Senegalese artisanal gold mining leads to elevated total mercury and methylmercury concentrations in soils, sediments, and rivers

Mercury (Hg) – predominantly in the form of methylmercury (MeHg) – is a potent neurotoxin that can significantly impair human health [Driscoll et al., 2013]. Prolonged exposure increases the risk for brain and neurological damage, which is especially harmful for pregnant women and children [Harada, 1995; Grandjean et al., 1997; Selin, 2014]. Although there are many potential sources of Hg, direct inhalation from anthropogenic emissions and consumption of fish constitute the major pathways of human exposure. Even when Hg concentrations are low in water, Hg concentrations can reach dangerously high levels in fish due to bioaccumulation and biomagnification [Morel et al., 1998; Driscoll et al., 2007]. Controlling Hg contamination in the environment has been a primary target of local and national health and environmental agendas for decades and, most recently, an issue of international concern addressed by the 2013 Minamata Convention on Mercury, which came into force in August 2017 [UNEP, 2013; Selin, 2014].

The largest source of Hg to the environment is artisanal and small-scale gold mining (ASGM), which releases 1,400 tons of Hg annually into the air, soil, and water [Schmidt, 2012] and accounts for an estimated 37% of global anthropogenic Hg emissions [Telmer and Veiga, 2009; UNEP, 2013]. In the process of ASGM, liquid elemental Hg (Hg⁰) is utilized to separate gold from sediments. After the formation of a Hg-gold amalgam, the mixture is heated to create pellets of gold, while the Hg is released into the atmosphere or deposited as mining waste directly into soil and water. Despite the health effects, many of which are unknown to ASGM communities, Hg is used in this process because it is inexpensive to obtain, easy to use, does not require time-consuming processing, and results in a high gold recovery rate [Telmer and Veiga, 2009]. For example, in Senegal, the cost of liquid Hg⁰ is approximately 0.40 USD per gram, while gold sells for approximately 32 USD per gram.

It is estimated that more than 10 million people, including over 3 million women and children, work in ASGM across more than 70 developing countries in Asia, South America, and sub-Saharan Africa. Families directly involved in the mining are exposed daily to levels of Hg vapor that far exceed the World Health Organization’s guidelines [Cordy et al., 2011], while communities located downstream and downwind consistently consume Hg-contaminated fish [Arifin et al., 2015; Diringer et al., 2015; Ha et al., 2017]. Mercury contamination has a legacy impact; since it cannot be degraded, it persists perpetually within the environment, undergoing localized transformations, and impacting future generations [Thomas et al., 2002; Veiga and Hinton, 2002].

Despite its pervasive use in the extraction of gold, little is known about the fate of Hg from local ASGM waste including the extent of contamination, its pathway through Hg environmental media, and uptake by aquatic organisms. An understanding of Hg biogeochemistry in these ecosystems is vital to the creation and implementation of national goals to reduce Hg contamination, as mandated by the Minamata Convention on Mercury [UNEP, 2013]. In fact, Article 7 of the Convention emphasizes the need for more scientific studies of ASGM-derived Hg in the environment.

Previous studies on ASGM have found elevated concentrations of total Hg (THg) and MeHg in soils at ASGM sites compared to reference sites removed from ASGM activities [Appleton et al., 1999; Carling et al., 2013; Nyanza et al., 2014]. Additionally, many studies have observed elevated Hg concentrations in downstream water and sediments, with patterns of decreasing THg concentration with distance from ASGM communities [Gray, 2002a; Nartey et al., 2011; Yin et al., 2013; Diringer et al., 2015]. These findings provide strong evidence that Hg utilized in ASGM is entering the environment directly from mining tailings as well as indirectly through local atmospheric deposition (predominantly as inorganic Hg) of the recently emitted Hg.

While many studies have investigated Hg use in tropical regions such as the Amazon, particularly in relation to the naturally Hg-rich soils [e.g., Cordy et al., 2011; Terán-Mita et al., 2013; Balzino et al., 2015], little quantitative work has been conducted in an environment with the climatic, edaphic, and geologic conditions of West Africa where crude technology and approaches are utilized to extract the gold [but see Taylor et al., 2005a; Niane et al., 2014; Rajaee et al., 2015]. Despite the prevalence of gold mining in Senegal, to our knowledge only one study has quantified Hg from ASGM in Senegal in environmental media [Niane et al., 2014]. Differences in edaphic, aeolian, and climatic conditions in West Africa compared to other locations can influence patterns and pathways of atmospheric Hg deposition, runoff and fluvial transport of Hg compounds, and conditions for the conversion of inorganic Hg to MeHg – the more bioaccumulative form. While the few studies of ASGM in Africa (mostly eastern Africa) have affirmed that Hg concentrations are high near mining sites [e.g., Taylor et al., 2005; Rajaee et al., 2015], research is critically needed that better characterizes the spatial extent and magnitude of Hg contamination.

In this study, we evaluated THg and MeHg concentrations in soil, sediment, and water samples collected from four ASGM communities in Senegal, West Africa. We compared THg and MeHg concentrations in surface soil samples collected within the huts where ASGM miners burn the Hg-gold amalgam (hereafter referred to as burning huts) to soils collected along a transect connecting each hut to the closest segment of the adjacent river, and to sediments collected alongside (riparian zone) and within the river channels. We also analyzed THg and MeHg concentrations in water samples collected from each downstream river reach and compared these levels to those measured in a river segment well upstream of any ASGM mining, to purified drinking water sold in a mining village, and to water collected from a drinking well in a mining village. Our goal was to ask three simple but important questions about the magnitude and spatial extent of Hg contamination in the region: (1) Are high THg and MeHg concentrations in environmental media limited to the vicinity of the burning hut in ASGM areas? (2) Do THg and MeHg concentrations decline predictably with distance from the burning huts? and (3) What is the distribution of THg and MeHg across various environmental media in mining communities?

Artisanal and small-scale gold mining in Senegal occurs in the southeastern region of Kedougou. The Senegalese Sabodala Deposit is the largest deposit in West Africa, extending over a 230 km² region and estimated to contain over 30 tons of gold [Savornin et al., 2007; Niane et al., 2015]. The geological composition of the Sabodala Deposit represents that of the Greenstone Gold Belt, which extends further into West Africa and contains over 400 tons of gold [Bassot, 1997]. The gold deposits in Senegal are mined predominately by artisanal methods (Figure 1) using gravity filtration, with an estimated 1–2 Mg/year of Hg used in ASGM in 2007 [Telmer and Veiga, 2009]. It has been suggested that Hg in Senegal originates from Hg mines in Europe [Veiga et al., 2006; COMTRADE, 2016].

Kedougou is located in the Sudanian zone and receives more rainfall than the rest of the country [Fall et al., 2006]. The region has a monsoonal climate, with heavy rains beginning in May and continuing until October. Average annual rainfall is approximately 1000 mm [DGPRE, 2011]. Mining occurs predominantly in the dry season from November to May, with mining activities decreasing drastically once the rainy season begins. Temperatures in Kedougou peak at the end of the dry season in May (average of 36°C) and reach a minimum at the end of the wet season in October (average of 23°C) [DGPRE, 2011]. Dominant vegetation includes tropic semi-deciduous lowland forest, drought-deciduous lowland woodland, flat-leaved savanna with isolated palms and deciduous trees, narrow-leaved savanna with isolated deciduous trees, and evergreen woody vegetation [Pruetz et al., 2002].

Gold mining has occurred in Senegal since the 1970s, but evidence suggests that ASGM activities have intensified since 2000 [Savornin et al., 2007; Persaud et al., 2017]. With 71% of the population in the Kedougou region living below the poverty line (<1 USD per day), the average of 4–7 USD earned per day from ASGM has popularized the activity [Persaud et al., 2017]. In 2007, it was estimated that 10,000 people in Senegal were working as miners and that at least 30,000 additional people lived in close proximity to ASGM activities or had family members who worked in ASGM [Savornin et al., 2007]; this amounted to 20% of the Kedougou regional population mining and 50% indirectly impacted by ASGM [WTO, 2009; Boyer, 2011]. ASGM has increased dramatically since 2007, now estimated to involve 77.5% of the regional population and hundreds of villages [Doucouré, 2014; Persaud et al., 2017] and resulting in a 10-fold increase in gold production [Reichl et al., 2017]. Note that the number of people involved in ASGM activities at each village varies seasonally and annually; more people are involved in the dry season and when the village produced a high total gold extraction the previous year.

Samples were collected from four mining villages (Bantako, Kharahenna, Kolya, and Sabodala) and one reference village (Saraya; Figure 2). The reference village was located in a non-ASGM watershed with similar forest cover and precipitation patterns to the mining villages. Using ArcGIS Version 10.4.1, a digital elevation model (DEM) with 30 arc second resolution obtained from ArcGIS online and hydrologic data obtained from ArcGIS online [ESRI, 2016], the watershed area was calculated for each ASGM area [Sarangi et al., 2003]. Watershed sizes ranged from small for Sabodala and Kharahenna (106 km² and 160 km², respectively) to large for Bantako and Kolya (7070 km² and 8800 km², respectively).

All villages studied were located next to a river to allow collection of terrestrial and riverine samples. River flow was highest in Bantako and lowest in Kharahenna. Soil samples were collected from the living room, front yard, and backyard of the burning hut in Bantako, Kharahenna, and Sabodala, as well as along a transect to the river and in the riparian zone in all the mining villages. Samples taken along the transect are representative of surficial flowpaths. Note that soil samples were not taken from a burning hut in Kolya due to inability to identify one, although burning huts do occur in Kolya and were likely located equidistant along the transect (e.g., 200 m from the first transect sample). Soil samples collected along the transect were spaced equally. In Kolya, the transect was 800 m; in Bantako, the transect was 1200 m; in Kharahenna, the transect was 2500 m; and in Sabodala, the transect was 8000 m. Sediment samples were collected from the riverbed and water samples were taken from the river. Three additional water samples were collected from an upstream site in Bantako, a drinking well in Kharahenna, and a sealed bag of clean drinking water sold in Kharahenna. In total, 123 soil, 22 sediment, and 29 water samples were collected.

Samples were collected in May 2016, before and after the first heavy monsoonal rains which fell on May 27. All soil, sediment, and water samples were collected using clean hands-dirty hands protocol (EPA Method 1669). Duplicate surficial (0–3 cm) soil and sediment samples were collected at each site and placed into a sealed ziplock plastic bag, which was then sealed in a second plastic ziplock bag and immediately placed on ice. Samples were frozen within 24 hours of collection and transported and stored frozen until analysis. Water samples were collected in new polyethylene terephthalate copolyester glycol (PEGT) bottles, and a field blank was taken. All water samples were acidified to 0.4% using trace metal grade hydrochloric acid (HCl), doublebagged, transported cold to Duke University, and stored at 4°C until analysis.

Soils and sediments were lyophilized for five days, homogenized, and analyzed for THg on a Milestone Direct Mercury Analyzer (DMA-80) via thermal decomposition, catalytic reduction, amalgamation, desorption, and atomic absorption spectroscopy (EPA Method 7473). Before sample analysis, calibration was performed using Brooks Rand Instruments Total Mercury Standard (1.0 ng/L). Continuous calibration verification (CCV) was performed using National Institute of Standards and Technology (NIST) standard reference material (SRM) 2709a (San Joaquin Soil, 1100 ng/g), and quality control standard (QCS) was performed using NIST certified reference material 1633c (coal fly ash, 1005 ng/g) with a detection limit of 0.25 ng Hg. All samples were run in duplicate, with relative percent difference within 10%. Instrument detection was 0.5 ng. All standards had recoveries of 90.4–123.9% (mean = 100.7%, n = 102), and all blanks were below detection limit (BDL).

For MeHg concentrations in soils and sediments, dichloromethane extraction was performed [Bloom et al., 1997]. For MeHg concentrations in water, samples reacted with trace grade sulfuric acid for a minimum of 24 hours [Munson et al., 2014]. Soil, sediment, and water samples were analyzed by direct aqueous ethylation with sodium tetraethylborate, purge and trap, cold vapor atomic fluorescence spectroscopy (CVAFS), gas chromatographic (GC) separation, and inductively coupled plasma mass spectrometry (ICP-MS) on a Tekran 2600 Automated Total Mercury Analyzer and Agilent 770 (EPA Method 1630) [Hintelmann and Evans, 1997]. All samples were spiked with an internal standard of Me²⁰²Hg (Oakridge National Laboratory) [Imura et al., 1971] prior to extraction; the recovery of this internal standard in each sample was used to calculate MeHg concentrations. Calibration and CCV were performed using Brooks Rand Instruments Methylmercury Standard (1 ng/L). Initial calibration verification (ICV) was performed using European Reference Material (ERM) CC580 (estuarine sediment, 0.075 µg/g), and QCS was performed using ERM CC580. The instrument detection limit was 1 pg Hg. All standards had recoveries of 73–117% (mean = 93.6%, n = 36). The field blank, digestion blanks, and analysis blanks were BDL.

For THg concentrations in water, samples were analyzed via oxidation with bromine chloride for a minimum of 24 hours, purge and trap, CVAFS, and GC (EPA Method 1631, revision E) on a Tekran 2600. Calibration, CCV, and MS were performed using Brooks Rand Instruments Total Mercury Standard (1.0 ng/L), ICV was performed using SPEX Centriprep ICP-MS Multi-Element in Solution Standard 2A, and QCS was performed using ERM CC580. Instrument detection limit was 0.5 ng/L. All standards had recoveries of 90.2–109.2% (mean = 103.1%, n = 10), the field blank was BDL, and analysis blanks were BDL.

Ash-free carbon content in soils and sediments was determined by the mass loss upon heating the samples at 500°C for 4 hours. All samples were run in duplicate, with relative percent difference within 10%.

All reported values in the text represent the average value of the duplicate samples. Since the data were not normally distributed, all statistical analyses were performed using non-parametric methods. To compare between ASGM sites and reference sites, the Mann-Whitney Wilcoxen Test was used. For multiple comparisons, the Kruskal-Wallis Test was used. Significance was tested at an alpha value of 0.05, and marginal significance at an alpha value of 0.1. All statistical analyses were performed using R 3.1.1 statistical software [R Core Team, 2014]. Analyses were also performed using normalized THg/C and MeHg/C values, which allowed examination of trends in mercury values not influenced by organic matter.

The concentrations of THg and MeHg in soil, sediment, and water collected from ASGM sites were nearly always much greater (all but 2 transect points for MeHg) than those measured in corresponding samples from the reference village (Figures 3 and 4, p < 0.05 for all). Percent Hg as MeHg did not vary between ASGM sites and reference sites (p > 0.3 for all). Despite considerable increases in soil moisture between our first and second sampling dates, we did not observe differences in THg, MeHg, or percent Hg as MeHg between samples collected before and after early monsoonal rains (p > 0.1 for all); we thus merged samples from both dates in all analyses.

Generally, THg concentrations and normalized values based on organic carbon content (THg/C) in soils and sediments were greatest in the burning hut (median of 7.5 µg THg/g, 190 µg THg/g C, p < 0.00001) and lower but still elevated, though variably so, in soils and sediments collected along the transect between each burning hut and the adjacent river (median of 1.5 µg/g). An exception to this occurred in a soil sample collected from Kharahenna 200 m from the burning hut, where we measured the highest THg concentration in the survey (130 µg/g).

Methyl Hg concentrations in soils and sediments were not correlated with THg concentrations (p = 0.35). Instead, across all four villages, MeHg concentrations, normalized MeHg/C, and percent Hg as MeHg were generally highest in river sediments (median of 4.2 ng MeHg/g, 157 ng MeHg/g C, 0.41% Hg as MeHg, p < 0.00002 for all) and lower but still elevated, though variably so, in soils and sediments collected along the transect between each burning hut and the adjacent river (median of 0.79 ng MeHg/g, 26.8 ng MeHg/g C, 0.067% Hg as MeHg). An exception to this pattern occurred in a sample collected from a burning hut in Sabodala, where we measured the highest MeHg concentrations (44 ng/g).

While the highest median river water THg and MeHg concentrations were measured in Kharahenna (1100 ng THg/L, 21 ng MeHg/L), the highest THg and MeHg concentrations in a single sample occurred in Bantako (2400 ng THg/L, 68 ng MeHg/L). Concentrations of Hg in the sample we collected from a drinking well in Kharahenna (28 ng THg/g, 0.10 ng/g MeHg) were higher than 50% of all the ASGM-impacted river water samples we collected (n = 25). In contrast, the commercially available purified water sampled had THg and MeHg concentrations (7.9 ng THg/L, 0.0066 ng MeHg/L) comparable to the reference site.

While the highest median THg and MeHg concentrations in soil and sediment were found in Kolya (n = 141, 2.3 µg THg/g, 2.7 ng MeHg/g), the highest median THg and MeHg concentrations in river water were from Kharahenna (n = 25, 1100 ng THg/L, 21 ng MeHg/L; Supplementary Figure 1).

Our results clearly show that Hg is being transported and deposited well beyond the walls of ASGM burning huts in Senegal. Despite the semi-arid environment of Senegal, both THg and MeHg concentrations were elevated in nearly every water, soil, and sediment sample collected near ASGM sites. As expected, the highest concentrations of THg measured in each village were from samples collected inside the burning huts, but the neurotoxic form of MeHg was highest in river sediments. There was no consistent difference in percent Hg as MeHg in soils between reference and ASGM sites, despite substantial increases in the concentrations of both Hg forms. Thus, the large increases in MeHg concentrations in soils result from higher THg inputs, rather than variation in Hg methylation efficiency.

For sediments, Kharahenna and Bantako include samples that exceed the USEPA THg sediment standard of 0.2 µg/g by an order of magnitude [USEPA, 1985]. For soils, all sites exceed the USEPA THg soil standard of 0.1 µg/g, with some sites exceeding this value by up to two orders of magnitude [USEPA, 1985]. Maximum soil THg concentrations at Senegalese ASGM sites were greater than those observed at all ASGM sites except one site in Venezuela [Santos-Francés et al., 2011], though median soil THg concentrations at our sites were comparable to reported values in the literature (Table 1) [e.g., Donkor et al., 2006; Feng et al., 2006; Loredo et al., 2009; Santos-Francés et al., 2011]. To our knowledge, only two other studies have examined soil MeHg concentrations at ASGM sites; our values were within the range reported in Ghana [Donkor et al., 2006], while our maximum soil MeHg concentrations exceeded those reported in China [Feng et al., 2006]. Sediment THg and MeHg concentrations at Senegalese ASGM sites were comparable to median and mean values reported in previous ASGM studies, though the maximum values we report in Kolya are lower than some of the maximum values in the literature [e.g., Donkor et al., 2006; Santos-Francés et al., 2011; Nyanza et al., 2014; Marrugo-Negrete et al., 2015].

For river water, all four ASGM sites exceed the USEPA THg standard of 12 ng/L for the protection against toxic levels of bioaccumulation in aquatic organisms, including fish consumed by humans [USEPA, 1985]. One site, Kolya, exceeds the World Health Organization (WHO) THg standard of 1000 ng/L [WHO, 1976]. This value is particularly elevated given the high river flow in Kolya, though it does have a large watershed that likely includes other ASGM sources. Though this river water is generally not consumed directly by people, it is sometimes used for drinking by children who swim in the river and is used to wash dishes. Note that drinking water and well water THg concentrations were greater than THg concentrations in half of the river water samples; though the concentration of THg was below the WHO drinking water guideline, if ASGM activities intensify in these communities and utilize increased amounts of Hg, the concentrations of THg in water could approach the WHO drinking water guideline. Maximum water THg concentrations at Senegalese ASGM sites were the second highest reported for other ASGM sites in Africa [Adjei-Kyereme et al., 2015] with the range and median values comparable to those reported in the literature (Table 1) [e.g., Gray, 2002b; Taylor et al., 2005; Donkor et al., 2006; Feng et al., 2006; Rajaee et al., 2015]. Maximum water MeHg concentrations at Senegalese ASGM sites were exceeded in only one study [Feng et al., 2006].

We did not see a consistent decline in soil THg and MeHg concentrations with distance from burning huts, which suggests that Hg distribution is not solely a function of direct local atmospheric deposition. It is likely that Hg used to process gold is either supplied to soil via tailings or atmospherically deposited locally after burning the amalgam. If the former, it is likely that the tailings are not deposited next to the hut; rather, the tailings may be deposited closer to the river, potentially leading to the observed spikes in Hg concentrations at some positions along the transect. If the latter, atmospheric deposition of Hg – likely as the short-lived ionic Hg (Hg²⁺) [Driscoll et al., 2007] – may not be uniformly distributed or predictable, even within several kilometers from the source of emission. Note also that atmospheric depositional patterns of Hg are also influenced by regional Hg transport; other ASGM upwind sites could influence the concentrations we observed. Additionally, since Hg tailings are in the form of Hg⁰ and atmospheric Hg methylation is minimal [Grigal, 2003], we expect that most MeHg is formed in-situ on the landscape.

Each of the four river segments sampled contained large concentrations of THg and MeHg. Concentrations of MeHg and percent Hg as MeHg were greatest in river sediments, which are typically habitats well-suited for anaerobic microbes that produce MeHg [Compeau and Bartha, 1985; Benoit et al., 1998; Skyllberg, 2010; Driscoll et al., 2013]. We expect that this Hg was delivered both from the local ASGM source we sampled, as well as from upstream and upwind ASGM activity. As there are no data available for the locations of all ASGM sites in the region, it is difficult to assess the potential cumulative impact of the many ASGM mines that may occur within each watershed.

High THg and MeHg concentrations in soil and sediment were found in Bantako, Kharahenna, and Kolya, with lower concentrations found in Sabodala. This pattern may be a result of distance between burning hut and river; the first transect point in Sabodala was 2 km from the burning hut. It is possible that Hg tailings and atmospheric deposition are reduced at this greater distance, thereby leading to the lower concentrations observed, compared to the other ASGM sites. It is also possible that the smaller watershed size reduced inputs of Hg from other ASGM sources. Differences might also be due to the number of ASGM miners in different villages or to the amount of Hg used, but this information is difficult to obtain since miners do not readily disclose information regarding their practices. Nevertheless, most of the concentrations of THg and MeHg observed in Sabodala were still larger than those at the reference site.

Since percent Hg as MeHg did not differ among ASGM sites or the reference site and since MeHg deposition is likely low [Grigal, 2003], the rate of methylation between sites is comparable. This pattern suggests that Senegalese aquatic and terrestrial ecosystems are increasing the net amount of methylation in proportion to the amount of THg. Thus, as the sites continue to increase in size, concurrently increasing Hg usage and resultant Hg introduction to the environment, it is also likely that MeHg concentrations will continue to increase in the soil, sediment, and water. This is especially important in rivers, with river sediment at the ASGM sites exhibiting the greatest MeHg concentrations, percent Hg as MeHg, and MeHg/C values. While we did not observe differences in THg and MeHg concentrations between samples collected in the dry season and after the first rains, there is a need for additional wet season sampling to investigate methylation increases that might occur later in the season once the soils are saturated.

The results from this study show that THg and MeHg derived from ASGM is extremely elevated in the adjacent ecosystems, resulting in local-scale contamination and likely contributing to regional-scale contamination. In fact, for every kg of gold produced, 1.1–1.5 kg of Hg is lost to the environment [Van Straaten, 2000], both in tailings (which represent 20–30% of Hg lost) and in amalgam burning (which represent 70–80% of Hg lost) [Van Straaten, 2000; Cordy et al., 2011].

With Senegal exporting approximately 345.6 million USD of gold per year predominantly from ASGM activities [COMTRADE, 2016], and gold selling at approximately 32 USD per gram, we estimate that Senegal releases 12–16 Mg of Hg into the environment annually. This value is an order of magnitude larger than previous estimates [Telmer and Veiga, 2009] and is indicative of the increased practice of ASGM in Senegal over the past decade; gold production has increased 10-fold in Senegal over the past decade [Reichl et al., 2017]. This calculation also indicates a gap between Hg used (likely in excess of 12–16 Mg Hg) and reported Hg imported (744 kg in 2010) [COMTRADE, 2016], which likely results from the clandestine and illegal importation of Hg via the black market. Given similar increases in ASGM activities globally over the past decade, we expect that Hg emissions – and thus Hg entering the environment – has also increased in other locations. This study shows that high THg concentrations are also likely resulting in high MeHg concentrations in soils, sediments, and water.

With such elevated concentrations of MeHg in the soil, plants and animals in ASGM regions have increased risk for Hg bioaccumulation. High MeHg concentrations in soils can also have detrimental impacts on humans, both directly and indirectly. First, Hg and MeHg in the soil can be taken up into crops and rice [Appleton et al., 2006; Zhang et al., 2010], as well as into domestic livestock, such as cattle and fowl [Chibunda and Janssen, 2009]. When these foods are ingested by people as part of their daily diet, they can result in elevated Hg concentrations. Additionally, Hg in soils can be ingested directly by people. This occurs both inadvertently – as inhaled dust particles or ingested dust-covered food – and intentionally – as a consequence of nutrient deficiency. In fact, it is estimated that children consume 25–81 mg soil/day inadvertently and 9–96 mg soil/day purposely, while adults consume 5–517 mg soil/day inadvertently and up to 25 mg soil/day purposefully [Taylor et al., 2005b; Hagan et al., 2013].

Mercury from ASGM is also being transported into aquatic ecosystems and can enter the aquatic food web. The high concentration of MeHg and percent Hg as MeHg in river sediments, consistent with studies of Hg from other ASGM sites [Compeau and Bartha, 1985; Benoit et al., 1998; Skyllberg, 2010; Driscoll et al., 2013], suggests the bioavailability of Hg to aquatic organisms. Additionally, small pellets of Hg⁰ in riverbed sediments derived from ASGM-tailings can become bioavailable; in microcosm experiments, introduced liquid Hg⁰ has been transferred to dissolved Hg⁰ and Hg²⁺, which can then be microbially transformed into MeHg [Dominique et al., 2007; Hu et al., 2013; Balzino et al., 2015]. Methyl Hg in aquatic environments near ASGM can negatively impact local biota, as evidenced by reduced body mass of invertebrate larvae in sediments downstream of ASGM in Tanzania [Chibunda et al., 2008]. This Hg can also bioaccumulate across the food chain, resulting in high concentrations of Hg in piscivorous fish [Driscoll et al., 2007]. High concentrations of Hg in fish near ASGM in Senegal have been correlated with high concentrations of Hg in human hair, a biomarker for MeHg exposure, resulting in concentrations of Hg that exceed the US EPA recommended threshold [Niane et al., 2015].

The impacts of Hg from ASGM on human populations can be significant. In Senegal, similar to many other ASGM countries, over 70% of the local population is estimated to be directly involved in ASGM activities, with even more indirectly impacted [Persaud et al., 2017]; thus, Hg-induced health effects are a source of concern for these communities. While most studies have focused on inhalation of Hg vapors and ingestion of fish as the major pathways of exposure, this study suggests that terrestrial pathways can also be an important source of exposure.

Though soil MeHg concentrations have only been measured in two other ASGM studies [Donkor et al., 2006; Feng et al., 2006], it is likely that similar elevated MeHg concentrations would be found at other ASGM sites globally. With long residence times of THg and MeHg in soils [Thomas et al., 2002; Veiga and Hinton, 2002], legacy impacts of Hg on the environment and humans will continue even once Hg amalgamation is reduced. For example, high Hg has been found in soils and sediments near abandoned gold refineries in Australia, Canada, the United Kingdom, and the United States (California, Nevada, and North Carolina) [de Lacerda and Salomons, 1998].

Scientific evidence of Hg transport is the first step in encouraging miners to reduce their use of Hg in ASGM. Previous work at ASGM sites in Ecuador suggest that using evidence to increase awareness of ASGM-associated risks among miners can lead to increased action to reduce these risks [Adler Miserendino et al., 2013]. In fact, education of miners is likely the most influential means for enacting change from these traditional methods of mercury-gold amalgamation to less risk-intensive mining practices that can simultaneously increase gold recovery [Spiegel et al., 2006; Adler Miserendino et al., 2013; Clifford, 2014], particularly when this education blends technical knowledge with traditional knowledge [Spiegel et al., 2015]. In Senegal, many miners are already organized in a mining collective group; this infrastructure could provide a means for changing miners’ practices, as miner organization in Portovelo-Zaruma, Ecuador has shown to be effective at reducing environmental and occupational risks associated with ASGM [Velásquez-López et al., 2010]. Thus, it is important to use the results from this study, and other similar studies, to inform ASGM miners and communities about the transport and fate of Hg that lead to human exposure.

This study provides evidence that Hg from ASGM in Senegal is entering terrestrial and aquatic ecosystems, is not limited to the burning hut vicinity, and is extremely elevated in all environmental media. It also shows that though Hg emitted is in the form of Hg⁰, it is being converted to the neurotoxic, bioavailable form of MeHg in the environment. The high concentrations of MeHg that we found in soils, sediments, and water in this arid environment warrant additional investigation; MeHg concentrations are likely similar or higher in other countries, such as those in South America and Asia, where ASGM activities are more intense and mesic conditions promote microbial production of MeHg. With ASGM activities in Senegal, and globally, increasing dramatically and predicted to continue to increase [Pacyna et al., 2010], it is important that we gain a greater understanding of the transport and fate of THg and MeHg in the environment. These data can then be used to inform global Hg models [Gustin et al., 2016; Sundseth et al., 2017], to better understand the transport and fate of Hg, as well as to inform local policy, such as through the creation and implementation of National Action Plans – associated with the Minamata Convention – that are directly targeted to locations with Hg transport and transformation in the bioavailable form of MeHg.

Data are available by contacting the author at jacqueline. [email protected].

We thank M. Montesdeoca, S. Todorova, and C. Johnson for assistance with sampling design; F. Danfakha, M. Beye, Senegalese community leaders, and ASGM miners for assistance with field sampling and for providing expertise at sampling sites; U. Ndu and N. Rivera for assistance in conducting laboratory analyses; and M. Simonin for assistance with translating the manuscript into French.

Funding for this work was provided by an Explorer’s Club Mamont Scholarship, American Association of University Women Fellowship, Syracuse University Graduate Fellowship, and NSF Graduate Research Fellowship to J. Gerson.

The authors have no competing interests to declare.

• Contributed to conception and design: JG, CTD

• Contributed to acquisition of data: JG

• Contributed to analysis and interpretation of data: JG, ESB

• Drafted and/or revised the article: JG, CTD, HH, ESB

• Approved the submitted version for publication: JG, CTD, HH, ESB