The Role of Climate: 71 ka of Atmospheric Mercury Deposition in the Southern Hemisphere Recorded by Rano Aroi Mire, Easter Island (Chile)

: The study of mercury accumulation in peat cores provides an excellent opportunity to improve the knowledge on mercury cycling and depositional processes at remote locations far from pollution sources. We analyzed mercury concentrations in 150 peat samples from two cores from Rano Aroi (Easter Island, 27 ◦ S) and in selected vegetation samples of present-day ﬂora of the island, in order to characterize the mercury cycling for the last ~71 ka BP. The mercury concentrations showed values ranging between 35 and 200 ng g − 1 , except for a large maxima (~1000 ng g − 1 ) which occurred at the end of the Last Glacial Maximum (LGM, ~20 ka cal BP) in both peat cores. Low temperatures during the LGM would accelerate the atmospheric oxidation of Hg(0) to divalent mercury that, coupled with higher rainfall during this period, most likely resulted in a very efﬁcient surface deposition of atmospheric mercury. Two exceptional short-lived Hg peaks occurred during the Holocene at 8.5 (350 ng g − 1 ) and 4.7 (1000 ng g − 1 ) ka cal BP. These values are higher than those recorded in most peat records belonging to the industrial period, highlighting that natural factors played a signiﬁcant role in Hg accumulation—sometimes even more so than anthropogenic sources. Our results suggest that wet deposition, linked to atmospheric oxidation, was the main process controlling the short-lived Hg events, both in the mire and in the catchment soils. accumulation Under the dry peatland stage (i.e., ca. 40.0–42.0 the peat-organic matter decomposition increases Hg content. The effect of vegetation and volcanic activity is not well addressed in this study. Modified from previous research [31].


Introduction
Mercury (Hg) is a metal of environmental concern due to its high volatility, long atmospheric residence time (1-2 years), and the bioaccumulation of its methylated forms. It is released and dispersed in the atmosphere by natural emission sources such as volcanoes, geothermal vents, and Hg-enriched soil, as well as by anthropogenic activities such as mining, coal-fired plants, and chlor-alkali plants [1][2][3]. Its transport and deposition in different environments on Earth depends on the type of Hg. objectives of the present research were (i) to reconstruct the Pleistocene Hg flux variability at millennial to centennial resolution and (ii) to determine the main factors that controlled peat Hg concentrations over long time scales.

Regional Setting
Easter Island (27 • 07 S, 109 • 22 W) is a Chilean volcanic island, Miocene in age, located in the southern Pacific Ocean. The nearest continental point from the South American continent lies 3512 km away. The island has a triangular shape and the topography is characterized by more than 70 volcano craters and the rolling surfaces of lava flows [33][34][35]. The climate is humid subtropical, with average monthly temperatures between 18 • C (August) and 24 • C (February) [36] and highly variable annual rainfall ranging from 500-1800 mm (mean of 1130 mm).
Rano Aroi is a small mire located in an ancient Pleistocene volcano crater close to the most elevated area of the island (27 • S, 108 • W, 430 m elevation, 0.13 km 2 ; Figure 1). The crater slopes form a small catchment area (0.158 km 2 , maximum dimension 0.21 km North to South). The watershed lithology mainly consists of very porphyritic olivinic tholeite, basalt and hawaiite lava flows, which are rich in iron [34]. The mire is covered by Scirpus californicus, Polygonum acuminatum, Asplenium polyodon var. squamulosum, Vittaria elongata and Cyclosorus interruptus [37], while the catchment area is covered by grassland and planted eucalyptus. There is a small ravine acting as a natural outflow that infiltrates before reaching the coast. An artificial outlet built in the 1960s has been used to control the maximum outflow from the peatland since then. source; for example, algae or terrestrial and aquatic plants [32]. Isotopic organic matter data, δ 13 C (‰), was used to indicate the C3 and C4 vegetation dominance and thus the main climatic conditions. The objectives of the present research were (i) to reconstruct the Pleistocene Hg flux variability at millennial to centennial resolution and (ii) to determine the main factors that controlled peat Hg concentrations over long time scales.

Regional Setting
Easter Island (27°07′ S, 109°22′ W) is a Chilean volcanic island, Miocene in age, located in the southern Pacific Ocean. The nearest continental point from the South American continent lies 3512 km away. The island has a triangular shape and the topography is characterized by more than 70 volcano craters and the rolling surfaces of lava flows [33][34][35]. The climate is humid subtropical, with average monthly temperatures between 18 °C (August) and 24 °C (February) [36] and highly variable annual rainfall ranging from 500-1800 mm (mean of 1130 mm).
Rano Aroi is a small mire located in an ancient Pleistocene volcano crater close to the most elevated area of the island (27° S, 108° W, 430 m elevation, 0.13 km 2 ; Figure 1). The crater slopes form a small catchment area (0.158 km 2 , maximum dimension 0.21 km North to South). The watershed lithology mainly consists of very porphyritic olivinic tholeite, basalt and hawaiite lava flows, which are rich in iron [34]. The mire is covered by Scirpus californicus, Polygonum acuminatum, Asplenium polyodon var. squamulosum, Vittaria elongata and Cyclosorus interruptus [37], while the catchment area is covered by grassland and planted eucalyptus. There is a small ravine acting as a natural outflow that infiltrates before reaching the coast. An artificial outlet built in the 1960s has been used to control the maximum outflow from the peatland since then.

Sampling
Two sediment cores (ARO 06 01, 13.9 m long and ARO 08 02, 4 m long) were retrieved from Rano Aroi on March 2006 and October 2008, respectively. The core ARO 06 01 was sampled with a UWITEC corer from the central part of the mire, and the ARO 08 02 with a Russian corer.
The uppermost two meters of the core ARO 06 01 were rejected at the field to avoid potential anthropic remobilization, as described in previous studies [31,32]. Meanwhile, ARO 08 02 was retrieved with a Russian corer from the eastern boundary in order to recover the previously discarded sediments, since the mire's margins were potentially not affected by recent human activities [38,39]. None of the cores retrieved in either campaign reached the bedrock. The cores were sealed, transported to a core repository, and stored in a cold room at +4 °C until sampling. Both cores were lithologically described and sampled for smear slides every 5 cm [31].

Sampling
Two sediment cores (ARO 06 01, 13.9 m long and ARO 08 02, 4 m long) were retrieved from Rano Aroi on March 2006 and October 2008, respectively. The core ARO 06 01 was sampled with a UWITEC corer from the central part of the mire, and the ARO 08 02 with a Russian corer.
The uppermost two meters of the core ARO 06 01 were rejected at the field to avoid potential anthropic remobilization, as described in previous studies [31,32]. Meanwhile, ARO 08 02 was retrieved with a Russian corer from the eastern boundary in order to recover the previously discarded sediments, since the mire's margins were potentially not affected by recent human activities [38,39]. None of the cores retrieved in either campaign reached the bedrock. The cores were sealed, transported to a core repository, and stored in a cold room at +4 • C until sampling. Both cores were lithologically described and sampled for smear slides every 5 cm [31].

Age Model
The chronology for both cores was fully described by Margalef, et al. [31]. A total of 27 AMS 14 C dates were obtained from pollen concentrates of the ARO 06 01 and ARO 02 08 cores in the Poznan Radiocarbon Laboratory (Poland). Only 18 dates were used for the age-depth model. The AMS ages were calibrated using CALIB 6.02 software, and the INTCAL 98 curve [40] and CalPal [41] for samples older than 20,000 radiocarbon year BP. The age model was built by simple linear interpolation [31].

Geochemical Analysis
Both cores were sampled every 5 cm (286 samples) for total carbon and nitrogen (TC and TN) and stable isotope (δ 13 C) analyses. Samples were dried at 60 • C for 48 h, frozen with liquid nitrogen, and ground in a ring mill. Analyses were performed using a Finnigan delta Plus EA-CF-IRMS spectrometer, located at the Centres Tècnics i Tecnològics of the Universitat de Barcelona (CCiTUB) [31].
Mercury concentrations were determined in 150 samples of the two cores collected in Rano Aroi and in selected vegetation samples of present-day flora located in the island. Dried, milled samples were analysed for Hg using a Milestone DMA-80 analyser hosted at the Departamento de Bioloxía Funcional (University of Santiago de Compostela). The analysis of one in every five samples was undertaken in duplicate to judge precision. A standard reference material of the moss Pleurozium schreberi [42] M3 was run with each set of samples. The quantification limit [average of the blanks + (10 × standard deviation of the blanks)] was 1.34 ng g −1 , and the mean recovery of the reference material was 90 ± 4%.

Mercury
Mercury concentrations in the Rano Aroi samples varied in the same range for both cores (Figures 2  and 3). Core ARO 06 08 has values that range between 35 and 1023 ng g −1 (147 ± 49, mean ± standard deviation), with the maximum concentration in the sample at 370 cm of core depth. The Hg minimum is located at 580 cm ( Figure 2). Core ARO 08 02 displays Hg values that range between 37 and 1088 ng g −1 (104 ± 56 ng g −1 ), and two maxima (1088 and 910 ng g −1 ) are located at 195 and 285 cm of core depth. According to Hg concentrations, this profile shows three distinct sections: (i) from the base of the core to 290 cm, values increase with depth but with small oscillations (122 ± 26 ng g  Table 1). The lowest value (341 ng g −1 ) corresponds to the rhizome of Scirpus californicus and the maximum value (11,175 ng g −1 ) to an unidentified rush sample.

Mercury Concentrations in Rano Aroi Records
The two Rano Aroi records (ARO 08 02 and ARO 06 01) cover different temporal ranges (38.7 ka cal BP-present day and 71 ka BP-8.5 ka cal BP, respectively) and were retrieved less than 50 m apart ( Figure 1). Previous studies using multiple short cores from the same peatland showed that spatial variability can be a problem for an accurate Hg study, since vegetation and micro-topography changes can affect the interception and retention of atmospheric deposition over relatively short timescales [43,44]. However, the short distance between the cores, the small size of the mire (0.13 km 2 ), and the age-depth model based on both cores [31] seem to have largely solved this problem. Therefore, we discuss the processes that controlled Hg concentrations in Rano Aroi using both cores together; the differences between them are illustrated based on other proxies (as TC, TN and δ 13 C).
To the best of our knowledge, there is only one comparable record-the Pleistocene peat record from a tropical-sub tropical area-to Rano Aroi in the literature. This is the Pinheiros mire, located in Minas Gerais state, Brazil. The range of Hg concentrations in Rano Aroi records (~35 and 333 ng g −1 ) is, in general, comparable with that of Pinheiros mire (~36-370 ng g −1 ) [29], although the minima found in the Brazilian mire are considerably lower than those in Rano Aroi (<2 vs. 35 ng g −1 ), probably due to the large quartz content of the corresponding peat sections in the former [29]. On the other hand, the ARO cores show two extraordinary Hg peaks of~1000 ng g −1 at~20.0 ka cal BP and~4.7 ka cal BP (Figures 4 and 5) that are more than two-fold the maximum found in the Brazilian record.   Maxima Hg values in the Pleistocene and Holocene sections of Rano Aroi are comparatively high with respect to other peatlands worldwide, even in sections affected by anthropogenic emissions. For example, for the Industrial Period, maxima of ~800 ng g −1 were reported for the Czech Republic [45], >400 ng g −1 for Scotland [46,47], ca. 600 ng g −1 in Northeast China [48], ca. 130 ng g −1 for Canada [49], and ca. 300 ng g −1 in Spain [43]. Our two maxima of natural Hg concentrations are, however, similar to those (up to 1100 ng g −1 ) found in four peat cores from Belgium, dating to AD Maxima Hg values in the Pleistocene and Holocene sections of Rano Aroi are comparatively high with respect to other peatlands worldwide, even in sections affected by anthropogenic emissions. For example, for the Industrial Period, maxima of~800 ng g −1 were reported for the Czech Republic [45], >400 ng g −1 for Scotland [46,47], ca. 600 ng g −1 in Northeast China [48], ca. 130 ng g −1 for Canada [49], and ca. 300 ng g −1 in Spain [43]. Our two maxima of natural Hg concentrations are, however, similar to those (up to 1100 ng g −1 ) found in four peat cores from Belgium, dating to AD 1930-1980, which were interpreted to be associated with coal burning and smelter emissions [50]. Pre-Industrial Hg concentrations in the previously mentioned records are quite heterogeneous (such as 40-50 ng g −1 in the peatlands studied in the Czech Republic [45] or 111 ± 64 ng g −1 in the records from Northeast China [48]), but are in the range of the ARO records. These comparisons suggest that, under natural conditions, as expected in Easter Island at~20 and~4.7 ka cal BP, large amounts of Hg can accumulate over relatively low background values.

Environmental Conditions in Rano Aroi and Related Proxies
In previous studies on Rano Aroi cores, a combination of proxies was used to improve the knowledge on tropical and subtropical peat dynamics and the climatic history of Easter Island. The peatland demonstrated itself to be a good archive for reconstructing the environmental history of the area, which was mainly driven by climate [31,32], although without ruling out the effect of human activity in the last few millennia [51].
The δ 13 C variations indicated two main climatic periods [31,32]. From~70 to~55 ka BP, high δ 13 C values (mean −15‰, C 4 -type plant remains) suggest the predominance of drier-colder climatic conditions, while from~47 to~8.5 ka cal BP low δ 13 C values (−26 ‰, C 3 -type plant remains) indicate a general dominance of wetter climate conditions (Figure 4). A gradual transition between both climates took place between~55 ka BP and~47 ka cal BP, and this is reflected by progressively lighter values of δ 13 C (Figure 4). ARO 08 02 does not show this transition ( Figure 5) because it only covers the last 38 ka cal BP, but the δ 13 C values shifted to lower values in agreement with the ARO 06 01 tendency. The long-term δ 13 C trend was interrupted by negative excursions, interpreted as short wetter periods.
The same general pattern of δ 13 C data is followed by PC1, the first component of a Principal Component Analysis performed on the inorganic fraction of Rano Aroi [32]. PC1 is characterized by large positive loadings (>0.7) of typically lithogenic elements and some metals. This component was identified as a proxy of the long-term fluxes of inorganic particulate material to the peatland. Its similarity with the δ 13 C background trend suggests higher erosion during drier periods (~70 tõ 55 ka BP)-probably due to lower vegetation cover-and an increasing mineral input during the short rainfall events previously mentioned (Figure 4). Another process, identified by the second component (PC2), is the occurrence of enhanced precipitation events related to the strong runoff and delivery of large amounts of terrigenous particles (probably rich in organic matter) from the soils of the catchment to the mire. Matching with these events, the C/N ratio and total nitrogen (TN) values show abrupt decreases and increases, respectively. Higher TN values and lower C/N ratio were attributed to a higher contribution of lacustrine algal material, in contrast to high C/N values that indicate higher proportions of terrestrial or aquatic plant (versus algae) organic matter [32].

Temperature
Temperature has a significant effect on Hg oxidation. Atmospheric oxidation is currently thought to occur through a two-step oxidation mechanism, favored at cold temperatures, and based on an initial recombination of Hg(0) and Br followed by the addition of a second radical (e.g., Br, I, Cl, BrO, ClO, OH, NO 2 or HO 2 ) [4,52] in competition with the thermal dissociation of the HgBr intermediate [53]. Low temperatures stabilize the intermediate HgBr species, allowing the second oxidation step that yields divalent Hg to occur. As a result, the oxidation, and associated surface deposition, of Hg is accelerated during cold periods. The~20 ka cal BP Hg maximum corresponds to the coldest phase, which occurred at the end of the Last Glacial Maximum [32]. Furthermore, the early Holocene Hg peak recorded in Aroi at~8.5 ka cal BP (ARO 06 01) coincided with an unusual cold period known as the 8.2 cooling event [54], given chronological uncertainties in the Aroi age-depth model [31]. Thus, the abrupt temperature shifts may have played a significant role in the Hg concentrations at the end of the LGM and the early Holocene. Similarly, at the time of the cold event Henrich 1 (H1,~17.5), Hg content in Pinheiros reached the maxima [29]. None of the four identified processes controlling Hg content in Pinheiros mire (i.e., wet deposition, aeolian dust, catchment soil erosion and OM decomposition) were able to explain the peak of Hg during H1 [29]. However, the results obtained in Rano Aroi suggest that the increase during the H1 was at least partially due to surface deposition associated with Hg atmospheric oxidation under cold conditions.

Effective Moisture
Increases in Hg concentration seem to match, at least partially, with strong runoff events, based on PC2 scores peaks (Figure 4). Six of these main events (centered at~60.5,~56.6 ka BP and~46.5, 43.5,~20.0 and~10.0 ka cal BP) coincide with peaks in Hg concentration, while PC2 peaks centered at~54.0 and~49.5 ka cal BP do not show correspondence with Hg rises (Figure 4). The PC2 peak at 10 ka cal BP agrees also with an increase in Hg (333 ng g −1 ) in ARO 06 01. Increased rainfall, which is the main driver for runoff events, should be the common link with Hg rises in Rano Aroi through enhanced Hg-wet deposition. However, it is unclear whether this is the only mechanism controlling the increase in Hg concentration in peat during wetter periods, or whether other mechanisms like the catchment erosion effect, the mire vegetation, peat decomposition and volcanic activity may have played a role.

Dry vs. Wet Deposition in Rano Aroi
The oxidation and deposition of Hg from the atmosphere through rainfall is a well-known process in the Hg cycle, however the importance of this process in environmental records is still under debate. Mercury isotopic analyses in living sphagnum mosses, upper peat layers, and rainwater samples were used in a recent publication to investigate the dominant modern Hg depositional processes in peat bogs [55]. The results of the three-year field experiment indicated that Hg deposition was dominated by gaseous elemental Hg dry deposition (79%) rather than wet deposition (21%). However, a later paleolimnological study conducted in a Pyrenean lake [24] showed that increased precipitation and more humid conditions controlled Hg accumulation in the area at a centennial scale, suggesting that wet deposition was the main driver at longer time-scales. In the tropical Pacific, modeling exercises during the pre-industrial period suggest dry deposition of Hg(II) as the main process governing Hg accumulation onto the land surface, with approximate values of~4-5 µg m −2 year −1 Hg(II) dry deposition in the Easter Island region [56]. Therefore, dry deposition may explain background Hg values in Rano Aroi. Nevertheless, other physico-chemical factors may also be considered when interpreting the interplay between dry and wet deposition onto land surfaces (i.e., peatlands and lakes). Large uncertainties exist regarding the atmospheric reduction of newly formed oxidized Hg(II). Although clouds enhance the scavenging and deposition of Hg to the surface, cloud droplets have also been suggested to be the medium where Hg(II) is reduced to elemental Hg, which is then re-emitted back to the gas phase [57]. Therefore, in the presence of clouds during rainy periods the dry deposition of atmospheric Hg may be slowed down. This would compete with wet deposition of Hg via precipitation, although it is almost certain that the latter would dominate the overall net atmospheric deposition of Hg to the surface. Thus, variable cloudiness might partially explain the discrepancies between the Pyrenean records.
At the Pleistocene scale, evidence supports the role of rainfall in intensifying Hg deposition. In Pinheiros mire, higher Hg concentrations found during a humid period from~26.7 to~2.8 ka cal BP were interpreted to result from increases in Hg concentration driven by precipitation (wet deposition-rainfall) [29].
Several differences exist in the Rano Aroi Hg cycle with respect to Pinheiros. First, the increases in Hg concentration are related to abrupt rainfall events (throughout ARO 08 02 and ARO 06 01) instead of being a continuous process over a long time period. Second, there is apparently no dilution effect, due to the input of inorganic matter from the catchment despite the increased delivery of terrigenous material to the peatland during high-rainfall episodes [32]. Third, considering the extraordinary Hg peaks in the Rano Aroi cores (~1000 ng g −1 ), there is a wider range in Hg concentrations than in Pinheiros (36-370 ng g −1 ). This difference may be due to the high interannual rainfall variability in Easter Island (from 500-1800 mm), or indicate that other processes may have played a role in Hg deposition or accumulation.
The good match between the increases in Hg during wet periods in the mentioned long-term records suggests that Hg wash out from the atmosphere and deposition should be the main driver. In spite of dry deposition being found to be the main mechanism in the three-year study of the French bogs [55], it is possible that the perspective provided by the Pleistocene records (as Rano Aroi and Pinheiros) might highlight the dominance of the longer-term wet deposition processes ( Figure 6). in Hg concentrations than in Pinheiros (36-370 ng g −1 ). This difference may be due to the high interannual rainfall variability in Easter Island (from 500-1800 mm), or indicate that other processes may have played a role in Hg deposition or accumulation. The good match between the increases in Hg during wet periods in the mentioned long-term records suggests that Hg wash out from the atmosphere and deposition should be the main driver. In spite of dry deposition being found to be the main mechanism in the three-year study of the French bogs [55], it is possible that the perspective provided by the Pleistocene records (as Rano Aroi and Pinheiros) might highlight the dominance of the longer-term wet deposition processes ( Figure  6).

Effect of Catchment Erosion: Geologic Materials and Catchment Soils
The direct contribution of local geologic materials to Hg concentration in the peat during high-rainfall events is negligible. The small catchment (15.82 km 2 ) is composed of highly porphyritic olivinic tholeite and basalt and hawaiite lava flows [34]. Specific information for Hg in basaltic materials is sparse in the published scientific literature. The most recent reviews of available data indicate that Hg in basalts is variable, generally in the range of 2-35 ng g −1 [58]. Data from basaltic reference materials showed that the Hg content in samples from island arcs or the circum-Pacific belt is lower than that of samples from the continental origin, with concentrations generally at the lower end of the mean for the basaltic samples analyzed (7-34 ng g −1 ) [59]. According to this, we would expect a low contribution of Hg from the catchment materials due to weathering.
However, catchment soils may maximize the effect of increased Hg concentration during rainfall events. Volcanic soils can retain Hg released form the parent material, as well as that

Effect of Catchment Erosion: Geologic Materials and Catchment Soils
The direct contribution of local geologic materials to Hg concentration in the peat during high-rainfall events is negligible. The small catchment (15.82 km 2 ) is composed of highly porphyritic olivinic tholeite and basalt and hawaiite lava flows [34]. Specific information for Hg in basaltic materials is sparse in the published scientific literature. The most recent reviews of available data indicate that Hg in basalts is variable, generally in the range of 2-35 ng g −1 [58]. Data from basaltic reference materials showed that the Hg content in samples from island arcs or the circum-Pacific belt is lower than that of samples from the continental origin, with concentrations generally at the lower end of the mean for the basaltic samples analyzed (7-34 ng g −1 ) [59]. According to this, we would expect a low contribution of Hg from the catchment materials due to weathering.
However, catchment soils may maximize the effect of increased Hg concentration during rainfall events. Volcanic soils can retain Hg released form the parent material, as well as that deposited from the atmospheric pool [60]. As mentioned above, the type of geological material in the Rano Aroi catchment does not seem to have significant amounts of Hg so, most probably, all the Hg retained in the soils may ultimately originate from atmospheric deposition. The mobilization of Hg due to soil erosion [61] could also take place in volcanic areas, since Hg is associated with metal (Al-Fe)-humus complexes. Margalef et al. [31,32] also identified increases of Fe content and high values of PC2 (large positive loading of Fe) during enhanced rainfall periods. At least part of the Fe in (andic) volcanic soils may be as Fe-humus compounds [62], to which the Hg could be bound [60,63].These compounds may be transported to the mire by superficial runoff (Figure 6).
The comparison of Hg concentration in ARO 06 01 with the PC1 and PC2 components (the long-term background fluxes of inorganic particulate material and the delivery of large amounts of terrigenous particles, respectively) only shows agreement during the shorter wet events and not for the long-term trends (see Figure 4). This suggests that matter fluxes do not affect the concentration of Hg, but precipitation may be the common driver.
It is worth mentioning the presence of a short rise of Hg and PC2 in ARO 06 01 at~37 ka cal BP that, according to the δ 13 C record, does not correspond to a wet period (Figure 4), pointing to other processes as an increase of Hg concentration due to terrigenous Hg-rich input. However the existence of a chronological uncertainty in this section [32] invites caution in the interpretation.
The data support the idea that wet deposition-directly on the mire or through the catchment circuits-was the dominant process controlling Hg concentration in the Rano Aroi peatland sincẽ 71.0 ka, essentially during high-rainfall events. The extremely high Hg concentration (1023 and 910 ng g −1 for ARO 06 01 and 08 02, respectively) at~20 ka cal BP in both cores, and the Holocene major peaks at~4.7 ka cal BP (ARO 08 02) and at 8.5 ka cal BP (ARO 06 01), seem to have been caused by temperature and/or precipitation variability (that eventually control catchment erosion) since they coincide with colder periods with an increase of PC2 (higher run-off) and a decline in δ 13 C (ARO 06 01 and ARO 08 02, respectively). For these peaks Hg is 3 and 4.5-fold (ARO 06 01 and 08 02, respectively) that of the next highest concentration (~300 ng g −1 ), suggesting the short-lived nature of these Hg increase episodes, which could indicate other processes may have enhanced Hg concentration in the mire.

Mire Vegetation
It has been shown that vegetation type and degree of decomposition can affect the Hg content in peat [64]. Table 1 reveals that different vegetation present in the island may involve different Hg concentrations in plant remains. According to the data presented here (Table 1) an increase in the abundance of rush, which showed large (~11,000 ng g −1 ) Hg concentrations, coupled with more humid conditions by~20 ka cal BP, might help to explain these increases in Hg ( Figure 6). However, we have no certainty of the presence of rush in the mire catchment at present, and the available pollen and macro remains data [31,32,65] do not allow us to support this speculation.

Peat Decomposition
The effect of organic matter decomposition in Hg accumulation is still a subject of debate. For short-term periods, it has been proposed that increased decomposition leads to an increase in Hg concentration [12,66]. This could be the situation during the extreme oxidation event caused by the drought period which occurred at~40.0 cal ka BP (Figure 4), when the Hg concentration slightly increased (Figures 4 and 6). On the other hand, the available data for longer Pleistocene-Holocene records, such as that of the Pinheiros mire [29], suggest that the effect of peat decomposition on Hg concentrations exponentially decreases to be almost negligible in peats with ages older than 10-11 ka [29]. There is no general decomposition trend in Rano Aroi and the short-term changes (reflected by changes in C/N ratio) seem to respond to enhanced rainfall periods and associated changes in mire vegetation [31], that may not have affected the Hg concentrations.

Aeolian Dust
Wind-borne dust particles derived from remote sources with higher Hg concentrations may also constitute a plausible explanation for the Hg peaks in the Aroi sedimentary sequence. Enhanced atmospheric dust deposition during the LGM may have also controlled the increase in Hg concentration in Aroi at 20 ka BP. This effect has been recently shown [28] to explain the high Hg accumulation rates in Caço lake (NE Brazil) between~20 and 17 ka BP.
LGM atmospheric elevated dust loadings in the Southern Hemisphere had a most likely source area in the Australian region, where increased aridity and less-efficient scavenging during this period would have enhanced Hg-enriched wind-born particles transported by westerlies to the South Pacific ocean [67]. Similarly, dust deposition together with cold conditions were suggested to be the cause of the increase of Hg during the 8.2 event in the Pinheiros mire record [29].
The mid-Holocene Hg peak in the Aroi peat record is also in agreement with the maximum Hg accumulation recorded at 4.2 ka cal BP in varved Elk Lake (North America) [68]. These authors found a causative factor between increased aeolian dust transport and deposition from remote locations with Hg-enriched soils. This period recorded a severe aridification at a global scale [69] that might have increased dust fluxes and transport in a similar pattern as during the LGM. Thus, Hg depositional patterns may have varied following aeolian dust fluxes from remote locations, demonstrating the significance of Hg bound to mineral particles in the Hg load in terrestrial ecosystems.

Oceanic Evasion
Biologically productive regions in the tropical Pacific Ocean are a significant source of atmospheric Hg, mainly relating to phytoplankton's conversion of seawater Hg(II) to Hg(0) [70]. Evasion of Hg from surface ocean Pacific waters in the region is further supported by results that account for oceanic evasion fluxes of 4-5 × 10 −6 g m −2 year −1 in oceanic areas near Easter Island, based on the modeling of the Pre-Industrial period [56]. During the LGM, a three-fold increase in Hg concentration in Dome C in Antarctica was attributed to higher oceanic productivity [71] increasing Hg emissions to the atmosphere and further deposition. However, Hg peaks in the Rano Aroi did not occur during periods of major oceanic productivity in the equatorial and tropical Pacific [72,73]. Thus, oceanic productivity will most likely have a significant contribution to the background Hg levels measured in the Aroi peat record.

Volcanic Activity
Individual volcanic eruptions have been identified for a long time as important natural sources of Hg. In addition to the direct emission of Hg, mainly as Hg(0) species [74,75], volcanic activity can increase the deposition of Hg to the environment by the oxidation of Hg(0) to Hg(II) [76]. The co-emission of Hg(0) and Br and SO 2 in volcanic plumes leads to an increase of Hg(II) formation that can easily be scavenged from the atmosphere by precipitation [76]. Furthermore, as the background atmospheric Hg is affected by the plume chemistry, the background Hg concentrations would be triggered by this mechanism and the total Hg deposit will be higher [76]. Thus, volcanic activity should also be considered in order to understand the triggers of the short-lived Hg peaks in Easter Island. Despite there being no volcanic lava flows older than two or three thousand years from Terevaka (one of the main volcanic structures in the island) [77], the Rano Aroi cores do not show evidence of recent volcanic activity (i.e., tephra layers). The absence of recorded volcanic eruptions on the island does not rule out the possibility of Hg degassing or Hg depletion by plume chemistry by active fumaroles, especially on an island containing more than 70 volcano craters [33,34]. Data (tephra layers) from South Patagonia indicated volcanic eruptions at~4.2 ka cal BP and~7.7 ka cal BP, identified as the Mt Burney (52 • S 72 • W) and Hudson volcanic eruptions (45 • S 72 • W), respectively [78,79]. We have no data to support the effect of these volcanic events for the Hg peak at~4.7 ka cal BP (ARO 08 02) and/or at~8.5 ka cal BP (ARO 06 01) in Easter Island, but they cannot be discarded as a possible Hg source. Moreover, although there is some mismatch between the Hg peak and the Hudson volcanic eruption, it has been shown that this specific eruption (at~7.7 ka cal BP) modified the atmospheric deposition dust during a period of~700 years [80]. A higher resolution would be needed in this section of the cores, as well as more cores, to confirm the hypothesis of the effect of volcanism and differences in intensity (i.e., differences in Hg concentration) between both events.

Fire Regime
Biomass burning is an important source of Hg emissions to the atmosphere [81]. Although Hg(0) represents the greatest fraction of Hg from biomass burning, modeling studies have shown that Hg(p) (with a shorter life-span in the atmosphere) also represent a significant amount of the Hg released during fire episodes [82]. Thus, we cannot exclude the effect of fire regimes to explain the short-nature Hg peaks through increasing atmospheric Hg concentration. For example during the LGM, charcoal data indicates an increase in fire activity at the tropical latitudes of South America [83]. What might have been an extra Hg emission to the atmosphere, was then eventually deposited at Easter Island during heavy rainfall episodes.

Conclusions
The Hg record in the Rano Aroi cores shows a huge range of values between 35-1000 ng g −1 , mostly driven by wet precipitation-directly on the mire or through the catchment circuits-essentially during high-rainfall events. Two large maxima occurred at the end of the LGM (~20 ka cal BP) and at~8.5 ka cal BP, both periods of general cold climatic conditions and wetness in Rano Aroi. Low temperature would accelerate the atmospheric oxidation of Hg(0) to divalent Hg, and if combined with high precipitation this would result in very efficient surface deposition of atmospheric Hg. Thus, both colder and humid conditions would have favored Hg accumulation in Rano Aroi, since Hg deposition is controlled by temperature and humidity variations [14,24].
Other processes such as mire vegetation, peat decomposition, aeolian dust, oceanic evasion, and volcanic activity (direct Hg emission and Hg oxidation) may have played a secondary role in Hg content. Oceanic evasion might have played a role as a general background Hg source, and increases in the fire regimen during the LGM at the tropical latitudes of South America might have been an extra Hg emission to the atmosphere that was eventually deposited on Easter Island during heavy rainfall episodes. Moreover, wind-borne dust particles derived from remote sources with higher Hg concentrations might also play a role in enhanced Hg concentrations during the LGM, and especially during the mid-Holocene Hg peak (4.2 ka cal BP), in agreement with previous works.
The combinations of these processes cause a range of variation of Hg concentrations of 28-fold in a remote and undisturbed area, at least during the studied period. The maximum values are higher than those recorded in most peat records belonging to the Industrial Period, highlighting that natural factors played a significant role in local Hg accumulation-sometimes even more so than anthropogenic sources.