake Saracuri (Figs
3.7 and 3.8).
S1 (860 cm – 740 cm; 8480 yrs BP – 7490 yrs BP). Lithology is composed of
3 different sedimentary units from bottom to top: black clay, a layer of dark-gray
clay, sandy dark clay, and dark-gray clay. Pollen concentration, primarily Mauritia,
peaks in the basal clay band (at 860 cm) with 496,573 grains/cm3 and then drops
rapidly to ca. 24,000 grains/cm3 (Fig 3.7). The percentage of Mauritia peaks at 860
cm followed by a peak of Cecropia at 840 cm. Forest taxa predominate, although the
maximum abundance of charcoal particles occurs in this zone (ca. 8.5 mm2/cm3) (Fig
3.8).
S2 (740 cm – 470 cm; 7490 yrs BP – 4960 yrs BP). Sediments are composed
of dark-gray clay, and pollen concentration increases fluctuating around 40,000
grains/cm3. This zone is differentiated from S2 by increased concentrations of
Cecropia, Mauritia, and Poaceae. The percentages of the remaining taxa display
minor fluctuations. Charcoal accumulation is diminished in the beginning of S2, but
increases upward.
S3 (470 cm – 190 cm; 4960 yrs BP – 2100 yrs BP). Sediments change from
dark-gray clay on the bottom to a layer of dark-gray gyttja with clay on the top.
Pollen concentration fluctuates around 35,000 grains/cm3, and the zone is
characterized by decreasing percentages and concentrations of Mauritia and
Poaceae. Concentration and percentage values of Cecropia pollen increase

89
throughout the zone, while most other taxa are essentially stable. Charcoal is present
throughout, but at relatively low concentrations.
x 100
Depth (cm)
Cal years BP
700
4170
4960
6920
8480
0
50
100
150
200
250
300
350
400
450
500
550
600
650
700
750
800
850
900
020
Utricula
ri
a
0204060
Spores
0500
1000
Total
Concentration
496,573
020
Tapirira
0204060
Virola
0204060
Zonorate
020406080
Cyperaceae
02040
Sapium
020
Solanaceae
0 102030
Spermacoce
020
Symphonia
12,01 3 25,362
020
Sapindaceae
6674
030
Rubiaceae
S1
S2
S3
S4
x 100
e
e
x 100
Depth (cm)
Cal years BP
700
4170
4960
6920
8480
0
50
100
150
200
250
300
350
400
450
500
550
600
650
700
750
800
850
900
02040608
0
Poaceae
11,954
22 , 69 2
030
Mel/Comb
020
Mor/Urt2
01020304050
Mor/Urt3
010
Mor/Urt4
01020304050
Myrtaceae
0102030
Pa
pi
llionacea
020
Phyllanthus
020406080100
Euphorbiacea
010
Euterpe
020
Hymenaea
020
Hyptis
020
Machaerium
020
Macrolobium
0100
Mauritia
228,26 4 9344
020
Polygala
020
Protium
53 39
S1
S2
S3
S4
020
Celtis
010
Astrocarium
010
A
stronium
020
B
ignoniaceae
020
Byrsonima
020406080100
Caesalpiniaceae
020406080100
Cassia
020
Acalypha
020406080
A
lchornea
020
Anacardiaceae
020
Arace
ae
020406080
Asteraceae
25, 362
02
04
060 80
1
00 1201401
60 1
80 2
00
Cecropia
27 , 7 0 0
Dep th (c m)
Cal years BP
700
4170
4960
6920
8480
0
50
100
150
200
250
300
350
400
450
500
550
600
650
700
750
800
850
900
S1
S2
S3
S4
Figure 3. 7. Pollen concentration diagram of Lake Saracuri, Pará (Brazil), showing
selected taxa (scale is x 103 grains/cm3 of sediment), lithology, and
radiocarbon dates (calibrated years BP). The hollow curves are
exaggerated 5 times.

90
0
50
100
150
200
250
300
350
400
450
500
550
600
650
700
750
800
850
900
010
Acalypha
010
Alchornea
010
Anacardiaceae
010
Araceae
010
Asteraceae
010
Astrocari
um
010
Astronium
010
Bignoniaceae
010
Byrsonima
010
Caesalpiniaceae
010
Cassia
0204060
Cecropia
010
Celtis
0102030
Euphorbiaceae
010
Euterpe
0102030
Poaceae
010
Hymenaea
010
Hyptis
010
Machaerium
010
Macrolobium
01020304050
Mauritia
Depth (cm)
S1
S2
S3
S4
0
50
100
150
200
250
300
350
400
450
500
550
600
650
700
750
800
850
900
Depth (cm)
010
Mel/Comb
010
Mor/Urt2
010
Mor/Urt3
010
Mor/Urt4
010
Myrtaceae
010
Papillionaceae
010
Phyllanthus
010
Polygala
010
Protium
010
Rubiaceae
010
Sapindaceae
010
Sapium
010
Solanaceae
010
Spermacoce
010
Symphonia
010
Tapirira
010
Virola
010
Zonorate
01020
C
yperaceae
010
Utricularia
010
Spores
Charcoal
S1
S2
S3
S4
0510
Figure 3. 8. Pollen percentage diagram of Lake Saracuri, Pará (Brazil), showing the
most representative taxa and area of charcoal particles (mm2/cm3 of
sediment). Hollow curves are exaggerated 5 times).
S4 (190 cm – 0 cm; 2100 yrs BP until present). Lithology is composed of 2
sedimentary units: gray gyttja with clay on the bottom, and dark-brown gyttja on the
top. Pollen concentrations decline from ca. 70,000 grains/cm3 to 25,000 grains/cm3

91
in this zone. As Cecropia pollen declines in abundance from a peak concentration at
the lower boundary of the zone, pollen of a broad array of forest taxa become more
abundant. Charcoal values increase to a peak of ca. 4 mm2/cm3 at 50 cm, and then
decline to background levels in the uppermost samples.
MULTIVARIATE ANALYSIS – LAKE SANTA MARIA
The resulting DCA scores were plotted in two different ways. First Axis 1 vs.
Axis 2, then sample scores from both axes were plotted with corresponding depths
and ages in cal years BP.
AXIS 1 vs. AXIS 2. Grouped samples that represent interpreted zones are
evenly spread over axes 1 and 2 (Fig 3.9). Axis 1 polarizes the samples of SM2 and
SM5 with the topmost samples of the core having the lowest scores, and the basal
samples the highest scores. The species that characterize the positive extreme of Axis
1 were Mauritia, Desmodium, Hymenaea, Croton, and Protium. The scores
characterizing samples at the negative extreme of the Axis were yielded by Alibertia,
Sapindaceae, Spermacoce, Sapium, and Dioclea.
Axis 2 polarizes samples from zones SM1 and SM4. Samples that represent
SM4 were placed on the positive side of Axis 2, and samples that characterize SM1
were placed on the negative side. The highest scores were yielded by Cordia,
Apeiba, Symplocos, Alibertia, and Hymenaea, and the lowest scores were yielded by
Poaceae, Desmodium, Asteraceae, Dioclea, and Croton.

92
AXIS 1 vs. DEPTH. The DCA sample scores plotted against depth show a
pattern that suggests a relatively uniform transition among zones (Fig 3.10). The
bottommost samples (SM1) are placed in the middle of the range of Axis 1, while the
above grouped samples (SM2, SM3, SM4, and SM5) are distributed from right to
left upward.
AXIS 2 vs. DEPTH. After excluding the variance accounted for in Axis 1,
SM4 stands out as being different from the other zones (Fig 3.10).
0
0
80
20
60
80
SM-180
SM-198
SM-220SM-2 9 7
SM-240
SM-120
SM-80 SM-3 40
SM-140
SM-1 6 0
SM- 10 0
SM-60
SM-40
SM-20
SM-615
SM-6 30
SM-650
SM-594
SM-400
SM-320 SM-44 0 SM-460
SM-520
SM-360
SM-260
SM-260
Axis 2
Axis 1
SM-574
SM-500
SM-554
SM-480
SM-534
SM-380
SM-3
40
40
SM1
SM2
SM3
SM4
SM5
Figure 3. 9. DCA scores from Lake Santa Maria showing Axis 1 vs. Axis 2. Samples
are grouped according to interpreted phases. Axis 1 polarizes samples
into 3 groups that represent phases SM1, SM3 and SM4, phase SM2, and
phase SM5. Axis 2 separates phases SM1 from SM3 and SM4.

93
SM1
SM2
SM3
SM4
SM5
Mauritia
Desmodium
Hymenaea
Croton
Protium
Alibertia
Spermacoce
Sapium
Dioclea
Sapindaceae
Cordia
Apeiba
Symplocos
Alibertia
Hymenaea
Poaceae
Asteraceae
Desmodium
Dioclea
Croton
SM1
SM2
SM3
SM4
SM5
7630
7060
5390
3120
3710
7630
7060
5390
3120
3710
0
50
100
150
200
250
300
350
400
450
500
550
600
650
-20 0 20 40 60 80 100 120 140 160 180 200
Dca scores - axis 1
Depth (cm)
0
50
100
150
200
250
300
350
400
450
500
550
600
650
-20 0 20 40 60 80 100 120 140 160
Dca scores - axis 2
Depth (cm)
Figure 3. 10. DCA scores from Lake Santa Maria, axes 1 and 2 plotted vs. depth
(cm). Also showing dates in cal years BP, the 5 most important species
associated with each axis of DCA, and interpreted zones.

94
MULTIVARIATE ANALYSIS – LAKE GERAL
The DCA sample scores from Lake Geral were also plotted in two different
ways. First, Axis 1 vs. Axis 2 (Fig 3.11), then the sample scores from Axis 1 were
plotted with the corresponding depths (Fig 3.12).
0
0
40 8 0
20
40
60
80
Axis 2
Ax is 1
G-4 10
G-555
G-390
G-5 20
G-4 35
G-5 0
G-3 0
G-265
G-2 00
G-3 20
G-3 50
G- 1 1 0
G-1 70
G-3 80
G-3 05
G-4 0
G- 1 9 0
G-2 80
G-150
G-70
G-2 0
G-1
G-6 0
G-2 20
G- 8 0
G-1 0
G-427 G-4 65
G-4 81
G-5 05
G-403
G-5 45
G-535
G-596
G-130
G-3 35
G1
G2
G3 and G4
Figure 3. 11. DCA scores from Lake Geral showing Axis 1 vs. Axis 2. Samples are
grouped according to phases interpreted. Axis 1 polarizes samples into
2 main groups that represent phases G1 and G2 (bottom of the core),
and G3 and G4 (top of the core). Axis 2 polarizes bottom samples into 2
groups that represent phases G1 and G2. Samples G-410 and G-390
appear to be outliers. Distinction between phases G3 and G4 is nearly
impossible.
AXIS 1 vs. AXIS 2. The DCA scores display a strong polarization of samples
on Axis 1, dividing them into two main groups: the bottom samples of the core that

95
represent zones G1 and G2 are on the positive side of the axis, and the top samples
of the core on the left side representing zones G3 and G4 (Fig 3.11). The highest
scores for Axis 1 were yielded by Didymopanax, Sapotaceae, Symphonia, Sapium,
and Proteaceae, placing the bottom samples on the positive end of the axis. Mabea,
Byrsonima, Mor/Urt2, Cecropia, and Cyperaceae scored lowest, bringing top
samples to the negative end of Axis 1.
Even though polarization of samples on Axis 2 is relatively weaker than on
Axis 1, samples from G1 are separated from samples from G2. Because Axis 2 did
not allow a further separation of samples into zones, only the first axis of the
ordination was plotted against depths.
6570
8260 G1
G2
G3
G4
Didymopanax
Symphonia
Sapium
Sapotaceae
Proteaceae
Mabea
Byrsonima
Cecropia
Mor/Urt2
Cyperaceae
0
50
100
150
200
250
300
350
400
450
500
550
600
-40 -20 0 20 4 0 60 80 100 120 1 40
Dca scores (axis 1)
Depth (cm)
Figure 3. 12. DCA scores from Lake Geral, Axis 1 vs. depth (cm). Also showing
dates in cal years BP, the 5 most important species associated with each
end of the axis of DCA, and interpreted zones.

96
AXIS 1 vs. DEPTH. Plotting Axis 1 DCA sample scores against sample
depth reveals a striking pattern that suggests an abrupt transition from zone G2 to G3
at 390 cm (6700 cal years BP – interpolated) (Fig 3.12). Little sample variation is
observed within zones.
MULTIVARIATE ANALYSIS – LAKE SARACURI
The DCA scores from Lake Saracuri were plotted in two different ways: Axis
1 vs. Axis 2, and sample scores from both axes vs. depth.
AXIS 1 vs. AXIS 2. Sample groups that characterize zones are evenly
distributed along the two axes (Fig 3.13). Axis 1 presents a strong polarization of
samples separating S1 (on the right) from S4 (on the left), while Axis 2 places
samples from S2 on the bottom, and samples from S3 on the top. The highest scores
on Axis 1 were yielded by Mor/Urt2, Zonorate, Araceae, Solanaceae, and
Papillionaceae, while the lowest scores were yielded by Astrocarium, Sapium,
Phyllanthus, Bignoniaceae, and Astronium. On Axis 2 Hyptis, Araceae,
Bignoniaceae, spores, and Hymenaea yielded the lowest values, which brought S2
samples to the negative end of the axis, while the highest scores were yielded by
Utricularia, Symphonia, Celtis, Macrolobium, and Acalypha.
AXIS 1 vs. DEPTH. The DCA sample scores plotted against depth show a
pattern that is similar to the one displayed by Axis 1 from Lake Santa Maria (Fig
3.10), but at Lake Saracuri the range of change is much smaller and the rate of
change also appears to be slower (Fig 3.14). At the beginning of zone S4 samples

97
reach the lowest scores, however above that, samples are placed in the middle of the
axis.
0
0
40 80
20
40
60
Axis 2
Axis 1
S80 0
S7 60
S690
S610
S570 S6 70
S590 S72 0
S4 70
S710
S630
S530
S550
S5 10
S1
S78
S2 0
S40 S60
S1 80
S1 60
S1 20
S1 40
S100
S270S230
S410
S250
S430
S210
S390
S370
S350
S190
S2 90
S330
S490
S650 S840
S7 40 S780 S8 20
S450
S8 60
S310
S1
S2
S3
S4
Figure 3. 13. DCA scores from Lake Saracuri showing Axis 1 vs. Axis 2. Samples
are grouped according to phases interpreted. Axis 1 displays a sample
distribution that is relatively coincident with phase depths. Bottom
samples (S1) yielded high scores (right), top samples (S4) yielded low
scores (left), and intermediate samples were placed in the middle. Axis
2 polarizes samples from phases S2 (bottom) and S3 (top).

98
700
4170
4960
6920
8480 S1
S2
S3
S4
700
4170
4960
6920
8480 S1
S2
S3
S4
Mor/Urt2
Zonorate
Aracea e
Solana ceae
Papillionaceae
Utricu laria
Symphonia
Celtis
Macrolobium
Acalypha
Astrocarium
Sapium
Phyllanthus
Astronium
Bignoniaceae
Hyptis
Hymenaea
Araceae
Bignoniaceae
Spores
0
50
100
150
200
250
300
350
400
450
500
550
600
650
700
750
800
850
900
-60-50-40-30-20-10 0 1020304050607080
DCA scores - axis 1
Depth (cm)
0
50
100
150
200
250
300
350
400
450
500
550
600
650
700
750
800
850
900
-40 -30 -20 -10 0 10 20 30 40 50 60
DCA scores - axis 2
Depth (cm)
Figure 3. 14. DCA scores from Lake Saracuri, axes 1 and 2 vs. depth (cm). Also
showing dates in cal years BP, the 5 most important species associated
with each axis of DCA, and interpreted zones.

99
AXIS 2 vs. DEPTH. The DCA scores from Axis 2 when plotted against depth
also display a striking similarity to the second axis from Lake Santa Maria (Fig
3.10). As in Lake Santa Maria record, the samples from immediately below the
uppermost zone, i.e. S3 (SM4 in Santa Maria) are displaced towards the positive end
of Axis 2, while the remaining samples display smaller variation in the middle region
of the axis (Fig 3.14).

100
DISCUSSION
The three lakes, Santa Maria, Geral, and Saracuri, provide a complex yet
consistent record of Holocene environmental history for the Prainha region.
Although there is local variation in the fine detail of the records, broad
depositional trends are evident in the cores from Lakes Santa Maria, Geral, and
Saracuri (Fig 3.15).
0
500
100 0
150 0
200 0
250 0
300 0
350 0
400 0
450 0
500 0
550 0
600 0
650 0
700 0
750 0
800 0
850 0
900 0
Inter polated age (cal years BP)
S1
S2
S3
S4
Forest regrowth
Lake formation
Human impa ct
Savanna expansion
and huma n impact
Scrub savanna
Wetl and e xpans ion
SM3
SM 4
SM 5
SM1
SM2
G1
G2
G3
G4
Lake
Santa Maria Lake Geral Lake Saracuri
Figure 3. 15. Phases of environmental changes interpreted from the paleoecological
records of Lakes Santa Maria, Geral, and Saracuri. Also showing
chronology and pollen zones of each record. Human impacts are
inferred from high fire frequency and agriculture.

101
The zones SM1 (Lake Santa Maria), G1 (Lake Geral), and S1 (Lake Saracuri)
were inferred to reflect the formation of the lakes (Figs 3.4 - 3.8), which took place
relatively synchronously. The basal sands of Geral and Saracuri support this
interpretation. At Santa Maria sands were present below 800 cm, but the sediments
deposited between 800 cm and 650 cm did not preserve pollen. A probable
explanation is that although sediments were accreting, they were oxidized frequently,
thus SM1 represents the first formation of a permanent water body.
Ponding at Lakes Santa Maria, Geral and Saracuri was essentially
synchronous. The calibrated ages of 8260 yrs BP and 8480 yrs BP from Geral and
Saracuri, when expressed with 95% confidence intervals are 8140 – 8370 and 8355 –
8600, respectively, and provide ages that overlap at ca. 8300 cal years BP. The
timing of lake formation is coincident with the Holocene sea-level rise (Bezerra and
others 2003; Suguio and others 1985), which affected the hydrology of several
Amazonian rivers and lakes by rising local water tables (Behling 1996; Behling
2002; Behling and Costa 2001; Bush and others 2000).
The relatively high percentages of Poaceae at the beginning of the records
probably indicate the early stages of lake formation with the presence of grasses on
the lakeshore or as floating mats, as they are accompanied by Cyperaceae.
Additionally, the occurrence of taxa such as Alchornea, Anacardiaceae,
Caesalpiniaceae, Cassia, Croton, Desmodium, Didymopanax, Euphorbiaceae, Ilex,
Hymenaea, Melastomataceae/Combretaceae, Moraceae/Urticaceae, Sapotaceae,
Solanaceae, and Zonorate type, suggests the presence of dry forest and woody

102
swamp vegetation near the sites at that early stage (Marchant and others 2002).
However, at Lake Saracuri, patches of scrub savanna are also suggested by the
presence of Byrsonima (Behling and Hooghiemstra 1999).
The co-occurrence of wetland taxa simultaneous with charcoal particles at
Santa Maria and Saracuri suggests human impacts since the lake formation (Figs 3.5
and 3.8). The temporal discontinuity of fires among the three lakes is another
indication of the anthropogenic origin of these fires.
Following lake formation, wetter conditions between ca. 8000 years BP and
6500 years BP allowed an expansion of wetland taxa, e.g. mostly Mauritia at Santa
Maria (SM2) and Saracuri (S2), and Symphonia and Amanoa at Geral (G2). As
swamp forest elements increased in abundance, terra firme forest elements such as
Desmodium, Didymopanax, Moraceae/Urticaceae, Papillionaceae, Protium,
Solanaceae, and Zonorate type declined. It is notable that wet forest still occurs near
these sites, but is absent from the catchment of Lake Santa Maria. The wetter
conditions lasted 700 years longer (until ca. 5800 years BP) at Santa Maria perhaps
reflecting a less leaky basin, or a higher ratio of catchment to lake surface area.
Charcoal accumulation begins near the base of the core at Santa Maria and
Saracuri. The amount of charcoal deposited increases until 7340 years BP at Santa
Maria, and until 7490 years BP at Saracuri. However, no charcoal was recorded
during that time at Geral (Bush et al. 2000), suggesting that the fires were very local,
for the other lakes are only 4 km apart.

103
After ca. 6700 years BP, intensified land use, in addition to a savanna
expansion, are suggested by increased percentages of Cecropia, Poaceae, Byrsonima
and Asteraceae, and the marked decrease of other forest taxa at Saracuri and Geral.
At Santa Maria, the transition from a dry forest with Mauritia in the wetlands to
more open landscapes took place at 5800 BP. The increased abundances of Poaceae
pollen, but not Cyperaceae and charcoal are consistent with savanna formation at this
time. Furthermore, the marked increase of Sagittaria following the peak in charcoal
(at Santa Maria) is consistent with a growth response of aquatic plants to increased
nutrient input caused by erosion of minerals following fires. The small quantities of
charcoal particles accumulated at Santa Maria between 7340 and 5800 years BP, and
Saracuri between 7490 and 6260 years BP, contrast to a peak of charcoal at Geral at
6700 years BP (Bush et al. 2000), suggesting a shift in land use.
A period of forest re-growth between ca. 5000 and 3200 years BP at Saracuri,
and between ca. 4300 and 3200 years BP at Santa Maria, is suggested by the
increased abundances of Anacardiaceae, Apeiba, Astronium, Caesalpiniaceae,
Cassia, Cecropia, Euphorbiaceae, Macherium, and Macrolobium. This forest
recovery could be due to wetter conditions coupled with less anthropogenic impacts,
as suggested by low charcoal accumulation. The apparent contrast between SM4 and
S3, and the other pollen zones, is especially pronounced when the DCA scores from
Axis 2 of Santa Maria and Saracuri are plotted against depth (Figs 3.10 and 3.14).
Because the negative extreme of Axis 2 is dominated by savanna indicators (e.g.
Asteraceae, Desmodium, Hyptis, and Poaceae), and in the case of Saracuri, the

104
former dry forest indicator (Hymenaea), a period of forest re-growth under wetter
conditions is supported by the second axis of DCA.
At Geral, the constant presence of Zea (maize), a marked increase in charcoal
accumulation, as well as the saw-tooth pattern displayed by Cecropia, Mauritia, and
charcoal after ca. 3900 years BP suggest that disturbance was more intense and
frequent nearby this lake. That Geral appears to be the most impacted of the three
lakes, is probably the reason why a forest re-growth was not observed at this site.
The earliest establishment of a scrub savanna with somewhat less
anthropogenic impact is dated at ca. 3200 years BP at Santa Maria and Saracuri. The
reduced fire frequency, and increased abundances of Asteraceae, Byrsonima, Cassia,
Myrtaceae, Rubiaceae, and Sapium support this interpretation and also suggest the
presence of dry forest nearby. The presence of a few swamp taxa such as Ludwigia,
Macrolobium, Polygonum, Pontederia, and Virola implies that climate could be
relatively wet in this phase. At Geral, this vegetation change took place only after
2300 years BP.
The local human impacts associated with fire were diminished at Santa Maria
after ca. 4200 years BP, and even though fire frequency at Geral started to decrease
after ca. 2300 years BP, agriculture, as suggested by presence of Zea, lasted at least
until 800 years BP. At Lake Geral, some forest re-growth is observed following
reduced human activity, although Poaceae abundances are still higher than pre-
disturbance levels (Bush and others 2000).

105
Increased fire frequency accompanied by relatively high abundances of
Astrocaryum at Saracuri after 2100 years BP suggests another shift in land use
(Roosevelt and others 1996). Anthropogenic impacts dramatically decreased after ca.
400 years BP (~1550 years AD), which would be consistent with a sharp reduction in
indigenous populations coincident with the arrival of Europeans (Denevan 1978,
Roosevelt et al. 1991; Bush et al. 1992; Roosevelt 2000). The subsequent
establishment of a scrub savanna with less human impacts is observed. At Santa
Maria, the last peak in fire frequency (close to the present time) is probably related to
the modern land use around the lake, where small farms are scattered.
Even though the three paleoecological records revealed some site specific
variation, a similar pattern is apparent. This pattern of vegetation shift is well
illustrated by plotting the DCA scores from Axis 1 with depth, which displays a
general trend from a more shaded landscape that is consistent with the presence of
dry forest and swamp forest, to a more open and disturbed landscape, which is
consistent with a scrub savanna (Figs 3.10, 3.12, and 3.14). The rate at which the
transition took place varied from site to site, and is probably related to the intensity
of human impacts. The most drastic change is inferred to have occurred at Lake
Geral (Fig 3.12), where large quantities of charcoal particles suggest strong
anthropogenic impacts at the same time of the vegetation change. That Santa Maria
showed the most gradual transition (Fig 3.10) could be just a function of less intense
human use in addition to the lake being located closer to the forest edge.

106
The DCA Axis 1 is therefore interpreted as a vegetation openness gradient,
meaning more forest cover on the right hand side, and scrub savanna on the left. This
gradient is probably a function of climate change from wetter (right) to drier (left)
coupled with the effects of human impacts on the landscape (Figs 3.10, 3.12, and
3.14).
The paleoecological records from Lakes Santa Maria, Geral, and Saracuri
provided evidence of regional vegetation changes during the Holocene in eastern
Amazonia, with a savanna expansion between 8000 and 5000 years BP. Therefore,
hypothesis 1 was accepted.
H1: Amazonian savannas expanded between 8000 and 5000 years BP.
A short period of forest re-growth was recorded between ca. 5000 years BP
and 3200 years BP, and 4300 years BP and 3200 years BP at Saracuri and Santa
Maria, respectively. However savanna vegetation continued to expand after 3200
years BP. For that reason, hypothesis 2 was rejected.
H2: Forests expanded and savanna vegetation contracted in the last 4000 years.
The Paleoecological record of Prainha region is consistent with an overall
increase in savanna extent as human exploitation of the landscape intensified. The
apparent lack of sychroneity of vegetation change between the sites suggests a
positive correlation between savanna expansion and human activity. However,
whether the savannas near Prainha are the result of severe land-use alone, or if the
intensification of human activities was simply favored by climate change is unclear.
Therefore, more records of vegetation changes from ecotonal areas with Pre-

107
Columbian human disturbance are needed to test Prance and Schubart’s hypothesis
regarding the anthropogenic origin of white sand savannas (Prance and Schubart
1977).

108
CHAPTER IV
PALEOECOLOGICAL RECORD OF RORAIMA (BRAZIL)
INTRODUCTION
The State of Roraima forms the northernmost spur of eastern Amazonia.
Flanked to the north and east by the Venezuelan and Guyanan shields the
northernmost portion of Roraima falls within the rain shadow of these highlands.
Consequently, northern Roraima is an unusually dry portion of Amazonia that
supports extensive savannas. As precipitation increases southward, there is a
transition from savanna to forest with the modern ecotone lying at about 2o45’ N.
Although the earliest palynological work conducted in Amazonia was an analysis of
varzea lakes in Roraima, the Holocene history of the savannas has not been
documented.
The oldest Holocene sedimentary lake record for Roraima comes from Lake
Caracaranã (Simões Filho and others 1997), the largest lake in the region (3.8 km of
circumference). This core has a basal age of ca. 10,200 years before present (BP).
Changes in sedimentary units between ca. 10,200 years BP and 8170 years BP were
interpreted as dry events of short duration. Sediment cores from Lakes Galheiro,
Redondo, and Fazenda São Joaquim provided paleoecological records for central
Roraima (Absy 1979, Absy et al. 1997). However, the basal radiocarbon date for the
oldest of those lakes (Fazenda São Joaquim) yielded a calibrated age of ca. 3960

109
years BP, thus providing only a Late-Holocene record. Absy et al. (1997) concluded
that no significant vegetation change took place in the savannas, at least in the last
3960 years.
Isotopic analysis (δ13C) of soil organic matter from several soil profiles
within the savanna and the forest areas near Boa Vista provided the first data on
savanna-forest dynamics for Roraima ecotones (Desjardins and others 1997). In soil
profiles within the savanna, values of δ13C are low (between -24 and -22 ‰) below
ca. 230 cm depth, and reach modern values characteristic of open savannas (between
-16 and -14 ‰) above 40 cm depth. These data suggest a transition from C3
dominated (forest or woodland) to C4 dominated (grassland) communities.
Conversely, in soil profiles within the forest, values of δ13C (between -23 and -22
‰) are recorded between 50 and 200 cm depth, which are intermediate between pure
C3 and C4 values, suggesting a mixed community of trees and savanna grasses. Near
the top of the forest profile, the δ13C values decrease markedly (-28 ‰), indicating
the expansion of trees at the expense of grasses. The presence of charcoal derived
from forest taxa at the bottom of soil profiles within savanna vegetation near Boa
Vista suggests that ancient forests covered areas that support open savannas at the
present (Desjardins and others 1997). The oldest pieces of charcoal found in savanna
profiles were derived from Connarus sp. and Vitex sp., and yielded the ages of 7470
cal years BP and 8460 cal years BP, respectively (Desjardins and others 1997). Even
though these taxa can also be found in dry forests and woodland savannas, they may

110
provide an approximate age for the on set of fires, and a savanna expansion for the
region near Boa Vista.
Despite these various paleoecological records, no detailed fossil pollen
analyses spanning the majority of the Holocene have previously been reported. In
this study, the Holocene dynamics of the forest-savanna ecotone in Roraima are
presented from a sediment core obtained from Lake Jacaré.
STUDY SITE
Lake Jacaré (3°48’N, 59°44’W) is located in northern Roraima (Fig 4.1),
about 160 km north from the State Capital (Boa Vista). The lake lies ca. 90 m above
sea level. This lake is a small, eutrophic water-body with about 300 m x 150 m of
open water, and is about 4 m deep. The open water is surrounded by a ca. 15 m wide
fringing floating swamp. Lake Jacaré is located 7 km east from Lake Caracaranã, and
just 8 km west of the Brazilian-Guyanan border.
Even though Jacaré is located within the savanna, scattered patches of
seasonal semi-deciduous forest are also present in the region, as well as gallery
forests and swamps along the water drainages, locally called igarapés (Silva 1997).
The climate in this region is tropical semi-humid, and is strongly influenced
by the Intertropical Convergence Zone (ITCZ) and the trade winds coming from
northeast (Barbosa 1997). Mean annual temperatures are ca. 26 °C, and mean annual
precipitation is between 1100 and 1400 mm with a dry season of 5-6 months (IBGE
1990, Barbosa 1997).

111
N
Lake J acaré
Lake Caracara nã
Boa Vista
20 Km
Figure 4. 1. Map of South America showing Roraima. The inset is a false color
Landsat TM image showing bands 7, 4, and 2. Pink represents highly
reflective surfaces such as savannas, urban areas and bare ground,
whereas green represents the more complex structure of forests.

112
HYPOTHESES
H1: Amazonian savannas expanded between 8000 and 5000 years BP.
H0: No savanna expansion was recorded between 8000 and 5000 years BP.
H2: Forests expanded and savanna vegetation contracted in the last 4000 years.
H0: No forest expansion was recorded in the last 4000 years.

113
METHODS
The sediment core was raised using a Colinvaux-Vohnout piston corer from a
platform attached to inflatable boats (Colinvaux and others 1999). The platform was
anchored in the middle of the lake at the deepest point. The sealed core tubes were
transported unopened to the laboratory and stored in a dark cold room until opened
and the sediments described. A total of 71 samples (0.5cm3), 24 for pollen, 24 for
charcoal, and 23 for mineralogy were collected from Lake Jacaré sediment core.
Samples for C14 AMS dating were sent to the INSTAAR – AMS Radiocarbon
Laboratory at the University of Colorado at Boulder, and the resulting dates were
calibrated to calendar years using the software CALIB 4.0 (Stuiver and Reimer
1993).
Standard pollen extraction procedures with KOH, HF, and acetolysis
followed Faegri and Iversen (1989) and Stockmarr (1971) for the addition of tablets
of Lycopodium spores to calculate pollen concentrations. Sodium pyrophosphate was
also used on clay-rich samples. Pollen samples were mounted in glycerol and counts
of at least 300 grains were conducted at 400x and 1000x magnification on a Zeiss
Axioskop equipped with a digital camera. Pollen grains were identified using the
Florida Institute of Technology reference collection of modern pollen, and published
catalogs with photographs and morphological descriptions of pollen types
(Colinvaux et al. 1999, Hooghiemstra 1984, Roubik and Moreno 1991).

114
Samples for charcoal analysis were disaggregated in 10% KOH, and the
resulting slurry washed through a 170 µm sieve (particles >170 µm were retained).
The material in the sieve was transferred to a Petri dish and charcoal counts
performed under an Olympus dissection microscope (20x magnification) equipped
with a video camera (Clark and Patterson 1997). Charcoal particles were manually
identified and then digital measurements were made via the video system and image
recognition (NIH-IMAGE) software. This software provides the area of charcoal
particles according to the number of pixels occupied by the fragment on the screen.
The pollen sum, percentage, and concentration were all calculated in TILIA (Grimm
1992), and plotted in C2 1.3 (Juggins 2003) and Corel Draw 8.
The characterization of the mineral fraction was performed at the University
of Bondy (France) by Dr. Bruno Turcq (IRD). The samples were prepared using the
KBr disc method (Bertaux and others 1998).
Detrended Correspondence Analysis – DCA (Hill and Gauch 1980) was
performed using PC-ORD 4.0 for Windows (McCune and Mefford 1999). This
technique was chosen because of 2 main reasons: The first one is that it prevents the
problems caused by the arch effect, which are a constant on other techniques such as
Principal Component analysis (PCA) and Correspondence Analysis (CA) (McCune
and Grace 2002). The second reason is that DCA allows the axes to be re-scaled and
shown as Standard Deviation units, which can be used to estimate the degree of
species turnover (McCune and Grace 2002). As very complex data sets with many
rare taxa can hinder DCA, the data set was reduced to include only pollen taxa with

115
percentage values > 0.5% and present at least in 3 samples throughout the sediment
core. The resulting matrix from Lake Jacaré had 40 pollen types. The percentage
values were then transformed using square root transformation, as this technique is
less drastic than log transformation (Kovach 1992). The resulting sample scores from
Axis 1 of DCA were plotted against sample depths in EXCEL.

116
RESULTS
RADIOCARBON DATES
The sedimentary chronology for Lake Jacaré is derived from 3 AMS
radiocarbon dates (Table 4.1). For each calibrated age a maximum probability
solution (Calib 4.3 type 2 solution) was adopted. All ages used henceforth will be
interpolated calibrated years before present (BP), unless noted otherwise.
The basal age of 9040 cal yrs BP indicates that Lake Jacaré formed at
approximately the same time of Lake Caracaranã (Simões Filho and others 1997),
and that its sediments provide a Holocene history for northern Roraima. The age
versus depth relationship (Fig 4.2) shows that sedimentation rates have remained
unchanged (~0.016 cm yr -1) between 105 cm (9040 years BP) and 50 cm (5620
years BP). However, the sediments above 50 cm provide much lower estimated
sedimentation rates of 0.009 cm yr -1, assuming that the top sediments (0 cm) were
deposited in the last 20 years.
Table 4. 1. Radiocarbon dates from Lake Jacaré sediment core.
Sample Depth (cm) 14C yr BP 13C/12C ratio Age (cal years BP)
CURL-5387 50 4870 + 40 -26.60 5579 – 5661
CURL-5388 81 6610 + 70 -28.13 7417 – 7593
CURL-5389 105 8080 + 45 -26.36 8975 – 9094

117
0
10
20
30
40
50
60
70
80
90
100
110
120
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Age (
14
C years BP)
Depth (cm)
Figure 4. 2. Radiocarbon ages from Lake Jacaré plotted with depths (cm). Dashed
line represents extrapolated ages.
STRATIGRAPHY
The 132 cm-long sediment core from Lake Jacaré (Fig 4.3) is mostly
composed of a mix of sandy silts and clays in the basal 32 cm, overlain by 100 cm of
dark silt (Table 4.2).
Table 4. 2. Sediment description of core from Lake Jacaré (Roraima - Brazil).
Depth (cm) Sediment description
100 - 0 Dark silt
106 - 100 Dark-sandy silt
116 - 106 Transition from blue-gray clay to dark-sandy silt
132 - 116 Blue-gray clay with sand

118
Dark silt Dark- sa ndy si lt Dark -sandy silt
with
Blue-gr ey clay
Blue-grey clay
0
10
20
30
40
50
60
70
80
90
100
110
120
130
140
Depth (cm)
5620 cal years BP
7510 cal years BP
9040 cal years BP
Figure 4. 3. Lithology of sediment core from Lake Jacaré, Roraima (Brazil). Also
showing location of dated samples in calibrated years BP.
PALEOECOLOGICAL RECORD – LAKE JACARÉ
Four pollen zones, J1, J2, J3, and J4 were identified in the record from Lake
Jacaré (Figs 4.4, 4.5, and 4.6). Samples from below 105 cm depth contained no
pollen.

119
Depth (cm)
5620 years BP
7510 years BP
9040 year s B P
0
10
20
30
40
50
60
70
80
90
100
110
120
130
J1
J2
J3
J4
0 1020 30405060
%Quartz
01020304050
%Amorphous Silica
0102030
%Kaolinite
Figure 4. 4. Diagram showing mineral fractions, lithology, and radiocarbon dates (cal
years BP).
J1 (132 cm – 65 cm; ca. 9040 yrs BP – 6500 yrs BP). Sediments are a
transition from sandy silts and clays to dark silt. Percentages of quartz are high in the
bottom 30 cm (ca. 52%) but markedly decrease to ca. 15% above 100 cm, while
amorphous silica (mostly sponge spicules) increases from ca. 7% to ca. 36%, and
kaolinite recovers from low values fluctuating around 15% (Fig 4.4). Total pollen
concentration oscillates around 20,000 grains/cm3 with as much as ca. 17,000 grains
coming from Poaceae, Zonorate, Cyperaceae, and Typha (Fig 4.5). A few forest taxa
such as Anacardium, Mor/Urt2, Urt/Mor3, Protium and Tapirira are present at
relatively low percentage values, while Asteraceae and Poaceae pollen are found at
relatively high abundances of 15% and 45%, respectively (Fig 4.6). Zonorate, Typha,

120
and Utricularia pollen types increase in abundance upward, whereas Cyperaceae and
Macrolobium decrease. Charcoal accumulation is insignificant at the beginning of
the zone, but increases rapidly upward (Fig 4.4).
J2 (65 cm – 50 cm; 6500 yrs BP – 5620 yrs BP). Sediments are composed of
dark silt. Percentages of quartz and kaolinite fluctuate around 12% whereas
amorphous silica increases from 36% to 46%. Total pollen concentration declines
(ca. 14,000 grains/cm3) but recovers at the top of the pollen zone. This zone can be
distinguished from the previous one by the contrasting increase of Bursera simaruba,
Caesalpiniaceae, Cecropia, Mimosa, Poaceae, and Cyperaceae abundances and the
decrease of Asteraceae, Zonorate and Typha. The abundance of charcoal particles
decreases slightly toward the top of this zone.
J3 (50 cm – 30 cm; 5620 yrs BP – 3400 yrs BP). Sediments are
indistinguishable from the previous zone. Percentages of mineral fractions display
insignificant variation except for the peak of quartz at 30 cm (ca. 3400 years BP) that
corresponds to lower values of amorphous silica and kaolinite. Pollen concentration
increases reaching ca. 30,000 grains/cm3. Concentration and percentage values of
Anacardiaceae and Anacardium pollen peak at 40 cm (ca. 4510 years BP). The peak
of Mor/Urt2, Mor/Urt3, and Sagittaria pollen contrasts to lower abundances of
Mimosa, Poaceae, and Cyperaceae, though Cyperaceae recovers quickly at the top of
the zone. Pollen of Ludwigia and Mel/Comb are recorded frequently. Charcoal
accumulation oscillates displaying the lowest values at 35 cm (3960 years BP) and
25 cm (2850 years BP).

121
x 1 00
x 100
A - 5620 cal years BP
B - 7510 cal years BP
C - 9040 cal years BP
01020 010 010 01020010010 010 010 010010 010010 010 010010 010010 010 010010 010 0102030 010 0102030 0 1 020 010 010 020406080100120
Depth (cm)
0
10
20
30
40
50
60
70
80
90
100
110
120
130
A
B
C
010 010 010 010 010 010 0100 102030 0204060801000200 1020 3040 506070 010 0 100 200 300 400
Depth (cm)
0
10
20
30
40
50
60
70
80
90
100
110
120
130
A
B
C
J1
J2
J3
J4
J1
J2
J3
J4
024681012
Alchornea
Anacardiaceae
Anacardium
Asteraceae
Bignoniaceae
Borreria
Bursera simar
uba
Caesal
piniaceae
Cassia
Cecropia
Celtis
Didymopanax
Euterpe
Geonoma
Hyptis
Ludwigia
Machaerium
Macr
olobium
Malvaceae
Mauritia
Mel/Comb
Mimosa
Mimosaceae
Mor/Urt2
Mor/Urt3
Myrtaceae
Papillionaceae
Poaceae
Pr
otium
Sapindaceae
Solanaceae
Spermacoce
Spondias
T
apirira
Zanthoxylum
Zonorate
Cyperaceae
Total Concentration
Charcoal
Sagittaria
T
ypha
Utricularia
Figure 4. 5. Pollen concentration diagram of Lake Jacaré, Roraima (Brazil), showing
selected taxa (scale is x 100 grains/cm3 of sediment), lithology,
radiocarbon dates (cal years BP), and charcoal particles (mm2/cm3). The
hollow curves are exaggerated 5 times.

122
01020010010010200100100100100100100100100100100100100100100100100100102001001020
0
10
20
30
40
50
60
70
80
90
10 0
11 0
12 0
13 0
A
B
C
010010010010203040506001001001001001001001001020
300 102030405060700100 102030405060
70 010
0
10
20
30
40
50
60
70
80
90
100
110
120
130
A
B
C
A - 5620 cal years BP
B - 7510 cal years BP
C - 9040 cal years BP
J1
J1
J2
J2
J3
J3
J4
J4
Alchornea
Anacardiaceae
Anacardium
Asteraceae
Bignoniaceae
Borreria
Burs
era simaruba
Caesalpini
aceae
Cassia
Cecropia
Celtis
Didymopanax
Euterpe
Ge
onoma
Hyptis
Ludwigia
Machaerium
Macrolobi
um
Malvaceae
Maurit
ia
Mel/Comb
Mim
osa
Mimosaceae
Mor/Urt2
Mor/Urt3
Myrtaceae
Papillionaceae
Poaceae
Pr
otium
Sapindaceae
Solanaceae
Spermacoce
Spondias
Tapirira
Zanthoxylum
Zonorate
Cyperaceae
Sagittar
ia
Typha
Utri
cularia
Figure 4. 6. Pollen percentage diagram of Lake Jacaré, Roraima (Brazil), showing
the most representative taxa, lithology, and radiocarbon dates (cal years
BP). The hollow curves are exaggerated 5 times.

123
J4 (30 cm – 0 cm; 3400 yrs BP until present). Sediments are composed of
dark silt. A peak of kaolinite at 15 cm (ca. 1730 years BP) is followed by a peak of
quartz at 10 cm (1180 years BP). Pollen concentration peaks at 15 cm reaching
36,000 grains/cm3 and then decreases markedly upward. Pollen of Mimosa and
Poaceae are abundant again, while Mor/Urt2 and Mor/Urt3 reach their lowest values.
Ludwigia, Machaerium, Macrolobium, Mauritia, and Mel/Comb have higher pollen
percentage values than in the previous zone. Asteraceae pollen abundance increases
in the top 5 cm, and Cyperaceae has relatively high values in the top sample (0 cm).
Percentages of Cecropia are very low in the middle of this zone. Charcoal
accumulation decreases markedly after reaching the highest values of the core at 10
cm.
MULTIVARIATE ANALYSIS – LAKE JACARÉ
The resulting DCA scores were plotted in two different ways. First Axis 1 vs.
Axis 2, then sample scores from Axis 1 were plotted with corresponding depths and
ages in cal years BP.
AXIS 1 vs. AXIS 2. The DCA scores from axes 1 and 2 showed a strong
polarization on Axis 1 dividing the samples into two main groups: the bottom
samples of the core on the positive extreme (J1 and J2) and the top samples of the
core on the left, representing zones J3 and J4 (Fig 4.7). The most characteristic
species of the positive side of Axis 1 were Zonorate, Typha, Protium, Solanaceae,

124
and Hyptis. The scores that characterize the negative extreme of Axis 1 were yielded
by Spermacoce, Euterpe, Mauritia, Ludwigia, and Malvaceae.
Axis 2 displays a somewhat weaker polarization, placing samples
representative of zones J1 and J2 in the middle, while samples from upper zones (J3
and J4) are evenly spread across the axis. For this reason, only scores from Axis 1
were plotted against depth.
J-0
J-5 J-10
J-15
J-20
J-2 5
J- 3 0
J- 35
J-40 J-45
J-50
J-55
J-60
J-65
J-70
J-75
J-80
J-100
J-105
0
0
20 40 60 80 100
10
20
30
40
50
60
70
80
Axis 1
Axis
2
J1
J2
J3 and J4
Figure 4. 7. DCA scores from Lake Jacaré showing Axis 1 vs. Axis 2. Samples are
grouped according to interpreted phases.

125
AXIS 1 vs. DEPTH. The DCA sample scores plotted against depth show a
pattern that suggests a relatively uniform and unidirectional transition between zones
(Fig 4.8). The bottommost samples (J1) are placed on the positive side of the axis,
whereas the above samples (J2, J3, and J4) drift continuously toward the left side of
Axis 1.
DCA scores - Axis 1
-40 -20 40 60 80 100200
0
10
20
30
40
50
60
70
80
90
10 0
110
Depth (cm)
Zonorate
Typha
Protium
Solanaceae
Hyptis
Spermacoce
Euterpe
Mauritia
Ludwigia
Malvaceae
5620
7510
9040
J1
J2
J3
J4
Figure 4. 8. DCA scores from Lake Jacaré, Roraima (Brazil). Axis 1 plotted with
depth (cm). Also showing dates (cal years BP), the interpreted pollen
zones, and the five most characteristic taxa of each axis extreme.

126
DISCUSSION
The sediment core from Lake Jacaré provides the longest Holocene
paleoecological record from Roraima (Brazil) yet. As the basal date of the core
yielded a calibrated age of 9040 years BP, this record spans almost the entire
Holocene period.
The mix of sandy silts and clays on the bottom of the core suggests a phase of
unstable water levels. Even though the sediment core is 132 cm long, pollen is only
preserved in sediments above 105 cm (ca. 9040 years BP). The contrasting decrease
in the high percentages of quartz minerals and the increased values of amorphous
silica, which come mostly from sponge spicules, indicate a change in the
depositional environment, favoring sedimentation of autochthonous material instead
of detrital minerals (Sifeddine and others 2001). Such a change is consistent with a
rising water-table, longer periods of inundation, and increased precipitation around
9040 years BP. At Carajás, decreased influx of detrital minerals, indicating less
erosion, is connected to the development of dense, humid forest, which is attributed
to wet conditions between 11,400 cal years BP and 8900 cal years BP (Absy et al.
1991, Sifeddine et al. 2001). But thereafter, lake levels at Carajás fell, leaving a
depositional hiatus from 8000 cal years BP to 5000 cal years BP in some cores (Absy
et al. 1991, Sifeddine et al. 2001).
At Jacaré, the pattern appears to be somewhat different from that of Carajás.
The early Holocene is marked by a progressive rise in lake level, and stabilization of

127
lake level just as the Carajás record is beginning to decline. Increasing abundances of
Zonorate and Typha pollen toward the top of J1 (ca. 6500 cal years BP) indicate a
transition from a Typha swamp to an open water body as lake level progressively
rises suggesting wet conditions. The contrasting high abundances of Asteraceae and
Poaceae, and the relatively low percentages of some woodland taxa (e.g. Alchornea,
Anacardium, Mor/Urt2, Mor/Urt3, Protium, and Spondias) suggest that even though
savanna vegetation was already dominant in the region, there were forest patches
near the site and probably gallery forests along the rivers and wetlands.
Fire frequency increases after ca. 7500 years BP toward the top of zone J1.
The oldest radiocarbon dated charcoal particles, derived from woody taxa, found in
soil profiles within savannas near Boa Vista (ca. 170 km southwest of Lake Jacaré)
provided ages in the range of 7470-8460 years BP (Desjardins and others 1997).
Because old records of pre-Columbian human occupation in Amazonia were
inexistent, which biased the perceptions of the earliest human occupation on the Rio
Negro region, Sanford et al. (1985) ascribed fires recorded at ca. 6500 years BP to
natural causes rather than human origin. However, as new archaeological data from
elsewhere in Amazonia (Roosevelt et al. 1991, Roosevelt et al. 1996, Roosevelt
2000) provided strong evidence that pre-Columbian human impacts were more
extensive and older than previously thought (more than 11,000 years old), the early
views of an untouched Amazonia became obsolete. As the oldest dated record of
human occupation in Roraima is from ca. 4400 years BP, even though dates could
reach ca. 6850 years BP in the bottom of the profile (Ribeiro 1997), and the northeast

128
Roraima is one of the driest regions of Amazonia (Fig 1.6), it would be possible that
the fires recorded at Jacaré between 7500 years BP and 6500 years BP could be of a
natural cause. Nevertheless, as new archaeological sites that may provide earlier
dates of human occupation for this region may be found with intensified research
(Ribeiro 1997), the possibility that the on-set of fire at Jacaré was anthropogenic in
origin cannot be totally excluded.
In zone J2, the increased pollen percentage and concentration values of
Mimosa, which are herbaceous plants present in modern savannas of Roraima
(Miranda and Absy 1997), and Poaceae, and the lower values of Asteraceae imply a
change within the savanna community at ca. 6500 years BP. The marked decrease in
abundances of Zonorate and Typha pollen suggests the establishment of a true lake.
The transition from a Typha swamp to an open-water system was probably caused by
higher lake levels, which could be the result of an increase in precipitation or a
shorter dry season. By contrast, a depositional environment sufficiently dry to
exclude Typha would not have produced the observed changes in mineral, carbon or
silica content, nor would it have allowed pollen preservation. Despite deepening of
the lake, abundances of pollen from forest taxa remain constant from J1 to J2. The
largest change in the pollen record is the transition of wetland taxa from Typha to
Cyperaceae suggesting that the deeper or more stable water depth may have resulted
from a change in basin hydrology rather than a change in climate. One explanation
for such an event is that the clays accumulating in the Typha marsh sealed the lake
basin. Reduced leakage could account for changing lake level without invoking

129
climatic change. As the area of the catchment increased, the lake would be expected
to record a more regional pollen and charcoal signal instead of just the local swamp.
The presence of large amounts (>20%) of Cecropia pollen in sediments from
humid lowland tropical systems is usually interpreted as a disturbance indicator
(Behling and Hooghiemstra 1999, Bush et al. 2000, Berrio et al. 2002). However,
Cecropia trees occur regularly as gap-filling constituents of gallery forest bordering
rivers and lakes, or in other edaphically moist pockets within savanna. To find
evidence of fire in the Jacaré record that does not induce a response in Cecropia is
not unusual as the savanna succession following fire is commonly dominated by
Poaceae, Mimosaceae, Byrsonima and Curatella rather than Cecropia (Hoffmann
and Moreira 2002, Miranda et al. 2002). When Cecropia pollen does show an
increase in representation, about 1000 years after the onset of fire and coincident
with lake expansion, increased lake size allowing capture of a regional, rather than
local, pollen input, may be a better explanation than a vegetation change as a
response to disturbance.
Zone J3 is particularly interesting because it harbors two apparently opposing
events: A period of forest development followed by a dry event with maybe
associated erosion. Between 50 cm (5620 years BP) and 35 cm (3950 years BP), a
period of forest expansion is suggested by markedly decreased abundances of
Mimosa (especially) and Poaceae pollen (savanna indicators) and increased
abundances of pollen from forest taxa such as Anacardium, Anacardiaceae, and
Myrtaceae at 40 cm followed by Mor/Urt2 and Mor/Urt3 at 35 cm. These alternating

130
peaks of pollen abundances imply that succession was taking place within the forest
community.
Even though the overall pattern of fire frequency displays insignificant
variations, the accumulation of charcoal at 35 cm is the lowest since the on-set of
fire, which could suggest the wettest period of the record, as it is coincident with the
peak of the forest expansion. Furthermore, because the increasing pattern of fire
frequency is interrupted only at 35 cm (ca. 3950 years BP) and forest expansion
started at 50 cm (ca. 5620 years BP), it is reasonable to assume that during this
period of time fires were anthropogenic. Overlapping dates of the oldest record of
human occupation in savannas of Roraima (Ribeiro 1997) and this period of charcoal
accumulation at Lake Jacaré provide further support to this hypothesis. Because this
is the only phase of forest expansion observed in the core, the climate conditions
during this period (after ca. 5620 years BP) are interpreted to be relatively wetter
than the present.
Because the vegetation around Jacaré has been a savanna throughout the
history of the lake, no savanna expansion was recorded between 8000 and 5000
years BP. Therefore, hypothesis 1 was rejected.
H1: Amazonian savannas expanded between 8000 and 5000 years BP.
Although the record from Lake Jacaré shows a period of forest expansion
from 5700 to 4700 years BP, no forest expansion was recorded in the last 4000 years.
Therefore, hypothesis 2 was rejected.
H2: Forests expanded and savanna vegetation contracted in the last 4000 years.

131
At 30 cm (ca. 3400 years BP), the percentage of quartz increases abruptly
from 5% to ca. 37% indicating a strong dry or erosional event. The relatively high
abundances of Ludwigia, Mauritia, Sagittaria, and Utricularia pollen at 30 cm
indicate an expansion of swamp/marsh vegetation supporting an interpretation of
lake contraction. As sedimentation rates between 50 cm and 0 cm are markedly
lower than those deeper in the core (Fig 4.2), it is possible that at 30 cm the drier
conditions caused a hiatus in deposition, which would account for the apparent
slowing of sedimentation. Assuming that sedimentation rates remained constant
between 81 cm and 35 cm and between ca. 27 cm and 0 cm, the new estimated age
(through linear interpolation) for the sedimentary layer at 35 cm would be ca. 4750
cal years BP, and the roughly estimated time of the duration of the sedimentary
hiatus would be of ca. 3050 years (4750 years BP – 1700 years BP = 3050 years BP)
(Fig 4.9). However, two additional AMS radiocarbon dates, one at 35 cm and the
other at 28 cm would be the most reliable and definite way to confirm or refute the
sedimentary hiatus and provide a better estimate of its duration.

132
hiatus
Break of sedimentation
Resume of sedimentation
0
10
20
30
40
50
60
70
80
90
100
110
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Ag e ( cal yea rs B P)
Depth (cm)
Figure 4. 9. Estimated sedimentation rates through rudimentary linear interpolation
reveal a sedimentary discontinuity at 30 cm. The duration of the hiatus is
estimated to be of ca. 3050 years, assuming constant sedimentation rates.
In zone J4, the influence of forest taxa is markedly reduced (the lowest
abundances of Urt/Mor2 and Urt/Mor3 pollen) while abundances of Mimosa and
Poaceae pollen increase again, suggesting a savanna expansion. At 10 cm (ca. 1180
years BP), a peak of quartz abundance also coincides with the largest values of
charcoal since the on-set of fire. Together, these data may be explained by increased
erosion caused by higher fire frequency in the surrounding landscape. Lower
abundances of Alchornea, Asteraceae, Cecropia, Mimosa, Poaceae, and Cyperaceae
simultaneous with relatively higher abundances of Celtis, Machaerium,
Macrolobium, Mel/Comb, and Urt/Mor2 indicate climatic conditions similar to

133
modern ones. Therefore, the coincident increase in quartz mineral and charcoal
accumulation probably indicates a period of increased human impacts.
In the top part of the record, modern climate conditions were probably
established in the region. Additionally, there seems to be less influence of forest taxa
near the site, and more influence of marsh/swamp taxa growing on the lake’s
shoreline. Markedly reduced charcoal accumulations in the top samples probably
indicate less human impact near the present times, perhaps reflecting post-
Columbian site abandonment.

134
CONCLUSIONS
The paleoecological record of Lake Jacaré provides a history of savanna-
forest dynamics in the northern part of Roraima. A period of forest expansion
associated with climate conditions wetter than the present is recorded at ca. 5620
years BP. The estimated age for the maximum of forest expansion and wet
conditions is of 4750 years BP. A markedly dry event, when the lake probably dried
out interrupting sedimentation, followed the forest expansion. The duration of the
sedimentary hiatus can be reliably estimated and confirmed only by additional AMS
dates on the sedimentary layers below and above the discontinuity layer (35 cm).
That the previous paleoecological records from Roraima, especially the one
from Lake Fazenda São Joaquim, did not reveal any significant vegetation change
(Absy 1979, Absy et al. 1997) could be due simply to the short period of time
spanned by the cores. In the case of the Lake Fazenda São Joaquim record,
additional explanations could be the local absence of forest patches and a small
catchment area, which would make the local pollen signal much stronger than the
regional.

135
CHAPTER V
HOLOCENE VEGETATION AND CLIMATE HISTORY OF SAVANNA-
FOREST ECOTONES IN NORTHEASTERN AMAZONIA
The overall paleoecological evidence from this study suggests that the
landscape in Eastern Amazonia has not been stable during the Holocene. Climatic
changes are evident in the records from the 5 lakes studied in this research. However,
the timing of those events is not precisely synchronous and leads to a consideration
of factors that influence lake level and local climates in Eastern Amazonia.
Despite all the lakes lying relatively close to one another, just 5° of latitude
separates the southernmost (Prainha) from the northernmost site (Jacaré) (Fig 5.1),
Lake Jacaré is climatically distinct from the others. Only Jacaré lies north of the
equator with the consequence that its climate follows the northern hemisphere
pattern, i.e. wet season from June to August (JJA), and dry season from December to
February (DJF). In contrast to this site, the others have their wet season from
December to February (DJF) and their dry season from June to August (JJA).
Furthermore, because the climate at Lake Jacaré is the driest of the study sites, slight
increases in precipitation should have stronger effects at this site than on the others.
The combined histories of the lakes reveal both local variation and some
regional commonality (Fig 5.1). The lakes are all Holocene in age and exhibit rising
water levels until ca. 6500 years BP, 5300 years BP in the case of Lake Marcio. All

136
the pollen records document this event with an expansion of wetland/swamp taxa.
Afterwards, relatively drier environmental conditions are inferred, as some lakes
record expansion of savanna taxa while others present sedimentary hiatuses. Another
striking similarity is the period of forest expansion observed in some of the records
(e.g. Lakes Santa Maria, Geral, Saracuri, and Jacaré). The timing of the
environmental changes, their nature and extent are probably the direct result of
several factors acting at very local to regional geographical scales.
There follows a review of major causes of climatic variation in Eastern
Amazonia and predictions for how each of these factors could influence the study
sites.

137
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
5500
6000
6500
7000
7500
8000
8500
9000
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
5500
6000
6500
7000
7500
8000
8500
9000
Interpol ated ag e (cal year s BP)
Interpolated age (cal years BP)
Forest regrowth
Lake formation
Hum an imp ac t
Savanna expansion
Scrub savanna
Wetland
Swa mp for est
Lake Santa Mari
a
Lake Geral
Lake Saracuri
Lake Marcio
hiatus
Lake Tapera
hiatus
hiatus
hiatus
hiatus
hiatus
hiatus
Lake Jacaré
hiatus?
hiatus?
hiatus?
Flooded savanna
Figure 5. 1. Environmental changes interpreted from the paleoecological records of
Lakes Santa Maria, Geral, and Saracuri (ca. 1°S), Lakes Marcio and
Tapera (ca. 0°N), and Lake Jacaré (ca. 3°N). Also showing chronology
(cal years BP) and human impacts, as suggested by large charcoal
deposits and agriculture indicators (e.g. maize pollen grains).

138
SOLAR FORCING
Changes in regional insolation are caused by variations in the orbit of Earth
around the Sun. These orbital patterns, known as Milankovitch cycles, include
precessional rhythms that last ca. 22,000 years, and obliquity rhythms, lasting ca.
40,000 years (Imbrie and others 1984). Orbital forcing affects temperature, and to
some extent precipitation in the tropics. A 650,000-year record of orbital forcing is
evident from the Sabana de Bogota, Colombia (Van't Veer and Hooghiemstra 2000).
At El Valle (Panama), changes in lake level, as inferred from pollen abundance of
marsh taxa for the last 40,000 years (Bush 2002), are relatively synchronous with the
GISP2 record (Blunier and Brook 2001). Similarly, sedimentary data from the
Bolivian Altiplano show wet and dry phases in the last 50,000 years that are related
to orbital forcing. Wet events inferred from deposits of mud layers rich in diatoms
coincide with periods of maxima of austral summer insolation (Baker and others
2001a). The sediments of 2 lakes from western Amazonian lowlands clearly record
the orbital cycles (Bush and others 2002). Peaks of K+ concentration, used as a proxy
for lake productivity, which are coincident with lower lake levels, follow the minima
of austral summer insolation at 0° N. A key observation is that while precession
cycles correlate with lake level at all these sites, in Colombia and Panama the timing
of highstands coincides with lowstands in the southern hemispheric sites.
Peaks in insolation at 0° latitude (Laskar 1990) result in higher temperatures,
which are translated into increased precipitation through complex mechanisms (Bush
and others 2002). As the rhythm induces changes in the strength of seasonality

139
through time, it is important to consider the insolation curves for both wet and dry
seasons.
Based only on the mean insolation curves calculated for December and June
at 0° latitude (Fig 5.2) some predictions can be made. The maximum June insolation
and minimum December insolation in the last 20,000 years are observed between
14,000 and 9000 years BP. For Lake Jacaré, which lies in the northern hemisphere,
this peak of summer insolation and dip of winter insolation result in wet seasons
wetter than normal and dry seasons drier than normal. Consequently, the period of
the strongest seasonality for this region in the last 20,000 years would be predicted to
be between 14,000 and 9000 years BP. Contrastingly, the other study sites, which lie
in the southern hemisphere, probably experienced a very weak seasonality, with wet
seasons drier than normal, and dry seasons wetter than normal at this time.
Unfortunately, and perhaps not coincidently, the studied lakes in the southern
hemisphere started forming only after 8500 years BP (Fig 5.1).
For the last 5000 years, the sites in the southern hemisphere have been
experiencing a stronger seasonality, as dry seasons became relatively drier, and wet
seasons relatively wetter. Conversely, climate at Jacaré would have become less
seasonal, with wetter dry seasons and drier wet seasons.

140
Time (years BP)
Insolation (W/m )
2
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
11000
12000
13000
14000
15000
16000
17000
18000
19000
20000
2100
0
0380 390 395 400 405 410 415 42 0385
December
June
Norther
n
Hemisphere
Wetter dry season
Drier wet seaso n
Wetter dry season
Drier wet season
Drier dry season
Wetter wet season
Drier dry season
Wetter wet season
Souther
n
Hemisphere
Figure 5. 2. Mean insolation (W/m2) during the months of June and December at 0°
latitude for the last 20,000 years (Laskar 1990). Between 7000 and 6000
years BP the difference between June and December mean insolation
was the smallest within the Holocene.
ITCZ MIGRATION
The Inter-Tropical Convergence Zone (ITCZ) is a narrow latitudinal belt of
atmospheric convection with associated cloud formation due to warm sea surface
temperatures (SSTs) and the convergence of the northeast and southeast trade winds
in the eastern Pacific (Cronin 1999). The seasonal migrations of the ITCZ coupled
with convective activity over the continent have a major influence on precipitation in
South America (Martin et al. 1997, Marengo and Rogers 2001, Bush et al. 2002). At

141
present, during the Austral summer (December-February) the ITCZ reaches its
southernmost position (Fig 5.3), and during the Austral winter (June-August) the
ITCZ is displaced to its northernmost position.
Although the importance of the ITCZ on the Amazonian climate is widely
accepted, there is no agreement yet about its range of migration or its mean width.
Some authors represent the ITCZ reaching as far south as 20-25°S (Martin et al.
1997, Ledru et al. 2002) in the Austral summer, and as far north as 10-20°N (Haug et
al. 2001, Haug et al. 2003) in the Austral winter, though a maximum northward
extension to ca. 5°N would be more widely accepted. As the ITCZ is primarily an
oceanic phenomenon (Cronin 1999), it is possible that these authors are mistaking
convective activity caused by different factors for the ITCZ. During the Austral
summer, precipitation over the continent is due mainly to the combination of deep-
cell convective activity, inputs of moisture from the ITCZ and South Atlantic
Convergence Zone (SACZ), and cold air from Antarctic (Marengo and Rogers
2001). For this reason, precipitation over the continent should be treated as
convective activity instead of the ITCZ.
Past climate conditions have been ascribed to variations in the range and
position of the ITCZ throughout time. Some of the suggestions of past ITCZ position
are improbable, but all make no distinction between the ITCZ and convective
activity. Martin et al. (1997) suggested that between 12,400 and 8800 cal years BP
the southern boundary of the ITCZ must have been further north during the Austral
summer. Because several sites below 10°S have recorded a dry phase at that time,

142
e.g., Salitre (Ledru 1993) and Brasilia (Salgado-Labouriau 1997, Salgado-Labouriau
et al. 1998), while sites north of 10°S recorded a wet phase, e.g. Carajás (Absy and
others 1991), and Lake Valencia (Leyden 1985). Martin et al. (1997) inferred that the
southernmost position of the ITCZ during the Austral summer was at 10°S. In
contrast, Ledru et al. (2002) suggested that the southern boundary of the ITCZ was
farther south (ca. 25°S) during the Younger Dryas (between ca. 12,500 and 11,000
cal years BP), which supposedly allowed polar advections from the northern
hemisphere to reach as far as 2°58’S at the Lagoa do Caçó. Clearly these 2
suggestions are contradictory.
Figure 5. 3. Schematic map of South America showing the study sites and the
present mean positions of the ITCZ (white clouds) during the Austral
summer (December-February) on the left, and during the Austral winter
(June-August) on the right. The pink clouds over the continent represent
increased convective activity.

143
A forest expansion at the expense of savanna in the last 3000 years is inferred
from a paleoecological record at ca. 13°S in the Bolivian Amazonia (Mayle and
others 2000). This environmental change was attributed to the increased Austral
summer insolation, which was believed to have caused an overall migration of the
ITCZ further south during the southern hemisphere summer (Mayle and others
2000).
Northern hemispheric oscillations and the position of the ITCZ have also
been invoked to explain climate change in the Caribbean. A high resolution
sedimentological record from the Cariaco Basin, Venezuela (10°N) is interpreted to
indicate shifts in the mean position of the ITCZ during the Holocene (Haug and
others 2001). In this record, sedimentary concentrations of iron (Fe) and titanium
(Ti) were measured at ca. 5-year resolution for the last 14,000 years. High
concentrations of these metals between 10,000 and 5000 years BP are related to
increased input of terrigenous sediment, and hence higher precipitation than in the
periods before and after. Such changes in precipitation pattern were linked to shifts
in the mean latitudinal position of the ITCZ, which were, in turn, caused by
insolation. As the wettest period recorded, inferred to be between 10,500 and 5400
years BP, is coincident with the time of maximum summer insolation (JJA), the
mean position of the northern boundary of the ITCZ must have been further north
bringing more seasonality (Haug and others 2001). Conversely, after 5000 years BP,
the Cariaco Basin recorded increasingly dry conditions, probably related to a

144
southern displacement of the northern boundary of the ITCZ, caused by changes in
insolation.
As insolation is the major driver of ITCZ migration and the associated
convective activity over the continent, we could expect the sites located in the
southern hemisphere to record increased precipitation during the wet season over the
last 5000 years, as the Austral summer insolation has been at its maximum (Fig 5.2)
during this time, which would have driven ITCZ and the convective activity further
south. Conversely, the Jacaré site should record increasingly dry conditions in the
last 5000 years.
ENSO EVENTS
The term ENSO incorporates El Niño, La Niña and the Southern Oscillation
oceanic/atmospheric phenomena. The warm current that appears off the coast of Peru
around Christmas time was named El Niño (after the birth of Christ) by Peruvian
fishermen. El Niño is basically a disruption of the ocean-atmosphere system in the
tropical Pacific that affects the weather around the globe (Cronin 1999). El Niño
starts with the weakening of the easterly trade winds that interrupts the transport of
the surface water from the eastern Pacific to the western Pacific, which stops the
upwelling off the coast of Peru (Beaufort and others 2001). The sea surface
temperatures (SSTs) at the eastern Pacific increase, which intensifies evaporation,
causing precipitation in the west coast of South America, and southern United States.

145
Even though El Niño variability through time and its effects on the
ecosystems are still far from being completely understood, there seems to be some
consensus about increased ENSO activity following the middle Holocene. A
sedimentary record from the Galápagos Islands, located within the core region of
ENSO, provides evidence of increased ENSO activity (frequency and intensity) after
4600 cal years BP (Riedinger and others 2002). Another regional ENSO record
comes from Laguna Pallcacocha in southern Ecuador. Imagery of the sediment core
was digitized, and the reflectance at 3 wavelengths was measured (Moy and others
2002). The wavelet analysis used red color intensity to provide a record of ENSO
activity for the last 14,000 years. ENSO variance becomes statistically significant
after 7000 cal years BP, but its frequency increased dramatically after 5500 cal years
BP (Fig 5.4). Based on beach-ridge records from southeastern Brazil, Martin et al.
(1993) suggested that changes in the direction of the drift current were due to strong
ENSO events after 5000 years BP.

146
1000
0010203040
ENSO band
Events per 100 yr
Time (cal years BP)
2000
3000
4000
5000
6000
7000
8000
9000
10000
Figure 5. 4. Variation in event frequency of ENSO during the Holocene (modified
from Moy et al. 2002)
One of the major effects of modern El Niño in eastern Amazonia is a
reduction in precipitation. Reductions in Amazon River discharge over the last
century have been linked to EL NIÑO (Richey and others 1989). However, only a
portion of the variance in the river discharge could be explained by El Niño events,
as the relationship between precipitation and runoff is complex (Richey and others
1989). Marengo et al. (1998) found that during El Niño events the northeast trade
winds are weakened, reducing the flow of moist air from the Atlantic into Amazonia,
which would probably reduce the formation of convective centers. Conversely,
episodes of increased river discharge are often related to La Niña events (Marengo
and others 1998). Episodic events of sediment deposition in the Beni and Mamoré

147
River floodplains (western Amazonia) have been associated with rapid floods caused
by La Niña (Aalto and others 2003).
Considering only the climatic effects of ENSO on eastern Amazonia, as the
frequency and intensity of El Niño events increased in the last 5000 years, an overall
trend to drier conditions caused by this disturbance should be recorded by the lakes
near Prainha and in Amapá. As Lake Jacaré is the driest of the study sites, and is also
the northernmost site (northern hemisphere), the effect of El Niño events is unclear.
RELATIVE SEA-LEVEL CHANGE
Factors influencing post-glacial sea-level changes include eustasy, glacio-
isostatic rebound, wave and wind pattern, climate and tectonics (Bezerra and others
2003). Several studies have provided records of Brazilian sea-level changes, but they
focused mainly on the central and southern portions of the coast (Suguio et al. 1985,
Martin et al. 1986).
The northernmost sea-level reconstruction comes from Rio Grande do Norte
(Bezerra and others 2003). During the last glaciation, sea-level was ca. 125 m
lower than today. At the beginning of the Holocene, sea-level rose rapidly, and
reached modern levels between 7000 and 6000 cal years BP (Bezerra and others
2003). At 5000 cal years BP, due probably to isostatic movements, relative sea-level
was higher than at the present, but it quickly fell to present levels around 4000 cal
years BP. A last sea-level rise is centered at 2000 cal years BP (Bezerra and others
2003).

148
The impacts of Holocene sea-level changes in several Amazonian sites are
summarized by Behling (2002). The grade of the Amazon River is so gentle
(~1:100,000) that the water level in Manaus (1200 km inland) is only 14 m above
sea-level (Behling 2002). Consequently, even small changes in sea-level can have a
strong effect on the landscape. The formation of Lake Calado, in central Amazonia,
was ascribed to sea-level rise at ca. 7700 years BP (Behling and others 2001). Near
the mouth of the Amazon River, in the Caxiuanã region, the Curuá River changed
from an active to a passive river around 8000 years BP (Behling and Da Costa 2000).
A marked increase of inundated forests, inferred by fossil pollen analysis, is
coincident with the second Holocene sea-level rise around 2500 years BP (Behling
and Da Costa 2000). The paleoecological record of Lake Crispim (coast of Pará)
provides further evidence of Holocene sea-level changes in Amazonia (Behling and
Costa 2001). The evolution of the mangrove vegetation near the lake appears to be
related to fluctuations of relative sea-level. Nevertheless, the most striking feature of
this record is the sharp increase in abundance of Poaceae pollen in the top of the
core. The interpolated age for this event, based on 3 dates only, is ca. 3880 cal years
BP. However, as top-bottom and bottom-up sedimentation rates do not provide
overlapping interpolated ages, and the sedimentary layers are different, it is probable
that there is a sedimentary hiatus in the Lake Crispim core. The dramatic change in
the pollen assemblage is consistent with a sedimentary gap. Unfortunately, more
radiocarbon dates are needed to verify the existence and duration of this hiatus. The
pollen record from Lake Aquiri, Maranhão (NE Brazil), presents a dramatic

149
vegetation change from a mangrove to a swamp savanna sometime after 7180 cal
years BP (Behling and Costa 1997). However a sedimentary hiatus accompanied by
an erosional event is also apparent from this record, and although there are no
radiocarbon dates above this sedimentary layer, the authors estimated that the top 22
cm would be representative only of the last 150 years (Behling and Costa 1997).
It can be expected that the sites near the Atlantic Ocean would have
experienced the effects of sea-level fluctuation more strongly than the sites further
west. However, as the topography along the Amazon River is very low, it is possible
that even ca. 500 km inland, water-tables were influenced by the sea-level rise,
which is reflected in the initial ponding of the lakes in Prainha.
HUMAN OCCUPATION
Archaeological data have provided strong evidence that humans have lived in
Amazonia for more than 11,000 years (Roosevelt et al. 1991, Roosevelt et al. 1996,
Roosevelt 2000). The effects of such long-term occupation on the Amazonian
ecosystem are still poorly known. Nevertheless, a few archaeological records have
suggested that pre-Columbian peoples have significantly modified the Amazonian
landscape (Erickson 2000, Erickson 2001, Heckenberger et al. 2003). Roosevelt
(2000) suggested that savanna sites near Monte Alegre are anthropogenic, and C4
plants (mainly Poaceae) were virtually absent from these areas, at least until ca. 6000
cal years BP. Testing this hypothesis should be fairly simple with the paleoecological
records from the region of Prainha. Finding fossil pollen of Poaceae in sediments

150
older than 6000 cal years would be the first step. The second and more difficult step
would be establishing the source of the Poaceae pollen, as grasses can grow on a
wide variety of environments such as savannas, swamps, and lakeshores. However,
this task could be accomplished by comparing the variation in the pollen abundance
(concentration and percentage) of Poaceae with that of Cyperaceae, as the latter
family is often used as a proxy for marsh taxa.

151
INTEGRATION OF ALL THE FACTORS
THE AMAPÁ LAKES
All the factors previously described have influenced the study sites, at least to
some extent. However, it is obvious that some factors had a much stronger effect on
certain sites than on others.
Variations on the relative sea-level probably had a much stronger impact on
the Amapá lakes, as they lie so close to the Atlantic (Fig 1.3) and within the range of
influence of the macro tides, than on the other sites. The initial ponding of Lakes
Tapera and Marcio was probably due to river flooding induced by rapid sea-level rise
(Suguio et al. 1985, Martin et al. 1986). Both lake records appear to have a
sedimentary hiatus. At Lake Tapera the hiatus lasts from ca. 7510 until 1670 cal
years BP while at Lake Marcio the sedimentary gap lasts only from 5300 until 4750
cal years BP. Such a discontinuity suggests that sea-level was not the primary control
of lake level, as the relative sea-level at 5000 cal years BP was the highest of the
Holocene and yet both lakes were dry. The difference between the estimated duration
of both sedimentary gaps is apparently due to other factors. However, it is important
to note that Lake Tapera is slightly higher than Marcio, as shown by the approximate
calculation based on the stratigraphy of both lakes (Fig 5.5).

152
Lake Tapera
Water depth
250 cm
Gyttja
65 cm
Total depth 450 cm
Gray clay
135 cm
Lake Marcio
Water depth
165 cm
Total depth 635 cm
Gyttja
118 cm
Gray clay
352 cm
Figure 5. 5. Schematic representation of stratigraphic profiles of Lakes Tapera and
Marcio. Blue represents the water column, black represents the layer of
gyttja, and gray represents the basal clays. Notice that Lake Marcio’s
basin is at least 185 cm deeper than that of Lake Tapera.
As it has been demonstrated that sea-level change was not the primary
control of lake levels in Amapá, alternative hypotheses must be explored in order to
explain the dry event. Both lakes probably relied on flood water to keep a high stand,
although their initial ponding was probably caused by the rapid sea-level rise
(Bezerra and others 2003). The fact that the sedimentary hiatus at Tapera started ca.
2200 years earlier than at Marcio, and also lasted ca. 3000 years longer, suggests that
flood waters stopped reaching Tapera earlier than at Marcio. A further possibility is
that Tapera’s basin was not completely closed yet, so that the lake could be losing
water through seepage. It is plausible to suppose that the hiatus at Lake Marcio (from
ca. 5300 until 4750 cal years BP) indicates the maximum isolation from riverine
flood waters. Several records show that the Andes were dry from ca. 8000 to 5000
years BP. The records from Lake Titicaca (Peru-Bolivia) show low lake stands in

153
this period of time (Wirrmann and others 1988), with the maximum aridity estimated
to be between 6000 and 5000 years BP (Baker and others 2001b). Additionally, the
ice core from Sajama (Bolivia) recorded increased concentrations of dust (Thompson
and others 1998), related to drier conditions during the mid-Holocene. As the climate
at the Andes was dry between ca. 8000 and 5000 years BP, it is probable that the
flow of the Amazon River was reduced to a great extent, as a considerable proportion
of its waters comes from the Andean region. Therefore, as Lakes Tapera and Marcio
probably relied heavily on flood waters from the Amazon River, they both dried out,
even though relative sea-level was high at ca. 5000 years BP. This hypothesis is
supported by the pollen records of Marcio and Tapera, which show a greater
abundance of Podocarpus pollen, an Andean taxon, before the sedimentary hiatus
than after, when sedimentation resumed (Figs 2.4 and 2.5). Although the climate
change at the Andes was probably the major influence on the dry event recorded by
Lakes Marcio and Tapera, local sedimentary processes, such as a river channel
migration could have also played an important role. As relative sea-level was high,
and the Amazon flow was reduced, it is possible that a build up of sediments
facilitated a “coastal progradation” within the Amazon estuary, intensifying the
effects of flood water isolation already experienced by the lakes. A similar process
was described for a mangrove in French Guiana, in which a coastal progradation at
ca. 5000 years BP was suggested to explain the decline in mangrove vegetation at the
peak of sea-level rise in the pollen record of a coastal lake (Tissot and others 1988).

154
That the relative sea-level fell to its modern position after 5000 cal years BP
(Bezerra and others 2003) and the resumption of sedimentation in Lake Marcio are
contemporaneous, suggest that sea-level did not play an important role in
determining mid-Holocene lake levels in Amapá. Therefore, a climatic forcing must
be invoked to account for this change. The local wetter conditions recorded by Lake
Marcio after 4750 cal years BP were probably caused by climatic changes
experienced in the Andes. As increasingly wet conditions are inferred from Andean
records after ca. 5000-4500 years BP (Wirrmann et al. 1988, Baker et al. 2001b), the
volume of the Amazon River, and hence seasonal flooding probably increased
rapidly, which compensated for the sea-level fall. Furthermore, the additional
influence of increased frequency of ENSO events after 5000 cal years BP (Moy et al.
2002, Riedinger et al. 2002) cannot be dismissed, as El Niño events are known to
cause drought in eastern Amazonia, the opposite is true for La Niña events (Marengo
et al. 1998, Aalto et al. 2003). Even though changes in insolation cannot account for
rapid climate changes, such as the one observed at the Amapá record, it is reasonable
that the stronger seasonality that the southern hemisphere has been experiencing in
the last 5000 years had some background influence on this paleoecological record.
As Lakes Tapera and Marcio stopped receiving contribution of river flood waters, it
is possible that they were relying on wet season precipitation more heavily to
maintain a high stand.
While the wet event that caused the resumption of sedimentation at Lake
Marcio at 4750 cal years BP can be explained by increased precipitation in the

155
Andes, the dramatic vegetation change that took place between 5300 and 4750 cal
years BP cannot. I propose that the replacement of the swamp forest by the savanna
was triggered by the very same coastal progradation (river channel migration) that
intensified the isolation of both lakes from flood waters during the Andean dry spell.
When the sediments transported by the Amazon were blocked by the high sea-level,
they accumulated near the mouth of the river, which caused the channel to migrate
further from the coast, adding more suitable area for colonization. As the local
climatic conditions were relatively dry, for the level of the Amazon River was
probably lower, the expansion of flooded savannas was facilitated.
The only fossil pollen records that present a similarly marked vegetation
change are located relatively near Amapá (Guyana, French Guyana, Suriname and
Belém) close to the ocean, and were subjected to the same effects of sea-level
changes (van der Hammen 1963, Wijmstra 1971, Tissot et al. 1988, Behling and
Costa 2001). Even though all these records show the same overall pattern of
vegetation change (forests and mangroves being replaced by savannas), the
interpretation of each record may be slightly different. While the vegetation changes
inferred from a sediment core from French Guyana were interpreted to be the result
of a coastal progradation due to sea-level changes (Tissot et al. 1988), the records
from Guyana (van der Hammen 1963), Suriname (Wijmstra 1971), and coastal Pará
(Behling and Costa 2001) were simply attributed to sea-level fall, which would cause
a similar effect. The apparent synchronism of these vegetation changes makes the
argument of a coastal progradation even more appealing.

156
Following the resumption of sedimentation at Lake Marcio, the dramatic
increase in the concentration of charcoal particles strongly suggests human
occupation near the site after 4700 years BP (Fig 2.5). However, because Poaceae
pollen percentages and concentrations increase before charcoal increases in
abundance, fire may not have induced the vegetation change. Consequently, it is
plausible to suppose that human occupation was not the primary trigger of the
vegetation change. Rather, the shift from flood forest toward a swamp savanna (or
flooded savanna) during the sedimentary hiatus is consistent with a lowering of local
water-tables, inferred from the loss of sedimentation between 5400 and 4700 years
BP. Nevertheless, it is clear that human impacts played an important role in the
maintenance of the new vegetation, as the return of wet conditions did not lead to the
re-establishment of local forests.
Finally, the resumption of sedimentation at Lake Tapera could have been
triggered by a combination of factors. First, a last sea-level oscillation is recorded at
the coast of Rio Grande do Norte around 2000 years BP (Bezerra and others 2003),
which would have caused the local water-table to rise. Second, for the last 2000
years the Austral summer insolation has been the highest of the Holocene (Fig 5.2),
which would cause increased precipitation during the wet season. Finally, even
though the influence of La Niña events on climate of eastern Amazonia is not well
established, they are known to increase white water discharge and to increase
precipitation in the headwaters of Beni and Mamoré Rivers (affluents of the Amazon
River) (Aalto and others 2003). Therefore, the intensification of ENSO in the late

157
Holocene (Moy et al. 2002, Riedinger et al. 2002) may have contributed to the wet
conditions that favored the resumption of sedimentation at Lake Tapera.
THE PRAINHA LAKES
Although the lakes near Prainha lie ca. 550 km upstream from the Amazon
mouth, they are only 40-70 m elevation above sea-level. Consequently, these lakes
are also influenced by Holocene sea-level variability. The overlapping ages of
ponding of the Prainha and Amapá lakes suggest that the rising water table caused by
sea-level rise in the early Holocene played an important role on the formation of both
lake systems (Fig 5.1). However, the Prainha lakes differ from those of Amapá as
they are not connected to river systems, and therefore, are not influenced by flood
waters or coastal progradation.
Even though it is apparent that the formation of the lakes near Prainha was
due to Holocene sea-level rise, it is very difficult to infer from these pollen records
further environmental changes caused by sea-level variation. However, the overall
record of Prainha shows increasingly wet conditions until ca. 6700 and 5700 years
BP, which would be consistent with a sea-level rise, although the forest expansion
recorded after 5000-4500 years BP would be contradictory to the sea-level fall.
Consequently, as the most striking feature recognized from the Prainha record is the
pattern of charcoal accumulation, in addition to the presence of maize pollen at Lake
Geral (Bush and others 2000), it is possible that the record of human occupation in
this region since at least 8200 years BP overwhelms the influence of sea-level

158
variation (Figs 3.5, 3.6, 3.8, 5.1). Charcoal particles are recorded in the lowermost
samples of Lakes Santa Maria and Saracuri (ca. 8300 years BP), and charcoal
abundance increases until 7340 and 7490 years BP, respectively. At Lake Geral, an
isolated peak of charcoal is recorded for the first time at ca. 6700 years BP. As peaks
of charcoal alternate between lakes through time, suggesting very local fires (as the
lakes are no more than 8 km apart), the possibility of these fires being natural can be
discarded. Because pollen of cultivars was found only in the sediments of Lake Geral
(maize), it is possible that Geral was the center of local agricultural activity, and
supported a more permanent settlement near the lake. The less consistent record of
charcoal and lack of cultivar pollen found at the other lakes suggest an irregular use
of these habitats.
As the Prainha site presents such a complex and strong record of human
occupation, it becomes very difficult to identify the influence of climatic factors such
as frequency of ENSO (El Niño and La Niña) events and insolation, as they too
could probably be overwhelmed by the anthropogenic signal. Nevertheless, both
Lakes Santa Maria and Saracuri record a phase of forest recovery between ca. 5000
and 3200 years BP. That this period coincides with relatively increased summer
insolation (Fig 5.2) suggests that the forest re-growth took place under increasingly
wet conditions, although a concurrent decrease in fire frequency is also observed at
this time (Figs 3.5 and 3.8). Wetter conditions could inhibit forest fires, but it is more
probable that the regrowth represents a quiet period in human activity rather than
climatic change. However, the accumulation of charcoal particles does not stop

159
completely, only decreases, which suggests that there were fires in the background.
That the resumption of sedimentation at Lake Marcio (Amapá) around ca. 4750 years
BP (Fig 5.1) coincides with the forest expansion at Lakes Santa Maria and Saracuri,
is further evidence that, in addition to the decreased human influence, as inferred
from the charcoal record, there maybe a climatic force driving the short period of
forest expansion.
After ca. 3200 years BP, there are no more records of forest expansion at
Lakes Santa Maria and Saracuri, although human impacts (based on charcoal
accumulation) appear to be generally reduced. It is possible that a regional drought
could have prevented the continued forest expansion in this area. The origin of such
a drought might lie in the concentration of El Niño events between ca. 3500 and
2500 years BP (Fig 5.4) (Moy et al. 2002). While La Niña brings flooding to areas in
floodplains, the direct local effect of El Niño in eastern Amazonia may be to induce
drought (Marengo and others 1998).
LAKE JACARÉ
It is obvious that Lake Jacaré, due to its location, was the least affected by
sea-level changes. Therefore, it is unlikely that the initial ponding (ca. 9040 years
BP) was due to the Holocene sea-level rise. Consequently, as the Boreal summer
insolation, although starting to decrease, was still high at 9040 years BP (Fig 5.2),
the overall wet conditions provided by a relatively wetter wet season probably
supplied enough moisture to initiate the lake formation (Fig 5.1). The Cariaco Basin

160
off the coast of Venezuela records wet conditions from ca. 10,000 to 5000 years BP
(Haug and others 2001). Increasingly wet conditions were also recorded at Lake
Valencia (Venezuela) when the saline lake turned into a freshwater lake at ca. 9500
years BP, and overflowed around 5500 years BP (Bradbury and others 1981; Leyden
1985; Salgado-Labouriau 1997). The Lake Miragoane in Haiti, recorded dry
conditions between 12,400 and 9300 years BP (Hodell and others 1991). However,
after 9200 years BP, climate became increasingly wet, which induced forest
expansion in the region (Hodell and others 1991). That several sites in the northern
hemisphere record wet conditions from ca. 9000 years BP until ca. 5000 years BP,
strongly suggests that climate was wetter than the present at this time, which
supports the influence of climate forcing on ponding of Lake Jacaré.
Increased quantities of carbonized particles after 7500 cal years BP suggest
the onset of fire in the region. The lack of records of human occupation this early in
Roraima (Ribeiro 1997) points to a natural origin for these fires. However, the
increasingly wet conditions between ca. 9000 and 6500 years BP as inferred from
this pollen record and from others elsewhere in the northern hemisphere suggest
otherwise (Bradbury and others 1981; Haug and others 2001; Hodell and others
1991; Leyden 1985; Salgado-Labouriau 1997). The marked decrease of Typha and
Zonorate pollen suggesting the transition from a swamp lake to a true lake at 6500
cal year BP (Figs 4.5 and 4.6), is consistent with a period of increasingly wet
conditions. Consequently, despite the lack of earlier records of human occupation in
Roraima, the fires are probably of anthropogenic origin. At present, the dry season in

161
this region lasts ca. 5 months, when precipitation is reduced to ca. 36 mm/month
(IBGE 1990, Barbosa 1997). Therefore, it is plausible to suppose that dry season
between 9000 and 4700 years BP was shorter than at present.
As a brief period of forest expansion is inferred from the pollen record
between ca. 5600 and 4700 years BP, it is probable that this period represents the
maximum of humid conditions. Alternatively, as the lowest charcoal accumulation
coincides with the peak of forest expansion, decreased human impacts probably
played an important role at this time. That the short period of forest expansion
recorded at Lake Jacaré (5700-4700 years BP) overlaps to some extent with that
recorded at Lakes Saracuri (ca. 5000-3200 years BP) and Santa Maria (ca. 4800-
3200 years BP), could be an indication of the climatic origin of this vegetation
change, although the importance of decreased human impacts cannot be disregarded.
However, as the climate of northern Roraima follows the northern hemisphere
pattern, which is opposite of the southern hemisphere, this apparent synchronism of
Lake Jacaré with the other sites is unclear.
Changes in sedimentation rates accompanied by a peak in quartz
accumulation following the period of forest expansion indicate a dry event and
possibly a sedimentary hiatus at 30 cm in the core of Lake Jacaré. Rough estimates
of sedimentation rates suggest that the sedimentary gap could have lasted from 4700
until 1700 years BP (Fig 4.9). The relatively drier conditions recorded at Lake Jacaré
would be consistent with an overall reduction in precipitation due to decreased
summer insolation in the northern hemisphere (Haug and others 2001). However, if

162
the dry conditions observed at Jacaré were triggered by changes in insolation, a
general trend toward dry conditions should have been recorded, instead of a marked
dry event, as changes in insolation are not likely to account for sudden climatic
shifts. A more likely explanation would be that increased frequency of ENSO events
in the last 5000 years (after ca. 4700 years BP) strongly affected climate in northern
Roraima (Moy et al. 2002, Riedinger et al. 2002). Because annual precipitation in
this region is the lowest of all the studied sites, it is probable that reduced
precipitation due to strong ENSO events would have caused more drastic
environmental changes than on the other sites. Nevertheless, only with additional
radiocarbon dates it will be possible to verify the existence, and perhaps duration of a
hiatus in the Lake Jacaré sediments.
In a wide context, the Jacaré record emphasizes the heterogeneity of
Holocene Neotropical savanna climates. That the majority of paleoecological records
from Colombia suggest increasingly wet conditions in the last 6000 years, with
savannas being replaced by forests after ca. 4000 years BP (Behling and others 1999;
Behling and Hooghiemstra 1998; Behling and Hooghiemstra 1999; Behling and
Hooghiemstra 2000; Behling and Hooghiemstra 2001; Berrio and others 2002), is a
clear indication that the study sites, even Lake Jacaré, behave in an opposite manner
to those of the Colombian savannas.
Paleoecological records from central Brazil show increasing rainfall after
7000-6000 years BP, as inferred from formation of swamps and marshes (Salgado-
Labouriau 1997). The continuing humid conditions allowed forest patches to expand

163
into cerrado areas until ca. 1400 years BP, although the cerrado vegetation was not
completely replaced by forests.
The fact that the study lakes show unique records of environmental changes
that do not agree with savanna records from either hemisphere suggests that regional
climatic controls may not be robust enough to generate the same pattern of
vegetation changes throughout such a wide region as the Amazonia.

164
CONCLUSIONS
In conclusion, even though each lake presents an individual environmental
history, as the factors influencing each site vary in importance, a regional climatic
pattern of increasingly wet conditions can be inferred from the paleoecological
records of Amapá, Prainha and Roraima sites. Despite the sedimentary hiatus
identified at Lakes Tapera and Marcio, and the possible hiatus at Jacaré, none of the
records suggest a widespread mid-Holocene dry event (contra Absy et al. 1991,
Ledru et al. 1998, Alexandre et al. 1999).
As changes in the vegetation during the Holocene were inferred from the
pollen records, it is clear that the proportions of savanna and forest did not remain
constant. That the studied sites present a pattern that is out of phase with the
Colombian sites it is also clear.
Even though the Amapá, Prainha and Roraima sites are relatively near each
other, no synchroneity was found between the vegetation changes recorded, with
Lakes Santa Maria and Saracuri being the only exception. Therefore, hypothesis 3
must be rejected.
H3: If vegetation changes are recorded, they will be synchronous between the study
sites.
Such a lack of synchroneity can be explained by the number of interacting
factors influencing each site at different scales. Additionally, as the Prainha site

165
presents such a strong record of human occupation, it is probable that the interaction
between all the factors became much more complex.
The fact that most of the records show human impacts as early as 8000 - 7500
years BP, with the exception of the Amapá site, supports the hypothesis that some
regions in Amazonia have been occupied for a long time (Roosevelt et al. 1991,
Roosevelt et al. 1996, Roosevelt 2000). However, only the record from Geral
contained unequivocal evidence of local agriculture, and therefore, was the only
record suggesting long-term occupation and exploitation.
As the records from Prainha and Roraima show human occupation
throughout the history of the lakes, it is still unclear if these savanna areas were
human made, as suggested by Prance and Schubart (1977). Even though Lake
Marcio records human impacts starting only after 4700 years BP, the vegetation
change still precedes the record of human occupation.
The final, and perhaps the most important, conclusion drawn from this study
is that Amazonian savannas have complex and independent histories. Consequently,
no unifying hypothesis can provide for the formation of all Amazonian savannas. As
complex and important climatic teleconnections exist between distant geographic
areas, the use of single paleoecological records to understand the climatic and
vegetational histories of such a large region should be strongly discouraged. As more
data emerge a better picture of Amazonian paleoecological history will become
available.

166
LITERATURE CITED
Aalto R, Maurice-Bourgoin L, Dunne T, Montgomery DR, Nittrouer CA, Guyot J.
2003. Episodic sediment accumulation on Amazonian flood plains influenced by
El Niño/Southern Oscillation. Nature 425:493-497.
Absy ML. 1979. A palynological study of Holocene sediments in the Amazon basin.
[dissertation]. Amsterdam: University of Amsterdam. 86 p.
Absy ML. 1985. The Palynology of Amazonia: The history of the forests as revealed
by the palynological record. In: Prance GT, Lovejoy TE, editors. Key
Environments of Amazonia. Oxford: Pergamon Press. p 72-82.
Absy ML, Clief A, Fournier M, Martin L, Servant M, Sifeddine A, Ferreira da Silva
M, Soubiès F, Suguio KT, van der Hammen T. 1991. Mise en évidence de quatre
phases d'ouverture de la forêt dense dans le sud-est de L'Amazonie au cours des
60,000 dernières années. Première comparaison avec d'autres régions tropicales.
Comptes Rendus Acad des Sci Paris, Series II 312:673-678.
Absy ML, Prance GT, Servant M, Miranda IS. 1997. Registros Palinológicos em
Sedimentos do Holoceno e Vegetação Atual em Roraima. In: Barbosa RI,
Ferreira EJG, Castellón EG, editors. Homem, Ambiente e Ecologia no Estado de
Roraima. Manaus: INPA. p 463-480.
Alexandre A, Meunier JD, Mariotti A, F. S. 1999. Late Holocene phytolith and
carbon-isotope record from a latosol at Salitre, South Central Brazil. Quat Res
51:187-194.
Anderson AB. 1981. White-sand vegetation of Brazilian Amazonia. Biotropica
13(3):199-210.
Athens JS, Ward JV. 1999. The late Quaternary of the western Amazon: climate,
vegetation and humans. Antiquity 73:287-302.
Bailey RC, Head G, Jenike M, Owen B, Rechtman R, Zechenter E. 1989. Hunting
and gathering in tropical rain forest: Is it possible? Amer Anthrop 91:59-82.
Baker PA, Rigsby CA, Seltzer GO, Fritz SC, Lowenstein TK, Bacher NP, Veliz C.
2001a. Tropical climate changes at millennial and orbital timescales on the
Bolivian Altiplano. Nature 409:698-701.

167
Baker PA, Seltzer GO, Fritz SC, Dunbar RB, Grove MJ, Tapia PM, Cross SL, Rowe
HD, Broda JP. 2001b. The History of South American Tropical Precipitation for
the Past 25,000 Years. Science 291:640-643.
Barbosa RI. 1997. Distribuição das chuvas em Roraima. In: Barbosa RI, Ferreira
EJG, Castellón EG, editors. Homem, Ambiente e Ecologia no Estado de
Roraima. Manaus: INPA. p 325-335.
Beaufort L, Garidel-Thoron T, Mix AC, Pisias NG. 2001. ENSO-like forcing on
oceanic primary production during the late Pleistocene. Science 293:2440-2444.
Behling H. 1996. First report on new evidence for the occurrence of Podocarpus and
possible human presence at the mouth of the Amazon during the late-glacial.
Veg Hist and Archaeobot 5:241-246.
Behling H. 2002. Impact of the Holocene sea-level changes in coastal, eastern, and
central Amazonia. Amazoniana 17(1/2):41-52.
Behling H, Berrio JC, Hooghiemstra H. 1999. Late Quaternary pollen records from
the middle Caquetá River basin in central Colombian Amazon. Palaeogeogr
Palaeoclimat Palaeoeco 145:193-213.
Behling H, Costa ML. 1997. Studies on Holocene tropical vegetation mangrove and
coast environments in the state of Maranhão, NE Brazil. Quat of S Amer and Ant
Pen 10(7):93-118.
Behling H, Costa ML. 2001. Holocene vegetational and coastal environmental
changes from the Lago Crispim record in northeastern Pará State, eastern
Amazonia. Rev of Paleobot and Palyno 114:145-155.
Behling H, Da Costa ML. 2000. Holocene environmental changes from the Rio
Curuá record in the Caxiuanã region, eastern Amazon Basin. Quat Res 53:369-
377.
Behling H, Hooghiemstra H. 1998. Late Quaternary paleoecology and
paleoclimatology from pollen records of the savannas of the Llanos Orientales in
Colombia. Palaeogeogr Palaeoclimat Palaeoeco 139:251-267.
Behling H, Hooghiemstra H. 1999. Environmental history of the Colombian
savannas of the Llanos Orientales since the Last Glacial Maximum from lake
records El Pinal and Carimagua. J of Paleolim 21:461-476.

168
Behling H, Hooghiemstra H. 2000. Holocene Amazon rainforest-savanna dynamics
and climatic implications: high resolution pollen record from Laguna Loma
Linda in eastern Colombia. J of Quat Sci 15(7):687-695.
Behling H, Hooghiemstra H. 2001. Neotropical savanna environments in space and
time: Late Quaternary interhemispheric comparisons. Interhemispheric Climate
Linkages:307-323.
Behling H, Keim G, Irion G, Junk W, Nunes de Mello J. 2001. Holocene
environmental changes in the central Amazon Basin inferred from Lago Calado
(Brazil). Palaeogeogr Palaeoclimat Palaeoeco 173:87-101.
Berrio JC, Hooghiemstra H, Behling H, Botero P, van der Borg K. 2002. Late
Quaternary savanna history of the Colombian Llanos Orientales from Lagunas
Chenevo and Mozambique: a transect synthesis. The Holocene 12(1):35-48.
Berrio JC, Hooghiemstra H, Behling H, van der Borg K. 2000. Late Holocene
history of savanna gallery forest from Carimagua area, Colombia. Rev of
Paleobot and Palyno 111:295-308.
Bertaux J, Fröhlich F, Ildefonse P. 1998. A new application of FTIR spectroscopy
for the quantification of amorphous and crystallized mineral phases. Examples of
organic-rich sediments. J Sed Res 68(3):440-447.
Bezerra FHR, Barreto AMF, Suguio K. 2003. Holocene sea-level history on the Rio
Grande do Norte State coast, Brazil. Marine Geology 96:73-89.
Binford MW, Kolata AL, Brenner M, Janusek J, Seddon MT, Abbott MB, Curtis JH.
1997. Climate variation and the rise and fall of an Andean civilization. Quat Res
47: 235-248.
Blunier T, Brook EJ. 2001. Timing of Millennial-scale climate change in Antarctica
and Greenland during the last glacial period. Science 291:109-112.
Bradbury JP, Leyden BW, Salgado-Labouriau ML, Lewis WMJ, Schubert C,
Binford M, Frey DG, Whitehead DR, Weibezahn FH. 1981. Late Quaternary
environmental history of Lake Valencia,Venezuela. Science 214:1299-1305.
Brown Jr. KS. 1977. Centros de evolução, refúgios Quaternários e conservação de
patrimônios genéticos na região neotropical: padrões de diferenciação em
Ithomiinae (Lepidoptera: Nymphalidae). Acta Amazônica 7(1):75-137.

169
Brown KSJ, Prance GT. 1987. Soils and Vegetation. In: Whitmore TC, Prance GT,
editors. Biogeography and Quaternary history in tropical America. Oxford:
Oxford Science Publications. p 19-45.
Bush MB. 2002. On the interpretation of fossil Poaceae pollen in the humid lowland
neotropics. Palaeogeogr Palaeoclimat Palaeoeco 177:5-17.
Bush MB, Colinvaux PA. 1988. A 7000-year pollen record from the Amazon
lowlands, Ecuador. Vegetatio 76:141-154.
Bush MB, Miller MC, De Oliveira PE, Colinvaux PA. 2000. Two histories of
environmental change and human disturbance in eastern lowland Amazonia. The
Holocene 10:543-554.
Bush MB, Miller MC, De Oliveira PE, Colinvaux PA. 2002. Orbital-forcing signal in
sediments of two Amazonian lakes. J of Paleolim 27:341-352.
Bush MB, Piperno DR, Colinvaux PA, Krissek L, De Oliveira PE, Miller MC, Rowe
W. 1992. A 14,300 year paleoecological profile of a lowland tropical lake in
Panama. Eco Monogr 62:251-276.
Clark JS. 1988. Particle motion and the theory of charcoal analysis: source area,
transport, deposition, and sampling. Quat Res 30:67-80.
Clark JS, Patterson WAI. 1997. Background and local charcoal in sediments: Scales
of fire evidence in the Paleorecord. In: Clark JS, Cachier H, Goldammer JG,
Stocks B, editors. Sediment Records of Biomass Burning and Global Change.
NATO ASI Series. Series 1: Global Environmental Change vol. 51. Berlin:
Springer. p 23-48.
Clark JS, Royall PD, Chumbley C. 1996. The role of fire during climate change in an
eastern deciduous forest at Devil's Bathtub, New York. Ecology 77(7):2148-
2166.
Cochrane MA, Schulze MD. 1999. Fire as a recurrent event in tropical forests of the
eastern Amazon: effects on forest structure, biomass, and species composition.
Biotropica 31(1):2-16.
Colinvaux PA, Bush MB. 1991. Tropical rainforest ecosystems as a resource for
hunters and gatherers. Amer Anthrop (93):153-160.
Colinvaux PA, De Oliveira PE, Bush MB. 2000. Amazonian and neotropical plant
communities on glacial time-scales: The failure of the aridity and refuge
hypothesis. Quat Sci Rev 19:141-169.

170
Colinvaux PA, De Oliveira PE, Moreno JE. 1999. Amazon Pollen Manual and Atlas.
New York: Harwood Academic Press. 332 p.
Colinvaux PA, De Oliveira PE, Moreno JE, Miller MC, Bush MB. 1996. A long
pollen record from lowland Amazonia: Forest and cooling in glacial times.
Science 274:85-88.
Colinvaux PA, Irion G, Räsänen ME, Bush MB, Nunes de Mello JAS. 2001. A
paradigm to be discarded: geological and paleoecological data falsify the Haffer
and Prance refuge hypothesis of Amazonian speciation. Amazoniana 16:609-646.
Cooper A. 1979. Muri and white sand savannah in Guyana, Surinam and French
Guiana. In: Specht RL, editor. Heathlands and related shrublands of the world, a
descriptive study. Amsterdam: Elsevier. p 471-481.
Costa LP. 2003. The historical bridge between the Amazon and the Atlantic forest of
Brazil: a study of molecular phylogeography with small mammals. J of Biogeogr
30:71-86.
Cronin TM. 1999. Principles of Paleoclimatology. Bottjer DJ, Bambach RK, editors.
New York: Columbia University Press. 560 p.
De Oliveira PE. 1992. A palynological record of late Quaternary vegetational and
climatic change in southeastern Brazil. [dissertation] Columbus: The Ohio State
University.
Desjardins T, Carneiro Filho A, Chauvel A. 1997. Flutuações do Limite Floresta-
Cerrado Durante o Holoceno em Roraima. In: Barbosa RI, Ferreira EJG,
Castellón EG, editors. Homem, Ambiente e Ecologia no Estado de Roraima.
Manaus: INPA. p 307-323.
Eiten G. 1972. The cerrado vegetation of Brazil. Bot Rev 38:201-341.
Erickson CL. 1999. Neo-environmental determinism and agrarian "collapse" in
Andean prehistory. Antiquity 73(281):634-642.
Erickson CL. 2000. An artificial landscape-scale fishery in the Bolivian Amazon.
Nature 408(9):190-193.
Erickson CL. 2001. Pre-Columbian roads of the Amazon. Expedition 43(2):21-30.
Faegri K, Iversen J. 1989. Textbook of Pollen Analysis. Chichester: Wiley.

171
Ferraz-Vincentini KR, Salgado-Labouriau ML. 1996. Palynological analysis of a
palm swamp in Central Brazil. J of S Amer Earth Sci 9:207-219.
Furley PA. 1999. The nature and diversity of neotropical savanna vegetation with
particular reference to the Brazilian cerrados. Glob Eco and Biogeogr 8:223-241.
Grimm E. 1992. TILIA Software, Version 1.12: Illinios State University.
Haberle SG. 1997. Upper Quaternary vegetation and climate history of the Amazon
Basin: Correlating marine and terrestrial pollen records. In: Flood RD, Piper
DJW, Klaus A, Peterson LC, editors. Proceedings of the Ocean Drilling Program,
Scientific Results. College Station, Texas: Ocean Drilling Program. p 381-396.
Haberle SG, Maslin MA. 1999. Late Quaternary vegetation and climate change in
the Amazon basin based on a 50,000 year pollen record from the Amazon fan,
PDP site 932. Quat Res 51:27-38.
Haffer J. 1969. Speciation in Amazonian forest birds. Science 165:131-137.
Haffer J, Prance GT. 2001. Climatic forcing of evolution in Amazonia during the
Cenozoic: On the refuge theory of biotic differentiation. Amazoniana.
Haug GH, Günther D, Peterson LC, Sigman DM, Hughen KA, Aeschlimann B.
2003. Climate and the collapse of Maya civilization. Science 299:1731-1735.
Haug GH, Hughen KA, Sigman DM, Peterson LC, Röhl U. 2001. Southward
migration of the Intertropical Convergence Zone through the Holocene. Science
293:1304-1308.
Heckenberger MJ, Kuikuru A, Kuikuru UT, Russell JC, Schmidt M, Fausto C,
Franchetto B. 2003. Amazonia 1492: pristine forest or cultural parkland? Science
301:1710-1714.
Hill MO, Gauch HG. 1980. Detrended correspondence analysis: an improved
ordination technique. Vegetatio 42:47-58.
Hodell DA, Brenner M, Curtis JH, Guilderson T. 2001. Solar forcing of drought
frequency in the Maya lowlands. Science 292:1367-1370.
Hodell DA, Curtis JH, Jones GA, Higuera-Gundy A, Brenner M, Binford MW,
Dorsey KT. 1991. Reconstruction of Caribbean climate change over the past
10,500 years. Nature 352:790-793.

172
Hoffmann WA, Moreira AG. 2002. The role of fire in population dynamics of
woody plants. In: Oliveira PS, Marquis RJ, editors. The cerrados of Brazil. New
York: Columbia University Press. p 159-177.
Hooghiemstra H. 1984. Vegetational and climatic history of the high plain of
Bogota, Colombia: A continuous record of the last 3.5 million Years. Vaduz:
Gantner Verlag. 368 p.
IBGE 1990. Unidades climáticas do Brasil. Rio de Janeiro: Instituto Brasileiro de
Geografia e Estatistica.
IBGE 1992. Vegetação atual. Rio de Janeiro: Instituto Brasileiro de Geografia e
Estatística.
Imbrie JD, Hays J, Martinson DG, McIntyre A, Mix A, Morley JJ, Pisias NG, Prell
WL, Shackleton NJ. 1984. The orbital theory of Pleistocene climate: support
from a revised chronology of the marine 18O record. In: Berger AL, Imbrie J,
Hays J, Kukla G, Saltzman B, editors. Milankovitch and Climate. Dordrecht,
Netherlands: Reidel. p 269-305.
Juggins S. 2003. C2 User guide. Software for ecological and palaeoecological data
analysis and visualisation. Version 1.3. Newcastle upon Tyne: University of
Newcastle.
Kastner TP, Goñi MA. 2003. Constancy in the vegetation of the Amazon basin
during the late Pleistocene: Evidence from the organic matter composition of
Amazon deep sea fan sediments. Geology 31(4):291-294.
Kennedy LM, Horn SP. 1997. Prehistoric maize cultivation at the La Selva
Biological Station. Biotropica 29:368-370.
Kipnis R. 1998. Early hunter-gatherers in the Americas: perspectives from Central
Brazil. Antiquity 72:581-592.
Kovach WL. 1992. Multivariate Data Analysis. In: Gauch Jr. HG, editor.
Multivariate Analysis in Community Ecology.: Cambridge University Press.
Laskar J. 1990. The chaotic motion of the solar system: A numerical estimate of the
chaotic zones. Icarus 88:266-291.
Ledru M-P. 1993. Late Quaternary environmental and climatic changes in Central
Brazil. Quat Res 39:90-98.

173
Ledru M-P, Bertaux J, Sifeddine A, Suguio K. 1998a. Absence of last glacial
maximum records in lowland tropical forest. Quat Res 49:233-237.
Ledru M-P, Mourguiart P, Ceccantini G, Turcq B, Sifeddine A. 2002. Tropical
climates in the game of two hemispheres revealed by abrupt climatic change.
Geology 30:275-278.
Ledru M-P, Salgado-Labouriau ML, Lorscheitter ML. 1998b. Vegetation dynamics
in southern and central Brazil during the last 10,000 yr B.P. Rev of Palaeobot and
Palyno 99: 131-142.
Leyden B. 1985. Late Quaternary aridity and Holocene moisture fluctuations in the
Lake Valencia Basin, Venezuela. Ecology 66:1279-1295.
Listopad C. 2001. Vegetational changes, fire history and human impact during the
last 6000 years: a paleoecological study of the Madre de Dios Province, Lowland
Peru. [thesis]. Melbourne: Florida Institute of Technology. 190 p.
Marchant R, Almeida L, Behling H, Berrio JC, Bush M, Cleef A, Duivenvoorden J,
Kappelle M, De Oliveira P, Oliveira-Filho AT and others. 2002. Distribution and
ecology of parent taxa of pollen lodged within the Latin American Pollen
Database. Rev of Paleobot and Palyno 121(1):1-75.
Marengo JA, Nobre CA, Sampaio G. 1998. On the associations between
hydrometeorological conditions in Amazonia and the extremes of the southern
oscillation. Bull Inst Franc Étud And 27(3):789-802.
Marengo JA, Rogers JC. 2001. Polar air outbreaks in the Americas: Assessments and
impacts during modern and past climates. In: Markgraf V, editor.
Interhemispheric Climate Linkages. San Diego: Academic Press. p 31-51.
Martin L, Bertaux J, Correge T, Ledru M-P, Mourguiart P, Sifeddine A, Soubiès F,
Wirrmann D, Suguio K, Turcq B. 1997. Astronomical forcing of contrasting
rainfall changes in tropical South America between 12,400 and 8800 cal yr B.P.
Quat Res 47:117-122.
Martin L, Fournier M, Mourguiart P, Sifedine A, Turcq B, Absy ML, Flexor J-M.
1993. Southern oscillation signal in South American paleoclimatic data of the
last 7000 years. Quat Res 39:338-346.
Martin L, Mörner NA, Flexor JM, Suguio K. 1986. Fundamentos e reconstruções de
antigos níveis marinhos do Quaternário. Boletim de Geociencias da Universidade
de São Paulo 4:1-161.

174
Maslin MA, Burns SJ. 2000. Reconstruction of the Amazon basin effective moisture
availability over the past 14,000 years. Science 290:2285-2287.
Mayle FE, Burbridge R, Killeen TJ. 2000. Millennial-scale dynamics of southern
Amazonian rain forests. Science 290:2291-2294.
McCune B, Grace JB. 2002. Analysis of Ecological Communities. Gleneden Beach,
Oregon: MJM. 300 p.
McCune B, Mefford MJ. 1999. PC-ORD. Multivariate Analysis of Ecological Data.
Version 4.0. Gleneden Beach, Oregon: MjM Software.
Meggers BJ. 1971. Amazonia, Man and Culture in a Counterfeit Paradise. Chicago:
Aldine.
Meggers BJ. 1987. The early history of man in Amazonia. In: Whitmore TC, Prance
GT, editors. Biogeography and Quaternary history in tropical America. Oxford:
Clarendon Press. p 151-174.
Miranda HS, Bustamante MMC, Miranda AC. 2002. The fire factor. In: Oliveira PS,
Marquis RJ, editors. The Cerrados of Brazil. New York: Columbia University
Press. p 51-68.
Miranda IS, Absy ML. 1997. A flora fanerogâmica das savanas de Roraima. In:
Barbosa RI, Ferreira EJG, Castellón EG, editors. Homem, Ambiente e Ecologia
no Estado de Roraima. Manaus: INPA. p 445-462.
Moy CM, Seltzer GO, Rodbell DT, Anderson DM. 2002. Variability of El
Niño/Southern Oscillation activity at millenial timescales during the Holocene
epoch. Nature 420:162-165.
Myers N, Mittermeier RA, Mittermeier CG, da Fonseca GAB, Kent J. 2000.
Biodiversity hotspots for conservation priorities. Nature 403:853-858.
Nelson BW, Ferreira CAC, da Silva MF, Kawasaki ML. 1990. Endemism centers,
refugia and botanial collection density in Brazilian Amazonia. Nature 345:714-
716.
Oliveira-Filho AT, Ratter JA. 2002. Vegetation physiognomies and woody flora of
the cerrado biome. In: Oliveira PS, Marquis RJ, editors. The Cerrados of Brazil.
New York: Columbia University Press. p 91-120.
Patterson WA, Edwards KJ, Maguire DJ. 1987. Microscopic charcoal as a fossil
indicator of fire. Quat Sci Rev 6:3-23.

175
Pennington RT, Prado DE, Pendry CA. 2000. Neotropical seasonally dry forests and
Quaternary vegetation changes. J of Biogeogr 27:261-273.
Piperno DR, Ranere AJ, Holst I, Hansell P. 2000. Starch grains reveal early root crop
horticulture in the Panamanian tropical forest. Nature 407:894-897.
Piperno DR, Stothert KE. 2003. Phytolith evidence for early Holocene Cucurbita
domestication in southwest Ecuador. Science 299:1054-1057.
Pitman NCA, Terborgh JW, Silman MR, Nunez PV, Neill DA, Cerón CE, Palacios
WE, Aulestia M. 2001. Dominance and distribution of tree species in upper
Amazonian terra firme forests. Ecology 82:2101-2117.
Prance GT. 1973. Phytogeographic support for the theory of Pleistocene forest
refugia in the Amazon basin based on evidence from distribution patters in
Caryocaraceae, Chrysobalanaceae, Dichapetalaceae and Lecythidaceae. Acta
Amazônica 3(3):261-263.
Prance GT, Schubart HOR. 1977. Nota preliminar sobre a origem das campinas
abertas de areia branca do baixo Rio Negro. Acta Amazonica 7:567-570.
Ribeiro PAM. 1997. Arqueologia em Roraima: histórico e evidências de um passado
distante. In: Barbosa RI, Ferreira EJG, Castellón EG, editors. Homem, ambiente
e Ecologia no Estado de Roraima. Manaus, AM: INPA. p 3-24.
Richey JE, Nobre C, Deser C. 1989. Amazon River discharge and climate
variavility: 1903 to 1985. Science 246:101-103.
Riedinger MA, Steinitz-Kannan M, Last WM, Brenner M. 2002. A 6100 14C yr
record of El Nino activity from the Galapagos Islands. J of Paleolim 27:1-7.
Roosevelt AC. 2000. The Lower Amazon: A dynamic human habitat. In: Lentz DL,
editor. Imperfect balance - Landscape transformations in the PreColumbian
Americas. New York: Columbia University Press. p 455-486.
Roosevelt AC, Housley RA, Imazio da Silveira M, Maranca S, Johnson R. 1991.
Eighth millennium pottery from a prehistoric shell midden in the Brazilian
Amazon. Science 254:1621-1624.
Roosevelt AC, Lima da Costa M, Lopes Machado C, Michab M, Mercier N,
Valladas H, Feathers J, Barnett W, Imazio da Silveira M, Henderson A and
others. 1996. Paleoindian cave dwellers in the Amazon: the peopling of the
Americas. Science 272:373 -384.

176
Roubik DW, Moreno E. 1991. Pollen and Spores of Barro Colorado Island.
Monographs in Systematic Botany, v. 36: Missouri Botanical Garden.
Salgado-Labouriau ML. 1973. Contribuição à Palinologia dos Cerrados. Rio de
Janeiro, Brazil: Academia Brasileira de Ciências.
Salgado-Labouriau ML. Pollen and spore rain in central Brazil; 1979; Schmidt,
Berlin. Federal Environmental Agency.
Salgado-Labouriau ML. 1997. Late Quaternary paleoclimate in the savannas of
South America. J of Quat Sci 12:371-379.
Salgado-Labouriau ML, Barberi M, Ferraz-Vicentini KR, Parizzi MG. 1998. A dry
climatic event during the Late Quaternary of tropical Brazil. Rev of Palaeobot
and Palyno 99:115-129.
Sanford RL, Saldarriaga J, Clark KE, Uhl C, Herrera R. 1985. Amazon rain-forest
fires. Science 227: 53-55.
Servant M, Fontes J-C, Rieu M, Saliège X. 1981. Phases climatiques arides
holocènes dans le sud-ouest de l'Amazonie (Bolivie). Comptes Rendus Acad Sci
Paris, Series II 292:1295-1297.
Sifeddine A, Martin L, Turcq B, Volkmer-Ribeiro C, Soubies F, Cordeiro RC,
Suguio K. 2001. Variations of the Amazonian rainforest environment: A
sedimentological record covering 30,000 years. Palaeogeogr Palaeoclimat
Palaeoeco 168:221-235.
Silva ELS. 1997. A vegetação de Roraima. In: Barbosa RI, Ferreira EJG, Castellón
EG, editors. Homem, Ambiente e Ecologia no Estado de Roraima. Manaus:
INPA. p 400-415.
Silva JMC, Bates JM. 2002. Biogeographic patterns and conservaton in the South
American cerrado: a tropical savanna hotspot. Bioscience 52(3):225-233.
Simões Filho F, Turcq B, Carneiro Filho A, De Souza AG. 1997. Registros
sedimentares de lagos e brejos dos campos de Roraima: implicações
paleoambientais ao longo do Holoceno. In: Barbosa RI, Ferreira EJG, Castellón
EG, editors. Homem, Ambiente e Ecologia no Estado de Roraima. Manaus:
INPA. p 295-305.
Smith TB, Wayne RK, Girman DJ, Bruford MW. 1997. A role for ecotones in
generating rainforest biodiversity. Science 276:1855-1857.

177
Stockmarr J. 1971. Tablets with spores used in absolute pollen analysis. Pollen
Spores 13:615-621.
Stuiver M, Reimer PJ. 1993. Extended 14C database and revised CALIB radiocarbon
calibration program. Radiocarbon 35:215-230.
Suguio K, Martin L, Bittencourt ACSP, Dominguez JML, Flexor JM, Azevedo AEG.
1985. Flutuações do nível relativo do mar durante o Quaternário Superior ao
longo do litoral Brasileiro e suas implicações na sedimentação costeira. Rev Bras
de Geoci 15:273-286.
Thompson LG, Davis ME, Mosley-Thompson E, Sowers TA, Henderson KA,
Zagorodnov VS, Lin P-N, Mikhalenko VN, Campen RK, Bolzan JF and others.
1998. A 25,000-year tropical climate history from Bolivian ice cores. Science
282:1858-1864.
Tissot C, Djuwansah MR, Marius C. Evolution de la mangrove en Guyane au cours
de l'Holocène etude palynologique.; 1988; Bordeaux. Inst Fran Pond. p 125-137.
van der Hammen T. 1963. A palynological study on the Quaternary of British
Guiana. Leidse Geologische Mededelingen 29:125-180.
van der Hammen T, Absy ML. 1994. Amazonia during the last glacial. Palaeogeogr
Palaeoclimat Palaeoeco 109:247-261.
van der Hammen T, Hooghiemstra H. 2000. Neogene and Quaternary history of
vegetation, climate and plant diversity in Amazonia. Quat Sci Rev 19:725-742.
Van't Veer R, Hooghiemstra H. 2000. Montane forest evolution during the last 650
000 yr in Colombia: a multivariate approach based on pollen record Funza-I. J of
Quat Sci 15:329-346.
Vanzolini PE. 1970. Zoologia sistemática, geografia e a origem das espécies.
Instituto Geografico Universidade de São Paulo Teses e Monografias 3:56.
Weng C, Bush MB, Athens JS. 2002. Holocene climate change and hydrarch
succession in lowland Amazonian Ecuador. Rev of Paleobot and Palyno 120:73-
90.
Wijmstra TA. 1971. The Palynology of the Guiana coastal basin. Amsterdam:
University of Amsterdam. 62 p.

178
Wijmstra TA, van der Hammen T. 1966. Palynological data on the history of tropical
savannas in northern South America. Leidse Geologische Mededelingen 38:71-
90.
Wirrmann D, Mourguiart P, De Oliveira Almeida LF. 1988. Holocene
sedimentology and ostracods distribution in Lake Titicaca - Paleohydrological
interpretations. Quat of S Amer and Ant Pen 6:89-128.