Long-term vegetation dynamics of a tropical megadelta: Mid-Holocene palaeoecology of the Orinoco Delta (NE Venezuela)

Coastal wetlands have been proposed as highly threatened by the ongoing and future climatic change, including projected sea-level changes as an additional forcing factor compared to more inland locations. The limited knowledge generated to date in this topic has been primarily focused on those areas attaining a high population density, and rarely deals with long-term (>50 years) dynamics. Here we present the first Holocene palaeoecological study carried in the Orinoco Delta, in NE Venezuela. The record presented here contains sediments from the last 6200 years and is located in a river-shore swamp dominated by the palm Mauritia flexuosa. Current human occupation is almost restricted to small settlements of the Warao indigenous culture, closely related to the use of M. flexuosa and other palm species present in the zone. The results show the occurrence of three well-distinguished palynological zones: (i) from 6200 to 5200 cal yr BP, characterised by mixed rainforest and other taxa related to salinity (coastal-like), low (negative) values of magnetic susceptibility and magnetic grain size, absence of transported clays, and the highest macrocharcoal particles abundance; (ii) from 5200 to 2950 cal yr BP, marked by a replacement of the mangrove-like vegetation by a more inland mixed-swamp forest community with low levels of charcoal, and (iii) from 2950 cal yr BP to present-day, characterised by the establishment of the current vegetation community, dominated by M. flexuosa, and an increasing trend in the charcoal curve since the last 700 years. A combination of regional (climatic changes) and local (sediment ontogeny) has been proposed as the key drivers influencing the vegetation succession recorded. The stabilisation of the sea-level that occurred during the mid-Holocene would have favoured the transgression of the coastal line, with the migration of the coastal-like vegetation seawards. Synchronous to this event, a trend towards drier conditions has been reported in the close Cariaco record, that could have also influenced the vegetation replacement. Between 3800 and 2800 years ago, the increased ENSO variability registered in Cariaco may have played a key role in the expansion of the Mauritia palm community. It is suggested that in our location, the potential inhabiting human populations were differently influenced by these environmental changes. First, the disappearance of the coastal resources could have favoured land abandonment, whereas the increase in the abundance of the palm might be influential for the arrival of other inland cultures that were previously used to manage Mauritia. This sequence shows the importance of the ecosystem services for the location inhabitants, highlighting the abandonment of the mid-Holocene culture coeval with the disappearance of its ecosystem. These results also provide information about the sensitivity and resilience in facing external stressors of both humans and vegetation, and will be valuable tools for managing the future of this ecosystem.


Introduction
Current climate change is a global phenomenon that will have a devastating influence on the planet and the way we perceive it nowadays (IPCC, 2013). As an attempt of anticipating, the IPCC last report highlights the importance of studying the dynamics occurred in the past that are preserved in the fossil sedimentary records as analogues to infer potential future directions that the Earth system total nitrogen (TN) (relative standard deviation: 5% of the measurements), δ 13 C and δ 15 N analyses (0.2‰ analytical precision). These analyses were conducted using a Finnigan DELTAplus TC/EA-CF-IRMS spectrometer at the Centres Científics i Tecnològics of the Universitat de Barcelona (Barcelona, Spain). The carbonate content of the samples was below the detection limit (<1%) of the XRD analysis. Therefore, the total carbon (TC) was considered equivalent to the total organic carbon (TOC).
Statistical analyses were performed in R version 3.82 using the packages "vegan" (Oksanen et al., 2013), "rioja" (Juggins, 2017), and "analogue" (Simpson, 2007). For statistical analyses, biological data consisted of the percentages data squared root transformed, with downweight of rare taxa in the ordination analyses. Geochemical data (XRD, rock-magnetism) were standardised. Stratigraphic diagrams were plotted with R and Psimpoll 4.27 (http://chrono.qub.ac.uk/psimpoll/psimpoll.html) and zonation was performed by "Optimal Splitting by Information Content" (OSIC), using the broken stick method to determine the significant zones (Bennett, 1996). Diversity measures include N0 (richness or species number, also called S), N1 (number of common taxa) and N2 (number of dominant taxa), calculated following Hill (1973), and the evenness ratio N2/N0 after Finsinger et al. (2017). For calculating these indices, the dataset used was the percentage data without downweight of rare taxa in order to capture the total diversity values. A Principal Component Analysis (PCA) was performed considering only the XRD data to characterise the main sedimentary processes that have ruled the sedimentation in the coring site that has not been included in the present paper (scores of axis 1 represented in Fig. 8).

Sediment description, chronology and geochemical data
The retrieved sediments are composed by greyish to black clays and silty clays with variable proportions of organic matter, which can form brownish centimetre-thick peat layers (Fig. 2). These sediments have been deposited within a floodplain environment. Locally, fine sand centimetre-thick layers have been found, mostly related to punctual avulsion episodes. According to the age-depth model, this sedimentary record contains the environmental history for the last 6200 cal years BP (Fig. 2). Sedimentation rate varies between 0.016 and 0.067 cm yr -1 , providing a sample resolution of 78-154 years per sampling interval. Fig. 3 shows the geochemical analyses (including different interparametric ratios based on rockmagnetism properties) and will be explained following the statistically significant zones found based on the XRD results. As expected, rock-magnetism indicates that the dominant magnetic mineralogy in the entire sequence is magnetite, as suggested by the IRM/K values obtained and the different concentration parameters' trends observed (Fig. 3), with some punctual inputs of a highcoercivity minerals (most probably hematite) around 25 and 45 cm of core depth. The coincidence of illite trends with the magnetite content confirms the detrital origin of the second mineral. The base of the record (Zone OR18-A: 141-106.5 cm depth) is characterised by high values of kaolinite and chlorite and absence of illite. Negative values of magnetic susceptibility (MS), characteristic of organic-rich sediments dominated by diamagnetic minerals, is observed in this zone, confirming the geochemical data. Total Carbon (TC) and Total Nitrogen (TN) show the highest values of the entire sequence during this zone, especially from 130 cm upwards the top of the zone. Together with the increase in TC content a sharp increase in the concentration of low-coercivity minerals (magnetite) is observed. Altogether the results indicate that the ARM (magnetite content) increase observed in the lower part of the sequence cannot be interpreted as a detrital magnetite. Zone OR18-B (106.5-0 cm depth) is marked, in the lower section (106.5-82 cm), by a decrease in kaolinite and chlorite and increase in illite and quite homogeneous magnetic properties dominated by low-coercivity minerals (e.g. magnetite) of detrital origin. This is confirmed by the clear correlation between the illite content and magnetite concentration in the middle part of the section (Fig. 3). This lower interval is

Modern samples
The modern samples show distinct palynological assemblages that reflect the different environment and vegetation communities in which they were collected ( particles cc -1 ), while CT3 attains a high value (2690 particles cc -1 ; Fig. 4).

Sedimentary record PATAM18_A12
The fossil pollen assemblages obtained in the sedimentary record of PATAM18_A12 delimits three significant zones (Fig. 5). Microcharcoal values follow the same trends than the macro-charcoal analysis, and given the broader geographical provenance of the microcharcoal particles, only the small particles (indicative of fires occurring from a larger spatial scale) have been included in the diagram for comparison.
Since 2950 cal yr BP onwards, the vegetation of Caño Tigre is represented by a Mauritia-dominated swamp, which is the presentday plant community (indicated by the similarity to the modern samples; Fig. 6). However, once established, several shifts have occurred maintaining the dynamism of the plant community until shaping the present landscape. Noteworthy are: i) the appearance and increase of aquatic elements during the last 1400 years (Fig. 5), high enough to change the isotopic and magnetic signals of the sediment (Fig. 7), and ii) the increase in Combretaceae/Melastomataceae and Spathiphyllum 700 cal yr BP. coeval to the increment in the fire occurrence (Fig. 5). Such synchrony points to a possible causal relationship, especially regarding the last taxon, inhabitant of the understory, that will be discussed in the last section of the discussion.
Caño Tigre has shown three different stable plant communities, in the sense of being present in the location without major structural changes during several millennia. It is noteworthy that although the changes occurred were dramatic and probably had profound consequences in the forest structure or ecosystem services provided, they were neither abrupt nor fast. Thus, the coastal-like forest was replaced by a mixed forest between 5400 and 4900 cal yr BP, whereas the expansion of Mauritia and the establishment of the palm swamp took almost 1000 years, between 3800 and 2900 cal yr BP. Nevertheless, these long-term changes are referred to plant community shifts. Regarding individual taxa responses, it can be observed the abruptness in the appearance and expansion of Mauritia, Combretaceae/Melastomataceae or Spathiphyllum, and in the disappearance of Rhizophora, Cassipourea or Urticales, all faster than the centennial sampling resolution performed ( Fig. 2). These results are in agreement with previous studies highlighting the need of long-term research in order to observe real plant communities' dynamics applicable to answer todays' ecological and conservation questions (Rull et al., 2013;Vegas-Vilarrúbia et al., 2011). Within the dynamics observed, the expansion of Mauritia occurred around 3000 years ago and the establishment of the current landscape resulted in a less diverse plant community. This is clearly appreciated in the diversity indices included (Fig. 8), where despite the total number of species being regular through the entire record (N0), the almost monospecific character of the palm-swamp drastically reduced the evenness of the plant community (N2/N0). Local conditions have been proposed as key drivers of the shift from mixed mangrove-like communities to monospecific assemblages in Colombian coastal records (Urrego et al., 2013), although referred to changes towards mangrove species dominance, not Mauritia's as in the present study. In the following sections, we will infer some of the key drivers responsible for the observed vegetation responses.

Ecological consequences of Holocene geomorphology, sea-level and climate changes
The onset of the Holocene was characterised by a eustatic sea-level around 60 m below present, and rose at a relatively stable rate until 7000 cal yr BP, with a progressive decrease afterwards to present (Khan et al., 2017). Holocene sea-level changes in the Caribbean show that for the Orinoco region, there was a rapid relative sea-level rise (SLR) in the early to mid-Holocene (from -6.8 ± 3 m at 7.8 cal kyr BP to -1 ± 0.9 by 5.9 cal kyr BP), remaining relatively constant afterwards (Khan et al., 2017). Most of coastal Amazonian palaeoecological studies highlight the importance of Holocene sea-level trends in the past vegetation responses observed (Cohen et al., 2012). Due to the proximity of the coring site to the present coastline ( Fig. 1), it is logical to explore the potential effects of sealevel rise on the vegetation dynamics, as it has been highlighted by Behling and Lima da Costa (2000) for Amazonian wetlands not strictly coastal. We suggest that our coring site was located closer to the coastline prior the stabilisation of the sea-level and thus during the occurrence of the coastal-like plant community, according to previous geomorphologic works that have suggested that the northwest region of the Orinoco Delta prograded around 20e30 km during the late Holocene (Aslan et al., 2003). The stabilisation of the sea-level likely provoked the transition from marinedominated sedimentation environment to a fluvial floodplain one with the formation of new soils and the progradation of the coast shoreline (Yulianto et al., 2005). This environmental shift was probably involved in the migration of the coastal-like elements seawards as it has been previously reported (Behling and Lima da Costa, 2001;Tissot and Marius, 1992). Considering the potential effect of the sea-level change, there was a delay or time-lag between the driver (sea-level stabilisation) and the ecological response (migration of the coastal-like taxa and appearance and establishment of Mauritia) of about 500 years in Caño Tigre (from 5900 to 5400 cal yr BP).
However, it is highly unlikely that the sea-level change acted as the sole force driving the vegetation change. Under this scenario of sea-level and soil stabilisation, regional climatic trends were likely also involved. The Cariaco basin record has proven its usefulness and accuracy in registering supracontinental past climatic trends, and is located very close to our study area (Fig. 1). The titanium (Ti) curve of Cariaco shows a decreasing trend in precipitation and strong El Niño signal around mid-Holocene starting around 5400 cal yr BP (Haug et al., 2001), coinciding with the onset of the vegetation replacement. This is also observed more locally in the scores of PCA first axis of the XRD data (Fig. 8) as well as the geochemical and rock-magnetic data (Fig. 3), that shows the absence of Illite and can be interpreted as some humidity in the soils and negative MS values related to organic-rich sediments. The absence of transported clays and the observed enhanced magnetite content in the lowermost part of the core suggests the presence of authigenic magnetite.
We suggest that the replacement of coastal-like elements by a mixed-swamp forest, with the arrival and establishment of Mauritia was influenced by the joint action of at least two interconnected, scale-dependent processes: i) sea-level stabilisation promoting stable conditions for new soil formation (local process), suitable for the arrival of new taxa, and ii) registered precipitation decrease with strong El Niño acting as a trigger (regional process), favouring its establishment with regards to other species (Urrego, 2018). As already mentioned, mid-Holocene mangroves colonisation has been reported in many Caribbean works highlighting the importance of local factors compared to regional sea-level rise in the establishment of coastal plant communities (Urrego et al., 2013). In this sense, Emilio et al. (2014) showed the importance of soil conditions to explain site-to-site variability in the observed current plant communities. These authors revealed in a macro-study that young, weakly structured and poor soils favour higher palm basal area compared to trees (Emilio et al., 2014). The importance of the soils has been also proved in a local scale by Vegas-Vilarrúbia and López Laseras (2008) who tested the importance of edaphic conditions to interpret the large b-diversity of the lower Orinoco Delta plain that can be observed today. It is possible that the nature and quality of the formed soils favoured the arrival of Mauritia and the increase of Euterpe. In this sense, finer-grain sediments accumulation resulted from the delta progradation could have formed water-saturated (or poorly drained) sediments, and the combination with a decrease in organic matter (Fig. 3) might have favoured the Mauritia establishment (Aslan et al., 2003;Urrego, 2018). Whatever the nature of the sediment prior and after the sea-level stabilisation, a change in its characteristics is evident by the zonation resulted in the geochemical data (Fig. 3). The potential socioeconomic and ecological threats of current and future sea-level rise to deltas have been already highlighted by Vegas-Vilarrúbia and Rull (2016). Our study shows the local habitat disappearance resulted at least partially from the Holocene sea-level rise, which could be used as past analogue for future scenarios (Vegas-Vilarrúbia et al., 2011;Rull et al., 2013).
After the disappearance of coastal-like elements, another dramatic event in the Caño Tigre record was registered with the expansion of Mauritia at the expenses of the mixed-forest between 3800 and 2900 cal yr BP. Based on the Cariaco Ti curve, this period registered the maximum amplitude of the climatic trends during the Holocene caused by an increased ENSO variability that included several precipitation minima during this period, and was preceded by a strong El Niño (Haug et al., 2001). According to Haug et al. (2001), such amplitudes are comparable to the difference between the Younger Dryas and the Holocene Thermal Maximum, and it is striking that the Mauritia expansion occurred exactly coeval to this climatic event (Fig. 8). These results are in agreement with Urrego (2018), which has suggested a positive relationship between ENSO driven shortduration droughts and Mauritia. Similar palm swamp flourishment as the one occurred in the present study has been also previously reported and interpreted as evidence of relatively high water table (Behling and Lima da Costa, 2001). Thus, a potential influence of the soil type present (watersaturated or poorly drained) should be taken into account. Moving onwards on the temporal frame, the Cariaco record is characterised by wetter conditions during the Medieval Warm Period (MWP: 1.05-0.7 cal kyr BP), and this is also reflected in Caño Tigre record, although more subtly, by the synchronous increase in aquatic elements (Figs. 5, 7 and 8). However, the trends observed in the PCA axis scores do not agree with the climatic conditions inferred by the Cariaco record (Fig.   8). It is suggested that once the Mauritia swamp expanded around 3800 cal yr BP, the clays trend is responding to the fine sediment captured by the Mauritia roots, obscuring any regional climate signal. The importance of the soils ontogeny (and therefore, of the local processes) in the present record is also shown in the RDAs made with the XRD and rock-magnetism data, with the biological samples (fossil pollen) ordered by age (Fig. 7). Our data are also in agreement with

Human occupancy in the NW Orinoco Delta
The presence of large indigenous populations in the Orinoco region was noticed by the earliest European explorers in the 16th and 17th centuries (Huber, 1995a). Even nowadays, Orinoquia is the homeland of 26 different Amerindian groups (Gassón, 2002). Palaeoecological studies can provide important evidence of past human presence or activities, and this information is especially valuable for those regions where archaeological research is limited (Montoya, 2018), as it is the Orinoco Delta. Amongst the most used proxies to infer past human impact in palaeoecological research is the fire record, even in the absence of impacts on the vegetation (Behling and Lima da Costa, 2000).
Nevertheless, human presence and impact in the landscape has also been evidenced in the absence of fire use in coastal savannah environments (Iriarte et al., 2012). In this sense, multi-proxy studies (including for instance phytoliths, macro-fossils, or biomarkers) providing independent lines of anthropic evidence are encouraged to avoid too simplistic or even wrong inferences, as fire can have a climatic origin too (Siegel et al., 2015). An exception to this climatic origin of fires would be the wettest regions of northwestern Amazonia, where fire ignition is often undoubtedly humanmade (Montoya et al., in press).
Last 700 yr were characterised in the Caño Tigre record by a high fire incidence. As already mentioned, this increase coincided with a climatic shift towards drier conditions so a climate role should not be disregarded. Given the high charcoal values attained in the record and the current precipitation regime in the zone (>1500 mm yr -1 without a marked dry season), we propose that the drier conditions could have favoured the fuel biomass availability. However, as showed by the continuous and increasing trend, we suggest that the ignition source of these fires was probably human occupation, with the burning of the plant material being favoured by the drier climatic conditions. Our suggestion of coupled human and climatic forces as responsible for the fire increase is also in agreement with the research performed so far in nearby ancient sites. In this sense, archaeological literature shows an important movement for eastern South American peoples starting at 500 Common Era (CE), with the Arauquinoid peoples migrating from Middle Orinoco towards the delta, the Antilles and the Guianas coast around 600e700 CE (Navarrete, 2008). Once the Arauquinoid started their migration down through the river, they mixed with Cedeñoid groups in the Delta, and mixed Cedeñoid-Arauquinoid-Valloid groups sailed toward Trinidad and the Antilles (Rostain, 2008). Closer to the present study area, the Mora site in central Delta was characterised by integrated elements of the Barrancoid and Arauquinoid traditions dated around 1000-1500 CE (Voorhies et al., 1981).
On the contrary, the oldest interval registered in the Caño Tigre record was related to more stable climatic conditions, and the presence of a mixed forest with coastal-like vegetation and some disturbance elements (i.e., Chenopodiaceae/Amaranthaceae, Cecropia, Monolete psilate). The most likely driver of such vegetation opening or clearing recorded on Zone OR18-1 is fire, as this period is characterised by the highest values of macrocharcoal of the entire sequence (Fig. 5). Despite the lack of additional proxies, it is suggested that this fire record was highly likely human-made, as it coincided with the wettest period of the interval studied (Fig. 8). The southeastern Caribbean (including southern islands and the mainland coast of Venezuela) has provided some evidence of ancient peoples (Siegel et al., 2015). The Archaic period has been described as a period of sparse but continuous inhabitation of the coastal areas forming an Archaic arc that included the southern Caribbean islands and the Venezuelan mainland coast (including the Gulf of Paria) until at least the Guianas and Brazil at the South (Antczak et al., 2019;Sanoja and Vargas, 1999). It has been suggested that the Archaic period in the coastal areas could have lasted longer than in the more inland locations due to geographic isolation (with stepped slopes and hills in the Venezuelan coast and islands). However, they might have had sporadic or even regular contacts with (inland) potterymakers depending on the proximity. During the Mesoindian period (7000-5000 yr BP, equivalent to Archaic) there was a transition from mammals' hunters to socio-economic tribal formation, defining diverse economic and sociocultural strategies. Within the subsistence strategies described for the eastern Venezuelan coast during the Archaic period, the following could have been adopted by the potential inhabitants of Caño Tigre: (i) gathering of endemic species in the Sucre coast mangroves (7000-4000 yr BP), (ii) fishermen and hunters related to eastern Venezuelan coast gatherers, or (iii) sea shore and deep water fishermen and sea shell gatherers, although the first remains for this last type appeared later (Navarrete, 2008). Towards the south, palaeoindian and archaic shell mounds were commonly found in western Guiana littorals, especially between the Orinoco Delta and Essequibo river, with relatively sparse inhabitation during the initial period and main subsistence based on shellfish, fishing and hunting, but also using a wide range of plants including palms (Plew, 2009;Rostain, 2008Rostain, , 2009). In addition, previous palaeoecological records in N Brazil have argued possible Archaic human occupancy based on the charcoal record in similar environments to the present study site (Behling and Lima da Costa, 2001). It is suggested that in Caño Tigre record, coastal peoples were likely occupying the location, and once the seaward progradation of the sediment started, these ancient people abandoned the area. However, more multi-and interdisciplinary research is needed in order to validate or reject this working hypothesis. In case the coastal-peoples occupation is proved, the land abandonment following the combination of drivers including sea-level rise and drier conditions would have profound consequences in terms of the projected climatic scenarios and the current inhabitants of the delta (IPCC, 2013;Vegas-Vilarrúbia and Rull, 2016). On the other hand, based on the evidence obtained so far, a causal link in the synchronicity between the disappearance of fires (land abandonment) and the appearance and establishment of Mauritia in the surroundings remains elusive (Rull and Montoya, 2014).
Given the abundance of archaeological remains near the study site altogether, it is proposed that Caño Tigre location was probably populated at least in the mid-Holocene and during the last millennium by different cultures with diverse subsistence strategies. The above suggestions of human presence and their relationship between their dynamics and the environment are not examples of environmental determinism, as formerly proposed for some lowland South American cultures (Meggers, 1954;Lathrap, 1970). Such interpretation should be taken as a working hypothesis of some of the potential strategies used by past peoples when facing environmental events and the interplay with the natural dynamics of the ecosystems in where they lived.
Nevertheless, given the absence of shell mounts or other archaeological evidence, this hypothesis should be considered with caution.

Conclusions
The palaeoecological study of the mid-Holocene dynamics of Caño Tigre vegetation has shown the forested nature of the landscape for the last 6200 years. However, this forest dramatically changed its composition twice during the interval studied, showing a rapid response of the individual species, but more gradual shifts as plant community. Swamp or rainforest with coastal-like elements characterised the location until approximately 5400 cal yr BP, when drier conditions were inferred from Cariaco and the sea-level was already stabilised, which promoted sediment accumulation and the transition to a fluvial floodplain environment. The shoreline progradation favoured seaward migration of the coastal-like elements of the forest, and the nature of the sediment formed likely allowed the arrival and establishment of Mauritia stands and the occurrence of a mixed-swamp forest in this floodplain. Between 3800 and 2900 cal yr BP, during a period of climatic instability, Mauritia undertook an expansion that culminated with its absolute dominance upon the vegetation, establishing a less diverse palm-swamp in the study site. This community was completely established around 3000 yr ago, when the present-day main tributaries of the northwestern sector of the delta (i.e., Mánamo) appeared. Although similar to the present-day community, additional plant elements arrived to the study site around 700 years ago that finally shaped the current landscape.
This last period was also marked by an increase in the fire regime. Archaeological evidence and the climatic inferences made from previous works point to a likely human contribution to the origin of this fire trend, possibly the arrival of the Arauquinoid culture to the study site. An ancient human occupancy in the area has been also suggested for the Mesoindian or Archaic period, between 6200 and 5400 cal yr BP, when the coastline was about 30 km closer than nowadays, which would probably have had some coastal affinities and subsistence lifestyle. Nevertheless, more research with multi-proxy and interdisciplinary approach is mandatory in order to validate these working hypotheses. Given the importance of local conditions including delta ontogeny in both present and previous works on wetlands, careful must be taken before extrapolating results of a given wetland/delta to a supra-regional scale in terms of palaeoclimate and biological responses.
Concerning climate during the 6200 yr interval studied, it is remarkable the lack of significant delays of Caño Tigre's plant communities showing fast ecological responses to the climatic events or shifts registered in Cariaco. This research shows the high dynamism and sensitivity of these tropical plant communities in facing climatic shifts, so special attention to the potential responses they may develop in the near future as consequences of the ongoing climate change scenario should be paid. In this sense, the inferred socio-ecological shifts of increased ENSO and sea-level changes have been noteworthy. Finally, the present study shows the importance of undertaking multi-proxy approaches and frame them in an interdisciplinary context for acquiring a more precise picture of past landscape dynamics. Fig. 1. Study area. A: North of South America showing the locations mentioned in the text (1: Cariaco Basin, 2: Gulf of Paria, 3: Trinidad, 4: Guianas coastal region, and 5: Amazon mouth). B: Orinoco Delta region, with main towns shown in yellow and main river channels or caños named in blue and italics. EVD: El Volcán Dam, 6: Mora archaeological site (Voorhies et al., 1981). C: Specific location of the study site, showing the closest village (in white) and the name of the main caños (blue and italics). Black dots and numbers in yellow refer to the location of the samples collected, being 15: modern sample CT0, 16: Modern sample CT3, 17: modern sample CT1, and 18: modern sample CT2 and sedimentary archive PATAM18A-12. Images modified from Google Earth. Fig. 2. Sediment description, radiocarbon ( 14 C) dates, and age-depth model of the sedimentary record PATAM18_A12. Colour legend used in the sediment description refers to subtle differences in the facies (composed by clays and peat) found, being silty clays with high content of organic matter (black), grey clay without peat (grey), peaty clay (brown), and light coloured clay with some peat (white). Fig. 3. Geochemical analyses. A: Identified mineral species and their temporal evolution. Data expressed in percentage with respect to dry weight (% wt). From B to F: Downcore variation of concentration of low-and high-coercivity magnetic minerals together with the S and the ARM/SIRM ratios that are proxies of the relative concentration of low versus high-coercivity magnetic minerals and of the magnetic grain size, respectively. G: Total carbon and nitrogen as well as their respective stable isotope composition of the bulk organic matter present in the studied record. Data expressed in per thousand (‰). Two values of δ 13 C were too high and saturated the curve (shown in grey bars). Fig. 4. Diagram of the palynomorphs registered in the modern samples analysed, expressed in percentages, and including macro-and microcharcoal particles expressed in concentration (particles cc -1 ) values. All the taxa percentages are shown with the same scale (equal width means equal % values). Only taxa with percentages higher than 5% have been represented.

Fig. 5.
Pollen diagram of the sedimentary record PATAM18_A12 expressed in percentages, including the stratigraphic column and charcoal particles expressed in concentration (particles cc -1 ) values: referring macro-and micro-charcoal to the different methodologies explained in section 2.2. Only taxa occurring at percentages higher than 10% are represented. All the taxa percentages are shown with the same scale (equal width means equal % values) and outline values represent x10 exaggeration. Colours refer to features of the taxa (Green: trees; Orange: palms; Purple: treelets, lianas or vines; Yellow: herbs; Brown: pteridophytes; Blue: algal and aquatic plants; Black: undefined). Fig. 6. Modern analogues analysis. A: AMCA, B: DCA, and C: Timetracker (in PCA). Sample colours in B refers to the main palynological zones and associated forest types: Blue (Zone OR18-1; swamp or rainforest) with coastal-like elements; Green (Zone OR18-2; mixed swamp forest); and Orange (Zone OR18-3; Mauritia palm-swamp). Fig. 7. RDA plot of the geochemical data (A: XRD; B: rock-magnetism) as environmental variables with biological data (fossil pollen samples), and C: Meyers (1994) plot for stable isotopes. Samples' colours follow Fig. 6.   Fig. 8. Discussion-summary figure that includes: (A) the pollen zones obtained for the sedimentary archive PATAM18_A12; (B) the Ti curve obtained in the Cariaco basin record, previously published in Haug et al. (2001) (data downloaded from NOAA database). LIA: Little Ice Age; (C) the diversity values calculated following Hill (1973); (D) the scores obtained for the PCA first axis of the XRD data, which represents the relationship between kaolinite and chlorite (K + C; positive) and illite (negative) values; (E) the macrocharcoal curve expressed in influx values (particles cm -2 yr -1 ); and (F) the pollen sum of terrestrial pollen taxa (in %), with colours legend following Fig. 6.