A review of global environmental mercury processes in response to human and natural perturbations: Changes of emissions, climate, and land use

Atmosphere

Atmospheric total gaseous mercury (TGM) concentrations have been measured since the late 1970s, with more reliable observations emerging since approximately 1990 (Slemr et al. 2003; Lindberg et al. 2007; Temme et al. 2007; Sprovieri et al. 2010; Slemr et al. 2011; Sprovieri et al. 2016). Mean TGM concentrations in background areas in the northern hemisphere, the tropics, and the southern hemisphere are 1.3–1.6, 1.1–1.3, and 0.8–1.1 ng/m³, respectively (Slemr et al. 2011; Sprovieri et al. 2016). Many observations show that in the past three decades global background TGM concentrations have declined, particularly in North America, Europe, and over the North Atlantic Ocean (Slemr et al. 2003; Lindberg et al. 2007; Temme et al. 2007; Slemr et al. 2008; Ebinghaus et al. 2011; Slemr et al. 2011; Soerensen et al. 2012; Cole et al. 2013; Weigelt et al. 2015; Sprovieri et al. 2016), with rates of declines between − 1.5 and − 2.2% per year. Because anthropogenic emission inventories did not support such declines, proposed drivers for declines include decreased Hg concentrations and subsequent evasion in the upper North Atlantic Ocean and/or changes in terrestrial surface–atmosphere fluxes (Slemr et al. 2011; Soerensen et al. 2012). Others have shown that declining point-source emissions of divalent Hg (Hg^II) in the northern hemisphere and from commercial products (Horowitz et al. 2014) were sufficient to reproduce observed declining trends in TGM from 1995 to present (Zhang et al. 2016d). TGM concentrations have decreased at lower rates in parts of the Arctic (approximately 0 to − 0.2% per year) (Cole et al. 2013). Several studies have suggested that the recent reversal in global TGM trends in China and India are due to increasing emissions from coal-fired power plants in addition to artisanal and small-scale gold mining (ASGM) activity (Slemr et al. 2014; Martin et al. 2017; Streets et al. 2017). At Cape Point in South Africa, TGM has increased during the last decade, although a clear explanation for this trend is lacking (Martin et al. 2017).

The global atmospheric Hg reservoir is estimated to be between 4400 and 5300 Mg, and is enriched by more than an order of magnitude relative to natural levels and by approximately three- to fivefold relative to 1850 levels (Amos et al. 2013; Engstrom et al. 2014; Horowitz et al. 2017; Streets et al. 2017). The anthropogenic contribution to atmospheric Hg has been inferred from observed Hg deposition in polar firn air, peat bogs, and lake sediments, all of which reveal large-scale impacts of anthropogenic emissions, particularly in the second half of the twentieth century (Biester et al. 2007; Faïn et al. 2009; Farmer et al. 2009; Outridge et al. 2011; Goodsite et al. 2013; Engstrom et al. 2014; Amos et al. 2015; Blais et al. 2015; Drevnick et al. 2016). Combining deposition records with Hg stable isotope analysis now allows differentiation between atmospheric Hg^II and Hg⁰ deposition, and reconstruction of past Hg⁰ levels (Enrico et al. 2016, 2017). Enrico et al. (2017) suggest that the maximum twentieth-century atmospheric Hg⁰ concentration was 15 times larger than natural background at various locations, which agrees well with modeling analyses by Amos et al. (2015). Reconstructions suggest that historical atmospheric Hg⁰ concentrations peaked around 1970 at 3–4 ng/m³ and then declined to the current levels of 1.3–1.6 ng/m³ (Faïn et al. 2009; Enrico et al. 2017), consistent with atmospheric Hg⁰ concentration measurements of the last few decades. Lake sediment archives, reflecting total atmospheric Hg deposition, show similar time trends, with maximum Hg accumulation rates around 1970–1990 (Engstrom et al. 2014; Amos et al. 2015; Drevnick et al. 2016).

Direct anthropogenic Hg emissions account for approximately 30% of total annual inputs (i.e., total anthropogenic and natural emissions plus re-emission) to the atmosphere (Pirrone et al. 2010; UNEP 2013; Streets et al. 2017). While several different emission inventories exist for the present day, few are comparable across the past decades (Muntean et al. 2014) or address all-time human emissions (Streets et al. 2011; Horowitz et al. 2014; Streets et al. 2017). Global anthropogenic Hg emissions have been estimated at approximately 2000 Mg/year by most studies (AMAP/UNEP 2008, 2013; Pacyna et al. 2010; Pirrone et al. 2010; Streets et al. 2011) except for one that reported a lower estimate (EDGARv4) of 1287 Mg/year (Muntean et al. 2014). While reference years of studies are within a common period of 2005–2010, the sectoral distributions among some studies are inconsistent. The Arctic Monitoring and Assessment Programme and United Nations Environment Programme (AMAP/UNEP) (2008) and studies by Pacyna et al. (2010), Pirrone et al. (2010), and Streets et al. (2011) found stationary fossil fuel combustion (SFFC), mainly coal combustion, to be the leading source of Hg (800–900 Mg/year), followed by ASGM (400 Mg/year). By assuming a higher activity level of unregulated and illegal ASGM in some countries, UNEP’s (2013) global Hg assessment (AMAP/UNEP 2013) estimated ASGM emissions of more than 700 Mg/year. SFFC accounted for less than 500 Mg/year of Hg in the UNEP inventory, while the EDGARv4 inventory (Muntean et al. 2014) estimated lower contributions from SFFC and ASGM. Given the global importance of ASGM releases, this source warrants further constraints. National-level emission inventories rely on local field emission tests, with an increasing amount of data emerging from newly industrialized countries. For example, Zhang et al. (2015a) estimated anthropogenic Hg emissions in China to be 538 Mg for 2010, somewhat lower than a previous estimate of 609 Mg/year (Pirrone et al. 2010). The discrepancy was attributed to better Hg removal efficiencies of air pollution control devices for nonferrous metal smelting (Wang et al. 2010; Zhang et al. 2012). Emissions in India were recently estimated at approximately 240 Mg/year (Pirrone et al. 2010; Burger Chakraborty et al. 2013). National anthropogenic Hg emission estimates are available for several other countries (e.g., Kim et al. 2010; Masekoameng et al. 2010; Nansai et al. 2012; Nelson et al. 2012; US EPA 2015). As per the UNEP emission inventory, the eight highest emitting countries are (in order) China, India, Indonesia, Columbia, South Africa, Russia, Ghana, and the U.S. These countries emit a total of 1095 Mg/year of Hg, or 56% of the global anthropogenic Hg emissions, to the atmosphere (AMAP/UNEP 2013; UNEP 2013). More accurate emission inventories are expected in the future, as parties of the UNEP Minamata Convention on Hg improve national Hg use and emission inventories.

Uncertainties remain in global anthropogenic Hg emission inventories, mainly because of a lack of local field test data for key sectors such as cement production, iron and steel production, waste incineration, and gold production, and a lack of accurate activity-level data for ASGM and intentional Hg use. Horowitz et al. (2014) found that more than 400 Mg/year of atmospheric Hg emission was from the intentional use of Hg in products and processes (excluding gold production), previously unaccounted for in inventories. Top-down constraints and inverse modeling from atmospheric observations are currently being explored as additional tools to better constrain emissions (Song et al. 2015, 2016; Denzler et al. 2017).

Hg speciation profiles of emissions are crucial to assess the environmental impacts of atmospheric emissions (Zhang et al. 2016b). Streets et al. (2005) reported overall Hg⁰:Hg _gas ^II :Hg _particulate ^II ratios of 20:78:2 and 80:15:5 for Chinese coal-fired power plants (CFPPs) and cement plants (CPs), respectively, based on North American and European field test data. A recent study by Zhang et al. (2015a) updated the speciation profiles to 79:21:0 and 34:65:1 for Chinese CFPPs and CPs, respectively, based on onsite measurements. Such changes in speciation profiles for large point sources lead to large variations in Hg transport distances and local deposition fluxes.

Hg⁰ oxidation to Hg^II is considered a key step in removing Hg from the atmosphere (Selin et al. 2008; Lyman and Jaffe 2011; Subir et al. 2011), although increasing evidence shows that oxidized Hg^II deposition is dominant only in oceans, while Hg⁰ deposition is dominant in many terrestrial ecosystems (see below). A recent controversy has arisen around the techniques and artifacts in measuring atmospheric Hg oxidation and atmospheric Hg^II concentrations. We refer to published reviews on atmospheric Hg and Hg^II measurement methodologies (Lyman et al. 2010b; McClure et al. 2014a, b; Jaffe et al. 2014; Gustin et al. 2015; Angot et al. 2016; Mao et al. 2016; Marusczak et al. 2017). Earlier work suggested a primary role for the oxidants OH and O₃ in Hg⁰ oxidation, but observations and theoretical calculations now suggest that the Br radical is likely the dominant oxidant (Hynes et al. 2009; Gratz et al. 2015; Shah et al. 2016). Current thinking is that the mechanism is a two-step process, where Hg⁰ reacts with the Br radical to form HgBr, and HgBr reacts with multiple potential oxidants (Br, I, OH, BrO, IO, NO₂, etc.) to generate different Hg^II species (Goodsite et al. 2004; Dibble et al. 2012; Wang et al. 2014; Ye et al. 2016; Horowitz et al. 2017; Jiao and Dibble 2017). Recent aircraft observations showed substantial atmospheric Hg^II concentrations and detectable BrO in the subtropics, suggesting that subtropical anticyclones are significant global Hg^II sources (Gratz et al. 2015; Shah et al. 2016). Photolytic formation of halogen radicals and Hg oxidation may occur at low light conditions, and recent Antarctic wintertime Hg oxidation may suggest a possible dark oxidation process (Nerentorp Mastromonaco et al. 2016). Hg oxidation chemistry has recently been updated in the GEOS-Chem model by Horowitz et al. (2017), which describes in-cloud Hg^II photoreduction as potentially important in Hg redox chemistry (Holmes et al. 2010; Horowitz et al. 2017). A recent model comparison (Travnikov et al. 2017), not including the updated chemistry of Horowitz et al., showed that models with a Br oxidation mechanism reproduced the observed near-surface seasonal variations in the Hg^II fraction, but did not capture seasonal variations in wet deposition observed at monitoring sites in North America and Europe. These findings suggest that more complex Hg oxidation chemistry and multi-oxidation pathways may be occurring in different regions of the atmosphere.

Terrestrial

In the terrestrial environment, the largest Hg pools are located in soils (Grigal 2003; Obrist 2012). Using global soil carbon inventories, Hg pools have been estimated at 300 Gg (Hararuk et al. 2013) and 240 Gg (Smith-Downey et al. 2010). The latter study estimated a 20% modern-day increase in overall soil storage and an even greater increase of Hg associated with the labile soil carbon fraction (Smith-Downey et al. 2010). Newer simulations using additional observational constraints (Streets et al. 2011; Horowitz et al. 2014) estimate higher present-day organic soil Hg pools (250–1000 Gg with a best estimate of 500 Gg) and propose that anthropogenic activities have doubled the Hg stored in organic soils (Amos et al. 2013, 2015). Hg in organic-rich upper soils is predominantly from atmospheric deposition (Grigal et al. 2000; Schwesig and Matzner 2000; Guedron et al. 2006; Peña-Rodríguez et al. 2012; Demers et al. 2013; Peña-Rodríguez et al. 2014; Jiskra et al. 2015; Enrico et al. 2016; Zheng et al. 2016; Obrist et al. 2017; Wang et al. 2017). Atmospheric Hg deposits are retained in humus-rich upper soils and bound to organic matter (Meili 1991; Grigal 2003; Obrist et al. 2011; Jiskra et al. 2015). In mineral horizons, a significant component of Hg also stems from release from natural geologic sources.

New spatial soil datasets demonstrate that landscape Hg distribution is correlated to soil organic matter, latitude, annual precipitation, leaf area index, and vegetation greenness (Obrist et al. 2011; Richardson et al. 2013; Navrátil et al. 2014, 2016; Obrist et al. 2016). Obrist et al. (2016) explained soil Hg variability across 1911 sites in the western U.S., largely by vegetation patterns, with high soil Hg accumulation in productive forests and 2.5 times lower soil concentrations in unproductive deserts and scrublands. Such patterns are due to a dominance of plant-funneled atmospheric Hg⁰ deposition in many terrestrial ecosystems (see below), or by higher re-emission of Hg from deserts compared to forests (Eckley et al. 2016). Soil Hg accumulation and retention, however, are also determined by soil morphology and genesis as well as soil properties, including soil organic matter stability, content, texture, and pH (Obrist et al. 2011; Richardson and Friedland 2015; Navrátil et al. 2016). Landscape gradients with the highest inorganic Hg concentrations in forested watersheds have also been reported in river and lake sediments (Fleck et al. 2016), highlighting strong connectivity between upland Hg deposition and aquatic loading. This is consistent with earlier studies that linked high biological Hg hotspots in the eastern U.S. to watersheds with high forest densities (Driscoll et al. 2007).

Elevation gradients in Hg, largely associated with shifts in vegetation type, have also been observed, with vegetation and organic soil Hg concentrations increasing by up to fourfold with increasing elevation (Zhang et al. 2013; Townsend et al. 2014; Blackwell and Driscoll 2015; Wang et al. 2017). Blackwell and Driscoll (2015) reported a shift in deposition patterns along an alpine forest gradient, from deposition dominated by litterfall (i.e., uptake of Hg by plant foliage and transfer to ecosystems after senescence/leaf shedding) to throughfall deposition (i.e., Hg uptake to plant surfaces and wash-off by precipitation) and deposition via cloudwater. They also noted that cloudwater accounted for up to 71% of total deposition at the highest altitudes. Hg stable isotope studies are inconsistent, reporting both higher contributions of Hg^II at high-elevation sites (Zhang et al. 2013; Zheng et al. 2016) and lower contributions of Hg^II with increasing altitude (Wang et al. 2017).

Aquatic

Oceans contain a substantial fraction of the global Hg reservoir and strongly affect atmospheric concentrations through air–sea exchange (Soerensen et al. 2010; Zhang et al. 2015b). Sunderland and Mason (2007) estimated the total ocean Hg to be 350 Gg with a 90% confidence limit of 270–450 Gg. More recent work (Lamborg et al. 2014; Zhang et al. 2015b) suggests a reservoir of 260–280 Gg. Lamborg et al. (2014) estimated the upper ocean (top 1000 m) reservoir, based on observations from the GEOTRACES cruise series, to be 63 Mg, which is on the low end of the range reported by Sunderland and Mason (2007) (63–120 Mg; 90% confidence interval), and may indicate either declining concentrations in some basins or improvements in analytical techniques (Lamborg et al. 2012). In the pelagic marine environment, trends in total Hg (THg) concentration are difficult to infer because of large inter-laboratory variability in measurements (Lamborg et al. 2012). Concentrations in the North Atlantic appear to be declining during the last several decades, from mean vertical profile concentrations of above 5 pM in the 1990s to consistently below 1 pM on recent cruises (Bowman et al. 2016). In the North Pacific, however, Sunderland et al. (2009) suggest that the THg concentration in the North Pacific Intermediate Water (NPIW) mass has increased since the mid-1980s due to enhanced atmospheric deposition from the Asian continent. Other cruises that sampled different locations (Hammerschmidt and Bowman 2012; Munson et al. 2015) did not see statistically different THg concentrations than those measured on earlier cruises, but current data coverage is sparse.

In freshwater ecosystems, predominant Hg sources include direct release from Hg-containing effluents, river runoff that contains atmospheric Hg deposits that accumulated in terrestrial environments, and direct atmospheric Hg deposition. The latter is particularly important in lakes with large surface area-to-volume ratios and small catchment-to-lake surface areas. The relative importance of atmospheric and watershed Hg sources varies [see the “Atmosphere–aquatic interactions” section and Hsu-Kim et al. (2018)] depending on the degree of development and land use change, hydrology, and dissolved organic carbon (DOC) content and composition (Engstrom and Swain 1997; Knightes et al. 2009; Lepak et al. 2015). Kocman et al. (2017) recently developed a first inventory of Hg inputs to freshwater ecosystems and estimated that 800–2200 Mg/year of Hg enters freshwater ecosystems. This is lower than the amount of Hg entering coastal ecosystems estimated by Amos et al. (2014) (5500 ± 2700 Mg/year), in part because it does not include natural mobilization from terrestrial ecosystems. Earlier estimates of global riverine Hg release to the ocean were between 1000 and 2000 Mg/year (Cossa et al. 1997; Sunderland and Mason 2007). ASGM is the largest direct Hg source globally to both land and water (approximately 880 Mg/year); although the proportion released to aquatic ecosystems is not certain, it is estimated to be 50% (Kocman et al. 2017). Streets et al. (2017) suggest that approximately 40% of combined Hg releases to land and water globally are sequestered at the release site rather than traveling in rivers to the ocean. In India and China, industrial Hg sources (from the chlor-alkali industry, Hg and large-scale gold production, nonferrous metal production, and Hg-containing waste) release approximately 86 Mg/year to lakes and rivers, accounting for 51% of global aquatic releases from these sources (Kocman et al. 2017).

Methylmercury (MeHg) production is a critical process that occurs within aquatic ecosystems. The largest source of MeHg to freshwater lakes and wetlands is in situ microbial production, with surface sediments producing larger amounts than the water column. However, the relative contributions of water column versus sediment productions depend on specific characteristics of lakes such as stratification, depth of the anoxic hypolimnion, and organic carbon content of sediments. The activity of methylating microbes is controlled by temperature, redox conditions, pH, and the presence of suitable electron donors (e.g., organic carbon) and acceptors (e.g., sulfate, Fe^III, methane). The primary controls on inorganic Hg bioavailability include DOC, sulfur, and Hg^II concentrations and speciation (Benoit et al. 1999; Boening 2000; Ullrich et al. 2001; Gilmour et al. 2013; Hsu-Kim et al. 2013). A breakthrough paper by Parks et al. (2013) identified a two-gene cluster (hgcA and hgcB) in microbes involved in Hg methylation. The hgcA gene encodes a putative corrinoid protein capable of transferring a methyl group to Hg^II, and the HgcB protein returns HgcA to a redox state that enables it to receive a new methyl group. Subsequent work identified these genes in many organisms, including sulfate and iron-reducing bacteria and methanogens (Gilmour et al. 2013; Yu et al. 2013; Podar et al. 2015). Several studies have developed hgcA probes to screen aquatic sediments for Hg-methylating microorganisms (Bae et al. 2014; Schaefer et al. 2014; Bravo et al. 2015; Du et al. 2017) and to demonstrate that different methylating microbes inhabit various niches, such as regions of wetlands with varying sulfate concentrations (Bae et al. 2014; Schaefer et al. 2014). Recent development of a broad range of probes spanning all known hgcAB genes (Christensen et al. 2016) may allow future quantification of the methylating potential in environmental samples.

One conundrum that remains is how to explain high levels of MeHg in marine seawater where MeHg production is still poorly understood. Gionfriddo et al. (2016) found that the micro-aerophilic, nitrite-oxidizing bacterium Nitrospina, present in Antarctic sea ice and brine, contains genes that are slight rearrangements of the hgcAB genes, suggesting that diverse microbial communities may be capable of methylation. Fitzgerald et al. (2007) suggested that most marine MeHg is produced in ocean margin sediments, but a variety of studies across most major ocean basins have since produced strong evidence for water column MeHg production (Cossa et al. 2009; Sunderland et al. 2009; Heimbürger et al. 2010; Cossa et al. 2011; Lehnherr et al. 2011; Blum et al. 2013). In the last several years, the abundance of ocean water concentration measurements has increased dramatically with new data from the North Pacific (Kim et al. 2017); North Atlantic (Bowman et al. 2014); Equatorial, South Atlantic, and Pacific (Munson et al. 2015; Bowman et al. 2016); Arctic (Heimburger et al. 2015); and Antarctic Oceans (Gionfriddo et al. 2016). In the coastal marine environment, Schartup et al. (2015) used enriched Hg isotope incubation to detect active Hg methylation for the first time in oxic estuarine seawater. Methylation was facilitated by the presence of labile terrestrial DOC and a shift in ionic strength and microbial activity that accompanies the transition into saline waters in estuaries. Similarly, Ortiz et al. (2015) measured active methylation in laboratory experiments with marine snow aggregates.

Marine species are responsible for a large fraction of human exposure to MeHg; for example, Sunderland (2007) found that for the U.S. more than 90% of MeHg exposure is from marine and estuarine species. Thus, understanding MeHg dynamics in marine environments is particularly important for public health (Gribble et al. 2016). Advances have been made in understanding MeHg uptake to aquatic food webs (see Eagles-Smith et al. 2018). The photochemical demethylation of MeHg has been shown experimentally to result in a large enrichment of odd-mass-numbered Hg isotopes (Δ¹⁹⁹Hg; Fig. 3, Box 1) in the residual MeHg pool that accumulated through the food web (Bergquist and Blum 2007). Odd-mass-independent fractionation (MIF) anomalies in aquatic organisms were used as source tracers to distinguish between MeHg derived from sediments with small odd-MIF isotopes and MeHg from the open ocean with large odd-MIF isotopes (Senn et al. 2010; Blum et al. 2013). Odd-MIF anomalies in biota have been hypothesized to serve as source tracers and as a proxy for ecological parameters such as foraging depth; for example, the largest odd-MIF anomalies are found in fish feeding in surface waters where photochemical demethylation is most active (Blum et al. 2013). The extent of odd-MIF anomaly is reduced in aquatic organisms where light penetration in the aquatic water column is inhibited [e.g., by sea ice (Point et al. 2011; Masbou et al. 2015)] or with high DOC (Sherman and Blum 2013). Thus, odd-MIF anomalies can track changes in climatological/environmental parameters such as sea-ice cover. Field measurements and experimental data suggest that DOC controls MeHg availability for uptake in freshwater. French et al. (2014) observed a water column DOC threshold of 8.6–8.8 mg/L, above which MeHg uptake was inhibited in Arctic lakes, while Jonsson et al. (2014) showed that in estuarine microcosms DOC-bound MeHg was 5–250 times more available for bioaccumulation into lower food web organisms than newly produced MeHg.

Open image in new window

Atmosphere–terrestrial interactions

Large uncertainties in terrestrial Hg⁰ sinks and source strengths, particularly over background areas and vegetated areas, make atmosphere–terrestrial interactions one of least constrained processes in the global Hg cycle. Major emissions from terrestrial ecosystems include biomass burning and volatilization of Hg⁰ from natural and Hg-enriched soils, while deposition from the atmosphere includes dry Hg⁰ deposition to plants and soils and wet Hg^II deposition.

Recent global atmospheric models have reported global biomass burning emissions of Hg between 210 and 680 Mg/year (Amos et al. 2014; De Simone et al. 2015; Song et al. 2016). Webster et al. (2016) recently provided an update on atmospheric Hg emissions from western U.S. wildfires and suggest a fivefold lower total (3.1 ± 1.9 Mg/year across 11 states) compared to previous estimates (14.7 Mg/year) (Wiedinmyer and Friedli 2007). The lower Hg emission estimates from Webster et al. are attributed mainly to the inclusion of fire severity and the notion that low-severity burns (52% of wildfires) result in lower soil heating and less Hg emissions. Other studies also found that Hg losses are related to fire intensity (Mitchell et al. 2012; Kolka et al. 2017). Wildfire studies consistently show that fires mobilize Hg primarily from biomass, litter, and organic soils (Friedli et al. 2001, 2003; Engle et al. 2006; Biswas et al. 2007; Obrist 2007; Wiedinmyer and Friedli 2007; Biswas et al. 2008; Burke et al. 2010; Woodruff and Cannon 2010), although Hg released from upper mineral soils can comprise up to 10% of total emissions (Biswas et al. 2007; Burke et al. 2010; Melendez-Perez et al. 2014). Significant post-fire soil evasion of Hg⁰ has been reported to occur years after fires, leading to an additional 50% in Hg emissions (Melendez-Perez et al. 2014), an effect which is not included in wildfire emission estimates.

A relatively well-constrained component of the atmosphere–terrestrial exchange is wet deposition of Hg^II, at least over land where extensive monitoring networks exist, including the Mercury Deposition Network (MDN) for the U.S. and Canada (Prestbo and Gay 2009; Weiss-Penzias et al. 2016; Sprovieri et al. 2017), the European Monitoring and Evaluation Programme (EMEP) for Europe (Tørseth et al. 2012), and a new Asia–Pacific Mercury Monitoring Network (2017) and the Global Mercury Observation System (GMOS) for other parts of the world. At background sites, mean atmospheric Hg wet deposition globally decreased from 10–15 μg/m²/year in the early 1990s to the current level of 5–10 μg/m²/year (Zhang et al. 2016d). Declines in atmospheric Hg^II wet deposition are reported from Europe and the U.S. with average deposition declines of − 1.5% per year and − 2.3% per year between 1990 and 2013, respectively (Zhang et al. 2016d), although increasing wet Hg^II deposition has been reported in East Asia. Mean background Hg wet deposition fluxes in the northern hemisphere, the tropics, and the southern hemisphere are approximately 5–10, 2–8, and 1–5 μg/m²/year, respectively (Prestbo and Gay 2009; Fu et al. 2016a; Sprovieri et al. 2016, 2017). There is, however, large spatial and temporal variability in wet Hg^II deposition (Butler et al. 2008; Prestbo and Gay 2009; Risch et al. 2012b; Zhang and Jaeglé 2013; Cole et al. 2014; Muntean et al. 2014; Sprovieri et al. 2016).

Estimated median dry deposition fluxes of Hg _gaseous ^II and Hg _particulate ^II in Asia and North America are 10.7 and 6.1 μg/m²/year, respectively. A particular challenge with dry Hg^II deposition is that measurements and modeling schemes have uncertainty factors greater than two (Wright et al. 2016), in addition to large measurement uncertainties (see the “Atmosphere” section). Atmospheric dry deposition flux measurements can be broadly divided into micrometeorological methods based on gradient measurements, dynamic gas flux chambers, and surrogate surface passive samplers, which have been more widely used in recent years (Lyman et al. 2009; Zhang et al. 2009; Lyman et al. 2010a; Gustin et al. 2011; Pirrone et al. 2013; Gustin et al. 2015; Huang and Gustin 2015). To model deposition, resistance approaches are used in chemical transport models that apply dry deposition velocities for Hg^II and measured or estimated Hg^II concentrations (Lin et al. 2006; Zhang et al. 2009). For simulations of net Hg⁰ deposition, bidirectional soil–vegetation–air exchange models are needed (Bash et al. 2007; Zhang et al. 2009); current models are not well constrained. Models have been used to estimate total Hg deposition compared with existing measurements and to attribute deposition to sources in the context of policy. Different models show similar global Hg deposition patterns, especially near source regions, and can reproduce major features of observed wet deposition (Travnikov et al. 2017). However, in some regions, there are substantial differences in simulated attributions of Hg deposition to local versus international sources.

Atmospheric Hg⁰ emissions from land surfaces was previously estimated between 1600 and 2900 Mg/year (Lindberg et al. 2007; Selin et al. 2008; Smith-Downey et al. 2010), but more recent estimates are considerably lower at 607 Mg/year (− 513 to 1353 Mg/year) (Agnan et al. 2016). Estimates of large terrestrial Hg⁰ evasion, or remobilization of legacy Hg from previous depositions, are in part based on enriched isotope tracing studies that showed that 45–70% of applied labile HgCl₂ was subject to atmospheric evasion (Hintelmann et al. 2002; Graydon et al. 2012). However, recent global data assimilations of 20+ years of terrestrial Hg⁰ flux measurements (Agnan et al. 2016; Zhu et al. 2016) now show that in most of the world Hg⁰ evasion from background soils is generally low, particularly when soils are covered by litter or plants. In fact, many background soils also show periods of net Hg⁰ deposition (Gustin et al. 2000; Choi and Holsen 2009) with recent work showing net Hg⁰ deposition in 20% of measurements over bare soils and in 45% over covered soils (Agnan et al. 2016). On the other hand, Agnan et al. (2016) and Zhu et al. (2016) showed an increased importance of soil Hg⁰ emissions in areas with high atmospheric Hg exposures (such as East Asia), which according to Agnan et al. (2016) now accounts for 261 Mg/year (range of 114–359 Mg/year) of Hg⁰ emissions. They also estimated Hg⁰ evasions from naturally Hg-enriched and contaminated areas (217 Mg/year, range of 202–258 Mg/year) that were lower than previous estimates (Ericksen et al. 2006) but still the second largest terrestrial Hg⁰ source. Another terrestrial Hg⁰ source includes global croplands (201 Mg/year; range 195–208 Mg/year). When considering vegetation uptake of atmospheric Hg, most terrestrial ecosystems turn into net atmospheric Hg⁰ sinks (see below). Recent global model simulations now reflect reduced soil Hg⁰ evasion (Amos et al. 2014, 2015); one global atmospheric model simulation (Song et al. 2016) showed reduced global soil evasions from 2200 to 1360 Mg/year, which is still in the higher range based on flux measurements by Agnan et al. (2016).

Over land, Hg^II deposition is minor compared to Hg⁰ deposition, and the dominant pathway of atmospheric deposition to terrestrial ecosystems is through litterfall. Globally, this deposition pathway is estimated at 1020–1230 Mg (Obrist 2007; Risch et al. 2012a; Fu et al. 2016a, b; Wang et al. 2016; Wright et al. 2016; Zhang et al. 2016c). In addition, throughfall deposition may contribute an additional 50% of litterfall deposition (Wang et al. 2016), which would result in a global deposition flux of approximately 560 Mg/year, although this needs further observational confirmation. In addition, further plant depositions occur via woody tissues, including by tree blowdown (Mitchell et al. 2012). Given that Hg contained in woody biomass can exceed Hg contained in foliage biomass by 67–100% (Obrist et al. 2012; Yang et al. 2017), woody tissue deposition is likely an important, yet not well-constrained, source of deposition as well (Obrist 2007). Plant contributions (litterfall plus throughfall wet deposition) dominate deposition loads in forests globally (70–85%), of which approximately 75% (with the exception of South American rainforests) is litterfall (Wang et al. 2016). Wright et al. (2016) and Wang et al. (2016) show large continental differences with high litterfall fluxes in the Amazon and in China that are greater than fivefold those in North America and Europe. This suggests that deposition fluxes decrease spatially from tropical to temperate/boreal regions, with 30% of total deposition in the temperate and boreal regions and 70% of total deposition in the tropical/subtropical regions. Unfortunately, there is a lack of plant and associated Hg litterfall and throughfall deposition data outside of forest ecosystems, such that estimates for global grasslands, savannas, and shrublands remain highly uncertain. Hg stable isotope approaches now suggest that Hg⁰ deposited via plants comprises the majority of Hg deposition to soils in central North America (57–94%) (Demers et al. 2013; Zheng et al. 2016), Alaskan tundra soils (71%) (Obrist et al. 2017), central European peat soils (79%) (Enrico et al. 2016), boreal forest soils in North Sweden (90%) (Jiskra et al. 2015), and high mountain soils in Tibet (Wang et al. 2017). This is likely also true for other areas of the world, for example, the tropics, where some of the highest litterfall Hg deposition is observed (Wang et al. 2016). Recent field studies also show evidence for an active Hg⁰ sink in soils (Sigler and Lee 2006; Moore and Castro 2012; Obrist et al. 2014, 2017) and possibly direct deposition of Hg⁰ to forest litter (Zheng et al. 2016). The finding that litterfall is the dominant terrestrial Hg⁰ source fundamentally reshapes our understanding of Hg deposition to terrestrial landscapes and Hg mobility, demonstrating the need for monitoring strategies to measure net ecosystem Hg⁰ deposition, including direct exchange/deposition measurements. The dominant role of foliar Hg⁰ uptake also results in a specific “terrestrial” Hg isotope signature that can be used to trace terrestrial contributions and transfer to aquatic ecosystems (Tsui et al. 2012; Kwon et al. 2015; Jiskra et al. 2017), aquatic biota (Li et al. 2016), and lake and ocean sediments (Grasby et al. 2015; Lepak et al. 2015; Enrico et al. 2016; Araujo et al. 2017; Gleason et al. 2017). Finally, whole-ecosystem flux measurements using micrometeorological techniques can further confirm the dominant Hg⁰ deposition source to terrestrial ecosystems. The longest data set spans 2 years in the Arctic tundra (Obrist et al. 2017) and shows that Hg⁰ deposition (6.5 ± 0.7 μg/m²/year) dominates other deposition processes (71% of total deposition), in agreement with Hg stable isotope analysis. Previous studies over grasslands also reported a net annual dry Hg⁰ deposition [17.5 μg/m²/year (Fritsche et al. 2008a) and 3.3 μg/m²/year (Castro and Moore 2016)] and shorter-term measurements showed a net Hg⁰ deposition during vegetation periods (Lee et al. 2000; Bash and Miller 2008; Fritsche et al. 2008b; Bash and Miller 2009; Sommar et al. 2016). Other studies, however, have reported inconsistent results, with net terrestrial Hg⁰ emissions even in the presence of vegetation (Lindberg et al. 1998; Cobos et al. 2002; Obrist et al. 2005; Cobbett and Van Heyst 2007; Baya and Van Heyst 2010).

Atmosphere–aquatic interactions

A new global model suggests that approximately 50% of total global wet Hg^II deposition may occur to tropical oceans within 30° northern and southern latitudes due to a high abundance of free tropospheric Hg^II reservoirs scavenged by precipitation in these regions (Horowitz et al. 2017). These important global hotspot regions for Hg^II deposition need to be confirmed by direct deposition measurements.

Air–sea exchange of Hg⁰ is critical to extending the lifetime of anthropogenic Hg in the atmosphere, terrestrial ecosystems, and the ocean (Strode et al. 2007; Soerensen et al. 2010). The magnitude of net Hg⁰ evasion from the ocean is approximately 2900 Mg/year (range 1900–4200 Mg/year) (Horowitz et al. 2017; Streets et al. 2017), compared to approximately 2000 Mg/year from primary anthropogenic emissions; thus, the ocean is considered the largest source of atmospheric Hg globally. There is, however, substantial uncertainty in the magnitude of air–sea exchange from different ocean regions. Simultaneous measurements of dissolved Hg⁰ in seawater and atmospheric concentrations are used to assess this source, with recent latitudinal gradients in both the Atlantic and Pacific Oceans showing distinct spikes in concentrations around the Intertropical Convergence Zone (ITCZ). These spikes are thought to result from deep convection and intense precipitation in these regions that increase atmospheric Hg^II deposition (Soerensen et al. 2014). Using high-frequency measurements of atmospheric and aquatic Hg⁰ concentrations, Soerensen et al. (2013) inferred that terrestrial DOC can inhibit Hg^II reduction and ambient Hg⁰ production in seawater. Lamborg et al. (2016) measured concentrations of Hg on settling organic particulate matter in the Atlantic Ocean and suggested that partitioning to particles may be up to 1000 times greater than assumed in some models. This would effectively lower the magnitude of the global evasion flux of Hg⁰ from the ocean because scavenging of Hg from surface waters would reduce the pool of Hg^II available for reduction and conversion to Hg⁰.

Terrestrial–aquatic interactions

In watersheds of the Americas, with historic Hg and precious metal mining, runoff from contaminated sites dominates aquatic Hg loads and is substantial (44 Mg/year) (Kocman et al. 2017). In contrast, in developed regions, including Europe and North America, catchment inputs and atmospheric deposition are generally the primary Hg sources, with background terrestrial environments accounting for approximately 230 Mg/year (170–300 Mg) of riverine loads. Hydrology and DOC mobility are the major controls on delivery of catchment Hg to aquatic ecosystems, with episodic high-flow periods from storms and/or snowmelt often representing the majority of annual catchment inputs (St. Louis et al. 2016).

Recent Hg stable isotope analyses have been able to differentiate among inorganic Hg sources to freshwater. The waters of remote Canadian lakes share the positive Δ²⁰⁰Hg signature observed in precipitation, suggesting that atmospheric deposition to these lakes is the major Hg source (42 ± 26%) (Chen et al. 2016). Similarly, analyses of sediment of the Laurentian Great Lakes indicate that atmospheric Hg^II deposition is the dominant Hg source to Lakes Huron, Superior, and Michigan, while terrestrial catchment and industrial effluent are the dominant Hg sources to Lakes Erie and Ontario (Jackson et al. 2015; Lepak et al. 2015). Contrasting results have been reported for a boreal forest catchment in Sweden, where Hg stable isotope analyses suggest that Hg in forest runoff originated from the deposition of Hg⁰ through foliar uptake, rather than from precipitation (Jiskra et al. 2017). Analyses of dated lake sediment cores and catchment soils using geochemical tracers can also be used to tease apart changes in atmospheric and terrestrial Hg^II inputs over time (Fitzgerald et al. 2005; Perry et al. 2005; Landers et al. 2008; Muir et al. 2009; Kirk et al. 2011; Hermanns and Biester 2013). Recent analyses of Hg stable isotopes in dated lake sediment cores have also been used to examine temporal changes in Hg sources (Sonke et al. 2010; Gehrke et al. 2011; Cooke et al. 2013; Donovan et al. 2013).

For pelagic ocean regions, the dominant source of Hg is atmospheric deposition (Soerensen et al. 2010). Rivers comprise only a small fraction of Hg input to most ocean basins because the majority of Hg is in the particle phase and settles in ocean margins. Zhang et al. (2015b) showed that the fraction of riverine Hg exported to the open ocean is 6–25% for all basins except the Arctic. The large shelf and relatively smaller surface area of the Arctic Ocean make riverine inputs of Hg a more important source. For example, modeling studies have used atmospheric observations to infer a large missing Hg source from rivers and coastal erosion (Fisher et al. 2012; Zhang et al. 2015b). Other work using measured Hg:DOC ratios and a synthesis of available observations suggests a riverine Hg source to the Arctic of approximately 50 Mg/year (Dastoor et al. 2015; Zhang et al. 2015b), which is slightly larger than atmospheric deposition of Hg (approximately 45 Mg/year).

Atmosphere

Critical factors affecting future anthropogenic emissions include energy use and Hg emission control strategies [e.g., in response to the Minamata Convention, see Selin et al. 2018 and Giang et al. 2015] and mitigation measures for global climate change. One of the largest anthropogenic sources of Hg, ASGM emissions, will critically affect future Hg emissions, but is also associated with the largest uncertainties in future emission estimates. Streets et al. (2009) projected the range of 2050 global Hg emissions to be 2390–4860 Mg, compared to the 2006 level of 2480 Mg. For combustion sources, large-scale deployment of advanced Hg sorbent technologies, such as activated carbon injection (ACI), may lower the 2050 emission range to 1670–3480 Mg/year. A study by Giang et al. (2015) suggested 90 and 150 Mg/year of avoided power sector emissions for China and India in 2050, respectively, due to Minamata Convention control technology requirements. Declining Hg emissions in China through 2050 could, however, be offset by increasing Indian emissions (Giang et al. 2015), leading to a shift in source areas globally toward the tropics. Rafaj et al. (2013) evaluated the co-benefit of global greenhouse gas mitigation efforts and found that anthropogenic Hg emissions under a global climate mitigation regime would be reduced in 2050 by 45% compared to the case without climate measures. Pacyna et al. (2016) estimated that a best-case scenario of Hg emission decreases by 2050 could result in a decrease in Hg of up to 50% in the northern hemisphere and 35% in the southern hemisphere.

Lei et al. (2014) compared the influence of projected emission changes and climate change on U.S. atmospheric Hg levels in 2050. They found that anthropogenic emissions would contribute 32–53% of projected changes in Hg air concentration, while climate and natural emission changes would account for 47–68%. Of the effects of climate change, influences via secondary emissions and ecosystem structure (see the “Atmosphere–terrestrial interactions” section) will likely exert the largest changes on atmospheric Hg pools. Although less important, atmospheric processes can also be directly impacted by climate change; for example, the oxidation rate constant of the Hg–Br reaction is sensitive to temperature. Zhang et al. (2016a) estimated a 5–7% increase in northern hemisphere atmospheric surface Hg⁰ concentrations by 2050 because of reduced oxidation and increased in-cloud reduction. Changes in precipitation patterns can change spatial distribution, magnitude, and seasonal variation of Hg^II deposition. Megaritis et al. (2014) projected increases in Hg^II deposition due to climate change to the eastern U.S. of 6% during summer and 4% during winter by 2050, with significant spatial variability. Zhang et al. (2016a) predicted that land use and land cover changes may lead to increases in Hg⁰ dry deposition fluxes under future climate. Hansen et al. (2015) projected climate-induced decreases in Hg deposition to the Arctic, while Zhang et al. (2016a) estimated increases in northern high latitudes. Large-scale meteorological patterns such as El Niño-Southern Oscillation may also influence interannual variability of atmospheric Hg concentrations (Slemr et al. 2016).

Terrestrial

Major impacts on global Hg cycling are expected due to shifts in global biomes, hydrology, fire patterns, water table depth, soil moisture, and redox conditions. In addition, forestry practices and deforestation can affect watershed Hg processes and MeHg exposures and biomagnification, although responses can be variable between different watersheds (discussed in detail in Hsu-Kim et al. (2018)). Based on the importance of vegetation for atmospheric Hg deposition, predicted shifts in biomes will directly impact terrestrial and aquatic Hg distribution and impacts. Modeling studies have predicted a strong sensitivity of Hg⁰ dry deposition to changes in vegetation (Krabbenhoft and Sunderland 2013; Zhang et al. 2016a). Zhang et al. (2016a) projected that by 2050 the annual mean Hg⁰ dry deposition flux over land will increase by 20% in northern mid-latitudes, driven by a combination of increased atmospheric Hg⁰ exposure and increased vegetation and foliage density induced by CO₂ fertilization. Regionally decreased Hg⁰ dry deposition may occur in South Asia and Africa, for example, where projected increases of agricultural land area will lead to losses of leafy areas. Strong shifts in soil Hg accumulation have been predicted as well. Hararuk et al. (2013) suggested that, along with changes in soil organic carbon, soil Hg sequestration in the U.S. will increase by 2100, mainly induced by changes in precipitation patterns. They also predicted that increases in CO₂ concentrations to 700 ppm would stimulate both soil carbon and soil Hg accruals, while increased air temperatures would have small negative effects on both. The combined effects of increased CO₂, increased temperature, and increased or decreased precipitation will likely lead to pronounced regional differences in soil Hg changes. Anticipated shifts in coniferous versus deciduous forest abundance will also likely impact terrestrial Hg distribution. Richardson and Friedland (2015) suggested that anticipated losses of 2.2 million ha of coniferous forests will lead to 27 Mg less Hg sequestered in organic horizons across this region, either as a function of reduced atmospheric Hg deposition or increased release into watersheds. These anticipated effects of climate change and ecosystem properties on Hg cycling can be seen clearly in archive studies. In a remote lake in Patagonia, for example, pre-anthropogenic changes in sediment loads up to a factor of four were found, comparable to recent anthropogenic forcing (Hermanns et al. 2012), and were attributed to past changes in deposition related to climate-induced shifts in terrestrial organic matter flux and DOM leaching from catchment soils. Similarly, Rydberg et al. (2015) documented how a climate-induced change from deciduous to coniferous forests around 5700 years BP led to reduced soil erosion and increased transport of DOC-bound Hg from a catchment in the area of Germany, resulting in increased Hg concentrations and accumulation rates.

Other ecosystem disturbances are expected to affect Hg cycling, but in unknown directions. For example, changes in wildfire frequency and abundance will impact atmospheric Hg emissions and watershed fluxes in a nonlinear fashion given the complex relationships between Hg emissions and fire intensities, pre-fire Hg accumulation, and post-fire Hg mobilization. Wildfire Hg emissions now likely include substantial emissions of anthropogenic legacy Hg accumulated in litter and soils, and climate and land use change impacts on fire frequency and activity need to be considered as part of human impacts (Westerling et al. 2006). A recent modeling study (Kumar et al. 2017) estimated a 14% increase in wildfire-related Hg emission due to climate change. Corresponding land use changes may amplify or alleviate this effect, e.g., due to reductions in forest fires by agricultural land expansion in Africa or increased forest abundance in North America and Asia. In addition, land management practices such as burning of agricultural fields, grassland, and brush fires contribute to Hg emissions, although little information is available on their global contribution. Land use changes such as shifts in land management and forestry practices also have the potential to mobilize terrestrial Hg pools, via increased erosion, changes in hydrological pathways, and changes in yields (Kocman et al. 2017; Kronberg et al. 2016). Kocman et al. (2017) summarized studies that show pronounced Hg leaching when soils are converted to agricultural land and predicted that global deforestation of 13 million ha/yr in the 2000s may have yielded up to 260 Mg Hg per year entering local freshwaters.

Aquatic systems

Both oceans and freshwater ecosystems will be affected by changing Hg emissions and climate-induced alterations. Modeling studies show that response times for lakes can range from a few years to many decades (Harris et al. 2007; Knightes et al. 2009). Recovery of freshwater ecosystems to decreased atmospheric Hg emissions is expected when reduced atmospheric Hg loadings translates into a substantial decrease in DOC-bound MeHg and inorganic Hg catchment runoff (Graham et al. 2013; Chiasson-Gould et al. 2014; French et al. 2014; Jonsson et al. 2014; Lepak et al. 2015; Ndu et al. 2015). A recent long-term mass balance study in New York state suggested that Hg inputs to lakes from forested catchments could be very responsive to decreased Hg emissions (Gerson and Driscoll 2016). They reported that a 25% reduction in atmospheric Hg⁰ concentration resulted in a 40% reduction in litterfall Hg fluxes, which in turn decreased inorganic Hg and MeHg stream inputs (45 and 29%, respectively) and lake concentrations over a 10-year period (Gerson and Driscoll 2016).

A common disturbance of freshwater Hg cycling results from widespread reservoir creation, which generally amplifies MeHg production; reservoir effects are discussed in detail in Eagles-Smith et al. (2018) and Hsu-Kim et al. (2018). Most studies of climate change impacts on freshwater Hg cycling have been conducted in sensitive ecosystems experiencing accelerated changes (e.g., high-altitude and high-latitude lakes, ponds, and peatlands). In the western Canadian Arctic, where approximately 15% of the water bodies have shoreline retrogressive thaw slump lakes, lake sediment cores demonstrated increased sedimentation of inorganics and decreased sediment concentrations of DOC, THg, and MeHg (Deison et al. 2012). In contrast, a survey of MeHg accumulation in invertebrates in tundra lakes suggested that lakes with low DOC are experiencing increased DOC input because of humification and these lakes may see a two- to threefold increase in Hg bioaccumulation as the DOC concentration approaches a threshold of approximately 8 mg/L (MacMillan et al. 2015). Jonsson et al. (2017) predicted a three- to sixfold increase in MeHg concentration in estuarine zooplankton in association with large inputs of terrestrial DOC intended to mimic potential future changes in runoff. Peat mesocosm experiments suggest that changes in hydrological regimes and shifts in vascular plant communities may have a significant impact on Hg cycling in peatlands (Haynes 2017). For example, lower, more variable water tables and the removal of Ericaceae shrubs significantly enhanced inorganic Hg and MeHg mobility in peat pore waters and MeHg export from snowmelt, likely from enhanced peat decomposition and internal regeneration of electron acceptors related to water table changes (Haynes et al. 2017).

The effect on Hg of large-scale changes in the marine environment expected from climate change was explored by Krabbenhoft and Sunderland (2013). Increased seawater temperatures may enhance organic matter remineralization and the propensity for MeHg production in some regions of the ocean. Rapid changes in sea-ice cover and seasonality in the polar oceans are likely to exert a major impact on air–sea exchange of Hg in these regions, and further work may be needed to understand how temperature and sea-ice dynamics alter Hg dynamics in the Arctic (Angot et al. 2016). Changes in microbial community structure and ocean productivity will propagate through marine food webs in potentially unexpected ways, altering bioaccumulation. Fisher et al. (2013) used 30 years of satellite data to quantify the impacts of interannual variability in meteorology on inorganic Hg dynamics in the Arctic. They concluded that the dominant climate mode of the future in the Arctic may result in a lower reservoir of Hg in the Arctic Ocean because of enhanced air–sea exchange. This work did not consider the effects of a changing (melting) terrestrial landscape, which may dramatically increase Hg input to the Arctic from the terrestrial environment. Soerensen et al. (2016) suggested that rivers are likely to become more dominant sources of Hg and MeHg in the future Arctic Ocean. Stern et al. (2012) reviewed a variety of changes that are occurring and expected to occur in the Arctic Ocean in the future, and suggested that future effects are likely to be large; they also highlighted current limitations in understanding multiple, simultaneously occurring impacts on Hg cycling and bioaccumulation. Finally, recent modeling suggests that the response times of marine fish tissue burdens to changes in Hg input will depend on the locations of MeHg production (Li et al. 2016). Benthic sediment will respond much more slowly than the upper Ocean and estuarine surface waters (Sunderland et al. 2010; Schartup et al. 2015; Soerensen et al. 2016). Amos et al. (2013) predicted that legacy Hg will be present in global ecosystems for periods ranging from decades to millennia, suggesting a large time lag in response to changes in anthropogenic emissions and climate.