Possible linkages of palaeofires in southeast Amazonia to a changing climate since the Last Glacial Maximum
A 200 cm-long high resolution macro-charcoal and pollen record from the Lagoa da Cachoeira in Serra Sul dos Carajás (Serra Sul) in southeast Amazonia reveals insights into local palaeofire over the last 26,200 years. Local fires in Serra Sul were most frequent in transition periods from dry to wet environmental conditions between 11,000 and 10,200 years ago, and under seasonal climatic conditions after 5,000 years ago. During pronounced dry periods fires were not a substantial component of the environment in Serra Sul. An anthropogenic influence on fire in Serra Sul may have played a role since the beginning of the Holocene, but is not a likely driver of palaeofire variability. Charcoal records for southern Amazonia coupled with proxy data for precipitation and changing Atlantic Sea Surface Temperature (SST) suggest that Holocene palaeofires in southern Amazonia are driven by changes in climate.
KeywordsAmazonian forest and savanna Late Pleistocene Holocene Palaeofire Climate change
Predictions for the next centuries highlight the growing threat to Amazonian ecosystems by climatically amplified wildfires coupled with anthropogenic influence. Increased droughts resulting from global warming and land use change may lead to increased flammability and mortality of tropical rainforests (Nepstad et al. 2004). Clarifying connections between past fire and vegetation under changing climatic conditions will aid in understanding long-term ecosystem dynamics of southeast Amazonia.
The analysis of charcoal fragments from lake sediments offers a tool for calculating past fire frequency at high temporal resolution. In Amazonia however, there are few high-resolution records. In southeast Amazonia, records mainly represent regional, low resolution fire histories of the past 8,000 years (Behling 2001; Behling and Costa 2000, 2001; Bush et al. 2000, 2007a, b; Cordeiro et al. 2008; Hermanowski et al. 2012; Irion et al. 2006; Sifeddine et al. 2001; Soubiès 1979). In eastern Amazonia fires were recorded from 6 to 3 ka bp (calibrated: 6.7–3 ka cal bp) (Soubiès 1979), in Serra Norte dos Carajás charcoal fragments are suggested as having derived from forest fires under dry climates between 7.45 and 4.75 ka cal bp (Cordeiro et al. 2008). At Rio Tapajós charcoal is more frequently recorded between 5.2 and 6 ka cal bp (Irion et al. 2006). In summary these studies recognize more frequent regional fires between ca. 7.5 and 3 ka cal bp. For these periods drier climates are suggested to be the driving factor of increased fire activity.
Palaeofire records from the Global Charcoal Database (Power et al. 2008) included in palaeofire analysis
Rain forest and edaphic savanna
Bush et al. (2000)
Lago Santa Maria
Bush et al. (2007b)
Bush et al. (2007b)
Behling and Costa (2000)
Disturbed coastal vegetation
Behling and Costa (2001)
Lagoa da Curuçá
Pasture and secondary forest, but formerly rainforest
Savanna and cerrado patches
Bush et al. (2007a)
Bush et al. 2007a)
Lake Santa Rosa
Old-growth forest, disturbed (cultivated farmland nearby)
Pontes e Lacerda
Gouveia et al. (2002)
Virgin, tall humid evergreen rainforest
The region experiences strong seasonal conditions with a severe dry season from June to October (Sifeddine et al. 2001) with rainfall of 150 mm/month (INMET 2011). In the peak of the wet season from December to February, the mean precipitation is 1,400 mm (INMET 2011). According to Stott (2000), natural fire would be an annual event under this ‘savanna-forest-climate’ because the dry season is long enough to dry the fuels (i.e. dried/dead grass, litter) that are needed for the combustion of fire, whereas the amount of supplied moisture during the rainy season is high enough for the production of biomass (source of fuel load). Several factors play a role in the seasonality of precipitation throughout the year, including the migration of the Intertropical Convergence Zone (ITCZ), changing Atlantic Sea Surface Temperature (SST), moist easterly trade winds from the tropical Atlantic, and the onset and intensification of Amazon convection (Nobre and Shukla 1996; Marengo et al. 1993, 2001; Fu et al. 2001; Liebmann and Marengo 2001; Garreaud et al. 2009). Moist trade winds from the tropical Atlantic move across the Amazon Basin to the eastern flanks of the Andes, where the South American Low Level Jet (SALLJ) develops and transports moist air further south. It crosses the Bolivian Amazon, and subsequently reaches Southeast Brazil, where the South American Convergence Zone (SACZ) begins to form (Cook 2009; Fig. 1). The amount of moisture from the tropical Atlantic is controlled by the zonal (NE-NW) and meridional (NW-SW) SST gradient (zonal/meridional ΔSST) in the Atlantic Ocean (Fig. 1). A high value of zonal SST gradient is correlated with an enhancement of northeasterly trade winds and consequently with increased advection to Amazonia (Baker et al. 2001a). Interhemispheric (meridional) SST gradients produce cross equatorial winds that influence the latitudinal position of the ITCZ (Hastenrath and Greischar 1993; Nobre and Shukla 1996; Chiang et al. 2002). Today higher rainfall in Amazonia south of the equator is the result of anomalously low SST in the northern tropical Atlantic, as the ITCZ then remains south of the equator (Nobre and Shukla 1996; Baker et al. 2001b). The latitudinal shifting of the ITCZ is responsible for the modern seasonal rainfall patterns in Amazonia. During the wet season (Nov–Apr) enhanced convection in eastern Amazonia is coupled with an ITCZ positioned south of the equator. During the dry season (May–Oct) decreased convection is coupled with a northerly position of the ITCZ. Precipitation in Amazonia is also affected by the El Niño-Southern Oscillation (ENSO) phenomenon, which is coupled with cyclic changing SST in the equatorial Pacific. During the El Niño stage (warmer SST in east Pacific) the western Amazon experiences lower rainfall than usual, whereas the La Niña (cooler SST in east Pacific) stage is associated with higher precipitation than usual (Cheng et al. 2013).
Vegetation at Serra Sul
Local vegetation in Serra Sul is susceptible to changes in precipitation due to edaphic conditions on the plateau. The modern vegetation of the area is comprised of a mosaic of edaphic dense shrubby and open shrub-bush savanna (Cleef and Silva 1994; Silva et al. 1996; Sifeddine et al. 2001; Rayol 2006; Nunes 2009) on the plateau associated with a thin lateritic crust (claylike ferruginous soil), and a transition zone between dense and open ombrophilous (evergreen) tropical forest (IBAMA 2003) in the surrounding lowlands with thicker soils and a higher availability of nutrients and accumulated water. Small forest islands within the savanna vegetation occur where organic material and water accumulates in depressions on top of the plateau. Their floristic composition is comparable to the forest on the slopes of the plateau (Rayol 2006; Nunes 2009). The Lagoa da Cachoeira vegetation is dominated by savanna and the lake is situated only 50 m from the northwestern flank of Serra Sul covered with tropical humid rainforest.
Archaeological evidence in southeast Amazonia
For the late Glacial and early Holocene, anthropogenic activity in southeast Amazonia is reported from Monte Alegre (Roosevelt et al. 1996) and Lagoa da Curuça (Behling 1996, 2001) (Fig. 1). In Serra Sul archaeological evidence indicates human occupation after 10 ka cal bp (Kipnis et al. 2005; Magalhães 2009).
Materials and methods
Radiocarbon dates from bulk marsh sediments of Lagoa da Cachoeria
14C year bp
Age (cal year bp)
2,374 ± 41
3,619 ± 40
12,414 ± 38
14,542 ± 95
21,723 ± 190
In total 200 sediment subsamples were taken at 1 cm intervals for high-resolution macro-charcoal analysis. The samples were processed following the sieving method of Stevenson and Haberle (2005). Two sieves with a mesh width of 250 and 125 µm were used to facilitate counting when large amounts of charred particles were present. For later calculations both fractions were summed to estimate past local fire activity (Whitlock and Larsen 2001). Charcoal accumulation rates (CHAR) were then calculated with Psimpoll (Bennett 1998). Subsequently, local palaeofire data were compared with the prevailing vegetation deriving from fossil pollen and fern spore analysis. For a regional comparison of palaeofire records in southern Amazonia data from 13 records from the Global Charcoal Database (Power et al. 2008) were used (Table 1). Charcoal data were selected based on their sample resolution. To allow for a comparison of the variable datasets, the data were standardized after Power et al. (2010) using standard Box–Cox transformation (homogenising variance) and a further rescaling to Z-scores. 250 year time-steps were chosen consistent with the temporal resolution of the Lagoa da Cachoeria core.
In total 24 sediment subsamples (0.5 cm3), taken every 8 cm, were used for pollen analysis for the radiocarbon dated core section from 200 to 0 cm. The samples were prepared using standard methods (Fægri and Iversen 1989) including 70 % HF treatment, addition of tablets with an exotic marker Lycopodium clavatum (Stockmarr 1971), and mounting in glycerine gelatine. Almost all samples were counted to a minimum of 300 terrestrial pollen grains. Spores and aquatic taxa are not included within the main pollen sum but expressed as percentages of the terrestrial pollen sum. The percentage pollen diagram and the zonation by CONISS (Grimm 1987) were conducted using Psimpoll (Bennett 1998). Pollen and spore identification is based on appropriate literature (Colinvaux et al. 1996; Carreira and Barth 2003; Carreira et al. 1996; Roubik and Moreno 1991) and a pollen reference collection held at the Department of Palynology and Climate Dynamics.
Chronology and zonation
Macro-charcoal and pollen data
Lower pollen concentrations are recorded from 200 to 160 cm depth with 11,000–45,000 pollen grains cm−3. Between 160 and 0 cm concentrations range from 25,000 to 60,000 pollen grains cm−3 (min. 3,500, max. 140,000). In three samples (108, 116, 124) very high concentrations between 270,000 and 640,000 pollen grains cm−3 could be recorded. Two samples (140 and 154 cm core depth) did not contain enough fossil pollen for counting and were discarded for pollen analysis (Fig. 3).
Zone LDC 1 (200–158.5 cm; 26.2–19.6 ka cal bp)
During this period almost no charcoal particles were recorded. The pollen diagram is dominated by pollen of savanna vegetation systems, mainly Poaceae, Spermacoce (max. 40 %), Amaranthaceae (incl. Chenopodiaceae; max. 16 %) and Asteraceae (max. 12 %). Also Cuphea and Myrtaceae (both < 5 %) are present. Cold adapted taxa (Podocarpus, Myrsine, Meliaceae, Hedyosmum, Euplassa) are recorded in low quantities (max. 2 %). Aquatic taxa are represented by Cyperaceae (max. 10 %). High amounts of Isoëtes (95 %) spores are recorded.
Zone LDC 2 (158.5–120 cm; 19.6–11.5 ka cal bp)
During this period almost no charcoal particles were deposited. The zone is characterized by low pollen concentrations of 3,000–37,000 pollen grains cm−3 and contained samples with little or no pollen at about 18.5 and 16.5 ka cal bp. In this period, savanna taxa are dominant, represented by Amaranthaceae (incl. Chenopodiaceae), Asteraceae, Borreria and Poaceae, which latter dominate the pollen spectra (30 %). Between 15.2 and 13.5 ka cal bp Melastomataceae/Combretaceae (max. 22 %), Moraceae/Urticaceae (max. 12 %), and cold adapted taxa like Podocarpus, Myrsine, Meliaceae and Hedyosmum are present with low values. Cyperaceae pollen is frequently present (12 %), but decrease after 15.2 ka cal bp.
Zone LDC 3 (120–97 cm; 11.5–6.7 ka cal bp)
Higher amounts of charcoal (500–1,000 particles cm−3) are recorded between 11 and 10.2 ka cal bp. These occur together with a decreased abundance of pollen from Melastomataceae/Combretaceae. Pollen of Moraceae/Urticaceae is virtually absent. In contrast, values of Anacardiaceae (5 %), Bignoniaceae (5 %), Fabaceae (10 %), Asteraceae (10 %), Rubiaceae (4 %) and Poaceae (50 %) increased. During this time the aquatic taxon and open water indicator Nymphaea also increases (max. 6 %). From 10.2 to 6.7 ka cal bp there is a decrease in Nymphaea coupled with an increase in Sagittaria (max. 20 %). This is also the time period when Poaceae pollen (80 %) dominates the pollen spectra, accompanied by Arecaceae (max. 25 %, only one sample) and low amounts of Asteraceae, Fabaceae and Cyperaceae.
Zone LDC 4a and b (97–37 and 37–0 cm; 6.7–3.4 and 3.4–0 ka cal bp)
A high abundance of charcoal is recorded for this zone, and is accompanied by a clear increase of tropical forest taxa. Pollen of the early secondary forest taxon Zanthoxylum (Marchant et al. 2002; Martins and Rodrigues 2002) is recorded with low values. The decrease in savanna taxa is mainly attributed to the decline of Poaceae. Between 6.7 and 3.4 ka cal bp declining Poaceae pollen (40 %) is still dominant together with high amounts of aquatic Echinodorus-type (75 %). After 3.4 ka cal bp Poaceae decrease to relatively low values (max. 20 %), accompanied by increasing values of Amaranthaceae (incl. Chenopodiaceae, 15 %), Myrtaceae (12 %) and Spermacoce (10 %), as well as Anacardiaceae (6 %), Melastomataceae/Combretaceae (10 %) and Moraceae/Urticaceae (12 %). From 2.2 ka cal bp onwards higher percentages of Alchornea/Aparisthmium (20 %), Celtis-type (max. 12 %), Melastomataceae/Combretaceae (max. 10 %), Moraceae/Urticaceae (15 %) and Mimosaceae (6 %) are recorded. Myrtaceae and Spermacoce pollen levels are low, as well as the aquatic Echinodorus-type.
Interpretation and discussion
Changes of palaeovegetation at Lagoa da Cachoeira are suggested to record past conditions on the plateau and at its slopes. It is uncertain how far lowland vegetation could also be represented in the pollen record, because comparative modern pollen rain studies for this area are still lacking. The recorded palaeofires are of (extra) local origin, as large charred particles of >125 and 250 µm are not transported over long distances and therefore are a reliable record for extralocal (nearby) and local fires (within the watershed) (e.g. Clark 1988; Withlock and Millspaugh 1996; Whitlock and Larsen 2001; Carcaillet et al. 2001) less than 7 km away (i.e. mainly restricted to the plateau).
Late Pleistocene and transition to early Holocene (LDC 1 and 2)
During the late Pleistocene from 26.2 to 15 ka cal bp the rare occurrence of forest taxa suggests less forested areas, possibly at the lower slopes of the plateau. A scrub-bush savanna is the dominant vegetation type, reflected by the frequent occurrence of Poaceae and Spermacoce, as well as Asteraceae and Amaranthaceae (incl. Chenopodiaceae). The occurrence of the semi-aquatic Isoëtes refers to a lower lake level until the beginning of the Holocene, indicating dry climatic conditions, which are also recorded by former studies at Serra Sul (Hermanowski et al. 2012). Between 15 and 10.2 ka cal bp the tropical forest area increased in the study area of Serra Sul, indicated by the higher occurrence of Melastomataceae/Combretaceae and Moraceae/Urticaceae. Savanna vegetation with Poaceae, Asteraceae and Myrtaceae was still present in reduced concentrations. This overlaps with the increased occurrence of forest at Serra Sul under increasingly wetter conditions between 15.4 and 11.4 ka cal bp as suggested by Sifeddine et al. (2001) from records CSS2 and CSS10. At Pantano de Mauritia about 2.3 km distant from Lagoa da Cachoeira, the forested area increased first between 11.4 and 10.2 ka cal bp (Hermanowski et al. 2012).
Early to mid-Holocene (LDC 3)
After 10.2 ka cal bp grass became a larger component of the vegetation, and the open vegetation became more floristically similar to modern savannas found in Serra Sul today. This is indicated by a decreasing sum of tropical forest taxa and an increase in the sum of savanna and seasonal dry tropical forest taxa (Pennington et al. 2001; Mayle 2006; Gosling et al. 2009) including Fabaceae, Bignoniaceae and Anacardiaceae. These taxa are able to tolerate longer dry periods than tropical forest taxa. Between 10.2 and 6.7 ka cal bp a marked reduction of tropical forest taxa and the increase of grass-dominated savanna around Lagoa da Cachoeira indicate a change to dry climatic conditions. Possibly forested areas were present on the slopes of the plateau. These mainly consisted of Arecaceae, Fabaceae, Anacardiaceae and Bignoniaceae. The low abundance of charcoal indicates local fires were not that frequent or were even absent at Lagoa da Cachoeira, whereas former studies from Pántano da Maurítia (Serra Sul) suggest regional fire activity under long phases of climatic dry conditions (Hermanowski et al. 2012).
Measurements of the molecular marker levoglucosan, which is emitted in high amounts during the burning of fuel containing cellulose, point to fire events at Serra Sul at 7 and 5 ka cal bp (Elias et al. 2001), coincident with forest fires at Serra Norte dos Carajás from 7.5 to 4.7 ka cal bp (Cordeiro et al. 2008). An opening of the forest was also suggested by Sifeddine et al. (2001) interpreted as the result of alternating dry and short humid periods between 7.9 and 9.4 ka cal bp.
Mid- to late Holocene (LDC 4)
After 6 ka cal bp both tropical forests and savannas are present. Forests are more diverse, with a relatively high abundance of Fabaceae and Anacardiaceae. The presence of Alchornea/Aparisthmium and the pioneer Celtis, together with low abundance of Melastomataceae/Combretaceae and Moraceae/Urticaceae could be indicative of a first repopulation of a formerly open habitat (Hermanowski et al. 2012), but also of an increasing human influence through the presence of hunter-gatherers (Kipnis et al. 2005; Magalhães 2009). After 3.4 ka cal bp the increased abundance of tropical rainforest taxa indicates the establishment of modern rainforests in Serra Sul. The pioneer trees Celtis and Trema (Marchant et al. 2002) point to an extension of the forested area, though moderate forest clearance by humans, e.g. for hunting-gathering purposes or forest cultivation, cannot be completely excluded. High concentrations and accumulation rates of charcoal point to frequent fires in this period of modern rainforest establishment.
The increased abundance of Moraceae/Urticaceae and Alchornea/Aparisthmium suggests reduced water stress due to intensified precipitation. Absy et al. (1991) recorded forest development around 3.1 ka cal bp at Serra Sul about 3.5 km distant from Lagoa da Cachoeira. Changes in precipitation are supported by rising lake levels at Serra Norte dos Carajás from 2.8 to 1.3 ka cal bp (Cordeiro et al. 2008).
Linkage between local fires, vegetation, humans, and climate
The most challenging part in the reconstruction of past fire activity is the determination of the driving factor for recorded fires. Especially when the archaeological evidence suggests that humans could have been involved, it is difficult to differentiate between a climatic and/or anthropogenic origin of palaeofires.
Peak fire at the glacial-interglacial transition
Climatic influence The climate during this period was seasonal with a pronounced wet season. Thunderstorm lightning ignitions were frequent at the transition from dry to wet season. Taxa known from seasonal dry forests with adaption to low intensity fires (Pinard and Huffman 1997) would then have had a competitive advantage over other tropical forest taxa.
Anthropogenic influence Humans could have influenced the vegetation species composition by setting fire to clear land to potentially improve hunting grounds. Archaeological evidence indicates human occupation at Serra Sul after 10 ka cal bp (Kipnis et al. 2005; Magalhães 2009). It is possible however that humans arrived even earlier in Carajás, as evidence for anthropogenic activity in eastern Amazonia during the late Glacial and early Holocene is reported from activity in Monte Alegre (Roosevelt et al. 1996) and Lagoa da Curuça (Behling 1996, 2001).
Combination of climatic and anthropogenic influence Seasonal climatic conditions with pronounced wet seasons were coupled with the growth of forest and the temporary presence of humans who additionally influenced local fire history in Serra Sul. The combination of favorable climate and anthropogenic influence facilitated the growth of pioneers, and plants with adaptions to low intensity fires.
Such environments could have existed in Serra Sul during transition periods between drier to wetter climates, and under a seasonal climate with distinct dry and wet seasons when available soil moisture was sufficient to provide the growth of tropical forest and shrubby savannas. The source of ignition could be provided by thunderstorms with lightning strikes, as they are common during wet seasons, and at the transition from dry to wet season, when enough climatically dried fuels would be available to catch fire after lightning strikes (Stott 2000).
We note when local palaeofires at Lagoa Cachoeira are compared with previous studies from Pántano da Maurítia (both Serra Sul), the recorded fires between 11 and 10.2 ka cal bp at Lagoa da Cachoeria are coupled with a short-term occurrence of wet tropical forest. At the same time and under the same vegetational conditions no fires (local and regional) were recorded at Pántano da Maurítia (Hermanowski et al. 2012), located 2 km south of Lagoa da Cachoeira on the same plateau. It is difficult therefore to make a direct correlation between a higher or lower fire frequency and dry or wet climatic conditions.
Low local fire activity during the mid-Holocene
Climatic influence If we assume a natural origin of local fire activity, the reason for low or absent local fires could be a decreased frequency of thunderstorms (source of ignition) in a period with longer dry seasons (Hermanowski et al. 2012). Under drier conditions fire prone forest margins would be also situated further downslope, out of the reach of a fireline that would be hindered by the sloping hillside (Stott 2000). Additionally, due to the absence of ligneous savanna shrubs as a significant fuel load, a natural ground fire would not produce such a large amount of charcoal particles as would be the case in a shrubby savanna.
Anthropogenic influence If local fires are primarily coupled with human activity one reason could be the abandoning of this area because of unbearable dry conditions. Site abandonment due to dry conditions during the mid-Holocene is also suggested by Irion et al. (2006) at Rio Tapajós (eastern Amazonia, Fig. 1).
Frequent fires in the late Holocene
Climatic influence Under a seasonal climate wet seasons were long enough for the growth of forest and shrub-bush savanna known from today in Serra Sul. During the dry season drying of this biomass would provide enough fuel load for natural fires, when thunderstorms at the transition from dry to wet season provided sufficient sources of ignition for the combustion of dried fuel (grass, litter) load.
Combination of climatic and anthropogenic influence The above mentioned boundary conditions are also favourable for pre-Columbian societies that could have already been using forest management strategies comparable to the modern Kayapó Indians of the Brazilian Amazon who also use fire for the establishment and protection of tropical forest patches (apêtê) in a campo/cerrado environment (Posey 1985). Therefore, it is difficult to identify whether palaeofires in Serra Sul were primarily of natural or anthropogenic origin.
Regional fire history of southern Amazonia
Proxy records included in the reconstruction of fire history in southern Amazonia
Tropical north Atlantic
12°05′N, 61°15′W; 1,299 m b.s.l.
Rühlemann et al. (1999)
Tropical north Atlantic
20°44′60.00″N, 18°34′59.99″W; 2,263 m b.s.l.
De Menocal et al. (2000)
Tropical south Atlantic
4°36′48.00″S, 36°38′12.00″W; 830 m b.s.l.
Weldeab et al. (2006)
Cariaco basin, N-Venezuela
10°41′60.00″N, 65°56′30.01″W; 790 m b.s.l.
Lea et al. (2003)
Cariaco basin, N-Venezuela
10°42′N, 65°10′ W; 893 m b.s.l.
Haug et al. (2001)
Santa Catarina, S-Brazil
27°13′S, 49°09′W; 250 m a.s.l.
Wang et al. (2007)
Cueva del Tigre Perdido
Nueva Cajamarca, San Martín, Peru
van Breukelen et al. (2008)
Speleothem δ18O values from Peru (van Breukelen et al. 2008) and southern Brazil (Wang et al. 2007) are used as a record for moisture input from Amazonia to the system of South American Summer Monsoon (SASM), as high (low) δ18O values point to weaker (strengthened) SASM due to less (more) moisture input from Amazonia (Cruz et al. 2005; Cheng et al. 2013).
Lower (higher) Ti values from the Cariaco Basin in northern Venezuela reflect weaker (stronger) rainfall in this region due to a southward (northward) shift of the ITCZ (Haug et al. 2001).
Reconstructed SSTs from the Cariaco Basin (Lea et al. 2003) and from the Atlantic Ocean (Rühlemann et al. 1999; De Menocal et al. 2000; Weldeab et al. 2005, 2006) are incorporated because of their influence on Amazonian rainfall regimes. The comparison is used to explain differences between the fire histories of southwest and southeast Amazonia which currently also exhibit differences in rainfall distribution (Marengo et al. 2001).
The reconstructed meridional (NW-SW) and zonal (NE-NW) ΔSSTs of the Atlantic show that the southern Atlantic Ocean was generally warmer between 11 and 10.2 ka cal bp (Fig. 5d). When meridional ΔSST was rather high during this time, zonal ΔSST was slightly weaker. This weaker zonal gradient would suggest weakened northeast trade winds coupled with weakened Amazon Basin convection (Baker et al. 2001b). We suggest that the described Atlantic SST conditions could be the reason for a more southerly position of the ITCZ, which would correspond to wetter conditions in southeastern Amazonia.
At Cachoeira during the mid-Holocene from 9 to 6 ka cal bp, local fire activity was at a minimum. On a regional scale, fire activity was higher in southwest Amazonia than in the southeast (Fig. 5a), which coincided with a weakened SASM in southern Brazil (Wang et al. 2007), higher SST in the Cariaco Basin (Lea et al. 2003) and wetter conditions in northern Venezuela (Haug et al. 2001) (Fig. 5b, c).
Between 9 and 8 ka cal bp more frequent fires in the southeast were accompanied by a lesser fire frequency in the southwest, whereas in both regions fires were generally frequent after 6.3 ka cal bp. During both periods SASM was intensified, only slightly between 9 and 8 ka cal bp, but all the more after 6.3 ka cal bp. Zonal ΔSSTs in the tropical Atlantic around 8.25 ka cal bp show a clear cold pool in the western equatorial Atlantic, roughly coinciding with the abrupt climate change around 8.2 ka cal bp (Alley et al. 1997; Alley and Ágústsdóttir 2005) and a Bond event around 8.1 ka cal bp (Bond et al. 1997), respectively. For these conditions a southward displacement of the ITCZ is suggested (Alley and Ágústsdóttir 2005). Around 8 ka cal bp meridional ΔSST (Fig. 5d) shows a shift from a formerly warmer southwestern to a generally warmer northwestern tropical Atlantic. A warmer tropical north Atlantic could suggest a general northward shift of the ITCZ, and together with still cold SSTs in the eastern equatorial Atlantic (Weldeab et al. 2005), suggesting a delayed onset of central Amazon convection (Fu et al. 1999, 2001), this would result in longer dry seasons in southeast Amazonia after 8 ka cal bp. Differences between southeast and southwest Amazonian fire history may also be attributed to weaker ENSO (El Nino/Southern Oscillation) activity (Sandweiss et al. 1996; Keefer et al. 1998; Rodbell et al. 1999; Clement et al. 2000; Sandweiss et al. 2001; Otto-Bliesner et al. 2003) that influenced the southwest more than the southeast, but the temporal resolution of the present record (ca. 250 years) is not high enough to recognize connections between fire and El Niρo events in the past. Also weaker low level jets from central Amazonia to the eastern Andes and a weaker Bolivian high (Dias de Melo 2007) may have played a role for different fire histories in SE- and SW-Amazonia.
After 6.3 ka cal bp, when local fires were recorded at Lagoa da Cachoeira after a long period of markedly less local fire activity, palaeofires were frequent in both southeast and southwest Amazonia. In the southeast no distinctive pattern of palaeofire activity can be recognized. Here, climatic conditions were increasingly seasonal with severe dry seasons until 3.7 ka cal bp (Hermanowski et al. 2012). Intensified SASMs (Wang et al. 2007) and wetter conditions in northern Peru (van Breukelen et al. 2008) refer to generally wetter conditions in the Amazon Basin, corresponding to slightly drier conditions in northern Venezuela (Haug et al. 2001).
After 3.7 ka cal bp enhanced fire frequency in southeastern and southwestern regions overlaps with the expansion of modern rainforest (Mayle et al. 2000; Burbridge et al. 2004; Sifeddine et al. 2001; Cordeiro et al. 2008; Hermanowski et al. 2012). In the tropical Atlantic meridional ΔSST slightly decreased around 6 ka cal bp, whereas zonal ΔSST increased (~1 °C). Afterwards zonal ΔSST decreased markedly about 4 °C contemporaneously with a steadily increasing meridional ΔSST (Fig. 5d). Due to lower zonal ΔSST northeast trades were possibly extenuated which would weaken convection over the Amazon basin.
Differences between the composite palaeofire data from southeast Amazonia with our record from Serra Sul may be partly attributed to the latitudinal distribution of the available records (Fig. 1; Table 2). With the exception of Lagoa da Cachoeira (6°S), the other records used for this study are situated near the equator (0.6–2.8°S). Only two of them are older than 8,000 years (Behling 2001; Bush et al. 2007c; Irion et al. 2006), and human induced forest fires during pre-Colombian settlements could have played an important role there.
For the reconstruction of long-term fire history in southern Amazonia the various interactions between fire, prevailing vegetation, regional climate, and anthropogenic influence need be addressed. The new study from Serra Sul dos Carajás in southeast Amazonia shows, that a clear correlation between wetter (drier) climates and lower (higher) frequency of palaeofires cannot be easily drawn. Most frequent fires recorded at Lagoa da Cachoeira occurred during seasonal and generally wetter periods from 11 to 10.2 and after 5 ka cal bp, when moist tropical forest and savanna vegetation coexisted at Serra Sul. The crucial factor for local fires seems to be the amount of time, when sufficient moisture is available to foster the growth of woody vegetation. This vegetation then serves as fuel for fire, naturally and/or anthropogenically induced. We suggest that these conditions are both favorable for natural fires and the presence of human societies in Serra Sul. Also the archaeological evidence suggests that humans could have influenced vegetation by using fire, as presence of hunter-gatherers in Serra Sul is recorded after 10 ka cal bp (Kipnis et al. 2005; Magalhães 2009). Markedly less frequent fires at Lagoa da Cachoeira under dry early-mid-Holocene conditions between 10.2 and 6.7 ka cal bp could be attributed to less fuel load in a more grass dominated savanna and less sources of ignition (lack of thunderstorms). On the other hand it could be also explained by site abandonment by humans due to unbearably dry conditions.
In both cases changing precipitation patterns would be the driving factor for the presence/absence of humans and both natural and/or anthropogenic fire. The comparison of the present study with fire records from Amazonia (southeast and southwest) and proxy records of past precipitation suggests connections between fire history in southern Amazonia and the seasonality of precipitation. Composite fire records from southeast and southwest Amazonia also suggest a sensitivity of these regions to changing conditions in the Atlantic Ocean at least since 11.5 ka cal bp, due to the strong influence of these on precipitation regimes. We note that paleofire activity is antiphased in southeast Amazonia in the time window around 8.2 ka cal bp, when for a short time (ca. 200 years) climate was cooler due to freshwater input to the North Atlantic (Bond et al. 1997; Urrego et al. 2009).
The different fire histories of southeast versus southwest Amazonia show that in the southwest palaeofire activity is generally higher throughout the Holocene, which may be also partly attributed to an increased frequency of ENSO warm events (Moy et al. 2002). But the underlying factor for Holocene fire frequency seems to be the influence of zonal and meridional SST gradients on Amazonian precipitation regimes, which could have not only influenced past vegetation composition, but also Pre-Colombian human societies entering Serra Sul. For a comprehensive reconstruction of fire history further high-resolution palaeofire studies south of 5°S in eastern Amazonia are needed. To clarify the role of Pre-Colombian societies in the Carajás region archaeological sites in Serra Sul need to be studied intensively.
Thanks to Martin Zweigert for assistance in pollen preparation, and Malte Semmler for the introduction to Clam software. The CNPq is thanked for fieldwork support and funding of the second author (Proc. 471 109/03-7). We also thank the Vale do Rio Doce company for logistical support and IBAMA for the permission to carry out fieldwork in the reserve Serra Sul dos Carajás. Funding was provided by the German Research Foundation (DFG project BE-2116/11-1).
- Absy ML, Cleef A, Fournier M, Martin L, Servant M, Sifeddine A, Da Silva F, Soubiès F, Suguio K, Turcq B, Van der Hammen T (1991) Mise en évidence de quatre phases d’ouverture de la forêt dense dans le sud-est de L’Amazonie au cours des 60,000 dernières années. Première comparaison avec d’autres régions tropicales. CR Acad Sci Paris II 312:673–678Google Scholar
- Alley RB, Ágústsdóttir AM (2005) The 8 k event: cause and consequences of a major Holocene abrupt climate change. Quat Sci Rev 24(123–1):149Google Scholar
- Bennett KD (1998) Psimpoll 4.10 and Pscomb 1.03—C programs for plotting pollen diagrams and analysing pollen data. http://www.chrono.qub.ac.uk/psimpoll/psimpoll.html
- Carreira LMM, Barth OM (2003) Atlas de Pólen da vegetação de canga da Serra de Carajás (Pará, Brasil). Museu Paraense Emílio Goeldi, BelémGoogle Scholar
- Carreira LMM, Da Silva MF, Lopes JRC, Nascimento LAS (1996) Catálogo de Pólen das Leguminosas da Amazônia Brasileira. Museu Paraense Emílio Goeldi, BelémGoogle Scholar
- Cheng H, Sinha A, Cruz FW, Wang X, Edwards RL, d’Horta FM, Ribas CC, Vuille M, Stott LD, Auler AS (2013) Climate change patterns in Amazonia and biodiversity. Nat Commun 4:1–6Google Scholar
- Chiang JCH, Kushnir Y, Giannini A (2002) Deconstructing Atlantic ITCZ variability: influence of the local cross-equatorial SST gradient, and remote forcing from the eastern equatorial Pacific. J Geophys Res 107:1–19Google Scholar
- Cleef A, Silva MFF (1994) Plant communities of the Serra dos Carajás (Pará), Brazil. Boletim do Museu Paraense Emílio Goeldi, série Botânica 10:269–281Google Scholar
- Cook KH (2009) South American climate variability and change: remote and regional forcing processes. In: Vimeaux F, Sylvestre F, Khodri M (eds) Past climate variability in South America and surrounding regions. Developments in Paleoenvironmental Research 14. Dordrecht, pp 193–212. doi:10.1007/978-90-481-2672-9
- de Melo MLD (2007) Simulacões de clima para o Holoceno Médio usando o MCGA do CPTEC, com ênfase sobre a América Sul. Tese de Doutorado do Curso de Pós-Graduação em Meteorologia, Instituto Nacional de Pescisas Especiais 15206-TDI/1306. http://mtc-m17.sid.inpe.br/col/sid.inpe.br/mtc-m17@80/2007/220.127.116.11/doc/publicacao.pdf
- Fægri K, Iversen J (1989) Textbook of pollen analysis, 4th edn. Wiley, New YorkGoogle Scholar
- Garreaud R, Vuille M, Compagnucci R, Marengo J (2009) Present-day South American climate. Palaeogeogr Palaeoclimatol Palaeoecol 281:80–195Google Scholar
- Gouveia SEM, Pessenda LCR, Aravena R, Boulet R, Scheel-Ybert R, Bendassoli JA, Ribeiro AS, Freitas HA (2002) Carbon isotopes in charcoal and soils in studies of paleovegetation and climate changes during the late Pleistocene and the Holocene in the southeast and centerwest regions of Brazil. Glob Planet Chang 33:95–106CrossRefGoogle Scholar
- IBAMA (2003) Instituto Brasileiro do Meio Ambiente e dos Recursos Naturais Renováveis. Plano de manejo para uso múltiplo da floresta nacional de Carajás. http://www.ibama.gov.br/
- INMET (2011) Instituto Nacional de Meterologia, Monitoramento das Estações Convencionais Ministério da Agricultura, Pecuária e Abastecimento. http://www.inmet.gov.br/
- Kipnis R, Caldarelli SB, De Oliveira WC (2005) Contribuição para a cronologia da colonização amazônica e suas implicações teóricas. Rev Arqueol 18:81–93Google Scholar
- Magalhães MP (2009) Evolução antropomorfa da Amazônia. Rev Hist Arte Arqueol 12:5–38Google Scholar
- Mayle FE (2006) The Late Quaternary biogeographical history of South American seasonally dry tropical forests: insights from palaeo-ecological data. In: Pennington RT, Lewis GP, Ratter JA (eds) Neotropical Savannas and seasonally dry forests: plant diversity, biogeography, and conservation. Systematics Association Special Volume Series 69. CRC, Boca Raton, pp 395–416Google Scholar
- Nunes JA (2009) Florística, estrutura e relações solo-vegetação em gradient fitofisionômico sobre canga, na Serra Sul, FLONA de Carajás – Pará. Dissertação apresentada á Universidade Federal de Viçosa, como parte das exigências do Programa de Pós-Graduação em Botânica, para obtenção do titulo de Magister Scientiae. http://www.bibliotecaflorestal.ufv.br/bitstream/handle/123456789/2813/163101_c.pdf?sequence=2
- Otto-Bliesner BL, Brady EC, Shin S-I, Liu Z, Shields C (2003) Modeling El Niño and its tropical teleconnections during the last glacial-interglacial cycle. Geophys Res Lett 30:1–4Google Scholar
- R Development Core Team (2008) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0. http://www.R-project.org
- Rayol BP (2006) Análise florística e estrutural da vegetação xerofítica das savannas metalófilas na Floresta Nacional de Carajás; subssídios à conservação. Dissertação de mestrado em Botânica com area de concentração em Botânica Tropical pela Universidade Federal Rural da Amazônia e Museu Paraense Emílio Goeldi. http://marte.museu-goeldi.br/zoologia/turma2004/dissertacaoBrenoRayol.pdf
- Reimer PJ, Baillie MGL, Bard E et al (2004) IntCal04—terrestrial radiocarbon age calibration, 0–26 cal kyr bp. Radiocarbon 46:1,029–1,058Google Scholar
- Roosevelt AC, da Costa ML, Lopes Machado C, Michab M, Mercier N, Valladas H, Feathers J, Barnett W, da Silveira MI, Henderson A, Sliva J, Chernoff B, Reese DS, Holman JA, Toth N, Schick K (1996) Paleoindian cave dwellers in the Amazon: the peopling of the Americas. Science 272:373–384CrossRefGoogle Scholar
- Roubik DW, Moreno E (1991) Pollen and Spores of Barro Colorado Island. Monographs in systematic botany from the Missouri Botanical Garden 36. Missouri Botanical Garden, St. Louis, MissGoogle Scholar
- Silva MFF, Secco R, Lobo MGA (1996) Aspectos ecológicos da vegetação rupestre da Serra dos Carajás, Estado do Pará, Brasil. Acta Amazonica 26:17–44Google Scholar
- Soubiès F (1979) Existence d’une phase seche en Amazonie Bresilienne datee par la presence de charbons dans les sols (6.000–8.000 ans bp). Cah. O.R.S.T.O.M., sér. Géology 1979–1990:133–148Google Scholar
- Stevenson J, Haberle SG (2005) Macro charcoal analysis: a modified technique used by the Department of Archaeology and Natural History. PalaeoWorks Technical Report 5:8. http://palaeoworks.anu.edu.au/paltr05.pdf
- Stockmarr J (1971) Tablets with spores used in absolute pollen analysis. Pollen Spores 13:615–621Google Scholar
- Urrego DH (2006) Long-term vegetation and climate change in Western Amazonia. Ph.D. Dissertation. Department of Biological Sciences, Florida Institute of Technology Melbourne, FL, USA. 278 pGoogle Scholar
- Urrego DH, Bush MB, Silman MR, Correa-Metrio A, Ledru M-P, Mayle FE, Paduano G, Valencia BG (2009) Millennial-scale ecological changes in tropical South America since the Last Glacial Maximum. In: Vimeux F, Sylvestre F, Khodri M (eds) Past climate variability in South America and surrounding regions. Developments in Paleoenvironmental Research 14, Berlin, Heidelberg, pp 283–300Google Scholar
- Wang X, Auler AS, Edwards RL, Cheng H, Ito E, Wang Y, Kong X, Solheid M (2007) Millennial-scale precipitation changes in southern Brazil over the past 90,000 years. Geophys Res Lett 34:1–5Google Scholar
- Whitlock C, Larsen C (2001) Charcoal as a fire proxy. In: Smol JP, Birks HJB, Last WM (eds) Tracking environmental change using lake sediments. Terrestrial, algal, and siliceous indicators, vol 3. Kluwer Academic Publishers, DordrechtGoogle Scholar