Speciated atmospheric mercury on haze and non-haze days in an inland city in China

Long-term continuous measurements of speciated atmospheric mercury were conducted from July 2013 to June 2014 in Hefei, a midlatitude inland city in eastern central China that experiences frequent haze pollution. The mean concentrations (±standard deviation) of gaseous elemental mercury (GEM), gaseous oxidized mercury (GOM) and particle-bound mercury (PBM) were 3.95± 1.93 ng m−3, 2.49± 2.41 and 23.3± 90.8 pg m−3, respectively, on non-haze days, and 4.74± 1.62 ng m−3, 4.32± 8.36 and 60.2± 131.4 pg m−3, respectively, on haze days. Potential source contribution function (PSCF) analysis suggested that atmospheric mercury pollution on haze days was caused primarily by local emissions, instead of via longrange transport. The poorer mixing conditions on haze days also favored the accumulation of atmospheric mercury. Compared to GEM and GOM, PBM was especially sensitive to haze pollution. The mean PBM concentration on haze days was 2.5 times that on non-haze days due to elevated concentrations of particulate matter. PBM also showed a clear seasonal trend; its concentration was the highest in fall and winter, decreased rapidly in spring and was the lowest in summer, following the same order in the frequency of haze days in different seasons. On both non-haze and haze days, GOM concentrations remained low at night, but increased rapidly just before sunrise, which could be due to diurnal variation in air exchange between the boundary layer and free troposphere. However, non-haze and haze days showed different trends in daytime GEM and GOM concentrations. On nonhaze days, GEM and GOM declined synchronously through the afternoon, probably due to the retreat of the free tropospheric air as the height of the atmospheric boundary layer increases. In contrast, on haze days, GOM and GEM showed opposite trends with the highest GOM and lowest GEM observed in the afternoon, suggesting the occurrence of photochemical oxidation. This is supported by simple box-model calculations, which showed that oxidation of GEM to GOM does occur and that the transport of free tropospheric GOM alone is not large enough to account for the observed increase in daytime GOM. Our results further postulate that NO2 aggregation with the HgOH intermediate may be a potential mechanism for the enhanced production of GOM during daytime.


Introduction
Mercury (Hg) is an environmental pollutant that has received much global attention because of its toxicity and bioaccumulation in the aquatic ecosystems.The most important transport pathway of mercury is via the atmo-sphere (Schroeder and Munthe, 1998;Lindqvist and Rodhe, 1985).Operationally, atmospheric mercury is commonly differentiated into three forms: gaseous elemental mercury (GEM), gaseous oxidized mercury (GOM) and particlebound mercury (PBM).The sum of these three types of atmospheric speciated mercury is defined as total atmospheric mercury (TAM = GEM + GOM + PBM), and the sum of GEM and GOM is known as total gaseous mercury (TGM = GEM + GOM) (Gustin and Jaffe, 2010;Gustin et al., 2015).Globally, GEM is the dominant form of atmospheric mercury, accounting for over 95 % of the total.GEM is stable in the troposphere with a long residence time (0.5-2 years) and can be transported at the regional to global scale (Schroeder and Munthe, 1998;Lindberg et al., 2007).GEM can be photochemically oxidized to GOM, which can be converted to PBM upon adsorption on aerosol surfaces.Different from GEM, GOM and PBM can be readily removed form the air by wet and dry deposition as a result of their high surface affinity and water solubility (Lindqvist and Rodhe, 1985).Thus, chemical transformation between GEM, GOM and PBM will directly influence the atmospheric lifetime of mercury.
As a result of rapid industrial development and economic growth in recent decades, China has become one of the major contributors to anthropogenic mercury emissions to the environment (Wu et al., 2006;Pacyna et al., 2006Pacyna et al., , 2010;;L. Zhang et al., 2015).Atmospheric mercury emissions from anthropogenic sources in China have been estimated to be in the range of 500-700 t year −1 , accounting for 25-30 % of the total global anthropogenic mercury emissions (Streets et al., 2005;Wu et al., 2006).Studies of atmospheric mercury in China are therefore critical to the understanding of mercury cycling at both regional and global scales.Long-term observation of atmospheric mercury has been conducted in numerous urban and remote areas in China.TGM concentrations in urban and industrial areas were observed to be in the range of 2.7-35 ng m −3 , higher than the values reported for North America and Europe, and for the nearby Asian countries such as Korea and Japan (Weigelt et al., 2013;Fang et al., 2009;Marumoto et al., 2015).TGM and PBM concentrations in remote areas of China were also found to be higher than those observed in North America and Europe (Fu et al., 2008a(Fu et al., , b, 2012;;Liu et al., 2010).
In recent years, haze pollution has become a major concern in China due to its impacts on visibility, air quality and climate.It is well known that haze formation is mainly dependent on the atmospheric relative humidity (RH) and the concentration of airborne particles (Chen et al., 2003;Sun et al., 2013).Most studies on haze have focused on the measurements of airborne particulate matter; few have examined the influence of haze on the chemistry of atmospheric mercury, especially PBM.Here we report a 1-year real-time measurement of speciated atmospheric mercury in Hefei, an inland city of China, which experiences frequent haze events.The comparison of atmospheric mercury on haze days and non- haze days allows us to examine the formation and deposition mechanisms of mercury, as well as their temporal variations.

Study site
Hefei (31 • 52 N, 117 • 17 E) is the capital of Anhui province in eastern central China, between the Changjiang (Yangtze River) and the Huaihe (Huai River).The region has a humid subtropical climate with four distinct seasons: summer (June-August), fall (September-November), winter (December-February) and spring (March-May).The prevailing wind is southeasterly in summer and northwesterly in winter.Like many Chinese cities, Hefei has experienced rapid growth in the past 20 years, with a present-day total permanent population of about 7.7 million.The city has also been witnessing an increasing frequency in haze pollution, especially in winter months.
The monitoring site was located on Science Island, a small peninsula on the Dongpu Reservoir in the northwestern outskirts of Hefei (Fig. 1).The sampling and analytical instruments were installed 1.5 m above the rooftop (∼ 20 m above the ground) of the main building of the Anhui Institute of Optics and Fine Mechanics.Further information about the monitoring site can be found in a previous study (Hu et al., 2014).We chose this area as the monitoring site because it is not near to any direct pollution sources such as power plants or iron and steel works.

Measurements of speciated atmospheric mercury
From July 2013 to June 2014, simultaneous measurements of GEM, GOM and PBM were carried out by an automated Tekran ™ mercury speciation system.The system consisted of a Model 2537B mercury analyzer combined with a Model 1130 GOM unit and a Model 1135 PBM unit.The system was configured to measure GEM every 5 min, and GOM and PBM every 2 h.
The details about the Tekran ™ -based mercury speciation system can be found in Landis et al. (2002).In general, the automated measurement process can be summarized as sample collection, thermal desorption and determination.During the collection period, ambient air is drawn to the system at a typical flow rate of 10 L min −1 .GOM and PBM in the air are captured by a KCl-coated quartz annular denuder in the 1130 unit and a quartz filter in the 1135 unit, respectively, whereas GEM passes through the denuder and filter and is quantified on the 2537B analyzer by cold vapor atomic fluorescence spectroscopy (CVAFS).After 1 h of sampling, the 1135 quartz filter and the 1130 denuder are switched to the thermal decomposition mode at 800 and 500 • C, respectively, with the resulting Hg 0 quantified by the 2537B unit in the next hour, while the 1135 and 1130 components are flushed with zero-mercury gas for the next sampling.
The instrument maintenance followed typical protocols used in similar studies (Landis et al., 2002;Hu et al., 2014).The quartz annular denuder was recoated every 2 weeks, the quartz filter was replaced once a month and the Teflon filter (pore size 0.2 µm) in the sample inlet was changed every 2 weeks.Automated recalibration of the Tekran ™ 2537B was performed every 25 h using an internal mercury permeation source.No calibration standards were available for GOM and PBM, but the 1σ precision for GOM and PBM was about 15 % (Landis et al., 2002).The detection limit in ambient air is about 0.5 ng m −3 for GEM (or TGM) at a resolution of 5 min, and 1 pg m −3 for GOM and PBM at a resolution of 2 h (Gustin et al., 2015).Although the Tekran ™ -based mercury speciation technique has been widely used around the world, recent studies have shown that the technique does not efficiently collect all GOM, and thus may substantially underestimate the concentration of reactive mercury (Huang et al., 2013;Gustin et al., 2013).Therefore, the GOM values reported in this study should be considered as the lower limit of GOM in the air (Wang et al., 2014).

Ancillary data
Standard meteorological measurements including air temperature, pressure, RH, wind direction and speed were observed with a 5 min resolution.CO was measured by an automated infrared carbon monoxide analyzer (Model EC9830T, Ecotech Inc., Australia), with a detection limit of 40 ppbv.O 3 was measured every 5 min by an ozone analyzer (Model EC9810B, Ecotech Inc., Australia); its detection limit and accuracy were 0.5 and 1 ppbv, respectively.NO 2 was measured by a multi-axis differential optical absorption spectroscopy (MAX-DOAS) instrument.The collected spectra were analyzed using the QDOAS spectral fitting software suite developed at BIRA-IASB (http://uv-vis.aeronomie.be/software/QDOAS/).PM 2.5 (particulate matter of less than 2.5 µm in diameter) data were collected from the China air quality online analysis platform (http://www.aqistudy.cn/historydata/index.php).

Potential source contribution function (PSCF) analysis
To identify the possible influence of long-range transport on the distribution of atmospheric mercury in Hefei, we calculated backward trajectories of air masses using the HYSPLIT (Hybrid Single-Particle Lagrangian Integrated Trajectory) model with the Global Data Assimilation System (GDAS 1 • ) developed by the National Oceanic and Atmospheric Administration (NOAA) (http://www.ready.noaa.gov)(Draxler and Hess, 1998).Considering atmospheric pollutants are mainly concentrated in low altitudes during heavy pollution days, the trajectory arrival heights were set at 500 m to represent the boundary layer where atmospheric pollutants were well mixed.In this study, 3-day back-trajectories were generated hourly by the TrajStat software, which employs HYSPLIT for trajectory calculation (Wang et al., 2009).
The contributions of other source regions to the atmospheric mercury in Hefei were identified by the potential source contribution function (PSCF) analysis with TrajStat.PSCF analysis has been shown to be useful in spatially identifying emission sources for pollutants with a long lifetime such as elemental mercury and CO (Xu and Akhtar, 2010).The study domain is divided into grid cells, and the PSCF value for each cell was calculated by counting the trajectory segment endpoints that terminate within the cell.The PSCF value for the ij th cell is defined as where N ij is the number of endpoints that fall in the ij th cell, and M ij is the number of endpoints in the same cell that has a GEM concentration higher than an arbitrarily set criterion; in this study the criterion was set to be 4 ng m −3 , which is the mean GEM concentration during the entire study period.W ij is an arbitrary weight function introduced to reduce the effect of small values of N ij to better reflect the uncertainty in the values for these cells (Polissar et al., 2001).The weight function reduces the PSCF values when the total number of endpoints in a particular cell is less than 3 times the average value of the end points per cell: (2)

Results
We intended to continuously monitor speciated atmospheric mercury concentrations over the course of 1 year; however, interruptions were inevitable due to instrument maintenance, which resulted in loss of data for the following four periods: (1) 25 September to 9 October 2013; (2) 5-14 November 2013; (3) 9-25 February 2014; and (4) 1-14 April 2014.The rest of the data were grouped into haze days and nonhaze days according to the China Meteorological Administration's haze standard (QX/T 113-2010).Haze days refer to the days when the atmospheric visibility was < 10 km and the RH was < 80 % (Duan et al., 2016), and non-haze days refer to clear days with an atmospheric visibility > 10 km.
The visibility and RH information was collected from the weather history data at the Luogang Airport of Hefei (http: //www.wunderground.com/).Throughout the study period of almost 1 year, a total of 56 days were identified to be haze days, and 253 days to be non-haze days.All the times reported herein are local time (UTC + 8 h).

Overall characteristics of speciated atmospheric mercury
The time series of GEM, GOM and PBM concentrations at the study site throughout the study period are shown in Fig. 2; their frequency distributions are shown in Fig. S1 in the Supplement.The mean (±standard deviation) GEM, GOM and PBM concentrations during the entire study period were 4.07 ± 1.91 ng m −3 , 3.67 ± 5.11 and 30.0 ± 100.3 pg m −3 , respectively (Table 1).GEM concentrations in different seasons did not differ much, with the highest in fall (4.51 ± 2.10 ng m −3 ) and the lowest in spring (3.89 ± 1.79 ng m −3 ).GOM concentrations varied greatly, with much higher concentrations in fall and the lowest in winter.A similar seasonal variation in the GOM concentra-tion was observed at a remote site in Mt.Gongga of southwestern China (Fu et al., 2008b).PBM showed the highest degree of seasonal variability; its concentration decreased in the following order: fall ≈ winter > spring > summer.The mean PBM concentrations in fall and winter were about 20 times that in summer, similar to the findings from many previous studies in China (Zhang et al., 2013;Fu et al., 2011Fu et al., , 2008b;;Fang et al., 2001).
Comparisons of speciated atmospheric mercury concentrations with other urban and rural areas in China and a few other countries are shown in Table 2.The mean GEM concentration in Hefei is slightly higher than the concentrations reported from many remote areas in China (Fu et al., 2008a(Fu et al., , b, 2012;;Wan et al., 2009a, b;H. Zhang et al., 2015), but is much lower than the concentrations from urban areas of heavily industrial cities such as Guiyang and Changchun, where large point sources of mercury exist (e.g., nonferrous metal smelting, coal-fired power plants and residential coal burning) (Feng et al., 2004;Fu et al., 2011;Fang et al., 2004).Although Hefei is geographically close to Shanghai, a mega urban center in China, it is interesting to note that the TGM concentration of Shanghai is much lower than that of Hefei.This may be due to the fact that Shanghai is a coastal city that is influenced more by cleaner marine air masses (Friedli et al., 2011).Table 2 also shows that the average concentration of GEM in Hefei is double the typical values reported from the urban and rural areas in Europe and North America (Liu et al., 2010;Li et al., 2008;Brooks et al., 2010).

Speciated atmospheric mercury on non-haze days
As shown in Table 1 and Fig High concentrations of GOM (exceeding 10 pg m −3 ) only accounted for 1.4 % of the total data points.The mean GOM concentration in Hefei on these non-haze days is much smaller than the concentrations reported from other study sites in China (Table 2), but is comparable to the values observed at many European and North American sites (Peterson et al., 2012;Cheng et al., 2014;Ren et al., 2016).The mean PBM concentration on non-haze days was 23.3 ± 90.8 pg m −3 , with an exceptionally large range of 0.5-1827 pg m −3 ; high PBM concentrations (i.e., > 50 pg m −3 ) accounted for 6.4 % of the total data points.
The PBM concentration under non-haze conditions in Hefei is generally similar to values reported from remote areas in western China, such as Mt.Gongga, Mt.Waliguan and Shangri-La.Diurnal variations of GEM, PBM and GOM concentrations on non-haze days are shown in Fig. 3.Both GEM and PBM concentrations exhibited similar patterns, with elevated concentrations at night.The GOM concentration remained relatively constant at night, but increased rapidly just before sunrise and reached its peak value at ∼ 10:00 UTC + 8 h, followed by a synchronous decline with GEM through the afternoon (10:00-18:00).

Speciated atmospheric mercury on haze days
Haze pollution mainly occurred in December and January at our monitoring site.The four major haze pollution periods were identified in gray in Fig. 2. The mean concentrations of GEM, GOM and PBM on these haze days were 4.74 ± 1.62 ng m −3 , 4.32 ± 8.36 and 60.2 ± 131.4 pg m −3 , respectively (Table 1).The frequency distributions of GEM, GOM and PBM on the haze days are shown in Fig. S1 (in gray).GEM, GOM and PBM concentrations show significant differences between haze and non-haze days (p < 0.001, t test) (Fig. S2).On average, the concentration of GEM on haze days was 1.2 times that on non-haze days.Similarly, the concentration of GOM on haze days was about 1-1.7 times that on non-haze days.The largest impact of haze pollution is, however, on PBM, with the mean PBM concentration on haze days about 2.5 times that on non-haze days.High concentrations of GOM (exceeding 10 pg m −3 ) and PBM concentrations (exceeding 50 pg m −3 ) were also more frequently observed on haze days, accounting for 5.9 and 25 %, respectively, of the total haze days.
As shown in Fig. 3, on haze days the GEM concentration was higher at night and lower during daytime.PBM typically peaked just before sunrise, with the lowest values occurring in the afternoon (14:00-16:00).The opposite pattern was observed for GOM, which showed higher concentrations during daytime than at night.Although on both haze and nonhaze days GOM showed rapid increase just before sunrise, they exhibited different trends during daytime from 10:00 to 18:00.On haze days, GOM peaked in the afternoon when GEM was the lowest; the duration of the afternoon GOM peak was also longer on haze days.

Influence of atmospheric mercury emission source
With the year-long data seasonal mercury emission sources could be inferred from the PSCF analysis.Figure 4a showed the overall spatial contribution of mercury emission sources in China.As Hefei is located in eastern central China, its atmospheric mercury concentration could be affected by both northern and southern emission sources, including those from the North China Plain (especially Shandong province) and the neighboring provinces of Henan, Jiangsu, Jiangxi and Hubei.The total mercury emissions from Henan and Shandong provinces were estimated to be over 50 and 45 t in 2010, respectively, making them two of the largest Hg emitters in China (L.Zhang et al., 2015).Long-range transport could also impact seasonal variations of atmospheric mercury in Hefei.As shown in Fig. 4, in spring, the major contributors of atmospheric mercury to Hefei were from the southwestern region, including the local area and Jiangxi and Henan provinces.In summer, the main contributors were from north of Anhui, as well as Henan and Jiangxi provinces, and even from the Pearl River Delta region in the far south.Since the number of haze days only accounts for 5.6 % of the total days in spring and summer, we did not provide haze and non-haze PSCF results for spring and summer seasons.As fall and winter are the prevalent seasons for haze pollution, one PSCF result for haze days and another for non-haze days are shown for fall and winter, respectively.The statistically significant difference (p < 0.001) in the GEM concentration between non-haze days and haze days suggests that haze pollution could directly affect the concentration of elemental mercury.As shown in Fig. 4d and f, higher GEM concentration was mainly influenced by local emission sources on haze days.On non-haze days, the most important mercury sources to the monitoring site were not only the local emission sources, but also those from the neighboring regions of Shandong, Henan and Jiangxi provinces (see Fig. 4e and g).Therefore, the increase in the GEM concentration on haze days was mainly caused by local emissions.GEM and CO often share similar anthropogenic emission sources, such as industrial coal combustion, domestic coal combustion, iron and steel production and cement production (Wu et al., 2006;Wang et al., 2005).However, they also have their distinct emission sources.For instance, power plants and nonferrous metal smelters emit mercury but hardly any CO; while automobiles contribute greatly to CO emission, they are not a major emitter for mercury.The correlation coefficients and slopes between GEM and CO con-centrations during mercury pollution events are shown in Table 3.These mercury pollution episodes were defined as when the hourly average GEM concentration was higher than seasonal average GEM concentration for at least 10 consecutive hours.These episodes could be classified into longrange transport episodes or local episodes by using the coefficients of determination (R 2 ) of linear regression between Hg and CO: a significant positive correlation indicates longrange transport episodes, and a poor correlation signals local episodes (Jaffe et al., 2005;Weiss-Penzias et al., 2006;Kim et al., 2009).Using this approach, we identified three local episodes (events: 1-3) characterized by poor correlations between GEM and CO concentrations (R 2 : 0.23-0.29),and four long-range transport episodes (events: 4-7) characterized by positive correlations between GEM and CO concentrations (R 2 : 0.51-0.79).These local episodes tend to occur in fall and winter.The slope of the trend line represents the Hg / CO ratio, which could aid in the identification of specific emission sources.Emissions from power plants typically have a higher Hg / CO ratio (Wu et al., 2006), whereas residential coal and biomass burning combustion have a lower Hg / CO ratio (0.0013-0.0046 ng m −3 ppbv −1 ) due to incomplete combustion (Weiss-Penzias et al., 2007).The Hg / CO ratio for vehicles is close to zero (Zhang et al.,  Notes: these episodes were identified when the hourly average GEM concentration was higher than the seasonal average GEM concentration for more than 10 consecutive hours. 2013).The Hg / CO ratios during the pollution episodes in our study ranged from 0.0001 to 0.005 ng m −3 ppbv −1 , suggesting that mercury emission in fall and winter in Hefei could be related to local incomplete combustion sources, such as residential coal and biomass burning.

Impacts of meteorological factors for atmospheric mercury on haze days
Meteorological conditions, especially wind direction and speed, could also impact atmospheric mercury on haze days.The wind rose for the monitoring site during the study period is shown in Fig. 5. Easterly and southeasterly winds represented the prevailing wind directions at the study site.A wind rose diagram of GEM concentrations above the 90th percentile value is shown in Fig. 5b.We found that 67 % of the high GEM concentrations occurred at low wind speed (below 1.5 m s −1 ); however, wind speed below 1.5 m s −1 accounted for only 1.7 % of the study duration.High GOM and PBM concentrations appear not to be related to high wind speed (wind speed: 3-5 m s −1 ); only 1.4 and 2.6 % of the high GOM and PBM concentrations were observed under high wind speed conditions, respectively (Fig. 5c and d).The occurrence of high atmospheric mercury levels under low wind speed conditions is to be expected, as this slow wind speed condition is not conducive to the spread and mixing of mercury especially on haze days, and thus favors mercury accumulation in the air.This further supports the finding that atmospheric mercury on haze days is mainly due to local emissions.
Both GEM and PBM concentrations exhibited diurnal variations with elevated concentrations at night or in early morning, regardless of the presence of haze.Such a diurnal variation could be related to changes in the height of the urban boundary layer.The diurnal trend of the boundary layer height (BLH) in Hefei is shown in Fig. S3, which is typically low in the morning and night, and high during the daytime on both non-haze and haze days.Such diurnal changes in BLH in Hefei and nearby cities have also been observed in previous studies (Yuan et al., 2005;Mao et al., 2006).The maximum PBM concentration (observed at 06:00) was more than 4 times higher than the minimum value (observed at 16:00) on both non-haze and haze days, with a 76 % decrease from early morning to the afternoon.However, the reduction of PBM as a result of deposition during haze days was 62.7 pg m −3 , which was about 2.4 times that on nonhaze days, suggesting that haze pollution could increase the removal of PBM.Although PBM is not the major form of atmospheric mercury, it is crucial in atmospheric mercury transport and removal processes due to its short atmospheric lifetime.As shown in Fig. 6, the highest PBM and PM 2.5 concentrations were observed in January, which is most likely due to a shallower boundary layer in January than in other months.The co-variation in February is weaker, possibly due to the loss of PBM data because of instrument maintenance.The PBM concentration co-varied with the PM 2.5 concentration, especially in January when all the four PBM peak events were associated with increased PM 2.5 concentrations (Fig. 6c).These results suggest that PM 2.5 may play an important role in the formation of PBM.Thus, elevated PBM concentrations in fall and winter might be due to the combination of poor mixing conditions and higher PM concentrations.

Enhancement in GOM and the potential GEM oxidation mechanism
Diurnal variations of GEM, GOM, O 3 and CO concentrations on non-haze and haze days are shown in Fig. 7.The weak correlation (r = 0.164) between GOM and CO suggests that the CO-producing primary emission is not a major source of GOM in the air.This is clearly shown in Fig. 7 (haze days), where the peak value of GOM coincided with the lowest value of CO.As mentioned earlier, on both nonhaze and haze days, GOM concentrations remained relatively constant during night, but increased rapidly prior to sunrise.However, GOM showed different trends during daytime (10:00-18:00) between non-haze and haze days.On non-haze days, a synchronous decline was found between GEM and GOM in the afternoon, but the opposite trend was observed between GEM and GOM on haze days.This difference indicates that different GOM formation mechanisms might be at work on non-haze and haze days.Two processes can affect the GOM concentrations in the boundary layer air.
The first is the change in the atmospheric boundary layer height, which could change the transport of GOM-enriched free tropospheric air.Secondly, in situ photochemical oxidation of GEM would increase the GOM concentration during daytime.Various atmospheric oxidants are capable of oxidizing GEM to GOM, including halogen radicals, ozone, hydroxyl radicals (OH), among others (Holmes et al., 2010;Wang et al., 2014).
It is well established that free troposphere (FT) contains higher GOM concentrations than in the boundary layer (e.g., Murphy et al., 2006;Lyman and Jaffe, 2012;Timonen et al., 2013;Brooks et al., 2014;Shah et al., 2016).On both non-haze and haze days, it is thus possible that the higher GOM concentrations observed prior to sunrise are due to enhanced admixing of the free tropospheric air as the boundary layer increases in the morning (see Fig. S3).On non-haze days, a synchronous decline in GEM and GOM throughout afternoon (10:00-18:00) might be related to the higher atmospheric boundary layer height during this period.However, on haze days, the opposite variation was observed between GEM and GOM from 10:00 to 18:00, along with the elevated boundary layer height.This suggests that other than FT transport, photochemical oxidation of GEM might also play an important role in the enhancements of GOM.To determine the relative importance of FT transport and in situ photochemical oxidation, we examined the relationship between GOM and the changes in the height of the atmospheric  boundary layer and the odd oxygen (O x = O 3 + NO 2 ) concentrations.We used O x because it is a more conserved tracer of the extent of photochemical processes in the urban atmosphere (Herndon et al., 2008;Wood et al., 2010), as O 3 reacts with NO emitted from automobiles to form NO 2 .Example results are shown in Fig. 8 for 20 November 2013 (haze day).As can be seen from the figure, GEM and GOM showed opposite trends in the afternoon (12:00-16:00), along with higher O x concentrations during this period.The height of the atmospheric boundary layer changed very little (less than 0.1 km) over the same period (see Fig. S4).This simple comparison suggests that the transport of FT GOM might be limited and that at least some of the GOM was formed from in situ oxidation of GEM.Note that in our studies we could only calculate daytime O x concentrations because NO 2 concentrations from MAX-DOAS were only available during daytime.
We further investigated the mechanism of GEM oxidation to GOM.Ozone itself is not an efficient oxidant for GEM oxidation due to the low reaction rate (Hall, 1995;Holmes et al., 2010).Instead, halogen radicals (especially bromine atoms) and OH radicals are believed to be the primary oxidants for GEM in the global troposphere (Holmes et al., 2010).Unfortunately, we did not measure halogen radicals in this study.OH radicals are known to be present in the early morning urban boundary layer, primarily from the photolysis of HONO, which accumulates during night (Kleffmann et al., 2005).Therefore, here we consider the oxidation of GEM by OH radicals only.The formation of HgOH as an intermediate product of the Hg 0 (g) + OH oxidation reactions has been proposed by Sommar et al., 2001, although HgOH is highly unstable and could decompose back rapidly to Hg 0 and OH (Sommar et al., 2001;Goodsite et al., 2004).It has been proposed that the presence of other gases X (X = NO 2 , HO 2 , RO, RO 2 , or NO) could assist the formation of Hg(II) by forming X-HgOH, which outcompetes the decomposition of HgOH (Calvert and Lindberg, 2005;Dibble et al., 2012;Wang et al., 2014).As an example, we calculated the transformation between GEM and GOM under the influence of NO 2 , using the reactions and rate con-stants shown in Table S1.As shown in Fig. S5, the production rate of NO 2 HgOH, d[NO 2 HgOH]/dt, increased almost linearly with increasing NO 2 under low NO 2 concentrations, and eventually reached a steady state when the NO 2 concentration was high enough.
Based on the production rate of NO 2 HgOH, we can estimate the production of NO 2 HgOH during the 1 h sampling period when GOM was captured by the KCl-coated denuder in the Tekran ™ 1130 unit.The production of NO 2 HgOH and d[NO 2 HgOH]/dt corresponding to different NO 2 concentrations is shown in Table 4.With the increase of the NO 2 concentration, the contribution of the NO 2 HgOH production to GOM will increase.If the NO 2 concentration is within 100 ppbv (from 0 to 100 ppbv), the production of NO 2 HgOH would be in the range of 0.058-4.81pg m −3 during the 1 h sampling period.As illustrated in Table 4, the level of NO 2 observed in our study is high enough to account for the increase in the observed GOM production.Our results thus support a recent study in the tropical equatorial Pacific (Wang et al., 2014) that NO 2 aggregation with HgOH provides a possible mechanism for enhanced production of GOM.If that is true, NO 2 would be expected to play an even more important role in the urban air because of its higher concentration.More laboratory and modeling studies on the mercury oxidation mechanism in the presence of NO 2 and other gases are thus warranted.

Summary
Continuous measurements of speciated atmospheric mercury were conducted in Hefei, a midlatitude inland city in central China, from July 2013 to June 2014.Measurements of other trace gases (e.g., CO, O 3 , NO 2 ) and meteorological parameters were employed to better understand the sources and oxidation pathways of atmospheric mercury.The mean GEM, GOM and PBM concentrations during haze days were 4.74 ± 1.62 ng m −3 , 4.32 ± 8.36 and 60.2 ± 131.4 pg m −3 , respectively.Potential source contribution function (PSCF) analysis suggested that the local mercury emission rather than long-range transport is the most important contributor of atmospheric mercury pollution on haze days at our monitoring site.The low GEM / CO ratio in Hefei could be indicative of local incomplete combustion sources such as residential coal and biomass burning.Haze pollution has a more profound impact on PBM than on GEM and GOM.PBM showed a remarkable seasonal pattern, with higher concentrations in cold seasons and lower concentrations in warm seasons.Elevated PBM concentrations might be due to both the high loadings of particle matter and poorer mixing conditions on haze days, especially in cold months.Both GEM and PBM concentrations exhibited great variations with elevated concentration during the night.The diurnal variations of GEM and PBM might be related to the boundary layer depth; a lower boundary layer depth in the morning and night could elevate the mercury concentration.
Different from the diurnal variations of GEM and PBM, GOM concentration remained relatively constant at night, and then increased rapidly prior to sunrise.The enhancement of GOM during daytime could be due to both the transport of GOM-enriched free tropospheric air to the boundary layer and in situ oxidation of GEM in the boundary layer.Simple photochemical modeling supports the occurrence of daytime oxidation of GEM to GOM.Based on HgOH as an intermediate product, our calculations suggest that NO 2 aggregation with HgOH is a potential mechanism for the enhanced production of GOM in the inland urban air.
The Supplement related to this article is available online at doi:10.5194/acp-16-13807-2016-supplement.

Figure 1 .
Figure 1.Location of the study site in Hefei, China.

Figure 2 .
Figure 2. Time series of GEM, GOM and PBM concentrations, along with visibility and relatively humidity, at the monitoring site in Hefei from July 2013 to June 2014.The GEM data were at a 5 min resolution, and the GOM and PBM data were 2 h averages.The gray columns show the major haze pollution episodes that occurred during the study period.

Figure 3 .
Figure 3.Diurnal trends of GEM, GOM and PBM concentrations in Hefei on non-haze and haze days (local time = UTC + 8 h).The data were 2 h averages.
. S1 (in blue), the mean concentration of GEM on non-haze days was 3.95 ± 1.93 ng m −3 .Its distribution was characterized by large variations ranging from 0.2 to 23.8 ng m −3 , although more than half of the GEM values were in the narrow range 2-4 ng m −3 .The mean concentration of GOM on non-haze days was 2.49 ± 2.41 pg m −3 , with a range of 0.5-33.5 pg m −3 , although most of the values were in the range of 1-4 pg m −3 .

Figure 4 .Figure 5 .
Figure 4. Likely emission source areas of GEM identified by PSCF analysis.(a) Overall (from July 2013 to June 2014), (b) spring, (c) summer, (d) haze days in fall, (e) non-haze days in fall, (f) haze days in winter, (g) non-haze days in winter.

Figure 6 .
Figure 6.Monthly variation of (a) PBM and (b) PM 2.5 concentrations from January to June 2014; (c) average daily PM 2.5 and PBM concentrations in January 2014.

Figure 7 .
Figure 7.Diurnal variations of GEM, GOM, O 3 and CO concentrations on non-haze and haze days.

Figure 8 .
Figure 8.A case study of diurnal variations of GEM, GOM, O x and NO 2 in Hefei (20 November 2013).The top panel shows the hourly averaged GEM and GOM concentrations, and the bottom panel shows the O x (O x = NO 2 + O 3 ) and the NO 2 concentrations.The error bars for NO 2 refer to the NO 2 standard errors.

Table 1 .
Summary of GEM, GOM and PBM concentrations measured in Hefei from July 2013 to June 2014.

Table 3 .
Coefficients of determination and slopes between GEM and CO concentrations during atmospheric mercury pollution episodes ( * p < 0.01).