Air – surface exchange of gaseous mercury over permafrost soil : an 1 investigation at a high-altitude ( 4700 m a . s . l . ) and remote site in the central 2 Qinghai-Tibet Plateau

The pattern of air–surface gaseous mercury (mainly Hg(0)) exchange in the Qinghai-Tibet 9 Plateau (QTP) may be unique because this region is characterized by low temperature, great 10 temperature variation, intensive solar radiation, and pronounced freeze-thaw process of permafrost soils. 11 However, air–surface Hg(0) flux in the QTP is poorly investigated. In this study, we performed field 12 measurements and controlled field experiments with dynamic flux chambers technique to examine the 13 flux, temporal variation and influencing factors of air–surface Hg(0) exchange at a high-altitude (4700 14 m a.s.l.) and remote site in the central QTP. The results of field measurements showed that surface soils 15 were net emission source of Hg(0) in the entire study (2.86 ng m -2 h -1 or 25.05 μg m -2 y -1 ). Hg(0) flux 16 showed remarkable seasonality with net high emission in the warm campaigns (June 2014: 4.95 ng m -2 17 h -1 ; September 2014: 5.16 ng m -2 h -1 and May–June 2015: 1.95 ng m -2 h -1 ) and net low deposition in 18 winter campaign (December 2014: –0.62 ng m -2 h -1 ), and also showed the diurnal pattern with emission 19 in daytime and deposition in nighttime, especially on days without precipitation. Rainfall events on the 20 dry soils induced large and immediate increase in Hg(0) emission. Snowfall events did not induce the 21 pulse of Hg(0) emission, but snow melt resulted in the immediate increase in Hg(0) emission. Daily 22 Hg(0) fluxes on rainy or snowy days were higher than those of days without precipitation. Controlled 23 field experiments suggested that water addition to dry soils significantly increased Hg(0) emission both 24 in short (minutes) and relatively long (hours) timescales, and also showed that UV radiation was 25 primarily attributed to Hg(0) emission in the daytime. Our findings imply that a warm climate and 26 environmental change could facilitate Hg release from the permafrost terrestrial ecosystem in the QTP. 27

Abstract.The pattern of air-surface gaseous mercury (mainly Hg(0)) exchange in the Qinghai-Tibet Plateau (QTP) may be unique because this region is characterized by low temperature, great temperature variation, intensive solar radiation, and pronounced freeze-thaw process of permafrost soils.However, the air-surface Hg(0) flux in the QTP is poorly investigated.In this study, we performed field measurements and controlled field experiments with dynamic flux chambers technique to examine the flux, temporal variation and influencing factors of air-surface Hg(0) exchange at a high-altitude (4700 m a.s.l.) and remote site in the central QTP.The results of field measurements showed that surface soils were the net emission source of Hg(0) in the entire study (2.86 ng m −2 h −1 or 25.05 µg m −2 yr −1 ).Hg(0) flux showed remarkable seasonality with net high emission in the warm campaigns (June 2014: 4.95 ng m −2 h −1 ; September 2014: 5.16 ng m −2 h −1 ; and May-June 2015: 1.95 ng m −2 h −1 ) and net low deposition in the winter campaign (December 2014: −0.62 ng m −2 h −1 ) and also showed a diurnal pattern with emission in the daytime and deposition in nighttime, especially on days without precipitation.Rainfall events on the dry soils induced a large and immediate increase in Hg(0) emission.Snowfall events did not induce the pulse of Hg(0) emission, but snowmelt resulted in the immediate increase in Hg(0) emission.Daily Hg(0) fluxes on rainy or snowy days were higher than those of days without precipitation.Controlled field experiments suggested that water addition to dry soils significantly increased Hg(0) emission both on short (minutes) and relatively long (hours) timescales, and they also showed that UV radiation was primarily attributed to Hg(0) emission in the daytime.Our findings imply that a warm climate and environmental change could facilitate Hg release from the permafrost terrestrial ecosystem in the QTP.

Introduction
Soils represent the largest Hg reservoirs in ecosystems and play a major role in the global Hg cycle (Selin, 2009;Agnan et al., 2016).Background soils receive Hg input from atmospheric deposition, which is mainly retained in organic-rich layers of upper soils (Schuster, 1991;Khwaja et al., 2006).Hg in soils can be reduced to Hg(0) and then emitted to the overlaying air because of its high volatility (Schlüter, 2006).Therefore, soils can serve as both sources and sinks of atmospheric Hg (Pirrone and Mason, 2009;Amos et al., 2013;Agnan et al., 2016).
The Qinghai-Tibet Plateau (QTP) is located in western China with an area of 2.5 million km 2 and mean altitude of > 4000 m.Due to the high altitude and subsequent low temperature, a significant portion (∼ 1.5 million km −2 ) of the QTP is underlain by permafrost (Kang et al., 2010).Because of the harsh natural environment, limited research resources and difficulty of access and sampling logistics, studies on Hg biogeochemistry in the QTP are limited.The role of the QTP in the regional and global Hg biogeochemical cycle is poorly understood (Ci et al., 2012;Agnan et al., 2016).At present, Hg studies in the QTP have mainly focused on investigations of Hg concentration, speciation and distribution in environmental samples, such as air (Fu et al., 2008b;2012;Yin et al., 2015), snow and glaciers (Loewen et al., 2007;Wang et al., 2008;Zhang et al., 2012;Huang et al., 2012), and rain water (Huang et al., 2013).The knowledge of Hg(0) dynamics at the air-surface interface in the QTP is extremely poor.The unique climatic conditions, land cover and soil properties suggest the need for the specific air-soil Hg(0) flux data and a mechanism representative of the environmental setting in the QTP to better constrain global natural sources inventories (Ci et al., 2012;Agnan et al., 2016).
In this study, we the applied dynamic flux chamber (DFC) technique to investigate the flux, temporal variation and influencing factors of air-surface Hg(0) exchange at a representative research station in the central QTP.At the same time, controlled field experiments were performed to explore the effect of rainfall and different wavebands of solar radiation on air-soil Hg(0) flux.Combining the results of this study and other knowledge, we discuss the effect of future climatic and environmental change on air-surface Hg(0) dynamics in the QTP.

Study site
The study was performed at the Beiluhe Permafrost Engineering and Environmental Research Station affiliated with the Cold and Arid Regions Environmental and Engineering Research Institute, Chinese Academy of Sciences (CAREER-CAS).The elevation of the Beiluhe region is about 4700 to 4800 m a.s.l.The station (34 • 49 45 N, 92 • 56 06 E) lies over the continuous permafrost zone in the central QTP (Fig. 1).The terrain is undulating with sparse vegetation and surface fine sands or gravels.The thickness of the active layer and permafrost around the station is 2.0-3.2 and 60-200 m, respectively; the active layer begins to freeze in September and thaw in May (Peng et al., 2015a).The Beiluhe region experiences a continental climate with cold winter (up to −30 • C) and warm summer (up to 25 • C).The magnitude of daily air temperature is up to 30 • C, and the annual mean surface air temperature is about −2 to −3 • C (Peng et al., 2015a).The solar radiation is high and characterized by intense UV radiation (Wei et al., 2006).The mean annual precipitation is about 300 mm, and the annual potential evapotranspiration (∼ 1300 mm) greatly exceeds the precipitation (Peng et al., 2015a).As it is a remote region, there is no direct human activity to influence the local Hg cycle.
The precipitation at the Beiluhe station mainly occurs during May to October under the influence of the southern Asian monsoon (Peng et al., 2015a).Due to the highaltitude location of the Beiluhe station, snow events commonly occur in May to June and late September to October.Because of intensive solar radiation and surface temperature, the snow melts or sublimates on a short timescale (1-100 h), i.e., little/no snow accumulation occurs for a long time (> 3 day).Therefore, the Beiluhe region provides an unique opportunity to investigate the different effects of rain, snow, and snowmelt on the air-surface Hg(0) flux over different timescales.

Measurement of air-surface Hg(0) flux
The dynamic flux chamber (DFC) technique was widely used to investigate Hg(0) flux between the air and surface because it is inexpensive, portable, easy to set up and operate (e.g., Kim and Lindberg, 1995;Carpi and Lindberg, 1998;Gustin et al., 2006;Wang et al., 2006;Dommergue et al., 2007;Fu et al., 2008a;Kocman and Horvat, 2010;Edwards and Howard, 2013).Air-surface Hg(0) flux obtained by DFC technique was calculated using Eq.(1), where F is the Hg(0) flux (ng m −2 h −1 ), Q is the flushing flow through the chamber (m 3 h −1 ), A is the footprint of the chamber (m 2 ), C 0 and C i (ng m −3 ) are air Hg(0) concentrations at outlet and inlet of the chamber, respectively.Positive flux values indicate Hg(0) emission from the surface into the air; negative flux values represent Hg(0) deposition to the surface from the air.In this study, quartz chambers were constructed for measuring Hg(0) flux and exploring the effect of different rainfall depths and radiation condition on the Hg(0) flux.Quartz glass has many advantages as a construction material used in chambers for determining Hg(0) flux in background soils.First, it has high transmittance of the full spectrum of solar radiation, especially the UV waveband (Fig. S1 in Supplement).Therefore, a quartz chamber is suitable for determining the more "actual" Hg(0) flux because the short wavelength of solar radiation has been found to have an important effect on Hg(0) dynamics at the air-soil interface (Moore and Carpi, 2005;Bahlmann et al., 2006).Second, it has low po-tential for Hg(0) adsorption and is easy to clean by heating to remove Hg bonding on the surface (Ci et al., 2016a).The low systematic blank of quartz chamber is critical for investigating Hg(0) flux over background soils (Carpi and Lindberg, 1998).
Our semi-cylindrical quartz chamber was 8 cm high and 24 cm long with a footprint of 0.0384 m −2 (0.16 m × 0.24 m) and an internal volume of 2.41 L, which is similar to previous studies (Eckley et al., 2010, and references therein).The chamber had nine inlets (8 mm in diameter) and three outlets which were on the two opposite sides of the chamber.The inlet sampling tube was placed near the ground surface (3 cm above the surface) directly near the inlet of the chamber.A flushing ambient air was drawn by vacuum pump (KNF Inc., Germany) with 3.0 L min −1 (0.18 m 3 h −1 ) through the chamber.
Due to the harsh environmental conditions and the unstable power supply, use of the commercial automatic Hg analyzer (such as Tekran 2537) to conduct field measurements of Hg(0) flux is challenging in the Beiluhe station.Therefore, air Hg(0) concentrations in the both inlet and outlet of the chamber were monitored manually by a gold trap simultaneously at 2-3 h intervals (Ci et al., 2016b).The air was pumped through gold trap using air pump (KNF Inc., Germany) with 0.50 L min −1 (0.03 m 3 h −1 ).Hg(0) collected on gold traps was quantified on site by a cold vapor atomic fluorescence spectrophotometer (CVAFS, model III, Brooks Rand, USA) using two-stage gold amalgamation method (Fitzgerald and Gill, 1979;Ci et al., 2011Ci et al., , 2016a)).Gold trap efficiencies were determined in laboratory (Hg(0) concentration: 3.2 to 13.4 ng m −3 ) and field (Hg(0) concentration: ∼ 1 to 2 to 9.4 ng m −3 ), and multiple measurements using gold traps in series showed no breakthrough at the sample flow rate of 0.50 L min −1 for 5 h.The method detection limit was 0.03 ng m −3 and precision was 3 ± 2 %.
The turnover time obtained from this protocol was 0.68 min, which is similar to previous studies (Eckley et al., 2010, and references therein).The flow rates of air both the inlet and outlet of the chamber were adjusted by a needle valve and controlled by a rotameter.Prior to the measurement, the rotameters were calibrated by a mass flow meter and a volumetric gas meter.The accuracy of flow rate was ±3 %.The sampling control system was installed in the tent near the soil plot (< 2 m).
In this study, a bare soil plot of 2 m × 2 m was chosen and separated into four subplots (1 m × 1 m, labeled as Subplot A to D) to measure Hg(0) flux and further to investigate the effects of different wavelengths of solar radiation on the Hg(0) exchange.The soil parent materials in the study site is mainly composed of sand (> 99 %) (Peng et al., 2015a).Geologic unit of this soil plot is representative of major lithologic units of the Beiluhe region.The study was conducted in four campaigns (June, September, and December 2014 and May-June 2015), covering typical intra-annual meteorological condition in the study region.It is well known that the application of the chamber isolates the soil surface from air turbulence and rain, dew, frost or snow, altering many environmental parameters that influence Hg(0) production/consumption and deposition/emission at the air-soil interface (Sommar et al., 2012, and references therein).In the Beiluhe region, rain, dew, frost or snow commonly occurs on a short timescale (10 1 -10 2 min) because of the highaltitude location and great variation in temperature and unstable weather conditions.To minimize the effect of great variation in soil surface condition on Hg(0) flux measurements, when only measuring Hg(0) flux, we used the same chamber to measure Hg(0) flux in four subplots (Subplot A to D) in turn.For the rainy and snowy days, if possible, the chamber was moved to next subplot generally after the rain or snow because studies confirmed the significant influence of precipitation on Hg(0) flux on a short timescale (Lindberg et al., 1999;Gustin and Stamenkovic, 2005;Johnson et al., 2003;Song and Van Heyst, 2005;Lalonde et al., 2001Lalonde et al., , 2003;;Dommergue et al., 2003Dommergue et al., , 2007Dommergue et al., , 2012;;Faïn et al., 2007;Bartels-Rausch et al., 2008;Brooks et al., 2008;Mann et al., 2014Mann et al., , 2015)).For the days without precipitation, the chamber was moved to the next subplot before sunrise to capture the effect of frost or dew on air-surface Hg(0) flux in the light.
In this study, all materials in contact with Hg(0) were quartz, Teflon or borosilicate glass.The chambers and tubing were rigorously acid-washed (Ci et al., 2016a).The quartz chambers were heated to 650 • C for 2 h prior the measurement to further remove all Hg (Ci et al., 2016a).System blanks of the four chamber systems were systemically inspected on site by placing acid-cleaned Teflon filter beneath the chamber and routinely inspected before and after the measurements under field conditions.The overall blank results were taken as the average of all chamber blanks for each particular day or from the entire period if continuous monitoring was conducted.The blank values of the four chambers were found to be very low (mean ± SD: −0.02 ± 0.03 ng m −2 h −1 ; range: −0.20-0.07ng m −2 h −1 ) and were not significantly different (p> 0.05) throughout the entire study.We also found that the mean blank values were not significantly different (p> 0.05) from zero and were insignificant compared to the measured Hg(0) fluxes (see below); thus, Hg(0) flux data were not blank-corrected in this study.

Water addition
To explore the effect of water addition in detail, we chose another similar soil plot with homogeneous soil properties to conduct controlled field experiments in order to investigate the effect of different rainfall depths on Hg(0) flux over different timescales (from minutes to hours).The controlled experiments were performed during the May-June 2015 campaign since this period had high surface temperature and low precipitation (Fig. 2).In the Beiluhe station, hourly rainfall depth rarely exceeds 15 mm (Peng et al., 2015a).Therefore, we designed four different treatments of rainfall depth (0, 1, 5, and 15 mm).The water addition to the dry soils commenced at night (01:40 LT) on 30 May 2015 to exclude the effect of photochemical process in the first hours of experiments.We added the Milli-Q water (Hg concentration < 0.2 ng L −1 ) to the inside and outside of the chamber with a pre-cleaned plastic syringe within 10 min to simulate the three different rain depths.Four chambers were used to simultaneously measure Hg(0) flux over these four treatments for 22 h (from 01:00 to 23:00 LT) with the same protocol described in the Methods section.Hg(0) flux was measured with 20 min intervals in the first hours (from 01:00 to 04:20 LT) of the experiments to investigate the temporal variation in Hg(0) flux in the short timescale, and with 1 h intervals for the rest period of the experiments.

Different wavebands of solar radiation
The QTP is characterized by high solar radiation with intense UV radiation.We performed the controlled experiment to quantify the role of different wavebands of solar radiation (UVB, UVA and visible light) in air-soil Hg(0) flux.Four chambers with different exposure treatments were used to measure Hg(0) flux simultaneously in the daytime.Chamber A was used to measure the Hg(0) flux in the natural light.Chamber B and chamber C were covered with UVB and UV filters to remove the corresponding wavebands from the natural light.Chamber D covered with foil was used to measure Hg(0) flux in the dark.The experiments were performed in 4 days without precipitation (21-22 December 2014 and 29-30 May 2015) to exclude the effect of precipitation.Hg(0) flux triggered by UVB, UVA and visible light was equal to difference of flux between chamber A and chamber B, between chamber B and chamber C, and between chamber C and chamber D, respectively.The transmittance of UVB filter and UV filter was shown in Fig. S1.

Determination of soil Hg
Surface soil samples (0-2 cm) were collected from four soil subplots during the June 2014 campaign.Soil samples were freeze-dried and homogenized for total Hg determination using a Milestone DMA direct Hg analyzer (detection limit: 0.01 ng Hg or 0.15 µg kg −1 ) following EPA Method 7473 (Briggs and Gustin, 2013), and the analytical accuracy was 3 %.

Measurements of environmental variables
A meteorological station that located 60 m from the soil plot was used to collect the following environmental variables: air temperature ( • C), relative humidity (%), wind speed (m s −1 ), precipitation (mm), photosynthetically active radiation (PAR, µmol m −2 s −1 ), and surface soil temperature ( • C).The surface soil temperature was monitored by an ASI-111 precision infrared radiometer (Campbell Scientific Inc., USA).This sensor was installed on a bar 0.8 m above the soil surface and provided a non-contact measurement of the surface temperature.We also measured the soil temperature at 1.0 cm soil depth on the inside of the chamber, and no significant difference was found between the soil temperature in the outside and inside of the chamber.Details on the measurements of environmental variables are given in Peng et al. (2015a).The thickness of snowpack was measured manually.

Soil Hg
Soil Hg concentrations of four subplots varied from 13.11 ± 0.51 to 12.83 ± 0.81 µg kg −1 , suggesting that the study region is a typical background soil for Hg.Soil Hg concentrations and air-soil Hg(0) flux of four subplots were not statistically different (p> 0.05), indicating the properties of the soil plot were homogeneous.

Air
Figure 2 shows that the Hg(0) flux was highly variable.The highest Hg(0) emission fluxes of 28.46 ng m −2 h −1 were observed at 13:00-15:00 LT on 13 September 2014 after a rain event on dry soils.The highest Hg(0) deposition fluxes of −6.24 ng m −2 h −1 were observed at nighttime over the cold and dry surface soil during the December 2014 campaign (03:30-06:30 LT, 24 December 2014).
Hg(0) flux generally showed a diurnal pattern with high emission in the daytime and remarkable deposition in nighttime, especially on days without precipitation (Fig. 2).Many studies have confirmed that solar radiation is one of the most important drivers for soil Hg(0) emission (Xin and Gustin, 2007;Choi and Holsen, 2009;Kocman and Horvat, 2010); high surface temperature also facilitates Hg(0) production and subsequent emission (Park et al., 2014).Therefore, the two environmental variables jointly regulate the diurnal pattern of Hg(0) flux.An in-depth discussion on synergistic effects of solar radiation and surface temperature on Hg(0) flux is provided below.Interestingly, the diurnal pattern of Hg(0) flux of each day during the December 2014 campaign was almost identical, which may be associated with very similar weather conditions throughout the entire campaign.

Effect of precipitation (rain and snow)
Firstly, the effect of rain events on Hg(0) flux was investigated.We found that the Hg(0) emission flux increased immediately following the rainfall (Fig. 2), which is consistent with many other studies (Gustin and Stamenkovic, 2005;Johnson et al., 2003;Lindberg et al., 1999;Song and Van Heyst, 2005).Previous studies have suggested that the dramatic increases in Hg(0) emission may be attributed to the physical displacement of Hg(0) present in soil air and desorption of loosely bound Hg(0) on soil particles by the infiltrating water (Johnson et al., 2003;Gustin and Stamenkovic, 2005).Notably, Fig. 2 shows that the pulse of Hg(0) emission after the rainfall was also observed at nighttime (such as 0:00 to 01:00 LT on 4 September 2014).A similar phenomenon was also documented by our controlled experiments (see below).This indicates that the immediate increase in Hg(0) emission might not be controlled by photochemical processes but by physical processes.
Soil moisture condition may also significantly regulate Hg(0) flux over relatively long timescales (from hours to several days).Therefore, many experiments studied the effect of water addition on the magnitude and pattern of air-soil Hg(0) flux over different timescales (Johnson et al., 2003;Gustin and Stamenkovic, 2005).However, most of studies were performed in controlled laboratory or mesocosm settings under certain well-defined but not necessarily environment relevant conditions (Johnson et al., 2003;Gustin et al., 2004;Gustin and Stamenkovic, 2005;Song and Van Heyst, 2005;Kocman and Horvat, 2010;Corbett-Hains et al., 2012;Park et al., 2014).In this study, it was also challenging to reveal the effect of rainfall on Hg(0) flux over relatively long timescales via field measurement since the intermittent rain events occurred irregularly during the June 2014 and September 2014 campaign (Fig. 2).Therefore, we performed the controlled field experiments (Sect.2.3.1) to explore the effect of different rainfall depths on Hg(0) flux over different timescales (from minutes to hours).
The high-time-resolution measurements captured the immediate and dramatic increases in Hg(0) emission flux after the watering of dry soils (Fig. 4).The baseline Hg(0) flux of the 0 mm treatment was used as the benchmark for the different rainfall depth treatments to be compared against.Obviously, the higher amount of water addition resulted in longer duration and higher accumulative flux of Hg(0) emission pulse.The duration of the Hg(0) emission pulse for the 1 mm and 5 mm treatment was < 20 min (from 01:40 to 02:00 LT) and ∼ 40 min (from 01:40 to 02:20 LT), respectively, which was lower than that of the 15 mm treatment (∼ 80 min, from 01:40 to 03:00).The duration of Hg(0) emission for the 15 mm treatment in the daytime was also longer than that of the 1 and 5 mm treatment (Fig. 4).
As shown in Fig. 4, the cumulative flux of Hg(0) emission during the entire study period mainly included two fractions: the pulse of Hg(0) emission after the watering (i.e., emission flux by watering) and the Hg(0) emission during the daytime (i.e., emission flux by radiation).Figure 5 shows that both "emission flux by watering" and "emission flux by radi-ation" for the 15 mm treatment were significantly higher than those of the 1 and 5 mm treatment.As mentioned above, the dramatic increase in Hg(0) emission after the simulated rain can be explained by physical displacement of interstitial soil air by the infiltrating water.The long emission duration and large immediate emission flux for soil plot with high water addition can be explained by more water needing a longer time to percolate the soil column and displacing more soil Hg(0).Many previous studies have suggested that the magnitude of Hg(0) emission with a rainfall or stimulated rain depended on soil moisture condition -i.e., if the amounts of water received by the soils were less than needed to saturate, the soil surface showed an immediate increase in Hg(0) emission, and after the soil became saturated, Hg(0) emission from surface soil was suppressed (Klusman and Webster, 1981;Lindberg et al., 1999;Johnson et al., 2003;Gustin and Stamenkovic, 2005).In this study, the pulse of Hg(0) emission flux for the 15 mm treatment was significantly higher than that of the 5 and 1 mm treatment (Fig. 5).The field water capacity and bulk density of soil in the Beiluhe region is about 28 % and 1 g cm −3 (Peng et al., 2015b), indicating that the 5 mm treatment may cause the upper soil to saturate on a short timescale since the duration of water addition was short (< 10 min).However, the pulse of Hg(0) emission for the 15 mm treatment was significantly higher than that of the 5 mm treatment.The surface soils with high sand con- tent in the Beiluhe region have a high rate of water infiltration and subsequently great infiltration depth.This process potentially increases the displacement of soil Hg(0) and facilitates Hg(0) emission, as mentioned above.Therefore, under the field conditions, the duration and flux of pulse Hg(0) emission following water addition depends not only on the amount of water received and soil moisture condition but also soil texture and soil water dynamics.
The water addition also increased the Hg(0) emission in the daytime, showing more water added, longer duration of Hg(0) emission, and more Hg(0) emitted (Figs. 4 and 5).After the surface soil was visibly dry, Hg(0) flux over soil plots with water addition had no significant difference from that of the soil plot without water addition (i.e., 0 mm treatment).This result is consistent with many other controlled studies.For example, Johnson et al. (2003) and Gustin and Stamenkovic (2005) found that once the soil water content became less than saturated, Hg(0) emission flux would be significantly enhanced, especially during the daytime, and once sufficient drying occurred, the magnitude of Hg(0) emission flux tended to gradually decrease.Investigators have suggested that as the water evaporates and soil dries, capillary action drives the upward movement of water and chemicals (including Hg components) and recharges the Hg pool in surface soils (i.e., the "wick effect") and subsequently favors the Hg(0) production and emission via photochemical processes in the light (Gustin and Stamenkovic, 2005).In our study, even for the wettest soil plot (i.e., 15 mm treatment), the surface soils were visually unsaturated in the daytime because of the low water retention, high infiltration rate of local soils and intensive solar radiation.Therefore, the pattern of Hg(0) emission for soil plots with high water addition is comparable to those of the above-mentioned studies.
Secondly, the effect of snow events on Hg(0) flux were investigated.One of the most significant differences between the rainfall and snowfall on the effect of Hg(0) exchange was that the snowfall did not induce the remarkable pulse of Hg(0) emission.For example, at 10:10 LT on 11 June 2014, a heavy snowfall event occurred, which continued until 11:20 LT.The highest thickness of the snowpack reached ∼ 12 cm.However, no remarkable pulse of Hg(0) emission was observed during the snowfall.Instead, the Hg(0) dynamics at the air-snow interface showed a clear diurnal pattern with high emission in the daytime and deposition or emission, albeit rather small, in the nighttime.This finding is consistent with previous studies on the air-snow interface (Cobbett et al., 2007).It can be seen that the pattern of Hg(0) dynamics at the air-snow interface was similar with that at the air-soil interface, indicating that Hg(0) emission from surface snow was also mainly regulated by photochemical processes (Ferrari et al., 2005;Dommergue et al., 2003Dommergue et al., , 2007)).However, it is well known that the snowpack is a porous matrix, and gases are subjected to diffusion in the snowpack.Therefore, our measurements of Hg(0) flux at the air-snow interface did not exclude the effect of Hg(0) dynamics at the soil-snow interface, especially the low thickness (< 12 cm) of snowpack in the study.
We found that the snow melting led to the remarkable peak of Hg(0) emission.For example, during the sunrise of 12 June 2014, a precipitation event with rain and snow caused the snowpack (12 cm) to melt suddenly and completely (i.e., bare soil with no surface snow) and a pulse of Hg(0) emission (∼ 8 ng m −2 h −1 ) was observed, which was the largest Hg(0) emission flux during the June 2014 campaign.We supposed that the great increase in Hg(0) emission by snowmelt in this study was consistent with the effect of rainfall, i.e., the displacement of soil Hg(0) during the snowmelt permeation of the soil column resulted in the dramatic increase in Hg(0) emission.At present, the study on the effect of snowpack melting on Hg(0) emission is limited.Cobbett et al. (2007) also found the remarkable increase in Hg(0) emission in Canadian Arctic during the snowmelt, although the Hg(0) flux was relatively small.
Finally, the effect of precipitation (including rain and snow) on daily Hg(0) flux was investigated.The abovementioned results and discussion suggest that the precipitation has great potential to facilitate soil Hg(0) emission over different timescales via physical, chemical and biological processes.The main processes include the displacement of soil Hg(0) by water, the "wick effect" to increase the photoreducible Hg(II) pool in surface soils, and the increased soil moisture to promote the biotic and abiotic reduction of Hg(II).Another well-documented process is that the atmospheric wet deposition of Hg will increase the Hg pool in surface soils and the newly deposited Hg is very active in reducing to Hg(0) (Hintelmann et al., 2002), although our study did not focus on this issue.During the June 2014 campaign, no precipitation occurred in the first 2 days (6-7 June 2014), but the rest of the days were rainy or snowy days (Fig. 2).We attempted to use the daily Hg(0) flux of the two sunny days as the benchmark to compare with those of rainy or snowy days in order to investigate the effect of precipitation on the Hg(0) flux over the timescale of one day.Figure 6 shows that the daily Hg(0) flux for rainy/snowy days was higher (ranging from 16 to 154 %) than the mean of the two sunny days.The result indicates that the precipitation increased soil Hg(0) emission on the timescale of 1 day, although the low solar radiation and temperature on rainy/snowy days would potentially decrease soil Hg(0) emission, as mentioned above.

Effect of solar radiation and soil temperature
Almost all laboratory experiments and field measurements, including this study, show that the high solar radiation and elevated soil temperature synergistically facilitate the soil Hg(0) emission (Edwards and Howard, 2013;Park et al., 2014).The following hypotheses have been proposed to explain the role of solar radiation and temperature in promoting soil Hg(0) emission: (1) solar radiation promotes the photoreduction of Hg(II) in surface soils to form Hg(0) on a short timescale, (2) solar radiation and high soil temperature reduce the apparent activation energy of Hg(0) desorption and increase Hg(0) emission from surface soils, and (3) the high soil temperature favors the Hg(0) production in soil column by biotic and abiotic processes (Carpi and Lindberg, 1998;Gustin et al., 2002).Many studies used the Arrhenius equation (Eq.2) to quantitatively investigate the relationship between soil temperature and Hg(0) flux.
where F is the Hg(0) flux (ng m −2 h −1 ), R is the gas constant, T is the soil temperature (K), A is the pre-exponential factor, and E a is the apparent activation energy.A plot of ln(F ) vs. 1/T obtains a straight line with an intercept equal to the log of the A, and the slop equal to −E a /R.Theoretically, the concept of the apparent E a refers to the thermally controlled reaction.Therefore, Hg(0) flux induced by light and precipitation should be excluded from the correlation analysis.However, in many previous studies, especially for the field measurements, the bulk Hg(0) flux in the light was generally used to explore the contribution of solar radiation or temperature to Hg(0) flux and did not isolate the respective effect of the two factors (Fu et al., 2008a).This will systematically overestimate or underestimate the contribution of solar radiation and temperature on soil Hg(0) emission depending on the source or sink of soil for Hg(0) in the dark.Only in some controlled experiments were the separated data used to explore the respective role of radiation and temperature in soil Hg(0) emission (e.g., Kocman and Horvat, 2010).In this study, for determining the respective contributions of solar radiation and temperature to the Hg(0) flux, in addition to Hg(0) flux being measured in natural light, Hg(0) flux in the dark was also measured simultaneously with a foilcovered chamber.The temperature-corrected Hg(0) flux (i.e., bulk Hg(0) flux in the light minus Hg(0) flux in the dark) in the daytime (PAR > 0) was considered to be the contribution of the solar radiation.As mentioned above, the effect of precipitation should be excluded from the data set; therefore, we only collected Hg(0) flux data on days without precipitation during the December 2014 and May-June 2015 campaign.
Figure 7 displays the temporal variation in bulk Hg(0) flux in the light, Hg(0) flux in the dark, net Hg(0) flux in the light (i.e., bulk Hg(0) flux in the light -Hg(0) flux in the dark) and the environmental variables.Clearly, changes in solar radiation had a greater influence on soil Hg(0) flux than did changes in soil temperature.The data showed that the soil served as a Hg( 0 the soils served as a very Hg(0) source at midday with relatively high soil temperature.This finding is consistent with many studies in background soils (Ericksen et al., 2006;Gustin et al., 2006;Fu et al., 2008a;Edwards and Howard, 2013).It indicates that the soil temperature plays an important role in Hg(0) dynamics at the air-soil interface, i.e., low soil temperature favors absorbing Hg(0) or reducing Hg(0) emission (e.g., Park et al., 2014, and references therein).
After the temperature was corrected, except for at midday on study days during the May-June campaign, the net Hg(0) flux in the light was higher than the bulk Hg(0) flux.A positive linear correlation was found between cumulative PAR and cumulative Hg(0) flux in the daytime, although cumulative PAR only explained ∼ 28 % of variation in cumulative Hg(0) flux in the daytime (Fig. S2).
We used the Hg(0) emission data set in the dark to calculate the E a using the Arrhenius equation.Since the soils in the dark were the sink of atmospheric Hg(0) in most of the study period, only a limited data set (n = 9) can be used (Fig. S3).The E a /R for Hg(0) emission from our remote soils with extremely low Hg concentrations (∼ 12 µg kg −1 ) was 30.40.Table S1 in Supplement lists the E a /R for different soils with large variation in soil Hg concentrations, including this study, and shows that the E a /R of Hg(0) emission from soils with low Hg concentrations was higher than that of soils with high Hg concentrations and significantly lower than that of the theoretical value (7.31) of elemental Hg.The result indicates that surface soils with high Hg concentrations has great potential to emit Hg(0).This trend is consistent with the laboratory study of Bahlmann et al. (2006), although the availability of Hg in soils also significantly regulates Hg(0) emission (Bahlmann et al., 2006;Kocman and Horvat, 2010).
Figure 8 shows that UV radiation was the dominant waveband of solar radiation for Hg(0) emission in the daytime, contributing > 80 % of Hg(0) emission in the light, and the contribution of UVB radiation accounted for > 50 % in all study days.This finding is consistent with previous laboratory studies (Moore and Carpi, 2005;Bahlmann et al., 2006;Xin et al., 2007).

Conclusions and implications
In this study, we measured the Hg(0) flux between the air and surface permafrost soil in the QTP.We also performed controlled field experiments to explore the effect of precipitation and different wavebands of solar radiation on the air-soil Hg(0) exchange.The results showed that the environmental conditions, including solar radiation, soil temperature and precipitation, greatly influenced the Hg(0) exchange between air and surface.This study and other field measurements and laboratory experiments have clarified that the fate and transport of soil Hg is very sensitive to the environmental variables (Krabbenhoft and Sunderland, 2013).Therefore, our results have several important implications to the Hg biogeochemical cycle in the soils of the QTP under the rapid climate warming and environmental change.Firstly, the increased surface temperature in the QTP will potentially promote the remobilization of soil Hg.Field measurements and a modeling study have revealed that the surface temperature in the QTP is increasing, and the warming trend exceeds those for the Northern Hemisphere and the same latitudinal zone (Kang et al., 2010).Secondly, the increased UV radiation in the QTP may enhance Hg(0) emission from surface soils.UV radiation reaching the surface of the QTP is estimated to increase because of the decrease of stratospheric O 3 (Zhou et al., 2013).Our result and many of the above-mentioned studies show that UV radiation plays the primary role in promoting the surface Hg(0) emission in the daytime.Thirdly, the temporal-spatial pattern of precipitation in the QTP is also changing (Kang et al., 2010), which potentially alters the flux and temporal-spatial pattern of air-soil Hg(0) exchange in this region because of the importance of precipitation on the Hg(0) exchange.However, this study was just the beginning to explore the effect of climate change on the terrestrial Hg cycle in the QTP.More studies are needed in the future.

Data availability
The data are available on request from the first author.
The Supplement related to this article is available online at doi:10.5194/acp-16-14741-2016-supplement.

Figure 2 .
Figure 2. Temporal variation in environmental variables, air Hg(0) concentrations inside and outside of the chamber, and air-surface Hg(0) flux at the Beiluhe station in the central QTP during four campaigns in 2014-2015.

Figure 3 .
Figure 3. Seasonal variation in Hg(0) concentration in ambient air and air-surface Hg(0) flux during four campaigns at the Beiluhe station in the central QTP in 2014-2015.

Figure 5 .
Figure 5. Increased Hg(0) emission for three different treatments (1, 5 and 15 mm addition of water) compared with the 0 mm treatment during the controlled experiment on 30 May 2015.

Figure 6 .
Figure 6.Daily Hg(0) flux and daily precipitation in the June 2014 campaign.

Figure 7 .
Figure 7. Temporal variation in bulk Hg(0) flux in the light, Hg(0) flux in the dark, and net Hg(0) flux in the light (bulk Hg(0) flux in the light-Hg(0) flux in the dark) in six study days without precipitation during the December 2014 campaign and May-June 2015 campaign.

Figure 8 .
Figure 8. Cumulative Hg(0) emission flux in the daytime triggered by UVB, UVA and visible light in four study days during the December 2014 campaign and May-June 2015 campaign.