Field Sampling and Sediment Description

In order to obtain a continuous pollen record throughout the Holocene, 10 cores were sampled depths up to 10 m using a Russian sampler. However, only four cores presented suitable pollen concentration to a consistent statistical analysis. Three cores were taken from old fluvial terraces located about 10 m higher than the modern floodplain at the margin of the Branco River (Figure 1), and they were named RBN 30 (0°54′11.9′′N, 61°23′41.8′′W), RBN 82 (0°26′42.7′′N, 61°43′16.1′′W), and RBN 83 (0°24′29.4′′N, 61°43′47.0′′W). One core, RBN 87 (0°29′19.9′′N, 61°40′04.3′′W) was sampled from an abandoned channel (Figure 1). These sampling sites are dominated by dense rainforest.

Facies descriptions included observations of lithology, texture and structure, and they aimed interpretation of depositional environments. These sedimentary characteristics were recorded on a measured lithostratigraphic profile. Grain size for the interval with pollen data was determined in more detail by laser diffraction using a Laser Particle Size SHIMADZU SALD 2101 in the Laboratory of Chemical Oceanography/UFPA. Prior to grain size analysis, approximately 0.5 g of each sample was immersed in H₂O₂ to remove organic matter, and the residual sediments were disaggregated by ultrasound (França et al. Reference França, Francisquini, Cohen, Pessenda, Rossetti, Guimarães and Smith2012). The grain-size scale of Wentworth (Reference Wentworth1922) was used in this work with sand (2–0.0625 μm), silt (62.5–3.9 μm), and clay fractions (3.9–0.12 μm).

Pollen Analysis

For pollen analysis, 1.0 cm³ of sediment were collected in intervals of 2.5 (RBN 30) and 5 cm (RBN 82, 83, and 87), which totaled 85 samples. All samples were prepared using standard pollen analytical techniques including acetolysis (Faegri and Iversen Reference Faegri and Iversen1989). Sample residues were mounted on slides in glycerin jelly. Pollen and spores were identified by comparison with reference collections of about 4000 Brazilian forest taxa and various pollen keys (Salgado-Labouriau Reference Salgado-Labouriau1973; Absy Reference Absy1975; Markgraf and D’Antoni Reference Markgraf and D’Antoni1978; Roubik and Moreno Reference Roubik and Moreno1991; Colinvaux et al. Reference Colinvaux, De Oliveira and Patiño1999), jointly with the reference collections of the Laboratory of Coastal Dynamics–Federal University of Pará and ¹⁴C Laboratory of the Center for Nuclear Energy in Agriculture (CENA/USP) to identify pollen grains and spores. A minimum of 300 pollen grains were usually counted for each sample, but for some specific depths 200 or 250 grains were counted due to their low pollen concentrations. Microfossils consisting of spores, algae, and some fungi were also counted, but they were not included in the sum.

All pollen taxa present were identified, but only the most abundant were plotted in pollen diagrams in order to illustrate the different ecological groups. The pollen diagrams were statistically subdivided into assemblage zones using a square-root transformation of the percentage data, followed by stratigraphically constrained cluster analysis (Grimm Reference Grimm1987). The intervals not analyzed by pollen content along the stratigraphic profiles exhibit absolute absence of pollen.

Radiocarbon Dating

Based on lithostratigraphic changes, seven samples of sedimentary organic matter (10 g each) were selected for ¹⁴C analysis, two in each of the cores Rbn82, Rbn83, Rbn30, and one in core Rbn87. Dating was carried out at the BETA Analytic Laboratory and LACUFF (Fluminense Federal University–Brazil) using accelerator mass spectrometers (AMS).

At the BETA Analytic Laboratory, sample pretreatment and dating followed the Beta Analytic for Radiocarbon Dating Laboratory standard technique for accelerator mass spectrometry (AMS) dating. The pretreatment consisted of dispersing the sediment deionized water, followed by sieving through 180-micron sieve to recover any available plant material and to remove any stones, gravel or other course inorganic grains. The less than 180-micron size fraction was used in each case. Two 0.1N HCl baths where applied for 2 hr each at 70ºC to remove any carbonates. Samples were rinsed to neutral and dried at 100ºC for 12 hr, cooled and weighed. After homogenizing with mortar and pestle, a small portion of the sediment was tested under a microscope with 12M HCl to ensure complete removal of all carbonate. Each sample was subsampled and combusted to CO₂ in a closed chemistry line which has been purged of any CO₂ to a level below 10e–15 atoms (background levels). The CO₂ was then introduced into a reaction vessel containing an aliquot of cobalt metal catalyst. Hydrogen was introduced such that when the cocktail was heated to 500ºC, the CO₂ cracked to carbon (graphite). The graphite was pressed into a target for measurement in an AMS system. The AMS was calibrated to provide an accurate ratio of the ¹⁴C/¹³C ratio between the sample graphite and a modern reference (NIST–4990C, oxalic acid). Quality assurance (QA) samples were reacted simultaneously in the chemistry lab and measured simultaneously in the AMS. The analytical result was obtained as a fraction of the value of modern reference, corrected for isotopic fractionation using ¹³C/¹²C (δ¹³C) and ¹⁴C age calculated according to the conventions cited in Stuiver and Polach (Reference Stuiver and Polach1977). The QA samples were checked for accuracy and observed to fall within expectations for the laboratory to accept and report the sample results. The QA acceptance was defined as the total laboratory error known to be within 2σ. The equipment used in these analyses were the Thermo Delta-Plus isotope ratio mass spectrometer (δ¹³C precision of ± 0.3‰) and the 250 kV NEC single stage particle accelerator (AMS precision of ±0.001 – 0.004 fractions modern) (http://www.radiocarbon.com/PDF/Beta-AMS-Methodology.pdf).

The samples dated at the LACUFF were previously examined and physically cleaned under a stereomicroscope. The organic matter (humin) from sediment samples were extracted according to the laboratory standard pretreatment with the acid–alkali–acid method (Pessenda et. al. 1996). At LACUFF, the samples were converted to carbon dioxide organic samples and all gas samples were cryogenically purified. Graphitization method and other procedures in the LACUFF laboratory can be obtained in Macario et al. (Reference Macario, Gomes, Anjos, Carvalho, Linares, Alves, Oliveira, Castro, Chanca, Silveira, Pessenda, Moraes, Campos and Cherkinsky2013). Results are expressed as radiocarbon ages (¹⁴C yr BP) normalized to a δ¹³C of –25‰ PDB, presently denoted as VPDB, or percentage of modern carbon (pMC) (Stuiver and Polach Reference Stuiver and Polach1977). ¹⁴C ages were also reported as calibrated years [cal yr BP, 2σ using CALIB 7.1 software and SHcal13 curve (Reimer Reference Reimer2013) and CALIBomb (Reimer et al. 2004)].

Causes for Vegetation Stability During the Middle and Late Holocene

The modern plant distribution in the Amazonian region has been widely debated, with main hypotheses having considered the influence of climate (ter Steege et al. 2003; Ruter et al. Reference Ruter, Arzt and Vavrus2004), topography (Vormisto et al. Reference Vormisto, Phillips and Ruokolainen2000), soil (ter Steege et al. Reference ter Steege, Jetten and Polak1993), and geology (Räsänen et al. Reference Räsänen, Salo and Kalliola1987; Van der Hammen et al. Reference Van der Hammen, Duivenvoorden, Lips, Urrego and Espejo1992; Zani et al. Reference Zani, Rossetti, Cohen, Pessenda and Cremon2012). In this context, patches of open vegetation that are commonly found intermingled with the Amazonian rainforest have been interpreted as remnants of a vegetation cover formed due to the presence of a dry period during the late Pleistocene (Ledru et al. Reference Ledru, Ceccantini, Gouveia, Lopez–Sáez, Pessenda and Ribeiro2006) and early to middle Holocene (Freitas et al. Reference Freitas, Pessenda, Aravena, Gouveia, Ribeiro and Boulet2001; Pessenda et al. Reference Pessenda, Boulet, Aravena, Rosolen, Gouveia, Ribeiro and Lamotte2001, Reference Pessenda, Ribeiro, Gouveia, Aravena, Boulet and Bendassoli2004).

Some works propose changes in the physical environment as the main factor to the establishment of vegetation patterns in Amazonia through time, as well as other areas around the globe (Van der Hammen et al. Reference Van der Hammen, Duivenvoorden, Lips, Urrego and Espejo1992; Behling and Hooghiemstra Reference Behling and Hooghiemstra2000; Maslin and Burns Reference Maslin and Burns2000; Freitas et al. Reference Freitas, Pessenda, Aravena, Gouveia, Ribeiro and Boulet2001; Sifeddine et al. Reference Sifeddine, Marint, Turcq, Volkmer-Ribeiro, Soubiès, Cordeiro and Suguio2001; Mayle and Power Reference Mayle and Power2008). For instance, the climate changed gradually from dry to wetter after the Last Glacial Maximum (LGM), which appears to have been accompanied by changes in the fluvial discharge (Latrubesse and Ramonell Reference Latrubesse and Ramonell1994; Latrubesse Reference Latrubesse2002; Rigsby et al. Reference Rigsby, Hemric and Baker2009). An increment of up to 40% of humidity after the Younger Dryas was proposed for this region (Maslin and Burns Reference Maslin and Burns2000).

The prevalence of wet conditions since the middle Holocene has been proposed for several areas of the Brazilian Amazonia (Maslin and Burns Reference Maslin and Burns2000; Pessenda et al. Reference Pessenda, Boulet, Aravena, Rosolen, Gouveia, Ribeiro and Lamotte2001; Sifeddine et al. Reference Sifeddine, Marint, Turcq, Volkmer-Ribeiro, Soubiès, Cordeiro and Suguio2001; Rossetti et al. Reference Rossetti, Toledo, Moraes-Santos and Santos2004, Reference Rossetti, Toledo and Góes2005; Jacob et al. Reference Jacob, Huang and Disnar2007; Cordeiro et al. Reference Cordeiro, Turcq, Sifeddine, Lacerda, Silva Filho, Gueiros, Cunha, Santelli, Pádua and Pachinelam2011), which coincided with the expansion of tropical forest as we know it today. The melting of the global ice volume during the transition from the last glacial stage into the current interglacial phase culminated with the global rise in sea level of up to 5 m above the modern sea level in the middle Holocene (Shackleton Reference Shackleton1988), which consequently caused an increase of humidity. Climatic and sea-level changes during the Holocene influenced significantly the development of mangrove along coastal Amazonian areas (e.g., Cohen et al. Reference Cohen, Pessenda, Behling, Rossetti, França, Guimarães, Friaes and Smith2012).

The data presented here indicate a mixture of forest and herbaceous communities comparable to the modern vegetation mosaic of the study area (Figure 6). They also suggest a wet period at least over the last 4808–4886 cal yr BP, without any indication of an expressive vegetation change as a consequence of any eventual middle Holocene dry event.

Some hypotheses may be raised to justify the stability of vegetation patterns in the study site since the middle Holocene. Hence, one possibility is that many Amazonian areas were not under the effect of dry climates during this time frame. In fact, speleothem oxygen isotope indicates that ecosystems in western Amazonia have not experienced prolonged drying over the past 94,000 years (Mosblech et al. Reference Mosblech, Bush, Gosling, Hodell, Thomas, Van Calsteren, Correa-Metrio, Valencia, Curtis and Van Woesik2012). In addition, an integrated approach combining pollen and diatom data from the Marajó Island, located at the mouth of the Amazon River in northeastern Amazonia, recorded no significant change in vegetation during the late Pleistocene through the early/mid Holocene (Castro et al. Reference Castro, Oliveira, Rossetti and Pessenda2013). A drastic middle Holocene change from forest to grassland vegetation was related by these authors to sedimentary influence resulting from coast progradation due to a drop in relative sea-level rather than a drier climate, because these are dominantly C₃ herbs typical of wet areas (see also Miranda et al. Reference Miranda, Rossetti and Pessenda2009 and Castro et al. Reference Castro, Rossetti and Pessenda2010). C and N isotope analyses of organic matter from sediments from an area of northern Amazonia located only around 30 km to the east of the study area (Zani et al. Reference Zani, Rossetti, Cohen, Pessenda and Cremon2012) led to state asyncronical development of open vegetation during the middle to late Holocene which was controlled by sedimentary dynamics, rather than climatic changes. In addition, no evidence of mid-Holocene dry climate was found in lakes near Prainha, central Amazonia (Bush et al. Reference Bush, Miller, de Oliveira and Colinvaux2000). Furthermore, a continuous record of forests was documented in the Caxiuaná area, northeastern Amazonia, for the last 7800–7940 cal yr BP (Behling and Costa Reference Behling and Costa2000). Lake Calado, near Manaus, displayed dense tropical forest since 8280–8380 cal yr BP (Behling et al. Reference Behling, Cohen and Lara2001). Several works stated that the Amazonia rainforest was most likely undisturbed during the last 50,000 years (e.g., Colinvaux et al. Reference Colinvaux, De Oliveira, Moreno, Miller and Bush1996, 2001; Bush et al. 2004; Irion et al. Reference Irion, Bush, Nunes de Mello, Stüben, Neumann, Müller, Morais de and Junk2006; Toledo and Bush 2007; Mayle and Power Reference Mayle and Power2008).

Another possibility to consider is that the impact of the dry period, claimed by several authors to have occurred in this region during the early and middle Holocene, did not influence this particular site of the Amazonian forest. It is interestingly to mention that, in contrast to many other Amazonian areas, this particular region currently displays a peculiar climatic regime strongly influenced by the periodic dislocation of the ITCZ. This results from changes in the Atlantic meridional overturning circulation (Chiang and Bitz Reference Chiang and Bitz2005) and/or tropical air–sea interactions, such as the El Niño and La Niña events (Broecker Reference Broecker2003). Therefore, one possibility is that, while a dry climate was in effect in other Amazonian areas, this site may have remained wet at least since the mid-Holocene under influence of the ITCZ dislocation dynamics.

On the other hand, considering the hypothesis of a dry Amazonia during the early and mid-Holocene, then a further hypothesis to be considered is that stabilization of the sea-level about 6000–5000 cal yr BP (Cohen et al. Reference Cohen, Souza Filho, Lara, Behling and Angulo2005b) may have attenuated the climate change, causing no impact on the vegetation cover. Existing RSL curves present dissimilar patterns for the late Holocene in Brazil. The first suggests a RSL fall in the last 5100 yr BP (5647–5999 cal yr BP) with two oscillations (Suguio et al. Reference Suguio, Martin, Bittencourt, Dominguez, Flexor and Azevedo1985a). Angulo and Lessa (Reference Angulo and Lessa1997) and Angulo et al. (Reference Angulo, Giannini, Suguio and Pessenda1999) disagree with the low RSL between 4100–3800 yr BP (~4560– ~4130 cal yr BP) and 3000–2700 yr BP (~3100–~2800 cal yr BP). Tomazelli (Reference Tomazelli1990) suggests a RSL rise in the last 1000 years in the State of Rio Grande do Sul State.

Regarding the northern Brazilian littoral, the RSL stabilized at its current level between ~7000 and ~5000 cal yr BP (e.g., Behling and Costa Reference Behling and Costa2001; Behling et al. Reference Behling, Cohen and Lara2001; Behling Reference Behling2002, Reference Behling2011; Cohen et al. Reference Cohen, Behling and Lara2005b; Souza Filho et al. Reference Souza Filho, Martins and Costa2006; Vedel et al. Reference Vedel, Behling, Cohen and Lara2006). The history of RSL fluctuations on northern Brazilian littoral seems to have been affected by tectonic activity during the late Pleistocene and Holocene (Rossetti et al. Reference Rossetti, Valeriano, Goes and Thales2008, Reference Rossetti, Souza, Prado and Elis2012). The post-glacial sea-level rise, combined with tectonic subsidence, caused a marine transgression during the late Pleistocene and early Holocene along the northern Brazilian coast (Cohen et al. Reference Cohen, Pessenda, Behling, Rossetti, França, Guimarães, Friaes and Smith2012). Between ~5100 and ~1000 cal yr BP, the RSL at the northern Brazil was probably never higher than 0.6 m above the current level (Cohen et al. Reference Cohen, Souza Filho, Lara, Behling and Angulo2005b). A work from the Lago do Crispim in the northeastern State of Pará, which is located next to the present coastline and only 1 m above msl, suggested no flooding during the last ~8800 cal yr BP (Behling and Costa Reference Behling and Costa2001).

At the global scale, sea-level is the dominant driver of the water table depth (Fan et al. Reference Fan, Li and Miguez-Macho2013). Then, sea-level changes during the Holocene might have influenced the shallow groundwater effects on terrestrial ecosystems by sustaining river base-flow and root-zone soil water in the absence of rain (Fan et al. Reference Fan, Li and Miguez-Macho2013). Consequently, regarding a stable relative sea-level along the northern Brazilian littoral during the last 6000 cal yr BP (e.g., Behling et al. Reference Behling, Cohen and Lara2001; Cohen et al. Reference Cohen, Behling and Lara2005b, Reference Cohen, Pessenda, Behling, Rossetti, França, Guimarães, Friaes and Smith2012; Vedel et al. Reference Vedel, Behling, Cohen and Lara2006), the water table depth may have favored the establishment and maintenance of the Amazon lowland forest during the middle and late Holocene. Such event may have created an ecological barrier between a narrow aquatic zone, favoring the development of Amazonian lowland forests at the feet of a large belt of topographic higher hinterlands.

Eustatic sea-level changes since the Pleistocene have caused significant impact on the extent and composition of the Amazonian floodplain. As a consequence of the increased gradient, caused by the decline of ca. 120 m in sea level during the LGM, the rivers cut deeply into their beds and a large volume of sediments were deposited in the Amazon fan (Junk Reference Junk1997). At this time, river valleys became incised, and may have filled with fluvial deposits. Alternatively, the valleys may have remained essentially empty until the next transgression (Walker Reference Walker1992). Geomorphological, sedimentological, and hydrological studies conducted along the Amazonian region upstream and downstream of the Negro River indicate that the várzea (floodplain) reached its present extent after sea-level stabilization, at 5000–6000 cal yr BP ( Irion et al. Reference Irion, Junk and Mello JASN1997; Behling et al. Reference Behling, Arz, Pätzold and Wefer2002). Similar conclusion emerged from detailed examination of riverbed morphologies at the Amazon mouth (Junk Reference Junk1994; Irion et al. Reference Irion, Junk and Mello JASN1997; Vital et al. Reference Vital, Stattegger, Posewang and Theilen1998).

Regarding the present work, in order to obtain a pollen record since the early Holocene, some cores were sampled with depths up to 10 m. However, all ¹⁴C dates obtained to the studied deposits indicated mid- and late Holocene. Although this study has not presented a pollen record since the early Holocene, the sediment deposits accumulated since ~4800 cal yr BP, from the base of the core RBN–83, may be an evidence for the time of the flood plain stabilization in a new level along the Branco River. Considering this hypothesis, the sea level rise during the early Holocene and its stabilization at the current level in the northern Brazilian coast, between 6000 and 5000 cal yr BP, may have caused the establishment of a fluvial terrace for accumulation of that sediments since the middle Holocene. Then, the absence of early Holocene sediments in the studied fluvial terrace may be related to the stabilization of the relative sea level in its modern level just in the middle Holocene. Older sedimentary deposits may have been eroded during the relative sea-level rise. Similar process was recorded for the Cardoso Island, São Paulo (Pessenda et al. Reference Pessenda, Vidotto, De Oliveira, Buso, Cohen, Rossetti, de Ricardi-Branco and Bendassolli2012).

Today, the Amazon River and its tributaries are affected by dynamic tides, but without saltwater intrusion due to the high freshwater discharge, as far as 1100 km upstream of its mouth (Kosuth et al. Reference Kosuth, Calléde, Laraque, Filizola, Guyot, Seyler, Fritsch and Guimarães2009; Freitas et al. Reference Freitas, Silva, Souza Filho, Ogston, Nittrouer and Asp2014). On the coastal zone, the stabilization of sea-level allowed development of wetlands (Cohen et al. Reference Cohen, Pessenda, Behling, Rossetti, França, Guimarães, Friaes and Smith2012), while the rainfall inland favored expansion of forest vegetation, the development of perennial streams, and the diversification of terrestrial and aquatic fauna (Meggers Reference Meggers2007).

Therefore, if a dry climate occurred in Amazonia during the Holocene, the most likely is that the stabilization of the relative sea-level since the middle Holocene, may have contributed to the development of the Amazonian lowland forest, having attenuated the impact of the middle Holocene dry period on vegetation in areas under most fluvial influence. In this case, no major vegetation changes are expected after the stabilization of relative sea-level in areas affected by the Amazon drainage system. Studies carried out in the Amazon have shown a buffer effect for areas close to rainforest, where the forest reveals mechanisms of protection against drier periods (Pessenda et al. Reference Pessenda, Valencia, Martinelli and Cerri1996).