Insights from mercury stable isotopes on terrestrial–atmosphere exchange of Hg(0) in the Arctic tundra

. The tundra plays a pivotal role in the Arctic mercury (Hg) cycle by storing atmospheric Hg deposition and shuttling it to the Arctic Ocean. A recent study revealed that 70 % of the atmospheric Hg deposition to the tundra occurs through gaseous elemental mercury (GEM or Hg(0)) uptake by vegetation and soils. Processes control-ling land–atmosphere exchange of Hg(0) in the Arctic tundra are central, but remain understudied. Here, we combine Hg stable isotope analysis of Hg(0) in the atmosphere, interstitial snow air, and soil pore air, with Hg(0) ﬂux measurements in a tundra ecosystem at Toolik Field Station in northern Alaska (USA). In

Abstract.The tundra plays a pivotal role in the Arctic mercury (Hg) cycle by storing atmospheric Hg deposition and shuttling it to the Arctic Ocean.A recent study revealed that 70 % of the atmospheric Hg deposition to the tundra occurs through gaseous elemental mercury (GEM or Hg(0)) uptake by vegetation and soils.Processes controlling land-atmosphere exchange of Hg(0) in the Arctic tundra are central, but remain understudied.Here, we combine Hg stable isotope analysis of Hg(0) in the atmosphere, interstitial snow air, and soil pore air, with Hg(0) flux measurements in a tundra ecosystem at Toolik Field Station in northern Alaska (USA).In the dark winter months, planetary boundary layer (PBL) conditions and Hg(0) concentrations were generally stable throughout the day and small Hg(0) net deposition occurred.In spring, halogen-induced atmospheric mercury depletion events (AMDEs) occurred, with the fast re-emission of Hg(0) after AMDEs resulting in net emission fluxes of Hg(0).During the short snowfree growing season in summer, vegetation uptake of atmospheric Hg(0) enhanced atmospheric Hg(0) net deposition to the Arctic tundra.At night, when PBL conditions were stable, ecosystem uptake of atmospheric Hg(0) led to a depletion of atmospheric Hg(0).The night-time decline of atmospheric Hg(0) was concomitant with a depletion of lighter Hg(0) isotopes in the atmospheric Hg pool.The enrichment factor, ε 202 Hg vegetation uptake = −4.2‰ (±1.0 ‰) was consistent with the preferential uptake of light Hg(0) isotopes by vegetation.Hg(0) flux measurements indicated a partial re-emission of Hg(0) during daytime, when solar radiation was strongest.Hg(0) concentrations in soil pore air were depleted relative to atmospheric Hg(0) concentrations, concomitant with an enrichment of lighter Hg(0) isotopes in the soil pore air, ε 202 Hg soil air−atmosphere = −1.00‰ (±0.25 ‰) and E 199 Hg soil air−atmosphere = 0.07 ‰ (±0.04 ‰).These first Hg stable isotope measurements of Hg(0) in soil pore air are consistent with the fractionation previously observed during Hg(0) oxidation by natural humic acids, suggesting abiotic oxidation as a cause for observed soil Hg(0) uptake.The combination of Hg stable isotope fingerprints with Hg(0) flux measurements and PBL stability assessment confirmed a dominant role of Hg(0) uptake by vegetation in the terrestrial-atmosphere exchange of Hg(0) in the Arctic tundra.

Introduction
Mercury (Hg) is a high priority pollutant that causes neurodevelopmental deficits in children and cardiovascular disease in adults (Budtz-Jørgensen et al., 2000;Roman et al., 2011).Arctic populations are particularly exposed to high Hg levels, despite few local anthropogenic Hg emission sources, due to their traditional diet consisting of high trophic level seafood (Sheehan et al., 2014).Anthropogenic Hg emissions from midlatitudes reach remote ecosystems, such as the Arctic, via the long-range transport of gaseous elemental mer-Published by Copernicus Publications on behalf of the European Geosciences Union.cury (GEM or Hg(0)) (Douglas et al., 2012).It has long been thought that springtime oxidation of Hg(0) driven by photochemically produced bromine radicals Br q on surface snow, referred to as "atmospheric mercury depletion events" (AMDEs), leads to enhanced deposition of divalent mercury (Hg(II)) in Arctic regions (Steffen et al., 2008).Such Hg(II) deposition to the snowpack during AMDEs, however, can be subject to photoreduction and fast re-emission back into the atmosphere, minimising the net load of Hg by AMDEs to snow (Douglas et al., 2012;Johnson et al., 2008).
Tundra soils play a central role in the Arctic Hg cycle by storing atmospheric Hg deposition from where it can then be mobilised and transported to the Arctic Ocean (Obrist et al., 2017;Sonke and Heimburger, 2012;Sonke et al., 2018).Over millennia, tundra vegetation and soils have drawn Hg(0) out of the atmosphere, resulting in one of the largest pools of Hg (408-863 Gg Hg, top 1 m) stored at the Earth's surface (Obrist et al., 2017;Olson et al., 2018;Schuster et al., 2018).Arctic rivers deliver 44-50 Mg a −1 of Hg to the Arctic Ocean (Dastoor and Durnford, 2014;Sonke et al., 2018).In aquatic ecosystems, Hg can be transformed to methyl mercury that bioaccumulates in the aquatic food chain, resulting in elevated Hg concentrations in high trophic level fishes and mammals (Outridge et al., 2009;Douglas et al., 2012).
On a global scale, vegetation uptake of atmospheric Hg(0) represents the dominant pathway of atmospheric Hg deposition to terrestrial surfaces, resulting in strong seasonal variation of atmospheric Hg(0) concentrations, with minima in summer when vegetation activity is highest (Jiskra et al., 2018).Direct Hg(0) flux measurements over selected surfaces (e.g.soil, snow, or leaves), however, do not always provide conclusive answers about the direction and magnitude of Hg exchange between terrestrial ecosystems and the atmosphere.For example, a review of 132 terrestrialatmosphere Hg(0) flux studies conducted over various surfaces over the last 30 years estimated a wide range of net fluxes in the range of −513 to 1650 Mg a −1 (37.5th to 62.5th percentile) (Agnan et al., 2016).In recent years, a number of studies using Hg(0) flux measurements based on micrometeorological methods directly quantified net ecosystem exchange fluxes of Hg(0) over terrestrial ecosystems (i.e. at the ecosystem level), including atmosphere-vegetation exchange and underlying soil/litter contributions (Lindberg et al., 1998;Fritsche et al., 2008;Bash and Miller, 2009;Castro and Moore, 2016;Osterwalder et al., 2017).Measurements of multi-level Hg(0) gradients and interstitial snow air and soil pore air provided additional constraints on the terrestrial surface exchange flux (Sigler and Lee, 2006;Moore and Castro, 2012;Faïn et al., 2013;Obrist et al., 2014;Fu et al., 2016b;Agnan et al., 2018).
This paper is part of a larger study aiming to better understand the fate of Hg in Arctic tundra ecosystems, centred around a 2-year field campaign on the Arctic Coastal Plain of Alaska.In Obrist et al. (2017), we performed a 2-year mass balance of terrestrial-atmosphere exchange over the Arctic tundra.We also investigated the spatial distribution of Hg in tundra soils (Olson et al., 2018), and spatial and temporal patterns of Hg in snow (Agnan et al., 2018) and in vegetation (Olson et al., 2019).In our previous work, we showed that the uptake of atmospheric Hg(0) by vegetation and soil represents 70 % of total atmospheric deposition and has led to high Hg levels in Arctic soils (Obrist et al., 2017(Obrist et al., , 2018)).In this study, we explore the use of novel Hg stable isotope measurements of Hg(0) in interstitial snow air and soil pore air to identify the processes driving tundra Hg(0) deposition.We further discuss the effects of terrestrial-atmosphere exchange processes and planetary boundary layer stability on the Hg(0) concentration and Hg stable isotope signature measured in the atmosphere.

Study site
The study was conducted at Toolik Field Station (68 • 38 N, 149 • 36 W) on the Arctic Coastal Plain in northern Alaska, USA, 180 km inland from the Arctic Ocean coast.All measurements were conducted on an acidic tussock tundra, on Aquiturbels soils with an active layer of 60-100 cm (Obrist et al., 2017).The climate of Toolik Field Station is char-acterised by low mean annual temperatures of −8.5 • C and a mean annual precipitation of 312 mm a −1 (Cherry et al., 2014).During the 1-year Hg isotope campaign from October 2015 to September 2016, the tundra was snow-covered for a total of 248 d (Agnan et al., 2018), leading to a relatively short snow-free growing season.

Hg stable isotope sampling and measurement
Hg(0) was continuously sampled in the atmosphere (0.3 and 2 m above ground), in interstitial air of surface snow (0 and 0.1 m above ground), and in soil pore air (0.4 m below ground) at low flow rates of 0.2 L min −1 .Interstitial snow air for Hg(0) stable isotope measurements was sampled from a dedicated snow tower adapted from Seok et al. (2009), consisting of an aluminium construction with three gas inlets on horizontal bars at 0, 0.1, and 0.3 m above ground (Fig. S1a in the Supplement).Two soil wells as described in Obrist et al. (2017) were dedicated to Hg(0) stable isotope sampling of soil air.Each soil well had two gas inlets, positioned 1.5 m apart at 0.4 m depth, each inlet consisting of a 47 mm single stage filter assembly (Savillex, Eden Prairie, USA) with a Teflon ® filter membrane (Fig. S1b).
Hg(0) was trapped on iodated activated carbon (IAC) traps (Brooks Rand, 0.1 g in custom made 12 cm long glass tubes with an inner diameter of 4 mm, Fu et al., 2014) and samples were changed manually during site visits every 6 to 8 weeks.During site visits in March and June 2016, higher temporal resolution (2-4 d) sampling was conducted with higher flow rates of 1.5-2 L min −1 .Overall, total volumes of sampled air per sample ranged from 5.7 to 17.7 m 3 (Table S1 in the Supplement).During the growing season (June to September 2016), diel variation of atmospheric Hg(0) was assessed using two parallel sampling lines operated with a time switch.Daytime samples were collected from 06:00 to 22:00 AKST (Alaska standard time -used for all times referenced throughout the paper), and night-time samples were collected from 22:00 to 06:00.Of the 14 soil pore air samples taken during the 1-year Hg isotope campaign, only three samples contained sufficient Hg (> 2.5 ng) for isotopic analysis.The reason for this was that the Hg(0) in soil pore air was largely depleted (below the detection limit of ≈ 0.1 ng m −3 ) (Obrist et al., 2017;Agnan et al., 2018), in particular during winter months, making isotopic characterisation of the soil pore air Hg(0) pool impossible during this time period.
The protocol for Hg stable isotope measurements of Hg(0) was adapted from Fu et al. (2014).We used lower amounts of IAC trap material (0.1 g) to reduce possible matrix effects during cold-vapour generation.Breakthrough was tested in the lab and under field conditions by connecting a Tekran 2537 after the IAC trap, and Hg(0) measurements were always below the detection limit (< 0.1 ng m −3 ).IAC traps were combusted in a two-stage oven system and Hg was recovered in a 4.2 N HNO 3 , 1.2 N HCl oxidising acid trap.
Hg stable isotope ratios of trap solutions were measured by cold-vapour separation multi-collector inductively coupled plasma mass spectrometry (CV MC-ICP/MS) at the Observatory Midi-Pyrénées, Toulouse, France (Jiskra et al., 2019;Sun et al., 2013).Six process blanks and one field blank were measured during the sample processing and were 0.25 ng Hg/trap (max 0.85 ng Hg/trap) on average, representing 1 %-2 % of the typical Hg amounts collected during sampling periods.The amounts of Hg collected on IAC-traps sampling atmospheric Hg(0) were compared to Hg(0) concentration measurements using a Tekran 2537 and revealed sample yields of 107 ± 23 % (mean ± 1 SD, n = 22).
Mass-dependent fractionation (MDF) of Hg stable isotopes is reported in small delta notation (δ) in per mille (‰) deviation from to the reference NIST 3133 Hg standard: where "xxx" refers to measured isotope masses : 199, 200, 201, 202, and 204.Mass-independent fractionation (MIF) is reported in capital delta notation ( ), which is defined as the difference between the measured δ 199 Hg, δ 200 Hg, δ 201 Hg, and δ 204 Hg values and those predicted for MDF relative to δ 202 Hg using the kinetic MDF law: where SF xxx is the mass-dependent scaling factor of 0.252 for 199 Hg, 0.502 for 200 Hg, 0.752 for 201 Hg, and 1.493 for 204 Hg (Blum and Bergquist, 2007).Hg isotope enrichment factors associated with two pools (ε xxx Hg pool 1−pool 2 and E xxx Hg pool 1−pool 2 ) were calculated from the difference in the MDF and MIF signatures between two pools (pool 1 and pool 2) as follows: The MDF enrichment factors of a reaction (ε xxx Hg reaction ) were determined by fitting a linear regression model (lm function of R) to the observational data following Mariotti et al. (1981): where δ xxx Hg residual corresponds to the Hg isotope signature of the residual Hg(0), δ xxx Hg 0 corresponds to the initial Hg(0) isotope signature, and f corresponds to the fraction of Hg(0) remaining in the gas phase.Note that for high f > 0.4 the systematic error of this simplified approach is minimal.

Hg(0) flux measurements
Micrometeorological flux measurements to quantify Hg(0) exchange at the ecosystem level were conducted using the aerodynamic gradient flux method.Surface-atmosphere flux was calculated by measurement of concentration gradients in the atmosphere above the tundra in conjunction with atmospheric turbulence parameters as follows: where k denotes the von Kármán constant (0.4), u * is the friction velocity, z is the measurement height, φ h is the Monin-Obukhov scaling coefficient (dimensionless) (Monson and Baldocchi, 2014), and ∂c(Hg(0))/∂z is the vertical Hg(0) gas concentration gradient.Hg(0) concentrations at heights of 61 and 363 cm above the soil surface were measured through 0.2 µm Teflon ® inlet filters connected to ∼ 35 m of perfluoroalkoxy polymer (PFA) lines.A valve control system with three-way solenoid valves (NResearch, West Caldwell, NJ, USA) allowed switching between the gradient inlets to take place every 10 min.A set of trace gas analysers with a total sampling flow of 1.5 L min No gap filling was performed.For quality control, sampling line blanks and line intercomparisons, where the two gradient lines were put on the same height, were performed approximately every 6 to 8 weeks (Obrist et al., 2017).The planetary boundary layer (PBL) stability was assessed via the stability index (ζ , dimensionless), defined as In the above-mentioned equation, z represents the height of the Metek USA-1 sonic anemometer (Metek GmbH, Elmshorn, Germany), i.e. 236 cm above ground, and L represents the Monin-Obukhov length.The PBL was considered stable when ζ > 0.1, unstable when ζ < −0.1, and neutral when −0.1 < ζ < 0.1 (Peichl et al., 2013).During the sampling period, auxiliary variables showed the following daily average values: air temperature of −7.4 • C (from −40.6 to 20.4 • C), relative humidity of 74 % (from 37 % to 98 %), and a wind speed of 2.36 m s −1 (from 0 to 7.82 m s −1 ).

Backward trajectory modelling
Backward trajectories of air masses were modelled using the HYSPLIT transport and dispersion model (Stein et al., 2015) from the NOAA Air Resources Laboratory, which is accessible via the READY website (Rolph et al., 2017).

Results and discussion
We divide the presentation and discussion of results into three seasons of the year.In winter ) and occasional atmospheric mercury depletion events (AMDEs) were detected at the study site (Obrist et al., 2017;Agnan et al., 2018).During summer (3 May-9 September 2016), the air temperature was above freezing (mean = 6.7 • C, hourly max = 25.1 • C), solar radiation was high (mean = 0.19 kW m −2 , hourly max = 0.80 kW m −2 ), and the study site was predominantly free of snow.

Winter
Over the winter period, atmospheric Hg(0) concentrations and CO 2 mixing ratios were relatively constant and there was little diel variation (Figs.1a, S2).Low solar radiation led to relatively stable PBL conditions throughout the day (Fig. 1b).Hg(0) flux measurements revealed a small deposition (−0.34±5.8ng m −2 h −1 , mean ±1 SD, median = −0.2ng m −2 h −1 ) (Obrist et al., 2017).The Hg(0) net deposition flux is supported by observed depletions of atmospheric Hg(0) in interstitial snow air (0.69 ± 0.22 ng m −3 , mean ±1 SD, 0 and 0.1 m sampling height) relative to atmospheric levels (1.07±0.04ng m −3 -mean ±1 SD) (p = 0.02, two-sided t test), implying a net sink of atmospheric Hg(0) in the ecosystem (Fig. 2c).A wintertime Hg(0) sink can either occur due to (i) Hg(0) deposition to Arctic snow, (ii) uptake by soil or litter, or (iii) assimilation by vegetation still active under the snowpack.Depletion of atmospheric Hg(0) in interstitial snow air was associated with an increase in δ 202 Hg (1.08 ‰ ± 0.20 ‰ versus 0.77 ‰ ± 0.16 ‰ in ambient Hg(0), mean ± 1 SD, p = 0.02, two-sided t test) (Fig. 2a) and a decrease in 199 Hg (−0.31 ‰ ± 0.05 ‰ versus −0.23 ‰ ± 0.06 ‰ in ambient Hg(0), mean ± 1 SD, p = 0.04, two-sided t test) (Fig. 2b).Both Hg(0) dry deposition to surface snow (Douglas and Blum, 2019) and by vegetation uptake (Demers et al., 2013;Enrico et al., 2016;Obrist et al., 2017) have been reported to discriminate heavier Hg(0) isotopes, consistent with these observations in the interstitial snow air.From a mass balance perspective, however, Hg(0) dry deposition to snow is considered to only play a minor role in the interior Arctic tundra.For example, using snow data in Agnan et al. (2018), we calculated a total seasonal snow Hg pool of only 50 ng m −2 at Toolik Field Station.Assuming that all this Hg in the snow was originating from the dry deposition of Hg(0), this would account for < 10 % of the Hg(0) deposition during the snow-covered period (total of 2.4 µg m −1 a −1 ; Obrist et al., 2017).In contrast, Douglas and Blum (2019) recently suggested that Hg(0) dry deposition to snow was the major source of Hg in meltwater collected on the coast of the Arctic Ocean close to Utqiagvik (former Barrow), ∼ 400 km north-west of Toolik Field Station.In snow on the coast of the Arctic Ocean, concentrations of halogens, which are considered to mediate reactive Hg(0) uptake, are elevated compared with inland sites (Douglas and Sturm, 2004;Agnan et al., 2018;Douglas et al., 2017), leading to much higher snow Hg pools (> 2000 ng m −2 ) in coastal snowpacks (Douglas et al., 2017).
A major wintertime Hg(0) deposition pathway to soils would be inconsistent with the observed enrichment in heavier Hg(0) isotopes in the interstitial snow air (Fig. 2a), as Hg(0) oxidation by humic acids in soils would lead to more negative δ 202 Hg values (Zheng et al., 2018) (see discussion below in Sect.3.3.3).Hg(0) re-emission from snow was insignificant in winter due to the absence of sunlight, as indicated by lower Hg(0) concentrations in interstitial snow air than in the atmosphere.By excluding snow and soil uptake as dominant deposition pathway, we infer that the uptake of Hg(0) by ground vegetation and/or litter on the soil surface may be driving Hg(0) deposition during the winter months.MDF and MIF signatures observed in interstitial snow air Hg(0) would be consistent with Hg(0) uptake by lichen.Hg isotope signatures of lichen measured at the same site (δ 202 Hg = −0.80‰ ± 0.20 ‰, 199 Hg = 0.20 ‰ ± 0.21 ‰, mean ± 1 SD, n = 12) were in the opposite direction of interstitial snow air Hg(0) (Olson et al., 2019).Lichen have been reported to actively exchange CO 2 for photosynthesis under snow cover (Kappen, 1993), and could possibly also take up atmospheric Hg(0) during winter months.

Spring
During spring 2016, two major AMDEs occurred (events 1 and 3; Fig. 3).During both events air masses were transported from the Arctic coast or the Arctic Ocean to Toolik Field Station, as inferred from HYSPLIT backward trajectory modelling.During the 19 March-20 March 2016 period (event 1 in Fig. 3), atmospheric Hg(0) concentrations dropped below the detection limit (< 0.1 ng m −3 ), while atmospheric Hg(II) concentrations remained low (< 0.1 ng m −3 ), with the exception of individual spikes of up to 0.4 ng m −3 .During the 1-3 April 2016 period (event 3 in Fig. 3), atmospheric Hg(0) concentrations decreased below the detection limit at times, while Hg(II) concentrations remained high (around 0.4 ng m −3 ) for 2 days.During the similar AMDEs 1 and 3, O 3 mixing ratios dropped below 10 ppb.Van Dam et al. (2013) previously reported that during AMDEs ozone depletion events (ODEs) also occurred at Toolik Field Station, concurrent AMDEs and ODEs were also observed on the coast at Utqiagvik ∼ 400 km to the north-west.Van Dam et al. (2013) suggested that AMDEs and ODEs were driven by bromine emissions from the Arctic Ocean and transported to different extents to the interior tundra.During event 1, no elevated Hg(II) concentrations were observed; thus, we assume that air depleted in Hg(0) was transported to Toolik Field Station, while the deposition of Hg(II) likely occurred closer to the coast.During the 26-29 March 2016 period (event 2 in Fig. 3), when air masses were coming from the south, atmospheric Hg(0) concentrations temporally decreased to 0.75 ng m −3 , O 3 remained high (> 30 ppb), and Hg(II) concentrations remained enhanced around 0.2 ng m −3 for 2 days while turbulent PBL conditions prevailed.The source of these enhanced Hg(II) levels transported from the Brooks Range mountains remains unknown.
Snowmelt occurred in May in 2016, during which the snow height declined rapidly between 7 May (24 cm) and 13 May (0 cm) and a small Hg(0) deposition possibly occurred (−0.24 ± 3.9 ng m −2 h −1 , mean ± 1 SD, median: 0.08 ng m −2 h −1 ).In contrast to wintertime patterns, the interstitial snow air Hg(0) during snowmelt showed low 199 Hg values of −0.62 ‰ and −0.44 ‰ (n = 2) versus −0.23 ‰ ± 0.06 ‰ in ambient Hg(0) (Fig. S3b).The negative 199 Hg values in Hg(0) suggest a substantial contribution of Hg(0) re-emission after photoreduction of Hg(II) in snow, which exhibited negative 199 Hg values with a minwww.biogeosciences.net/16/4051/2019/Biogeosciences, 16, 4051-4064, 2019  imum of −1.37 ‰ (Obrist et al., 2017).This observation is consistent with previous chamber experiments, where a negative 199 Hg value of −2.08 ‰ was reported for Hg(0) reemission from snow (Sherman et al., 2010).Even-MIF ( 200 Hg) is considered to be unaffected by post-deposition processes such as re-emission (Sherman et al., 2010;Enrico et al., 2016), providing a conservative tracer for the pathway of atmospheric Hg deposition. 200Hg values measured in snow impacted by AMDEs at Toolik Field Station and other sites in Alaska (Obrist et al., 2017;Sherman et al., 2010Sherman et al., , 2012a) ) are similar to the 200 Hg values of atmospheric Hg(0) (Fig. 4) (−0.06 ‰ ± 0.06 ‰ versus −0.05 ‰ ± 0.04 ‰, mean ± 1 SD).This similarity can be explained by a quantitative oxidation of atmospheric Hg(0) to Hg(II) (e.g.event 1 in Fig. 3) that is deposited to snow.Thus, Hg(II) in snow inherits the isotopic composition of the source Hg(0) due to conservation of mass, irrespective of the isotopic fractionation factor associated with Hg(0) oxidation.Several samples exhibited 200 Hg values between that of atmospheric Hg(0) and Hg(II) in precipitation measured in temperate regions (Fig. 4).This intermediate 200 Hg signature can be explained by either AMDEs with non-quantitative oxidation due to limited Br oxidant availability (e.g.event 2 in Fig. 3), or a mixing of AMDE-derived Hg(II) with Hg(II) present in the overlying Arctic free troposphere.Even-MIF ( 200 Hg) has been suggested as a promising tracer to distinguish between atmospheric deposition of Hg(II) in precipitation, which exhibits positive 200 Hg anomalies, and direct Hg(0) deposition (e.g.uptake by vegetation), which exhibits small negative 200 Hg (Enrico et al., 2016;Sun et al., 2019).We caution that the presence of AMDEs complicates the use of 200 Hg for mixing model-based Hg deposition calculations in the Arctic (Obrist et al., 2017).

Drivers of diel cycling in atmospheric Hg(0)
Figure 5 represents a time series of the atmospheric Hg(0) concentration, the CO 2 mixing ratio, the PBL stability, and Hg(0) fluxes during midsummer (15-28 July 2016).Atmospheric Hg(0) concentrations generally declined during each night, and the strongest Hg(0) depletions (Hg(0) < 1 ng m −3 ) were observed when the PBL was stable (ζ > 0.1, events 2 and 3 and green bars in Fig. 5b).These Hg(0) night-time www.biogeosciences.net/16/4051/2019/Biogeosciences, 16, 4051-4064, 2019 minima coincided with maxima of atmospheric CO 2 mixing ratios of 410 to 420 ppm.These patterns are consistent with measured Hg(0) deposition fluxes during nights (daily minima at 00:00, −2.56 ± 0.22 ng m −2 h −1 (mean ± 1 SD), Fig. 1i), under corresponding CO 2 accumulation in the PBL driven by night-time CO 2 soil respiration (Wofsy et al., 1993;Schlesinger and Andrews, 2000;Grant and Omonode, 2018).On nights with unstable PBL conditions (e.g.event 1 in Fig. 5), diel Hg(0) and CO 2 variations were lower or absent due to increased mixing with free tropospheric air containing background levels of Hg(0) and CO 2 .During late summer (Fig. S4, 20-31 August 2016), the longer duration of stable nocturnal PBL conditions led to even more pronounced night-time depletions in Hg(0).During daytime under strong solar radiation, flux measurements showed a net Hg(0) emission around noon (daily maxima at 11:00, 2.9 ± 0.92 ng m −2 h −1 , mean ± 1 SD, Fig. 1i).Daytime Hg(0) reemission, however, did not lead to a build-up of atmospheric Hg(0) above the surface due to prevailing turbulent conditions, allowing for the efficient mixing with background free tropospheric air.These patterns demonstrate how atmospheric Hg(0) and CO 2 are both controlled by the magnitude and direction of the net ecosystem exchange fluxes in conjunction with the PBL stability.Overall, flux measurements showed the tundra ecosystem to be a net sink of atmospheric Hg(0) over the duration of the growing season (−0.76 ± 6.3 ng m −2 h −1 -mean ±1 SD, median: −1.2 ng m −2 h −1 ).
At Toolik Field Station, the difference between δ 202 Hg in vegetation relative to atmospheric Hg(0) was also considerably lower (range of −1.29 ‰ to −2.09 ‰, depending on vegetation species; Olson et al., 2019) than the fractionation factor derived from the atmospheric pattern.This discrepancy can be explained by the fact that δ 202 Hg signatures measured in vegetation not only reflect the isotopic fractionation during foliar uptake, but also contain a re-emission component.Hg(II) reduction is expected to lead to more positive δ 202 Hg values in the residual, foliar Hg(II) fraction, irrespective of the reduction mechanism (Bergquist and Blum, 2007;Zheng and Hintelmann, 2010;Kritee et al., 2007;Jiskra et al., 2015).Re-emission of foliar Hg is supported by the observed negative shifts in odd-mass isotope MIF ( 199 Hg) in vegetation relative to 199 Hg of atmospheric Hg(0), which have been observed at Toolik Field Station (Olson et al., 2019) and elsewhere (Enrico et al., 2016;Demers et al., 2013).Positive 199 Hg in atmospheric Hg(0) re-emitted from foliage has recently been constrained by flux bag experiments (Yuan et al., 2018).Therefore, it is expected that the fractionation factor of foliar uptake is larger than just the difference between δ 202 Hg in foliage and Hg(0) in the atmosphere.Instead, the difference reflects a net fractionation consisting of the isotopic fractionation during foliar uptake, as well as during foliar reduction and re-emission.Our observation that the fractionation factor derived from δ 202 Hg of atmospheric Hg(0) (Fig. 6) is larger than the difference of δ 202 Hg between vegetation and the atmosphere could also be associated with the diel variation of Hg(0) fluxes and PBL dynamics.During daytime, atmospheric turbulence is higher and, therefore, local signals of terrestrial re-emission are expected to be diluted by mixing with background Hg(0).At night, when the PBL is stable, foliar uptake of lighter Hg(0) isotopes is imprinted on the residual atmospheric Hg(0).

Conclusions
We document that the concentrations and the isotopic composition of atmospheric Hg(0) are strongly affected by terrestrial-atmosphere exchange, in particular by vegetation uptake, of Hg(0).While directions and magnitudes of the terrestrial-atmosphere exchange vary with season and time of the day, atmospheric stability and the dynamics of the PBL strongly affect atmospheric Hg(0) concentrations.During the vegetation period, atmospheric Hg(0) isotope systematics were dominated by vegetation uptake of Hg(0) discriminating heavy Hg(0) isotopes in the residual atmospheric pool, which manifested itself most strongly during stable PBL conditions at night.We found a larger discrimination of heavier Hg(0) isotopes during foliar uptake when deriving a fractionation factor from atmospheric Hg(0) observations, compared with deriving this factor based on the difference measured between bulk Hg in vegetation and atmospheric Hg(0).While this discrepancy is not fully understood, it may be attributed to photoreduction and re-emission of lighter Hg(0) isotopes during the day.Hg(0) concentrations were continuously depleted in interstitial snow and soil pore air.Hg sta- ble isotope measurements of Hg(0) in soil air indicated that the soil Hg(0) sink was driven by Hg(0) oxidation by natural organic matter.However, we observed no isotopic traces of this Hg(0) soil sink in interstitial snow air and the atmosphere above and concluded that soil uptake of Hg(0) made only a minor contribution to the net ecosystem exchange of Hg(0).Based on the stable isotope signatures of Hg(0) in soil and snow pore air and mass balance considerations, we infer that the uptake of Hg(0) by ground vegetation and/or litter on the soil surface was likely responsible for Hg(0) deposition during the winter months.The combination of stable isotope fingerprints with Hg(0) flux measurements and PBL stability assessment confirmed a dominant role of Hg(0) uptake by vegetation in Arctic tundra Hg cycling.
Atmospheric temperatures have increased twice as fast in the Arctic as in temperate regions over the last 2 decades (Cohen et al., 2014).The increase in Arctic temperatures has been accompanied by an earlier snowmelt in spring and, thus, a longer vegetation period and higher maximal greenness of Arctic vegetation (Box et al., 2019).Vegetation uptake is estimated to currently drive the deposition of 210 Mg Hg a −1 of atmospheric Hg(0) to the Arctic tundra (Obrist et al., 2017;Sonke et al., 2018).With an ongoing warming and greening trend associated with climate change, an amplification of the Arctic tundra Hg(0) vegetation pump can be expected.Contrarily, higher soil temperatures, leading to permafrost thaw and degradation and development of thermokarsts, increase the risk of the re-mobilisation of large amounts of Hg that is currently stored in Arctic tundra soils (St. Pierre et al., 2018;Olson et al., 2018;Schuster et al., 2018).The net effect of these complex alterations in Arctic Hg cycling driven by climate change is currently poorly understood and the risk needs to be assessed.

Figure 1 .
Figure 1.Diel variation during winter (a, b, c), spring (d, e, f), and summer (g, h, i): (a, d, g) average Hg(0) concentration (in red) and the CO 2 mixing ratio (in blue) in the atmosphere (2 m height).(b, e, h) The median planetary boundary layer stability parameter (ζ , grey), where positive values represent stable conditions, and solar radiation is shown in yellow.(c, f, i) The average Hg(0) flux (in red) and the average CO 2 flux (in blue).The shaded areas in (a) and (c) represent the mean ± 1 SD interval of concentration and flux measurements, respectively.

Figure 2 .
Figure 2. Hg(0) measurements in interstitial snow air profiles.(a) Mass-dependent fractionation stable isotope signature of Hg(0) (δ 202 Hg), (b) mass-independent fractionation stable isotope signature of Hg(0) ( 199 Hg), and (c) mean Hg(0) concentration.Note that the Hg(0) concentration marked with an asterisk ( * ) was calculated from Hg recovered on the IAC traps, whereas other Hg(0) concentration profiles were measured by an automated trace gas system deployed in the snowpack(Agnan et al., 2018).The dashed horizontal lines represent the average snow height during the respective period.

Figure 5 .
Figure 5.Time series during a midsummer period (15-28 July 2016) of the (a) atmospheric Hg(0) concentration and CO 2 mixing ratio; (b) median planetary boundary layer stability (ζ ), where the shaded areas in green represents stable conditions (ζ > 0.1) and shaded areas in red represent turbulent conditions (ζ < −0.1); (c) solar radiation and air temperature; and (d) Hg(0) flux, with Hg(0) deposition in green and Hg(0) re-emission in red.Midnight is indicated by dashed lines.

Figure 6 .
Figure 6.Mass-dependent Hg isotope signature (δ 202 Hg) of atmospheric Hg(0) versus Hg(0) concentration during the snow-free growing period (11 June-10 September 2016).The dashed line represents a non-linear Rayleigh fit and the shaded area is the 95 % confidence interval (see main text).Error bars represent the analytical precision determined by the 2 SD from multiple measurements of an in-house standard.

Figure 7 .
Figure 7. Hg stable isotope systematics of Hg(0) in the atmosphere during summer/fall and in soil pore air measured during the same time period.(a) Mass-dependent Hg isotope signature (δ 202 Hg) against Hg(0) concentration.(b) Mass-independent Hg isotope signature ( 199 Hg) against mass-dependent Hg isotope signature (δ 202 Hg).The dashed green lines represent the regression of a Rayleigh model (a) and a linear model (b).The straight red lines represent the trajectories for abiotic dark oxidation of Hg(0) by natural humic acids (HAs) from Zheng et al. (2018).For comparison with observations, the intercept of the linear regressions was adjusted to fit through the average of atmospheric Hg(0).The shaded areas represent the 95 % confidence interval.Error bars represent the analytical precision determined by the 2 SD from multiple measurements of an in-house standard.

Figure 8 .
Figure 8. Overview of the mass-independent Hg isotope signature ( 199 Hg) against the mass-dependent Hg isotope signature (δ 202 Hg) of atmospheric Hg(0) (red) and Hg(0) in the pore air of snow (blue) and soil (green) during winter (circles), the AMDE season (squares), and summer/fall (triangles).Bulk vegetation measurements (green diamonds) are reproduced from Obrist et al. (2017).The straight arrow represents the fractionation during Hg(0) uptake by vegetation, and the dashed arrow represents the expected development of the corresponding residual Hg(0) pool.Error bars represent the analytical precision determined by the 2 SD from multiple measurements of an in-house standard.