Substantial accumulation of mercury in the deepest parts of the ocean and implications for the environmental mercury cycle

Significance

Mercury is a globally ubiquitous pollutant that is harmful to humans and animals. Most mercury entering the environment is released from anthropogenic sources and then stored for some period in soils and water bodies before potentially being remobilized, which greatly facilitates its global distribution. Although the ocean is recognized as the largest ultimate sink of mercury, the distribution and burial of mercury in deep ocean sediments remains largely unknown because of the difficulty in obtaining samples from these ecosystems. This study found that although hadal trenches (the deepest parts of the oceans) occupy a very small portion (<0.5%) of the ocean area, they may receive a large amount of mercury and represent a disproportionately important, previously overlooked sink for mercury.

Abstract

Anthropogenic activities have led to widespread contamination with mercury (Hg), a potent neurotoxin that bioaccumulates through food webs. Recent models estimated that, presently, 200 to 600 t of Hg is sequestered annually in deep-sea sediments, approximately doubling since industrialization. However, most studies did not extend to the hadal zone (6,000- to 11,000-m depth), the deepest ocean realm. Here, we report on measurements of Hg and related parameters in sediment cores from four trench regions (1,560 to 10,840 m), showing that the world’s deepest ocean realm is accumulating Hg at remarkably high rates (depth-integrated minimum–maximum: 24 to 220 μg ⋅ m⁻² ⋅ y⁻¹) greater than the global deep-sea average by a factor of up to 400, with most Hg in these trenches being derived from the surface ocean. Furthermore, vertical profiles of Hg concentrations in trench cores show notable increasing trends from pre-1900 [average 51 ± 14 (1σ) ng ⋅ g⁻¹] to post-1950 (81 ± 32 ng ⋅ g⁻¹). This increase cannot be explained by changes in the delivery rate of organic carbon alone but also need increasing Hg delivery from anthropogenic sources. This evidence, along with recent findings on the high abundance of methylmercury in hadal biota [R. Sun et al., Nat. Commun. 11, 3389 (2020); J. D. Blum et al., Proc. Natl. Acad. Sci. U. S. A. 117, 29292–29298 (2020)], leads us to propose that hadal trenches are a large marine sink for Hg and may play an important role in the regulation of the global biogeochemical cycle of Hg.

Human activities have led to mercury (Hg) pollution on a global scale (1–4). The atmospheric lifetime of elemental Hg (Hg⁰, months) allows for long-distance transport and eventual deposition in many remote regions worldwide. The ocean receives most of its Hg from atmospheric deposition and river export (5). Hydrothermal vents, as well, can be sources of small quantities of Hg in the deep-open ocean (6). Human activities have directly emitted 260,000 megagrams (Mg) of Hg into the atmosphere since industrialization (7), while recent observations suggest that 60,000 Mg of this Hg is present in the global ocean (8). The transportation of Hg from the surface ocean to deep-sea sediments represents long-term sequestration of anthropogenic Hg, with a currently estimated flux of 200 to 600 Mg ⋅ y⁻¹, up to a 200% increase compared with that of the preindustrial period (SI Appendix, Table S1) (6, 9–14). Nevertheless, these model-based estimates do not extend beyond the abyssal zone (2,000 to 6,000 m), mainly owing to the scarcity of observations (Fig. 1A). Thus, the burial flux of Hg in deep-sea sediments remains inadequately constrained, in particular regarding the hadal zone (6,000 to 11,000 m), the deepest part of the ocean, where the Hg cycle is virtually unknown.

Fig. 1.

Global and regional Hg accumulations in oceanic surface sediments. (A) Distribution of Hg burial rates in surface sediments synthesized from published literature and our measurements. The grid-registered bathymetric map of the global seafloor was created by the National Centers for Environmental Information of the National Oceanic and Atmospheric Administration. The distributions of Hg concentration and SR are provided in SI Appendix, Fig. S6. In this panel, the SR for MST was assumed to be similar to that for the MT (20). (B) Comparisons of Hg concentration, SR, and Hg burial rate in surface sediments. Different letters (a to c) denote significant differences; that is, for two groups with different letters, their mean values are significantly different at the *P < 0.05 level. The results of the comparisons are provided in SI Appendix, Table S9. All data for surface sediments are classified as continental shelves (water depth <200 m, n = 33), bathyal (200 to 2,000 m, n = 43), and abyssal zones (2,000 to 6,000 m, n = 6) and sampling sites (n = 9) across four disparate trench regions in this study. Previously published data are provided in SI Appendix, Table S3.

Hadal zones are mainly composed of trenches formed along the subduction zones of oceanic plates (15). These trenches occupy 0.4 to 0.5% (∼1.5 × 10⁶ km²) of the total area of the global ocean and represent one of the least explored habitats on our planet. Since rapid attenuation of organic matter supply with increasing water depth is common (16), hadal trenches were traditionally considered as biological deserts. However, mounting evidence has shown higher–benthic biomass abundance, higher–organic carbon (OC) content, and more intense microbial activity in several hadal trenches compared to the surrounding abyssal plains (17–21). Such unusual characteristics are presumably attributable to enhanced particulate organic matter supply in the hadal trenches because of the V-shaped topography and frequent, mass-wasting events shuttling higher–organic matter sediments down the slope (15, 22). Considering that the majority of Hg deposited in the ocean in high-productivity areas is scavenged onto organic, rich particles, Hg might also be efficiently accumulated in hadal trenches. Additionally, recent studies have found that surface ocean–derived methylmercury is bioaccumulating in hadal trench biota (23–25), motivating an investigation of the potential role of the deepest–ocean realm in the marine Hg cycle.

We aim to understand whether hadal trenches are important for the global Hg cycle and whether the impact of human activity on Hg contamination in hadal trenches is significant. We collected nine sediment cores (water depth from 1,560 to 10,840 m) from four disparate hadal trench regions (trench axes and surrounding slopes) in the Pacific Ocean: the Mariana Trench (MT), Massau Trench (MST), New Britain Trench (NBT), and Bougainville Trench (BT; Fig. 1A). We conducted geochemical analyses of Hg, OC, total nitrogen (TN), grain size, and stable isotopic compositions of Hg and carbon (δ¹³C; SI Appendix, Table S2) and applied statistical tools to distinguish the potential influences of ancillary parameters on variations in Hg concentrations.

Results and Discussion

High-Hg Burial Rates in Hadal Trenches.

Our measurements reveal remarkably high-Hg burial rates in hadal trenches. We find that the surface burial rates (upper 1 cm) in MT1 (10,840 m), NBT1 (8,225 m), and BT6 (8,903 m) reach 27 ± 2.6 (1σ), 189 ± 12, and 241 ± 36 μg ⋅ m⁻² ⋅ y⁻¹, respectively. These values are greater than those modeled for the global deep sea (0.57 to 1.8 μg ⋅ m⁻² ⋅ y⁻¹) (6, 9–14). Despite the extreme water depth, the Hg burial rates in hadal trenches are close to or higher than those in many continental shelves (water depth 0 to 200 m) and inland lake sediments with significant terrigenous or anthropogenic impacts, such as the shelves of the Gulf of Mexico (26 μg ⋅ m⁻² ⋅ y⁻¹) (26), Hudson Bay (32 ± 13 μg ⋅ m⁻² ⋅ y⁻¹) (27), Middle East China Sea (4.2 to 90 μg ⋅ m⁻² ⋅ y⁻¹) (28), and Canadian lakes (35 ± 10 μg ⋅ m⁻² ⋅ y⁻¹) (29).

In addition to our measurements, we also conducted a metadata analysis of Hg burial rates in surface sediments in different marine settings, namely, continental shelves, the bathyal zone (200 to 2,000 m), and the abyssal zone, from published literature to compare the hadal trench data to other marine depositional environments (Fig. 1 and SI Appendix, Table S3 and Text S1). Our synthesis reveals that Hg burial rates in hadal trenches (median: 133 μg ⋅ m⁻² ⋅ y⁻¹ and average: 140 ± 78 μg ⋅ m⁻² ⋅ y⁻¹) are significantly higher than those in continental shelves (39 and 105 ± 173 μg ⋅ m⁻² ⋅ y⁻¹), the bathyal zone (34 and 94 ± 151 μg ⋅ m⁻² ⋅ y⁻¹), and the abyssal zone (2.5 and 3.2 ± 3.1 μg ⋅ m⁻² ⋅ y⁻¹; Fig. 1B). Correspondingly, the median and mean sedimentation rates (SRs) in hadal trenches (0.11 and 0.12 ± 0.055 cm ⋅ y⁻¹, respectively) are substantially higher than those in the abyssal zone (9.6 × 10⁻³ and 9.8 × 10⁻³ ± 8.4 × 10⁻³ cm ⋅ y⁻¹, respectively; Fig. 1B). This difference explains the high-Hg burial rates in the hadal trenches. However, the median SRs in the bathyal zone (0.13 cm ⋅ y⁻¹, average: 0.25 ± 0.36) and continental shelves (0.18 cm ⋅ y⁻¹, average: 0.25 ± 0.20) are slightly higher than those in trench regions (Fig. 1B). Therefore, SRs alone cannot explain the higher-Hg burial rates in hadal trenches compared to the bathyal zone and continental shelves. We also find that the Hg concentration is significantly higher in trench sediments (average: 89 ± 33 ng ⋅ g⁻¹) than in other marine settings (47 to 60 ng ⋅ g⁻¹; Fig. 1B), suggesting the enrichment of Hg in hadal sediments. Thus, both high-Hg concentrations and high SRs contribute to high-Hg burial rates in hadal trenches, making them hotspots for Hg burial in the ocean.

Source and Mechanism Resulting in High-Hg Burial Rates.

We propose that substantial amounts of Hg from the surface ocean can be transported to the hadal trenches. The MT and MST are remote from continents, while NBT and BT are situated along underdeveloped watersheds lacking anthropogenic Hg sources during the past century. Thus, Hg entering these trenches is likely derived from Hg emitted from distant industrial regions and transported by air and ocean currents. Positive values of odd and even mass-independent fractionation (MIF, Δ¹⁹⁹Hg and Δ²⁰⁰Hg) of Hg isotopes in the NBT (depth-integrated averages of 0.24 ± 0.07‰ and 0.06 ± 0.02‰, respectively; SI Appendix, Table S4) are consistent with those in bathyal and abyssal zones (30), indicating that Hg in hadal trenches is more likely associated with atmospheric input to the surface ocean than riverine Hg inputs (31, 32). Furthermore, the slope of the linear regression of Δ¹⁹⁹Hg versus Δ²⁰¹Hg (y = 0.92x + 0.08, R² = 0.42) highlights the strong photoreduction signature of Hg (33, 34). These results support that atmosphere-derived Hg can be transported down to hadal trenches through sinking particles, consistent with recent findings in deep-sea organisms (23–25).

The adsorption of Hg to the matrix of terrigenous particles may also play an important role in the efficient transport and deposition of Hg into hadal trenches. The OC/TN ratios and δ¹³C values are widely used as source indicators, given the distinct differences between terrigenous and marine OC (35). The higher OC/TN and depleted δ¹³C values suggest that terrigenous OC is significant in NBT1 (13 ± 1.4‰ and −24.8 ± 0.78‰, respectively) and BT6 (9.4 ± 0.81‰ and −22.1 ± 0.42‰, respectively; Fig. 2A). In contrast, the lower OC/TN and enriched δ¹³C values suggest that terrigenous OC is insignificant in MT1 (7.0 ± 0.93‰ and −19.9 ± 0.40‰, respectively) and MST1 (4.6 ± 0.22‰ and −19.7 ± 0.20‰, respectively). These source assignments are consistent with recent studies that investigated the distribution of branched tetraether lipids and also suggested that the terrigenous input is considerable in the NBT and BT but is minimal in the MST and MT (36, 37). Using a two-endmember mixing model (38), we estimate the fractional contributions of terrigenous OC to be 64 ± 9.6% in the NBT axis and 33 ± 4.9% in the BT axis, which is much higher than in MT and MST (5.0 ± 4.8% and 3.1 ± 2.3%, respectively). Assuming constant Hg/OC ratios in riverine particles (an average of 6.0 and 7.2 μg ⋅ g⁻¹ C in rivers flowing to the South and North Pacific oceans, respectively, for the NBT/BT and MT/MST, respectively) (5, 39), we calculate that the contributions of terrigenous Hg inputs are significant in the NBT (49 ± 20%) but negligible in oceanward slopes and other trench axes (0 to 2%). These results suggest that the inputs of terrigenous particles facilitate the delivery of surface ocean Hg to some hadal trenches.

Fig. 2.

Long-term variation in Hg accumulation in trench regions. (A) Comparisons of Hg, OC, OC/TN ratio, stable carbon isotopic composition (δ¹³C), and grain size for different sediment cores across four trench regions. The data for each sediment core are compared with the average and variation in the data for all cores. The results of the comparisons are provided in SI Appendix, Table S10; ns stands for not significant. (B) Trends of Hg burial rates for different sediment cores illustrated by smoothing curves, and significant fixed effects of anthropogenic release of Hg (AR), OC, and grain size (GS) on variations in Hg burial rates for different cores based on the multivariate regression model (SI Appendix, Table S6). The ordinate of MST1 represents the Hg concentration (ng g⁻¹) due to a lack of sedimentation rate. Error bars represent SDs. Shaded areas represent 95% CIs. Turbidites represent the potential impact of geological events such as earthquake turbidites on the sediment core. ^†P ≤ 0.10, *P < 0.05, **P < 0.01, ***P < 0.001.

We attribute the high burial rates of Hg in trench regions to the lateral transport from shelves and slopes and the vertical transport through the water column. Other sources, such as carrion fall, dust, and hydrothermal events, were likely minor. Infrequent carrion falls might subsidize the Hg flux to deep seas but are unlikely to be an important source (40). Dust input is only important in regions where other sediment supplies are low, and hydrothermal and diagenetic processes cannot explain the remarkable excess ²¹⁰Pb signals and large positive values of MIF in our study (23). Vertical transport is usually thought to be the major pathway for the deposition of Hg in deep-sea sediments. During vertical transport, Hg might become enriched in sinking particles at the expense of continuous OC remineralization and/or carbonate dissolution below the carbonate compensation depth (CCD), partly explaining the higher-Hg concentration and Hg/OC in hadal trenches (see Figs. 1B and 3) (41). However, vertical transport alone cannot explain the high SRs in hadal trenches compared with abyssal basins because the SR usually decreases with increasing depth (Fig. 1B). Thus, Hg in hadal trenches is also from lateral transport because of their V-shaped topography with steep slopes that can create gravity-driven downward transport and even mass-wasting events from shallow areas, such as slopes and edges of the continental shelves, some of which are significantly influenced by terrigenous input. During this process, part of the sediments can be resuspended into the water column, facilitating vertical transport. Because positive MIF is found on the outer shelves and slopes (30), either transport mechanism could explain the positive values of MIF. Several studies have shown strong excess ²¹⁰Pb content in trench sediments, such as in the Japan, Atacama, and Kermadec trenches (42, 43), and the preservation of carbonates far below the CCD in NBT, Palau, and Puerto Rico trenches (20, 44–48), additionally supporting rapid sedimentological processes. These multiple lines of evidence demonstrate that high-Hg burial rates are likely to be common in hadal trenches.

Fig. 3.

Physical and biogeochemical processes affecting the transport and fate of Hg in different types of trench regions. Hadal trenches were separated into two major types, namely, trenches adjacent to land (e.g., NBT and BT, respectively) and trenches distant from land (e.g., MT and MST, respectively) (54). Gray arrows represent major external sources of Hg. Light brown arrows and green arrows represent Hg transport dominated by the transport of terrigenous and marine particles, respectively, as indicated by the OC/TN ratio and stable carbon isotopic composition (δ¹³C) signal. Blue arrows represent materials transported by water exchanges. Brown and green dots represent Hg bound to particles that originate from terrigenous inputs and autochthonous production, respectively. The size of the dot represents the grain size of the particle. a-data from Motta et al. (31). b-data represents the terrigenous and marine endmembers of δ¹³C (38, 62).

Rapidly Increasing Hg Accumulation in Hadal Trenches since Industrialization.

Previous studies have shown that Hg contamination in continental shelves across a large latitudinal gradient has increased continually with accelerated human activity (26, 27, 49). Our work confirms that this trend also exists in the deepest and most remote parts of the ocean, with an enrichment factor of 1.6 ± 0.2 since the industrialization, consistent with many shelf areas (1.3 to 2.0) (26, 27, 49). Our observations for the hadal trenches are also consistent with many model-based predictions (6, 9–14). Based on Hg concentration, SRs, and dry sediment density, we estimate an average Hg burial rate of 132 ± 70 μg ⋅ m⁻² ⋅ y⁻¹ in the four trench regions from 1950 to the present, a period previously defined as the acceleration in global industrialization (50). This mean rate is 30% greater than the averages during 1900 to 1950 (98 ± 59 μg ⋅ m⁻² ⋅ y⁻¹) and 60% greater than pre-1900 (81 ± 54 μg ⋅ m⁻² ⋅ y⁻¹). These increasing trends are more significant on oceanward slopes. For example, averages of 90 ± 14, 109 ± 24, and 270 ± 142 μg ⋅ m⁻² ⋅ y⁻¹ for post-1950 in NBT2, NBT3, and BT10, respectively, which are two- to threefold greater than those in pre-1900 (Fig. 2B). Although Hg burial rates differ among trench regions (range of depth-integrated rates: 24 ± 3.6 to 222 ± 61 μg ⋅ m⁻² ⋅ y⁻¹) because of geographic differences in Hg inputs and topography, eight out of nine sediment cores (excluding BT7) exhibit sharp increases in Hg concentrations post-1900 (Fig. 2B and SI Appendix, Table S5) (e.g., from 63 ± 13 to 94 ± 35 ng ⋅ g⁻¹ at trench axes [NBT1, BT6, MT1, and MST1], 43 ± 8.9 to 90 ± 7.3 ng ⋅ g⁻¹ at oceanward slopes [NBT2, NBT3, and BT10], and 50 ± 27 to 79 ± 68 ng ⋅ g⁻¹ at landward slopes [NBT4 and BT7]). These ubiquitously increasing trends suggest that hadal trench sediments record changes in Hg inputs. We attributed the distinct temporal profile of Hg concentration in core BT7 to its proximity to land (<40 km) and the shallow water depth (1,560 m), in which a recent large flux of biogenic particles from the surface ocean could dilute anthropogenic Hg in sediments.

We find that changes in OC content and particle grain size cannot satisfactorily explain the observed increasing trends in Hg concentration and that other factors are influencing the Hg concentration in trench sediments. Positive correlations between Hg concentrations and sedimentary OC found in some cores (e.g., NBT2 to ΝΒΤ4) confirm that OC content influences Hg concentrations in trench sediments. Additionally, the changes in grain size also partly explain the variations in Hg concentrations (e.g., NBT1, BT6, and MT1; SI Appendix, Figs. S1 and S2). However, a paradox exists: A significant decrease in Hg concentrations coincides with decreasing grain size (e.g., NBT2 and NBT3), and significant increases in Hg concentrations occurred without increasing trends in sedimentary OC (e.g., NBT1, NBT2, NBT4, and MT1). After we eliminated the effects of OC and grain size on Hg concentration using multivariate regression models, the residuals showed conspicuously increasing trends for all cores except for BT7 (SI Appendix, Fig. S3). Thus, we introduced the flux of anthropogenic Hg released into the environment into the regression models (Fig. 2B and SI Appendix, Fig. S2) (7), based on the assumption that anthropogenic release is associated with the ²¹⁰Pb content of sediment. The results show that anthropogenic release has an overarching positive influence on Hg concentrations in most cores from disparate regions. Given the combined effect of anthropogenic Hg release, OC, and grain size on Hg concentration, a linear mixed-effect regression model was further applied to disentangle the relative importance of each factor (R²_c = 0.83 when regional variance was included). This model confirms that anthropogenic Hg input explains larger variable components than does OC and grain size (F values: 70, 10, and 0.023, respectively; SI Appendix, Table S6). Thus, our results demonstrate that anthropogenic Hg has reached the deepest parts of the ocean and changed the vertical Hg profiles in sediments from multiple trenches.

Implications for Global Hg Cycle.

Our findings highlight the important role of hadal trenches in the global Hg cycle. Globally, there are 33 hadal trenches, many of which are along active continental margins (SI Appendix, Table S7). Assuming a linear change in the burial rate within a trench region, and that the MT/MST and NBT/BT represent the two endmembers of Hg burial rates in hadal trenches that are close to and distant from the continents, respectively, we scale Hg burial flux up to the global hadal trenches to an average of 157 ± 65 Mg ⋅ y⁻¹ from 1950 to the present or 164 ± 63 Mg ⋅ y⁻¹, based on the surface sediment data. The latter number equals 7.1 ± 2.7% of the annual anthropogenic Hg emissions (3) and 27 ± 10% of the annual Hg burial flux in deep-sea sediments suggested by the United Nations Environment Program (51). We further estimate that the amount of Hg buried in hadal trenches from 1900 to the present could reach 17,000 ± 5,000 Mg, which is of the same order as the amount of anthropogenic Hg in the global ocean (60,000 Mg) (8). The high-Hg burial flux in the hadal trenches is consistent with recent findings that the amount of Hg burial in global ocean sediments appeared to be underestimated (52), and approximately one-third of anthropogenic Hg that entered the ocean was in deep-sea waters (8). Given that the hadal trenches account for less than 0.5% of the global ocean surface, such a disproportionally large burial flux suggests that hadal trenches are hotspots for Hg burial in the deep sea, which is consistent with the fact that many trenches are less than 200 km away from continents.

Currently, hadal trenches are among the least explored ecosystems on Earth, mainly because of very limited sampling opportunities at extreme depths. Our sample size, although still limited, is unprecedently large (four trenches) and therefore provides an opportunity to better understand the Hg cycle in the deepest ocean. However, our estimate does not include either episodic depositional events, such as earthquake-induced, mass-wasting events that could bring enormous amounts of materials into the hadal trenches in a short time or some hadal trenches (such as Japan and Philippine trenches) that might receive more anthropogenic Hg from adjacent, well-developed continents than the NBT and BT (51). Thus, our estimate for Hg burial flux in hadal trenches is likely to be conservative.

Similar to high levels of Hg concentrations in our target trenches, extraordinarily high levels of persistent organic pollutants and bomb–¹⁴C signatures were recently observed in the Kermadec Trench and MT, which emphasizes that materials derived from surface oceans could be rapidly delivered to the deepest hadal trenches (40, 53). Taken together, hadal trenches represent another important sink for anthropogenic Hg and other pollutants, and the turnover time of anthropogenic pollutants from surface oceans to deep-sea sediments is likely shorter than previous estimates, for example, >1,000 y for Hg (13). This emphasizes that the hadal zone is not presently in a pristine condition and deserves more attention. Finally, although our cores generally do not record the substantial Hg release from gold and silver production during 1860 to 1900, our observations are consistent with the general increasing trend of anthropogenic Hg release since 1900 (7, 52). Our results, together with previous modeling work and observations in shallower marine settings, support the view that the postindustrial increase in Hg burial rates is ubiquitous in the global ocean.

Materials and Methods

Study Area and Sampling.

The MT, NBT, BT, and MST, all located in the Pacific Ocean, cover a latitude range of ∼7° S to 12° N. These four trenches differ significantly in net primary production and terrigenous influence (54). MT and MST underlie oligotrophic waters, whereas NBT and BT are overlain by moderately productive waters. Terrigenous input is insignificant in MT and MST because of the remoteness from any landmass. In contrast, NBT and BT are located near large landmasses and receive significant terrigenous inputs (38). NBT and BT are adjacent to New Britain Island and Bougainville Island, which are high standing islands of Oceania covered with volcanic ash, ash-derived alluvial soils, tropical lush rainforest, and numerous small mountainous rivers. These rivers efficiently transport terrigenous material into the coastal ocean and the NBT (55). We used autonomous, full-ocean depth landers to retrieve sediment cores from the trench axes of the MT (MT1, ∼10,840 m, Challenger Deep), NBT (NBT1, ∼8,225 m), BT (BT6, ∼8,903 m), and MST (MST1, ∼6,990 m) during an 11,000-m sea trial expedition aboard the research vessel (R/V) Zhangjian (December 2016 to February 2017). We also used a box corer to collect samples from the adjacent landward and oceanward slopes of NBT (NBT2 to NBT4, ∼4,130 to 5,860 m) and BT (BT7 and BT10, ∼1,560 to 4,700 m). We stored all sediment cores immediately at −20 °C and then sliced the cores at 1-cm intervals onboard. All the samples were freeze dried in a −40 °C condenser in the laboratory.

Experimental Procedures.

Following the US Environmental Protection Agency Method 7473, we determined the Hg concentrations in sediments using a Milestone direct Hg analysis system (DMA-80), which performs thermal decomposition followed by gold amalgamation and detects Hg by atomic absorption spectrometry (56). Quartz boats were used to introduce the samples. All samples were measured in triplicate, and ∼0.5 g of each sample was used. We analyzed one replicate standard for every 10 samples. The average recovery of certified reference material (sediment, BCR-580) was 91 ± 7.0% (sample size n = 19), and the method blank was 0.026 ± 0.013 ng ⋅ g⁻¹ (n = 17) for Hg in all samples. The detection limit for Hg was 0.065 ng ⋅ g⁻¹, calculated based on the average concentrations of the method blanks plus triple the SD of the blanks. All the samples had Hg concentrations above the detection limit. The average relative SD was 5.5 ± 6.0%, based on triplicate analysis.

We measured the isotopic composition of Hg using the methods reported in previous studies (24, 33). The Hg isotope ratio was measured using a multicollector inductively coupled plasma mass spectrometry (Nu Instruments Ltd.) at Tianjin University, China. We corrected the instrumental mass bias by adding an internal standard (NIST SRM 997, with a ²⁰⁵Tl/²⁰³Tl ratio of 2.387) and sample–standard bracketing. The analyses were run at 1 to 3 ng ⋅ mL⁻¹ Hg, and the bracketing standard concentrations were matched to those of samples within 5% (57). Mass-dependent fractionation (MDF) of Hg isotopes was reported using δ^xxxHg relative to the standard reference material (NIST SRM 3133) following the equation:

$${\delta }^{\mathit{xxx}}\text{Hg}\left(\text{‰}\right)=\left[\frac{{\left({}_{}{}^{\mathit{xxx}}\text{H}\text{g}/{}_{}{}^{198}\text{H}\text{g}\right)}_{\text{sample}}}{{\left({}_{}{}^{\mathit{xxx}}\text{H}\text{g}/{}_{}{}^{198}\text{H}\text{g}\right)}_{\text{NIST3133}}}-1\right]\times 1,000 \text{,}$$ [1]

where xxx represents different mass numbers of Hg isotopes (199 to 204 for ¹⁹⁹Hg to ²⁰⁴Hg). The MIF of Hg isotopes ($\mathit{\Delta }$^xxxHg), defined as the difference between the measured δ^xxxHg and the theoretical values, was calculated based on the kinetic MDF law:

$${\mathrm{\Delta }}^{\mathit{xxx}}\text{Hg}={\delta }^{\mathit{xxx}}\text{Hg}-\beta \times {\delta }^{202}\text{Hg} ,$$ [2]

where β represents a scaling constant to calculate the theoretical kinetic MDF of Hg isotopes, which are 0.252, 0.502, 0.752, and 1.493 for ¹⁹⁹Hg, ²⁰⁰Hg, ²⁰¹Hg, and ²⁰⁴Hg, respectively. To evaluate the accuracy and precision of the treatment protocol, we regularly pretreated and analyzed a certified reference material (soil standard, GBW07405), following the same procedures (58). The high recovery rate (98.5 ± 0.7%, n = 4) and isotope data are consistent with previous results [δ²⁰²Hg, Δ¹⁹⁹Hg, Δ²⁰⁰Hg, and Δ²⁰¹Hg of −1.75 ± 0.12‰ (2σ, n = 4), −0.28 ± 0.06‰, 0.01 ± 0.04‰, and −0.30 ± 0.07‰, respectively (58)], showing high-quality control during the pretreatment and measurements. Repeated measurements of the standard UM-Almadén Hg (n = 13) yielded values of −0.57 ± 0.08‰, 0.00 ± 0.02‰, 0.00 ± 0.03‰, −0.03 ± 0.03‰, and 0.01 ± 0.06‰ (2σ) for δ²⁰²Hg, Δ¹⁹⁹Hg, Δ²⁰⁰Hg, Δ²⁰¹Hg, and Δ²⁰⁴Hg, respectively, consistent with the published data (33, 57). The 2σ SDs of the UM-Almadén Hg measurements were considered to be analytical uncertainties for the samples (23, 33). Since MDF in marine sediments can occur through multiple physical, chemical, and biological processes (59), our discussion mainly focused on MIF.

A total of ∼1 to 2 g of each sample was treated with 1 N HCl for 48 h at room temperature to remove carbonates, rinsed to neutral pH, and freeze dried. After homogenization with an agate mortar and pestle, ∼35 to 40 mg decarbonated sediment was weighed. We analyzed the sediment elemental compositions (OC and TN) and carbon stable isotopes (δ¹³C) using a model 100 isotope ratio mass spectrometer (IsoPrime Corporation) and a Vario ISOTOPE cube elemental analyzer (Elementar Analysensystem GmbH). Stable carbon isotopic data are reported in δ notation relative to Vienna Pee Dee Belemnite. The intralab standard for normalizing δ¹³C was USG24 (graphite, −16.05‰). The average SD of each measurement, determined by replicate analyses of two samples, was ±0.004 weight percent (wt%) for OC, ±0.03 wt% for TN, and ±0.03‰ for δ¹³C. Dried sediment (∼0.5 g) was weighed from each sample, and the grain size distribution (0.01 to 2,000 μm) was analyzed using a Mastersizer 2000 diffraction grain sizer (Malvern Instruments Ltd.). Each sample was measured three times, with an average SD of ±0.7 μm. OC, TN, δ¹³C, and grain size values for the NBT were reported in our previous study (38).

We applied the molar ratios of OC and TN contents (OC/TN) and δ¹³C to distinguish between terrigenous and marine OC in hadal trench sediments (35). Marine algae and bacteria are protein rich and have OC/TN ratios of 4 to 10, while vascular land plants are cellulose and lignin rich and have OC/TN ratios of 20 or greater (Fig. 3) (60). We also used δ¹³C to identify the significance of terrigenous OC in marine sediments (35). Because of different carbon sources and biosynthetic pathways, the typical δ¹³C values are ∼−19‰ to −22‰ for marine algae, ∼−27‰ for C₃ plants, ∼−7.2‰ to −10‰ for coastal seagrasses, and ∼−44‰ to −77‰ for chemosynthetic organic matter (61). We applied a two-endmember mixing model to estimate the terrigenous fraction of the sedimentary OC pool in each trench:

$${\delta }^{13}{C}_{\text{sample}}=F_{\text{terr}}\times {\delta }^{13}{C}_{\text{terr}}+F_{\text{mari}}{\delta }^{13}{C}_{\text{mari}},$$ [3]

$$F_{\text{terr}}+F_{\text{mari}}=1 ,$$ [4]

where ${\delta }^{13}{C}_{\text{sample}}$, ${\delta }^{13}{C}_{\text{terr}}$, and ${\delta }^{13}{C}_{\text{mari}}$ are the δ¹³C signals of our samples and endmembers of terrigenous and marine OC, respectively (‰), and $F_{\text{terr}}$ and $F_{\text{mari}}$ are the terrigenous and marine fractions of OC (%), respectively. Following published literature values (38, 61), we used the heaviest δ¹³C values for NBT2 (−19.6‰) and BT10 (−19.2‰) from the oceanward slope as the marine OC endmembers in NBT and BT, respectively. We assumed the heaviest δ¹³C values of cores MT1 (−19.5‰) and MST1 (−19.5‰) to be the marine OC endmembers of the MT and MST, respectively. We selected the average δ¹³C value for OC-rich humus layers from forest soils (−27.8‰) in Papua New Guinea as the terrigenous OC endmember (62).

Data Collection and Hg Burial Rates.

We summarized published data on Hg burial rates in surface sediments and associated Hg concentrations and SRs at 82 sites across the world from peer-reviewed, published literature (SI Appendix, Table S3), following strategies in a previous study (63). Briefly, most of the Hg concentration measurements were conducted using a cold vapor atomic absorption spectrometer. We did not consider data obtained from infrequently used methods for measuring Hg concentrations nor data from studies in which the sampling site of the sediment core was not provided. Additionally, Hg data from heavily Hg-polluted sites (e.g., mining areas or coal-fired power plants) or nearby known anthropogenic point sources (e.g., wastewater releases) were not considered because the Hg concentrations in sediments near these sites may be notably higher, possibly leading to substantial overestimates. Here, we mainly compared data from surface sediments, since sediment cores from different studies may represent different time scales. Most previous studies reported surface sediment data in a 1-cm interval, while a few studies chose a larger interval, such as 2 cm. We assumed that the results of the aforementioned studies could represent contemporary Hg contamination and are comparable to those obtained using a 1-cm interval. Water depths of sediment cores were reported in the literature or calculated using ArcMap software (version 10.7.1) with a high-resolution bathymetric map of the global seafloor (1/60 × 1/60° resolution) created by the National Oceanic and Atmospheric Administration. We calculated the area of each hadal trench based on the bathymetric map in ArcMap. We did not include hadal troughs in the calculation.

We divided the previously published data into three groups, namely, continental shelves (<200 m water depth), the bathyal zone (200 to 2,000 m), and the abyssal zone (2,000 to 6,000 m). For the hadal trench regions, we divided the data into trench axes and surrounding slopes. Some studies have reported a mass accumulation rate (MAR, g ⋅ cm⁻² ⋅ y⁻¹) of sediment. To compare our SRs with those in the published literature, we transformed the MARs into SRs as follows:

$$SR=\frac{\text{MAR}}{D_{\text{average}}},$$ [5]

where $D_{\text{average}}$ is the average dry sediment bulk density reported in the literature. For studies without published data on Hg concentrations, we calculated the concentration as follows:

$${\text{Hg}}_{\text{conc}}=\frac{{\text{Hg}}_{\text{rate}}}{\text{MAR}}\times \frac{1}{10}.$$ [6]

We calculated the burial rate of Hg as follows:

$${\text{Hg}}_{\text{rate}, ij}={\text{Hg}}_{\text{conc}, ij}\times D_i\times S{R}_i\times 10 ,$$ [7]

where ${\text{Hg}}_{\text{rate}, ij}$ is the Hg burial rate in μg ⋅ m⁻² ⋅ y⁻¹ in layer j of core i; ${\text{Hg}}_{\text{conc}, ij}$ is the Hg concentration in ng ⋅ g⁻¹ in layer j of core i; $D_i$ is the dry sediment bulk density in g ⋅ cm⁻³ in core i; and $S{R}_i$ is the sedimentation rate in cm ⋅ y⁻¹ in core i. Average dry bulk densities of 1.3, 1.4, 1.4, and 1.4 g ⋅ cm⁻³ were used for the MT, MST, NBT, and BT samples, respectively. The ²¹⁰Pb_ex data for the NBT and BT were taken from Luo et al. (48), and the data for the MT were obtained from Glud et al. (20). The sediment cores studied in Luo et al. and this study were both collected during the expedition aboard R/V Zhangjian. For NBT and BT, ²¹⁰Pb_ex data show an exponential decrease with depth (R = 0.85 to 0.94), indicating that the SR is relatively constant over the years. Therefore, we applied a constant initial ²¹⁰Pb concentration model for the sediment cores (SI Appendix, Fig. S4) (48). Regressions of ²¹⁰Pb_ex versus core depth provide long-term average SRs of 0.16 cm ⋅ y⁻¹ for core NBT1, 0.08 cm ⋅ y⁻¹ for core NBT3, 0.09 cm ⋅ y⁻¹ for core NBT4, 0.13 cm ⋅ y⁻¹ for core BT6, 0.17 cm ⋅ y⁻¹ for core BT7, and 0.20 cm ⋅ y⁻¹ for core BT10. We assumed the SR for NBT2 to be similar to that for NBT3 because they are both located on the oceanward slope in the NBT. We noted that Hg in the upper layer of NBT1 (0 to 3 cm) and the bottom layer of BT10 (18 to 36 cm) was probably supplied by sporadic geological events such as earthquake turbidites. The former is indicated by the prominently older radiocarbon age of OC in the surface sediment compared to those below (38), as well as the stable Hg isotopic composition, while the latter is indicated by the abrupt change in ²¹⁰Pb_ex activity in the bottom layer. We estimated that the SR for core MT1 (11.43° N, 142.36° E, 10,840 m) is similar to that for another core in the Challenger Deep (0.038 cm ⋅ y⁻¹, 11.34° N, 142.43° E, 10,850 m) investigated by Glud et al. (20). We estimated the chronological framework of sediment cores based on the SRs and the assumption that the deposition year of surface sediment coincided with the sampling year, since the SRs for our sediment cores were relatively constant according to the strong exponential correlations of ²¹⁰Pb_ex data with depth through these short cores. The SR for MST1 was not determined because of the lack of samples, and the SR determined by Glud et al. was assumed to be similar to that for MST1 when estimating the Hg burial flux of trenches remote from the continents (20). The uncertainties of Hg burial rates in different trenches were independent. Thus, we added these uncertainties to the quadrature to represent the aggregated uncertainty.

Statistical Analyses.

All statistical analyses were performed using R software (version 4.0.2, Statistical Environment). No statistical methods were used to determine the sample size. Significance was determined at the *P < 0.05 and **P < 0.01 levels. We used a linear regression model to identify any ancillary parameters that could potentially influence the variation in Hg concentration of each sediment core (SI Appendix, Table S8). We used a multivariate regression model to further examine the possible effects of selected ancillary parameters on the variation in Hg concentration of each core (SI Appendix, Table S6) and to reproduce the concentration. In the multivariate regression model, we fitted each core using the fixed effects of OC + grain size + anthropogenic Hg release. We substituted the depth of the layer for anthropogenic Hg release into the environment (including air, soil, and water). According to the results of the linear regression model, anthropogenic Hg release can better explain the variation in Hg concentrations than the amount of anthropogenic Hg emitted to the atmosphere (SI Appendix, Table S8), which might be due to the release of anthropogenic Hg from contaminated soil and water (13, 64). Because of a lack of regional historical estimates of anthropogenic Hg release and the difficulty of identifying the anthropogenic source of Hg for each trench region, a global estimate was used (7). The modeling impact of anthropogenic Hg release is conservative because of the lack of regional historical estimates of the release flux, and further regional historical estimates of anthropogenic Hg release are desirable. Despite this uncertainty, our modeling exercise confirmed that anthropogenic sources strongly influence the vertical variations in Hg concentrations across all trench regions and shows that this assumption is reasonable. Overall, the multivariate regression model allowed us to successfully reproduce 83% of the variations in Hg concentrations in trench regions (SI Appendix, Fig. S5), which verifies that the current modeling method is reasonable.

A linear mixed-effect regression model was fitted to the Hg concentration across cores using the lme4 package (SI Appendix, Table S6) (65). We treated each sediment core as a random effect to identify any unique variances associated with localized geological properties related to Hg concentration (66). R²_m and R²_c values, defined as marginal (fixed effect) and conditional (random and fixed effects) R² values, respectively, and calculated using the MuMIn package, were reported as goodness-of-fit statistics. The ancillary parameters were tested using the ANOVA Satterthwaite approximation of degrees of freedom in the lmerTest package. We reported the F-value from the F test, which is the mean square of each independent variable divided by the mean square of the residuals, to interpret the relative importance of an independent variable to the variation in the Hg concentration. The Hg concentration and ancillary parameters were highly and positively skewed. Thus, we applied the Mann–Whitney U test to compare the difference in each variable between the two groups at a significance level of *P < 0.05.

Data Availability

All study data are included in the article and/or SI Appendix.

Note Added in Proof.

We submitted this manuscript on February 8, 2021. During the review process, Sanei et al. published an article entitled “High mercury accumulation in deep-ocean hadal sediments” on May 26, 2021 (42). They did a remarkable job reporting the high Hg accumulation in two hadal trenches from the Southern Hemisphere, which largely agrees with one of our conclusions. Based on the multiple geochemical measurements and statistical analyses on four different trenches from both Northern and Southern Hemispheres, we further illustrate the mechanism of high Hg accumulation and highlight the impacts of terrigenous input and anthropogenic Hg releases. More investigations on the hadal zone might help better comprehend the role of hadal trenches in the global Hg cycle.