Permafrost Thaw Increases Methylmercury Formation in Subarctic Fennoscandia

Methylmercury (MeHg) forms in anoxic environments and can bioaccumulate and biomagnify in aquatic food webs to concentrations of concern for human and wildlife health. Mercury (Hg) pollution in the Arctic environment may worsen as these areas warm and Hg, currently locked in permafrost soils, is remobilized. One of the main concerns is the development of Hg methylation hotspots in the terrestrial environment due to thermokarst formation. The extent to which net methylation of Hg is enhanced upon thaw is, however, largely unknown. Here, we have studied the formation of Hg methylation hotspots using existing thaw gradients at five Fennoscandian permafrost peatland sites. Total Hg (HgT) and MeHg concentrations were analyzed in 178 soil samples from 14 peat cores. We observed 10 times higher concentrations of MeHg and 13 times higher %MeHg in the collapse fen (representing thawed conditions) as compared to the peat plateau (representing frozen conditions). This suggests significantly greater net methylation of Hg when thermokarst wetlands are formed. In addition, we report HgT to soil organic carbon ratios representative of Fennoscandian permafrost peatlands (median and interquartile range of 0.09 ± 0.07 μg HgT g–1 C) that are of value for future estimates of circumpolar HgT stocks.


Brittany Tarbier, Gustaf Hugelius, A. Britta K. Sannel, Carluvy Baptista-Salazar, Sofi Jonsson
Corresponding Author: Sofi Jonsson, sofi.jonsson@aces.su.se Index Supplementary Information for Material and Methods S1 Study Sites S1 Figure S1. Karlebotn peat plateau coring site S2 Figure S2. Karlebotn collapse fen coring site S3 Figure S3. Tavvavuoma Table S5. One-way ANOVA and Tukey's post-hoc, core classes S11 Table S6. Two-way ANOVA and Tukey's post-hoc, core classes and depth S12 Figure S4. Depth-distribution of HgT conc. by site S13 Figure S5. Depth-distribution of %SOC by core class and by site S14 Figure S6. Depth-distribution of all parameters for each site S15 Figure S7. Depth-distribution of HgT conc. normalized to C by core class and by site S20 Figure S8. HgT as a function of %SOC (RHgTC) S21 Figure S9. Depth-distribution of %MeHg across all sites and core classes S21 Discussion -potential effect of sampling transportation S22 The mean annual ground temperature at all study sites is just below freezing, with an approximate late season thaw depth of 55 to 60 cm in Tavvavuoma and 40 to 70 cm in Finnmark [7][8][9] . In Tavvavuoma the mean annual precipitation for the period 2006 -2013 was 461 mm/yr (at the nearest meteorological station in Naimakka, ~35 km northeast of Tavvavuoma) 8 while the mean annual precipitation for the Finnmark sites is below 500 mm for the inland locales and above 1000 mm at the coast. 10 Across these sites, the principal vegetation of the relatively dry peat plateau surfaces consists of dwarf shrubs (Empetrum nigrum ssp. hermaphroditum, Betula nana, Vaccinium uligunosum ssp. uligunosum, V. vitis-idaea, V. microcarpum, Andromeda polifolia), Rubus chamaemorus, lichens, and mosses (e.g. Polytrichum juniperinum, Dicranum elongatum). 1,7 At the lake edges and in the fens both adjacent to and at a distance from the degrading permafrost plateaus, hydrophytic cotton grasses (Eriophorum spp.), sedges (Carex spp.) and Sphagnum species dominate.
S2 Figure S1. The Karlebotn peat plateau coring site with vegetation typical of a permafrost-raised plateau. In the middle-distance, raised ridges of peat are interspersed with small sunken "valleys" formed through ground subsidence. Photo: Marit Hichens-Bergström.

S4
Sample preparation and analysis. All samples were freeze-dried in a Heto Drywinner 6.55 freeze dryer then homogenized to a fine powder with an agate mortar and/or a scientific-grade IKA A11 Basic Mill. Thorough cleaning using ethanol was conducted between each sample to prevent contamination.
Geochemical analysis. Biogeochemical analysis of %SOC, %N, δ 13 C, and δ 15 N was carried out at the UC Davis Stable Isotope Facility. In order to test for the prevalence of inorganic carbon, ~80% of samples were baked in the lab at Stockholm University at 900° for 4 hours following standard loss on ignition protocols 11 . Inorganic carbon was found to contribute insignificantly to the overall soil composition, with values ranging from 0-3.4%, a mean of 0.82% and a standard deviation (SD) of 0.84. As a result, further analysis of inorganic carbon was not conducted and "carbon" is understood to refer to "organic carbon" for purposes of this study. All %SOC scatterplots were created in R. 12

Analysis of HgT.
Concentrations of HgT was determined through thermal decompositionatomic absorption spectrophotometry using a Milestone Direct Mercury Analyzer-80 (DMA) following the procedure described in U.S. EPA Method 7473 13 . Approximately 100 (± 5%) mg of homogenized and freeze-dried sample material was weighed out into preweighed and cleaned Ni sample boats and analyzed with the DMA. The instrument detection limit is 0.01 ng HgT, with a working range of 0.05 -600 ng. To minimize the risk of memory effects between samples and ensure DMA-calibration, between every 10 ± 2 samples, an empty sample boat, a boat containing an internal sediment reference standard (previously determined to contain 210 ± 10 ng/g soil), and a boat of replicate sample material was also tested. Throughout HgT testing for this study, the internal standard had a Hg concentration of 201.3 -221.8 ng/g soil, with a mean of 214.9 ng/g soil, a SD of 3.8, and a relative standard deviation (RSD) of 1.8%. The replicate material was a mineral-rich sample from the Karlebotn collapse fen. For this quality control sample, HgT, in ng/g soil, ranged from 4.3 -7.6, with a mean of 5.4, a SD of 0.7, and an RSD of 12.2%. All HgT scatterplots were created in R. 12

Analysis of MeHg.
In order to extract MeHg from the soil samples, between ~0.5 -1 g of material was measured into new 50 mL Falcon tubes and between 20 -100 μL of an internal standard-an isotopically enriched Me 200 Hg standard with concentration 1.1 ng/g-added and left to equilibrate for an hour. During a subsequent run, for 11 samples and 3 replicates (each analyzed twice), the amount of tracer was adjusted to between 100 μL and 2.5 mL to improve the analysis, as too little or too much tracer can "mask" the measured MeHg concentration. After equilibration, 10 mL KBr (1.4M), 2 mL CuSO4 (2M) and 10 mL dichloromethane, DCM (CH2Cl2) are added to each tube, which is capped and left for 45 minutes. In order to extract MeHg, the samples are rotated at 85 RPM on a sample rotor for 45 minutes, then centrifuged for 5 minutes at 3000 RPM. Glass Pasteur pipettes are used to manually transfer the lower (clear) layer containing DCM and the extracted S5 MeHg to a new 50 mL Falcon tube. After adding 10 mL of Milli-Q (MQ) water to the pipetted liquid, the DCM is purged in a warm water bath at 45 °C, and the extraction is complete.
MeHg was analyzed using a Tekran® Model 2700 Automated Methylmercury Analysis System connected to an Inductively Coupled Plasma Mass Spectrometer, Thermo-Fisher Xseries 2 (ICPMS). Prior to analysis, half the extracted sample was ethylated using sodium tetraethyl-borate (NaTEB) at pH 4.9 (using 225 μl 2M acetate buffer). The resulting data was manually adjusted in Excel to determine the MeHg peak area.
The concentration of MeHg for each sample was subsequently calculated using mass-bias (MB) corrected signals derived through signal deconvolution. 14 Three mass-bias vials, each containing 0.5 ppt ambient Hg ethylated in sodium tetraethyl-borate (NaTEB) at a pH of 4.9 (using 225 μl 2M acetate buffer), were analyzed and MeHg concentration for each sample adjusted by the calculated MB correction factor. Replicate testing was conducted using three peat-rich samples from the topmost sample horizons of the Alvi collapse fen. The mean (in ng MeHg/g soil) and %RSD of the replicates was: 4.30 ± 9.12%, 2.74 ± 10.40%, and 1.20 ± 10.74%. Five blanks, containing only reagents, were tested concurrently with the sampled material to ensure no contamination of MeHg occurred during the extraction process. Certified reference material (ERM-CC580, estuarine sediment) analyzed were on average 110% of the certified value (75 ± 4 ng g -1 ). All MeHg and %MeHg scatterplots were created in R. 12 6      As noted in the main text, the cores were transported from the field to the lab at temperatures above the ambient soil temperatures (Table S1), but these increased temperatures are unlikely to explain any of the trends observed. We base this assumption on the fact that: i) we observe lower concentrations of MeHg (and %MeHg) in the deeper portion of the active layer where the differences between in-situ and transported temperatures were the greatest; ii) we observe differences in %MeHg in the upper layer of the distant and collapse fens, although these were sampled and treated in parallel, and are assumed to have similar in-situ temperature (Figures 2 and 3); and iii) we observe no artificial methylation in the control experiment done to test for artificial methylation during transport (described below). Furthermore, the cores were transported intact prior to subdivision in the lab, and if methylation was caused by processes during transport, we would expect the cores to be affected equally across depths (which they are not). For the frozen permafrost layer, "artificial" methylation caused by thaw during transport cannot be categorically ruled out. However, the concentrations of MeHg in the frozen layer were low, suggesting limited methylation during transport.

Supplementary Information for Results and Discussion
The test performed to rule out the role of Hg methylation during transportation was done using similar peat samples collected from Abisko, Sweden in 2019 (n=2, the peat material used was kept at -20°C from the sampling day until the experiment). The materials were incubated at 17°C during ~1 week and then analyzed for MeHg. No artificial methylation was detected during the incubation of the two samples. For one of the samples, the concentration of MeHg observed in the incubated sample was ~40% lower in the sample that was not incubated, suggesting some demethylation to occur. This loss is, however, not significant in relation to the 10 and 13 times higher %MeHg we observed in the collapsed fen.