The nation of Vanuatu in the Southwest Pacific stretches 1000 km from north to south and comprises over 83 islands, which formed as a result of uplift and volcanism associated with the New Hebrides Arc, where eastward subduction of the Indo-Australia Plate beneath the Pacific Plate occurs (Stewart et al., 2010). Around 90% of the islands of Vanuatu are less than 1.8 Ma years old (Mallick, 1975). Here we focus on Efate, Central Province, the location of Port Vila, the capital city of Vanuatu. Efate has a volcanic core (Ash et al., 1978), but the surface geology is dominated by uplifted coral sequences and the Efate Pumice Formation, which form the main topography of the centre of the island (Raos and Mcphie, 2003). A series of now uplifted coral terraces formed within the last 300,000 years along the island coast (Taylor et al., 1987).

Sediment samples were recovered from Lake Emaotul (sometimes called Duck Lake; 17°43′57.48″S, 168°24′53.58″E) on Efate. Lake Emaotul was selected because it is a confined lake system likely to be sensitive to changes in the precipitation-evaporation balance. It is relatively insulated from marine processes compared with other sites on the island due to its elevation (119 m above sea level). The lake sits within a raised coral reef dated at c. 120,000 BP (Sear et al., 2020). It has a maximum depth of 7.1 m and is meromictic. A chemocline at 6 m results in a suboxic hypolimnion with dissolved oxygen at 5% saturation (Sear et al., 2020).

Hydrologically, the lake is a closed system. It measures 0.3 km² in area and has a catchment covering 0.98 km². The lake is fringed by grass-dominated wetlands to the east and west, with the swamp-forest taxa Barringtonia, Hibiscus, and Pandanus. The lake surface is mostly open but has some cover of water lilies, such as Nelumbo nucifera, Nymphaea spp., and Nymphoides spp. The wetland areas are bordered by Phragmites. Beyond the wetland, the lake catchment is covered by secondary tropical forest that supports patches of subsistence gardening and cattle grazing (Sear et al., 2020). Secondary lowland forest taxa include Metroxylon warburgii (Kalfatak and Jaensch, 2014).

The Lake Emaotul cores were retrieved between the 9th and 11th October 2015. The 3.4-m sedimentary sequence of overlapping cores was retrieved using a UWITEC gravity corer and a piston corer (Geo-Core, Columbus, OH, USA) from the deepest part of the lake. Deeper coring of the sediments was impeded by stiff clay. This precluded recovery of sediments that predated human arrival. All six recovered cores were stored in airtight tubes during transportation and subsequently at + 4°C at the School of Geography and Environmental Sciences (SOGES) at the University of Southampton (UK).

Three subsamples of the core that displayed discrete and high magnetic susceptibility were extracted from 173 to 171, 169 to 167, and 166 to 160 cm. To identify the depth range over which tephra was deposited small amounts of sediment were smeared on glass slides and examined under an optical microscope. All samples contained shards; however, particularly high counts of fresh volcanic glass particles occurred between 166 and 160 cm. To obtain the geochemical signature of volcanic glass samples from all three horizons were cleaned in deionised water, and the coarsest grains (∼63–125 μm) were hand-picked, mounted in an epoxy round, cut, and polished. The polished samples were sputter-coated in carbon. An initial suite of three samples was analysed using a Jeol JXA8230 Superprobe at the University of Iowa, with repeat analyses and an additional eight samples using a JXA8530F Hyperprobe at the University of Auckland. For both instruments, points were analysed using an accelerating voltage of 15 kV and a 2-nA electron beam current. A defocussed 10-μm beam was used in order to minimise Na loss, with a 5-μm diameter required for a few smaller grains. Analysis was completed with 10-s peak and 10-s total background dwell times for all elements. To test for Na migration, some analyses were collected in two 1-s accumulated peak measurements with 1-s dwell time on background for the second measurement. The accumulated Na measurements showed that post-measurement correction of Na migration was not required. Analyses were calibrated and quality cross-checked against secondary glass standards ATHO-G (MPI-DING), BCR-2G, (USGS), and VG-568 (NMNH 72854).

The age-depth model is based on the radiocarbon ages of seven plant macrofossils and 34 ²¹⁰Pb analyses; see Maloney et al. (2022). All ¹⁴C dates were re-calibrated using the Southern Hemisphere calibration curve (Hogg et al., 2020) to years before present (cal. years BP). To build the age-depth model we used Bacon v2.2 (Blaauw et al., 2021).

We analysed 188 sediment samples (every 2 cm) for carbon-to-nitrogen (C/N) ratio, which is used as an indicator of the source of organic material in a lake (Rice and Hanson, 1984). See Supplementary Information for full C/N method. The C/N ratio can vary, since algae are more enriched in nitrogen and depleted in carbon compared with terrestrial vascular plants (Tyson, 1995). A C/N ratio of >20 usually indicates a system in which terrestrial organic material is dominant (Jones et al., 2013), whereas a ratio of 10–20 likely signifies a mixture of terrestrial and aquatic organic inputs. A ratio of <10 is more likely to suggest more aquatic plants were present (García-Alix et al., 2012). C/N samples were measured at the British Geological Survey, in Nottingham, UK.

To determine plant community changes, we analysed 32 samples of 1-cm³ volume for pollen and spores. The sampling resolution is about 1 sample per ∼70 years (14 cm) for the majority of the sequence and one sample every ∼5 years (1 cm) above and below the Kuwae tephra layer (Supplementary Information). We followed a standard procedure to extract and count pollen and spores (Erdtman and Wodehouse, 1944).

Pollen grains were counted and identified at × 400 magnification with detailed identification and photography of grains at × 1000 magnification (Supplementary Information). Unidentified types were described, photographed, and assigned a code; obscured or damaged grains were grouped as unknown. Poaceae pollen grains were grouped as either <40 μm, >40 μm, >60 μm or as multiporate grains. We used the Australian National University (2007) Australasian Pollen and Spore Atlas to identify the pollen and spores.

Micro-particles of charcoal were quantified from 32 samples by counting in the size range 10–125 μm alongside exotic Lycopodium spores and pollen on pollen slides (Wang et al., 1999). Particles were identified in the pollen slides as opaque, black, and angular fragments. Macro-charcoal samples were prepared from the >125-μm sieved fraction created in pollen preparation and were bleached with 6% H₂O₂ until the reaction stopped. All macro-charcoal fragments in each 1 cm³ sample were counted using a stereo microscope (Whitlock and Larsen, 2001).

The head capsules of chironomid larvae were extracted by disaggregating 2 cm³ samples of sediment with 10% KOH (following Brooks et al., 2007). The disaggregated samples were then sieved to <90 μm. The head capsules were then picked and mounted with Hydro-Matrix onto microscope slides. The counting and identification was done at × 400 magnification and taxonomically identified using Brooks et al. (2007) and Andersen et al. (2013), and more specifically with a focus on the Southern hemisphere, following Dieffenbacher-Krall et al. (2007) and the New Zealand subfossil chironomid taxonomy which can be found at https://climatechange.umaine.edu/perl/nzguide.html. Ten samples were analysed, with an average temporal resolution of c. 200 years.

To identify the occurrence and timing of transitions, we created the pollen and spores and chironomid diagrams using Tilia 2.9.1 software and numerically zoned them using depth-constrained cluster analysis (CONISS) (Grimm, 1987; Grimm, 1990). All pollen taxa with an abundance of >1%, apart from unknown types, were included in the CONISS calculations. The same analysis was carried out for the chironomid data. In addition, the arboreal (trees and shrubs) to non-arboreal (herbs) pollen ratio was calculated for all depths.

Using pollen and spore percentages and chironomid percentages we performed detrended correspondence analysis (DCA) in R (RStudio Team, 2015; R Core Team, 2017) with the vegan package (Oksanen et al., 2007) with the assumption that the DCA axis 1 scores represent turnover (Bush et al., 2004; Blarquez et al., 2014; Castilla-Beltrán et al., 2021a,Nogué et al., 2021). Unknown pollen types, which were assigned a code (see section “Supplementary Information”), were included in the DCA. DCA axis 1 scores were used since they are considered optimal for analysing percentage data which contains many zeros (Hill and Gauch, 1980).

Canonical correspondence analysis (CCA) was conducted on the pollen and spore percentage data also using the vegan R package. Micro-charcoal, macro-charcoal (regional and local fire, respectively), reconstructed precipitation rate (mm day^–1), Ti/incoherent scatter (erosion), and magnetic susceptibility (tephra and erosional events) were incorporated into the CCA as environmental variables. To fill data gaps in the precipitation rate dataset (to facilitate comparisons with the other proxies) we interpolated between data points. See Maloney et al. (2022) for how precipitation rate values were calculated.

The opportunity to carry out a detailed tephra analysis allowed us to assign the source of the main tephra horizon in the record with some confidence. Samples between 166 and 160 cm contain dominantly clear, vesicular, angular, and un-weathered volcanic glass, with subordinate brown fresh volcanic glass, feldspar, and pyroxene grains (with adhering glass). The coarsest particles of 2–2.5 mm diameter, and highest concentration of volcanic particles were in the 162–160 cm core segment. While we recognise that a firm identification requires further data, the glass texture, associated minerals, and glass major-element composition shows a good match to glass analysed on the same instrument in this study from near-source sites around the Kuwae Caldera (Figure 2 and Supplementary Information). The age-depth model (Supplementary Information) suggests that the 6 cm layer of Kuwae tephra was deposited 650–510 cal. years BP (95% confidence) which is close to estimates of the Kuwae eruption occurring in the 1450s CE (Gao et al., 2006; Plummer et al., 2012).

The sample from 169 to 167 cm contained mainly weathered and highly altered crystalline rock fragments (andesitic to mafic based on mineralogy) as well as rare fragments of clear and highly vesicular glass. The composition of these rare glass fragments is not similar to the Kuwae tephra, but instead matches the Efate Pumice Formation tephra (∼1.7 Ma; Stewart et al., 2010), which is a common contributor to colluvial and alluvial sediments on Efate (Figure 2). These early Pleistocene glass and older weathered lithic grains from basement rocks, cf. Ash et al. (1978), were likely reworked into the lake following earthquakes or during major rain events or cyclones. The 173–171 cm samples contained no fresh glass, but inorganic particles were dominated by weathered and altered crystalline rock fragments (Figure 2). Most compositional analysis was unsuccessful on this material, but two analyses showed an andesitic composition, consistent with basement rock on Efate (Ash et al., 1978).

Over 100 pollen types were identified, around 5% of which were assigned a code since they could not be identified. The pollen diagram is divided into four pollen assemblage zones based on the CONISS dendrogram (Figure 3). The following percentage values are calculated as means for each zone.

Zone one (341–258 cm, >1620–1120 cal. years BP) comprises of nine samples and is dominated by secondary forest taxa Acalypha (30%), Pandanus (4%), Macaranga (20%) Trema (13%), and Homalanthus (1%) with occurrences of Poaceae (7%), Cyperaceae (2%), trilete psilate fern spores (7%), and monolete psilate fern spores (1%).

Zone two (258–155 cm, 1120–560 cal. years BP) consists of 13 samples and has relatively high proportions of Acalypha (21%), Macaranga (19%), Pandanus (6%), Trema (9%), and Homalanthus (3%). Cyperaceae (4%), Poaceae (8%), psilate trilete spores (12%), psilate monolete spores (2%) also exhibit relatively higher proportions.

Zone three (155–24.5 cm, 560 cal. years BP to 1930 CE) consists of seven samples and shows a decline in secondary forest taxa Acalypha (to 16%) Trema (to 6%), Macaranga (to 16%) and an increase in Persicaria-type (to 2%), Poaceae (to 17%), Nymphoides (to 1%), Pandanus (to 9%), and trilete psilate spores (to 13%). Cyperaceae remains at 4% and monolete psilate fern spores declines to 1%.

Zone four (24.5–1 cm, 1930–2015 CE) consists of three pollen samples. In this zone, Acalypha Trema, Macaranga, and Pandanus decline to 12, 0.5, 14, and 4%, respectively. Herbaceous taxa Asteraceae (5% abundance), Persicaria-type (2% abundance), and Poaceae (17% abundance) increase. The aquatic plant Nymphoides increases to 2%, as do monolete (to 2%) and trilete fern spores (to 26%) whereas Cyperaceae remains at 4%.

Ten chironomid taxa were identified. The chironomid diagram (Figure 4) is divided into three zones using CONISS (Grimm, 1987).

Zone one (357–252 cm, >1600–1090 cal. years BP) consists of four samples and is dominated by Procladius (46%) and Tanytarsus funebris- type (41%).

Sub-zone two (252–221.5 cm, 1090–910 cal. years BP) consists of one chironomid sample and has high proportions of Chironomus anthracinus-type (52%) and Orthocladiinae type 1 (similar to Pseudorthocladius) (8%). Procladius and Tanytarsus funebris- type decline to 4 and 24%, respectively.

Zone three (221.5–61 cm, 910–160 cal. years BP) consists of five samples and is similar to zone one where Procladius increases to 36% and Tanytarsus funebris- type increases to 55%. Chironomus anthracinus-type decreases again to 2%.

The core stratigraphy consisted of gyttja throughout. We structured this section following the pollen zones from Figure 3 (red dashed line). Micro-charcoal concentrations, as shown in Figure 5, are variable in pollen zone one and the mean value is 1,300,000 particles per cm³, macro-charcoal concentrations are lower (7 particles per cm³). In pollen zone two, micro-charcoal concentrations are relatively lower (1,190,000 particles per cm³) and macro-charcoal particles are still 7 particles per cm³. In pollen zone three, micro-charcoal particle concentrations decrease again to 790,000 particles per cm³ but there is a relative increase in macro-charcoal particle concentrations to 14 particles per cm³. In pollen zone four, micro-charcoal particle concentrations remain relatively low at 205,000 particles per cm³ but macro-charcoal concentrations increase to 57 particles per cm³. The ratio of arboreal to non-arboreal pollen varies throughout the record (between 0.4 and 1.96). The most notable shifts in the ratio occur after the Kuwae eruption and in the upper part of the record (pollen zone four). Following the Kuwae eruption, AP/NAP increases and then declines after ∼20 years. The lowest AP/NAP is within pollen zone four. The C/N ratio fluctuates thorough the record but does not display any change during the shift to drier conditions (Figure 5). However, the ratio largely varies around between 12 and 16 with one notable decline to 7.8 at 185 cm (∼700 cal. years BP), prior to the Kuwae eruption suggesting a greater proportion of algal material.

We used a turnover measure (DCA axis 1) to assess the degree of change in the composition of both chironomids and pollen due to the shift to drier conditions (Maloney et al., 2022) and the Kuwae eruption (Figure 5). Both chironomid and vegetation turnover increased during the shift toward drier conditions ∼1100 cal. years BP. While chironomid turnover showed no change following the Kuwae eruption, the vegetation values indicate greater taxonomic turnover following the eruption. There was also a sharp increase and then decline in the arboreal to non-arboreal pollen ratio (AP/NAP) following the eruption.

Canonical correspondence analysis (CCA) was used to determine which environmental drivers have dominant associations with each pollen zone. CCA axis 1 explained 39% of the variance (eigenvalue 0.05) and CCA axis 2 explained 22% (eigenvalue 0.04) (Figure 6). Environmental variables on the positive side of CCA axis 1 include regional fires, precipitation rate (mm day ^–1), erosion (Ti/incoherent scatter), and tephra/erosional events indicated by magnetic susceptibility (correlation scores 0.47, 0.61, 0.9, 0.01, respectively, see section “Supplementary Information”). One environmental variable, macro-charcoal particles (local fires), was located on the negative side of CCA axis 1 (correlation score −0.5). Pollen taxa from zone one (in red) are mostly positioned in the positive half of the CCA axis 1 and associated with precipitation and erosion. Pollen zone two (in orange) transitions from the positive to the negative half of CCA axis 1 with tephra/erosion (magnetic susceptibility) as the most influential environmental variable of this period and finally pollen zones three (in green) and four (in blue), which occur after 560 cal. years BP, are found in the negative side of CCA axis 1 and explained by local fires.

Drought in the SW Pacific islands is currently considered a threat to both island ecosystems and their human societies (e.g., Sear et al., 2020). After ∼1100 cal. years BP, there was a shift toward drier climate conditions (Figure 5; Maloney et al., 2022). Precipitation values changed from around 6.8 mm day^–1 to 4.4 mm day^–1 over a period of ∼70 years (Maloney et al., 2022), a significant shift in moisture availability. Both vegetation and chironomid turnover temporarily increased across the transition from wetter to drier climate conditions (Figure 5). The vegetation change was characterised by an increase in Cyperaceae, Poaceae, and ferns, which then declined to previous values within ∼80 years (Figure 3). According to the CCA analysis, precipitation, and erosion plot with samples from pollen zone one (>1620–1120 cal. years BP) prior to the change to drier conditions, linking both environmental drivers of change as the most significant. Indeed, erosion (Ti/inc) is well known to be linked with sedimentary in-wash (e.g., Sear et al., 2020), which here correlates with higher precipitation. While there is a higher turnover at the transition to drier conditions, the individual taxon changes are subtle and somewhat transient (Figure 5).

The changes in the chironomid community show a strong response to the moisture shift, then recovery, but with different abundances of the dominant taxa than prior to the shift. While all main taxa declined during the transition, Chironomus anthracinus type increased; this caused high turnover in the short term (Figures 4, 5). Chironomus anthracinus is known to be able to withstand low levels of oxygen and can tolerate unfavourable conditions (Hamburger et al., 1994, 1995). The increase in Chironomus anthracinus and relative decrease in other taxa may suggest a decrease in oxygen or even anoxia following the climate shift. Turnover decreased within ∼135 years, despite climate conditions remaining relatively dry. The C/N ratio shows no clear response to the climate transition, indicating that there was no shift in the balance of terrestrial and algal input to the lake with the shift to drier conditions. This possibly suggests that the overall biomass in the catchment did not change much, and also that the organic input to the lake was not strongly driven by precipitation-driven in-wash.

The Kuwae eruption is considered one of the largest eruptions of the last millennium (Neukom et al., 2014), and it had far more impact on the forest of Efate than did the moisture change. The CCA picks out the Kuwae eruption as the most significant association with the fossil pollen assemblage in pollen zone two (1120–560 cal. years BP) (Figure 6). It is notable that vegetation turnover not only increased sharply after the Kuwae eruption, but it remained high throughout the subsequent record (Figure 5). It is generally accepted that after a volcanic eruption there may be changes in vegetation composition, such as an increase in grass abundance (see examples in “Supplementary Information”). In Efate, after the Kuwae eruption sequence, an increase in grass pollen (Poaceae) is linked with a decline in AP/NAP, suggesting increased occurrence of open patches (Figure 5).

The tree and shrub post-eruption dynamics show Acalypha (native shrub or small tree) and Trema (native tree) decreasing within ∼20 years of the deposition of the Kuwae tephra, while Persicaria-type, Solanaceae, and Nymphoides increased together with grasses (Figure 3). Initially, damage to small trees and shrubs such as Acalypha and Trema could have created space for grasses and sedges to expand, and the increase in grasses may also be explained by its relatively short generation time compared with woody vegetation, which gives the group the ability to profit relatively quickly from nutrient enrichment following volcanic eruptions (Urrutia et al., 2007). It is important to note, however, that our results show that the forests surrounding Lake Emaotul were already disturbed at the start of this record, as indicated by the presence of Acalypha, Macaranga, and Trema, which are considered secondary, disturbance, and early successional taxa (Whittaker et al., 1989; Fall, 2010; Åkesson et al., 2021) and the continuous presence of fire, which is often associated with human occupation on Pacific islands (e.g., Fall, 2005; Stevenson et al., 2017). Although our results show that the Kuwae eruption impacted on the forest surrounding Lake Emaotul, local human impacts during the period of the Kuwae eruption cannot be ruled out as a driver for change. For example, there is archaeological evidence that human populations moved southwards from Sheppard and Epi islands to Efate following the eruption (Garanger, 1972), and an increase in population size could have led to some level of deforestation. However, there is no clear link between the declining taxa (Acalypha and Trema) and known human uses of them.

Despite numerous examples of change in vegetation composition (Supplementary Information), recovery rates of vegetation following volcanic eruptions are poorly understood. Palaeoecological evidence of tephra deposition events have typically shown short-lived impacts (decades to centuries, e.g., Lotter and Birks, 1993; Montoya et al., 2021). For example, the main patterns that emerge from analysing large datasets of pollen time-series from tropical forests suggest that forest recovery rates are faster for large, infrequent events, such as volcanic eruptions, than for human-induced burning (Cole et al., 2014). The analysis by Cole et al. (2014) also suggests that arboreal taxa typically recover to ∼95% abundance after >200 years. On Efate, the arboreal taxa Acalypha and Trema had largely recovered their pre-eruption abundance in ∼260 years. Prior to the eruption, Acalypha and Trema values were 23 and 9% of the total pollen sum, respectively. After 260 years, they reached 19 and 11%, respectively. These two taxa responded differently to other arboreal taxa such as Macaranga, Pandanus, Sapindaceae, and Sapotaceae, which showed no obvious decline after the eruption. Despite the partial recovery of Acalypha and Trema, turnover remained higher for all taxa ∼625 years following the eruption (Figure 5).

While the forests of Efate showed changes linked to the volcanic eruption, Lake Emaotul chironomid communities remained unperturbed (Figure 4). Addition of volcanic tephra of basaltic/andesitic composition to the lake could be expected to deliver nutrients, including substantial S, Fe, and other cations (Cronin et al., 1998), and this may have led to algal blooms (e.g., Harper et al., 1986; Roberts et al., 1997; Eastwood et al., 2002; Telford et al., 2004), which might be expected to show as a change in C/N. The Emaotul C/N record shows short-term variability, but no clear directional change is evident following the eruption, with ratio values before and after being ∼13 (Figure 5), a value indicative of a mixed source of organic matter (García-Alix et al., 2012); if there was a post-eruption increase in algal biomass, it is not detectable. As suggested for forested sites in Ecuador (see Montoya et al., 2021), the dense canopy of the forest trees surrounding Lake Emaotul may have partially shielded aquatic ecosystems from major impact by blocking some airborne tephra fall on to the lakes surface. The inconsistent responses of algal productivity to volcanic tephra-fallout recorded to date suggest site- or catchment-specific factors may be important in determining responses. At Lake Emaotul, chironomids also showed no significant change following the eruption. Either they recovered so quickly that changes were not identified in this record or they are not sensitive to volcanic tephra inputs at this site.

Archaeological evidence indicates that people had been actively modifying the landscape of Efate ∼1400 years prior to the start of this record (1600 cal. years BP), since the first human settled Vanuatu ∼3000 years ago (Petchey et al., 2014; Bedford et al., 2018). Notably, the CCA analysis (Figure 6) shows an association between samples from the period 560 cal. years BP to 1950 CE and local fires, indicating that the local use of fires, suggested by the increase in macro-charcoal particles (Figure 5) could have been an important driver of change (Clark, 1988; Whitlock and Millspaugh, 1996). During pollen zone four (1930–2015 CE), an increase in Asteraceae, Poaceae, Solanaceae, and Persicaria-type pollen, and fern spores, a decrease in woody taxa (e.g., Acalypha and Trema), together with higher concentrations of macro-charcoal concentrations spanning the last ∼90 years is strongly indicative of opening of the forest and transition toward the wetland areas and secondary forests we see today surrounding Lake Emaotul. These findings agree with other studies from Vanuatu; for example, human impacts are seen in pollen diagrams from ∼3000 cal. years BP onward from Aneityum island, Southern Vanuatu (∼320 km south of Efate). At Aneityum island, tree and shrub taxa (Eugenia, Celtis, and Trema) were replaced by grasses (Poaceae), sedges (Cyperaceae), ferns, accompanied by an increase in charcoal particles (Hope and Spriggs, 1982). Another example is the palaeoecological record from the island of Erromango (Figure 1, ∼150 km south of Efate), which shows evidence of burning throughout a 2000 year period during which humans were present (Hope et al., 1999). Human impacts may have been present prior to the last 1600 years, but despite their apparent impacts, the transition to drier climate ∼1100 cal. years BP and the Kuwae eruption were both strong enough disturbances to affect the vegetation surrounding Lake Emaotul.