Permafrost Stores a Globally Significant Amount of Mercury
Abstract
Changing climate in northern regions is causing permafrost to thaw with major implications for the global mercury (Hg) cycle. We estimated Hg in permafrost regions based on in situ measurements of sediment total mercury (STHg), soil organic carbon (SOC), and the Hg to carbon ratio (RHgC) combined with maps of soil carbon. We measured a median STHg of 43 ± 30 ng Hg g soil−1 and a median RHgC of 1.6 ± 0.9 μg Hg g C−1, consistent with published results of STHg for tundra soils and 11,000 measurements from 4,926 temperate, nonpermafrost sites in North America and Eurasia. We estimate that the Northern Hemisphere permafrost regions contain 1,656 ± 962 Gg Hg, of which 793 ± 461 Gg Hg is frozen in permafrost. Permafrost soils store nearly twice as much Hg as all other soils, the ocean, and the atmosphere combined, and this Hg is vulnerable to release as permafrost thaws over the next century. Existing estimates greatly underestimate Hg in permafrost soils, indicating a need to reevaluate the role of the Arctic regions in the global Hg cycle.
Plain Language Summary
Researchers estimate the amount of natural mercury stored in perennially frozen soils (permafrost) in the Northern Hemisphere. Permafrost regions contain twice as much mercury as the rest of all soils, the atmosphere, and ocean combined.
1 Introduction
Over thousands of years, sedimentation buried mercury (Hg) bound to organic material and froze it into the permafrost (Obrist et al., 2017). Permafrost is soil at or below 0°C for at least two consecutive years. The active layer is the surface soil layer on top of the permafrost that thaws in summer and refreezes in winter (Figure S1 in the supporting information). Hg deposits onto the soil surface from the atmosphere, where it bonds with organic matter in the active layer. Microbial decay then consumes the organic matter, releasing the Hg (Smith‐Downey et al., 2010). At the same time, sedimentation slowly increases soil depth such that organic matter at the bottom of the active layer becomes frozen into permafrost. The organic matter consists almost entirely of plant roots, and, once frozen, microbial decay effectively ceases, locking the Hg into the permafrost. However, permafrost has begun to thaw under a changing climate (Hinzman et al., 2005; Romanovsky et al., 2008; Smith et al., 2010). Once the permafrost and associated organic matter thaws, microbial decay will resume and release Hg to the environment, potentially impacting the Arctic Hg balance, aquatic resources, and human health (Dunlap et al., 2007; Jonsson et al., 2017; Obrist et al., 2017; USGS Fact Sheet, https://www2.usgs.gov/themes/factsheet/146-00/, 2016). Model projections estimate a 30–99% reduction in the area of Northern Hemisphere permafrost by 2100, assuming anthropogenic greenhouse gases emissions continue at current rates (Koven et al., 2013). In a novel approach, we make the first‐ever estimate of the storage of Hg in the Northern Hemisphere permafrost soils using empirical relationships based on in situ measurements of sediment total mercury (STHg) combined with published maps of soil organic carbon (Hugelius, Tarnocai, et al., 2013; Hugelius, Bockheim, et al., 2013).
2 Site Descriptions and Methods
To estimate Hg in permafrost regions, we drilled 13 permafrost soil cores of variable lengths in Alaska along a ~500 km north‐south transect representing a broad array of characteristics and ages typical of circumpolar permafrost soils (Figures 1 and S2–S10 and Tables S1 and S2). The maximum depths varied between 98 and 248 cm below the land surface, and the sites reflect a variety of depositional conditions. For each core we dug a pit down to the top of the permafrost and measured active layer depth (ALD). We extracted the cores using a gas‐powered Snow, Ice, and Permafrost Research Establishment (SIPRE) auger modified with carbide cutting blades for frozen soil (Figure S11). The SIPRE can drill to a maximum depth of ~2 m, but obstructing rocks typically limited the actual drilling depth at each site. We broke each core into smaller samples, which we photographed, wrapped, labeled, and stored in a portable freezer (Figure S12). We shipped the frozen cores to the U.S. Geological Survey (USGS) Research Laboratory in Boulder, Colorado and stored them at −20°C until cutting, processing, and analysis.
We cut the cores in half lengthwise using a band saw thoroughly cleaned with methanol and 18 MegaOhm deionized water and archived one half in airtight bags in a freezer for future analyses (Figures S13 and S14). We then sliced the frozen cores into 1.5 cm segments for a total of 588 samples (Figure S15). We followed standard techniques of trace metal sampling with laboratory technicians wearing Latex gloves at all times (U.S. Geological Survey, 2015). Using a ceramic knife on a Teflon surface, we cut each frozen segment into two rectangular subsamples from the center of the segment, one for STHg and one for bulk density (BD). We scraped off the outer few millimeters of the Hg subsample with the knife before measuring STHg (Figure S16). We weighed and measured the dimensions of the larger subsample to determine BD (Soil Survey Staff, 2014). We placed the STHg subsample in a preweighed, Hg‐free jar, recorded the mass, and freeze‐dried it (Figure S17). We thoroughly cleaned a mortar and pestle with 18 MegaOhm deionized water, homogenized the Hg subsample to a fine powder, and stored the powder in Hg‐free glass vials (Figure S18).
We measured STHg, BD, soil organic carbon (SOC), and Δ14C and calculated the Hg to carbon ratio (RHgC) (Table S3). Quality Assurance and Control accounted for more than 56% of all analyses (Texts S1–S3). We split every tenth segment into duplicate subsamples (Figure S19). For the CF and EG cores, we analyzed STHg at the USGS Mercury Research lab in Madison, Wisconsin using EPA Method 7473 (EPA Method 7473 (SW‐846), 1998; USGS Mercury Research Lab (MRL), http://wi.water.usgs.gov/mercury-lab/analysis-methods.html, 2013). For all other cores, we measured STHg using a Milestone Direct Mercury Analyzer‐80 (DMA‐80) at the USGS Research Laboratory in Boulder, Colorado (Figure S20, Text S4, and Table S4). For soil carbon dating, we give Δ14C concentration as the fraction Modern Δ14C and conventional radiocarbon age (Text S5). We measured loss on ignition (LOI) for the second subsample as the loss of weight of a dried sample ignited at 550°C for 5 h by a Muffle furnace (Soil Survey Staff, 2014). To estimate SOC, we multiplied LOI by a carbon fraction of 0.493 (Anderson & Sarmiento, 1994). We used the standard Redfield ratio of 0.493 based on the 117:14:1 C;N;P; ratio of marine biomass samples. We calculated RHgC as the ratio of STHg to SOC for each sample (Bargagli et al., 2007).
To create maps of soil Hg, we multiplied maps of soil carbon by a representative RHgC. We used soil carbon maps from the Northern Circumpolar Soil Carbon Database (NCSCD) at a spatial resolution of 0.5° latitude and longitude (Hugelius, Tarnocai, et al., 2013; Hugelius et al., 2014). We calculated separate Hg maps for several soil layers: 0–30 cm, 0–60 cm 0–100 cm, 0–300 cm, active layer, and permafrost. We assume 300 cm represents the typical soil accumulation since the last glacial maximum to capture 90% of the carbon in the near surface soils. The active layer Hg extends from the surface to ALD, and the permafrost Hg extends from the ALD to 300 cm. We used simulated ALD from the Community Land Model version 3.4 (Koven et al., 2015) and a linear fit of cumulative SOC with depth from the profiles in Harden et al. (2012). We ignore localized deposits in yedoma and deltaic soils known to extend below 300 cm. We account for the fact that permafrost occurs only under 50–90% of the land area in discontinuous zones and for the volume of soil taken up by excess ground ice (Hugelius, Tarnocai, et al., 2013; Hugelius et al., 2014).
We used the median as the representative RHgC with uncertainty as the 25th to 75th percentiles. We use quadrature to combine uncertainty of RHgC with the 14.5% uncertainty in soil carbon from the NCSCD. We statistically evaluated the representativeness of the median RHgC by subdividing the data and determining if the median RHgC changes between sites, depths, and soil types. We regressed RHgC against 14C age to evaluate the representativeness of the median RHgC over time. We evaluated the spatial representativeness of our measurements by comparison with 11,000 published measurements. Less than 2% of published data come from permafrost sites, so we included data from boreal and temperate sites (Table S5 and S6). We included only sites with both SOC and STHg measurements to calculate RHgC. The published data included extreme outliers indicative of modern Hg contamination that biased the statistics. Our samples stayed frozen for thousands of years and represent preindustrial conditions, so to make a fair comparison, we removed outliers exceeding the mean plus 2 times the standard deviation (73 STHg and 125 RHgC values). This resulted in 11,000 published measurements of STHg, SOC, and RHgC from 4,926 different sites.
3 Results
We find a median STHg of 43 ± 30 ng Hg g soil−1 and a median representative RHgC of 1.6 ± 0.9 μg Hg g C−1 (Figure 2). For both STHg and RHgC, the median and mean values for our data match each other and those from the published data within uncertainty. The published data show a higher mean RHgC because it has 11% or ~1,100 STHg values greater than 200 ng Hg g soil−1, compared to one sample in our data, indicating local contamination by anthropogenic sources at some sites. Our STHg data showed a bimodal distribution with a primary peak at 40 ng Hg g soil−1 and a weaker secondary peak at 100 ng Hg g soil−1 resulting from data from four of the cores, indicating strong variability between sites. The published data also show variability between sites, but the secondary peak does not appear at all. Our data and the published data both show a single dominant mode in RHgC, with 95.5% of the RHgC values falling between 0.0 and 5.0 μg Hg g C−1.
The median RHgC of 1.6 ± 0.9 μg Hg g C−1 appears representative of site, depth, soil type, and age. STHg showed high variability between sites, but the median RHgC is insensitive to site (Figure S21). Comparing our sites to maps of soil characteristics indicates the cores broadly represent ~69% of the soils found in Alaska (Hugelius, Tarnocai, et al., 2013; Hugelius et al., 2014). RHgC does not change with soil type: mineral and peat soils have roughly the same median RHgC. Although STHg and SOC both decrease with depth (Figures S22 and S23), RHgC is constant with depth: the RHgC median for any 20 cm range of depths is 1.6 μg Hg g C−1 (Figures S24 and S25). Because age decreases exponentially with depth, SOC and STHg also decrease with age. However, unlike STHg and SOC, RHgC, appears constant with age (Figures S26–S29).
Maps of soil Hg show great spatial variability reflecting different sedimentation histories (Figure 3). The relative uncertainty per pixel is 57%, which means there is a 95% probability that the actual value lies within ±57% of our estimated value. Soils with high carbon content and high sedimentation show high Hg mass, such as the North Slope of Alaska and the Mackenzie River basin in Canada. Areas with little sedimentary overburden and shallow soils such as the Rocky Mountains and the Canadian Shield in North America have low SOC and Hg. Slow decay due to freezing temperatures coupled with high sedimentation rates has buried substantial amounts of carbon and Hg over much of Siberia. The active layer depth (ALD) varies from 30 cm near the Arctic coastline to 100 cm in the southern permafrost regions, so much of the buried carbon and the Hg bound to it lies frozen and preserved in permafrost. Our 0–30 cm map agrees with a published 0–30 soil Hg map in areas with little sedimentary overburden but shows much higher soil Hg in areas with high sedimentation rates, particularly Siberia (Smith‐Downey et al., 2010). A large pool of Hg in the active layer leaching into Arctic Rivers might explain why the permafrost‐dominated terrestrial environment is the dominant source of Hg to the Arctic Ocean and why the Arctic Ocean is a net source of Hg to the Atlantic and Pacific Oceans (Fisher et al., 2012; Schuster et al., 2011; Soerensen et al., 2016).
4 Discussion
Several atmospheric Hg sources unique to the northern high latitudes have significant spatial and temporal variability to explain STHg variability within and between cores (Fitzgerald & Lamborg, 2003). Northern boreal forest fires release Hg into the atmosphere leading to spatial variability in Hg deposition (Homann et al., 2015; Rothenberg et al., 2010; Turetsky, et al., 2006). Spatial variations in temperature and moisture change microbial respiration rates (Wickland et al., 2006). Peaks in atmospheric Hg during summer resulting from atmospheric mixing with ozone enhance Hg deposition (Banic et al., 2003; Sonke & Heimbürger, 2012). Springtime atmospheric Hg depletion after the polar sunrise may elevate Hg deposition to the high latitudes (Berg et al., 2008; Fitzgerald et al., 2005; Lindberg et al., 2002). Although southeast Alaska contains geologic Hg deposits (Gray et al., 2000), we see no evidence of terrestrial geologic Hg sources within Alaska (Eberl, 2004; Williams, 1962). Volcanic eruptions release Hg into the atmosphere, leading to variability in Hg deposition (Pirrone et al., 2010; Pyle & Mather, 2003; Schuster et al., 2002), but we saw little evidence of Hg deposition from volcanic ash. These depositional processes are not exclusive to the 13 coring sites chosen for this study but rather inclusive to the high latitudes of the Northern Hemisphere.
Despite these processes leading to highly variable depositional environments, our STHg measurements appear consistent with similar measurements in permafrost regions. Peat deposits in Tomsk Oblast, west Siberia (Lyapina et al., 2009) show STHg values and vertical profiles similar to ours. Rydberg et al. (2010) measured STHg of 40 ng Hg g soil−1 below 25 cm depth, close to our median value of 43 ng Hg g soil−1. STHg in active layer soils along a ~970 kilometer north‐south transect of Alaska ranged from 100 ng Hg g soil−1 in the O horizon to 50 ng Hg g soil−1 in the A horizon (Wang et al., 2010). STHg ranged between 12 and 375 ng Hg g soil−1, and RHgC varied between 1 and 11.3 μg Hg g C−1 at a sub‐Antarctic site in Tierra del Fuego (Peña‐Rodríguez et al., 2014).
Moreover, our STHg measurements appear consistent with published data for nonpermafrost soils (Figure 4). Most of the 11,000 published data fall in the temperate midlatitudes, with 2088 in boreal forests and only 67 points in permafrost regions. Here we show a curve fit of the median STHg as a function of SOC with the 90% envelope defined as the 5th to 95th percentiles (Figures S30–S32). Superimposing our data indicates that 90% of our STHg measurements fall within the 90% envelope of the published data, with a slight shift toward higher STHg.
As SOC increases, STHg appears to shift from a receptor‐limited regime to a flux‐limited regime (Figure 4). Hg enters plants through the roots or by dry deposition from the atmosphere onto leaves (Obrist et al., 2017; Windham‐Myers et al., 2009), where it attaches to appropriate receptor sites in organic molecules in place of nutrients such as iron or magnesium. For mineral soils with SOC < 10%, the number of receptor sites limit the Hg plants can retain, and STHg increases strongly with SOC (R2 = 0.98). For organic soils with SOC > 10%, receptor sites appear unlimited and the flux of Hg from the atmosphere limits the STHg. Since atmospheric deposition is highly variable in space and time, the data become noisier and STHg appears nearly independent of SOC. The two regimes might explain why, like previously published data (Bargagli et al., 2007; Erickson, 2014), some sites show statistically significant correlations between STHg and SOC, whereas others did not. Sites with low SOC in the receptor‐limited regime tended to have statistically significant correlations, while sites with organic soils in the flux‐limited regime did not.
We estimate that soils in permafrost regions contain an estimated 1,656 ± 962 Gg Hg, of which half or 793 ± 461 Gg Hg is frozen in permafrost (Figure 5 and Table 1). We estimate the total mass of Hg in each layer by multiplying each pixel by the grid cell area and summing across the permafrost domain. The average of nine previously published estimates of global soil Hg is 454 ± 321, with a range from 235 to 1,000 Gg Hg (Table S7). However, these estimates generally limit soil depth to 30 cm, indicating our 0–30 cm soil Hg for permafrost regions (347 ± 196 Gg Hg; Table 1) rivals the global soil Hg in previous estimates. These studies leverage the known link between microbial decay and Hg release and often rely on biogeochemical models that tend to underestimate soil Hg in permafrost regions. To improve Hg estimates in permafrost soils, these models should account for the large drop in microbial activity under freezing conditions and sedimentation processes that bury and freeze organic matter into the permafrost. Our results indicate the active layer alone represents the largest Hg reservoir on the planet. The active layer and permafrost together contain nearly twice as much Hg as all other soils, the ocean, and the atmosphere combined.
| Depth range (cm) | Soil carbon (Pg) | Soil Hg (Gg Hg) |
|---|---|---|
| 0–30 | 217 ± 12 | 347 ± 196 |
| 0–60 | 333 ± 18 | 532 ± 301 |
| 0–100 | 472 ± 27 | 755 ± 427 |
| 0–200 | 827 ± 108 | 1323 ± 764 |
| 0–300 | 1035 ± 150 | 1656 ± 962 |
| Permafrost | 496 ± 72 | 793 ± 461 |
| Active layer | 539 ± 78 | 863 ± 501 |
Hg in permafrost soils represents an environmental risk as permafrost continues to thaw in the future. The turnover time associated with the microbial decay of frozen organic matter is ~14,000 years (Figure S28), making the Hg locked in permafrost effectively stable on human time scales. However, projections indicate a 30–99% reduction in near surface permafrost by 2100, and, once thawed, the turnover time for microbial decay drops to ~70 years (Koven et al., 2013; Schaefer et al., 2011). This makes the reservoir of Hg in permafrost soils vulnerable to release over the next century, with unknown consequences to the environment.
5 Summary
We measured a median STHg of 43 ± 30 ng Hg g soil−1 and a median RHgC of 1.6 ± 0.9 μg Hg g C−1 based on 588 samples from 13 soil permafrost cores from the interior and the North Slope of Alaska. These values appear consistent with published results of Hg concentrations for tundra soils and 11,000 nonpermafrost soil measurements from 4,926 different sites in North America and Eurasia. In a novel approach, we estimate that the entire Northern Hemisphere permafrost region contains 1,656 ± 962 Gg Hg, of which 793 ± 461 Gg Hg is frozen in permafrost. Northern Hemisphere permafrost soils contain nearly twice as much Hg as all other soils, the ocean, and the atmosphere combined, indicating a need to reevaluate the role of the Arctic regions in the global Hg cycle. This Hg is vulnerable to release as permafrost thaws over the next century.
Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.
Acknowledgments
The authors thank John Borg, John Crawford, and Bob Eaganhouse for drilling assistance. This article is dedicated to George R. Aiken, whose 40 year career as a U.S. Geological Survey scientist was cut short with his death from cancer in 2016. Without George's foundational scientific contributions to our knowledge of the world's organic matter, much of what we know about organic matter would not have been possible, including the findings from this study. George is missed by too many friends and family to count. However, his legacy as “Dr DOC” will live on in all of us. Supporting Information (SI) are available at http://agupubs.onlinelibrary.wiley.com/hub/journal/10.1002/(ISSN)1944-8007/. Data generated for this study are available at https://www.sciencebase.gov/catalog/domain repository and GeoPass: https://geopass.iedadata.org/josso/. Reprints and permission information are available at https://grl‐submit.agu.org/. The authors declare no competing financial interests. Readers are welcome to comment on the online version of this paper. Correspondence and request for materials should be addressed to pschuste@usgs.gov (https://orcid.org/0000‐0002‐8314‐1372) or kevin.schaefer@nsidc.org. Funding for this research came from the USGS Toxic Substances Hydrology Program (TSHP) and the USGS NASQAN/NAWQA and CEN Programs; NASA grants NNX10AR63G, NNX06AE65G, NNX17AC59A, and NNX13AM25G; NOAA grant NA09OAR4310063; CUHK Direct grant 4053206; the National Natural Science Foundation of China (91325202); the National Key Scientific Research Project (2013CBA01802) by the Ministry of Science and Technology of China; and National Science Foundation grant ARC1204167. G.H. received funding from the EU H2020 Nunataryuk project (773421). P. S. devised and supervised the study, drilled four cores, and supervised all laboratory work. K. M. S. supervised the collection of the other cores and analyzed data. D. P. K., R. G. S., K. P. W., and G. R. A. provided advice and guidance during writing. R. C. A. provided statistical support. J. F. D. analyzed and quality assured data from two cores. J. D. G. and C. W. processed all the cores. G. H. led map development. E. J. assisted with the writing process and references. N. H. M. contributed to map development. A. G., E. J., L. L., T. S., and T. Z. drilled one or more cores. D. A. R. wrote the standard operation procedure for the DMA‐80, maintained the instrument, and processed and quality assured 85% of the data.
