The Vanuatu Archipelago is located between the Australian and Pacific tectonic plates, at the eastern margin of the Vanuatu Arc (Fig. 1). It comprises both subaerial and submarine volcanoes (Ash et al., 1978), some of which are still active. These islands consist of lava formed by basalt volcanoes dating from Late Miocene to Holocene. Efate Island, located in the central part of Vanuatu, consists mainly of volcanic rocks levelled by erosion, and limestone terraces issued from tectonic uplifts. Emaotfer Swamp, located on the southern coast of Efate, lies on a Pleistocene limestone terrace (Ash et al., 1967–1970). It is close to the Teouma Graben, on the left side of Teouma River. The water depth is currently less than 1 m throughout the swamp. During the wettest season (December through April), the water level rises and decreases during the drier season (July through September).

The oceanic context and the oceanic-atmosphere coupling (West Pacific Warm Pool, WPWP and South Pacific Convergence Zone, SPCZ) mainly influence the subtropical climate of the archipelago (Vincent, 1994). The location and the magnitude variability of WPWP and SPCZ control the alternation of wet (summer) and relatively dry (winter) season, the wet season being often marked by strong cyclones. Annual rainfall on Efate Island varies, on average, between 2400 mm on the western coast and 3000 mm on the eastern coast (Cillaurren et al., 2001). The El Niño Southern Oscillation (ENSO) (Wyrtki, 1975), the primary cause of long-term climate variability in the western Pacific (Kilbourne et al., 2004 and Moy et al., 2002), influences rainfall and sea surface temperatures (SSTs). The wind-driven ocean currents move warm water in the ocean, eastward during the warm phase (El Niño) and westward during the cool phase (La Niña). The strengthening of El Niño-like conditions causes the northward shift of the SPCZ, consequently Vanuatu becomes relatively drier; conversely, under La Niña-like conditions, the SPCZ is shifted southward and precipitation increases on Vanuatu. Palaeo-ENSO records throughout the tropical Pacific region identify the onset of modern ENSO periodicities after 5000 yr BP, with abrupt increases in ENSO magnitude around 3700 and 3300 yr BP (Brijker et al., 2007, Donders et al., 2007, Donders et al., 2008, Gagan et al., 2004 and Griffiths et al., 2010).

During Late Quaternary, sea-level changes have occurred in relation to tectonic uplifts and eustatic variations. In Vanuatu, the sea-level has risen by 120 m since the Last Glacial Maximum to 6 ka due to eustatic variations, with a sudden acceleration after 11.3 ka (Cabioch et al., 2003). Important forearc tectonic effects vary with geographical position (Lecolle et al., 1990 and Pineda and Galipaud, 1998): in north Vanuatu, high uplift rates have been recorded (3.2 mm/yr on Malo), while they are weaker in central Vanuatu. Estimations of the last interglacial uplift rate of 0.2–0.6 mm/yr, and 0.8–1 mm/yr are reported on the northeast and southwest coast of Efate, respectively (Lecolle et al., 1990 and Pineda and Galipaud, 1998).

This archipelago is quite young, and its flora is principally derived from Southeast Asia by winds, sea and/or animals (Schmid, 1987).

Field trips conducted in September 2005 and October 2013 enabled us to characterize the present-day vegetation around the Emaotfer Swamp. Its shores consist of wooded areas, rich in creepers, and dominated by Barringtonia edulis, Pandanus tectorius and Hibiscus tiliaceus. Cyperaceae, Poaceae, Nymphaeaceae and ferns cover the flooded zones of the swamp. The surrounding plateau is an anthropogenic savannah, composed principally of Urticaceae, Moraceae, Burseraceae and Flacourtiaceae, as a result of cattle grazing.

As on other south Pacific archipelagos, Vanuatu abounds in archaeological sites (Bedford, 2009, Galipaud, 2004, Garanger, 1972, Shutler et al., 2002 and Valentin et al., 2011). Bearers of the Lapita culture began to settle Efate Island around 3000 cal yr BP, and one archaeological Lapita site have been uncovered on Efate, on the western side of Emaotfer Swamp (Bedford et al., 2006): nearly 70 burials features and remains of just over 100 individuals, some accompanied with pots, as well as a contemporary midden constitute the Teouma cemetery. Burial use of the site continued for up to 200–300 years, beginning ca. 3100–2900 cal yr BP or even slightly later ca. 2880–2800 cal yr BP (Petchey et al., 2014). The Teouma cemetery is an outstanding Lapita archaeological site due to the significant number of burials, which represents an early phase of Lapita migration into Remote Oceania (Bedford et al., 2009). The settlement expanded across the cemetery area during the late Lapita-Erueti transitional period (2700–2300 cal yr BP). But there are no traces of human occupation after 2300 cal yr BP, until the development of a coconut plantation, about one century ago.

Languages, material cultures and social practices remained similar during Lapita period, whereas the post-Lapita period was characterized by varied cultures, depending on time and geographic positions (Bedford, 2009). Subsistence behaviour also changed in the Southwest Pacific Islands: Lapita people consumed a large range of food items, taken from the reef, inshore and terrestrial environment, while post-Lapita people favoured lower trophic level terrestrial resources, suggesting the intensification of horticulture (Field et al., 2009, Kinaston et al., 2013, Kinaston et al., 2014 and Valentin et al., 2014). However, to stimulate tuber growth of introduced plants (taro and yam), the settlers consistently cut their flowers: hence, the paucity of these introduced taxa pollen may skew the palynological results.

Four cores were retrieved from the southwest shore of Emaotfer Swamp (Wirrmann and Sémah, 2006). The longest one, Tfer06, sliced into continuous 1-cm width sections was sampled along its longitudinal axis. Three lithological sequences were identified from the base to the top of the core: •

unit I (from 480 to 431 cm) is composed of a homogenous clay-rich organic sediment, and has the slowest sedimentation rate of 0.14 mm/year;

Hereafter, we present a detailed palynological study of 16 samples, 8 from each section 433–426 cm and 151–108 cm.

Each sample of 1 g was prepared by cleaning with KOH (this cleaning was repeated twice for rich organic samples), and by elimination of the mineral phase, with a standard method using hot HF and HCl (adapted from Sittler, 1955). The residue was then mixed with a known volume of glycerol, and 50 μl of this mixture was used to prepare pollen slides. On average 150–200 pollen grains and 30 taxa were identified on each slide, except for the barren samples (< 50 pollen grains counted). A total of more than 100 taxa were identified. The diverse pollen flora was determined by comparison with the collection of over 2000 slides held at the IRD (France), also with photographs and descriptions in Bulalacao (1997), Erdtman (1966), Ledru and Sémah (1992) and with regional reference collections currently held at the Department of Archaeology and Natural History of the Australian National University (http://www.apsa.anu.edu.au). Charcoals, (black, opaque and angular particles ≥ 10 μm), as fire indicators (Whitlock and Larsen, 2001), were also counted. For each sample, the microcharcoal surface was estimated according to the Clark method (Clark, 1982). However, this method only indicates changes in small microcharcoal particles abundance (< 160 μm), and does not fully represent fire patterns, due to the lack of coarser particles.

The chronology is based on AMS radiocarbon ages obtained on 18 samples: bulk disseminated organic matter, vegetal remains, wood fragments and gastropod shells (Wirrmann et al., 2011a). The samples were prepared in the LMC14 Laboratory (UMS 2572, CEA-CNRS-IRD-IRSN-MCC) at Saclay (France), under the laboratory's routine quality control procedures (Cottereau et al., 2007). For charcoal and wood, the classical acid-alkali-acid pretreatment was applied to remove any CaCO₃, humic acid contaminants and to ensure the removal of the modern atmospheric CO₂.

Radiocarbon ages, including those taken from the literature, were calibrated using Oxcal 4.2.2 with the Southern Hemisphere data set (Bronk Ramsey and Lee, 2013; McCormac et al., 2004 and Stuiver and Pearson, 1993) and the two-sigma probability ranges, noted cal yr BP (Table 1). The ¹⁴C division between the Northern and Southern Hemisphere is represented by the ITCZ (Inter-Tropical Convergence Zone). Even if the SPCZ, with merges with the ITCZ, moves over Vanuatu half the year, we chose to use the Southern Hemisphere calibration curve, in order to provide comparable results between the whole core Tfer06 and palaeoenvironmental data from the Southwest Pacific area, especially New-Caledonia, located at 22°S. The curve of the age-depth model was deduced by fitting a smoothed curve to the age by applying a Stineman function to the data (Fig. 2). The curve of the age-depth model generated on the same dates by Bayesian statistics (P_sequence model, Oxcal 4.2.2) matches the curve obtained from the smooth: that is why we kept the smooth polynomial model to present the interpretation of the palaeoenvironmental data obtained on core Tfer06.

The seven dates asterisked in Table 1 are considered as presenting sediment reworking (signs of transportation and/or allochthonous material, mostly roots), and thus were not taken into account in the age-depth model.

This sedimentary section is characterized by the occurrence of two consecutive pollen barren levels, which defines two subzones (Fig. 3, Fig. 4), from 3790 to 3760 and between 3680–3600 cal yr BP. The lower subzone, characterized by the end of the unit I (clay-rich organic sediment), presents the highest value for rainforest taxa (28–30%), dominated by Araliaceae, Geissois, Ardisia and Nauclea. The rainforest also reaches its maximum diversity. Rhizophora and Sonneratia dominate mangrove/littoral vegetation, also well represented (26%). Cyperaceae represent the only herbaceous taxa. This zone shows moderate levels of ferns, and a low charcoal value. In the upper subzone, the unit II replaces the unit I. The rainforest decrease (10-15%) is coeval with a markedly reduced diversity in mangrove taxa (15%). These previous vegetation types are replaced by a vegetation marked by disturbance (32 to 50%), dominated by Mallotus/Macaranga and a slight increase in herbaceous taxa is showed by the onset of Poaceae (Fig. 3).

Between 1500–1450 cal yr BP (Fig. 3, Fig. 4), corresponding to the onset of peat deposit, the vegetation remains relatively stable. Seasonal forest taxa reach maximum values (20 to 35%), with dominant Mimosoideae and Fabaceae. However, the following slight changes occur: •

Chenopodiaceae sometimes reach more than 50% of the total pollen sum;

With the development of the peat unit, a microcharcoal peak, coeval with a palynological richness peak, is noticed. Introduced taxa, dominated by Musaceae, appear, and rise toward the end of the zone.

Due to this relative stability of the vegetation, we also studied two younger samples (1200 and 990 cal yr BP), from peat section (unit III – Fig. 3, Fig. 4), to assess environmental changes. Around 1200 cal yr BP, an increase in rainforest taxa (15%), especially Geissois, Weinmania (Cunoniaceae) and Peperomia is observed. A decrease in rainforest taxa occurs in the uppermost sample, while markers of disturbance and introduced taxa, in particular, rise. Palynological richness declines and microcharcoal particles are less prevalent.

An obvious change in the pollen record over this interval is observed. There was a rapid drop in the rainforest and mangrove/coastal vegetation, which were replaced by indicators of disturbance. This pollen signal, in conjunction with sedimentological and micro-faunal/floral studies (Wirrmann et al., 2011a) suggests drier conditions at this time. The barren pollen zone, volcanic ash-free, is probably due to the sediment oxidation from exposure of the substratum.

The palynological variations correlate with ENSO variability documented by previous works (Donders et al., 2008, Gagan et al., 2004 and Moy et al., 2002). The replacement of rainforest by seasonal forest in 80 years could be linked to peak in sustained El Niño dated from 3700 yr BP (Brijker et al., 2007). The rainforest supported the first notable El Niño events until 3700 yr BP and then decreased. The Indonesian-Australian summer monsoon (IASM) decline from 4200 yr BP (Denniston et al., 2013 and Denniston et al., 2014) could also favour drier conditions in the area. A similar palaeoenvironmental pattern is observed in many Asian-West Pacific areas (Cabioch et al., 2008, Ellison, 1994, Haberle and Ledru, 2001, Haberle et al., 2001, Sémah and Sémah, 2012, Shulmeister and Lees, 1995 and Wirrmann et al., 2011b), although in some tropical Pacific islands, changes in the pollen record, coeval with increase in charcoal values, are observed only after 3000 yr BP, and are interpreted as signs of the onset of human impact (Hope et al., 1999, Hope et al., 2009 and Stevenson, 2004).

However, the mangrove forest decrease illustrates a drop in sea level rather than a climatic change. The relative sea level-change across the Pacific can be summarized as a post glacial eustatic rise until 6000–4000 yr BP (Cabioch et al., 2003), followed by a Late Holocene hydro-isostatic drawdown (Dickinson, 2001). In Vanuatu co-seismic uplift, due to the subduction of the D’Entrecasteaux Ridge, has also to be taken into account (Lecolle et al., 1990, Neef and Veeh, 1977). Around 3700 yr BP, the end of the eustatic rise was coeval with tectonic uplift rate close to 1 mm/yr in the South of Efate which in turn induced a sea-level decrease, marked by the loss of mangrove forest. The occurrence of former rolled-coral in several archaeological sites (Bedford et al., 2007, Dickinson, 2001 and Pineda and Galipaud, 1998) confirms that the sea level was higher than today when the first settlers arrived.

The emergence of the Lapita culture on Mussau (Papua New Guinea) is dated around 3400 cal yr BP (Denham et al., 2012). The dispersal of Lapita people occurred after the onset of regional drier conditions. Moreover, during El Niño phase, the easterly winds decline, facilitating the sail-powered transport from New Guinea to the eastern islands (Anderson et al., 2006). If the precise causes of this eastward migration remain unclear, yet El Niño events have to be taken into account in the settlement of Remote Oceania.

Except increases in Cyperaceae, Chenopodiaceae, fern spores and microcharcoal particles, there are little significant variations in the pollen record between 1500 and 1450 cal yr BP. The vegetation remained broadly stable, while Wirrmann et al. (2011a) show lithological, micro-faunal and -flora changes during this period: peat replaced poor-organic sediments and acidophilus diatom species replaced species of high conductivity water. The occurrence of these deposits, associated with an increase of fern spores and Cyperaceae, may correspond to a hydroseral succession. Adapted plants invade open water, reducing water flow, trapping sediment and contributing to the invasion by emergent vegetation. As peat accumulates, made up of the remains of this vegetation, the water body becomes progressively shallower. The high percentage of Chenopodiaceae pollen grains, often found in clumps, indicates close proximity of this vegetation type, and shows a decrease in water level, in agreement with the peat development. An increase in vegetation cover could prevent important runoff, resulting in a decline of sedimentation rate. However, the decrease in diatom species characteristic of saline conditions is inconsistent with shallower water (Van Dam et al., 1994). It could be explained by an increase in humic compounds, due to higher-level vegetation decomposition.

The peak in microcharcoal particles shows an increase in fire intensity and/or quantity. This was possibly because of further sustained El Niño events, from 2000 to 1400 yr BP, peaking at 1500 yr BP, associated with a period of IASM rainfall minimum (Denniston et al., 2013, Denniston et al., 2014 and Gagan et al., 2004). But Cyperaceae, Chenopodiaceae and fern variations more certainly mark local environmental change, likely variations in water level, than a climate event. Hence, ENSO and IASM rainfall variations seem to have a low impact on the vegetation. We propose that human populations took advantage of these local drier conditions, or even favoured them by setting fires too, to cultivate different Musaceae (bananas). The occurrence of Phyla is an additional evidence of human influence on vegetation: this plant has presumably been cultivated for ornamental and medicinal use, and is now considered as a weed (Smith, 1979).

A significant change in the pollen record is observed around 1200 cal yr BP. There is a sharp decline of seasonal forest taxa, coeval with an increase in rainforest taxa, suggesting relatively wetter conditions, as observed at the same time in New-Caledonia (Wirrmann et al., 2011b). This could be linked to the decline in El Niño events after 1500 yr BP (Denniston et al., 2013, Denniston et al., 2014, Gagan et al., 2004 and Moy et al., 2002). But rainforest taxa values increased only 250 years after the onset of decline in El Niño events. One could explain this fact by a rapid growth and reproduction of light-tolerant species (disturbed vegetation) compared to rainforest species (Chave, 1999, Prévost, 1983). Moreover, the increase in taxa such as Geissois and Weinmania, characteristic of higher altitudes, suggests lower regional temperatures compared to 3790 cal yr BP. At 990 cal yr BP, the significant rise in introduced taxa, coeval with a decrease in rainforest taxa, is interpreted as human impact; which suggests a more permanent settlement in this area, perhaps longer than during the Lapita period.