In 1997, sediment core 1076A was collected from the Lower Congo Basin (5° 4.132′S, 11° 6.092′E, 204.3-m-long sediment core, 1,404.2 m water depth) during Ocean Drilling Program (ODP) Expedition 175 using an advanced hydraulic piston corer (Figure 1). The core falls within the Angola–Benguela Current system, which is one of the great upwelling regions in the world (Wefer et al., 1998). This region is also influenced by freshwater input from the Congo River. The sediment along the core comprises bioturbated organic carbon-rich olive-gray clay and greenish-gray clay. Biogenic sections contain abundant diatoms with rare nanofossils, silicoflagellates, siliceous sponge spicules, phytoliths, and traces of radiolarian and foraminifer fragments. The core was sliced by 5 cm³ plastic scoops into 1 cm slices. Samples from the upper 1 m of the core were collected at ∼1 cm intervals, resulting in a total of 99 samples, representing the past ∼5,000 years (Wefer et al., 1998).

Sediment core 1228B was collected from the continental shelf of Peru (11° 3.862′S, 78° 4.664′W, 55.3-m-long sediment core, 262.3 m water depth) in 2002 during ODP Expedition 201 using a hydraulic piston corer (Figure 1). Site 1228 lies within the Peru upwelling zone. Winds along the coast of Peru maintain the upwelling process throughout the year. Strong upwelling of cold, nutrient-rich subsurface water causes extensive phytoplankton blooms and high productivity in this region, which is known as the Peruvian high-productivity upwelling system (Tarazona and Arntz, 2001; D’Hondt et al., 2003). The continental shelf of Peru has been dominated by upwelling over the last 40 kyr. Sediments are predominantly biogenic sediments and silt and clay-sized, predominantly siliciclastic sediments. Large, fast-growing diatoms dominate the phytoplankton biomass (De Mendiola, 1981; Scheidegger and Krissek, 1983; DiTullio et al., 2005). Samples from the upper 2.70 m of the core were collected at ∼10 cm intervals, resulting in a total of 27 samples, representing the past ∼4,300 years (D’Hondt et al., 2003).

Sediment core 941A was collected from the western side of the Amazon Fan, at the mouth of the Amazon River (5°22.384′N, 48°1.723′W, 177.66-m-long sediment core, 3,380.9 m water depth), in 1994 during ODP Expedition 155. The fan contains much material eroded from the continent within the Amazon drainage basin and fluvial sediments (Flood et al., 1995). The sediments consist of bioturbated nanofossil-foraminifera clay and carbonate. Samples from the upper 1 m of the core were collected at ∼1 cm intervals, resulting in a total of 93 samples, representing the past ∼19,400 years (Flood et al., 1995).

Total carbon (C), nitrogen (N), and sulfur (S) were analyzed using a EuroEA 3000 elemental analyzer (Hekatech GmbH, Germany). Carbonate in the samples was removed with HCl acid. The quality of the analysis was ensured by certified reference materials (CRMs), i.e., sulphanilamide (C = 41.75 ± 0.17%; N = 16.26 ± 0.22%; S = 18.64 ± 0.18%), BBOT 2.5-Bis(5-tert-butyl-benzoxazol-2-yl)thiophene (C = 72.52%; N = 6.51%; S = 7.44%), Canmet LKSD-4 lake sediment (C = 17.7 ± 0.8%; S = 0.99 ± 0.09%), and MOC soil standard (C = 3.19 ± 0.07%). The average results for the CRMs were sulphanilamide (n = 31) C = 42.0 ± 0.25%, N = 16.2 ± 1.01%, and S = 18.5 ± 0.54%; BBOT (n = 6) C = 72.2 ± 0.12%, N = 6.55 ± 0.04%, and S = 7.41 ± 0.03%; LKSD-4 (n = 5) C = 19.1 ± 0.05% and S = 1 ± 0.01%; and MOC (n = 6) C = 3.23 ± 0.02%.

Total Hg was determined by thermal decomposition followed by preconcentration of Hg on a gold trap and cold vapor atomic absorption spectrometry (CVAAS) Hg detection using a Milestone DMA-80 analyzer (EPA, 1998). The quality of the analysis was ensured by including a CRM (Canmet LKSD-4 = 190 ± 17 ng g^–1) alongside the analyzed samples. The average measured concentration for LKSD-4 (n = 37) was 190 ± 15 ng g^–1. The results of replicate analyses (n = 27) were always within a relative standard deviation (RSD) of 10% of the certified value.

The samples were analyzed for concentrations of silicon (Si), titanium (Ti), zirconium (Zr), sulfur (S), calcium (Ca), potassium (K), aluminum (Al), yttrium (Y), manganese (Mn), strontium (Sr), iron (Fe), lead (Pb), copper (Cu), zinc (Zn), arsenic (As), bromine (Br), nickel (Ni), chlorine (Cl), and rubidium (Rb) by energy dispersive X-ray fluorescence (ED-XRF). The methods used for calibration and accuracy and precision determination are described in detail in Cheburkin and Shotyk (1996). The CRMs (Canmet LKSD-4, NRC/CNRC-PACS-2, NRC/CNR-Mess-3, and NCS-DC75304) and replicates were measured in each set of samples for accuracy and precision control. Repeated analysis of CRMs gave an RSD of less than 10% for Si, K, Ca, Ti, Mn, Fe, Zn, Br, Rb, Sr, Y, and Zr; 8 –17% for Ni, Cu, Cl, and S; 6–15% for Al and Ba; 4–12% for Pb; and 9–20% for As.

The correlation matrix of elements and concentration profiles of Si, Zr, Al, K, Ti, Br, S, Sr, Ca, C, and Zn and C/N ratios are shown in Supplementary Table 1 and Figure 2, respectively. Si has the highest concentration of all elements in the sediments of the Congo Basin. On the one hand, Si is associated with the flux of terrestrially derived mineral components and, on the other hand, with siliceous phytoplankton, protozoans, protists, plant phytoliths, and sponge spicules (Croudace and Rothwell, 2015). The record of Si concentrations does not show a specific trend and varies between 13.3 and 20.7%, with a median of 15.9%. The records of Al, Ti, Cl, K, S, Mn, and Ca follow the Si concentration trend. Concentrations of Al, Ti, Cl, K, S, and Ca range between ∼3.24 and 8.89%, ∼3,136 and 6,432 mg kg^–1, 1.24 and 2.89%, ∼0.78 and 1.59%, ∼0.34 and 0.80%, and ∼0.59 and 0.91%, respectively. The positive covariance of Al and Ti, as lithogenic elements, with the abovementioned elements reveals that fluctuations in these elemental concentrations in Congo Basin sediments are likely related to changes in terrestrial inputs. This can be due to fluctuations in Congo River discharge. The records of Ni, Fe, Cu, Zn, Br, and Sr concentrations show similar trends throughout the core (Figure 2). Ni, Fe, Cu, Zn, Br, and Sr concentrations vary between ∼31.8 and 75.8 mg kg^–1, ∼6.65 and 13.1%, ∼19.7 and 48.3 mg kg^–1, ∼92 and 169 mg kg^–1, ∼167 and 248 mg kg^–1, and ∼87.3 and 105 mg kg^–1, respectively. The covariation of Fe, Ni, Cu, and Zn, as biologically essential elements (Morel and Price, 2003), suggests that fluctuations of these elements are mainly associated with the biological production and settling of biogenic debris. In the Congo Basin, biological productivity is mainly the result of upwelling (Wefer et al., 1998). The organic C concentration varies between ∼2.76 and 4.01%, with a median of 3.23%. This concentration is higher than that in most deep-sea sediments from the South Atlantic (∼0.3 wt%) (Premuzic et al., 1982; Keswani et al., 1984) and reflects a history of elevated primary production in this area caused by upwelling (Wefer et al., 1998). C/N ratios vary by a factor of ∼1.27 between 11.5 and 14.6 at a 100–56 cm depth and by a factor of ∼2.13 between 6.46 and 13.8 at a 50–1 cm depth of the core, with five more pronounced peaks at approximately 53, 51, 46, 44, and 43 cm depths. A shift in the C/N ratio from relatively higher values (median: 13.2) in the lower ∼50 cm of depth to lower values (median: 10.4) in the upper ∼50 cm depth of the core indicates a change in the source of organic matter from terrestrial sources (higher C/N ratios) to marine sources (lower C/N ratios). This change is likely attributable to the change in environmental conditions and increasing input from the Congo River in the lower 50 cm depth of the core. Stronger freshwater discharge from the Congo River is a result of a more intense monsoon over Central Africa and enhanced precipitation (Uliana et al., 2002).

The PCA for the Congo Basin data set resulted in five components (Table 1 and Supplementary Figure 1), explaining almost 82% of the total variance. The first component (Cp1) explains 33% of the variance and shows large positive loadings for Si, Ti, Al, K, Cl, S, and Mn. The second component (Cp2), which explains 21% of the variance, is characterized by large positive loadings for Fe, Cu, Ni, Sr, Zn, and Br. The third component (Cp3) explains 16% of the variance and shows large positive loadings for Hg, N, Ca, and Mn and negative loadings for Rb and Zr. The fourth and fifth components (Cp4 and Cp5) account for 6 and 5% of the variance, respectively. Cp4 is characterized by positive loadings for C, and Cp5 shows positive loadings for Pb. In general, the PCA results imply that Congo Basin sediments are formed by material from terrigenous (Congo River) and biogenic (Benguela upwelling) sources.

Briefly, Cp1, which mainly includes positive loadings of lithogenic elements, represents the variability of terrigenous and fluvial inputs from the Congo River. Cp2 mainly comprises loadings of elements of biogenic sources and appears to reflect biogenic productivity. Cp3 appears to reflect biogenic productivity, mainly calcite-producing microorganisms, as a result of upwelling. Increased biogenic fluxes are accompanied by relative dilution of the terrigenous content of the sediments. This is reflected by the negative loadings for Rb and Zr, as indicators of changes in the flux of lithogenic materials, in Cp3. Hg forms a group along Cp3 together with positive loadings for biogenic sources and negative loadings for terrigenous sources. Cp4, which includes positive loading for C, represents the variability of primary productivity as well.

The correlation matrix of elements and concentration profiles of Si, Zr, Al, K, Ti, Br, S, Sr, Ca, C, and Zn and C/N ratios are shown in Supplementary Table 2 and Figure 3, respectively. Similar to that in the Congo Basin, Si has the highest concentration of all elements in the sediments of the Peru Margin. Si can be detrital, derived from the weathering of crustal rocks, or biogenic, derived from siliceous phytoplankton (Croudace and Rothwell, 2015). The record of Si concentrations shows an increasing trend from the top to the bottom of the core and varies between 14.6 and 27.4%, with a median of 18.3%. The records of K, Zr, Ti, Al, and Y concentrations generally follow the Si concentration trend. Concentrations of K, Zr, Ti, Al, and Y range between ∼0.52 and 1.47%, ∼41.7 and 219 mg kg^–1, ∼1,219 and 3,057 mg kg^–1, ∼2.27 and 4.41%, and ∼11.6 and 24.5 mg kg^–1, respectively. Zr, as a conservative lithogenic element, is an indicator of changes in the flux of terrestrial material. The positive correlations of Zr with the abovementioned elements, therefore, indicate that they are associated with terrestrial inputs as well. Organic C was found to be another major component in these sediments and is associated with the Peruvian high-productivity upwelling system. Concentrations vary between ∼2.58 and 15.9%, with a median of 13.1%. The organic C concentrations record shows a decreasing trend from the top to the bottom of the core. Sea level fluctuations and the associated shift of bottom circulation and the oxygen minimum layer can be the main reasons for the decreasing trend in organic C (Suess and von Huene, 1988). The C/N ratios in the Peru Margin sediments vary between 8.89 and 10.5, with 2 more pronounced peaks at depths of approximately 200 and 190 cm in the core. Low C/N ratios indicate that marine organic matter is the source of organic matter in the Peru Margin sediments. This is a result of high primary productivity in the surface water driven by upwelling. The variability in the C/N ratio record is generally low in Peru Margin sediment (Figure 3), pointing to non-significant changes in terms of the dominant source of organic matter. The records of Ni, Br, Cu, Zn, and Cl concentrations show a similar trend to the organic C concentration. Ni, Cu, and Zn are known to have a biological role (Morel and Price, 2003). Concentrations of Ni, Cu, and Zn range between ∼12.8 and 180 mg kg^–1, ∼11 and 95 mg kg^–1, and ∼65.6 and 182 mg kg^–1, respectively. Br is also associated with marine organic matter and related productivity (Ziegler et al., 2008, 2009; Ren et al., 2009; Caley et al., 2011). The concentration of Br ranges between ∼45 and 1,577 mg kg^–1. Cl can travel through biological pathways (incorporation into algae) and reach the sediments by fast-sinking detritus (Leri et al., 2015). Concentrations vary between ∼1.14 and 5.21%. Correlations of the abovementioned elements with organic C reflect the same depositional process, which is accumulation by organic matter flux from surface waters.

The PCA results further helped explain the processes that affected Peru Margin sedimentation. The PCA yielded five components (Table 2 and Supplementary Figure 2), explaining almost 92% of the total variance. Cp1 explains 52% of the variance and shows large positive loadings for Cl, S, Cu, Ni, N, and C and negative loadings for Y and Zr. Cp2, which explains 18% of the variance, is characterized by large positive loadings for Hg, Rb, Br, N, Zn, and C and negative loading for Si. Cp3 explains 12% of the variance and shows large positive loadings for Si, Ti, K, Mn, and Al. Cp4 and Cp5 account for 7 and 4% of the variance, respectively. Cp4 is characterized by positive loadings for Ca and Sr and negative loading for Pb. Cp5 shows positive loadings for Fe and As.

The PCA results imply that mainly terrigenous material input, upwelling productivity, and associated particle export to the seafloor affected Peru Margin sedimentation. Briefly, Cp1 and Cp2, which include positive loadings for elements associated with biogenic productivity, appear to reflect the productivity in the upwelling system and associated particle export to the seafloor. The covariation of Ni and Cu with organic C along Cp2 implies that for this component, the main removal mechanism is settling with organic matter (Böning et al., 2015). Increased biogenic fluxes are accompanied by dilution of the terrestrial matter content, which is reflected by the negative loadings for Y and Zr along Cp1 and Si along Cp2. The variance of Hg was captured by this component. Cp3, which includes positive loadings for lithogenic elements, represents the variability of terrigenous inputs. The Peru Margin site, unlike our two other sites, i.e., the Congo Basin and Amazon Fan, is not affected by riverine discharge (D’Hondt et al., 2003). Sources of terrigenous sediments to the Peru Margin are controlled primarily by continental supplies (Jaisi and Blake, 2010). The covariation of Ca and Sr in Cp4 represents marine productivity, mainly that of foraminifers and coccolithophorids. Cp1 and Cp2 account for much of the variability/process that controlled the geochemical composition of these sediments. This information, along with knowing that the Peru Margin site is a major upwelling site and receives a very small amount of riverine discharge, suggests that the contribution of terrigenous inputs is small.

The correlation matrix of elements and concentration profiles of Si, Zr, Al, K, Ti, Br, S, Sr, Ca, C, and Zn and C/N ratios are shown in Supplementary Table 3 and Figure 4, respectively. Si has the highest concentration of all elements in the sediments from the Amazon Fan. The record of Si concentrations shows no specific trend with depth and varies between 17.3 and 23%, with a median of 20.5%. The records of Ti, Y, Rb, Zn, and Fe concentrations generally show a similar tendency and increase from the top to the bottom of the core. Concentrations of Ti, Y, Rb, Zn, and Fe range between ∼2,871 and 5,129 mg kg^–1, ∼16.7 and 26.2 mg kg^–1, ∼75.2 and 143 mg kg^–1, ∼62.8 and 146 mg kg^–1, and ∼2.21 and 8.24%, respectively. Covariations of Ti, as a proxy for the changes in the flux of lithogenic material, with the mentioned elements indicate an association of these elements with terrigenous inputs from the catchment of the Amazon River. Sr concentrations fluctuate by a factor of approximately two between ∼200 and 317 mg kg^–1 within a 97–70 cm depth range of the core. At a depth of 70 cm, the concentration increases by a factor of approximately 4 and reaches 223 to 1,033 mg kg^–1. Sr concentrations fluctuate by a factor of approximately 1.7 between 622 and 1,041 mg kg^–1 from a 42 cm depth to the top of the core. The record of Ca shows a similar pattern to the record of Sr. Ca concentrations fluctuate by a factor of approximately 2 between ∼1.96 and 4.09% within a 97–70 cm depth range. At a depth of 70 cm, the Ca concentration increases by a factor of approximately 7 and reaches 2.67–18.7%. Ca concentrations fluctuate by a factor of approximately 1.8 between 10.1 and 18.6% from 42 cm to the top of the core. Hence, Sr is a proxy for biogenic origin, covariation of Ca with Sr reflecting the fact that Ca is mainly biogenic and that they are associated with calcite-producing microorganisms. Variations in Ca and Sr reflect dilution by terrigenous material rather than productivity changes. This is indicated by the negative correlations of these two elements with lithogenic elements. Varying fluxes of terrestrial material resulted in the dilution of calcite organisms in the sediments. After Si, Ca has the highest average concentration and is found to be another major component in the sediments from the Amazon Fan. This can be associated with foraminifers (Flood et al., 1995) and indicates that calcite-producing microorganisms contribute significantly to the Amazon Fan sediments. Organic C concentrations vary between ∼0.51 and 0.94%, with a median of 0.64%. The average organic C concentrations in the Amazon Fan are a factor of 5 and 17 times lower than those in the Congo Basin and Peru Margin, respectively. The variability in C/N ratios is generally low. The ratios vary by a factor of ∼1.66 between 8.33 and 13.8, with a median of 10.1. Concentrations of biologically essential elements such as Zn, Ni, Cu, and Fe vary between ∼62.8 and 146 mg kg^–1, ∼14.4 and 46.8 mg kg^–1, ∼15.6 and 54.1 mg kg^–1, and ∼2.21 and 8.24%, respectively.

The PCA yielded five components (Table 3 and Supplementary Figure 3), explaining almost 83% of the total variance. Cp1 explains 44 % of the variance and shows large positive loadings for Rb, Y, Pb, Zn, Ti, Fe, Hg, and Ni and negative loadings for Ca, Sr, and Al. Cp2, which explains 14% of the variance, is characterized by large positive loadings for K, Si, Cl, and S. Cp3 explains 11% of the variance and shows large positive loadings for Ni, Mn, and Cu. Cp4 and Cp5 account for 8 and 6 % of the variance, respectively. Cp4 is characterized by positive loadings for C, N, and Br. Cp5 shows positive loadings for As and Zr.

Based on the age models and estimated sedimentation rate of 20 cm kyr^–1 for the Congo Basin (Wefer et al., 1998), sediments in the top 98 cm of the core correspond to approximately 5,000 years.

In this period, Hg concentrations range between 67.2 and 139 μg kg^–1, with a median of 93.4 μg kg^–1. Hg concentrations fluctuate between 84 and 92.5 μg kg^–1 within a 98–88 cm depth range of the core (Figure 2). At 87 cm, the Hg concentration decreases by a factor of approximately 1.2 and reaches 71 μg kg^–1. At this depth of the core, the Hg record shows an increasing trend to the top of the core and reaches 114 μg kg^–1 at a core depth of 38 cm. At a core depth of 37 cm, the Hg concentration decreases by a factor of approximately 1.1 and reaches 101 μg kg^–1. Hg concentrations fluctuate between 92.2 and 113 μg kg^–1 within a 37–1 cm depth range of the core and reach a maximum value of 139 μg kg^–1 at the very top of the core. The estimated average Hg accumulation rate in the Congo Basin is ∼46.7 μg m^–2 yr^–1.

To identify driving forces of the variations in Hg accumulation, we used the regression analysis of the C/N ratios and Hg concentrations (Supplementary Figure 4), our PCA results (Table 1), and the correlation matrix of elements (Supplementary Table 1). The variance in Hg in the Congo Basin is captured by Cp3. We interpreted Cp3 as representing the variability of biogenic fluxes from upwelling productivity. This is indicated by positive loadings for N and Ca and negative loadings for Rb and Zr. PCA demonstrated that upwelling and enhanced marine productivity, mainly of foraminifers and coccolithophorids (Croudace and Rothwell, 2015), controlled the flux of Hg to Congo Basin sediments. PCA further revealed that Hg shows no positive covariance with lithogenic elements (e.g., Al, Ti, and K along Cp1). This indicates that terrestrial influence, associated with the supply by the Congo River, on Hg accumulation in these sediments is of minor importance.

Comparison of the Hg concentration record with the C/N ratio record indicates that in the lower 50 cm of the core, where C/N ratios are higher (median: 13.2), Hg concentrations are lower (median: 80.9 μg kg^–1), and in the upper 50 cm of the core, where C/N ratios are lower (median: 10.3), Hg concentrations are higher (median: 101 μg kg^–1). As mentioned above, the lower C/N ratios indicate a marine source for the organic matter.

No positive covariance of Hg with lithogenic elements, together with the information regarding the organic matter source, which was revealed from C/N ratios, and their relations with Hg concentrations, indicate that atmospheric Hg deposition and binding of atmospheric-derived Hg to marine-derived organic matter is the main source of Hg in Congo Basin sediments. The upwelling-related productivity in the Congo Basin increases Hg scavenging by organic particles and Hg export to the sediment.

Considering a sedimentation rate of 85 cm kyr^–1 for the Peru Margin, i.e., based on radiocarbon dating (D’Hondt et al., 2003), sediments in the top 270 cm of the core likely correspond to the past 4,300 years.

The Hg record at this time shows an increasing trend from the bottom to the top of the core. Hg concentrations range between 49 and 172 μg kg^–1, with a median of 90.7 μg kg^–1. The Hg concentration reaches a maximum value of 172 μg kg^–1 at a depth of 20 cm (Figure 3). The average Hg accumulation rate in Peru Margin is ∼178 μg m^–2 yr^–1.

The results of the regression analysis of the C/N ratios and Hg concentrations (Supplementary Figure 5), the PCA (Table 2), and the correlation matrix of elements (Supplementary Table 2) helped identify the processes that controlled Hg sedimentation in the Peru Margin. Hg forms a group with high factor loadings along Cp2 together with Ni, Rb, Br, N, Zn, and C. The covariation of Hg with Ni, Cu, and organic C implies that the main removal mechanism of Hg in the Peru Margin is the sinking of plankton particles derived from productive surface waters of the Peru Margin. Hg shows no positive covariance with lithogenic elements (e.g., Al, Ti, and Zr). This indicates that terrestrial influences on Hg accumulation are negligible.

The C/N ratio record and its comparison with the Hg concentration record suggest that, similar to the pattern observed for Congo Basin sediment, the source of organic matter can influence Hg sedimentation. Low C/N ratios in the Peru Margin sediments, which point to a marine source for the organic matter, result from the settling of algae as a result of upwelling. Periods of low Hg concentrations, at approximately 200 and 190 cm in the core (Figure 3), correspond to phases of relatively high C/N ratios and relatively low marine organic matter. Similar to our conclusion for the Congo Basin, we conclude that the binding of atmospheric-derived Hg to sinking particles is the dominant process of Hg accumulation in these sediments.

Based on a sedimentation rate of 5 cm kyr^–1 for the Amazon Fan (Flood et al., 1995), sediments in the top 97 cm of the core likely correspond to the past 19,400 years.

During this period, the Hg record shows no clear pattern with depth and ranges between 16.4 and 54 μg kg^–1, with a median of 35.8 μg kg^–1 (Figure 4). The Hg concentration decreases from 40 to 20.7 μg kg^–1 within a 97–89 cm depth range of the core. The Hg concentration increases at an 89 cm depth in the core and fluctuates between 27.3 and 54 μg kg^–1 within an 88–47 cm depth range of the core. At a depth of 47 cm, the Hg concentration decreases by a factor of ∼2.9 and reaches 18.4 μg kg^–1. The Hg record then shows a general increasing trend to the top of the core and ranges between 16.4 and 52 μg kg^–1. The average Hg accumulation rate in the Amazon Fan is ∼4.52 μg m^–2 yr^–1.

To identify driving forces of the variations in Hg accumulation, we used the regression analysis of the C/N ratios and Hg concentrations (Supplementary Figure 6), our PCA results (Table 3), and the correlation matrix of elements (Supplementary Table 3). PCA results indicate that the variance of Hg in the Amazon Fan was captured by Cp1. Rb, Y, Pb, Zn, Ti, Fe, and Ni show positive factor loadings as well, whereas Ca, Sr, and Al show negative factor loadings. The covariation of Hg with terrigenous elements implies that Hg fluxes in the Amazon Fan are significantly influenced by changes in lithogenic inputs. Hg and Ca [Ca is attributed to biogenic carbonate in the Amazon Fan sediments (see section “Geochemical Processes Controlling the Distribution of Elements in the Sediments, Amazon Fan”)] are negatively correlated (Supplementary Table 3). We hypothesize that this is due to the dilution effect rather than the negative covariation of Hg with groups of marine plankton such as foraminifera and coccolithophorids. The records of Hg concentrations in Amazon Fan sediments are not consistent with variations in the C/N ratio. Therefore, we cannot relate changes in Hg concentration to the source of organic matter.