Holocene histories of biome stability in northern Amazonian savannas

Abstract

Paleoecological analysis of three lake sediment cores from the Roraima savannas in northern Brazil revealed systems dominated by Poaceae pollen throughout most of the Holocene. A slight increase of palms and woody taxa, probably linked to wetter conditions, is observed during the last 1000 years but is more noticeable after 300 and 150 calibrated years BP (yr BP) respectively. Charcoal was present throughout affirming the importance of fire for the landscape but showed the highest values mostly in the last millenium. Poaceae pollen size spectra varied considerably, showing more variability than the overall pollen record. Despite evidence of some climate change in the Holocene, these savannas were stable systems. The sedimentary records showed gaps in deposition between 10,000 and 7800 yr BP and between 2500 and 1200 yr BP, suggesting dry periods during which lakes most likely dried out or became impermanent and highlighting the sensitivity of these lakes to local water table variation. The establishment of the modern climate in the region is evident after 1000 yr BP.

Keywords

Amazonian savannas biome stability charcoal fossil pollen Holocene poaceae pollen size

Introduction

While most of Amazonia is covered by tropical rainforests, enclaves of edaphic and climatic savannas add to the diversity of the system. These disjunct savannas, which are often on very nutrient-poor soils, have long intrigued biogeographers as they have been variously hypothesized to be remnants of formerly more extensive savannas (Bueno et al., 2017; Pennington et al., 2004), stepping-stones for species dispersal (Gosling and Bush, 2005), and the source area of species should Amazonia become a semi-desert due to climate change (Cowling et al., 2004; Shukla et al., 1990).

For several decades there has been growing concern that the precipitation regimes of Amazonia could be destabilized by the synergy of deforestation and climate change (Lovejoy and Nobre, 2018). If the moisture flow that progresses from east to west across Amazonia via cycles of evapotranspiration and precipitation is interrupted, all areas of the western Amazon and much of the tropical Andes would become much drier (Salati, 1985; Weng et al., 2018). Should that happen, it has been suggested that large areas of Amazonia, especially in the east, could see a replacement of tropical rainforest with seasonal forest or savanna (Cox et al., 2004; Malhi et al., 2009). Under such a drying scenario, and should there be a major replacement of forest with grassland, the savannas that lie at the modern periphery of Amazonia would provide the pool of native species that would be favored by the new conditions.

The expansion of mid-Holocene grasslands at the expense of Amazonian forest cover was observed in paleoecological records from within forest-savanna transitional areas of eastern and southern Amazonia (e.g. Carajas Plateau) (Absy et al., 1991; Fontes et al., 2017; Hermanowski et al., 2012). Nevertheless, it is not certain that climate change alone caused the signal observed at Carajas, as it was accompanied by fire, which could have been set by people (Gosling et al., 2021; Nascimento et al., 2022). Another setting showing a profound mid-Holocene dry event was the Altiplano of Peru/Bolivia, which received its moisture from Amazonia. Lakes in this region fell to their lowest levels of the last 100,000 years about 5500 years ago (Baker and Fritz, 2015). Other lowland Amazonian settings, especially those in the northern hemisphere, did not show such a strong drying (Nascimento et al., 2019), and even shallow lakes maintained their volume (Bush et al., 2000) .

Probably the best-known controversy surrounding Amazonian savannas, however, was their purported expansion during ice-ages as envisaged in the Refugial Hypothesis of Amazonian speciation (Haffer, 1969). It was suggested that extreme ice-age aridity caused rainforests to be replaced by savanna, leaving only the wettest pockets with refugial forests. Allopatric speciation within resulting forest fragments was postulated to explain Amazonian diversity (Brown, 1987). Many studies have subsequently cast doubt on this hypothesis, as it became clear that glacials were not especially arid (Colinvaux and DeOliveira, 2001; Wang et al., 2017), speciation does not date to the period of most extreme ice ages (Ribas et al., 2012; Santos et al., 2009; Wang et al., 2017), there was no widespread fragmentation of forest and savanna expansion (Bush and Oliveira, 2006), and that river dynamics may have played a larger role than originally envisaged (Musher et al., 2022; Thom et al., 2020). Indeed, contra the Refugial Hypothesis, these driest times may have been interglacials rather than glacials (Cheng et al., 2013; Hanselman et al., 2011; Rodbell et al., 2022). For example, extremes of heat and drought have been inferred primarily during interglacial times in Andean records (D’Agostino et al., 2002; Rodbell et al., 2022). Savanna expansions occurred during both the Pleistocene and Holocene along ecotonal margins (Absy et al., 1991; Fontes et al., 2017; Hermanowski et al., 2012; Mayle et al., 2000), but there was no evidence of such expansion in less seasonal Amazonian forest.

For many years, biogeographers have postulated north-south savanna connections across or around the modern rain-forested areas of Amazonia (Toby Pennington et al., 2000; Wüster et al., 2005). Such connectivity would explain the origins of currently disjunct floras and faunas (Werneck et al., 2012). Rather than broad corridors and complete connectivity, the formation of spatially and temporally discontinuous habitats that could support some savanna taxa has been offered as an alternative explanation (De Oliveira et al., 2020; Gosling and Bush, 2005).

Paleoecological data and models confirm that dry periods during the Quaternary promoted savanna expansion at forest-savanna ecotonal boundaries especially along the southern edge of Amazonia (Mayle et al., 2000; Smith et al., 2022; Smith and Mayle, 2018). The most recent such event took place between c. 10,000 and 5000 years ago when orbital geometry provided reduced (increased) insolation to the southern (northern) hemisphere and induced the strongest seasonality of the Holocene c. 9000 years ago (Mayle and Power, 2008). These events caused the Intertropical Convergence Zone to migrate northwards, weakening the South American summer monsoon (SASM) making southern and eastern Amazonia drier, but probably having less effect on northern Amazonian precipitation (van Breukelen et al., 2008).

South American savannas mainly lie outside Amazonia, and an analysis of Holocene savanna extents revealed a general pattern of peak areal extent c. 10,000–5000 years BP . Southern hemisphere savannas showed a marked contraction of savannas c. 4000 yr BP; a pattern that strengthened in the last 1000 years as climates became wetter (Smith and Mayle, 2018). In their review of pollen records, Smith and Mayle (2018) had no data from the northern central Amazonian savannas, and so it was not clear how these savannas may have changed during the Holocene. Fossil pollen and limnological data indicated that the mid-Holocene (8000–4000 cal BP) in northwestern Brazil was relatively wet (Bush et al., 2002; Nascimento et al., 2019), with a trend toward drier conditions, while that of southern Amazonia was dry, becoming wetter (Absy et al., 1991; Fontes et al., 2017; Hermanowski et al., 2012; Mayle et al., 2000). Cave isotope data suggested that during the last ice age eastern and western Amazonian sites were anti-phased in terms of their precipitation histories (Cruz et al., 2009). As the savannas of northern Brazil lie centrally within the east-west axis identified by Cruz et al. (2009), no prediction of change can be made about their pattern of drying.

Here we present two new paleoecological records and a reinterpretation of a previously studied lake (de Toledo, 2010) from the savannas of Roraima, in northern Brazil. One of these lakes Caracaranã, was previously the subject of research into its carbon: nitrogen (C:N) ratios, δ¹³C and microcharcoal records (Cordeiro et al., 2014; Turcq et al., 2002). This analysis inferred a period of active fires in the early Holocene, C4 plants as the primary source of carbon, a wet mid-Holocene, and human-induced changes, including fire activity after c. 2500 cal BP. The nearest existing fossil pollen records to our sites that span most of the Holocene came from peat bogs in the Gran Sabana of Venezuela, a region also characterized by the presence of savanna-forest mosaics (Montoya and Rull, 2011; Rull, 2007). These bogs lay at c. 800–900 m elevation in much cooler and wetter conditions than those around our lakes (Fick and Hijmans, 2017; Huber, 1995) (Supporting Information – Appendix 1). Those savannas formed following an abrupt vegetation change from mesic forest to savanna c. 10,000 years ago as a result of drying (Rull, 2007; Rull et al., 2013). The Gran Sabana system appeared relatively stable throughout the rest of the Holocene (Montoya and Rull, 2011; Rull, 2007).

In the dense forests of Amazonia, fires are seldom, if ever, a natural phenomenon (Gosling et al., 2021). Savannas, however, are fire-maintained systems (Eiten, 1982) that should show a near-continuous history of fire. Sedimentary charcoal is well established as a proxy for fire in forested settings, though the quality of such records in savanna settings is less well documented. Lower burn temperatures and fine fuel (grasses rather than coarse woody material), which characterize savanna fires, may lead to less charcoal deposition than when a forest burns (Leys et al., 2017). At the same time, the occurrence of fire adapted taxa could imply few tree deaths and little or no charcoal presence thereafter (Cochrane and Laurance, 2002; Pivello, 2011). If forest and savanna could exist as alternative stable states, with moisture as the primary determinant, the presence of fire asserts a hysteresis on transitions between these systems (Sternberg, 2001). Consequently, even if precipitation increases past the point where forests could re-establish, it is not until considerably wetter conditions eliminate fire, that forest could colonize the area. Thus, once established, savannas would be expected to be rather stable ecosystems.

Grasses of the Poaceae family are an important floristic component of savannas (Eiten, 1982). Grassland floras can range from low-diversity swampy systems to high-diversity systems (Murphy et al., 2016) in which grazing, and fire reduce opportunities for competitive exclusion. A long-term frustration for palynologists has been the inability to determine differences between grass assemblages as, with the exception of crops, Poaceae pollen cannot be separated at the genus level based on their morphology (Beug, 1961). Thus, while perhaps 60%–80% of the pollen from a savanna is Poaceae, we have no insights into changes between samples in that portion of the assemblage. Poaceae pollen grain size analysis offers some potential insights. Different grass species produce pollen of a given size ± 2 µm (Salgado-Labouriau and Rinaldi, 1990). The pollen input from a given community of grasses should produce a multi-modal histogram that reflects the balance of individuals of all species contributing to the spectrum (Flenley et al., 1991). As species rise or fall in abundance it is likely that the overall histogram of Poaceae pollen sizes will reflect that change. This approach was used in Easter Island to reveal Poaceae pollen sizes changing with deforestation (Flenley et al., 1991). Subsequently, grass pollen size spectra were used to distinguish different types of grasslands across elevational gradients (Schüler and Behling, 2011a, 2011b), precipitational gradients in southern Brazil (Radaeski and Bauermann, 2018), vegetation types (Radaeski et al., 2016) and in identifying polyploid versus diploid grasses (Jan et al., 2015). Here we will use Poaceae pollen size spectra to elicit changes in Poaceae assemblages within the savannas of Northern Brazil during Holocene.

Our study seeks to answer three questions: (1) Were there significant changes in the composition of northern Amazonian savanna vegetation during the Holocene? (2) Does fire intensity increase during the course of the Holocene? (3) Were there significant changes in the Poaceae communities during the Holocene?

Site description

In Roraima, a Brazilian state in northern Amazonia, there is an area of edaphic savanna that occupies c. 67,000 km². Underlain by the Boa Vista Formation, this region is characterized by deflational basins between paleodunes that were probably active in the Pleistocene (Latrubesse and Nelson, 2001), and seasonal soil-moisture deficits (Frost, 1968; Meneses et al., 2007; Nimer, 1972) .

Lake Caracaranã (longitude: −59.78228, latitude :3.84375, elevation 105 m; Figure 1) is the largest lake in the savannas of Roraima. The lake is oval, 1 x 0.6 km in size, flat bottomed, with a maximum depth of 5 m. Lake Caracaranã has a complex watershed, receiving drainage during the brief wet season through a system of small impermanent lakes. This system is fed by the creek (igarape) Buritizal do Nambi which receives water from the Surumú hills in the northwest (Simões et al., 2010) . Lake Jacaré (longitude: −59.066667, latitude: 3.800, elevation 84 m, Figure 1), located just 7 km from Caracaranã, is a small eutrophic water body with 410 × 300 m of open water, flat-bottomed, and 4 m deep. The lake lies within savanna, however there is presence of scattered patches of seasonal semi-deciduous forest, gallery forests and swamps in the area (De Toledo, 2004). Lake Indigena (longitude: −60.73871, latitude:3.514917, elevation 88 m; Figure 1) is a small lake c. 300 m in diameter, 4 m deep, and rests c. 111 km SW from Caracaranã, located in the core of the actual savanna area, between the rivers Parime and Uraricoera; the hydrology of this lake is also reliant on inputs from small creeks (igarapes) that supply water during the wet season.

Figure 1.

(a) Location of the three studied lakes and other paleoecological records discussed in the text. Lakes Jacaré, Caracaranã and Indigena are denoted with red diamonds. Studies in the Gran Sabana, Mapauri and El Paují, are showed with light blue circles (Montoya et al., 2011; Rull, 2007; Rull et al., 2013). Records from the Llanos Orientales of Colombia-Venezuela are symbolized with blue dark crosses (Behling and Hooghiemstra, 1998, 2000; Berrio et al., 2002). Late Holocene studies of Roraima (Fazenda Cigana and Terra Indigena Aningal) are portrayed with a black square (da Silva Meneses et al., 2013) and paleoclimatic records are displayed with purple stars, Cariaco Basin and Paraiso Cave (Haug et al., 2001; Wang et al., 2017). (b and c) Satellite view of the coring locations at different scales. Maps were generated in ArcMap10.8. See Supporting Information-Appendix 2 for landscape pictures of Caracaranã and Indigena.

The study area receives c. 1500 and 1700 mm of precipitation annually (Nimer, 1972), with the former value for the eastern lakes and the latter for Indigena. According to the Köppen classification, the climate of the region is tropical Aw (semi-humid). The driest months occur between December and March and the rainiest between May and August (Barbosa and Miranda, 2005; Reinaldo and Ciro, 2011), with a strong dry season of up to 6 months (Cordeiro et al., 2014). The temperature regime is stable with a range of 26°C–28°C (average per year); with October and November being the hottest months.

The lakes lie near an edaphic ecotone where Amazonian rainforest suddenly shifts to grassland (Cordeiro et al., 2014). The vegetation of the region is typically open grassland, scrubby savanna and open woody savanna, however, gallery forests line watercourses, and patches of semi-deciduous forest can form in wet areas (Reinaldo and Ciro, 2011). The forested waterways are rich in Caesalpiniaceae, Mimosaceae, Arecaceae, Chrysobalanaceae, and Lauraceae (Barbosa et al., 2007). Human impacts on the system are high with extensive habitat destruction and increased fire frequency (Reinaldo and Ciro, 2011). The open vegetation is mainly dominated by Poaceae and Cyperaceae (da Silva Meneses et al., 2013) with shrubs of Curatella americana (Dilleniaceae) and Byrsonima spp (Malpighiaceae) (Barbosa and Miranda, 2005; Miranda et al., 2003). Marshes and palm swamps form in low-lying areas and are often dominated by the canopy palm Mauritia flexousa (da Silva Meneses et al., 2013).

Materials and methods

Lakes Caracaranã and Indigena were cored in 2019 using an Aquatic Research Instruments universal percussion core sampler, while Lake Jacaré was cored in 2000 using a Colinvaux-Vohnout piston corer. Data from Jacaré was previously published by de Toledo (2010), however the present analysis include two new radiocarbon dates which modify the age-model and therefore the interpretation. In all cases, the corer was deployed from a wooden platform supported by inflatable boats. The upper 15 cm of the cores from Caracaranã and Indigena were extruded in 1 cm increments in the field. All samples were sent to the Paleoecology Laboratory at the Florida Institute of Technology for analysis. Subsamples of the upper 15 cm were sent to the University of New Hampshire for ²¹⁰Pb dating, and four additional samples per core were sent to Direct AMS laboratories for ¹⁴C radiocarbon dating. In Lake Jacaré, three samples were sent to the INSTAAR – AMS Radiocarbon Laboratory at the University of Colorado at Boulder, and two new additional samples were sent to the National Ocean Sciences AMS Facility at the Woods Hole Oceanographic Institution. For each lake an age-depth model was generated using the bacon R-package version 2.5.8 (Blaauw and Christen, 2011) with the calibration curve IntCal20 (Reimer et al., 2020).

The pollen analysis for Indigena and Caracaranã cores were carried out at 2 cm intervals (volume = 0.5 cm³, 18 and 23 samples respectively). The Jacaré sediments were studied at 5–10 cm increments (19 samples). The pollen samples were processed using standard analytical methods (Faegri and Iversen, 1989). Pollen concentration and influx were calculated through the addition of Lycopodium spores as an exotic marker. Pollen samples were analyzed using a Zeiss Axioskop photomicroscope at a magnification of 630x. A minimum of 300 terrestrial pollen grains were counted for each sample, that is, excluding Cyperaceae and aquatic pollen. Pteridophytes were expressed as a percentage of the terrestrial pollen sum. Pollen identification was based on the Neotropical Pollen Database (Bush and Weng, 2007), regional pollen atlases (Fontes et al., 2020; Leal et al., 2011; Roubik and Moreno, 1991; Rull, 2003), and the pollen reference collection at the Florida Institute of Technology.

Pollen diagrams were created using C2 (Juggins, 2007) and CorelDraw ^®. Pollen taxa were grouped to support the interpretation based on their ecological association following Barbosa and Miranda (2005), and Miranda et al. (2003). These ecological assemblages comprised: (i) Forest-woody taxa, (ii) Savanna-grasses and (iii) Palms. Paleoecological zonation of the diagrams was based on sedimentary hiatuses in the record, that is, a zone before and one after, which matched with a CONISS (Grimm, 1987; see Supporting Information-Appendix 3) . The zonation of Lake Jacaré was based on an integration of sediment analysis, visual inspection of pollen diagrams, and DCA (detrended correspondence analysis) results.

A single non-metric multidimensional scaling (NMDS) was performed on the palynological data for all three lakes. To reduce noise and improve the stress performance of NMDS, data were filtered to include only taxa present in five or more samples per lake, and with values of abundance > 5% in at least one sample. The ordination was based on Bray-Curtis dissimilarities among pollen samples. The analysis were carried out using the R-Package “vegan”(Oksanen et al., 2019).

Charcoal was sampled continuously along the core, with subsamples every 1 cm (volume = 1 cm³) at Lake Indigena and Caracaranã. In Lake Jacaré discrete samples were analyzed at 5 cm intervals. Samples were filtered using a 180 μm mesh. The filtered sample was suspended in water and charcoal was identified, quantified, and imaged using a Nikon stereo microscope SMZ-645 at a magnification of 20x. The surface area of every individual charcoal fragment was measured using ImageJ software (Rasband, 2012). Charcoal influx was calculated as area (mm²) accumulated per year.

For Caracaranã, Indigena and Jacaré, the lengths of all Poaceae pollen grains encountered were recorded. Poaceae pollen size frequency distribution per depth was plotted for each lake using boxplots and the R-packages “graphics” and “ggplot”(Villanueva and Chen, 2019). In total 2030 grains were measured for Lake Jacaré, 2840 for Caracaranã and 2905 for Indigena.

Results

Age model

The cores from Lake Caracaranã, Indigena and Jacaré were 47, 35, and 132 cm in length, respectively. Sediments were organic rich in the upper portion of each record, becoming silty-sands in the middle part and with clays near the base (Figure 2).

Figure 2.

Age-depth model for Lake Caracaranã, Lake Indigena and Lake Jacaré. The calibrated ages were obtained with the rbacon package implemented in R software (Blaauw and Christen, 2011). The line inside the shaded area corresponds to the modeled mean age of each depth. The vertical line indicates an inferred hiatus in sedimentation. Core stratigraphy is summarized below the age model for each site.

The ²¹⁰Pb (Caracaranã and Indigena) and ¹⁴C (all lakes) data were used to generate the chronology (Table 1, Figure 2 & Supporting Information, Appendix 4). The sedimentary records for all lakes spanned much of the Holocene, but the apparently very slow sedimentation across some intervals strongly suggested discontinuities in the sedimentary record. Lake Caracaranã formed c. 11200 calibrated ¹⁴C years before present (hereafter yr BP), while Indigena formed at c. 7800 yr BP, and Jacaré at c. 8900 yr BP. Sandy lenses marked sedimentary gaps in Caracaranã between c. 10,200 and 6500 yr BP, between c. 6500 and 1000 yr BP at Indigena, and between c. 4400 and 1500 yr BP at Jacaré (Table 1, Figure 2). Sedimentation rates were low in all lakes ranging between 0.0019 and 0.048 cm/yr at Caracaranã, 0.008 and 0.029 cm/yr at Indigena, and 0.015 and 0.023 cm/yr at Jacaré (Figure 3).

Table 1.

AMS radiocarbon dates, ²¹⁰Pb dates, and calibrated ages for Lakes Caracaranã, Indigena and Jacaré.

LabID	Type	Age	Depth	Calibrated ages in cal yr BP	Average in cal yr BP
Lake Caracaranã
UNH-surf_cara	210Pb	−69 ± 2	0	−74 to −69	−64
UNH-210Pb-cara	210Pb	150 ± 2	5.5	117 to 191	145
D AMS-39792	14C	372 ± 21	16.5	307 to 484	415
D AMS-39793	14C	2526 ± 23	32.5	1803 to 2658	2356
D AMS-46077	14C	4904 ± 25	39.5	3765 to 5443	4983
D AMS-39794	14C	9997 ± 36	46.5	9603 to 11576	10,958
Lake Indigena
UNH-surf_indi	210Pb	−69 ± 2	0	−74 to −64	−69
UNH-210Pb-indi	210Pb	150 ± 2	6.5	98 to 153	130
D AMS-39788	14C	333 ± 20	12.5	268 to 451	361
D AMS-46076	14C	938 ± 21	19.5	540 to 849	708
D AMS-39789	14C	6158 ± 29	24.5	6430 to 7054	6792
D AMS-38989	14C	7162 ± 33	34.5	7364 to 7969	7715
Lake Jacare
OS-56603	14C	1380 ± 45	23.5	970–1413	1217
OS-56604	14C	4180 ± 35	33	4263–4812	4611
CURL-5387	14C	4870 ± 40	50	5188–5737	5522
CURL-5388	14C	6610 ± 70	81	7065–7623	7411
CURL-5389	14C	8080 ± 45	105	8510–9233	8914

All ages were based on bulk carbon samples.

UNH: University of New Hampshire; D AMS: Direct AMS; OS: National Ocean Sciences AMS Facility; CURL: INSTAAR – AMS Radiocarbon Laboratory.

Figure 3.

(a–c) left: Pollen diagrams of the three north Amazonia savanna lakes highlighting the taxa with the highest percentage values. (a) Lake Caracaranã, (b) Lake Indigena, and (c) Lake Jacaré. Right: Summary percentage values by ecological group, charcoal amount (mm²), pollen concentration (grains x cm³), NMDS second axis (NMDS2) and sedimentation rate (cm/yr). To see the complete pollen diagrams, go to Supporting Information (Appendix 2).

Paleoecological zones

For a full description of the pollen zones please see Supporting Information (Appendix 5)

Lake Caracaranã

The sediments of Lake Caracaranã span approximately 11,200 yr BP. Pollen samples generally had good preservation, but diversity was relatively low. A total of 78 pollen morphotypes were found, belonging to 42 families and 72 genera. Most families were represented by just one or two morphotypes. The majority of morphotypes were rare with only 19 occurring with an abundance > 3%. Although not a sedimentary hiatus, pollen was not preserved between c. 2500 and 1000 yr BP. Three palynological zones were recognized, reflecting changes in the relative increases of woody taxa relative to Poaceae. The zones coincided with sediments deposited between hiatuses in pollen preservation. While the record is consistently grass-dominated (Poaceae), there is a trend toward slightly more woody taxa after the hiatus at 6500 yr BP (Figure 3). Moraceae and Alchornea occur at > 10%, and Melastomataceae, Trema, Bursera, Cecropia, Byrsonima, and the palm Mauritia, all increase in abundance. Forest taxa increase again after 600 years, with the same groups plus Melastomataceae/Combretaceae reaching values up to 3%. About 300 years BP palms rise in abundance, especially Attalea, Euterpe, and Astrocaryum, but decline again in the uppermost sample. Pollen concentration show the higher values (>20,000 grains) in the early Holocene, and at ~200 yr BP (>15,000 grains) (Figure 3). Charcoal was abundant prior to c. 10,000 cal BP, and was intermittently abundant during the mid-Holocene, but the area and influx values document a marked increase in fire activity c. 300 yr BP (Supporting Information – Appendix 6 and Figure 3). Sedimentation rates were low and constant during most Holocene, but an abrupt increase is present after 600 yr (from 0.007 to 0.020 cm/year), reaching the higher values at c. 400-200 yr BP (~ 0.040 cm/year)(Figure 3).

Lake Indigena

Sediments span approximately 7800 yr BP at Lake Indigena. A total of 64 morphotypes were found, belonging to 35 families and 60 genera. Two palynological zones were recognized (Figure 3). Overall, the site showed a history of grassland (Poaceae) dominance with an increase in woody taxa in the last 1000 years after the hiatus in sedimentation. Byrsonima increased up to 5%, Moraceae/Urticaceae had values between 4% and 13%, while Celtis had as much as 12% of the pollen sum. Myristicaceae and Trema showed values between 3% and 5%. As at Caracaranã, forest and palms categories increased in the last c. 200 years. Pollen concentration shows the higher values in the bottom part of the core (>12,000 grains/cm3), with a decreasing tendency toward the top (Figure 3). The charcoal was found throughout but was most abundant c. 8000–7000 cal BP (Figure 3). A peak in charcoal area and influx was evident at c. 600 yr BP (Supporting Information – Appendix 6). Sedimentation rate was low in the mid-Holocene (~0.010 cm/year) but showed increased values in the last millenium especially after ~600 yr BP (~0.020–0.030 cm/year) (Figure 3).

Lake Jacaré

The sediments span approximately 8900 yr BP. Samples above 88 cm (c. 7800 yr BP) had good pollen preservation, but below this depth there was not enough pollen to analyze statistically, and none was seen below 105 cm (c. 8900 yr BP). Four pollen zones were identified from the record. The record begins at c. 7800 yr BP when Asteraceae (15%) and Poaceae (45%) were the most abundant types. Minor fluctuations in the abundance of woody taxa define the zone boundaries, but the environment was a grassland throughout these samples. Forest taxa showed a significative increase after 6500 yr BP linked to Fabaceae/Caesalpinoideae and Mimosa reaching values up to 6% and 18% respectively. During the last 1000 years there has been an increase in forest taxa after c. 800 yr BP due mainly to Mimosa and Melastomataceae/Combretaceae attaining values from ~3% to 8% (Figure 3 and Supporting Information – Appendix 5). Higher (lower) proportion of savanna (forest) elements are present after the sedimentary hiatus. Charcoal accumulation was insignificant at the beginning of the Jacaré record but increased rapidly at c. 7000 yr BP (Figure 3). Charcoal representation remained consistent before reaching the highest values of the core at c. 600 yr BP; thereafter it declined abruptly. As in the other lakes, the sedimentation rate was low at the bottom of the core but rise after the sedimentary hiatus, getting values higher than 0.020 after 800 yr BP (Figure 3).

Non-metric multidimensional scaling

The Non-Metric Multidimensional Scaling (NMDS) was run using 18 taxa that met the conditions, and a convergent solution was reached after 19 iterations (stress = 0.16). Sample ordination scores along NMDS Axes 1 and 2 varied between −0.441 and 1.01, and −0.537 and 0.664, respectively. The ordination largely separated the lakes on Axes 1 and 2, apart from the Caracaranã samples from the last 1000 years, which overlapped with those of Indigena. All lakes showed a similar trend of increasingly negative scores on Axis 2 as the Holocene proceeded (Figure 4).

Figure 4.

NMDS of the pollen assemblages at the savannas of Roraima. (a) The results of the sample ordination: red = Caracaranã, blue = Indigena, green = Jacaré. The age of each sample is labeled directly over each point. (b) The mean NMDS values for each palynologycal zone for the three lakes. Colors follow panel a, time direction is indicated by arrows.

Poaceae pollen sizes

The Poaceae pollen size distribution plots for all the lake records showed that most grains were between 25 and 40 µm. Through time the size distributions for the lakes all showed considerable changes in the range of Poaceae pollen sizes and modal peaks. Lake Caracaranã Poaceae pollen had median sizes of c. 31–33 µm during the early Holocene, followed by a trend toward larger sizes and higher variability (larger amplitude in the distribution, i.e. grains of diverse size not just grains of similar size). This trend reversed in the 410 yr BP as grass grains became smaller (Figure 5). Lake Indigena had abundant large grains at the bottom of the core with average median values between 35 and 40 µm, however after the depositional hiatus the interquartile range shifted toward smaller sizes within the range of 30–35 µm. At this lake, two timeframes stood out as having tight modal values (780 and 7340 yr BP, Figure 5), in both instances those values were close to 33 µm. At Indigena, the size spectrum of samples straddling the main sedimentary hiatus appeared to be very similar in composition, whereas others with no known discontinuity between them showed very large variations in their size distributions. Jacaré showed a similar behavior to Indigena, bigger grains in the Mid-Holocene and smaller grains after the hiatus, reaching individual median minima at 1060 and 1320 yr BP (27 µm, Figure 5); the samples from the last 600 years BP were generally small (median = 31 µm, Figure 5)

Figure 5.

Changes in the size distributions of Poaceae pollen through time at the three lakes in the savannas of Roraima. (a) Lake Caracaranã, (b) Lake Indigena, and (c) Lake Jacaré. Dotted line corresponds to the sedimentary gap. Inside arrow line in Caracaranã representing the period of no pollen deposition between 1000 and 2300 years BP. Note examples of substantial changes between neighboring samples, for example, Caracaranã 865–605 cal. yr BP (change in mean size), and at Indigena 7510–7340 cal. yr BP (change in range of size).

Discussion

Our paleoecological records from Roraima provided three discontinuous histories from one regional setting, each one with a different time of initiation. The first lake to form was Caracaranã at c. 11,000 yr BP, followed by Jacaré at c. 8900 yr BP, and Indigena at c. 7800 yr BP. It took a few hundred years for Caracaranã and Indigena to become permanent waterbodies depositing sediments that preserved pollen, that is, for the sediments to become consistently anoxic. For Jacaré, this process took over 1200 years. We detected at least one gap in each of the records, but there were probably more that went undetected. Changes in quartz fluxes and the increase of amorphous silica found in Caracaranã and Jacaré, at ~10,000 and ~7800 yr BP respectively, probably reflected unstable hydrologies as the lakes deepened (de Toledo, 2010; Simões et al., 2010)

Were there significant changes in the composition of northern Amazonian savanna vegetation during the Holocene?

That the gaps in sedimentation showed a broad commonality of no net deposition/pollen preservation between c. 10,000 and 7800 yr BP, and between c. 2500 and 1200 yr BP (Figure 6) highlighted times when water tables were probably lowest. Studies in the savannas of the Cerrado region in central Brazil during the early Holocene (Lagoa Feia), and in the Roraima region throughout the last 1000 yr (Lago Galheiro), reported sedimentary layers lacking pollen (Absy, 1979; Cassino et al., 2020), probably caused by dry conditions which generate lake desiccation and the pollen to oxidize.

Figure 6.

Summary pollen and charcoal diagrams of the Lakes Caracaranã, Indigena, Jacaré in the Roraima region and from Mapauri (Rull, 2007) and El Paují (Montoya et al., 2011) in the Gran Sabana region, compared with paleoclimatic reconstructions for the Cariaco Basin (Haug et al., 2001) and Paraiso Cave (Wang et al., 2017). Yellow = savanna, green = forest, orange = palms, and gray = other taxa. Highlighted pink area highlighted corresponds to the Mid-Holocene Dry Event (MHDE) or Holocene Thermal Maximum (HTM).

That the statistical pairwise comparisons used to identify pollen zones tended to pick out samples either side of the hiatuses as being most different was unsurprising. CONISS, the statistical tool that defines pollen zones, identifies a zone boundary where it detects the largest difference between adjacent samples (Grimm, 1987). If, for example, a vegetation community is sampled every 500 years and is found to be changing gradually but continuously, no zone boundary might be the correct outcome of the analysis. If, however, there was a 2000-year discontinuity between samples, the difference before and after the discontinuity, even with the same overall rate of vegetation change, would now stand out as being the strongest of the record and a zone boundary would be drawn. Thus, we do not find the zonation of these records to be particularly helpful, rather we see minor local changes in vegetation, superimposed on overall stability of a savanna ecosystem. In the last 1000 years, there is some evidence of an increasing woodland component, probably caused by an expansion of riparian corridors as the system became wetter (Figure 6. And Supporting Information – Appendix 7) (Flantua et al., 2016; Mayle et al., 2004).

These data for long-term stability appeared to conflict with long-term climate trends from Central Amazonian cave isotope data and from the Cariaco Basin (Haug et al., 2001; Wang et al., 2017), which suggested that climates of the last 5000 years were becoming drier (Figure 6). Our study sites were shallow, with large surface area to volume ratios, and the sandy substrate may have meant they were imperfectly sealed. Small changes in groundwater height and leaky basins, coupled with a delicate evaporitic balance, made them vulnerable to desiccation. We note that Turcq et al. (2002), Simões et al. (2010), and Cordeiro et al. (2014) cored a different locations in Caracaranã and obtained an 11,000-year record with no obvious discontinuities, that is, their location stayed much wetter than our coring location, indicates that these lakes were fluctuating in size and that a hiatus does not equate to strong drought. Erratic climates, with increased seasonality or just a few back-to-back dry years could easily alter the lake water balance such that they stopped accumulating sediment. We are reminded of a parallel in peat bogs and tundra surfaces in which oxidation balances organic deposition, resulting in long periods of no net peat accumulation (Clymo, 1984). Thus, the individual hiatuses do not indicate distinct episodes of strong climatic change, but when all three lakes were dry, between c. 10,000 and 7800 yr BP and again between c. 2500 and 1200 yr BP probably indicate sustained dry events. Our data were consistent with a drying inferred from a transition from forest to treeless savanna in the Venezuelan Gran Sabana at c. 10,000 yr BP, (Rull, 2007; Rull et al., 2013). As warming in equatorial South America was likely largely complete by c. 14,000–12,000 years BP (Bromley et al., 2016; Raczka et al., 2019; Shakun et al., 2015; Urrego et al., 2009) the savanna expansion in Venezuela probably reflected a reduction in precipitation. Holocene speleothem growth at Paraiso Cave in central Amazonia was interrupted between c. 8500 and 5200 yr BP (Wang et al., 2017), suggesting a dry period that slightly overlaps with our inferred drying between c. 10,000 and 7800 years BP (Rull, 2007; Rull et al., 2013); (Figure 6). Although Paraiso Cave offers some similarity to our record, it needs to be recognized that it is 1000 km from our study area and in a region that today is considerably wetter. The extent to which these systems do not overlap probably reflects regional heterogeneity of wet and dry events within Amazonia.

The return of wetter conditions in Caracaranã and Jacaré after ~6800 yr BP (Figures 3 and 6), is also consistent with another record from the Roraima region, Serra do Tepequem, where wetter conditions were inferred by a rise in the pollen percentages of forests and palm swamps of Mauritia flexuosa after 6500-6000 yr BP (Rodríguez-Zorro et al., 2017).

Our results also suggest commonalities with other savannas in Colombia and Venezuela. For example, sites in the Llanos Orientales of Colombia which are located at a similar latitude (above ~3°N) to our study region showed analogous conditions during the Pleistocene-Holocene transition and also during the early-mid Holocene. Lakes El Piñal, Sardinas and Angel, showed an increase in forest taxa just after the onset of the Holocene (ca. 13,000–11,000 years BP) suggesting a wetter climate (Behling and Hooghiemstra, 1999; Piraquive-Bermúdez and Behling, 2022), coincident with the establishment of Lake Caracaranã. This same time frame also seems to be coeval with the establishment of Mapauri and a period of forest dominance in the Gran Sabanas of Venezuela. In the case of the mid-Holocene, wetter conditions after c. 7000 yr BP in Llanos Orientales are evident in lakes Loma Linda, Sardinas, Angel, and Chenevo with the increase of gallery forests or Mauritia palms (Behling and Hooghiemstra, 1999, 2000; Berrio et al., 2002) .

In the same way, records from the savanna areas of the Cerrado biome of central and southern Brazil (usually below ~14°S) registered a dry period between ~11,000 and 7000 yr BP, with reduced abundances of gallery and moist forest taxa, more indicators of open landscapes, and many samples in which pollen was not preserved. These same southern hemisphere records report an increase in forest taxa after ~6800 yr BP and suggested wetter conditions (Cassino et al., 2018, 2020). The difference in the timing of Mid-Holocene climatic events across this latitudinal range (~3°N–14°S) probably relates to the latitudinal migration of the Intertropical Convergence Zone ITCZ (Haug et al., 2001). Other climatic influences, such as the South American Summer Monsoon (core Amazonia) or the South Atlantic Convergence Zone (SACZ, south, central and eastern Brazil) and their interactions with the regional climate (Bernal et al., 2016; Cheng et al., 2013; Cruz et al., 2009; Wang et al., 2017) would promote regional differences in climate histories.

In our Roraima records, the next major dry event causing an across-site hiatus was between 2500 and 1200 years BP. This period was interpreted as drier/slightly drier in two locations from the Gran Sabanas of Venezuela, in El Pauji as a decrease in rainforest taxa and algae and the increase in a specific dry forest taxon (Gran Sabana – Figure 6), and in Laguna Encantada as a decline in forest extent and an expansion of the herbaceous component (Montoya et al., 2009, 2011). However, this period does not stand out as being unusual in the Cariaco Basin data (Haug et al., 2001; Hughen et al., 1996) or in data from Paraiso Cave (Wang et al., 2017)

Increases in the sedimentation rates at all lakes and in the proportions of forest taxa during the last 1500-1000 yr (Figures 3 and 6), also, fluctuations in the second axis of the NMDS toward negative values (Figures 3, 4a and b) were all consistent with wetter conditions. Similarly wet late-Holocene conditions were reported from the forest-savanna ecotone about 60 km from Lake Indigena, da Silva Meneses et al. (2013) documented evidence of wetter conditions across a range of sites that included: the establishment of Mauritia flexuosa and a higher proportion of forest taxa between 1550 and 1440 years BP, an increase in forest pollen elements between 1100 and 900 years BP, and at another site between 1000 and 700 years BP. These periods coincided with pollen preservation after the sedimentary gap for Jacaré, Indigena and the no-pollen zone in Caracaranã (Figure 3). The geochemical analysis for Jacaré also supported wetter conditions after 1500 yr BP (de Toledo, 2010), and at Caracaranã after 2000 yr BP (Simões et al., 2010).

The palynological record for the lakes of Roraima showed the dominance of open grassland vegetation throughout the Holocene. These systems were fundamentally stable with a slight long-term trend toward more woody vegetation. This transition seems more pronounced in the last 300 years and could reflect human activity promoting the survival of palms (Heijink et al., 2020). Palms are disproportionately important to indigenous communities and the increase coinciding with a spike in fire activity may indicate human influence on riparian forest communities (Costa et al., 2009; Levis et al., 2017). We offer this inference tentatively, as we do not see a massive increase in Mauritia palms, which has been suggested as indicative of human modifications of savanna wetlands in Colombia and Venezuela (Behling and Hooghiemstra, 2000; Rull and Montoya, 2014). The increase in forest taxa in the last 150 yr could also be related to decreased burning as has been suggested for other neotropical areas (Bush et al., 1992; Bush and Colinvaux, 1994; Niemann and Behling, 2008). In their study of late-Holocene environments of Roraima ecotones, da Silva Meneses et al. (2013) also reported a marked expansion of forest and palm taxa during the last ~250 years, which they attributed to a decrease in the anthropogenic fire activity in the area.

The record from Indigena showed the longest hiatus (approximately 5500 yr BP, from 6500–1000 years BP) perhaps reflecting its central location within the northern Amazonian savannas, while sites closer to the ecotone with currently forested areas experienced less long-term drying. In no instance did we see a flip-flop between savanna and forest. Our data suggest that these northern Amazonian savannas were stable at the biome scale throughout the Holocene, that is, they were always grasslands. Nevertheless, some fluctuations in available moisture probably induced changes in species abundances within the biome. Similar timings of vegetation change were found in lowland savannas (<600 m.a.s.l.) in Colombia and the Gran Sabanas of Venezuela (<1000 m.a.s.l.). Responses to the mid-Holocene dry period appeared to differ most markedly between the northern and the southern hemisphere savannas. In Roraima, we did not observe an extreme vegetation turnover as has been reported in other savanna records, where grasslands were replaced by forests or by palm swamps (Mauritia flexuosa) (Behling and Hooghiemstra, 1998; Montoya et al., 2009, 2011). Mauritia, a semi-aquatic palm, was never a conspicuous element in our records, probably reflecting drier habitats that inhibited the full development of the dense stands common in wetland areas.

Did fire intensity increase during the course of the Holocene?

Even in this savanna region, many fires are the result of human activity because fire ignition by lightning in Amazonia is in most cases complemented with rain, reducing the probability of natural fire (Barbosa and Fearnside, 2005). Although fire is frequent in the Roraima savannas it is strongly related to the proximity to human settlements and managed areas (cattle ranches and indigenous villages) (Meneses et al., 2015).

Fire histories in Amazonia show a strong relationship between frequency of events and inferred presence of people (Gosling et al., 2021). A previous study of fossil charcoal influx to Lake Caracaranã found an order of magnitude more particles at c. 11,000 yr BP compared with late-Holocene levels, but that analysis was based on the number of microcharcoal particles found between 2 and 350 µm in size and did not have a standardized effort per sample (Cordeiro et al., 2014). Consequently, that study, because of the range of sizes considered, included both a long-distance dispersed component and a local component. In our analysis, which standardized effort by volume of material counted of local charcoal (fragments > 180 µm), we also found a peak of charcoal production in the basal samples from Caracaranã, but our estimates of late-Holocene fire activity showed the highest values during the last millenium, particularly at c. 300 yr BP. Lake Jacaré which is close in proximity to Caracaranã also showed the highest values during the last 600 yr (Figure 3).

During the mid-Holocene the savanna lakes all contained charcoal, affirming that fire was an important part of this landscape. Where charcoal values fell, it was unlikely that this represented a lack of fire, for savanna is a fire-maintained system, but rather that the fuel was either completely ashed during combustion, that is, charcoal was not produced, or the charcoal produced by burning fine fuels was small enough to pass through our filter or not as well fossilized as that of woody vegetation (Leys et al., 2017).

In the Late-Holocene, we and other authors found an increase in fossil charcoal. Montoya and Rull (2011) described an abrupt increase in the fire incidence during the last 2000 years in the Gran Sabana region of Venezuela, suggesting a landscape managed and altered by humans. The increase in fire in that setting was associated with the expansion of savannas, the decline of forest and shrublands and in some cases the appearance of Mauritia palm swamps (Montoya and Rull, 2011).

In general, charcoal was present in most of the samples for all the lakes studied, and the highest values were present in the last millenium for lakes Caracaranã and Jacaré; this outcome was consistent with intensifying human activities in the area. Nevertheless, in the last 300 years there was a steady reduction in charcoal abundance. Contrastingly, Lake Indigena showed the highest values during the mid Holocene and generally low values after 600 yr BP (<1 mm²), the later probably related to less human presence and impacts in the area (difficult access) and less forest cover in that specific location. Analysis of the modern landcover using remote sensing in a 20 km buffer for the sites Caracaranã and Indigena, showed the former one with higher presence of forest compared to the latter (Supporting Information – Appendix 8).

Indigenous communities of the Makuxi and Wapichana from this region rely on hunter-gathering, slash and burn agriculture, livestock and pig or sheep farming (Ricardo and Ricardo, 2011). Although these communities engaged in activities such as cattle raising since the 1700s (Barbosa, 1993), their earlier lifestyle and occupation sites are not known. A major European influence on the region in the 1700s was slavery. Depopulation of indigenous communities through warfare, slave capture, and remaining populations migrating to hide in deeper forest, could have caused local abandonment and reduced anthropogenic fire (Meneses et al., 2015). Such an explanation is consistent with the increase in woody species seen in the fossil pollen data (Figure 3).

Were there significant changes in the Poaceae communities during the Holocene?

Most of the Poaceae pollen found in the samples from the Roraima lakes lay in the range of 26–40 µm ( Figure 5). In southern Brazil, such a size range was described as typical of herbaceous pollen types, while large (>45 µm) Poaceae pollen were uniquely associated with “arboreal forest” habitats in the same study (Radaeski et al., 2016). Among these grains, those > 46 µm belonged to bamboos (Radaeski and Bauermann, 2018). Salgado-Labouriau and Rinaldi (1990) studied grasses in Venezuelan mountains and found that Poaceae pollen grains that were 50–60 µm in length belonged either to Bambusoideae, for example, Chusquea, or Pooideae, for example, Bromus spp. Bamboos were not typical of savanna floras, making it most likely that grains in this size range found at Roraima belonged to Pooid grasses. Another, and not mutually exclusive, possibility was that these large grass pollen types were from C4 polyploid grasses (Jan et al., 2015).

At the other extreme of the size range, small Poaceae grains (<22 µm) are suggested to be indicative of grasslands (Radaeski et al., 2016; Salgado-Labouriau and Rinaldi, 1990). Such grasslands can be characteristic of human disturbance in forested settings, but here in a savanna setting that relationship seems less certain.

All sites showed some changes in the grass community. Discontinuities were usual across hiatuses, but not always strong and never as strong as one of the other changes between some consecutive samples. The pollen size in the three lakes showed high variability based on the results of the Interquartile Range (IQR). General results point to a decrease in median sizes in the smallest lakes throughout the entire Holocene. In the case of lake Indigena from ~39 to 34 µm, and at Jacaré from 35 to 31 µm. These results could be interpreted based on the size of the lakes, with the smallest receiving mainly pollen from a small catchment area and a narrower range of habitat types (Prentice, 1985) . Thus, if there was a shift in local grass populations it was likely to be documented. Indeed, the smallest lake, Indigena, had samples with the clearest modal peak of Poaceae pollen size, for example, in samples from c. 7340 and 780 yr BP. Such tight clustering of sizes could suggest the dominance of a small cluster of species and was possibly consistent with a small lake contracting to a grassy swamp and providing a very local signal of dominance by aquatic species. Contrastingly, the much larger surface area of Lake Caracaranã would receive pollen from a larger area (Jacobson and Bradshaw, 1981) and the influence of any given habitat on the pollen signature would be small. Consequently, the pollen spectra would probably show less variance through time.

Overall, the Poaceae pollen size data indicated that there were considerable changes taking place within the composition of the grasses. The median size and size range of the grasses is constantly changing and provides a more dynamic record than that of the pollen flora as a whole. However, more research on Poaceae pollen size distributions and the best statistical methods to describe them would improve the application of this technique to the analysis of the grassland communities.

A recent study cautioned against using Poaceae pollen sizes to reconstruct past vegetation and climates due to the variability within genera and species and also because there is no obvious relationship between the size and certain explanatory variables (Wei et al., 2023). However, we suggest that the analysis of pollen sizes to reveal shifts in grass assemblages, which does not rely on species identification, can be informative. Changes in pollen size assemblages can yield valuable information on the timing of changes within the Poaceae flora that would otherwise be hidden and can indicate a much greater volatility in the savanna community composition than can be inferred without them.

Conclusions

The savannas of Roraima showed an overall biome stability during the entire Holocene, with little vegetation turnover apparent in the records. The lake sedimentary records appeared to be highly sensitive to local water table variation. Lake formation took place between 11,000 and 7800 years BP. The presence of sedimentary discontinuities from c. 10,000 to 7800 years BP and from c. 2500 to 1200 years BP suggested dry periods during which we infer that the lakes dried out or became impermanent. Our data were consistent with observations from other northern Amazonian settings of relatively wet conditions between 6500 and 3500 years BP, an opposite signal to that of the Altiplano and southern Amazonia. Wetter conditions, and the establishment of the actual climate is present after 1000 yr in the Roraima Savannas.

Charcoal was present in most of the samples we analyzed, but there was an increase in charcoal abundance during the last millenium probably triggered by anthropogenic land use. A decrease in charcoal and an increase in forest taxa in the last 300 years may reflect local land abandonment in response to European slaving.

Our investigation of Poaceae pollen size distributions revealed substantial changes taking place that were not evident in the broader pollen signature. Although size was not an identifying determinant for most Poaceae pollen there did seem to be useful information to be gained from analysis of a population of grass sizes, though the best metrics for such an analysis have yet to be established.

Supplemental Material

sj-docx-1-hol-10.1177_09596836231211878 – Supplemental material for Holocene histories of biome stability in northern Amazonian savannas

Supplemental material, sj-docx-1-hol-10.1177_09596836231211878 for Holocene histories of biome stability in northern Amazonian savannas by Julian Beltran, Mauro B de Toledo, Michael Palace, Jack Dibb and Mark B Bush in The Holocene

Footnotes

Acknowledgements

We thank the government and people of Brazil for allowing us to conduct this research. We thank the researchers Majoi Nascimento and Luciane Fontana for the time and effort necessary to extract the Indigena and Caracaranã cores from the field, and also for the pictures of the areas. At the same time, we thank Lula, the fieldwork assistant. We acknowledge the help from Erika Kurtzweil and Jake Labishak, students from Florida Institute of Technology in the charcoal analysis.

Data availability statement

The data that supports the findings of this present study will be made available as .xlsx file in the supporting on-line material and through NEOTOMA.org.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by grants from the National Geographic Society to MBB and from NSF to MBB and from NASA to MP, JD, and MBB (80NSSC20K1266).

ORCID iDs

Julian Beltran

Mark B Bush

Supplemental material

Supplemental material for this article is available online.

References

Absy

(1979) A palynological study of Holocene sediments in the Amazon basin. Ph.D. Dissertation, University of Amsterdam, Amsterdam.

Absy

Clief

Fournier

, et al. (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 Academie des Sciences Paris, Series II 312: 673–678.

Baker

Fritz

(2015) Nature and causes of Quaternary climate variation of tropical South America. Quaternary Science Reviews 124: 31–47.

Barbosa

(1993) Ocupação Humana em Roraima. I. Do Histórico Colonial ao Início do Assentamento Dirigido. Boletim do Museu Paraense Emílio Goeldi 9(1): 123–144.

Barbosa

Campos

Pinto

, et al. (2007) The “Lavrados” of Roraima: Biodiversity and conservation of Brazil’s Amazonian savannas. Functional Ecosystems and Communities 1(1): 29–41.

Barbosa

Fearnside

(2005) Fire frequency and area burned in the Roraima savannas of Brazilian Amazonia. Forest Ecology and Management 204(2-3): 371–384.

Barbosa

Miranda

(2005) Fitofisionomias e diversidade vegetal das savanas de Roraima. Savanas de Roraima: etnoecologia, biodiversidade e potencialidades agrossilvipastoris. Boa Vista, FEMACT, pp.61–78.

Behling

Hooghiemstra

(1998) Late Quaternary palaeoecology and palaeoclimatology from pollen records of the savannas of the Llanos Orientales in Colombia. Palaeogeography, Palaeoclimatology, Palaeoecology 139: 251–267.

Behling

Hooghiemstra

(1999) Environmental history of the Colombian savannas of the Llanos Orientales since the Last Glacial Maximum from lake records El Pinal and Carimagua. Journal of Paleolimnology 21: 461–476.

10.

Behling

Hooghiemstra

(2000) Holocene Amazon rainforest-savanna dynamics and climatic implications: High-resolution pollen record from Laguna Loma Linda in eastern Colombia. Journal of Quaternary Science 15(7): 687–695.

11.

Bernal

Cruz

Stríkis

, et al. (2016) High-resolution Holocene South American monsoon history recorded by a speleothem from Botuverá Cave, Brazil. Earth and Planetary Science Letters 450: 186–196.

12.

Berrio

Hooghiemstra

Behling

, et al. (2002) Late-Quaternary savanna history of the Colombian Llanos Orientales from Lagunas Chenevo and Mozambique a transect synthesis. The Holocene 12(1): 35–48.

13.

Beug

(1961) Leitfaden der Pollenbestimmung Fur Mitteleuropa und Angrendzende Gebeite. Stuttgart: Gustav Fischer Verlag.

14.

Blaauw

Christen

(2011) Flexible paleoclimate age-depth models using an autoregressive gamma process. Bayesian Analysis 6(3): 457–474.

15.

Bromley

GRM

Schaefer

Hall

, et al. (2016) A cosmogenic 10Be chronology for the local last glacial maximum and termination in the Cordillera Oriental, southern Peruvian Andes: Implications for the tropical role in global climate. Quaternary Science Reviews 148: 54–67.

16.

Brown

Jr (1987) Conclusions, synthesis and alternative hypotheses. In: Whitmore

Prance

(eds) Biogeography and Quaternary History in Tropical America. Oxford: Oxford Science Publications, pp.175–196.

17.

Bueno

Pennington

Dexter

, et al. (2017) Effects of Quaternary climatic fluctuations on the distribution of Neotropical savanna tree species. Ecography 40(3): 403–414.

18.

Bush

Miller

de Oliveira

, et al. (2002) Orbital forcing signal in sediments of two Amazonian lakes. Journal of Paleolimnology 27: 341–352.

19.

Bush

Colinvaux

(1994) Tropical Forest Disturbance: Paleoecological Records from Darien, panama. Ecology 75: 1761–1768.

20.

Bush

Miller

De Oliveira

, et al. (2000) Two histories of environmental change and human disturbance in eastern lowland amazonia. The Holocene 10: 543–553.

21.

Bush

Oliveira

PED

(2006) The rise and fall of the refugial hypothesis of Amazonian speciation: A paleoecological perspective. Biota Neotropica 6.

22.

Bush

Piperno

Colinvaux

, et al. (1992) A 14 300-Yr paleoecological profile of a lowland tropical lake in panama. Ecological Monographs 62(2): 251–275.

23.

Bush

Weng

(2007) Introducing a new (freeware) tool for palynology. Journal of Biogeography 34: 377–380.

24.

Cassino

Ledru

M-P

Santos

RDA

, et al. (2020) Vegetation and fire variability in the central Cerrados (Brazil) during the pleistocene-Holocene transition was influenced by oscillations in the SASM boundary belt. Quaternary Science Reviews 232: Article 106217.

25.

Cassino

Martinho

da Silva Caminha

SAF

(2018) A Late Quaternary palynological record of a palm swamp in the Cerrado of central Brazil interpreted using modern analog data. Palaeogeography, Palaeoclimatology, Palaeoecology 490: 1–16.

26.

Cheng

Sinha

Cruz

, et al. (2013) Climate change patterns in Amazonia and biodiversity. Nature Communications 4: 1411.

27.

Clymo

(1984) The limits to peat bog growth. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 303(1117): 605–654.

28.

Cochrane

Laurance

(2002) Fire as a large-scale edge effect in Amazonian forests. Journal of Tropical Ecology 18(3): 311–325.

29.

Colinvaux

DeOliveira

(2001) Amazon plant diversity and climate through the Cenozoic. Palaeogeography, Palaeoclimatology, Palaeoecology 166: 51–63.

30.

Cordeiro

Turcq

Moreira

, et al. (2014) Palaeofires in Amazon: Interplay between land use change and palaeoclimatic events. Palaeogeography, Palaeoclimatology, Palaeoecology 415: 137–151.

31.

Costa

FRC

Guillaumet

Lima

, et al. (2009) Gradients within gradients: The mesoscale distribution patterns of palms in a central Amazonian forest. Journal of Vegetation Science 20(1): 69–78.

32.

Cowling

Betts

Cox

, et al. (2004) Contrasting simulated past and future responses of the Amazonian forest to atmospheric change. Philosophical Transactions of the Royal Society of London 359: 539–547.

33.

Cox

Betts

Collins

, et al. (2004) Amazonian forest dieback under climate-carbon cycle projections for the 21st century. Theoretical and Applied Climatology 78: 137–156.

34.

Cruz

Vuille

Burns

, et al. (2009) Orbitally driven east–west antiphasing of South American precipitation. Nature Geoscience 2(3): 210–214.

35.

da Silva Meneses

MEN

da Costa

Behling

(2013) Late Holocene vegetation and fire dynamics from a savanna-forest ecotone in Roraima state, northern Brazilian Amazon. Journal of South American Earth Sciences 42: 17–26.

36.

De Oliveira

Raczka

McMichael

CNH

, et al. (2020) Climate change and biogeographic connectivity across the Brazilian cerrado. Journal of Biogeography 47: 396–407.

37.

de Toledo

(2010) Dinâmica das Savanas e Paleoincêndios em Roraima nos últimos 10.000 anos. In: Reinaldo

Valdinar

(eds) Roraima: Homem, Ambiente e Ecologia. Boa Vista: FEMACT, pp. 237–256.

38.

De Toledo

(2004) Holocene Vegetation and Climate History of Savanna-Forest Ecotones in Northeastern Amazonia. Melbourne: Florida Institute of Technology.

39.

D’Agostino

Seltzer

Baker

, et al. (2002) Late-Quaternary lowstands of Lake Titicaca: Evidence from high-resolution seismic data. Palaeogeography, Palaeoclimatology, Palaeoecology 179(1-2): 97–111.

40.

Eiten

(1982) Brazilian “savannas”. In: Huntley BJ and Walker BH (eds) Ecology of Tropical Savannas: Ecological Studies, vol 42. Berlin, Heidelberg: Springer, pp.25–47.

41.

Faegri

Iversen

(1989) Textbook of Pollen Analysis. Chichester: Wiley.

42.

Fick

Hijmans

(2017) WorldClim 2: New 1-km spatial resolution climate surfaces for global land areas. International Journal of Climatology 37(12): 4302–4315.

43.

Flantua

SGA

Hooghiemstra

Vuille

, et al. (2016) Climate variability and human impact in South America during the last 2000 years: Synthesis and perspectives from pollen records. Climate of the Past 12(2): 483–523.

44.

Flenley

King

ASM

Jackson

, et al. (1991) The Late Quaternary vegetational and climatic history of Easter Island. Journal of Quaternary Science 6: 85–115.

45.

Fontes

Cordeiro

Martins

, et al. (2017) Paleoenvironmental dynamics in South Amazonia, Brazil, during the last 35,000 years inferred from pollen and geochemical records of Lago do Saci. Quaternary Science Reviews 173: 161–180.

46.

Fontes

Jaramillo

Moreno

(2020) Pollen morphology of the Amacayacu Forest dynamics plot, western Amazon, Colombia. Palynology 44(1): 32–79.

47.

Frost

(1968) The Climate of the Rupununi Savannas. Montreal, Canada: Department of Geography, McGill University.

48.

Gosling

Bush

(2005) A biogeographic comment on: Wüster et al. (2005) tracing an invasion: Landbridges, refugia, and the phylogeography of the Neotropical rattlesnake (Serpentes: Viperidae: Crotalus durissus). Journal of Molecular Ecology 14: 3615–3617.

49.

Gosling

Maezumi

Heijink

, et al. (2021) Scarce fire activity in north and north-western Amazonian forests during the last 10,000 years. Plant Ecology & Diversity 14: 143–156.

50.

Grimm

(1987) CONISS: A fortran 77 program for stratigraphically constrained cluster analysis by the method of incremental sum of squares. Computers & Geosciences 13: 13–35.

51.

Haffer

(1969) Speciation in Amazonian forest birds. Science 165: 131–137.

52.

Hanselman

Bush

Gosling

, et al. (2011) A 370,000-year record of vegetation and fire history around Lake Titicaca (Bolivia/Peru). Palaeogeography, Palaeoclimatology, Palaeoecology 305: 201–214.

53.

Haug

Hughen

Sigman

, et al. (2001) Southward migration of the Intertropical Convergence Zone through the Holocene. Science 293: 1304–1308.

54.

Heijink

McMichael

CNH

Piperno

, et al. (2020) Holocene increases in palm abundances in north-western Amazonia. Journal of Biogeography 47: 698–711.

55.

Hermanowski

da Costa

Behling

(2012) Environmental changes in southeastern Amazonia during the last 25,000 yr revealed from a paleoecological record. The Quaternary Research (Daiyonki-Kenkyu) 77(1): 138–148.

56.

Huber

(1995) Conservation of the Venezuelan Guayana. Flora of the Venezuelan guayana 1: 193–218.

57.

Hughen

Overpeck

Peterson

, et al. (1996) The nature of varved sedimentation in the Cariaco Basin, Venezuela, and its palaeoclimatic significance. Geological Society London Special Publications 116(1): 171–183.

58.

Jacobson

Bradshaw

RHW

(1981) The selection of sites for paleovegetational studies. The Quaternary Research (Daiyonki-Kenkyu) 16: 80–96.

59.

Jan

Schüler

Behling

(2015) Trends of pollen grain size variation in C3 and C4 Poaceae species using pollen morphology for future assessment of grassland ecosystem dynamics. Grana 54(2): 129–145.

60.

Juggins

(2007) C2 Version 1.5: Software for ecological and palaeoecological data analysis and visualisation.

61.

Latrubesse

Nelson

(2001) Evidence for Late Quaternary aeolian activity in the Roraima–Guyana region. CATENA 43(1): 63–80.

62.

Leal

Berrío

Raimúndez

, et al. (2011) A pollen atlas of premontane woody and herbaceous communities from the upland savannas of Guayana, Venezuela. Palynology 35(2): 226–266.

63.

Levis

Costa

Bongers

, et al. (2017) Persistent effects of pre-Columbian plant domestication on Amazonian forest composition. Science 355(6328): 925–931.

64.

Leys

Commerford

McLauchlan

(2017) Reconstructing grassland fire history using sedimentary charcoal: Considering count, size and shape. Plos One 12(4): e0176445.

65.

Lovejoy

Nobre

(2018) Amazon tipping point. Science Advances 4(2): eaat2340. DOI: 10.1126/sciadv.aat2340

66.

Malhi

Aragão

LEOC

Galbraith

, et al. (2009) Exploring the likelihood and mechanism of a climate-change-induced dieback of the Amazon rainforest. Proceedings of the National Academy of Sciences 106: 20610–20615.

67.

Mayle

Beerling

Gosling

, et al. (2004) Responses of Amazonian ecosystems to climatic and atmospheric carbon dioxide changes since the last glacial maximum. Philosophical Transactions of the Royal Society of London 359: 499–514.

68.

Mayle

Burbridge

Killeen

(2000) Millennial-scale dynamics of southern Amazonian rain forests. Science 290: 2291–2294.

69.

Mayle

Power

(2008) Impact of a drier early–mid- Holocene climate upon Amazonian forests. Philosophical Transactions of the Royal Society A: 363(1498): 1829–1838.

70.

Meneses

Costa

Enters

, et al. (2015) Environmental changes during the last millennium based on multi-proxy palaeoecological records in a savanna-forest mosaic from the northernmost Brazilian Amazon region. Anais da Academia Brasileira de Ciencias 87: 1623–1651.

71.

Meneses

MENDS

Costa

MLD

Costa

JAV

(2007) Os lagos do lavrado de Boa Vista - Roraima: Fisiografia, físico-química das águas, mineralogia e química dos sedimentos. Brazilian Journal of Geology 37(3): 478–489.

72.

Miranda

Absy

Rebêlo

(2003) Community structure of woody plants of Roraima savannahs, Brazil. Plant Ecology 164(1): 109–123.

73.

Montoya

Rull

(2011) Gran Sabana fires (SE Venezuela): A paleoecological perspective. Quaternary Science Reviews 30(23-24): 3430–3444.

74.

Montoya

Rull

Nogué

, et al. (2009) Paleoecología del Holoceno en la Gran Sabana, SE Venezuela: Análisis preliminar de polen y microcarbones en la Laguna Encantada. Collectanea Botanica 28: 65–79–79.

75.

Montoya

Rull

Nogué

(2011) Early human occupation and land use changes near the boundary of the Orinoco and the Amazon basins (SE Venezuela): Palynological evidence from El Paují record. Palaeogeography, Palaeoclimatology, Palaeoecology 310(3-4): 413–426.

76.

Murphy

Andersen

Parr

(2016) The underestimated biodiversity of tropical grassy biomes. Philosophical Transactions of the Royal Society B, 371(1703): 20150319.

77.

Musher

Giakoumis

Albert

, et al. (2022) River network rearrangements promote speciation in lowland Amazonian birds. Science Advances 8(14): eabn1099.

78.

Nascimento

Heijink

Bush

, et al. (2022) Early to mid- Holocene human activity exerted gradual influences on Amazonian forest vegetation. Philosophical Transactions of the Royal Society B, 377(1849): 20200498.

79.

Nascimento

Martins

Cordeiro

, et al. (2019) Vegetation response to climatic changes in western Amazonia over the last 7,600 years. Journal of Biogeography 46: 2389–2406.

80.

Niemann

Behling

(2008) Late Quaternary vegetation, climate and fire dynamics inferred from the El Tiro record in the southeastern Ecuadorian Andes. Journal of Quaternary Science 23(3): 203–212.

81.

Nimer

(1972) Climatologia da região Nordeste do Brasil. Introdução à climatologia dinâmica. Revista Brasileira de Geografia 34(2): 3–51.

82.

Oksanen

Blanchet

Friendly

, et al. (2019) Vegan: Community Ecology Package (R package Version 2.5-5). Available at: https://CRAN.R-project.org/package=vegan.Community ecology package (accessed 7 August 2023).

83.

Pennington

Lavin

Prado

, et al. (2004) Historical climate change and speciation: neotropical seasonally dry forest plants show patterns of both Tertiary and Quaternary diversification. Philosophical Transactions of the Royal Society of London 359: 515–537.

84.

Piraquive-Bermúdez

Behling

(2022) Holocene paleoecology in the Neotropical savannas of northern South America (Llanos of the Orinoquia ecoregion, Colombia and Venezuela): What do we know and on what should we focus in the future? Frontiers in Ecology and Evolution 10: 165.

85.

Pivello

(2011) The use of fire in the Cerrado and Amazonian rainforests of Brazil: Past and Present. Fire Ecology 7(1): 24–39.

86.

Prentice

(1985) Pollen representation, source area, and basin size: Toward a unified theory of pollen analysis. The Quaternary Research (Daiyonki-Kenkyu) 23(1): 76–86.

87.

Raczka

Mosblech

Giosan

, et al. (2019) A human role in Andean megafaunal extinction? Quaternary Science Reviews 205: 154–165.

88.

Radaeski

Bauermann

(2018) Poaceae pollen grains from southern Brazilian grasslands: Pollen grain size in species from dry and humid environments. Neotropical Biology and Conservation 13(2): 111–123.

89.

Radaeski

Bauermann

Pereira

(2016) Poaceae pollen from southern Brazil: Distinguishing grasslands (Campos) from forests by analyzing a diverse range of Poaceae species. Frontiers in Plant Science 7: 1833.

90.

Rasband

(2012) ImageJ: Image processing and analysis in Java. Astrophysics Source Code Library. ascl: 1206.1013.

91.

Reimer

Austin

WEN

Bard

, et al. (2020) The IntCal20 Northern Hemisphere radiocarbon age calibration curve (0–55 cal kBP). Radiocarbon 62(4): 725–757.

92.

Reinaldo

Ciro

(2011) Detection and geographical distribution of clearing areas in the savannas ("lavrado") of Roraima using Google Earth web tool. Journal of Geography and Regional Planning 4(3): 122–136.

93.

Ribas

Aleixo

Nogueira

ACR

, et al. (2012) A palaeobiogeographic model for biotic diversification within Amazonia over the past three million years. Proceedings of the Royal Society of London B: Biological Sciences 279: 681–689.

94.

Ricardo

(2011) Povos Indígenas No Brasil: 2006/2010. São Paulo: Instituto Socioambiental.

95.

Rodbell

Hatfield

Abbott

, et al. (2022) 700,000 years of tropical Andean glaciation. Nature 607(7918): 301–306.

96.

Rodríguez-Zorro

Costa

MLD

Behling

(2017) Mid- Holocene vegetation dynamics with an early expansion of Mauritia flexuosa palm trees inferred from the Serra do Tepequém in the savannas of Roraima state in Amazonia, northwestern Brazil. Vegetation History and Archaeobotany 26(5): 455–468.

97.

Roubik

Moreno

(1991) Pollen and spores of Barro Colorado Island. Missouri Botanical Garden: Monographs in Systematic Botany 36.

98.

Rull

(2003) An illustrated key for the identification of pollen from Pantepui and the Gran Sabana (eastern Venezuelan Guayana). Palynology 27(1): 99–133.

99.

Rull

(2007) Holocene global warming and the origin of the Neotropical Gran Sabana in the Venezuelan Guayana. Journal of Biogeography 34(2): 279–288.

100.

Rull

Montoya

(2014) Mauritia flexuosa palm swamp communities: Natural or human-made? A palynological study of the Gran Sabana region (northern South America) within a neotropical context. Quaternary Science Reviews 99(0): 17–33.

101.

Rull

Montoya

Nogué

, et al. (2013) Ecological palaeoecology in the neotropical Gran Sabana region: Long-term records of vegetation dynamics as a basis for ecological hypothesis testing. Perspectives in Plant Ecology, Evolution and Systematics 15(6): 338–359.

102.

Salati

(1985) The climatology and hydrology of Amazonia. In: Prance

Lovejoy

(eds) Amazonia. New York: Pergamon, pp.18–48.

103.

Salgado-Labouriau

Rinaldi

(1990) Palynology of Gramineae of the Venezuelan mountains. Grana 29(2): 119–128.

104.

Santos

Coloma

Summers

, et al. (2009) Amazonian amphibian diversity is primarily derived from late Miocene Andean lineages. Plos Biology 7(3): e1000056.

105.

Schüler

Behling

(2011a) Characteristics of Poaceae pollen grains as a tool to assess palaeoecological grassland dynamics in South America. Vegetation History and Archaeobotany 20(2): 97–108.

106.

Schüler

Behling

(2011b) Poaceae pollen grain size as a tool to distinguish past grasslands in South America: A new methodological approach. Vegetation History and Archaeobotany 20(2): 83–96.

107.

Shakun

Clark

Marcott

, et al. (2015) Cosmogenic dating of Late Pleistocene glaciation, southern tropical Andes, Peru. Journal of Quaternary Science 30(8): 841–847.

108.

Shukla

Nobre

Sellers

(1990) Amazon deforestation and climate change. Science 247: 1322–1325.

109.

Simões

FILHOF

Turcq

Sifeddine

(2010) Mudanças Paleoambientais Do Contato Floresta-Savana de Roraima Durante o Holoceno. Roraima: Homem, Ambiente e Ecologia. Boa Vista: FEMACT.

110.

Smith

Mayle

(2018) Impact of mid- to late Holocene precipitation changes on vegetation across lowland tropical South America: A paleo-data synthesis. The Quaternary Research (Daiyonki-Kenkyu) 89(1): 134–155.

111.

Smith

Singarayer

Mayle

(2022) Response of Amazonian forests to mid-Holocene drought: A model-data comparison. Global Change Biology 28(1): 201–226.

112.

Sternberg

LSL

(2001) Savanna–forest hysteresis in the tropics. Global Ecology and Biogeography: A Journal of Macroecology 10: 369–378.

113.

Thom

Xue

Sawakuchi

, et al. (2020) Quaternary climate changes as speciation drivers in the Amazon floodplains. Science Advances 6(11): eaax4718.

114.

Toby Pennington

Prado

Pendry

(2000) Neotropical seasonally dry forests and quaternary vegetation changes. Journal of Biogeography 27: 261–273.

115.

Turcq

Albuquerque

ALS

Cordeiro

, et al. (2002) Accumulation of organic carbon in five Brazilian lakes during the Holocene. Sedimentary Geology 148(1-2): 319–342.

116.

Urrego

Bush

Silman

, et al. (2009) Millennial-scale ecological changes in tropical South America since the last glacial maximum. In: Vimeux F, Sylvestre F and Khodri M (eds) Past Climate Variability in South America and Surrounding Regions. Springer, pp.283–300.

117.

van Breukelen

Vonhof

Hellstrom

, et al. (2008) Fossil dripwater in stalagmites reveals Holocene temperature and rainfall variation in Amazonia. Earth and Planetary Science Letters 275(1-2): 54–60.

118.

Villanueva

RAM

Chen

(2019) ggplot2: Elegant graphics for data analysis (2nd ed). Measurement Interdisciplinary Research and Perspectives 17(3): 160–167.

119.

Wang

Edwards

Auler

, et al. (2017) Hydroclimate changes across the Amazon lowlands over the past 45,000 years. Nature 541(7636): 204–207.

120.

Wei

Jardine

Gosling

, et al. (2023) Is Poaceae pollen size a useful proxy in palaeoecological studies? New insights from a Poaceae pollen morphological study in the Amazon. Review of Palaeobotany and Palynology 308: 104790.

121.

Weng

Luedeke

MKB

Zemp

, et al. (2018) Aerial and surface rivers: Downwind impacts on water availability from land use changes inAmazoniaa. Hydrology and Earth System Sciences 22(1): 911–927.

122.

Werneck

Nogueira

Colli

, et al. (2012) Climatic stability in the Brazilian Cerrado: Implications for biogeographical connections of South American savannas, species richness and conservation in a biodiversity hotspot. Journal of Biogeography 39(9): 1695–1706.

123.

Wüster

Ferguson

Quijada-Mascareñas

, et al. (2005) Tracing an invasion: Landbridges, refugia, and the phylogeography of the Neotropical rattlesnake (Serpentes: Viperidae: Crotalus durissus). Molecular Ecology 14: 1095–1108.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

6.15 MB